Findings from the Second Texas Air Quality Study (TexAQS II)

Final Rapid Science Synthesis Report: Findings from the

Second Texas Air Quality Study (TexAQS II)

A Report to the Texas Commission on Environmental Quality

by the

TexAQS II Rapid Science Synthesis Team

Prepared by the Southern Oxidants Study Office of the Director at

North Carolina State University

by Ellis B. Cowling, Director of SOS Cari Furiness, Research Associate

Basil Dimitriades, Adjunct Professor and

David Parrish Earth System Research Laboratory,

National Oceanic and Atmospheric Administration

In cooperation with Mark Estes of TCEQ and 47 other members of the Rapid Science Synthesis Team

TCEQ Contract Number 582-4-65614

31 August 2007

Final Rapid Science Synthesis Report: Findings from TexAQS II

31 August 2007 2

Acknowledgments

The Texas Commission on Environmental Quality (TCEQ) and the Rapid Science Synthesis Team (RSST) for the Second Texas Air Quality Study (TexAQS II) gratefully acknowledge significant contributions to the success of the TexAQS II field research campaign by the many different organizations listed below. These contributions include significant financial and in-kind organizational support as well as the time and creative energy of individual scientists, engineers, post-doctoral fellows, and graduate students who voluntarily joined in this collaborative enterprise.

These contributions include: • Defining the objectives of TexAQS II, • Instrumenting the various air-quality measurement platforms, • Implementing and quality assuring the myriad field measurements during the course of

the study, • Completing the analysis, interpretation, and intercomparison of research results, and • Formulating carefully crafted statements of research findings to be used for policy

purposes by TCEQ and other stakeholders.

Cooperating State Government Organizations California Air Resources Board Texas Commission on Environmental Quality Chief Engineer’s Office (CEO) Air Quality Division (AQD) Air Modeling and Data Analysis (AMDA) Industrial Emissions Assessment (IEA) Toxicology Office of Compliance and Enforcement Field Operations Division Region 4 – Dallas/Fort Worth Region 12 – Houston Monitoring Operations (MonOps) Ambient Monitoring Air Laboratory and Mobile Monitoring Data Management and Quality Assurance

Cooperating Texas Local Government Organizations Alamo Area Council of Governments (AACOG) Capital Area Council of Governments (CAPCOG) City of Corpus Christi City of Victoria

Cooperating Federal Government Agencies Department of Defense (DOD) Navy Postgraduate School (NPS)


31 August 2007 3

Department of Energy (DOE) Pacific Northwest Laboratory (PNL) Argonne National Laboratory (ANL) Los Alamos National Laboratory (LNL) Department of Interior (DOI) Minerals Management Service (MMS) Environment Canada (EC) Meteorological Service of Canada (MSC) National Aeronautics and Space Administration Headquarters Science Mission Directorate and science teams for the Atmospheric Infrared Sounder (AIRS), Cloud-Aerosol Lidar Infrared Pathfinder Satellite Observations (CALIPSO), Moderate Resolution Imaging Spectroradiometer (MODIS), Multi-angle Imaging SpectroRadiometer (MISR) Ozone Monitoring Instrument (OMI), and Tropospheric Emission Spectrometer (TES) Jet Propulsion Laboratory (JPL) Langley Research Center (LRC) National Oceanic and Atmospheric Administration (NOAA) Office of Oceans and Atmospheric Research (OAR) Earth System Research Laboratory (ESRL) Pacific Marine Environmental Laboratory (PMEL) Atlantic Oceanographic and Meteorological Laboratory (AOML) Air Resources Laboratory (ARL) National Weather Service (NWS) National Environmental Satellite Data and Information Service (NESDIS) Office of Marine and Aviation Operations (OMAO) National Science Foundation (NSF) National Center for Atmospheric Research (NCAR) Tennessee Valley Authority (TVA) United States Environmental Protection Agency (USEPA) Office of Air Quality Planning and Standards Air Quality Analysis Division Air Quality Modeling Group

Cooperating Research Universities Baylor University California Institute of Technology Chalmers University of Technology, Göteborg, Sweden Fort Hayes State University Georgia Institute of Technology Howard University Lamar University


31 August 2007 4

Texas Air Research Center (TARC) North Carolina State University Pennsylvania State University Portland State University Rice University Scripps Institution of Oceanography Texas A&M University Texas Tech University University of Alabama in Huntsville University of California at Los Angeles University of California at Santa Cruz University of Colorado Cooperative Institute for Research in Environmental Science (CIRES) University of Houston University of Manchester, Manchester, UK University of Maryland Baltimore County University of Massachusetts University of Miami University of New Hampshire University of Rhode Island University of Texas in Austin University of Washington Joint Institute for Study of Atmosphere Oceans (JISAO) Valparaiso University Washington State University

Cooperating Private Sector Organizations Aerodyne Research Inc. Baron Advanced Meteorological Systems Exxon Mobil Corporation Houston Regional Monitoring Network Science Systems and Applications, Inc. SensorSense, The Netherlands Shell Oil Corporation Sonoma Technology, Inc. URS Corporation

Other Cooperating Organizations Houston Advanced Research Center (HARC) Texas Environmental Research Consortium (TERC)


31 August 2007 5

Executive Summary

The Rapid Science Synthesis Team (RSST) for the Second Texas Air Quality Study (TexAQS II) has been charged to address a series of 12 High Priority SIP-Relevant Science Questions identified by the Texas Commission on Environmental Quality (TCEQ). Answers to these important questions were needed by TCEQ and other stakeholders in Texas to help fulfill the Commission’s responsibility to develop scientifically sound State Implementation Plans (SIPs) by which to attain the recently implemented 8-hour National Ambient Air Quality Standard (NAAQS) for Ozone. SIPs for both the Houston-Galveston-Brazoria and the Dallas-Fort Worth Ozone Non-Attainment Areas were submitted to EPA in June 2007.

This Final Report of the Rapid Science Synthesis Team for TexAQS II is designed to provide statements of Findings in response to each of TCEQ’s 12 High Priority SIP-Relevant Science Questions. This report addresses: 1) Significant sources of ozone and aerosol pollution in eastern Texas, 2) Photochemical and meteorological processes involved in the production, transport, and

accumulation of these pollutants in various parts of Texas, 3) An assessment of the adequacy of emissions inventories for both biogenic and

anthropogenic sources of ozone and aerosol precursor chemicals, and 4) An assessment of the skill of current air quality modeling and forecasting systems and

recommendations for improvement of these systems. This Executive Summary provides a short introduction to the scientific Findings from TexAQS II research for use by TCEQ managers and other air-quality decision makers and stakeholders in Texas. It contains a complete list of the 12 High Priority SIP-Relevant Science Questions that TCEQ asked our Rapid Science Synthesis Team to address a series of carefully crafted statements of Findings that have been developed in response to each of these questions.

We emphasize that the statements of scientific Findings developed as part of the Rapid Science Synthesis effort are based on analysis and interpretation of results from TexAQS II research within the limited time that was available between completion of the last TexAQS II field measurements on 15 October 2006 and our Rapid Science Synthesis contract ending date – 31 August 2007. More thorough and comprehensive analyses are already underway and will yield a great deal of additional important information in the future.

The institutional affiliations of the scientists responsible for the analyses leading to these Findings are given in the Acknowledgments section, immediately preceding this Executive Summary. Each section of this report is structured to contain one of the 12 SIP-Relevant Science Questions, then a numbered sequence of Findings in response to that question. Following each Finding, acknowledgment of the individuals whose analyses and data contributed to that Finding are listed. It also provides a brief discussion of the evidence that supports each Finding. The concerned reader should carefully consider the caveats contained in these discussions. For the convenience of all readers, a glossary of acronyms is provided in Appendix 1.

Please note that questions printed in blue were designated by TCEQ for special emphasis.


31 August 2007 6

Findings

Question A Which local emissions are responsible for the production of high ozone in Houston, Dallas, and eastern Texas? Are different kinds of emissions responsible for transient high ozone and 8-hour-average high ozone (i.e., ≥84 ppbv)?

Finding A1: The highest (i.e. > 125 ppbv) ozone concentrations in the HGB area result from rapid and efficient ozone formation in relatively narrow, concentrated plumes, which originate from HRVOC and NOx co-emitted from petrochemical facilities. The Houston Ship Channel (HSC) is the origin of the plumes with the highest ozone concentrations.

Finding A2a: Winds carry the emission plumes from the Ship Channel throughout the Houston area. The wind direction determines the location of the ozone maximum relative to the Ship Channel; shifts in wind direction lead to the transient high ozone events observed at monitoring sites within the Houston area.

Finding A2b: The general characteristics of the highest (i.e. > 125 ppbv) ozone formation in Houston did not change between 2000 and 2006, although the maximum observed ozone concentrations were lower in 2006 than in 2000.

Finding A3: Emissions from the Houston Ship Channel play a major role in the formation of the highest 8-hour average ozone concentrations (and ozone design values) observed in the Houston area.

Finding A4: Observations during TexAQS 2006 showed that nitryl chloride (ClNO2) is formed within the nocturnal boundary layer when NOx emissions and marine influences are both present. Following sunrise, ClNO2 photolyzes to yield chlorine atoms, which may lead to earlier and more rapid O3 production in the Houston region.

Question B How do the structure and dynamics of the planetary boundary layer and lower troposphere affect the ozone and aerosol concentrations in Houston, Dallas, and eastern Texas?

Finding B1: Boundary layer structure and mixing height near and over Galveston Bay and the eastern Houston ship channel area are spatially complex and variable from day to day. Vertical mixing profiles often do not fit simple models or conceptual profiles. High concentrations of pollutants are sometimes found above the boundary layer.

Finding B2: Complex coastal winds, resulting from weak larger-scale winds over the area, occurred during many, but not all, ozone exceedance days in the Houston-Galveston-Brazoria nonattainment area. Almost no high-ozone days during TexAQS 2000 resembled any high ozone days from TexAQS 2006, but collectively the 2000 and 2006 field intensives sampled the full range of meteorological conditions associated with high ozone events.


31 August 2007 7

Finding B3: After sea breeze days, the Houston plume was broadly dispersed at night through the formation of low-level jets.

Finding B4: The Dallas ozone plume can extend well beyond the monitoring network.

Question C Are highly reactive VOC and NOx emissions and resulting ambient concentrations still at the same levels in Houston as they were in 2000? How have they changed spatially and temporally? Are there specific locations where particularly large quantities of HRVOC are still being emitted? Are those emissions continuous or episodic? How well do the reported emissions inventories explain the observed concentrations of VOC and NOx?

Finding C1: There are indications that ethene (the lightest HRVOC) emissions from industrial sources in the Houston area decreased by 40 (±20)%, i.e., by a factor of between 1.25 and 2.5, between 2000 and 2006.

Finding C2: Measurements of ethene emission fluxes from petrochemical facilities during TexAQS 2006 indicate that the 2004 TCEQ point source database underestimates these 2006 emissions by one to two orders of magnitude. Repeated sampling of the same petrochemical facility showed that the ethene emission flux remained constant to within a factor of two.

Finding C3: Close to petrochemical HRVOC sources, the OH reactivity of propene is generally greater than that of ethene.

Finding C4: The latest available emission inventories underestimate ethene emissions by approximately an order of magnitude.

Finding C5: Inventories for NOx point sources at petrochemical facilities equipped with CEMS appear to be relatively accurate. Substantial decreases in NOx emissions in the Houston Ship Channel are suggested by the inventories, and measurements from aircraft are qualitatively consistent with the NOx decreases.

Question D What distribution of anthropogenic and biogenic emissions of ozone and aerosol precursors can be inferred from observations?

Finding D1a: Several rural electric generation units (EGU) in the Houston area and in eastern Texas have substantially decreased their NOx emissions per unit power generated since the TexAQS 2000 study. With one exception, SO2 emissions have not changed appreciably since 2000 for the plants sampled in 2006.

Finding D1b: Comparisons of emissions derived from ambient observations with those measured by continuous emission monitoring systems (CEMS) indicate that the emissions from point sources equipped with CEMS are very accurately known.


31 August 2007 8

Finding D1c: Underreporting of CO emissions at several EGU noted in 2000 (Nicks et al., 2003) has been reconciled by large increases (by factors of 5 to 50) in the inventory values between 2000 and 2006, as a result of newly implemented CEMS monitoring of CO at these plants.

Finding D2: On-road mobile emission inventories developed from MOBILE6 have significant shortcomings. MOBILE6 consistently overestimates CO emissions by about a factor of 2. It accurately estimated NOx emissions in the years near 2000, but it indicates decreases in NOx emissions since then, while ambient data suggest NOx emissions have actually increased. Consequently in 2006, NOx to VOC emission ratios in urban areas are likely underestimated by current inventories.

Finding D3: Emissions from ships constitute a significant NOx source in the HGB region. Literature results provide accurate emission factors for inventory development.

Finding D4: Mixing ratios of isoprene over Texas measured from the WP-3D were used for evaluation of the BEIS-3 emission inventory. On average, the isoprene emissions from the inventory and emissions derived from the measurements agree within a factor of ~2. There may be areas south of Dallas-Fort Worth and southwest of Houston, where isoprene emissions are lower than indicated by the BEIS3 inventory based upon the biogenics emission land cover data (BELD-3.0).

Finding D5: The speciation of VOC from mobile sources in the Houston and Dallas-Forth Worth areas agrees with detailed measurements in the northeastern U.S. However, initial results suggest that the agreement with the NEI-99 emission inventory is poor.

Question E Are there sources of ozone and aerosol precursors that are not represented in the reported emissions inventories?

Finding E1: The observed mixing ratios and regional distribution of ambient formaldehyde are broadly consistent with daytime photochemical production from reactive VOC. An upper limit for primary formaldehyde emissions from mobile sources is obtained from nighttime measurements, and is small in comparison with the secondary, daytime formation.

Finding E2: Concentrated plumes of ammonia were observed occasionally in the Houston Ship Channel area. These plumes often led to the formation of ammonium nitrate aerosol.

Finding E3: Concentrated plumes of gaseous elemental mercury from at least one point source were observed repeatedly in the Houston Ship Channel area and once in the Beaumont-Port Arthur area. The sources of the plumes could not be identified with current inventory sources of mercury.


31 August 2007 9

Question F How do the mesoscale chemical environments (NOx-sensitive ozone formation vs radical-sensitive ozone formation) vary spatially and temporally in Houston, Dallas, and eastern Texas? Which mesoscale chemical environments are most closely associated with high ozone and aerosol?

Finding F1: Both Eulerian and Lagrangian plume modeling approaches indicate that in 2000 high ozone concentrations in the HGB area were sensitive to both VOC and NOx emission reductions.

Finding F2: An observation-based approach to determine the sensitivities of high ozone in the HGB non-attainment area to the precursor VOC and NOx emissions has been investigated; it has yielded ambiguous results.

Finding F3: At the highest ozone concentrations, the observed relationship between ozone and the products of NOx oxidation indicates less efficient ozone production in the Dallas area than in the Houston area. In the observation-based indicator species approach, this behavior corresponds to less NOx-sensitive and more VOC- or radical-sensitive ozone formation in Dallas compared to Houston.

Finding F4: Tests of the ability of models to reproduce observed relationships between ozone and other photochemical products have the potential to provide very fruitful approaches to improving models.

Question G How do emissions from local and distant sources interact to determine the air quality in Texas? What meteorological and chemical conditions exist when elevated background ozone and aerosol from distant regions affect Texas? How high are background concentrations of ozone and aerosol, and how do they vary spatially and temporally?

Finding G1: The maximum background ozone concentrations encountered in 2006 exceeded the 8-hour NAAQS. On average, air of continental origin had higher background concentrations than marine air. The average background ozone concentrations measured in 2006 in eastern Texas complement a previously developed climatology.

Finding G2: The net ozone flux transported out of Houston averages about a factor of two to three larger than the corresponding flux from Dallas. The fluxes from these urban areas are significant contributors to the background ozone in the eastern Texas region.

Finding G3: Elevated background ozone concentrations for urban areas can include contributions from the recirculation of locally produced ozone or local precursor emissions.


31 August 2007 10

Finding G4: Plumes from Texas urban areas make substantial contributions to the ozone, aerosol, and precursor concentrations in the rural regions of eastern Texas.

Finding G5: Dust of African origin and sulfate aerosol advected into the Houston area, under southerly flow conditions from the Gulf of Mexico, can make significant contributions to the background aerosol in the eastern Texas region.

Finding G6: Nighttime chemistry influences the availability of oxides of nitrogen (NOx), highly reactive VOC (HRVOC), and O3.

Finding G7: Low rural nighttime ozone concentrations have been observed at some, but not all, rural locations in northeast Texas; these low nighttime ozone concentrations are not replicated in the regulatory modeling.

Question H Which areas within Texas adversely affect the air quality of non-attainment areas in Texas? Which areas outside of Texas adversely affect the air quality of non-attainment areas in Texas?

Finding H1: Ozone can be transported into the Dallas area from the Houston area.

Finding H2: High ozone concentrations in eastern Texas result from both in-state sources and transport of continental air from the east and northeast.

Finding H3: A synthesis of satellite and in situ measurements with photochemical modeling and Lagrangian trajectory analyses provides a quantification of regional influences and distant sources on Houston and Dallas air quality during TexAQS 2006.

Finding H4: In the Dallas area, local emissions and transport each contributed about equally to the average 8-hr ozone exceedance in 2002. Transported ozone alone can bring the Dallas area close to exceeding the 8-hour ozone standard.

Finding H5: In the Houston area, local emissions and transport each contributed about equally to the average 8-hr ozone exceedances investigated by aircraft flights in 2000 and 2006. As in Dallas, transported ozone alone can bring the Houston area close to exceeding the 8-hour ozone standard.

Question I Why does the SAPRC chemical mechanism give different results than the carbon bond (CB-IV) mechanism? Which replicates the actual chemistry better?

Finding I1: Air quality modeling for both 2000 and 2006 shows substantial differences in ozone concentrations predicted by the SAPRC99 and CB-IV chemical mechanisms.


31 August 2007 11

Finding I2: In regions with very high VOC reactivity and high NOx emission density, differences in ozone formation and accumulation predicted by regional photochemical models using the SAPRC99 and CB-IV mechanisms are due to differences in: (1) the chemistry of aromatics, especially mono-substituted aromatics (e.g., toluene), (2) nitric acid formation rates, and (3) the rates of free radical source terms in the SAPRC and CB-IV mechanisms.

Finding I3: The differences in the predictions of the SAPRC99 and CB-IV mechanisms can be probed using simulations of model compounds and comparisons of the simulations to environmental chamber data. For simulations involving CO and NOx (inorganic chemistry), the predictions of the CB-IV and SAPRC99 mechanisms that are used in regional photochemical models (with no chamber wall corrections) converge if the rate constants for the OH + NO2 reaction are made consistent between the two mechanisms. The predictions of the Carbon Bond mechanism, version 5 (CB05) also converge to the same predicted values if the OH + NO2 rate constant is made consistent with the other mechanisms.

Finding I4: The performance of the mechanisms in simulating olefins chemistry can be improved through more explicit representation of internal olefin chemistry, which has been added in CB05. In addition, performance of the SAPRC99 mechanism in simulating chamber data was improved for some experiments when propene was modeled explicitly, as opposed to being represented by a lumped chemical species.

Finding I5: For high reactivity chamber experiments involving olefins, sensitivity analyses indicated that mechanism adjustments that would lead to increased radical concentrations (increasing the radical yield in olefin-ozone reactions and changing the OH+NO2 termination rate constant) had little impact on predicted ozone concentrations.

Finding I6: The SAPRC99 mechanism performed better than the CB mechanisms in simulating some chamber experiments with toluene; the mechanism performances were more comparable for xylenes and other multiply substituted aromatics.

Finding I7: The differences between the SAPRC and CB-IV mechanism predictions for toluene chemistry decrease substantially if the yield for the lumped species CRES, and rate constant for the OH + NO2, are made consistent between the two mechanisms. The predictions of the CB05 mechanism also converge to the same predicted values if the CRES yield and the OH + NO2 rate constant are made consistent with the other mechanisms.

Finding I8: For simulations of ambient surrogate mixture experiments in the UCR EPA chamber, all mechanisms underpredict O3 at low ROG/NOx ratios, with the bias decreasing as the ROG/NOx ratio increases. In simulations of mixture experiments in chambers without aromatics, CB05 performs the best, and with no dependence of bias on the ROG/NOx ratio.


31 August 2007 12

Finding I9: SAPRC99, CB-IV, and CB05 all successfully predicted concentrations of ozone and other species during simulation of a stagnation event, if the chemical mechanisms were initialized after initial high concentrations of olefins had reacted. None of the mechanisms successfully predicted a rapid rise in radical concentrations and ozone concentration concurrent with initially high C2, C3, and CMBO concentrations during the event.

Question J How well do air quality forecast models predict the observed ozone and aerosol formation? What are the implications for improvement of ozone forecasts?

Finding J1a: Most forecast models exhibit skill in predicting maximum 8-hr-average O3 but none of the models is better than persistence in predicting 24-hr-average PM2.5 levels.

Finding J1b: Seven air quality forecast models and their ensemble were generally unable to forecast 85 ppbv 8-hr-average O3 exceedances with any reliability.

Finding J2: Sophisticated data assimilation of meteorological and even chemical observations is essential for improving photochemical model forecasts.

Finding J3: Model performance evaluations and intercomparisons require a comprehensive, best-guess emissions inventory for the TexAQS 2006 Field Intensive.

Question K How can observation and modeling approaches be used for determining (i) the sensitivities of high ozone in the HGB non-attainment area to the precursor VOC and NOx emissions, and (ii) the spatial/temporal variation of these sensitivities?

Finding K1: A simple, heuristic model based upon the Empirical Kinetic Modeling Approach (EKMA) method suggests that the HGB region may ultimately require drastic NOx emissions controls to reach compliance with the NAAQS for ozone.

Finding K2: The same model suggests that in a projected future scenario with very strict VOC emission controls, but without drastic NOx emission controls, biogenic VOC emissions plus background concentrations of CO and methane may be sufficient to cause ozone exceedances.

Question L What existing observational databases are suitable for evaluating and further developing meteorological models for application in the HGB area? The Final Report for Question L can be found in Appendix 2 of this Report, as submitted 31 October 2006.


31 August 2007 13

Introduction

The Rapid Science Synthesis Team (RSST) for the Second Texas Air Quality Study (TexAQS II) was charged to address the series of 12 High Priority SIP-Relevant Science Questions listed in the text box below. The Texas Commission on Environmental Quality (TCEQ) posed these questions to elicit responses needed to help TCEQ fulfill its responsibility to develop State Implementation Plans (SIPs) by which to attain the recently implemented 8-hour National Ambient Air Quality Standards (NAAQS) for ozone in the 8-county Houston-Galveston-Brazoria and 8-county Dallas-Fort Worth ozone non-attainment areas. TCEQ’s interests and questions also deal, to a lesser extent, with fine particulate matter. This report presents the final Responses of the RSST to these 12 TCEQ Science Questions.

As part of the Rapid Science Synthesis effort, TCEQ and the Office of the Director for the Southern Oxidants Study (SOS-OD) established Working Groups to address each Science Question. Each Working Group consisted of 8-15 experts drawn from various university-, state-, federal-, and private-sector organizations. The members of the Working Groups are listed in the text box below; contact information is given in Appendix 3.

The Working Groups developed research approaches for responding to each of the 12 TCEQ Science Questions; the approaches are described in a Progress Report from the RSST dated 31 July 2006. On 12 and 13 October 2006, shortly before completion of the TexAQS II Field Study, the Rapid Science Synthesis Team participated in a day-and-a half-long meeting in Austin, Texas. Presentations from this initial Rapid Science Synthesis Workshop were posted on the TCEQ website, http://www.tceq.state.tx.us/implementation/air/airmod/texaqs-files/TexAQSII.html. These presentations formed the basis for an RSST report, Preliminary Findings from the Second Texas Quality Study (TexAQS II), dated 31 October 2006 [8 November revision].

The RSST participated in the Principal Findings and Data Analysis Workshop for TexAQS II/ GoMACCS 2006, held 29 May – 1 June 2007, in Austin, Texas. Presentations from this Workshop were also posted on the TCEQ website, http://www.tceq.state.tx.us/implementation/air/airmod/committee/scc.html#workshops. These presentations and subsequent analyses of research results by various TexAQS II researchers provided the foundation for this final report of the Rapid Science Synthesis Team, Final Rapid Science Synthesis Report: Findings from TexAQS II. The sections of this Final Report present carefully crafted statements of scientific findings in response to each of TCEQ’s High Priority SIP-Relevant Science Questions. In this connection, we found very helpful the “Guidelines for Formulation of Scientific Findings to be Used for Policy Purposes” shown in Appendix 4.

Electronic copies of the three RSS reports to TCEQ are available on websites maintained by TCEQ (listed above) and by the National Oceanic and Atmospheric Administration (NOAA) (http://esrl.noaa.gov/csd/2006/).

The following introductory information provides some background and perspective for the Science Questions and the Responses developed by the RSST.


31 August 2007 14

TCEQ’s High Priority SIP-Relevant Science Questions and Leaders (L), Participants (P) and Observers (O) in Working Groups within the Rapid Science Synthesis Team

A Which local emissions are responsible for the production of high ozone in Houston, Dallas, and eastern Texas? Are different kinds of emissions responsible for transient high ozone and 8-hour-average high ozone (i.e., ≥84 ppbv)? L – David Parrish, P – Tom Ryerson, Joost deGouw, Basil Dimitriades, David Allen, Mark Estes, Bernhard Rappenglück, O – Noor Gillani

B How do the structure and dynamics of the planetary boundary layer and lower troposphere affect ozone and aerosol concentrations in Houston, Dallas, and eastern Texas? Co-L – Robert Banta & John Nielsen-Gammon, P – Allen White, Christoph Senff, Wayne Angevine, Bryan Lambeth, Lisa Darby, Bright Dornblaser, Daewon Byun, Bernhard Rappenglück, O – Carl Berkowitz, Noor Gillani

C Are highly-reactive VOC and NOx emissions and resulting ambient concentrations still at the same levels in Houston as they were in 2000? How have they changed spatially and temporally? Are there specific locations where particularly large quantities of HRVOC are still being emitted? Are those emissions continuous or episodic? How well do the reported emissions inventories explain the observed concentrations of VOC and NOx? L – David Parrish, P – David Allen, Joost deGouw, Tom Ryerson, Mark Estes, David Sullivan, John Jolly, Eric Williams, Barry Lefer, O – Yulong Xie, Carl Berkowitz, Noor Gillani. Note: To answer the last part of question C, TCEQ must define the inventory to which the observations must be compared.

D What distribution of anthropogenic and biogenic emissions of ozone and aerosol precursors can be inferred from observations? Co-L – David Allen & David Parrish, P – Tom Ryerson, Charles Brock, Joost deGouw, David Sullivan, Mark Estes, John Jolly, Eric Williams, Barry Lefer, Bernhard Rappenglück, O – Yulong Xie, Carl Berkowitz, Noor Gillani

E Are there sources of ozone and aerosol precursors that are not represented in the reported emissions inventories? L – David Parrish, P – Tom Ryerson, Charles Brock, Joost deGouw, David Sullivan, John Jolly, David Allen, Eric Williams, Barry Lefer, Bernhard Rappenglück

F How do the mesoscale chemical environments (NOx-sensitive ozone formation vs radical-sensitive ozone formation) vary spatially and temporally in Houston, Dallas and eastern Texas? Which mesoscale chemical environments are most closely associated with high ozone and aerosol? Co-L – Basil Dimitriades & David Parrish, P – David Allen, Harvey Jeffries, William Vizuete, Daewon Byun, Mark Estes, Kenneth Schere, Barry Lefer, Bernhard Rappenglück, O – Yulong Xie, Carl Berkowitz

Note: Letter designations are for convenience only and do not denote priority. Questions in blue have been designated by TCEQ to receive special emphasis.


31 August 2007 15

TCEQ’s High Priority SIP-Relevant Science Questions and Leaders (L), Participants (P) and Observers (O) in Working Groups within the Rapid Science Synthesis Team

(continued) G How do emissions from local and distant sources interact to determine the air

quality in Texas? What meteorological and chemical conditions exist when elevated background ozone and aerosol from distant regions affect Texas? How high are background concentrations of ozone and aerosol, and how do they vary spatially and temporally? Co-L – David Allen & David Parrish, P –Bryan Lambeth, David Sullivan, Basil Dimitriades, Charles Brock, Michael Hardesty, Steve Brown, Joost deGouw, Bernhard Rappenglück, Brad Pierce, Wallace McMillan, Kevin Bowman, David Winker, Tim Bates

H Which areas within Texas adversely affect the air quality of non-attainment areas within Texas? Which areas outside of Texas adversely affect the air quality of non-attainment areas within Texas? Co-L – David Allen & David Parrish, P – Mark Estes, Greg Yarwood, Basil Dimitriades, David Sullivan, Charles Brock, Michael Hardesty, John Jolly, Bryan Lambeth, Brad Pierce, Wallace McMillan, Kevin Bowman, David Winker

I Why does the SAPRC chemical mechanism give different results than CB-IV? Which replicates the actual chemistry better? Co-L – David Allen & Greg Yarwood, P – Harvey Jeffries, William Vizuete, Bill Carter, David Parrish, Stuart McKeen, Daewon Byun, Joost deGouw, Barry Lefer, Bernhard Rappenglück, O – Mark Estes, Noor Gillani

J How well do forecast air quality models predict the observed ozone and aerosol formation? What are the implications for improvement of ozone forecasts? L – Stuart McKeen, P – Gregory Carmichael, Bryan Lambeth, Kenneth Schere, James Wilczak, Greg Yarwood, Daewon Byun, John Nielsen-Gammon, Michael Hardesty

K How can observation and modeling approaches be used for determining (i) the sensitivities of high ozone in the HGB non-attainment area to the precursor VOC and NOx emissions, and (ii) the spatial/temporal variation of these sensitivities? Co-L – Basil Dimitriades & David Parrish, P – Ted Russell, Harvey Jeffries, William Vizuete, Mark Estes, David Sullivan, Tom Ryerson, Greg Yarwood, Barry Lefer, Bernhard Rappenglück, O – Noor Gillani

L What existing observational databases are suitable for evaluating and further developing meteorological models for application in the HGB area? L – Lisa Darby, P – Robert Banta, John Nielsen-Gammon, Daewon Byun, Wayne Angevine, Mark Estes, Bryan Lambeth, Stuart McKeen

Note: Letter designations are for convenience only and do not denote priority. Questions in blue have been designated by TCEQ to receive special emphasis.


31 August 2007 16

Non-attainment Areas Requiring SIP Development TCEQ initiated TexAQS II to obtain additional SIP-relevant understanding of the processes that lead to formation and accumulation of ozone and particulate matter in two very different ozone non-attainment areas within Texas:

1) The Houston-Galveston-Brazoria (HGB) ozone non-attainment area is a coastal urban region of about 5 million people. It consists of eight counties in southeastern Texas and is subject to distinctive coastal (sea-breeze) meteorological conditions and large petrochemical sources of industrial emissions, especially the Houston Ship Channel (HSC) and other nearby industrial sites. In 2005 HGB was classified as a non-attainment area of “moderate” status, with an attainment deadline of June 2010. The most recent SIP revision (June 2007) does not demonstrate attainment by that date, and the state of Texas has requested a reclassification of the HGB non-attainment area to “severe,” which requires an attainment date of June 2019 and a SIP revision deadline of March 2010. HGB is in attainment of the PM2.5 NAAQS.

2) The Dallas-Fort Worth (DFW) ozone non-attainment area is an inland urban region, also of about 5 million people. The DFW non-attainment area includes 8 counties in north-central Texas, with relatively typical inland metropolitan meteorological conditions and only limited industrial sources within the non-attainment counties, but with several power plants in nearby locations within northeastern Texas. DFW also was classified as in “moderate” non-attainment of the ozone standard, with a June 2010 attainment date. The June 2007 SIP revision submitted to EPA demonstrates attainment of the 8-hour standard by that deadline. DFW is in attainment of the PM2.5 NAAQS.

Recent Air Quality Studies in Texas During recent years, the State of Texas provided substantial funding for scientific studies to improve understanding of ozone formation and accumulation in eastern Texas. These scientific studies have been focused around two major air-quality field research programs.

First Texas Air-Quality Study (TexAQS 2000) TexAQS 2000 was a relatively short-term (six-week-long) intensive field measurement program conducted during the summer of 2000 that included aircraft-based, tall tower-based, and ground-based field measurements; the study period was imbedded in a 16-month long study on particulate matter in the Houston area. The research results demonstrated that the extraordinarily stringent decrease in NOx emissions proposed in the 2000 SIP for the HGB non-attainment area of Texas was not an optimal approach; a more realistic plan for attainment of the NAAQS for ozone should involve decreases in emissions of both VOC and NOx – with a particular focus on the highly reactive VOC (HRVOC) emissions in the industrial areas surrounding the Houston Ship Channel and Galveston Bay.

Second Texas Air Quality Study (TexAQS II and TexAQS 2006) TexAQS II was a much longer-term (18-month-long) program focused on the photochemical and meteorological processes leading to the formation and accumulation of ozone and particulate matter air pollution in eastern Texas. The study began in June 2005 and extended through 15 October 2006. This period includes not only the 2005 and 2006 summer ozone seasons, but also the intervening fall, winter, and spring months when occasional exceedances of the 8-hour ozone standard did occur, both together with and separately from, occasional episodes of high


31 August 2007 17

concentrations of airborne particulate matter. The field measurements of TexAQS II constitute one of the most comprehensive air quality field research studies ever undertaken in the United States; the analysis and interpretation phases of TexAQS II will extend for years after completion of the field measurements.

TexAQS II culminated during the months of August, September, and October 2006 with an intensive series of coordinated chemical and meteorological measurements and modeling studies. In order to distinguish this relatively short-term but very intensive study from the earlier parts of TexAQS II, the 2006 summer intensive study has been dubbed TexAQS 2006.

The air-quality measurement platforms during TexAQS 2006 included: 1) Multiple instrumented aircraft operating in both the HGB and DFW areas of Texas, 2) A series of ground-based sites with continuing measurements of air chemistry, a network

of ground-based wind profiler and rawinsonde locations, a mobile solar occultation flux research van, and both aircraft-based and ship-based ozone and aerosol lidar measurements throughout eastern Texas,

3) NOAA’s Ronald H. Brown Research Vessel operating within the Houston Ship Channel and Galveston Bay,

4) The 200-foot-tall Moody Tower at the University of Houston, and 5) The National Aeronautics and Space Administration (NASA) Aura, Aqua, Terra, and

GOES satellites.

The multimillion-dollar 18-month-long TexAQS II field research study and its embedded short-term intensive study, TexAQS 2006, were conducted jointly by staff of TCEQ and by scientists and engineers working under contracts issued by TCEQ, the Texas Environmental Research Consortium (TERC), and through formal and/or informal agreements for cooperation with the many other research organizations listed in the Acknowledgments of this Report.

All the research studies and plans for analysis and interpretation of results obtained during TexAQS II and TexAQS 2006 were undertaken with specific scientific research objectives in mind. But many of these investigations also were designed, undertaken, and funded by various federal, state, and private-sector organizations with specific policy purposes in mind – and most particularly in order to be used by TCEQ in developing State Implementation Plans that were required by the US EPA in 2007 for the Houston-Galveston-Brazoria and Dallas-Fort Worth ozone non-attainment areas. The very limited time available between completion of many of the TexAQS II field measurements and the deadline for preparation and final submission of the SIPs required for Houston and Dallas-Forth Worth was extraordinarily short – thus the need for a rapid synthesis of SIP-relevant science.

History of Ozone Exceedances in Houston and Dallas Figure 1 shows that maximum ozone concentrations throughout HGB and DFW decreased from 1978 to 2005. It is notable that:

1) Most of the decreases in HGB occurred before 1990; 2) The decreases were greater in HGB (from over 150 ppbv to near 100 ppbv) than in DFW

(from about 120 ppbv to near 95 ppbv); and 3) The ozone design values in HGB are now approximately the same as in DFW. The large

variability in ozone maxima since 1990 – both among monitoring sites and between years – make it difficult to determine whether these apparent decreases in ozone maxima since 1999 are due to changes in emissions or to meteorological variations.


31 August 2007 18

Although the ozone design values for the HGB and DFW areas are now very similar, acute ozone episodes are much more frequent in Houston than Dallas. During 2000-2005, the HGB area had 219 8-hour ozone exceedance days while the DFW area had 203. In contrast, however, during this same period, the HGB area had 147 one-hour exceedance days (ozone concentrations

Figure 1. Temporal trends of 8-hour O3 design values at a) HGB and b) DFW monitoring sites. The 8-hour O3 design value for each monitor is the 3-year average of the fourth-highest daily maximum 8-hour-average ozone concentration measured at that monitor; these values are representative of the maximum ozone observed and the air concentrations of specific concern in developing control strategies. (Figure from Bryan Lambeth, TCEQ)


31 August 2007 19

> 120 ppbv) but the DFW area had only 24 such days. Over 60 percent of the HGB 8-hour exceedances were accompanied by 1-hour exceedances. In the HGB area the local design value maxima have not changed locations during that period, and have persisted in the vicinity of Deer Park, Bayland Park, and Aldine. There is a local minimum in ozone design value at sites located in the urban core of Houston, presumably because of the abundant fresh NO emissions there, which can suppress ozone formation, or titrate ozone transported into the area.

Regional Character of Ozone Exceedances in Eastern Texas Maximum daily ozone concentrations generally occur in multi-day episodes, and episodes often are region-wide, with similar patterns throughout eastern Texas (Figure 2). These region-wide episodes of high ozone concentrations are seen in all years. Two hypotheses can be suggested to explain this behavior:

1) The episodes occur during meteorological conditions under which elevated ozone concentrations, produced in upwind regions, are transported into the whole region of eastern Texas.

2) The high ozone episodes represent periods when the regional meteorology is particularly conducive (sunny, hot, stagnant) to local or regional formation of ozone from locally or regionally emitted precursors, and the intervening periods represent meteorological conditions that are particularly poor (cloudy, cool, windy) for ozone formation within the region.

Figure 2. Daily maximum 8-hour ozone averages for August-September 2004 for all monitoring sites in eastern Texas. Each column represents one day, and each row represents one monitor within each community (TLM-Tyler, Longview, Marshall; BPA-Beaumont, Port Arthur; Aus-Austin; SAo-San Antonio, CCV-Corpus Christi, Victoria). The increasing color intensity indicates increasing ozone concentrations (yellow ≥65 ppbv, orange ≥85 ppbv, red ≥105 ppbv, purple ≥125 ppbv. (Data from TCEQ; Figure adapted from David Sullivan-U. Texas)


31 August 2007 20

A decreasing temporal trend in regional ozone concentrations and exceedances has occurred over the past 13 years. A trend analysis covering 1994 to 2004 shows that the average of the maximum 8-hour-average ozone concentrations observed at upwind surface monitoring stations in the Houston area have declined since about 2000 (Figure 3). Values were approximately 80 ppbv in the 1990s and have dropped to between 60 and 70 ppbv in 2002 – 2004. This suggests that background ozone concentrations during high ozone episodes in the Houston area have been declining in recent years. There has been a concomitant decrease in the total number of exceedance days recorded each year in August and September at twelve representative eastern Texas monitoring sites (Figure 4). These sites were selected for completeness of data coverage (>75% each year) and regional representation (one site per county, except for two in Harris County). It is important to note that the year of the TexAQS 2006 field study had the fewest number of regional exceedances during this 13-year period.

Figure 4. Total number of monitor- days with maximum 8-hour-average ozone above 85 ppbv in August and September during the years between 1994 and 2006. Twelve regional monitoring sites in eastern Texas were considered in this analysis. (Data from TCEQ; Figure from David Sullivan-U. Texas)

Figure 3. 1994 to 2004 trend of highest 3 - 6 values of 8-hour background ozone from upwind surface stations in Houston area (from Nielsen-Gammon et al., 2005).


31 August 2007 21

An Important Distinction: Ozone Production versus Accumulation From the standpoints of understanding air quality science and the practical business of air quality management, it is important to recognize the distinction between ozone formation and/or ozone production (which are essentially synonymous), on the one hand, and ozone accumulation on the other hand. The ambient concentration of ozone in a particular, near-surface volume of air at a fixed location is the net result of six processes that occur in the lower atmosphere:

• In situ ozone formation (or production) from chemical precursors in an air parcel,

• In situ ozone destruction by chemical reactions in that air parcel,

• Deposition of ozone from the air to vegetation or other surfaces,

• Vertical transport of ozone from an ozone reservoir aloft,

• Horizontal transport of ozone from up-wind locations, and

• Dispersion and dilution as a result of mixing with cleaner air during advection or vertical mixing when the height of the planetary boundary layer increases.

In effect, the rate of change in the concentration of ozone in a particular, near-surface volume of air is an “algebraic sum” of the rates of all six of these processes. The time-integral of this “algebraic sum” then represents the accumulation of ozone in that volume of air at that specific location (if the integral is positive, or a decrease in ambient ozone concentration if the integral is negative.)

Atmospheric scientists and engineers are well aware of these important processes and distinctions. They devote much effort to experimentally determining the rates of each of these processes, and to developing models that accurately calculate the resulting accumulation of ozone to compare with ambient observations of ozone concentrations.

Air quality managers are concerned with management of near-surface ozone concentrations; thus, they are not concerned with the production of ozone as such, so long as the ambient concentration of ozone does not exceed the NAAQS. Air quality managers are required to develop State Implementation Plans that are aimed at keeping ozone from accumulating in the atmosphere in concentrations that exceed these NAAQS.

In the mixed dialog that occurs between atmospheric scientists and air quality managers, this important distinction can be lost, and a degree of confusion may result. Some of this confusion arises because, among the six processes listed above, only the rate of ozone formation can be effectively addressed through local emission control efforts. Thus, ozone formation receives particular attention from both atmospheric scientists and air quality managers.

In this report, the term “accumulation of ozone” is used when the distinction is important. However, in many situations, the distinction between “accumulation of ozone” and “ozone production” is not so important. These situations arise when the rate of production dominates the “algebraic sum” above. In such situations, the discussion will focus on “ozone production” and the contribution of the other five processes will not be discussed.


31 August 2007 22

Reactivity of VOC Ozone Precursors Photochemical production of ozone requires two classes of precursors: oxides of nitrogen (NOx, which include NO and NO2) and volatile organic compounds (VOC). During ozone formation the VOC are photochemically oxidized while NOx catalyzes the formation of ozone during that oxidation. Oxidation of carbon monoxide (CO) and methane also can contribute significantly to the formation of ozone.

Figure 5 shows the rate constants for the reactions of many different VOC with the hydroxyl radical (OH). This radical is the active agent in the atmosphere that initiates the photochemical oxidation of VOC. The OH concentration in ambient air varies widely, but a representative, 24-hour average value is 1 x 106 OH radicals per cm3 of air. The right hand axis in Figure 5 gives the lifetimes of various VOC at that representative OH concentration. The OH rate constants span such a wide range that the VOC lifetimes vary from many days to less than one hour. The effectiveness of any given VOC in ozone formation is approximately proportional to its OH reactivity, which is the product of the VOC concentration times its OH rate constant.

The TexAQS 2000 study highlighted the important role played by highly reactive VOC (HRVOC) in the HGB area. These species are the alkenes, five of which are included in Figure 5 (above the “Alkenes” label in the figure). Their importance is due to their high OH reactivities, which result from both their large OH rate constants and their high ambient concentrations. The high ambient concentrations result from large quantities of HRVOC emitted from petrochemical facilities in the HGB area.

The biogenic VOC, especially isoprene, are particularly important in forested regions of eastern Texas (Figure 6). Recently, new land cover and vegetation data have improved the spatial resolution and provided more quantifiable estimates of uncertainties for these emissions. Biogenic emissions derived from these new data are about 40% lower than previous calculations. It is important to recognize that some biogenic VOC, especially isoprene, are also emitted from anthropogenic sources in the industrial facilities in HGB.

Figure 5. Rate constants and representative atmospheric lifetimes for some VOC. CO is included for comparison. Rate constant data are from the literature.


31 August 2007 23

Figure 6. Biogenic VOC emissions in southeast Texas on a representative summer day. The emissions are calculated at 4 km x 4 km spatial resolution based on the latest University of Texas-Center for Space Research land cover data. (Figure adapted from Mark Estes, TCEQ)

A Reflective Note Although this is the Final Report of the Rapid Science Synthesis Team, all of us on the RSST recognize that we have not yet derived maximum benefit from all the research data and information that was accumulated during TexAQS II. The accelerated pace of our analyses and interpretations did not allow full utilization of the results. Fortunately, however, the future will provide many additional opportunities for more detailed analyses, for more intercomparisons of results, and for more integrative interpretations. The findings developed in this Final Report will be subject to more rigorous examination when TexAQS II results are presented at scientific meetings and when papers deriving from TexAQS II research are submitted for publication in refereed scientific journals.

We are proud of what has been accomplished so far – the new scientific insights obtained, and the new policy insights provided. We hope this Final Report will be useful to TCEQ and the people of Texas, as they continue the practical business of air quality management in Houston and Dallas-Fort Worth.

We wish to express our thanks to the many staff in TCEQ who organized our several workshop meetings and helped us gain access to additional data and information from existing databases – especially those that go back in time to observations made long before both TexAQS 2000 and TexAQS II were conceived and implemented. We have also enjoyed the excellent cooperation among so many of the diverse groups of scientists, engineers, graduate students, and post-docs that joined in various parts of this very large study. It was fascinating to have the chance to understand more fully the very challenging chemical, biological, physical, meteorological, and mathematical factors that are involved, as well as some of the social, cultural, industrial, political, professional, and economic forces involved in managing air quality – especially in the Houston-Galveston, and Brazoria ozone non-attainment area.


31 August 2007 24

Response to Question A

QUESTION A Which local emissions are responsible for the production of high ozone in Houston, Dallas, and eastern Texas? Are different kinds of emissions responsible for transient high ozone and 8-hour-average high ozone (i.e., ≥84 ppbv)?

BACKGROUND Question A is related to Questions F and K, which focus on the sensitivities of high concentrations of ozone to VOC and NOx emissions. Here the response to Question A gives an overview of the photochemical processes that produce the highest observed ozone concentrations, and the emissions that allow those processes to proceed rapidly. These highest ozone concentrations are limited to the HGB region, so the focus is on that area. Understanding 8-hour-average high ozone requires very different considerations and approaches than the understanding of transient high ozone; an initial investigation of the longer-term-average ozone in HGB is included.

FINDINGS

Finding A1: The highest (i.e. > 125 ppbv) ozone concentrations in the HGB area result from rapid and efficient ozone formation in relatively narrow, concentrated plumes, which originate from HRVOC and NOx co-emitted from petrochemical facilities. The Houston Ship Channel (HSC) is the origin of the plumes with the highest ozone concentrations.

Analysis: Trainer-NOAA; Data: Ryerson, de Gouw et al.-NOAA, Fried et al.-NCAR, Atlas et al.-U. Miami. The aircraft flights of TexAQS 2000 and 2006 reveal a persistent feature over Houston – on all days conducive to photochemical ozone production, narrow and concentrated plumes of ozone formed and were transported downwind from the HSC. Figure A1 shows measurements from one flight; on this day the wind was from the east-northeast. The top two graphs show the measurements of two primary ozone precursors. NOy represents the NOx emissions plus their oxidation products, primarily PAN and nitric acid. Ethene, a HRVOC, is an important ozone precursor. The concentrations of these two primary emissions were highest over the HSC, and defined a narrow downwind plume with decreasing concentrations. Figure A1 indicates that the NOx and ethene emission sources are not precisely collocated, but the primary sources of both are in the HSC region. The downwind decrease in the concentration of NOy was primarily due to dilution. Clearly, ethene concentrations decreased more rapidly than NOy, due to its much faster removal through photochemical oxidation. The bottom two panels of Figure A1 show the concentrations of two secondary photochemical products, ozone and formaldehyde, formed within the plume of primary emissions. Formaldehyde, produced by the photochemical oxidation of ethene and other HRVOC, formed during the downwind transport as the HRVOC were oxidized; formaldehyde itself is an important ozone precursor. Ozone accumulated as a concentrated (up to 147 ppbv), narrow plume within the much less concentrated (≤ 100 ppbv), but wider ozone plume from the larger Houston urban area.


31 August 2007 25

-96.5 -96.0 -95.5 -95.0 -94.5

161284

CH2O (ppbv)

12840C2H4 (ppbv)

30.5

30.0

29.5

29.0

28.5

403020100NOy (ppbv)

30.5

30.0

29.5

29.0

28.5-96.5 -96.0 -95.5 -95.0 -94.5

60 90 120 150

O3 (ppbv)

Figure A1. Measurements of NOy, ethene (C2H2), ozone, and formaldehyde (CH2O), plotted along the flight track of the WP-3D on 6 October 2006. The symbols are sized and color-coded according to the indicated concentrations. Only measurements below 1.5 km altitude are shown. (A close approach to the Parish power plant has been removed from the NOy plot.)

The high temporal resolution measurements of ethene made by the laser photo-acoustic spectrometer (LPAS) and the wide species coverage of VOC measurements conducted on the canisters from the whole air sampler (WAS) provided detailed characterization of the HRVOC sources in HSC. The map in Figure A2 shows the concentration of ethene measured by these two systems downwind from six petrochemical complexes; the plumes of ethene emissions from the complexes are clearly delineated. The four pie charts in Figure A2 indicate the total OH reactivity of the VOC in four of the WAS samples, and show how that reactivity was divided among four of the lightest HRVOC species, and the total of all of the other alkanes, aromatics, and biogenic VOC measured in the canister samples. These four canister samples illustrate the general finding that the HRVOC dominate the VOC reactivity immediately downwind of the petrochemical facilities, and thus are responsible for a large fraction of the high ozone


31 August 2007 26

production in the HGB area. The total reactivity and the fraction of the reactivity contributed by the HRVOC decrease downwind in the plumes, as expected from their very short lifetimes.

The pie charts in Figure A2 do not include aldehydes. These oxygenated VOC, which also account for a significant fraction of the OH reactivity, are primarily secondary products of the oxidation of the HRVOC. Thus, the contribution of the aldehydes to ozone production increases the fraction of the total VOC reactivity in the HSC area that is attributable to the emissions of the HRVOC.

Figure A2. Measurements of ethene along the section of the WP-3D flight track included in the rectangle in the upper right panel of Figure A1. The symbols are sized and colored according to the indicated concentrations. The triangles indicate the 20-second-average measurements by LPAS, and the circles indicate the measurements from WAS. The pie charts are sized according to the total OH reactivity for four of the WAS samples, and the colors indicate the OH reactivity contributed by the four indicated HRVOC.


31 August 2007 27

Finding A2a: Winds carry the emission plumes from the Ship Channel throughout the Houston area. The wind direction determines the location of the ozone maximum relative to the Ship Channel; shifts in wind direction lead to the transient high ozone events observed at monitoring sites within the Houston area.

Finding A2b: The general characteristics of the highest (i.e. > 125 ppbv) ozone formation in Houston did not change between 2000 and 2006, although the maximum observed ozone concentrations were lower in 2006 than in 2000.

Analysis: Ryerson et al.-NOAA; Data: Ryerson, Williams et al.-NOAA. Ryerson et al. (2006) examined the four Electra flights during TexAQS 2000 that encountered ozone concentrations above 150 ppbv. Figure A3 shows the flight track segments where the highest ozone concentrations were observed. In each case, back-trajectory analysis attributed these high ozone concentrations to plumes from industrial emission sources in the HSC area. Measured chemical characteristics of the plumes (simultaneous high NOx, SO2, CO2, and oxidation products of HVROC) confirmed this source attribution. The relationship between the transport times derived from the trajectory analysis and the observed enhancements in ozone concentration provided a measure of the net average ozone production rates in the plumes. Figure A3 also shows the observed relationship between ozone and the products of NOx oxidation for those four flights; the slopes of these relationships provide an estimate of the net ozone production efficiencies in these plumes.

Figure A3. Highest ozone (red points) observed by the Electra aircraft during four flights in TexAQS 2000. In the left panel flight track segments are color-coded by observed ozone, and in the right panel the dependence of ozone on the products of NOx oxidation are shown with approximate ozone production efficiencies estimated from fitted slopes (Ryerson et al., 2006).


31 August 2007 28

Figure A4 presents a similar analysis for the three WP-3D daytime flights and the RHB ship cruise segment during TexAQS 2006 that encountered ozone concentrations near or above 120 ppbv. A simple wind direction analysis, coupled with the chemical plume signatures, indicate that in each case these plumes also originated from industrial emission sources in the Houston Ship Channel area. Figure A4 also shows the relationship between ozone and the products of NOx oxidation in these plumes.

Comparison of Figures A3 and A4 show that the ozone production environment in the Houston area was similar in 2000 and 2006. Similar ozone vs. (NOy-NOx) slopes are seen in each year; the maximum concentrations of ozone observed in both years were associated with slopes near 7. Lower maximum ozone concentrations were observed in 2006, but this difference may be partially due to meteorological factors; the background ozone, as indicated by the y-intercepts in the two right panels in Figures A3 and A4, was lower in 2006 and lower background concentrations led to lower maximum concentrations. The lower 2006 background may indicate smaller effects of stagnation and recirculation that year. Finally, the maximum ozone concentrations were observed at greater distances from the HSC in 2006 than in 2000 (note difference in area covered in the two maps). This difference may be attributable to slower ozone production or to greater transport speeds in 2006 than in 2000.

Figure A4. Highest ozone observed by the WP-3D aircraft and RHB research vessel during TexAQS 2006. The figure is in the same format as Figure A3. The dotted rectangle in the left panel indicates area covered in Figure A3.


31 August 2007 29

Finding A3: Emissions from the Houston Ship Channel play a major role in the formation of the highest 8-hour average ozone concentrations (and ozone design values) observed in the Houston area.

Analysis: Sullivan et al.-U. Texas; Data: TCEQ. The general flow patterns that give rise to the highest 8-hour-average ozone concentrations in the HGB area have been investigated through back-trajectory analysis. Monitoring sites to the south and west of downtown Houston have the highest ozone design values in the HGB area (defined as the 3-year average of the fourth-highest daily maximum 8-hour-average ozone concentration). In Figure A5 the blue dots highlight the four highest design values: 104 ppbv at Tom Bass CAMS 558, 103 ppbv at Bayland Park CAMS 53, 102 ppbv at West Houston CAMS 554, and 98 ppbv at Monroe CAMS 406. The green dot indicates the Aldine site north of the city; as recently as 2002 Aldine had the maximum design value in the area, but it declined from 108 and 107 ppbv in 2001 and 2002, respectively, to 92 and 88 ppbv in 2005 and 2006, respectively.

Relationship between Ozone Production Efficiency and the Slope of the O3 versus NOy-NOx Correlation – A Cautionary Note

Many investigations of atmospheric photochemistry have interpreted the slope of the correlation of the measured O3 concentration versus the measured concentration of the oxidation products of NOx (i.e. the difference between measured concentrations of NOy and NOx) as a direct measure of the ozone production efficiency (i.e. the number of O3 molecules formed for each NOx oxidized.) The first paper that examined this correlation noted that nitric acid deposition will affect the observed slope between O3 and NOy-NOx correlations. If nitric acid is removed from the atmosphere, then a measurement of NOy-NOx will underestimate the concentration of NOx oxidation products, and consequently, the slope of O3 versus NOy-NOx will overestimate the ozone production efficiency. During TexAQS 2006, ozone production was studied under a variety of meteorological conditions. High wind speeds were found to enhance nitric acid loss, which caused the O3 versus NOx oxidation products correlation slope to increase, often dramatically. The observed slopes varied from 2 to 16 in the coalesced plume from Houston, which was followed up to 170 km downwind. In many cases, the particularly high O3 versus NOy-NOx slopes were primarily caused by particularly rapid loss of nitric acid from the transported plume, rather than from particularly high ozone production efficiency within the plume. Thus, when interpreting the correlation slope of O3 versus NOy-NOx as a measure of the ozone production efficiency, it is critical to consider the effect of nitric acid loss.


31 August 2007 30

Figure A5. Harris County monitoring sites with the 4 highest 2006 design values marked in blue, and Aldine marked in green. In 2006, Bayland Park monitoring site (C53) had the area-wide highest regulatory design value of 104 ppbv.

A set of near-surface back-trajectory ensembles was generated for 8-hour ozone exceedance days at Bayland Park and Monroe, the two regulatory sites with the highest 2006 design values, and at Aldine to provide geographical contrast. The analysis covers the period from 2000 through 2006. On each day that a given monitor recorded an ozone exceedance, 7-hour back trajectories were calculated for all hours in which the measured ozone was above 85 ppbv. Hourly wind speed and direction data from 14 locations covering Harris County provided the data for the back-trajectory calculations. Wind speeds were adjusted to account for varying anemometer exposure among sites using factors provided by Bryan Lambeth of TCEQ, and further increased by 20 percent to account for generally higher speeds above each monitor’s 10-meter anemometer tower. The data from all sites were pooled in an un-weighted average to provide one urban-scale wind speed and direction vector for each hour. Air parcel back trajectories were calculated in 10-minute time steps beginning from the start time of each appropriate hour. The resulting trajectory ensembles include 79 days at Aldine, 104 at Bayland Park, and 56 at Monroe. Since exceedance days generally had multiple hours with ozone concentrations over 85 ppbv, the number of trajectories totaled 252 at Aldine, 281 at Bayland Park, and 110 at Monroe. One composite trajectory was produced from each ensemble by averaging the x-y coordinates of all trajectories in the ensemble as a function of transport time. The resultant composites are expected to be biased short, but based on maps of the ensembles and their centerlines, this effect is judged to be minimal.

Figure A6 shows that each of these three composite back-trajectories reaches back to the HSC area earlier in the day, with approximately 7-hour transport times to the respective monitoring


31 August 2007 31

sites. For Aldine the trajectories cluster to the south-southwest, for Bayland Park they wrap around the south side of central Houston, and for Monroe they cluster to the northeast. The consistency of the trajectories coming from the HSC provides strong support for the conclusion that emissions from the petrochemical industrial facilities in the HSC play a major role in the 8-hour ozone exceedances in the HGB area.

Figure A6. Composite 7-hour near-surface back trajectories for three Houston monitoring sites. All hours with O3 ≥ 85 ppbv on 8-hour exceedance days for the 2000-2006 period were included in the composite analysis.

Finding A4: Observations during TexAQS 2006 showed that nitryl chloride (ClNO2) is formed within the nocturnal boundary layer when NOx emissions and marine influences are both present. Following sunrise, ClNO2 photolyzes to yield chlorine atoms, which may lead to earlier and more rapid O3 production in the Houston region.

Analysis and Data: Roberts, Osthoff et al.-NOAA. The first reported atmospheric observations of nitryl chloride, ClNO2, were made aboard the Ronald H Brown during TexAQS 2006. Simultaneous measurements of ClNO2, N2O5, and aerosol size and composition, show that ClNO2 is a general product formed when N2O5 reacts on chloride-containing aerosol. N2O5 is produced from nighttime reactions of NO2 with O3. ClNO2


31 August 2007 32

is relatively stable at night, but photolyzes upon sunrise to yield Cl atoms and NO2. These reactions link the nitrogen oxides to halogen activation.

An episode of high concentrations of ClNO2 was observed on 2 September 2006 when the RHB was anchored in the Barbour’s Cut inlet located off Galveston Bay near HSC. Figure A7 shows the measured N2O5 and ClNO2 concentrations along with the Cl atom production rate calculated from the measured ClNO2 and solar radiation flux. This Cl atom production reached almost 106 sec-1, which is highly significant since Cl atoms react with VOC up to 100 times more rapidly than OH radicals. Figure A7 compares O3 measured on 2 September with the average (±1 standard deviation) of the O3 levels observed during the other days that the RHB was in the Houston-Galveston area during TexAQS 2006. On 2 September the O3 concentration increased more rapidly and reached a higher peak concentration compared to the average increase. This observation, while not definitive, is indicative of a potentially significant role for Cl atoms in O3 production in polluted marine air.

A preliminary box model study was conducted to investigate the ClNO2 formation chemistry and to examine the effect of this Cl source on VOC-NOx photochemistry. The model consisted of the Master Chemical Mechanism (Jenkin et al., 2003; Saunders et al., 2003), to which the N2O5 aerosol chemistry and ClNO2 photochemistry were added. The model results show that a reaction efficiency of 25% for N2O5 on chloride-containing aerosol can account for the measurements in Figure A7. The Cl atoms produced upon sunrise led to a 2- to 3-fold increase in total peroxy radicals during the morning hours, which resulted in about a 15% higher O3 concentration at the end of the afternoon production period. These results indicate the need to add this Cl atom source to current regional photochemical models.

Figure A7. N2O5 and ClNO2 measured on 2 September 2006, and calculated Cl atom production in the atmosphere (a); and the measured O3 on 2 September, and the average from the other measurements in the Houston-Galveston area (b).


31 August 2007 33

KEY CITATIONS AND INFORMATION AND DATA SOURCES Jenkin, M.E., S.M. Saunders, V. Wagner, and M.J. Pilling. 2003. Protocol for the development

of the master chemical mechanism MCMv3 (Part B): Tropospheric degradation of aromatic volatile organic compounds. Atmospheric Chemistry and Physics, 3:181-193 (also Atmospheric Chemistry and Physics Discussions, 2:1905-1938 (2002)).

Ryerson, T.B., K.K. Perkins, M. Trainer, D.K. Nicks Jr., J.S. Holloway, J.A. Neuman, F. Flocke, A. Weinheimer, S.G. Donnelly, S. Schauffler, V. Stroud, E.L. Atlas, D.D. Parrish, R.W. Dissly, G.J. Frost, G. Hübler, R.O. Jakoubek, P.D. Goldan, W.C. Kuster, D.T. Sueper, A. Fried, B.P. Wert, R.J. Alvarez, R.M. Banta, L.S. Darby, C.J. Senff, and F.C. Fehsenfeld. 2006. Chemical and meteorological influences on extreme (>150 ppbv) ozone exceedances in the Houston metropolitan area. Draft Report to TCEQ, Contract No. 582-4-65613.

Saunders, S.M., M.E. Jenkin, R.G. Derwent, and M.J. Pilling. 2003. Protocol for the development of the master chemical mechanism MCMv3 (Part A): Tropospheric degradation of non-aromatic volatile organic compounds. Atmospheric Chemistry and Physics, 3:161-180 (also Atmospheric Chemistry and Physics Discussions, 2:1847-1903 (2002)).


31 August 2007 34

Response to Question B

QUESTION B How do the structure and dynamics of the planetary boundary layer and lower troposphere affect the ozone and aerosol concentrations in Houston, Dallas, and East Texas?

BACKGROUND Meteorology in the boundary layer affects ozone levels through regulation of near-source concentrations (wind speed, mixing height), background levels (transport winds), photochemistry (solar radiation, temperature), and air parcel history (local winds).

FINDINGS

Finding B1: Boundary layer structure and mixing height near and over Galveston Bay and the eastern Houston ship channel area are spatially complex and variable from day to day. Vertical mixing profiles often do not fit simple models or conceptual profiles. High concentrations of pollutants are sometimes found above the boundary layer.

Analysis: Nielsen-Gammon-Texas A&M; Senff, Darby, Banta, Angevine, Tucker, White-NOAA; Breitenbach, Dornblaser, Lambeth-TCEQ; Morris-Valparaiso U.; Rappenglück, Perna-U. Houston. Data: Nielsen-Gammon-Texas A&M; Senff, Tucker, White, Darby, Angevine, Banta-NOAA; Morris-Valparaiso U.; Rappenglück, Perna-U. Houston. Mixing depth (synonymous and used interchangeably here with the term “mixing height”) exerts an important control on the concentrations of pollutants including ozone and its precursors. For example, when pollutants are released into a shallow mixed layer, high concentrations of emissions can accumulate; this was observed using airborne ozone lidar during TexAQS 2000 (Banta et al., 2005). On the other hand, deep mixing can significantly dilute pollutants.

Well away from the coastal region, the mixed layer was relatively uniform over broad spatial areas. Well inland, peak afternoon mixing heights were generally between 1½ and 2½ km, but could reach as high as 4 km. The mixed-layer heights generally grew into late afternoon, often leveling off above 2 km. Over the Gulf of Mexico, shipboard lidar and platform radar wind-profiler measurements indicated a relatively constant maritime mixed-layer depth of 600 m, with weak positive heat fluxes at the surface, both day and night.

The Houston urban area and major industrial sources are located in the coastal region, where the mixing depth is highly variable in space and time, as determined by shipboard lidar and land-based sensors. Reasons for this variability include land-sea contrast and the sea-breeze cycle, land-use differences, and along-shore coastal irregularities, the major one in this area being Galveston Bay. The meteorology of the previous day and night is another influence on mixed-layer variability, but that effect is much more difficult to generalize. The coastal zone of southeast Texas is a transition region between the maritime boundary layer, with the relatively constant 600-m mixed-layer depths over the Gulf of Mexico, and the deeper daytime mixed layers inland. The coastal influence on Houston mixed-layer heights can be seen in their diurnal behavior. In contrast to the inland sites, where the mixed-layer height generally increases until late afternoon, in the coastal zone the mixing height was observed to reach its maximum earlier


31 August 2007 35

in the day (as early as late morning for sites near the shore) and then to decrease, as the sea-breeze front brought in cooler marine air through the afternoon.

Land-cover and land-use differences include the cross-region gradients in soil moisture and urban heat island processes. Climatologies and vegetation indicate moister conditions to the east of Houston and nearer to the coast, and drier conditions to the west and inland. Airborne lidar flights revealed a significantly deeper mixed layer over the arid regions to the west, consistent with mixing-depth data provided by the profiler array, which indicated increases in peak afternoon mixing depths with distance from the coast.

Urban heat island. The issue of the urban heat island (UHI) is also complicated in the coastal zone by the sea breeze. The effects of the sea breeze can be appreciated by contrasting Houston with an inland urban area, Dallas. Both Dallas and Houston have comprehensive networks of surface meteorology and chemistry sensors. The similarities of the networks and lack of terrain in Dallas and Houston allow for the comparison of their UHIs. The strength of the UHI is measured by taking the temperature difference from representative urban and rural sites. The Dallas UHI, unperturbed by thermal flows driven by land/sea temperature differences, was a well-defined phenomenon during the summers of 2000-2006 (Fig. B1a). Including all weather conditions, the average nighttime Turban – Trural temperature difference was between 1.5º and 2.0º C and the average daytime difference was ~ 1.0º C. Analysis of Houston temperature data, however, revealed a different picture due to the Bay and Gulf breezes (Fig. B1b). Although the Houston UHI was a distinct phenomenon, even when including all weather conditions, the Bay or Gulf breezes modified the Houston UHI by cooling the city. Average nighttime Turban – Trural temperature differences in Houston were between 1.75º and 2.75º C. However, during the day, the rural areas to the north and west of the city were often warmer than the downtown area during afternoon hours as a result of the sea breeze. Averaging the Houston Turban – Trural temperature differences over the summers of 2000-2006 indicated a very small urban-rural temperature difference between 1400 and 1600 local standard time (LST), in contrast to Dallas, which had a Turban – Trural temperature difference of ~1° C. In some individual years, such as 2000, 2003, 2005 and 2006, the Houston urban areas were actually cooler than the rural areas, on average, in the mid-afternoon. These years had more Bay-breeze/Gulf-breeze activity to cool the urban area.

Figure B1. Turban – Trural for each of the 7 summers analyzed, and the average for all summers. “Summer” includes all days from 1 June– 30 September. a) Dallas. b) Houston.


31 August 2007 36

The effects of along-shore irregularity combined with the other coastal effects can be seen in the distributions of daytime mixed-layer heights. Midday mixed-layer heights measured at an inland site at Moody, Texas (near Waco) were relatively consistent from day to day, thus exhibiting a rather narrow distribution with a mean of ~1.5 km. Near the shore of Galveston Bay at LaPorte, a much broader distribution of mixed-layer heights, including high frequencies below 1 km, indicates the much more complex influences on mixing depth in the coastal zone. Spatial variability of mixing depth was evident in airborne lidar flights, with shallow mixing layers over Galveston Bay and deeper mixing layer depths over the Houston urban area as a result of heat-island effects. Variations in mixing depth at nearby locations often reflected advection of high or low mixing layer depths downwind of the urban area or Galveston Bay, for example.

Sounding observations from the University of Houston (UH) campus (e.g., Fig. B2) and from the Ronald H. Brown (when stationed in Galveston Bay) often showed a very complex vertical ozone structure, with no well-marked transition between the local boundary layer and the air above; this was consistent with shipboard Differential Absorption Lidar (DIAL) and Doppler lidar profiles. This complexity is likely to be underrepresented in numerical models.

Figure B2. Ozonesonde profiles of ozone (left, 0-100 ppb) and temperature (right, 270-300 K) from the surface to 6,000 m above ground level (AGL), for a 1200 UTC ascent on 31 August 2006, released from the University of Houston campus. The minimum in the ozone profile near 1,000 m AGL was associated with a layer of high relative humidity and clouds.

Spatial variability of mixing height during TexAQS 2000 and TexAQS 2006. Multi-sensor spatial analyses of mixing heights, using data from the TexAQS field campaigns, illustrate the effects of these diverse influences, often showing mixing height varying by a factor of two or more across


31 August 2007 37

Houston. An example of large spatial variability in mixing heights is shown in Figure B3. A combination of fixed-location ground-based sensors and mobile airborne sensors is essential for determining the spatial and temporal mixing height variability across the Houston metropolitan area.

Figure B3. Mixing height measurements in early afternoon in Houston, 1 September 2000. Colors indicate observed or estimated mixing heights. Shown are mixing heights from airborne instruments (colored lines) and mixing heights from profilers, soundings, and aircraft ascents and descents (circles). Note the large spatial variability north and east of Houston (Nielsen-Gammon et al., 2007).

Mixing height and O3 concentrations. Attempts to determine general relationships between daytime mixing heights and ozone concentrations have been mostly fruitless. The expected relationship is that higher concentrations would be associated with lower mixed-layer heights. Northerly or northeasterly large-scale flow, however, is often associated with higher concentrations from distant continental sources, and continental air masses often have higher mixing heights. Conversely, and also opposing the expected mixing-height/ozone relationship, out over Galveston Bay, large-scale southerly flow off the Gulf of Mexico was observed to be associated with more clean and moist air but shallower mixed layers. On the other hand, when pollutants are emitted into the shallow mixed layer of Galveston Bay and carried over the Bay, pollutant concentrations have been shown to remain very high – in agreement with expectation. In another scenario, aircraft sampling of urban or other plumes some distance downwind of the sources finds well mixed regions outside and inside of the plumes where the ozone is low and high, respectively, on the same day, but the mixing heights are essentially the same, yielding no


31 August 2007 38

systematic relationship. All of these effects confound the search for general relationships, producing seemingly random or otherwise meaningless scatter plots of O3 and mixing heights

Nighttime mixing. Nighttime boundary layer structure in the Houston-Galveston area is less easy to characterize. UH tethersonde and ozonesonde data indicate that nighttime depletion can occur at elevations as high as 200 m, implying that the top of the urban boundary layer tends to be ~200 m or less at night. Thus, nighttime mixing and transport effects in the Houston-Galveston area remain an area of longer-term research needs.

Layering of pollutants. O3 and aerosol pollutants in the Houston-Galveston area were often confined to the mixed layer (Fig. B4a,b). But occasionally significant concentrations of pollutants were found above the mixed-layer height (Fig. B4c,d). Pollutants, especially aerosol, sometimes exhibited a complex layered structure, in which the origin of pollution in the individual layers was difficult to determine (Fig. B2). In the analysis performed to date, high concentrations of pollutants found in deep layers (more than a few hundred meters thick) above the mixed layer (Fig. B4d) were found to originate in distant regions outside of Texas to the northeast or east.

Figure B4. Airborne lidar time-height cross sections of aerosol backscatter (top, in units of 10-8 m-1 sr-1) and ozone concentrations (bottom, in ppb) for a), b) 30 August 2006, when high ozone and aerosol concentrations were confined to the mixed layer, and c), d) 4 September 2006, when the mixed layer was indicated by higher aerosol concentrations, but the high ozone was above the boundary layer.

Finding B2: Complex coastal winds, resulting from weak larger-scale winds over the area, occurred during many, but not all, ozone exceedance days in the Houston-Galveston-Brazoria nonattainment area. Almost no high-ozone days during TexAQS 2000 resembled any high ozone days from TexAQS 2006, but collectively the 2000 and 2006 field intensives sampled the full range of meteorological conditions associated with high ozone events.

Analysis: Banta-NOAA; Nielsen-Gammon-Texas A&M; Bryan Lambeth-TCEQ; Gary Morris-Valparaiso U.; Perna, Rappenglück-U. Houston; Senff, Darby, Angevine-NOAA; Breitenbach, Dornblaser-TCEQ. Data: Nielsen-Gammon-Texas A&M; Senff, Tucker, White, Darby, Angevine, Banta-NOAA; Morris- Valparaiso U.; Rappenglück, Perna-U. Houston. High ozone concentrations were most often associated with light winds and daytime sea-breeze reversals (often leading to stagnation occurrences; (Darby, 2005]), but some days with stronger


31 August 2007 39

winds also had O3 exceedances during both 2000 and 2006. The interplay between local wind patterns and large-scale transport produced a wide variety of meteorological scenarios for ozone events in the Houston area. Broadly speaking, when mean winds (here measured using 24-hr-mean resultant winds from nearshore buoy 42035) were light, recirculation was likely, whereas when mean winds were strong, the wind direction tended to be steady throughout the day. The direction of the mean wind determines when stagnation will occur if winds are light enough, and also determines the direction of transport of the Houston ozone plume.

By plotting the west-east and south-north components of the mean winds on a scatter diagram, the range of meteorological conditions during an ozone episode or an extended period may be determined. Such a diagram is shown in Fig. B5 for the late summer ozone season in Houston. During TexAQS 2000, most high-ozone events occurred when the mean wind was from the southeast (regime 1) or southwest (regime 2), but no light northeast wind events occurred at all. In contrast, during TexAQS 2006, wind conditions were much more representative of climatology, and several high-ozone events occurred under light northeast winds. Those days in which the wind was light from the southeast during 2006 tended to feature widespread clouds and precipitation. Thus, almost no high-ozone days during TexAQS 2000 resembled any high ozone days from TexAQS 2006, but collectively the 2000 and 2006 field intensives sampled the full range of meteorological conditions associated with high ozone events.

Figure B5. Scatter diagram of daily mean winds at buoy 42035, near Houston, during the period 1 August – 15 October, 1998-2006. Grid lines are every 5 m/s; a dot in the upper left quadrant implies winds from the southeast. The wind regimes most conducive to high 1-hr ozone in the Houston area are circled. Days during TexAQS 2000 are indicated with pink squares, and days during TexAQS 2006 are indicated with yellow triangles; blue dots represent data from all other years.

Peak ozone and wind speed. The overall relationship between peak O3 concentrations and wind speed is shown in Fig. B6, which indicates the expected strong negative correlation between


31 August 2007 40

maximum in-network ozone enhancements (or “add-on”) above background and wind speed in the Houston area (see red data and line on Fig. B6). The 10-hr trajectory displacement is a surrogate for mean wind speed. Stronger winds (larger displacements) produce greater dilution of the pollution emissions, and thus lower concentrations of pollutants. Analysis of ground-based measurements over several years indicates a threshold effect in Houston (Nielsen-Gammon et al., 2005). Large local add-ons are typically present when large-scale winds are weaker than about 3 m s-1, light enough to allow recirculation, and the add-on decreases rapidly for wind speeds larger than 3 m s-1. This effect is evident especially in the airborne data in Fig. B6. A steady 3 m s-1 wind would produce a 10-hr displacement of 108 km, and below this value is where peak O3 add-on concentrations become very large.

R2 = 0.5033

R2 = 0.3902

-20

0

20

40

60

80

100

120

140

160

180

200

0 100 200 300 400

10-hr displacement (km)

O3

add

on

Airb add on

Sfc add on

Linear (Airb add on)

Linear (Sfc add on)

Figure B6. Houston ozone enhancement (peak ozone values in urban plume minus background values) plotted vs. displacement of 10-hr trajectories, representing the vector-mean wind for the period, starting at Houston at 8:00 a.m. CST. Red symbols indicate data from surface measurement network (1-hr average), and blue symbols indicate data from airborne ozone lidar and WP-3D. “Add on” represents peak ozone concentrations observed minus background concentrations. Lines are linear best-fit lines.

The blue points in Fig. B6 show the airborne-lidar-determined peak ozone concentrations vs. 10-hr trajectory displacements. The concentrations again decrease with increasing speed, but are everywhere larger than the surface measurements. Reasons for this discrepancy include that the airborne data represent conditions aloft not subject to surface processes such as removal, that the airborne lidar data represent averaging over shorter intervals, and that on days with stronger winds, plumes are narrower and harder to sample by fixed networks and also high concentrations may be blown out of the surface network before being sampled.

The last issue was expected to be a greater source of discrepancy between airborne and ground-network sampling, because the aircraft samples the pollution plume wherever it is, even though the maximum values may be occurring beyond the surface measurement network before the photochemical reactions are complete. Indeed, individual days have been identified with large discrepancies. However, Fig. B6 shows that, in general, light winds (small 10-hr displacements)


31 August 2007 41

were associated with large O3 enhancements by the Houston area, and stronger wind speeds, with smaller enhancements. Several cases of stronger-wind days have been identified when the total O3 concentrations (background plus add-on) exceeded the 1-hr standard. Those days, most often associated with high background values, are being studied further to determine all processes responsible for the high concentrations.

On such stronger-wind days with high background levels, high pollutant concentrations were generally being advected into the Houston area under easterly or northerly continental flow. Airborne ozone lidar flights sampling air masses entering Texas from the east on three different days found concentrations of 50, 80, and 90 ppb in the inflow air. Another typical scenario is for post-cold-frontal northerly flow to be associated with high pollutant concentrations. When the offshore winds are weak enough, sea-breeze stagnation occurs, and when they are stronger, they are associated with high background. During 2006 all ozone episodes in the month of September occurred in a post-frontal environment. For example, a four-day sequence of ozonesonde profiles (Fig. B7) showed an increase in background ozone within the free troposphere up to 5000 m AGL, corresponding to transport from the continental United States after the passage of the front.

Figure B7. Ozonesonde profiles at ~1200 CST on 28, 29, 30, and 31 August 2006. Passage of cold front occurred on 29 August 2006. Before the passage of the front, marine air masses with O3 concentrations of 30 ppb had concentrations even lower than the EPA boundary-condition benchmark values (40 ppb, red dotted vertical line). In subsequent days, background ozone increased throughout the troposphere up to 5000 m AGL due to the shift in wind direction after the frontal passage.

Many of these episodes were associated with strong elevated inversions. Higher afternoon ozone peaks occurred where there was an extremely dry air mass aloft, although it was not always associated with the strongest inversion observed. The air masses aloft with the strongest winds


31 August 2007 42

(greater than 8 m s-1) and driest air did not have extremely high ozone values associated with them, mainly because these episodes occurred right after a frontal passage where the northerly flow was strong enough to keep ozone peaks in check.

Thus, the Houston-Galveston-Brazoria area produces a tremendous amount of ozone. Even under relatively strong wind conditions, the result can be high pollutant concentrations, although the very highest concentrations are produced on days with weak-wind or sea-breeze-reversal conditions. Even with the 8-hr standard, most, but not all, 8-hr exceedances during the TexAQS II field intensive (and other 2005-6 episodes) occurred when winds were light enough to allow stagnation/recirculation.

Finding B3: After sea breeze days, the Houston plume was broadly dispersed at night through the formation of low-level jets.

Analysis: Nielsen-Gammon-Texas A&M; Banta, Senff, Ryerson, Darby, Tucker, White-NOAA; Lambeth-TCEQ. Data: Nielsen-Gammon-Texas A&M; Senff, Tucker, White, Darby, Angevine-NOAA. Trajectory analysis of nighttime transport of the Houston plume, based on the wind profiler network, indicates that the Houston plume sometimes remains a coherent entity, subject to little wind shear (see Fig. B8a,b), but at other times is dispersed over a broad portion of Texas by strongly sheared winds (Fig. B8c,d). Broad dispersal is favored after sea breeze days when nighttime decoupling allows a strong low-level jet to form from the remains of the sea breeze circulation. The wind speed is strongest at 300-500 m, decreasing above and below.

Figure B8. Overnight trajectories for four elevations above ground level, based on hourly radar wind profiler measurements, for the nights of a) 16-17 August, b) 14-15 August, c) 31 August-1 September, and d) 1-2 September. Red trajectories were averaged over the vertical interval from 200 to 500 m;, black, from 500-800 m; green, from 800-1200 m;, and blue, from 1200-1800 m. Lower trajectories or those ending up to the north of their origin are not likely to be contaminated by signal from migrating birds.

An inland surge of winds found in the 200-600 m layer at night has been called the sea-breeze low-level jet. The essential characteristics of this jet are shown in a sequence of images from an MM5 model forecast from 8 June 2006 (Fig. B9). In the first panel, at 1500 local standard time (LST), the winds are light and onshore. This wind pattern is particularly conducive to the formation of a sea-breeze low-level jet. Along the coast, a sea breeze has developed, and onshore wind speeds are locally 4-7 m s-1. Around sunset, in the second panel, the sea breeze wind maximum has moved inland and has increased in speed to over 9 m s-1 in places. The LLJ occurs in a broad area along the coast, rather than a narrow along-wind band. In the lower left panel, at 2100 LST, the sea-breeze low-level jet is about 200 km inland and has continued to


31 August 2007 43

increase in intensity, with peak wind speeds over 12 m s-1. The leading edge of the jet is very sharp, implying a sudden increase in wind speed, whereas along the trailing edge the decline in wind speed is gradual. Note also that the wind direction, initially southeasterly, has veered to southerly under the influence of the Coriolis force. This veering continues through the night, and the wind direction is southwesterly by 0000 LST (final panel). By this time, with the winds no longer blowing toward lower pressure, the wind speed within the jet has begun to decrease. A similar wind evolution is simulated whenever winds are light and onshore during the day and precipitation is sparse or absent. Sea-breeze low-level jet passages are also evident in wind profiler observations from 2005-6.

Figure B9. Forecasted 300 m winds (indicated with barbs and isotach shading) and pressure (black isobars), 1500 LST 8 June 2006 (upper left), 1800 LST (upper right), 2100 LST (lower left) and 0000 LST, 9 June (lower right). The sea-breeze low-level jet is the band of strong wind speeds originating along the coast at 1500 CST and accelerating inland through the evening.

A dramatic example of nocturnal transport has been documented for 8 September 2006, when high daytime concentrations of ozone that had accumulated over Houston during the previous day traveled to Dallas overnight, contributing to enhanced concentrations in Dallas the next day.


31 August 2007 44

This episode is discussed in more detail in the Response to Question H. Other examples of overnight transport of the Houston plume were identified in data from the instrumented tower at Moody, Texas (near Waco) and in the Tyler-Longview areas of northeast Texas. In a further example from 1 September, the Houston plume moved to the south of the city overnight, and then was brought back in the sea breeze the next day to produce enhanced O3 concentrations in the western suburbs. Calculations of the O3 flux from Houston indicate that if these concentrations were spread over an area 100 by 100 miles by 2 km deep, it would raise the concentrations by 10 ppb over the entire volume. These calculations are further discussed in the Response to Question G.

Finding B4: The Dallas ozone plume can extend well beyond the monitoring network.

Analysis: Senff, Darby, Banta-NOAA; Breitenbach, Lambeth-TCEQ. Data: Senff, Darby, Banta-NOAA; Lambeth-TCEQ. An airborne O3 lidar flight, sampling the DFW plume on 13 September in northerly flow of ~ 5 m s-1, found peak O3 values of 90-95 ppb against a background of ~ 65 ppb (Fig. B10), representing an enhancement of 25-30 ppb, in agreement with previous estimates. These high O3 values were observed in a distinct urban plume extending past the southernmost cross-wind flight leg at a distance of 85 km downwind of DFW, indicating that the plume extended much farther downwind than this. It seems likely that the highest ozone occurs beyond the margin of the monitoring network on such occasions, as was previously noted for the Houston network. Under southerly large-scale flow, recently installed ozone sensors in southern Oklahoma also indicated elevated ozone concentrations resulting from the DFW plume, at a distance of at least 240 km (near Lawton, Oklahoma).

Figure B10. a) Flight track of the Twin Otter airborne ozone lidar, showing color-coded ozone concentrations averaged between 500 and 1000 m MSL (ozone scale from 50 to 100 ppb), for 13 September, a day with stiff north-northeasterly flow. b) Vertical time-height cross section of ozone [ppb, same scale as in a), vertical scale from 0 to 1800 m MSL, horizontal scale from 2145 to 2230 UTC] for southernmost cross-wind leg, showing plume of higher ozone from Dallas-Ft. Worth.


31 August 2007 45

KEY CITATIONS AND INFORMATION AND DATA SOURCES Banta, R.M., C.J. Senff, T.B. Ryerson, J. Nielsen-Gammon, L.S. Darby, R.J. Alvarez, S.P.

Sandberg, E.J. Williams, and M. Trainer. 2005. A bad air day in Houston. Bull. Amer. Meteor. Soc., 86:657-669.

Darby, L.S. 2005. Cluster analysis of surface winds in Houston, Texas and the impact of wind patterns on ozone. J. Appl. Meteor. 44:1788-1806.

Nielsen-Gammon, J.W., J. Tobin, and A. McNeel. 2005. A Conceptual Model for Eight-Hour Ozone Exceedances in Houston, Texas. Part II: Eight-Hour Ozone Exceedances in the Houston-Galveston Metropolitan Area. HARC/TERC/TCEQ Report, 79 pp. http://www.harc.edu/Projects/AirQuality/Projects/Projects/H012.2004.8HRA

Nielsen-Gammon, J.W., C.L. Powell, M.J. Mahoney, W.M. Angevine, C. Senff, A. White, C. Berkowitz, C. Doran, and K. Knupp. 2007. Multisensor estimation of mixing heights over a coastal city. J. Appl. Meteor. Clim., in press.


31 August 2007 46

Response to Question C

QUESTION C Are highly reactive VOC and NOx emissions and resulting ambient concentrations still at the same levels in Houston as they were in 2000? How have they changed spatially and temporally? Are there specific locations where particularly large quantities of HRVOC are still being emitted? Are those emissions continuous or episodic? How well do the reported emissions inventories explain the observed concentrations of VOC and NOx?

BACKGROUND Questions C, D, and E all deal with emissions. Here, Question C specifically addresses highly reactive VOC and co-located NOx emissions in the Houston area. Question D addresses all other ozone and aerosol precursor emissions, biogenic as well as anthropogenic, that are included in emission inventories. Question E addresses evidence for additional, unrecognized sources of precursor emissions.

FINDINGS

Finding C1: There are indications that ethene (the lightest HRVOC) emissions from industrial sources in the Houston area decreased by 40 (±20)%, i.e., by a factor of between 1.25 and 2.5, between 2000 and 2006.

Comparison of ethene/NOx emission ratios measured during TexAQS 2000, TexAQS 2006, and one 2002 aircraft flight

Analysis: de Gouw et al.-NOAA. Data: de Gouw, Ryerson et al.-NOAA; Atlas et al.–U. Miami The emissions of ethene from numerous point sources were individually characterized using the Electra and WP-3D aircraft measurements made in TexAQS 2000, TexAQS 2006, and on one flight in the southeast Texas region in April 2002. Determination of ethene/NOx emission ratios was most straightforward for the isolated petrochemical plants to the south of Houston. These facilities were extensively investigated in 2000 and 2006, and an additional research flight was made in 2002 when the WP-3D transited to a study in California. Figure C1 shows the flight track in 2002 (Ryerson et al., 2003) and an example of a flight made in 2006, along with results of the NOx and ethene measurements. Apart from Texas City, there appear to be changes in the ethene/NOx emission ratios. The results are summarized in Table C1, and show decreases in the ratio by a factor of 3-7 for Sweeny, Freeport, and Chocolate Bayou. Further research is necessary to determine if the lower emission ratios were systematic, or only happened to be lower on the 29 September flight; preliminary analyses have shown that emission ratios are variable within factors of 2-3 between different flights. In addition, the high-time-resolution ethene data from the WP-3D have indicated that ethene and NOx enhancements are not always well correlated. Evidently the sources of these species are sometimes not co-located, making it difficult to define emission ratios in those cases.


31 August 2007 47

Figure C1. Measurements of NOx and ethene downwind from 4 isolated petrochemical plants to the south of Houston. The top panels show the flight tracks of the WP-3D on 22 April 2002, (Ryerson et al., 2003) and on 29 September 2006. The lower panels show the measurement results plotted as a function of longitude.

Table C1. Ethene/NOx emission ratios for petrochemical complexes determined from measurements in 2000, 2002, and 2006, compared with emission inventories.

Ethene/NOx Emission Ratios Inventories Measured 1999 a 2004 b 2000 c 2002 c 2006 Sweeny 0.05 0.019 3.6 1.7 0.5 Freeport 0.05 0.030 1.5 0.62 0.32 Choc. Bayou 0.08 0.048 2.0 1.2 0.62 a TNRCC emission inventory. b TCEQ point source emission inventory with 1999 VOC speciation. c Ryerson et al. (2003.)


31 August 2007 48

Comparison of ethene and formaldehyde concentration distributions during TexAQS 2000 and 2006

Analysis: de Gouw et al.-NOAA. Data: de Gouw, Ryerson, Holloway, et al.-NOAA; Atlas et al.–U. Miami; Fried et al.-NCAR. Figure C2 compares the 2000 and 2006 airborne ethene measurements made below 1000 m altitude inside a box around Houston (Figure C2, panel A). The median ethene from the whole air samples (WAS) was 40% lower in 2006 than in 2000 (Figure C2, panel B). The median ethene from the laser photo-acoustic spectroscopy (LPAS) measurements was even lower; one can show that the targeted filling of WAS inside pollution plumes leads to a bias in the WAS median. A similar comparison of the 2006 data from the Ronald H. Brown while in Barbour’s Cut with the 2000 data from La Porte airport (these two locations are in close proximity) indicates a similar 40% decrease (not shown here).

Figure C2. Cumulative probability diagrams for ethene, wind speed, and temperature measured from the NOAA WP-3D inside the box around Houston shown in panel A and below 1000 m altitude.


31 August 2007 49

The question arises whether the differences between 2000 and 2006 are due to a decrease in HRVOC emissions or to a difference in meteorology. Average wind speeds in 2006 were higher than in 2000 (Figure C2, panel C), which would lead to a greater dilution of ethene emissions and lower mixing ratios in 2006. However, average temperatures in 2006 were lower than in 2000 (Figure C2, panel D). Higher temperatures can be associated with deeper boundary layers and thus a larger volume within which the emissions are mixed and diluted, which would imply higher mixing ratios in 2006.

All emissions are subject to the same meteorology. Comparison of other emitted species between 2000 and 2006 can provide additional insight into possible meteorological differences. If the decrease in ethene between 2000 and 2006 were due to meteorology, then other trace gases should exhibit similar trends. If, on the other hand, the decrease were due to a decrease in emissions, then other trace gases would show ambient concentration trends characteristic of their own emission trends. Figure C3 shows the difference in median mixing ratio between 2000 and 2006 for six species. On-road mobile emissions dominate the emissions of ethyne and CO; their mixing ratios showed only small changes between 2000 and 2006. (The CO mixing ratios were corrected for the observed background mixing ratios.) A decrease in CO is expected because of the modernization of the vehicle fleet between 2000 and 2006; the observed decrease of ~3% per year is slightly smaller than the reported historical trend of 4.6% per year (Parrish, 2006). NOy showed a significant decrease between 2000 and 2006; this is attributed to the installation of emissions reduction technologies in power plants and other industrial NOx point sources (see Finding C3 below). SO2 increased slightly between 2000 and 2006; reductions were not expected as emissions controls have focused on NOx. The largest decreases were observed in the mixing ratios of ethene and formaldehyde (HCHO), its main photoproduct. From this comparison we conclude that the meteorological differences had compensating effects on the two years, and that the decrease in ethene between 2000 and 2006 was due to a reduction in its industrial emissions.

Figure C3. Difference in median mixing ratio for several trace gases measured from the WP-3D inside a box around Houston and below 1000 m altitude.


31 August 2007 50

Long-term data sets

Analysis: Estes et al.-TCEQ; Data: TCEQ. In the Houston area there have been extensive VOC measurements made by as many as eight auto-GC systems and by canister-based methods. Measurements were begun at some sites as early as 1997, giving temporal coverage over one decade by the end of 2006. Figure C4 presents results from two sites near the Ship Channel. The median ambient ethene levels at both sites decreased by about a factor of two between 1997 and 2005. Similar decreases were also seen for propene (not shown). Analysis by Sather and Cavender (2007) also shows significant decreases of ambient concentrations of ethene and propene, in qualitative agreement with Figure C4.

Figure C4. Results of ethene measurements by auto-GCs at two sites near the Houston Ship Channel: 9 years of data from Clinton (on the western end) and 8 years of data from Deer Park (on the eastern end).

Summary of evidence Emission ratios of ethene relative to NOx from several isolated petrochemical plants in 2006 were lower in comparison with results from 2000 and 2002. This decrease in ratios must be due to a decrease in ethene emissions, since the NOx emissions have also decreased in many cases (see Finding G3). Further, lower mixing ratios of ethene and formaldehyde were observed in 2006 in comparison with 2000, and this change cannot be explained by differences in dilution rates between the two years.

Thus, analyses based on four different measured parameters (ethene/NOx emission ratios in plumes of petrochemical facilities, the ambient distribution of ethene concentrations, the ambient distribution of formaldehyde, and long-term auto-GC ethene measurements) have all found evidence for a significant decrease in ethene emissions. The unanimity of these four analyses


31 August 2007 51

increases our confidence that a significant decrease in HRVOC emissions from Houston area petrochemical facilities has actually occurred in the period between 2000 and 2006. Considering all of these analyses, the best estimate is that ethene emissions have decreased by about 40% (i.e., a factor of 1.7) between 2000 and 2006. The much more limited evidence available for propene suggests that emissions of this HRVOC may have decreased similarly.

It is important but difficult to assign confidence limits to the magnitude of the decrease. Our best estimate is that the decrease is 40 ± 20%, which corresponds to a decrease by a factor of 1.25 to 2.5. We consider it unlikely, but cannot categorically exclude the possibility that there has in fact been no significant decrease.

Finding C2: Measurements of ethene emission fluxes from petrochemical facilities during TexAQS 2006 indicate that the 2004 TCEQ point source database underestimates these 2006 emissions by one to two orders of magnitude. Repeated sampling of the same petrochemical facility showed that the ethene emission flux remained constant to within a factor of two.

Analysis and data: de Gouw et al.-NOAA; Mellqvist et al.–Chalmers U. Emissions of ethene were determined during TexAQS 2006 by two completely independent methods. First, column measurements of ethene by absorption of solar IR radiation were obtained from the solar occultation flux (SOF) mobile laboratory operated by Chalmers University (Gothenburg, Sweden). In this method, the ethene emission flux was estimated by combining the column data with measured vertical wind profiles and then integrating across the width of the plume in close proximity to the emission source. Second, ethene was measured onboard the WP-3D aircraft using a new laser photo-acoustic spectroscopy (LPAS) instrument. An example of the resulting data is shown from 25 September when the WP-3D and the SOF van both sampled the emissions from the Mont Belvieu complex (Figure C5).

Figure C5. Transect of the WP-3D just downwind from the Mont Belvieu complex to the northeast of HSC, color-coded by ethene measured by LPAS. Ethene sources from the 2004 TCEQ point source database are indicated by the blue triangles, with the size proportional to the source strength. The measured time series of ethene for this transect is shown on the right.


31 August 2007 52

The ethene flux was estimated from the aircraft measurements by integrating across the width of the plume and by assuming that the plume was homogeneously distributed over the height of the boundary layer. A flux of 280 kg h-1 is the result for this example. The result from the SOF van from the same complex on the same day was 500 kg h-1, i.e., agreement within a factor of 2. Flux estimates from the SOF and WP-3D measurements were compared for different sources (Sweeny, Freeport, Texas City) and generally agreed within a factor of 2, with no systematic difference as to which technique was higher or lower.

The WP-3D determined the flux from Mont Belvieu on a total of 10 downwind transects. Figure C6 shows a histogram of the WP-3D results (grey bars); the average calculated flux is 470 kg h-1 with a standard deviation of 160 kg h-1. Two considerations can serve to put these fluxes in perspective. First, the 2004 TCEQ point source database estimated the ethene emissions from Mont Belvieu to be 29 kg h-1, a factor of 10-40 lower than the values measured during TexAQS 2006. Second, Murphy and Allen (2005) investigated the role of large, accidental releases of HRVOC in ozone formation in the HGB area. They identified 763 HRVOC emission events in a one-year period (31 January 2003 to 30 January 2004). More than half of these events released less than 1000 lbs (454 kg) total HRVOC. Thus, the Mont Belvieu complex, as an example of petrochemical facilities in the HGB area, routinely emits more ethene each hour than the total released in most of the individual accidental release events considered by Murphy and Allen. These facilities represent a significantly larger source of HRVOC and more substantial contribution to ozone formation than indicated by current emission inventories.

Figure C6. Histogram of ethene fluxes determined from aircraft measurements of 10 transects. Also indicated are the SOF result from September 25, the average SOF result from 3 days of measurements, and the 2004 TCEQ point source database.

Finding C3: Close to petrochemical HRVOC sources, the OH reactivity of propene is generally greater than that of ethene.

Analysis: de Gouw et al.-NOAA; Data: Atlas et al.-U. Miami. Close to emission sources, the reactivity of propene generally outweighs that of ethene. To illustrate, Figure C7 shows the OH reactivity of propene (the product of the propene concentration and the rate coefficient for its reaction with OH) versus that of ethene from all the


31 August 2007 53

whole air sample measurements onboard the WP-3D. At high reactivity (>0.5 s-1) the data points are well above the blue line that indicates equal reactivity of ethene and propene. At lower reactivity (<0.1 s-1), the reactivity of ethene is higher than that of propene; the lifetime of ethene is longer by about a factor of 3, and therefore it persists further downwind from sources. Thus, VOC measurements made at routine monitoring sites, generally located relatively distant from the facilities, will not accurately reflect the propene OH reactivity contributions unless the relative lifetimes of the HRVOC are carefully considered.

Figure C7. OH reactivity of propene versus that of ethene from all the data from the WP-3D. The blue line indicates equal reactivity of ethene and propene; the gray area shows the range where the reactivities of ethene and propene differ by no more than a factor of 3, higher or lower.

Finding C4: The latest available emission inventories underestimate ethene emissions by approximately an order of magnitude.

Analysis: Jolly et al.-TCEQ; Data: TCEQ. The TexAQS 2000 study established that inventories underestimated emission fluxes of HRVOC from petrochemical facilities by one to two orders of magnitude (Ryerson et al., 2003.) A remarkable feature of Table C1 and Figure C6 is that appreciation of this finding has not been reflected in the inventory evolution. For example, total HRVOC emissions included in the Harris County Point Source EI for 2000-2004 were fairly steady across those years, with the lowest year (2002, at 3300 tons) being about 83 percent of the highest year (2004, 4000 tons). Total VOCs in the county in the 2000-2005 period differed approximately 13 percent between the lowest year (2003, ~29,000 tons) and the highest year (2000, ~33,500 tons). Consequently, the latest available emission inventories still underestimate HRVOC emissions (as judged by the ethene comparisons summarized in Table C1 and Figure C6) by at least an order of magnitude.

In an effort to improve this situation, TCEQ collected the special 2006 Hourly Emission Inventory during the TexAQS 2006 intensive from 15 August through 15 September. During this period, 141 sites in eastern Texas reported their hourly emissions of VOC, NOx, CO, and SO2 from predetermined industrial sources, which were selected based upon the following criteria: sources subject to HRVOC rules, NOx and sulfur dioxide sources equipped with CEMS,


31 August 2007 54

and emissions sources located near ambient air monitoring sites. For HRVOC the reported emissions were based upon process flow monitoring (flares, cooling towers) that was required beginning in January 2006. This inventory was the first in HGB conducted since compliance deadlines passed for key HRVOC regulations promulgated by the TCEQ. These hourly estimates were made only for a subset of process units and emission points within the reporting plants, and only a subset of plants that normally report in the HGB EI did so in the Hourly EI. Both of these factors add uncertainty to how well this inventory represents the total HGB VOC inventory. In HGB, about 10-11% of the process units and emission points that emit ethene and propene in the annual (periodic) emission inventory were included in the Hourly EI. However, these units accounted for 23-24% of all ethene/propene emissions reported in the annual EI.

0

50,000

100,000

150,000

200,000

250,000

300,000

350,000

400,000

450,000

500,000

Hourly EI OSD EI Hourly EI OSD EI Hourly EI OSD EI Hourly EI OSD EI

Cooling Towers Flares Equipment Leak Fugitives Unknown source type

Lb p

er 3

2 da

ys

Propylene

Ethylene

Butenes

_1_3__Butadiene

The results from the hourly EI suggests that reported emissions of total VOC and HRVOC species may increase in the coming years. Figure C8 shows emissions changes in inventoried emissions from cooling towers and flares, as well as for equipment leak fugitives (which were poorly represented in the hourly EI). The total HRVOC emissions from inventoried flares, increased about 2.4 fold (95 tons to 231 tons), although the much smaller emissions from cooling towers decreased by 28% (9.9 to 7.7 tons in the 32 day period). Overall, the total reported ethene and propene emissions in HGB approximately doubled.

In summary, the discrepancy between the HRVOC emission measured in the field and those included in inventories has improved to some degree between the 2000 and 2006 TexAQS studies. The measured emissions have decreased by about 40%, while the inventoried emissions (at least as indicated by the limited Special 2006 Hourly Emission Inventory) have increased by

Figure C8. HRVOC emissions, by process unit type and compound/group. The special 2006 Hourly Emission Inventory is compared to the Ozone Season Daily Emission Inventory (OSD EI) summed over the 32 days (15 August – 15 September) included in the special inventory.


31 August 2007 55

about a factor of 2. However, in spite of this progress, the latest available emission inventories still underestimate ethene emissions by approximately an order of magnitude.

Finding C5: Inventories for NOx point sources at petrochemical facilities equipped with CEMS appear to be relatively accurate. Substantial decreases in NOx emissions in the Houston Ship Channel are suggested by the inventories, and measurements from aircraft are qualitatively consistent with the NOx decreases.

Analysis: Trainer et al.-NOAA; Data: Ryerson, et al.-NOAA. Data from the WP-3D flights in 2006 compared to similar data from the Electra aircraft in 2000 indicate that substantial reductions in NOx emissions from certain petrochemical facilities have occurred in the intervening period. Figure C9 presents an example of a transect downwind of HSC that was flown under similar conditions during both field studies. Plumes with well-defined correlations between NOy and CO2 concentrations were seen in both years from the east end of the HSC and from the Cedar Bayou complex (at the northern end of Galveston Bay). The slopes of these correlations give the NOx/CO2 emission ratio from the respective sources. These observed slopes compared well to the ratios calculated from the CEMS systems for the times of emissions of these plumes. It is clear that the NOx/CO2 emission ratio decreased dramatically for the Cedar Bayou complex (by a factor of 6) as a result of NOx emission controls implemented at this plant. NOx emissions also decreased detectably (by about 30%) from the facilities at the eastern end of the HSC.

Figure C9. Relationships between NOy and CO2 from flights during TexAQS 2000 and 2006. Linear least squares fits are given for correlations with r2 > 0.80. Slopes of the fits give the NOx to CO2 emission ratios.


31 August 2007 56

KEY CITATIONS AND INFORMATION AND DATA SOURCES Murphy, C.F. and D.T. Allen. 2005. Hydrocarbon emissions from industrial release events in the

Houston-Galveston area and their impact on ozone formation. Atmos. Environ. 39:3785–3798.

Parrish, D.D. 2006. Critical evaluation of US on-road vehicle emission inventories. Atmos. Environ. 40:2288–2300.

Ryerson, T.B., M. Trainer, W.M. Angevine, C.A. Brock, R.W. Dissly, F.C. Fehsenfeld, G.J. Frost, P.D. Goldan, J.S. Holloway, G. Hübler, R.O. Jakoubek, W.C. Kuster, J.A. Neuman, D.K. Nicks, Jr., D.D. Parrish, J.M. Roberts, D.T. Sueper, E.L. Atlas, S.G. Donnelly, F. Flocke, A. Fried, W.T. Potter, S. Schauffler, V. Stroud, A.J. Weinheimer, B.P. Wert, C. Wiedinmyer, R.J. Alvarez, R.M. Banta, L.S. Darby, and C.J. Senff. 2003. Effect of petrochemical industrial emissions of reactive alkenes and NOx on tropospheric ozone formation in Houston, Texas. J. Geophys. Res. 108(D8), 4249, doi:10.1029/2002JD003070.

Sather, M.E. and K. Cavender. 2007. Trends analysis of ambient 8 hour ozone and precursor monitoring data in the south central U.S. J. Environ. Monit. 9:143–150.


31 August 2007 57

Response to Question D

QUESTION D What distribution of anthropogenic and biogenic emissions of ozone and aerosol precursors can be inferred from observations?

BACKGROUND Questions C, D, and E all deal with emissions. Question C specifically addresses highly reactive VOC and co-located NOx emissions in the Houston area. Here, Question D addresses all other ozone and aerosol precursor emissions, biogenic as well as anthropogenic, that are included in emission inventories. Question E addresses evidence for additional, unrecognized sources of precursor emissions.

FINDINGS

Finding D1a: Several rural electric generation units (EGU) in the Houston area and in eastern Texas have substantially decreased their NOx emissions per unit power generated since the TexAQS 2000 study. With one exception, SO2 emissions have not changed appreciably since 2000 for the plants sampled in 2006.

Finding D1b: Comparisons of emissions derived from ambient observations with those measured by continuous emission monitoring systems (CEMS) indicate that the emissions from point sources equipped with CEMS are very accurately known.

Finding D1c: Underreporting of CO emissions at several EGU noted in 2000 (Nicks et al., 2003) has been reconciled by large increases (by factors of 5 to 50) in the inventory values between 2000 and 2006, as a result of newly implemented CEMS monitoring of CO at these plants.

Analysis and Data: Ryerson et al.-NOAA. NOx, SO2, CO, and CO2 are emitted directly, in varying ratios, from electric generation units (EGU). Enhancement of the first three species, relative to CO2 enhancements when sampled in plumes immediately downwind of EGU point sources, provide a measure of pollutant emissions per unit energy generated by the plant (e.g., Ryerson et al., 2003). Comparisons of emissions ratios between the TexAQS 2000 and TexAQS 2006 studies permit an assessment of EGU emissions control strategies, intended primarily to reduce NOx emissions, that have been implemented since 2000.

Near-field plumes from numerous rural EGU in Texas were characterized using aircraft measurements in 2000 and 2006. Data were generated as shown in Figure D1a, which depicts the NOAA WP-3D ground track for a flight designed to assess several large EGU point sources in the eastern Texas area. The data are taken from the closest transects, within 10 km downwind of the plants, and plotted in Figure D1b as enhancement ratios versus CO2. The slopes of linear fits to these data provide a direct measure of the plant emissions ratios.

Analyses of rural EGU emissions ratios to CO2 have been performed for Monticello, Welsh, Martin Lake, Big Brown, and W.A. Parish power plants. Table D1 compares derived emission ratios from the 2006 WP-3D data with the ratios measured from the NCAR Electra aircraft in 2000.


31 August 2007 58

The TexAQS 2000 study demonstrated quantitative agreement between emissions estimates from the Electra aircraft data and the tabulated emissions from Continuous Emissions Monitoring Systems (CEMS) data, for NOx and SO2 at each plant. Some variability is expected on time scales of hours to years, and is reflected in the data in Table D1. Nonetheless, conclusions from the 2006 study can be drawn that are well outside of the uncertainties due to normal emissions variability over time.

34.0

33.5

33.0

32.5

32.0

Latit

ude

-95.5 -95.0 -94.5 -94.0

Longitude

MonticelloWelsh

Martin Lake

10

8

6

4

2

0N

Ox em

issions, 1024 m

olec/sec

Figure D1a. Sept. 16, 2006 WP-3D ground track (green) with observed SO2 enhancements (blue) plotted along the track.

120

100

80

60

40

20

0

NO

x an

d SO

2, p

pbv

7:48 PM9/16/06

7:50 PM 7:52 PM

Time, GMT

420

410

400

390

380C

O2,

ppm

v

200

180

160

140

120

100

80

60

CO

, ppbv

SO2 CO CO2 NOx

Welsh

Monticello

Figure D1b. Time series of CO2, SO2, NOx, and COmeasured immediately downwind of Welsh and Monticello.

Table D1. Measured emissions relative to CO2 for EGU in East Texas.

NCAR Electra aircraft data

2000

NOAA WP-3D aircraft data

2006

EGU name SO2 CO NOx SO2 CO NOx

NOx emissions decreased by factor of:

Monticello 3.5 6.4 1.0 2.8 5.4 0.80 1.25 Welsh 1.5 1.7 0.80 1.7 1.7 1.20 1.5 (increase) Martin Lake 1.4 4.0 1.3 3.0 6.1 0.80 1.6 Big Brown 4.8 2.9 1.5 7.8 6.8 0.66 2.3 W.A. Parish 2.1 (variable) 0.88 2.1 (variable) 0.25 3.5 Emissions values presented as molecules per 1000 molecules of CO2 emitted. Analysis of all available transects of the Parish power plant plume shows that inventory values derived from CEMS data and estimates from the WP-3D are in quantitative agreement. Figure D2 shows that both the magnitude and the variability in time of emissions are captured by the data in Texas 2006 for both SO2 and NOx.


31 August 2007 59

3.0

2.5

2.0

1.5

1.0

SO2/C

O2 (

ppb/

ppm

)

9/21/06 10/1/06 10/11/06Date

SO2/CO2 CEMSP-3 data

0.6

0.5

0.4

0.3

0.2NO

x/C

O2 (

ppb/

ppm

)

9/21/06 10/1/06 10/11/06Date

NOx/CO2 CEMSP-3 data

Conclusions from this body of evidence are: • NOx emissions have decreased substantially in several EGU, by factors of 2–4,

qualitatively consistent with NOx controls implemented since the 2000 study. NOx emissions from other EGU are essentially unchanged.

• SO2 emissions are generally unchanged between 2000 and 2006 in the plants studied to date. The small variability observed is within that expected from CEMS data due to normally changing plant loads. An exception is the Martin Lake plant, where SO2 appears to have increased by a factor of 2 relative to CO2 compared to the 2000 study.

• Large CO emissions discrepancies, noted in the 2000 study, have been reconciled by substantial increases in the inventory CO values for several EGU in east Texas.

• Quantitative agreement between inventory values and aircraft estimates of NOx/CO2 and SO2/CO2 emissions ratios from rural EGU suggests that emissions from point sources equipped with CEMS are very well known.

Finding D2: On-road mobile emission inventories developed from MOBILE6 have significant shortcomings. MOBILE6 consistently overestimates CO emissions by about a factor of 2. It accurately estimated NOx emissions in the years near 2000, but it indicates decreases in NOx emissions since then, while ambient data suggest NOx emissions have actually increased. Consequently in 2006, NOx to VOC emission ratios in urban areas are likely underestimated by current inventories.

Analysis: Parrish et al.-NOAA; Data: TCEQ, Lefer et al.-U. Houston. Figure D3 compares CO to NOx ratios from ambient measurements with those from emission inventories. The Dallas and Houston routine ambient data are in excellent agreement with the nationwide AIRS data. The TexAQS 2006 data from Moody tower agree reasonably well with the routine monitoring data. Significant differences are seen in El Paso and San Antonio, which have older vehicle fleets.

Figure D2. Ratios of SO2 and NOx to CO2 emitted from the Parish power plant plume during six weeks of the TexAQS 2006 study. The ratios derived from CEMS data are compared to the determinations from multiple near-field transects of the emission plume by the WP-3D aircraft.


31 August 2007 60

The HGB inventory overestimates the CO to NOx emission ratio, and that overestimate becomes worse with time as the inventory does not show a significant temporal decrease. The HGB points are from the “SIP Quality” emission inventories (www.tceq.state.tx.us/assets/public/implementation/air/am/committees/pmt_set/20070424/20070424-rthomas-0506EI_Modeling_Update.pdf; accessed 17 August 2007). It should be noted that the on-road emissions inventory value for 2000 was calculated with actual data for 2000, whereas the on-road mobile inventories for later years are projections, based upon projected population growth, economic data, and estimates of expected emission decreases. The projections, therefore, are more uncertain than the base case 2000 emissions. Parrish (2006) showed that the rapid decrease (6.6%/yr) in the ratio is partially due to a slower decrease in CO emissions (4.6%/yr), which implies a significant increase in NOx emissions (approximately 2%/yr). The large inventory overestimates in the CO to NOx ratio at the present time are attributed to a factor of 2 overestimate in CO emissions, and an underestimate in present NOx emissions. This will cause NOx to CO emission ratios in urban areas, which are often dominated by on-road mobile emissions, to be underestimated by current emission inventories.

Urban emission ratios sampled by the WP-3D aircraft in 2006 and the NCAR Electra aircraft in 2000 are consistent with measurements carried out at a Houston highway tunnel in 2000 (McGaughey et al., 2004). These measurements demonstrate the weekday increase in CO/CO2 and CO/NOx emission ratios from midday to the afternoon rush hour correlated with increases in the proportion of gasoline vehicles during rush hour. While aircraft and tunnel measurements were not made during the morning rush hour, their afternoon rush hour emission ratios are consistent with the routine monitoring data for the morning rush hour. Similar to the routine monitors, the aircraft and tunnel observations indicate that inventories overestimate mobile source CO by at least a factor of 2.

The VOC to NOx ratio in on-road emissions is of more critical importance to photochemical modeling in the HGB area than is the CO to NOx ratio. Parrish (2006) showed that VOC

Figure D3. Determination of CO to NOx ratio in Texas on-road mobile emissions from monitoring data (solid symbols) compared to the HGB emission inventory (open symbols), color-coded according to urban area. The black symbols are for all stations in the EPA AIRS network.


31 August 2007 61

emissions from on-road vehicles have decreased at a similar rate to CO emissions. This correspondence is expected because catalytic converters are the principal control measure for both species. Although MOBILE6 accurately calculated the VOC to NOx ratio in 2000, the decrease in the ratio is not accurately captured. Consequently, in 2006, NOx to VOC emission ratios in urban areas are likely underestimated by current inventories.

Finding D3: Emissions from ships constitute a significant NOx source in the HGB region. Literature results provide accurate emission factors for inventory development.

Analysis and Data: Williams et al.-NOAA. During TexAQS 2006, emissions in exhaust plumes from over 200 marine vessels in Galveston Bay and the Houston Ship Channel were measured on the Ronald H. Brown. Table D2 presents the average derived emission factors for slow speed diesel (SSD) engines, which are those with maximum power greater than ~10 MW, and medium speed diesel (MSD) engines, which are of lower power. The NOx values are within 20% of the average values reported in the Lloyd's (1995) study for both MSD and SSD engines although the measured variability is large in both cases. These data are sufficient to provide emission factors classified by ship type (e.g., freighters, container ships, tankers, tugs, etc.). It is concluded that Lloyd's (1995) provides an accurate characterization of NOx emissions from underway vessels in the HGB region.

Table D2. Summary of average marine vessel emission factors.

Engine type NOx CO SO2 H2CO LAC Medium speed diesel (MSD)

60 8.7 9.1 0.17 0.41

Slow speed diesel (SSD) 74 6.6 28 0.20 1.16 All units are grams of species per kilogram fuel consumed. Emission factors for CO are median data. The emission factors for CO are within 20% of the value reported by Lloyd's (1995). There is no trend of increasing CO at lower vessel speeds (used as a surrogate for engine load), which was seen in the Lloyd's data. Measurements of formaldehyde (H2CO) emissions from ships show little distinction between MSD and SSD engines; emission of H2CO is less than 5% of the emission of CO. Emission factors for SO2 vary with fuel sulfur content. In the HGB region the mean fuel S derived from the measurements is 0.46% for MSD engines and 1.4% for SSD engines. Measurements of light absorbing carbon (LAC; also known as black carbon) were also derived.

Given detailed activity data (i.e., marine fuel consumption) for ships in the HGB region, an emissions inventory for this source could be constructed. The 2007 report from Eastern Research Group (ERG) to TCEQ (Eastern Research Group, 2007) has such data, but there appears to be a significant underestimate (factor of 2-8) of fuel consumption, when compared to an estimate from a more comprehensive model (Wang et al., 2007). The NO2 emission factors used in the ERG report agree with our data to within 10%. Thus, for current ship emissions inventory modeling, there is less uncertainty contributed by emission factors than by activity data.


31 August 2007 62

Table D3 gives emissions of NOx, CO, and SO2, relative to the emission of CO2, from ships, compared to similar emission ratios from electric power generating units (EGU) from the 2004 point source emission inventory, updated to 2006 with CEMS data. This comparison shows that ships emit 10 to 100 times more NOx (and somewhat more CO, and SO2) per unit fuel burned (i.e. CO2 emitted) than large stationary sources. Though the emissions from an individual vessel might be 10-100 times lower than from an EGU, the volume of ship traffic in the HGB region is sufficient that emissions from commercial shipping, in aggregate, cannot be neglected. Accurate fuel consumption or other ship activity data are needed to accurately quantify these emissions. Importantly, while emissions from stationary sources are a focus of ongoing control measures, emission controls on commercial shipping are not likely to be implemented because of technical constraints and complications arising from international law.

Table D3. Comparison of marine vessel and power plant emissions.

Emission source NOx/CO2 CO/CO2 SO2/CO2 Commercial vessel, SSD 22 6.1 6.1 Commercial vessel, MSD 18 8.1 2.0 Electric power generating unit W. A. Parish 0.23 0.51 1.9 Welsh 0.88 1.4 1.8 All units are molecules of species per 1000 molecules of CO2.

Finding D4: Mixing ratios of isoprene over Texas measured from the WP-3D were used for evaluation of the BEIS-3 emission inventory. On average, the isoprene emissions from the inventory and emissions derived from the measurements agree within a factor of ~2. There may be areas south of Dallas-Fort Worth and southwest of Houston, where isoprene emissions are lower than indicated by the BEIS3 inventory based upon the biogenics emission land cover data (BELD-3.0).

Analysis and Data: de Gouw, Warneke et al.-NOAA. Isoprene was measured from the WP-3D aircraft both in-situ by PTR-MS and off-line by laboratory analysis of whole air samples (WAS) collected in flight. Both measurements agreed within the respective measurement uncertainties. The PTR-MS has the highest time resolution, and thus spatial resolution, and the data illustrate the large variability of the isoprene concentration due to the in-homogeneity of its sources and its extremely short photochemical lifetime.

The PTR-MS data allowed the isoprene emissions to be estimated along the flight track using the measured height of the boundary layer (BL) and the concentration of OH radicals calculated from the parameterization of Ehhalt et al. (1998). The emissions were then converted to standard temperature and photo-active radiation using the formalism from Guenther et al. (1995) and compared to the BEIS3 emission inventory extracted along the flight track. The results for a flight on September 16th over NE Texas are shown in Figure D4. On this flight, the isoprene emissions from BEIS3 and those estimated from the aircraft data agree within a factor of ~2. A better agreement cannot be expected given the uncertainties in some of the relevant factors for this comparison (BL height, OH concentrations, and vertical gradients in the BL).


31 August 2007 63

We also used the BEIS3 isoprene emission inventory as input for the Lagrangian particle dispersion transport model FLEXPART (Stohl et al., 2003). A comparison between isoprene measured by PTR-MS and modeled using FLEXPART after an average transport time of 1 hr is shown in Figure D5. As before, the model and measurements agree within a factor of ~2. Work is in progress to compare measured isoprene mixing ratios to 3-D chemistry-transport models; however, for a good comparison the models need to get boundary layer heights and OH levels correct, which is not true in all cases.

Figure D4. Isoprene emissions for a flight of the NOAA WP-3D on 16 September over NE Texas. Emissions extracted from the BEIS3 inventory based upon the biogenics emission land cover data (BELD-3.0) along the flight track are shown in red. Emissions derived from the PTR-MS data are shown in blue.

Figure D5. Comparison between isoprene mixing ratios measured by PTR-MS from the WP-3D on 16 September, and those modeled by FLEXPART.


31 August 2007 64

Comparisons between isoprene measurements and FLEXPART model results have been made for all flights. Figure D6 shows a map of the flight tracks in the boundary layer color-coded by the differences between measurements and model. The flight tracks in red indicate areas where measurement and model agree well. Flight tracks in yellow indicate areas where modeled isoprene was higher than the measurements; in particular to the southwest of Houston there is an area where this is the case. A possible explanation includes a difference between the actual land use and the land use assumed in the BEIS3 inventory; work is in progress to investigate this possibility. There is also an area in between Dallas-Fort Worth and Houston where the model indicates higher isoprene than the measurements. This area, however, was not as frequently sampled. Flight tracks in black indicate areas where the modeled isoprene was lower than the measurements. This is the case for NE Texas, but because isoprene was very high over this region, the difference of -400 pptv only represents a small relative difference between measurements and model.

Figure D6. Flight tracks of the NOAA WP-3D aircraft in the boundary layer color-coded by the difference between modeled and measured isoprene mixing ratios.


31 August 2007 65

Finding D5: The speciation of VOC from mobile sources in the Houston and Dallas-Forth Worth areas agrees with detailed measurements in the northeastern U.S. However, initial results suggest that the agreement with the NEI-99 emission inventory is poor.

Analysis and Data: de Gouw, Warneke et al.-NOAA. For specific wind directions, the emissions from urban, mostly mobile, sources in Houston are spatially separated from industrial emissions from the Ship Channel area. Specific WP-3D flights were selected based on wind direction to determine the speciation of VOC from mobile sources, using airborne data from the PTR-MS and WAS instruments. In addition, data from the flights downwind from Dallas-Fort Worth were studied. Emissions ratios of VOC versus CO were calculated from the airborne data and compared with the results of a similar study downwind from New York City in 2004 for different classes of compounds (Figure D7). There was a good quantitative agreement between urban emissions in Houston, Dallas, and New York City, which is not surprising. Warneke et al. (2007) showed previously that the VOC composition of urban emissions in New York City and Boston are not represented well in the NEI 99 emissions inventory. Initial results suggest that the same is true for Houston and Dallas-Fort Worth.

Figure D7. Emissions ratios (ER) of various VOC to CO from urban, mostly mobile, sources in Houston and Dallas-Fort Worth in 2006 versus the same ratios from New York City in 2004.


31 August 2007 66

KEY CITATIONS AND INFORMATION AND DATA SOURCES Eastern Research Group. 2007. Houston/Galveston Routine Vessel Identification and Traffic

Study, ERG Rep. No. 0196.00.023, TCEQ Contract No. 582-04-65564-23, January 26, 2007.

Ehhalt, D.H., F. Rohrer, A. Wahner, M.J. Prather, and D.R. Blake. 1998. On the use of hydrocarbons for the determination of tropospheric OH concentrations, J. Geophys. Res. 103:18981-18997.

Guenther, A., C. Hewitt, D. Erickson, R. Fall, C. Geron, T. Graedel, P. Harley, L. Klinger, M. Lerdau, W. McKay, T. Pierce, B. Scholes, R. Steinbrecher, R. Tallamraju, J. Taylor, and P. Zimmerman. 1995. A global model of natural volatile organic compound emissions. J. Geophys. Res. 100(D5):8873-8892.

Lloyd's Register Engineering Services. 1995. Marine Exhaust Emission Research Programme. Lloyd's Register House, London.

McGaughey, G.R., N.R. Desai, D.T. Allen, R.L. Seila, W.A. Lonneman, M.P. Fraser, R.A. Harley, J.M. Ivy, and J.H. Price. 2004. Analysis of motor vehicle emissions in a Houston tunnel during the Texas Air Quality Study 2000. Atmos. Environ. 38(20):3363-3372.

Nicks Jr., D.K., J.S. Holloway, T.B. Ryerson, R.W. Dissly, D.D. Parrish, G.J. Frost, M. Trainer, S.G. Donnelly, S. Schauffler, E.L. Atlas, G. Hübler, D.T. Sueper, and F.C. Fehsenfeld. 2003. Fossil-fueled power plants as a source of atmospheric carbon monoxide. J. Environ. Monit. 5:35-39.

Parrish, D.D. 2006. Critical evaluation of US on-road vehicle emission inventories. Atmos. Environ. 40(13):2288-2300.


Stohl, A., C. Forster, S. Eckhardt, N. Spichtinger, H. Huntrieser, J. Heland, H. Schlager, S. Wilhelm, F. Arnold, and O. Cooper. 2003. A backward modeling study of intercontinental pollution transport using aircraft measurements. J. Geophys. Res. 108, 4370, doi:10.1029/2002JD002862.

Wang, C., J.J. Corbett, and J. Firestone. 2007. Modeling energy use and emissions from North American shipping: Application of the ship traffic, energy, and environment model. Environ. Sci. Technol., doi:10.1021/es060752e.

Warneke, C., S.A. McKeen, J.A. de Gouw, P.D. Goldan, W.C. Kuster, J.S. Holloway, E.J. Williams, B.M. Lerner, D.D. Parrish, M. Trainer, F.C. Fehsenfeld, S. Kato, E.L. Atlas, A. Baker, and D.R. Blake. 2007. Determination of urban volatile organic compound emission ratios and comparison with an emissions database. J. Geophys. Res. 112, D10S47, doi:10.1029/2006JD007930.


31 August 2007 67

Response to Question E

QUESTION E Are there sources of ozone and aerosol precursors that are not represented in the reported emissions inventories?

BACKGROUND Questions C, D, and E all deal with the evaluation of emission inventories. Question C specifically addresses highly reactive VOC and co-located NOx emissions in the Houston area. Question D addresses all other ozone and aerosol precursor emissions, biogenic as well as anthropogenic, that are included in emission inventories. Here, Question E addresses evidence for ozone and aerosol precursor emissions not presently represented in inventories.

FINDINGS

Finding E1: The observed mixing ratios and regional distribution of ambient formaldehyde are broadly consistent with daytime photochemical production from reactive VOC. An upper limit for primary formaldehyde emissions from mobile sources is obtained from nighttime measurements, and is small in comparison with the secondary, daytime formation.

Analysis: de Gouw et al.-NOAA, Herndon et al.-Aerodyne, Rappenglück et al.-U. Houston; Data: Herndon et al.-Aerodyne, Fried et al.-NCAR, de Gouw et al.-NOAA, Atlas et al.-U. Miami, Rappenglück et al.-U. Houston. Diurnal variations in the concentrations of ethene, propene, and formaldehyde were measured at several locations in the HSC. Figure E1 shows the data from Barbour’s Cut, at the eastern end of the HSC. The mixing ratios of ethene and propene (VOC from direct emission) were highest at night, when photochemistry was absent and emissions of these species accumulated in a shallow boundary, and lowest during the day, when the boundary layer was deeper and alkenes were removed by reaction with OH. In contrast, formaldehyde is highest during the day – indicating that daytime photochemistry was the dominant source of this ozone precursor in the HSC area. Similar diurnal patterns also were observed at the Moody Tower site at the University of Houston.

Figure E1. Average diurnal variation of mixing ratios of ethene, propene, and formaldehyde measured from the Ronald H. Brown over approximately ten days of measurements in Barbour’s Cut.


31 August 2007 68

Nighttime measurements onboard the Ronald H. Brown (RHB) and the WP-3D and at the Moody Tower site provide a useful means to estimate the direct emissions of formaldehyde from mobile sources. The left panel of Figure E2 shows the relationship of formaldehyde to CO measured in the turning basin at the western end of the HSC with winds coming from the central urban Houston area. The data are color-coded by the measured photolysis rate of formaldehyde; the nighttime points, in black, show a clear correlation between formaldehyde and CO with a slope of about 3 pptv ppbv-1. After sunrise, the formaldehyde to CO ratio increased strongly due to photochemical formation of formaldehyde (daytime points in yellow). Similar observations were made onboard the WP-3D during a missed approach at Montgomery airport to the north of Houston during the night (Figure E2, right panel). In this graph the data are color-coded by altitude. Below 200 m the aircraft penetrated a shallow layer that was strongly impacted by local, primary emissions; the formaldehyde to CO ratio in this layer was 1.8 pptv ppbv-1. Above the layer, the aircraft sampled the remnants of the daytime boundary layer that is isolated from primary emissions, and the formaldehyde to CO ratio was much higher. The corresponding slope from the Moody Tower measurements is somewhat higher (5 to 7 pptv ppbv-1); the cause and implications of this larger slope is under investigation. From these observations we estimate the primary emissions of formaldehyde from mobile sources to be 0.18-0.3 percent of the CO emissions; this estimate is an upper limit, since it is not possible to exclude the possibility that the sampled air had been photochemically processed to at least some extent during the preceding daytime period. These findings are consistent with the conclusions of Anderson et al. (1996) who found formaldehyde to CO emission ratios of 0.10 to 0.14 percent in measurements made in Denver, Colorado.

Figure E2. Relationships of HCHO to CO concentrations obtained onboard the Ronald H. Brown in the turning basin at the western end of the HSC (left panel), and from the WP-3D during a nighttime, missed approach at Montgomery County airport to the north of Houston.

Finding E2: Concentrated plumes of ammonia were observed occasionally in the Houston Ship Channel area. These plumes often led to the formation of ammonium nitrate aerosol.

Analysis and Data: Nowak et al.-NOAA. Ammonium nitrate aerosol is formed from the reaction of gas-phase ammonia (NH3) and nitric acid (HNO3). Anthropogenic emissions of NH3 and NOx, which in sunlight is oxidized to form


31 August 2007 69

HNO3, can result in elevated concentrations of ammonium nitrate. Sources of NH3 in the Houston area are thought to include motor vehicles, industrial facilities, outlying agricultural activity, and possibly power plants. High-time-resolution (≈1 s average) NH3 measurements were made from the WP-3D aircraft by a Chemical Ionization Mass Spectrometry technique and from the Ronald H. Brown by quantum cascade laser absorption during TexAQS 2006 with the goals of characterizing sources and examining the effect of NH3 on atmospheric aerosol formation.

Figure E3 shows the altitude profile of all 1-second average NH3 mixing ratios measured aboard the WP-3D aircraft during the TexAQS 2006 study. Typically, NH3 mixing ratios over the urban area ranged from 0.2 to 3 ppbv, and generally decreased with increasing altitude. Though infrequent, plumes with NH3 mixing ratios from 5 to greater than 50 ppbv (highest values not indicated in Figure E3) were observed in the boundary layer below 1 km altitude. These plumes were encountered over the Houston metropolitan area in the vicinity of the HSC, around Beaumont, and in St. James Parish, Louisiana. NH3 mixing ratios as high as several hundred ppbv were measured in the HSC from the RHB.

It is difficult to trace the sources of plumes with high NH3 concentrations to particular industrial facilities. Figure E4 illustrates observations made in one of these plumes intercepted approximately 13 km downwind from the Cedar Bayou Station power plant. This plant is the location of the only significant upwind NH3 point source listed in the National Emission Inventory of 1999 (NEI99v3). However, the lack of a strong correlation of the concentration of NH3 with the enhancements of the CO2 and NOx concentrations (upper panel of Figure E4) suggests the NH3 is not released in the Cedar Bayou stack emissions. Future analysis will focus

Figure E3. TexAQS 2006 NH3 observations from the WP-3D plotted as function of altitude.


31 August 2007 70

on using the 2006 Toxic Release Inventory along with measurements of VOC made aboard the WP-3D aircraft to unambiguously identify this NH3 plume source.

The observed NH3 enhancements were typically found to be accompanied by corresponding increases in the particulate nitrate (NO3

-) and decreases in HNO3 mixing ratios as illustrated in the bottom panel of Figure E4. These correlated variations indicate the formation of ammonium nitrate. The magnitude of the observed HNO3 lost and NO3

- formed is consistent (within a factor of 2) with ammonium nitrate formation. The air quality implications of such ammonia plumes should be considered, particularly during cooler wintertime months when the ammonium nitrate will make a longer-lived contribution to the PM2.5 concentrations.

Power plants that have installed Selective Catalytic Reduction (SCR) units constitute a possible NH3 source. This process adds aqueous NH3 to the exhaust gases as a reagent to decrease NOx emissions. NH3 “slippage,” i.e., unwanted emissions of NH3 into the atmosphere, occurs when exhaust gas temperatures are too low for the SCR reaction to proceed to completion, or when too much NH3 is added. The W.A. Parish electric generating facility is equipped with these units. To characterize emissions from this plant, the WP-3D aircraft sampled the Parish plume on numerous flights during TexAQS 2006; Figure E5 shows a time series of measurements from one example transect of the plume. Clear enhancements of CO2, NOy, and SO2 coincide with a depletion in the ozone concentration due to its reaction with the freshly emitted NOx, while no difference in NH3 (black line) mixing ratios can be discerned in or out of the plume. The lack of NH3 enhancement in the power plant plumes sampled by the WP-3D indicates that NH3 slippage was not significant during any of the TexAQS 2006 plume transects.

Figure E4. Time series of concentration measurements of primary emissions (NH3, NOx, CO2) and secondary products (HNO3 nitrate aerosol) during a transect of the Cedar Bayou power plant plume.


31 August 2007 71

Finding E3: Concentrated plumes of gaseous elemental mercury from at least one point source were observed repeatedly in the Houston Ship Channel area and once in the Beaumont-Port Arthur area. The sources of the plumes could not be identified with current inventory sources of mercury.

Analysis and Data: Ryerson et al.-NOAA. High-time-resolution (≈1 s) measurements of gaseous elemental mercury (Hg) were made from the Ronald H. Brown during TexAQS 2006. Figure E6 shows examples of the plume encounters in the HSC and Beaumont industrialized area. The magnitude of the detected plumes varied widely; the HSC plume in Figure E6 was detected during each of four transects of the HSC under southerly to easterly winds, but the magnitude of the plume varied by a factor of approximately 25. During the plume transects, the measured Hg concentration did not correlate with the measured concentrations of any other measured species. This lack of correlation rules out most known sources of Hg emissions, which can be expected to co-emit one or more of the wide range of chemical or aerosol species that were measured concurrently on the RHB. Further, the measured Hg concentrations are not consistent with the latest TCEQ or EPA AIRS, eGRID, and TRI inventory source locations for known Hg sources in the HSC region. A potential source not included in these inventories would be re-release of Hg from historically contaminated soils. More research into this possibility is needed to better understand the unexpectedly high concentrations of Hg observed in coastal industrial areas in eastern Texas.

Figure E5. Time series of NH3, NOy, SO2, CO2, and O3 observations during transect of the W.A. Parish power plant plume.


31 August 2007 72

80

60

40

20

0

8:30 PM8/9/06

8:40 PM

Time, UTC

Beaumont-Port Arthur

1-second Hg(0) data

250

200

150

100

50

0

Hg,

ng/

m3

7:00 PM8/2/06

7:04 PM 7:08 PM

Time, UTC

HoustonShip Channel

1-second Hg(0) data

KEY CITATIONS AND INFORMATION AND DATA SOURCES Anderson, L.G., J.A. Lanning, R. Barrell, J. Miyagishima, R.H. Jones, and P. Wolfe. 1996.

Sources and sinks of formaldehyde and acetaldehyde: An analysis of Denver’s ambient concentration data. Atmos. Environ. 30:2113-2123.


Wert, B.P., M. Trainer, A. Fried, T.B. Ryerson, B. Henry, W. Potter, W.M. Angevine, E. Atlas, S.G. Donnelly, F.C. Fehsenfeld, G.J. Frost, P.D. Goldan, A. Hansel, J.S. Holloway, G. Hübler, W.C. Kuster, D.K. Nicks Jr., J.A. Neuman, D.D. Parrish, S. Schauffler, J. Stutz, D.T. Sueper, C. Wiedinmyer, and A. Wisthaler. 2003. Signatures of terminal alkene oxidation in airborne formaldehyde measurements during TexAQS 2000. J. Geophys. Res. 108(D3), 4104, doi:10.1029/2002JD002502.

Figure E6. Measurements of gas-phase elemental mercury (Hg). Note the change in scales between the two graphs. Periods of missing data are due to automatic instrument calibration or zero procedures.

Figure E6. Measurements of gas-phase elemental mercury (Hg). Note the change in scales between the two graphs. Periods of missing data are due to automatic instrument calibration or zero procedures.


31 August 2007 73

Response to Question F

QUESTION F How do the mesoscale chemical environments (NOx-sensitive ozone formation vs radical-sensitive ozone formation) vary spatially and temporally in Houston, Dallas, and eastern Texas? Which mesoscale chemical environments are most closely associated with high ozone and aerosol?

BACKGROUND

Determination of NOX versus VOC (or Radical-Formation) Sensitive Ozone Formation The accurate prediction of the relative response of ozone concentrations to future reductions in NOx and VOC emissions is a much sought, but very elusive goal. The prediction is central to the very important SIP-relevant question of “direction of control” – that is, should ozone control efforts in an ozone non-attainment area be focused on: a) decreasing emissions of NOx alone, b) decreasing emissions of VOC alone, or c) decreasing emissions of both NOx and VOC? The following paragraphs discuss some important considerations concerning this question.

In the period between the two TexAQS field studies, a change in the NAAQS for ozone was implemented, with at least two important implications for evaluating the “direction of control” question. The older standard, based upon a relatively high (120 ppbv) ozone concentration averaged over a short period (one hour), has been supplanted by one based on a lower (80 ppbv) concentration averaged over a longer period (eight hours). One implication is that the older, shorter-period average was much more amenable to analysis based upon in situ observations; the “snap shot” of the relationships between simultaneously measured concentrations of ozone, its precursors, and other photochemical products provided direct clues to the limiting precursor in an observed ozone exceedance. The longer-period average requires a more comprehensive analysis of the integrated accumulation of ozone over the full 8-hour exceedance period. A second implication is that the background ozone transported into an urban area will constitute a much larger fraction of the ambient ozone concentration constituting an 8-hour exceedance than was the case for a one-hour exceedance. Thus, much more attention must be paid to the effect of control strategies in upwind regions.

Some approaches to determining the “direction of control” question rely, in effect, upon comparing the response of ambient ozone to differential reductions in VOC versus NOx precursor emissions. Such approaches give useful guidance for effectively achieving incremental improvements in ozone air quality; however such approaches cannot determine if compliance with the NAAQS can be reached through such incremental emission reductions, and do not identify the most effective emission control approach for reaching compliance.

A rigorous approach to answering all aspects of the “direction of control” question would require a thoroughly tested Eulerian air quality model that treats the region of interest with sufficient accuracy. Multiple simulations would then be run with that “perfected” model to test possible NOx and VOC emission control strategies. These simulations would then provide the required ozone response information. Unfortunately however, it is not yet generally possible to develop such models because of deficiencies and uncertainties in many critical areas, including emissions, meteorological modeling, and photochemical mechanisms. The thorough testing of


31 August 2007 74

the model must verify that the model adequately reproduces the observed ozone for the correct reasons. This testing would include at least two model-measurement comparisons: first, the model’s ability to reproduce observed relationships between the concentrations of ozone and its precursors and, importantly, other secondary photochemical products (organic nitrates, nitric acid, formaldehyde and other oxygenated VOC, peroxides), and second, the model’s ability to reproduce the observed ambient concentrations of the radicals that drive the photochemical reactions. The first comparison should pay particular attention not only to the slopes and correlation coefficients of the relationships, but also to the range of the observed and modeled concentrations of ozone and the related species. From this perspective, the air quality community is still in the early stages of the testing process; a process that is limited both by resources for model development, and by the availability of observational data with which to compare the models.

The eastern Texas region provides particular advantages, as well as particular challenges, for achieving the required model development and application. The TexAQS 2000 and TexAQS II field studies have collected data sets that provide unprecedented opportunities for model testing. However, experience has shown that development of needed emission inventories is a challenge that has not been met in all respects (see discussion in Responses to Questions C, D, and E.) A further challenge is presented by the concentrated plumes that determine the highest ozone values in the Houston area (see discussion in Response to Question A.) Such resolution may be a severe challenge for Eulerian models, especially if they attempt to include the whole eastern Texas region. At this point it is not possible to provide a rigorous answer to the “direction of control” question.

In the absence of a perfected model, it has been necessary to take a variety of heuristic approaches to provide guidance to air quality managers. Heuristic implies a model that is simplified, but that is designed and used to learn more about some specific but important aspect of the more complex air-quality system that is to be managed. In this report the closely related Questions F and K both deal with aspects of this issue. Here in Question F the application of both observation and modeling based approaches for approximately determining the sensitivity of ozone production will be discussed; the photochemical environment of the entire eastern Texas region will be considered. Question K presents an EKMA model-based approach to address the same issue, but from a different perspective; its application is restricted to the HGB area.


31 August 2007 75

Figure F1. Relationship between ozone and formaldehyde concentrations observed from the Electra aircraft on 1 September 2000 and modeled by a Lagrangian plume model (Wert et al., 2003).

FINDINGS

Finding F1: Both Eulerian and Lagrangian plume modeling approaches indicate that in 2000 high ozone concentrations in the HGB area were sensitive to both VOC and NOx emission reductions.

Lagrangian Plume Model

Analysis: M. Trainer-NOAA; Data: Parrish et al.-NOAA, Fried et al.-NCAR, Atlas et al.-U. Miami. Wert et al. (2003) presented a Lagrangian plume model designed to closely reproduce the emissions, ozone formation, other secondary photochemical product formation, and plume dispersion observed during the TexAQS 2000 study. Figure F1 shows that the model accurately reproduced the rapid production of ozone and formaldehyde that was measured in a concentrated plume originating from the HSC region. (Figure A3 shows the location of this plume intercept). The model required only two HRVOC – ethene and propene. To successfully reproduce the highest observed ozone concentrations, it was critical that the model reproduced two factors: a high ozone production efficiency as shown by observations (e.g. Figures A3 and A4 of this report), and a rapid rate of ozone production. High ozone production efficiency assures that high ozone can be produced from the emitted precursors, and the rapid rate of ozone production assures that high ozone is produced before the plumes of emissions dilute and disperse.

200

150

100

50

0

Ozo

ne (p

pbv)

302520151050

CH2O (ppbv)

Measured

Model

Figure F2 shows the modeled sensitivity of the ozone accumulation to both NOx and VOC emission reductions for the plume illustrated in Figure F1. The upper panel indicates that decreases in NOx emissions lead to a more rapid ozone production, but a lower peak ozone concentration, while HRVOC emission reductions lead to both slower ozone production and a lower peak ozone concentration. The lower panel indicates that HRVOC and NOx emission reductions are almost equally effective in reducing the total ozone flux in this plume. It is noteworthy that the ozone response is highly non-linear; reduction of either precursor by one-half reduces the ozone flux by only about one-quarter.


31 August 2007 76

The ozone flux produced and transported in this plume from the HSC accounts for a substantial fraction of the total ozone produced in the entire Houston area. In Response to Question G of this report, the total flux from the Houston urban area is calculated from airborne ozone lidar measurements. The base case calculated plume flux in Figure F2 represents 45 to 80% of the total Houston flux determined on six days during TexAQS 2000 and 2006. These calculations led Wert et al. (2003) to conclude that targeted reductions in either or both emission categories would effectively reduce the highest observed ozone levels.

Eulerian Modeling TCEQ (2004, 2006) has conducted Eulerian modeling of the entire HGB nonattainment area in order to determine which controls are necessary to reach attainment. This modeling indicates that both VOC and NOx controls were effective in reducing ozone in 2000.

In particular, two tests have indicated sensitivity to both VOC and NOx in Houston: reductions of biogenic VOC emissions by 30%, and modeling of weekend/weekday differences in mobile source emissions. However, in both cases, the effects on ozone concentrations varied by location within the HGB area and meteorological conditions, suggesting that VOC and NOx sensitivity in HGB varies spatially and temporally.

TCEQ modeling for the HGB area indicates that in 2009 further NOx emission reductions will be by far more effective than VOC emission reductions in decreasing O3. In the 2009 future baseline scenario, the emissions of anthropogenic NOx and VOC are projected to be reduced by 56% and 24%, respectively, relative to the 2000 emission inventory. Despite these large projected decreases in emissions, the maximum ozone design value in HGB is predicted to be 97 ppbv, well above the 84 ppbv level of the NAAQS. Thus without additional emission reductions, the HGB area is predicted to remain out of compliance with the NAAQS in 2009. Figure F3 shows the predicted effects of additional anthropogenic emission reductions beyond the 2009 future baseline. Additional reductions in VOC emissions result in only modest improvement in the ozone design value, and compliance with the NAAQS cannot be achieved with VOC decreases alone. The modeling does suggest that reductions in VOC emissions would allow compliance with the NAAQS to be reached with somewhat less drastic NOx emission reductions (53% versus 58%) than is the case for NOx emission reductions alone.

Figure F2. Ozone sensitivity calculated by the Lagrangian plume model. The red lines give the base case calculation, and the green and red lines indicate the effect of a factor of two decrease in emissions NOx and VOC, respectively.


31 August 2007 77

“Radical Starvation” in Selected Modeling Scenarios A photochemical grid modeling scenario has shown that under certain conditions with high concentrations of HRVOC, ozone formation can be inhibited by "radical starvation." The radical starvation can be alleviated by adding large quantities of primary formaldehyde emissions, CO emissions, or aromatic emissions, or by changing to the SAPRC99 chemical mechanism. It is unclear at this point which, if any, of these solutions are appropriate for Houston modeling.

Figure F3. Predicted future ozone design values (O3 DVf; the 3-year average of the fourth-highest daily maximum 8-hour-average ozone concentration) for HGB as a function of reductions in emissions of NOx, VOC, and both. The reference calculation (i.e. 0 percent reduction in the table in the figure) is based upon the projected emissions for 2009, and the reductions from that reference case are assumed constant across all emission categories. (Analysis by TCEQ; Figure adapted from D. Karp)


31 August 2007 78

Finding F2: An observation-based approach to determine the sensitivities of high ozone in the HGB non-attainment area to the precursor VOC and NOx emissions has been investigated; it has yielded ambiguous results.

Analysis: Parrish-NOAA; Data: Ryerson, Neuman, Parrish et al.-NOAA. The observation-based approach selected is based upon the relationships between indicator species developed by S. Sillman (http://www-personal.engin.umich.edu/~sillman/obm.htm). This approach was selected because: 1) it addresses integrated total O3 produced (not instantaneous rate of O3 production), and is thus designed to answer the question of how maximum O3 responds to changes in VOC versus changes in NOx emissions; 2) it is arguably the most fully developed observation-based method and is widely used; 3) it is related to the ozone production efficiency relationships exemplified in Figures A3 and A4 of this report; and 4) it utilizes measurements made with high precision and accuracy on the Electra aircraft during TexAQS 2000 and the WP-3D and Ronald Brown during TexAQS 2006.

Figure F4 compares the measurements made from the Electra during TexAQS 2000 with the modeled indicator species relationships of Sillman. These two relationships are those he discusses most thoroughly. The data shown include all 5-second average measurements made in the greater Houston area (94.3-96.3 deg. E Long., 28.7-30.7 deg. N. Lat.), which includes the entire urban area, all petrochemical facilities including the isolated Gulf Coast facilities, and the Parish power plant.

Figure F4. Indicator species relationships from models [Sillman] and from the TexAQS 2000 Electra measurements (red symbols). The relationships expected for NOx-, mixed-, and VOC-sensitive ozone production are given by green, violet and black symbols, respectively. The blue highlighted symbols indicate air masses dominated by NOx titration.


31 August 2007 79

The indicator species relationships in Figure F4 give ambiguous results. In both figures most of the data are localized in the predominately NOx-sensitive region, but some points are in the mixed and predominately VOC-sensitive regions. Further, the model results themselves are interspersed between regions. Definitive conclusions cannot be drawn from these figures. This ambiguity is particularly clear in the examination of ozone data above 200 ppbv. These data were collected on three days in 2000: 25 and 30 August and 1 September (see Figure A3). The relationship of ozone with HNO3 suggests that all three days represent NOx-sensitive conditions, while the relationship of ozone with NOy - NOx suggests that the three days span the full range from NOx-sensitive to VOC-sensitive. It is clear that no definitive conclusions can be drawn. It may be that the indicator species relationships may be most useful as a basis of comparison of models with measurements.

Finding F3: At the highest ozone concentrations, the observed relationship between ozone and the products of NOx oxidation indicates less efficient ozone production in the Dallas area than in the Houston area. In the observation-based indicator species approach, this behavior corresponds to less NOx-sensitive and more VOC- or radical-sensitive ozone formation in Dallas compared to Houston.

Analysis: Parrish-NOAA; Data: Ryerson, Neuman, Parrish et al.-NOAA. Figure F5 shows the relationships between ozone and the oxidation products of NOx. These oxidation products include only nitric acid on the left plot and nitric acid plus organic nitrates on the right plot. The generally shallower slopes in the Dallas area indicate less efficient ozone production in that area. From an observation-based indicator approach, this behavior corresponds to less NOx-sensitive and more VOC- or radical-sensitive ozone formation in Dallas compared to Houston.

Figure F5. Indicator species relationships from the TexAQS 2000 Electra measurements. The gray and blue symbols indicate data collected in the Houston and Dallas areas, respectively.


31 August 2007 80

Finding F4: Tests of the ability of models to reproduce observed relationships between ozone and other photochemical products have the potential to provide very fruitful approaches to improving models.

Analysis: Parrish-NOAA; Modeling: TCEQ; Kim et al.-NOAA. Relationships between ozone and other photochemical products, such as those illustrated in Figure F5, have been found to be remarkably robust in observations. Examination of the ability of models to reproduce these observed relationships is expected to provide powerful tests of the models’ performance and to suggest possible model improvement. Figures F6 and F7 give some example model output for the relationships of O3 with HNO3 and with NOy-NOx. The figures are plotted in the same format as the observations in Figure F5 for ease of comparison.

The results of the WRF-Chem model (Figure F6), which covers most of the continental U.S. at relatively coarse (27 km x 27 km) resolution (Kim et al, 2006), shows some similarities and differences when compared to the observations (Figure F5). The model produces relatively well defined relationships, but the slopes of the relationships are significantly smaller than those observed. There is no marked difference in the modeled relationships in the different regions of eastern Texas. The very high O3 concentrations observed in the Houston area are not reproduced by the model. Significantly higher resolution, particularly over the urban areas, is an avenue of model development that would be expected to improve the model performance. Further, the HRVOC emissions for the petrochemical facilities are undoubtedly underestimated in these calculations, since they are based upon the NEI 1999 emission inventory.

The TCEQ modeling results (Figure F7) are focused on the HGB region with much higher resolution, and generally give a better reproduction of the observed relationships. The slopes of the relationships through the majority of the data are accurately reproduced. However, there is an indication that the slopes decrease at the highest O3 concentrations, so that the highest modeled O3 concentrations are significantly lower than those observed. One explanation that has been suggested for this behavior is that the HRVOC emissions are underestimated in the model emissions and/or that they are not properly collocated with the NOx emissions from the petrochemical facilities. The model also predicts relatively low concentrations of O3 at relatively high concentrations of HNO3 and NOy-NOx. Such behavior is not seen in the observations. The cause of the disagreement in these few points should be investigated.


31 August 2007 81

Figure F6. Indicator species relationships calculated by the WRF-Chem model for 2000 UTC each day between 1 April and 31 October 2004. The color coding indicates results for different areas of eastern Texas.

Figure F7. Indicator species relationships derived from 2006 modeling by TCEQ with the Base 1b configuration. The data points include all hours, 12:00-18:00 local time, on the days between 16 August and 6 September 2000. The color coding indicates two different monitoring sites in the HGB area.


31 August 2007 82

KEY CITATIONS AND INFORMATION AND DATA SOURCES Kim, S.-W., A. Heckel, S.A. McKeen, G.J. Frost, E.-Y. Hsie, M.K. Trainer, A. Richter, J.P.

Burrows, S.E. Peckham, and G.A. Grell. 2006. Satellite-observed U.S. power plant NOx emission reductions and their impact on air quality. Geophys. Res. Lett. 33, L22812, doi:10.1029/2006GL027749.

Sillman, S. Observation-based methods (OBMs) for analyzing urban/regional ozone production and Ozone- NOx-VOC sensitivity (http://www-personal.engin.umich.edu/~sillman/obm.htm)

TCEQ. 2004. Houston-Galveston-Brazoria Mid-Course Review SIP Appendixes (2004-042-SIP-NR). http://www.tceq.state.tx.us/implementation/air/sip/dec2004hgb_mcr.html

TCEQ. 2006. Weekday-Weekend Effect Analysis, Part Two. Presented by Jim Smith, June 8, 2006. http://www.tceq.state.tx.us/assets/public/implementation/air/am/committees/pmt_set/20060621/20060621-smith-weekday_weekend_effects_part2.pdf

Wert, B.P., M. Trainer, A. Fried, T.B. Ryerson, B. Henry, W. Potter, W.M. Angevine, E. Atlas, S.G. Donnelly, F.C. Fehsenfeld, G.J. Frost, P.D. Goldan, A. Hansel, J.S. Holloway, G. Hübler, W.C. Kuster, D.K. Nicks Jr., J.A. Neuman, D.D. Parrish, S. Schauffler, J. Stutz, D.T. Sueper, C. Wiedinmyer, and A. Wisthaler. 2003. Signatures of terminal alkene oxidation in airborne formaldehyde measurements during TexAQS 2000. J. Geophys. Res., 108(D3), 4104, doi:10.1029/2002JD002502.


31 August 2007 83

Response to Question G

QUESTION G How do emissions from local and distant sources interact to determine the air quality in Texas? What meteorological and chemical conditions exist when elevated background ozone and aerosol from distant regions affect Texas? How high are background concentrations of ozone and aerosol, and how do they vary spatially and temporally?

BACKGROUND Question G is closely related to Question H. Here Question G focuses on characterizing the background ozone and aerosol distributions, and the chemical and physical processes that affect the background concentrations of ozone and aerosol in Texas. Question H focuses on the transport processes and source-receptor relationships of those background concentrations.

FINDINGS

Finding G1: The maximum background ozone concentrations encountered in 2006 exceeded the 8-hour NAAQS. On average, air of continental origin had higher background concentrations than marine air. The average background ozone concentrations measured in 2006 in eastern Texas complement a previously developed climatology.

Analysis: Senff et al.-NOAA, Sullivan et al.-U. Texas; Data: Senff et al.-NOAA, TCEQ. Daily 8-hour-average ozone maxima from upwind suburban or rural sites are taken as indicators of the local background in a particular region. In Figure G1, the red line segments indicate the average background for four areas in Texas on days in April – July 2006 when the area maximum 8-hour average reached or exceeded 80 ppbv. These segments are higher than the corresponding average curves, which were compiled for all days, not just the higher ozone days. The marine influence in the Houston area accounts for the lower background (49 ppbv) compared to the Dallas (63 ppbv), Beaumont (61 ppbv), and Northeast Texas (60 ppbv) areas.

The NOAA lidar provided a different approach for obtaining mesoscale estimates of background ozone for four regions in eastern Texas: 1) offshore over the Gulf of Mexico, 2) Gulf of Mexico coastal areas (less than 50 km from shore), 3) southeast Texas including Houston (all of southeast Texas more than 50 km from shore), and 4) northeast Texas including Dallas. Background ozone was determined by averaging lidar ozone profiles between the surface and the top of the boundary layer over all data points that were identified to be outside of pollution plumes. The blue line segments in Figure G1 indicate the background ozone values for the four regions identified above using measurements from all suitable lidar flights. The average background values for the southeast Texas and Gulf coast areas compare well with the Houston area background on high-ozone days (red line segment). The northeast Texas lidar background ozone value is close to the Dallas and northeast Texas TCEQ network-based background values for high-ozone days. This good agreement is not surprising because the lidar was typically flown on days when an ozone exceedance was forecast. The average ozone background measured over the Gulf of Mexico was the lowest of all four areas (39 ppbv) and is close to the average curves for Houston-Galveston-Brazoria (HGA) and Beaumont-Port Arthur (BPA) for the 1 August to 15 September period of the measurements. The highest observed ozone background value was 86


31 August 2007 84

ppbv measured on 8 September in east central Texas near the Louisiana border, after several days of continuous easterly flow conditions.

Figure G1. Continuous curves give six-year (1998-2003) average background ozone in various regions in eastern Texas, smoothed with a 31-point running mean filter (Nielsen-Gammon et al., 2005). The blue (lidar data) and red line segments (TCEQ surface network data) indicate 2006 background ozone determinations with greater emphasis on exceedance days. The widths of the line segments approximately indicate the periods considered in 2006.

A characterization of synoptic flow based on classes of regional ozone concentrations showed higher background ozone concentrations under continental air compared to maritime air. Twelve sites were selected for completeness of data coverage (>75% each year in August and September from 2001 to 2006) and regional representation (one site per county, except for two in Harris County). Each of the days included in this data set was classified as “high regional ozone,” if and only if four or more stations reported maximum 8-hour-average O3 of 75 ppbv or higher, (this included 40 days) or “low regional ozone,” if and only if all stations reported maximum 8-hour-average O3 of at least 40 but less than 50 ppbv (this included 43 days). The probability distributions for the 1-hour time step points of the HYSPLIT back trajectories (72-hour, 300 meter above ground level/mid-day starting point) for the twelve sites are given in Figure G2. On “low regional ozone” days air parcel trajectories tended to cluster to the south and southeast with relatively longer fetches from the Gulf of Mexico, while on “high regional ozone” days the trajectories cluster to the northeast with relatively short fetches, indicating transport from within eastern Texas or from areas to the northeast of Texas.


31 August 2007 85

Finding G2: The net ozone flux transported out of Houston averages about a factor of two to three larger than the corresponding flux from Dallas. The fluxes from these urban areas are significant contributors to the background ozone in the eastern Texas region.

Analysis: Senff et al.-NOAA, Sullivan et al.-U. Texas; Data: Senff et al.-NOAA, TCEQ. The horizontal flux of O3 downwind of Houston and Dallas during TexAQS 2000 and TexAQS 2006 was calculated from airborne lidar measurements of O3. The O3 flux was computed by vertically integrating excess O3 concentration in the plume (plume O3 minus background O3) between the surface and the top of the boundary layer and horizontally between the plume edges (see Figure G3 for example of a plume cross section). The integrated plume excess O3 concentration was then multiplied by the horizontal wind speed (estimated from nearby wind profilers) to yield O3 flux, which is expressed in molecules per second. Using data from the TexAQS 2000 and 2006 studies, fluxes were computed for the Houston area on five days (8/28 and 9/06/2000; 8/12, 8/14, and 8/30/2006) and the Dallas/Fort Worth area on one day (9/13/2006). The ozone flux from Houston ranged from 3.2 * 1026 to 6.0 * 1026 molec s-1, and - within the uncertainties - was similar for the 2000 and 2006 cases. The average ozone flux from Houston of 4.6 * 1026 molec s-1, emitted over the 8-hour photochemically active part of one day, is sufficient to produce a 10-ppbv increase in ozone over an approximately 10,000 square mile area, assuming a 2-km deep boundary layer. The flux for the one Dallas case was estimated as 1.7 * 1026 molec s-1, about a factor of two to three smaller than that observed for Houston. Although this comparison is based on only one flux determination for Dallas, we expect this comparison to generally hold, since Dallas lacks the large ozone production from the petrochemical industry, which accounts for a large fraction of the ozone production in Houston.

Figure G2. Contour plots showing the probability distribution of origins of 72-hour HYSPLIT back trajectories ending at 12 representative TCEQ surface ozone monitoring stations in eastern Texas. Red color denotes high probability, yellow indicates low probability. Left panel: high regional ozone days; right panel: low regional ozone days.


31 August 2007 86

As a complement to airborne lidar data, which are only available for rather short time periods, a similar assessment was performed using data from TCEQ’s surface ozone monitoring network to estimate horizontal ozone fluxes. The horizontal plume integral was calculated from ozone measurements at downwind surface sites. The ozone was assumed constant through the mixed layer and the vertical plume extent was estimated from mixed layer height measurements from nearby wind profilers. For the days of the lidar analysis the differences between lidar- and surface-network-based flux estimates were generally on the order of 20 to 25%, except for one case where the difference was nearly 60%. This one large discrepancy may be due to uncertainties in the estimation of mixing height and horizontal plume dimensions when using surface network data. The agreement between the airborne lidar and surface monitoring results suggests that routine flux assessments using surface data may be useful.

Figure G3. Time-height cross section of ozone plume about 50 km downwind of the Houston and Ship Channel area measured with NOAA’s airborne ozone lidar on 8/14/2006 at about 17:00 CST. Note the strong variation of mixing height (shown as black line) across the plume.


31 August 2007 87

Finding G3: Elevated background ozone concentrations for urban areas can include contributions from the recirculation of locally produced ozone or local precursor emissions.

Analysis: Ryerson et al.-NOAA, Sullivan et al.-U. Texas; Data: Ryerson et al.-NOAA, TCEQ. Background ozone concentrations entering an urban area can be elevated by recirculation of the ozone produced in that same urban area on a previous day, or by ozone production from precursors emitted within that urban area and recirculated. A past example provided by TCEQ is a May 31, 2003 episode in DFW, for which O3 concentration contours at mid-afternoon are shown in Figure G4. In this event, the one-hour-average peak O3 was 161 ppbv, with the eight-hour-average peak at 130 ppbv. The event was influenced by a stalled frontal passage – note the contrasting direction at the wind vanes associated with monitoring sites on the north side versus the south side of DFW. The color code for ozone concentrations in the figure is: Gray 85-99 ppbv, Orange 100-124 ppbv, Red 125-149 ppbv, Purple > 149 ppbv.

Figure G4. (Left) Surface weather map for 6 CST 31 May 2003 shows stalled front in north Texas (http://weather.unisys.com/archive/index.html). (Right) Mid-day contour of O3 concentrations shows “pancake” of elevated O3 between Dallas and Fort Worth. Wind barbs show northerly winds north of the city and southerly winds south of the city, trapping local pollution (TCEQ, 2003).

A recirculation example was observed in the Houston area during TexAQS 2006 (Figure G5). On 26 September the Houston plume with ozone concentrations approaching 120 ppbv was carried to the southeast by the prevailing winds. On the following day, winds from the south brought air from over the Gulf of Mexico into the Houston area. However, this background air contained ozone concentrations much higher (64 to 84 ppbv, average 74 ppbv) than the more characteristic concentrations of about 40 ppbv (see Figure G1). These elevated ozone concentrations were accompanied by elevated concentrations of other photochemical products (e.g. formaldehyde and nitric acid) and CO, a tracer of primary emissions.


31 August 2007 88

Results from FLEXPART particle dispersion modeling (Figure G6) and back-trajectories calculated from wind profiler data (not shown) both indicate that the elevation in background ozone concentrations on 27 September were due to recirculation of the 26 September Houston area plume back into the HGB region.

Figure G5. WP-3D flight tracks within the boundary layer in the HGB area on two successive flights color-coded and sized according to the measured O3 concentration.

Figure G6. Footprint emission sensitivity for air sampled by WP-3D on 27 September at the position marked by the asterisk. The integral of the product of this quantity with the surface emission density gives the expected concentration that would be measured at the aircraft position. The numerals 1 and 2 indicate days of backward transport time. The near coincidence of the asterisk and the numeral 1 indicates that emissions from 26 September have been returned to near the original emission location. The shape of the red to violet colors indicates the primary transport pathways. The color scale is logarithmic in arbitrary units.


31 August 2007 89

Finding G4: Plumes from Texas urban areas make substantial contributions to the ozone, aerosol, and precursor concentrations in the rural regions of eastern Texas.

Analysis: Brock et al.-NOAA, Data: Middlebrook, de Gouw et al.-NOAA. Figure G7 shows one example of model output and measurements of transported plumes of emissions and their photochemical products from the Houston area to northeastern Texas. The model results are from the FLEXPART Lagrangian particle dispersion model, and the measurements are from the WP-3D aircraft on 16 September 2006. There is excellent agreement between the model and the measurements, with both showing that the SO2 plume, primarily from the Parish power plant, was transported parallel and to the west of the NOx and benzene plumes, primarily from the Houston urban and Ship Channel areas. Table G1 shows that these plumes significantly increased the total aerosol concentrations in rural northeast Texas (for reference, the NAAQS for PM2.5 aerosol are 15.0 µg m-3 annual mean, and 35 µg m-3 24-hour mean).

Table G1. Measured concentrations of aerosol and gas phase species in the Parish power plant and Houston area plumes in northeast Texas.

Background Parish Petrochemical Industry

Total mass (µg m-3) 4.3 7.9 5.7 Sulfate (µg m-3) 1.9 4.7 2.0 Organic (µg m-3) 1.2 1.9 2.3 Black Carbon (µg m-3) 0.08 0.12 0.17 Benzene (ppbv) 0.094 0.103 0.26 SO2 (ppbv) 0.29 0.97 0.53


31 August 2007 90

Figure G7. Modeled and measured concentrations of aerosol and gaseous species in transported plumes from the Houston area to northeast Texas. The upper panel shows the FLEXPART model calculations (Cooper et al.–NOAA) for SO2 (primarily Parish power plant) and NOx (primarily HSC) plumes, and the lower panel shows the measurements from the WP-3D of sulfate aerosol and benzene concentrations. A plume of sulfate aerosol associated with the Parish power plant SO2 emission plume is evident in the measurements, as is a plume of benzene associated with HSC emissions. (Lower panel covers only a fraction of the area in the upper panel.)


31 August 2007 91

Finding G5: Dust of African origin and sulfate aerosol advected into the Houston area, under southerly flow conditions from the Gulf of Mexico, can make significant contributions to the background aerosol in the eastern Texas region.

Analysis and Data: Bates and Quinn–NOAA. Figure G8 summarizes 2006 aerosol chemical composition measurements made on the Ronald H. Brown in the Houston area. The onshore southerly flow of background aerosol (low radon concentrations indicating no contact with land for several days) was substantially impacted by Saharan dust and what appear to be ship emissions (acidic sulfate and nitrate). Mean (median) mass concentrations of the total submicrometer and supermicrometer aerosol were 6.5 (4.6) µg m-3 and 17.2 (8.7) µg m-3, respectively. These mass loadings of “background” aerosol are much higher than typically observed in the marine atmosphere, and are large enough to substantially impact the PM loading in the Houston-Galveston area. The integrated PM2.5 mass at ambient relative humidity (Figure G9) includes the accumulation mode (primarily acidic sulfate and dust under southerly flow conditions, seen in Figure G8) and part of the coarse mode (primarily sea salt, dust, and the acidic nitrate and sulfate absorbed by these basic components seen in Figure G8). The average PM2.5 mass advecting into the Houston-Galveston area during TexAQS 2006 was 19 ± 11 µg m-3 (Figure G9). Aerosol composition and mass size distributions during northerly flow conditions are shown in Figures G8 and G9 for comparison with the “background” southerly flow. Air quality forecast models need to include ship emissions and dust transport to correctly characterize aerosol loadings in SE Texas. Compliance with PM2.5 regulations in the Houston-Galveston area may require stricter controls on upwind aerosol sources (e.g. ship emissions).

POM

(NH4)HSO4Nitrate

ECSeasalt

Dust

Northerly Flow20.8 µg m-3

36%

51%8%

POM

(NH4)HSO4

Nitrate

Seasalt

Dust

Northerly Flow7.4 µg m-3

9% 30%

11%

37%

13%

POM

(NH4)HSO4

NitrateECSeasalt

Dust

Southerly FlowGulf of Mexico6.5 µg m-3

61%

30%

6%

POM

(NH4)HSO4

Nitrate

Seasalt

Dust

Southerly FlowGulf of Mexico17.2 µg m-3

38%

33%

15%

7%

6%

Figure G8. Average submicrometer (top) and supermicrometer (bottom) aerosol composition in the marine boundary layer, measured (at 60% RH) from the Ronald H. Brown in TexAQS 2006 during periods of southerly and northerly flow.


31 August 2007 92

70

60

50

40

30

20

10

0

dM /

dlog

Dp (

ambi

ent R

H)

3 4 5 6 7 8 9

1022 3 4 5 6 7 8 9

1032 3 4 5 6 7 8 9

104

Dp (nm)

Southerly flow Northerly flow

Integrated PM 2.5 Mass Southerly flow 19 ± 11 µg/m3

Northerly flow 34 ± 20 µg/m3

Figure G9. Average aerosol mass size distributions at ambient relative humidity measured from the Ronald H. Brown during TexAQS 2006.

Finding G6: Nighttime chemistry influences the availability of oxides of nitrogen (NOx), highly reactive VOC (HRVOC), and O3.

Analysis and Data: Brown et al.-NOAA. Nocturnal measurements of key nitrogen oxide species, nitrate radical (NO3) and N2O5, were made on the Ronald H. Brown and WP-3D. These compounds are important to regional air quality because their formation and subsequent reactions affect the nocturnal loss rates for NOx, O3, and HRVOC. Analysis of the measured NO3 and N2O5 led to the following conclusions.

• Hydrolysis of N2O5, typically the most important reaction in the nocturnal conversion of NOx to HNO3, was generally inefficient in air masses sampled aloft around Houston and elsewhere in Texas. As a result, NOx emissions occurring late in the day or at night could be transported overnight in the form of N2O5 to regions distant from the NOx source regions. Transport of O3 was also efficient since N2O5 is a reservoir for odd oxygen (Ox) as well. This transport may affect O3 concentrations well downwind of major urban and industrial areas on the following day (Figure G10).

• The reduced rate of N2O5 hydrolysis aloft in Houston, and the relatively warm temperatures in this area, enhanced the availability of NO3 as an oxidant. This was important in plumes containing both NOx and HRVOC from industrial sources in Houston. Observed nocturnal loss rates for HRVOC due to reaction with NO3 from plumes originating in the Houston Ship channel were 0.5 – 4 ppbv hr–1 (Figure G11).

• Surface measurements of N2O5 from the RHB and vertical profiling to low altitude from the WP-3D indicated more rapid loss rates for N2O5 near the surface than aloft, possibly arising from concentration of sinks for NO3 and N2O5 from emissions within the shallow nocturnal boundary layer.


31 August 2007 93

Figure G11. Total VOC loss rates due to reaction with NO3 measured from the WP-3D on 11-12 October 2006. The flight track in the map is color-coded by NO3 mixing ratio; these values are among the largest ever observed. The pie chart insets show the relative contributions to NO3 loss on this flight and the relative contribution of NO3 loss to anthropogenic VOC, which is dominated by the reaction with alkenes. Apparent "Biogenic" NO3 losses on this chart are dominated by isoprene from anthropogenic sources.

Figure G10. Fraction of NOx [F(NOx) = (NO3 + 2N2O5)/(NO2 + NO3 + 2N2O5)] present in the form of N2O5 and predicted to be transported overnight from the Parish power plant near Houston to northeast Texas, based on the observed N2O5 loss rates within the plume sampled in the Houston area and a forward overnight meteorological trajectory. The WP-3D flight track is color coded by F(NOx) and the forward trajectory is color and size coded by the predicted F(NOx) during transport. Sunrise for the trajectories occurs over east central Texas where F(NOx) quickly drops to zero, indicating re-conversion of N2O5 to NOx and O3 in a rural area.


31 August 2007 94

Finding G7: Low rural nighttime ozone concentrations have been observed at some, but not all, rural locations in northeast Texas; these low nighttime ozone concentrations are not replicated in the regulatory modeling.

Analysis: Sullivan et al.-U. Texas; Data: TCEQ. Photochemical models fail to reproduce the low nighttime ozone concentrations observed at some rural sites in northeast Texas (Figure G12). Possible causes of this discrepancy include the presence of: 1) shallower nighttime boundary layers than predicted by the model at the affected sites, and 2) larger local NO emissions than included in emission inventories. If the latter is the cause, then corrections to the emissions inventory are needed to accurately assess NOx concentrations and atmospheric chemistry upwind of northeast Texas cities and the DFW area. Compressors at well-heads and on pipelines emit NOx, and are potential sources of local NO emissions that may be underestimated in inventories. The map in Figure G12 shows that northeast Texas is a region of intense natural gas exploitation. Four other monitors in that area also consistently show lower than expected nighttime O3 concentrations.

KEY CITATIONS AND INFORMATION AND DATA SOURCES Nielsen-Gammon, J.W., J. Tobin, A. McNeel, and G. Li. 2005. A Conceptual Model for Eight-

Hour Ozone Exceedances in Houston, Texas - Part I: Background Ozone Levels in Eastern Texas. HARC Report No. H012.2004.8HRA, January 29, 2005. http://www.harc.edu/Projects/AirQuality/Projects/Projects/H012.2004.8HRA.

Figure G12. Time series of ozone concentrations for 19-22 August 1999, an episode included in the DFW SIP modeling (left). Ambient data at the Cypress River site in northeast Texas are compared with modeling results. The map shows natural gas well regions color-coded for production (right), and the arrow locates the Cypress River site.


31 August 2007 95

Response to Question H

QUESTION H Which areas within Texas adversely affect the air quality of non-attainment areas in Texas? Which areas outside of Texas adversely affect the air quality of non-attainment areas in Texas?

BACKGROUND Question H is closely related to Question G. Here Question H focuses on the source-receptor relationships that determine the background concentrations of ozone and aerosol in Texas and the meteorologically driven transport processes. Question G focuses on characterizing the background concentrations and the chemical and physical processes that affect those background concentrations.

FINDINGS

Finding H1: Ozone can be transported into the Dallas area from the Houston area.

Analysis and Data: Senff et al.-NOAA. An example of direct transport of ozone from Houston to Dallas was observed during TexAQS 2006 and is depicted in Figures H1 and H2. During 4–8 September 2006, there was a regional buildup of background ozone in eastern Texas, indicated by Dallas-Fort Worth (DFW) surface station ozone monitors (not shown) and airborne ozone lidar measurements (Figure H1). From 4–7 September, large-scale winds tended to be northerly to easterly. While the ozone was building up in Dallas, Houston also experienced a daily increase in ozone, resulting in 8-hr ozone averages up to 110 ppbv in Houston on 7 September (not shown). Between 7 and 8 September, a shift in the position of a synoptic-scale high caused the transport winds to Dallas to change from a weak northeasterly component to a stronger southerly component. This major shift in transport had a strong impact on Dallas, as 24-hour forward trajectories from Houston, beginning at 3 pm local time, indicated transport from Houston to Dallas (Figure H2). Also, elevated overnight O3 concentrations were measured at two rural O3 stations sited for TexAQS II (Palestine and Italy). This transport brought additional ozone to a region that was already approaching 8-hr exceedance levels, resulting in 3 stations exceeding the 80 ppbv 8-hr-average O3 in the DFW network.


31 August 2007 96

Figure H1. Airborne ozone lidar measurements on 4 and 8 September 2006, showing the large increase in background ozone in eastern Texas between the two days.

Figure H2. Forward trajectories starting at 3 pm local time in Houston, 7 September, and ending at 3 pm local time on 8 September 2006. The trajectories show direct transport from Houston to Dallas. 7 September was an exceedance day in Houston; 8 September was an exceedance day in Dallas. (The trajectory map was created using the NOAA Physical Science Division (PSD) upper air back-trajectory tool [http://www.etl.noaa.gov/programs/2006/texaqs/traj/]).


31 August 2007 97

Finding H2: High ozone concentrations in eastern Texas result from both in-state sources and transport of continental air from the east and northeast.

Analysis: Sullivan et al.-U. Texas. Ensembles of historic HYSPLIT (Draxler and Rolph, 2003; Rolph, 2003) 5-day back trajectories have been run from major Texas cities on high (≥ 75 ppbv 8-hr max) and low O3 days to characterize the upwind “dirty” and “clean” typical fetches. A residence time analysis of the “dirty” fetch trajectories yields high O3 air residence maps for northeast Texas, as exemplified in Figure H3. These maps (particularly the right one) show that transport of air from the east and northeast support many high O3 episodes in August through October, but this analysis also shows that on high O3 days in eastern Texas, the upwind air is resident within the state for several days, potentially building up background ozone from in-state source emissions.

Using a new statistical analysis tool in HYSPLIT, an ensemble of back-trajectories can be categorized into clusters of similar behavior. Back-trajectories from East Texas from August and September 1997–2006 cluster into 5 categories: “short fetch” within Texas, Arkansas, and Louisiana; “E-SE” running through the southeastern US and the northern Gulf of Mexico; “S-SE” running through the Gulf to the Yucatan; “NE” running to the Midwest; and “north” running through the north-central US. An assessment of back-trajectory clusters by year compared with regional O3 levels for August and September 1997–2006 shows that the “NE” cluster is best associated with high regional ozone, with the “short fetch” having the second highest correlation. There is some evidence that the year-to-year variation in the distribution of back-trajectory clusters helps explain the relative severity of O3 seasons.

Figure H3. Contouring of HYSPLIT 500 meter 5-day back-trajectories on moderate and worse O3 days in San Antonio (left) and Tyler/Longview/Marshall (right), color coded by density of end point frequency.


31 August 2007 98

Finding H3: A synthesis of satellite and in situ measurements with photochemical modeling and Lagrangian trajectory analyses provides a quantification of regional influences and distant sources on Houston and Dallas air quality during TexAQS 2006.

Analysis: Pierce et al.–NASA; Data: MODIS, AIRS, CALIPSO, and TES teams–NASA, NOAA/NESDIS, UMBC. Realtime Air Quality Modeling System (RAQMS) (http://rossby.larc.nasa.gov/RAQMS/ accessed October 2006) chemical analyses provide a means to quantify regional influences and distant sources on local air quality in Texas. Total column ozone from the Ozone Monitoring Instrument (OMI) and ozone and carbon monoxide profiles from the Tropospheric Emission Spectrometer (TES) onboard the NASA Aura satellite constrain these analyses, and fire counts from the Moderate Resolution Imaging Spectroradiometer (MODIS) instrument onboard the NASA Terra and Aqua satellites are utilized to generate biomass burning emissions. The fidelity of the RAQMS ozone analysis for the TexAQS 2006 period was assessed with independent satellite observations and aircraft data. The resulting bias-corrected RAQMS chemical analyses were used to provide estimates of background composition along ensemble back trajectories initialized daily at 18Z from surface EPA AIRNow ozone monitoring stations within the Houston and Dallas metropolitan statistical areas (MSA). Lagrangian averaged ozone production minus loss (O3 P-L) rates along the back trajectories were used as a metric to classify back trajectories. The Lagrangian averages were computed during time periods when the back trajectories were outside the respective MSA, defined as more than 2o in longitude or latitude away from central Houston or Dallas. Three trajectory classes were considered: Class 1 (O3 P-L > 10ppbv/day); Class 2 (0 < O3 P-L < 10ppbv/day); and Class 3 (O3 P-L < 0ppbv/day).

A time series depiction of the RAQMS back-trajectory analysis of regional influences on Houston and Dallas O3 is shown in the upper panels of Figure H4. The red line shows the observed mean and variability of surface O3 measurements in the urban area at 18 UTC extracted from EPA’s AIRNow data system (http://www.airnow.gov/ accessed October 2006). The blue lines show the RAQMS chemical analysis with the solid blue for the AIRNow mean (bias-corrected), and the dashed blue for the background mean (bias-corrected) immediately prior to entering the urban area. The color bar along the upper part of each time series indicates the regional influence classification code for the modeled extent of O3 production and loss upwind (see key). The Lagrangian analysis shows that periods of enhanced (P-L >10ppbv) regional O3 production preceded 6 out of 9 Houston periods and 7 out of 15 Dallas periods with elevated O3 (greater than 60 ppbv) during the 15 July – 15 October 2006 TexAQS II period. Maps of source regions for days with enhanced regional O3 production are shown in the lower panels of Figure H4. Houston enhanced regional O3 production events have a Midwest/Ohio River Valley source with significant O3 P-L (40ppbv/ day) due to NOx sources along the southern Great Lakes. Dallas enhanced regional O3 production events have a broad Great Plains/Midwest/Ohio River Valley source with significant O3 P-L (30ppbv/ day) due to Chicago NOx sources along the southern Great Lakes.


31 August 2007 99

Case studies were conducted to investigate the processes responsible for enhanced regional O3 production rates during the TexAQS II period. One such case is early September 2006. Twelve-day boundary layer back trajectories were initialized at locations where the High Spectral Resolution Lidar (HSRL) instrument (used to look down from the NASA King Air airplane to

=Net O3 Loss

=0< O3 P-L <10 ppb/day

= O3 P-L >10 ppb/day

Trajectory Classification

Figure H4. Lagrangian analysis of regional influences on Houston/Dallas O3. Left, time series of Houston and Dallas observed, analyzed, and background ozone. Color bar along upper part of each time series indicates regional influence trajectory classification. Below, maps of Houston and Dallas Class 1 (see text) NOy source contributions (ppbv/day). Red=AIRNow MSA mean+Std

Blue=RAQMS MSA (solid) Regional (dashed) bias corrected mean

DallasHouston


31 August 2007 100

assess the aerosol loading of the vertical column of air below) observed enhanced aerosol optical depths (AOD) indicating the presence of elevated particulate matter concentrations in the vicinity of Houston on 4 September 2006. Based on calculations of regional 24-hour rolling averages of TCEQ’s hourly PM2.5 concentrations, 4 September was one of only three days in the TexAQS 2006 period with widespread rolling 24-hour concentrations above 20 μg m-3. In addition, a speciation monitoring site in Deer Park showed elevated sulfate and carbon material for that date. The back trajectories link the local 4 September HSRL measurements to satellite column and profile measurements from various sensors on low-earth orbit satellites (see Table H1 and Figure H5) on 23 August.

The 12-day back trajectories indicated two primary source regions: southern Canada and the eastern US. The MODIS sensor and AIRS sensor show column AOD and carbon monoxide (CO) enhancements associated with Pacific Northwest wild fire emissions within the Canadian branch, and CO enhancements associated within the eastern US branch of the back trajectories. CALIPSO attenuated aerosol backscatter cross-sections through the central US show an elevated aerosol layer associated with wild fire emissions and boundary layer aerosol enhancements over the eastern US. TES CO vertical cross-sections, which follow the same orbit as CALIPSO, show both lower and upper tropospheric CO enhancements. This case study illustrates the influence of remote emissions from the southeastern US and Pacific Northwest on Houston air quality. This analysis underscores the importance of integrating satellite, aircraft, and surface measurements of aerosol and trace gases in conjunction with advanced modeling techniques, for characterizing the impact of emissions from remote sources on local Texas air quality. Additional case studies are presented on the TCEQ website (http://www.tceq.state.tx.us), and a detailed case study is available from the 29-31 May Principal Findings Data Analysis Workshop for a 20 July 2004 O3 and PM2.5 episode in the Dallas area.

Table H1. Satellite sensors and resulting data products utilized for 4 September 2006 case study.

NASA Sensor Product MODIS - Moderate Resolution Imaging Spectro-radiometer Aerosol optical depth CALIPSO - Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation Aerosol optical depth profile TES - Tropospheric Emission Spectrometer Carbon monoxide profile estimate AIRS - Atmospheric Infrared Sounder Carbon monoxide column


31 August 2007 101

Yellow here at 30 deg. latitude is suspected sulfate from eastern U.S.

Here, aloft at 45 deg. latitude, is suspected Canadian fire aerosol.

Here also at 45-50 deg. latitude is elevated CO, believed to be from fire.

Figure H5. MODIS and AIRS satellite sensor observations, with CALIPSO and TES orbits given by white lines. Center CALIPSO/TES ground track in two top panels represents 8/23/06 orbit, with color-coding for AOD (left) and CO (right).

Bottom two panels show corresponding vertical profiles along the center ground tracks, with x-axis roughly the north-south direction.


31 August 2007 102

Finding H4: In the Dallas area, local emissions and transport each contributed about equally to the average 8-hr ozone exceedance in 2002. Transported ozone alone can bring the Dallas area close to exceeding the 8-hour ozone standard.

Analysis: Kemball-Cook et al. –ENVIRON; Parrish et al.-NOAA. Data: Ryerson et al.–NOAA. Transport contributions to Dallas area O3 were quantified for June–September 2002 using the CAMx photochemical model with emissions from EPA’s 2002 National Emissions Inventory (NEI) (updated by the Central Regional Air Planning (CENRAP) consortium) and meteorology from the MM5 model (Kemball-Cook et al., 2006). The Dallas area had 35 days in 2002 with monitored 8-hour ozone levels of 85 ppbv or higher (Figure H6). Averaged over these days, Dallas area emissions contributed about 48 ppbv and other sources 54 ppbv to the total modeled ozone 8-hour maximum of 102 ppbv on the average exceedance day. The modeled average transport contribution from other parts of Texas was 6 ppbv, and there were days when northeast Texas, Houston/Beaumont, south Texas, and central Texas individually contributed 9 ppbv or more. The average modeled transport contribution from other US states was 28 ppbv, and the largest contributing states were Louisiana, Arkansas, Mississippi, Alabama, and Tennessee. These findings are consistent with back trajectory and residence time analyses. The example shown in Figure H7 compares CAMx O3 transport contributions to back trajectories for 7 August 2002. The 5-day back trajectories cross NE Texas, Arkansas, and Tennessee, extending to the mid-Atlantic region; the CAMx modeling finds these same regions contributing to the Dallas area O3 exceedance at the Ft. Worth NW monitor on 7 August 2002.

Estimates of the O3 imported into the DFW area from aircraft data are in general agreement with the trend from the CAMx model, although the aircraft data suggest that the transported background ozone contributed an even larger fraction of the ozone in the DFW region than suggested by the model calculations. The Electra aircraft in TexAQS 2000 and the WP-3D aircraft in TexAQS 2006 measured O3 upwind, across, and downwind of Dallas on two days during each field study. The green symbols in Figure H6 show the results from the 23 August and 7 September 2000 flights, and the orange symbols the results from the 13 and 25 September 2006 flights. The peak 8-hour-average O3 for these four flights are from the monitoring data, and the DFW contribution is estimated from the difference between these peak values and the background O3 measured upwind of DFW. 23 August 2000 is the one day investigated by the aircraft flights that had a significant exceedance of 98 ppbv. The background ozone transported into Dallas was about 72 ppbv, or 73% of the total, on that day. On some days transported ozone can bring the Dallas area close to an 8-hour exceedance without any added contribution from local emissions.


31 August 2007 103

August 7, 2002--106 ppb at Fort Worth NW (CAMS 13), 95 ppb in CAMx

05

10152025303540

DFW

NN

A

Nor

thTe

xas

Arka

nsas

Ala

bam

a

Tenn

esse

e

Mid

Atla

ntic

Stat

es

Geo

rgia

Mis

siss

ippi

Biog

enic

s

Boun

dary

Con

ditio

ns

Ozo

ne C

ontr

ibut

ion

(ppb

)

Figure H7. Modeled contributions to one monitor’s ozone total concentration. Local is largest single contributor, but transport from upwind source areas significantly increase concentrations (analysis by Kemball-Cook et al., 2006).

Trajectories from 3 different altitudes over Ft. Worth to characterize upwind areas for the column of air 2 pm CST on 7 August 2002.

Figure H6. Measured peak 8-hour-average O3 in the DFW area as a function of the DFW contribution to that peak. The DFW contribution is derived from the difference between the measured peak from the regional monitoring network, and the background ozone transported into the DFW region. The small symbols are from the model results from the CAMx model for each day 1 June -30 September 2002. The larger symbols are from aircraft transects upwind and across the DFW region on four days in 2000 and 2006.


31 August 2007 104

Finding H5: In the Houston area, local emissions and transport each contributed about equally to the average 8-hr ozone exceedances investigated by aircraft flights in 2000 and 2006. As in Dallas, transported ozone alone can bring the Houston area close to exceeding the 8-hour ozone standard.

Analysis Parrish et al.-NOAA; Data: Ryerson et al.–NOAA. Houston area emissions contributed about 48 ppbv and transported ozone about 59 ppbv to the average total ozone 8-hour maximum of 107 ppb on the seven exceedance days investigated by aircraft measurements during 2000 and 2006. Figure H8 shows the apportionment of the measured maximum 8-hour-average ozone for all days characterized by aircraft measurements. The linear least-square fits give slopes near unity for the separate years, and for the total data sets, which indicates that the local and transported background contributions are largely independent of each other. The background contribution was similar between the two years, but the local contribution was considerably larger in 2000 than 2006. The transported background ozone reached a maximum of 70 ppbv, which can lead to an exceedance with a relatively small additional contribution from local sources.

140

120

100

80

60

40

20

0

HG

B M

axim

um 8

-hr A

vera

ge O

3 (p

pbv)

806040200

Transported Ozone Background (ppbv)

y = (0.98 ± 0.19)x + 28 ± 10 r 2 = 0.80 y = (0.77 ± 0.66)x + 57 ± 35 r 2= 0.19

y = (1.00 ± 0.30)x + 35 ± 16 R2 = 0.42

806040200

HGB Ozone Contribution (ppbv)

y = (1.00 ± 0.22)x + 50 ± 9 R2 = 0.58

y = 0.91x +51 r 2 = 0.17 y = 0.92x +56 r 2 = 0.69

HGBTexAQS 2006TexAQS 2000

Figure H6. Measured peak 8-hour-average O3 in the HGB area as a function of the background ozone transported into the region (left) and the local HGB contribution to that peak (right). The background ozone was determined from aircraft transects upwind and across the HGB region on seventeen days in 2000 and 2006. The HGB contribution is derived from the difference between the measured peak and the background.


31 August 2007 105

KEY CITATIONS AND INFORMATION AND DATA SOURCES Draxler, R.R. and Rolph, G.D. 2003. HYSPLIT (HYbrid Single-Particle Lagrangian Integrated

Trajectory) Model access via NOAA ARL READY Website (http://www.arl.noaa.gov/ready/hysplit4.html). NOAA Air Resources Laboratory, Silver Spring, MD.

Kemball-Cook, S., M. Jimenez, and G. Yarwood. 2006. Seasonal Model of Northeast Texas Ozone for Summer 2002. http://www.netac.org/0405closeout/Task_6.1.pdf (accessed July 6, 2007).

Rolph, G.D. 2003. Real-time Environmental Applications and Display sYstem (READY) Website (http://www.arl.noaa.gov/ready/hysplit4.html). NOAA Air Resources Laboratory, Silver Spring, MD.


31 August 2007 106

Response to Question I

QUESTION I Why does the SAPRC chemical mechanism give different results than the carbon bond (CB-IV) mechanism? Which replicates the actual chemistry better?

BACKGROUND Gridded, regional photochemical models, used in developing State Implementation Plans (SIPs), use simplified photochemical reaction mechanisms. The two mechanisms that are most commonly used are the [California] Statewide Air Pollution Research Center (SAPRC) mechanism and the Carbon Bond (CB) mechanism. Both mechanisms are approved for use by the US EPA and are updated periodically to incorporate new experimental findings. For most urban areas, the CB mechanism, version IV (CB-IV) and the 1999 SAPRC (SAPRC99) mechanism yield similar results, but for conditions found in Houston, the SAPRC99 mechanism leads to concentrations of ozone that are up to 30-50 ppbv greater than in CB-IV and is more sensitive to reductions in NOx emissions, especially on days with high predicted ozone concentrations (Figure I1). These differences in the sensitivity of chemical mechanisms to emission reductions could have significant consequences for determining the magnitude of decreases in emissions of ozone precursors that will be required to demonstrate attainment of the current 8-hour, 80 ppbv NAAQS for ozone.

Figure I1. Predictions of domain-wide maximum O3 concentrations in Comprehensive Air Quality Model, with extensions (CAMx) on 30 August 2000. (Faraji et al., 2007a)


31 August 2007 107

FINDINGS

Finding I1: Air quality modeling for both 2000 and 2006 shows substantial differences in ozone concentrations predicted by the SAPRC99 and CB-IV chemical mechanisms.

Analysis: Faraji et al. (2007a,b); Byun-U. Houston. Simulations performed using multiple regional photochemical modeling packages (Comprehensive Air Quality Model, with extensions [CAMx] and Community Multi-scale Air Quality model [CMAQ]), and multiple emissions preprocessing systems (Sparse Matrix Operator Kernel Emissions [SMOKE] and Emissions Preprocessing System version 2 [EPS2]) for 2000 and for 2006 predicted higher ozone concentrations when the SAPRC99 mechanism was used than when the CB-IV mechanism was used. The differences in predicted ozone concentrations were greatest when predicted ozone concentrations were high.

Finding I2: In regions with very high VOC reactivity and high NOx emission density, differences in ozone formation and accumulation predicted by regional photochemical models using the SAPRC99 and CB-IV mechanisms are due to differences in: (1) the chemistry of aromatics, especially mono-substituted aromatics (e.g., toluene), (2) nitric acid formation rates, and (3) the rates of free radical source terms in the SAPRC and CB-IV mechanisms.

Analysis: Faraji et al., 2007a. In reactions of mono-substituted aromatics, the CB-IV mechanism predicts a higher proportion of ring-retaining products, such as cresols, than SAPRC99. These ring-retaining products are less reactive than the ring opening products. The CB-IV mechanism also predicts: (1) more extensive nitric acid formation, (2) less extensive formation of peroxyacetyl nitrate (PAN), and (3) decreased formation of free radicals than the SAPRC99 mechanism. The SAPRC mechanism also has several source reactions for higher aldehydes that are not present in CB-IV. The enhanced free radical production in SAPRC99, as compared to CB-IV, leads to accumulation of higher concentrations of radicals.

Finding I3: The differences in the predictions of the SAPRC99 and CB-IV mechanisms can be probed using simulations of model compounds and comparisons of the simulations to environmental chamber data. For simulations involving CO and NOx (inorganic chemistry), the predictions of the CB-IV and SAPRC99 mechanisms that are used in regional photochemical models (with no chamber wall corrections) converge if the rate constants for the OH + NO2 reaction are made consistent between the two mechanisms. The predictions of the Carbon Bond mechanism, version 5 (CB05) also converge to the same predicted values if the OH + NO2 rate constant is made consistent with the other mechanisms.

Analysis: Faraji et al., 2007b. Equating the rate constant for the OH + NO2 reaction in CB-IV, CB05, and SAPRC99 mechanisms, in the form in which they are used in regional photochemical modeling, led to convergence in the concentrations of O3, OH, HO2, NO, and NO2. This suggests that in photochemical modeling simulations, the major differences in the inorganic chemistry between the mechanisms are due to differences in the rate parameters of the OH + NO2 reaction.


31 August 2007 108

However, CO - NOx experiments may not be sensitive to all aspects of inorganic mechanisms that affect simulations under ambient conditions.

Finding I4: The performance of the mechanisms in simulating olefins chemistry can be improved through more explicit representation of internal olefin chemistry, which has been added in CB05. In addition, performance of the SAPRC99 mechanism in simulating chamber data was improved for some experiments when propene was modeled explicitly, as opposed to being represented by a lumped chemical species.

Finding I5: For high reactivity chamber experiments involving olefins, sensitivity analyses indicated that mechanism adjustments that would lead to increased radical concentrations (increasing the radical yield in olefin-ozone reactions and changing the OH+NO2 termination rate constant) had little impact on predicted ozone concentrations.

Finding I6: The SAPRC99 mechanism performed better than the CB mechanisms in simulating some chamber experiments with toluene; the mechanism performances were more comparable for xylenes and other multiply substituted aromatics.

Finding I7: The differences between the SAPRC and CB-IV mechanism predictions for toluene chemistry decrease substantially if the yield for the lumped species CRES, and rate constant for the OH + NO2, are made consistent between the two mechanisms. The predictions of the CB05 mechanism also converge to the same predicted values if the CRES yield and the OH + NO2 rate constant are made consistent with the other mechanisms.

Analysis and Data: Faraji et al., 2007b. The lumped group of chemical species, CRES, as used in the photochemical reaction mechanisms, is often referred to as cresol, and comparisons between chamber experiments involving aromatics and observed cresol concentrations are generally better for the SAPRC mechanism than for the CB mechanisms. However, in the CB mechanism, the CRES species is used to represent both cresol yields and other products (such as unsaturated dicarbonyls) that react rapidly with the nitrate radical. Therefore direct comparisons of CRES yields with cresol concentrations for the CB mechanisms are not appropriate. It is clear that there are substantial differences in predictions of aromatics chemistry between the CB and SAPRC mechanisms, that the toluene mechanisms all need to be revised to improve their consistency with chamber data, and that much still remains unresolved about aromatics chemistry.

Finding I8: For simulations of ambient surrogate mixture experiments in the UCR EPA chamber, all mechanisms underpredict O3 at low ROG/NOx ratios, with the bias decreasing as the ROG/NOx ratio increases. In simulations of mixture experiments in chambers without aromatics, CB05 performs the best, and with no dependence of bias on the ROG/NOx ratio.

Analysis and Data: Faraji et al., 2007b. Figure I2 shows the results of comparisons of modeled simulations of environmental chamber experiments performed at University of California, Riverside (UCR). The performance of SAPRC99 in simulating the UCR EPA mixture experiments (both with and without aromatics) improved considerably if compounds are modeled explicitly, rather than in lumped groups.


31 August 2007 109

Full Surrogate - Arc Lights Full Surrogate - Blacklights

ROG/NOx Ratio ROG/NOx Ratio

Non-Aromatic Surrogate - Blacklights

ROG/NOx Ratio

Δ([O

3]-[N

O])

Mod

el E

rror

Note: Full surrogate experiments include experiments with formaldehyde removed.

-90%

-75%

-60%

-45%

-30%

-15%

0%

15%

30%

1 10 100 1000

-90%

-75%

-60%

-45%

-30%

-15%

0%

15%

30%

1 10 100

-90%

-75%

-60%

-45%

-30%

-15%

0%

15%

30%

1 10 100 1000

Detailed SAPRC-99CAMx SAPRC-99CB05CB4

Figure I2. Comparison of CB-IV, CB05, and SAPRC99 model errors for UCR EPA surrogate runs. (Faraji et al., 2007b)

Findings I3-I8 were arrived at by analyzing model compound simulations and environmental chamber data. These six findings are consistent with Findings I1 and I2, which are based on sensitivity analyses performed using the regional photochemical models. Additional analyses were performed using ambient data from Houston collected during a stagnation event with high olefin concentrations.


31 August 2007 110

Finding I9: SAPRC99, CB-IV, and CB05 all successfully predicted concentrations of ozone and other species during simulation of a stagnation event, if the chemical mechanisms were initialized after initial high concentrations of olefins had reacted. None of the mechanisms successfully predicted a rapid rise in radical concentrations and ozone concentration concurrent with initially high C2, C3, and CMBO concentrations during the event.

Analysis and Data: Faraji et al., 2007b. Analyses have been performed on ambient data collected during TexAQS 2000 at the heavily instrumented LaPorte site (Faraji et al., 2007b). Specifically, a period of stagnation at the LaPorte airport was simulated. This stagnation period had high concentrations of C2 and C3 alkanes and alkenes, as well as high concentrations of a marker of chlorine chemistry (CMBO – 1-chloro-3-methyl-3butene-2-one) that lasted for approximately one hour. The data suggest that a source of radicals, not accounted for in the current mechanisms was present, but no definitive source of the radicals was identified.

RECOMMENDATIONS Model compound analyses and chamber data, as well as sensitivity analyses in CAMx, indicate that major causes for the differences in SAPRC and CB mechanisms under conditions encountered in eastern Texas are the chemistry of aromatics, and to a lesser extent, radical termination rates. In choosing a mechanism that will be most effective for eastern Texas, several guiding principles emerge.

1. Over the past several decades, comprehensive updates to the CB and SAPRC chemical mechanisms have been released roughly every 5-10 years. These updates incorporate changes to the mechanisms that reflect evolving knowledge about rate constants and key chemical pathways. For example, the rate parameters for the critical OH + NO2 rate constant have been adjusted in CB05, as compared to CB-IV, and in SAPRC07, as compared to SAPRC99. In addition, the updates often provide explicit representations of chemical species that have recently been identified as being especially important. For example, explicit isoprene chemistry was incorporated into both SAPRC and CB mechanisms in the 1990s and the newly released CB05 incorporates internal olefins as an explicit compound class. These updates generally reflect more current knowledge, and, as a guiding principle, should be used when possible.

2. Recognizing that it may be difficult to unambiguously define the most appropriate mechanism for use in Houston, the effectiveness of proposed control strategies should be evaluated using multiple chemical mechanisms.

3. As new comprehensive updates to the CB and SAPRC mechanisms are developed, more explicit representation of both key chemical pathways (e.g., the low NOx routes for aromatics) and key chemical species (propene, toluene, and possibly other aromatics) would likely make the mechanisms more useful in analyzing the complex industrial emission sources in the Houston area.


31 August 2007 111

KEY CITATIONS AND INFORMATION AND DATA SOURCES Carter, W.P.L. 2004. Evaluation of a Gas Phase Atmospheric Reaction Mechanism for Low NOx

Conditions. Final Report to the California Air Resources Board, Contract No. 01-305. http://pah.cert.ucr.edu/~carter/bycarter.htm.

Faraji, M., Y. Kimura, E. McDonald-Buller, and D. Allen. 2007a. Comparison of the Carbon Bond and SAPRC photochemical mechanisms under conditions relevant to southeast Texas. Atmospheric Environment, in press.

Faraji, M., G. Heo, Y. Kimura, E. McDonald-Buller, D. Allen, G. Yarwood, G. Whitten, and W. Carter. 2007b. Comparison of the Carbon Bond and SAPRC photochemical mechanisms. Report to the Texas Commission on Environmental Quality, August, 2007.

Yarwood, G., S. Rao, M. Yocke, and G.Z. Whitten. 2005. Updates to the Carbon Bond Mechanism: CB05. Report to the U.S. Environmental Protection Agency (December). www.camx.com/publ/pdfs/CB05_Final_Report_120805.pdf.


31 August 2007 112

Response to Question J

QUESTION J How well do air quality forecast models predict the observed ozone and aerosol formation? What are the implications for improvement of ozone forecasts?

BACKGROUND An assessment of seven air quality forecast models (AQFMs) operating in real time during TexAQS 2006 focused on skill at predicting maximum 8-hour-average O3 at 119 sites and 24-hr-average PM2.5 levels at 38 sites in eastern Texas, western Louisiana, Arkansas and Oklahoma. These models included: the NCEP CMAQ-WRF model (12-km horizontal resolution), two versions of the NOAA/ESRL WRF-Chem model (12- and 36-km res.), the Canadian CMC AURAMS (28-km) and CHRONOS (21-km res.) models, one version of the Baron-AMS MAQSIP-RT model (15-km res.), and the University of Iowa STEM model (12-km res.). Details on each of these modeling systems can be found in McKeen et al. (2005), and internet web links listed below. Three standard statistical parameters, Mean Bias, Root Mean Square Error (RMSE), and correlation coefficients were evaluated in this preliminary work. Skill for these three statistics is measured relative to a persistence forecast, or the forecast that tomorrow’s AQ levels are the same as today’s observed levels. In addition, bias-corrected model forecast values are calculated based on the mean O3 or PM2.5 bias at each site at each hour of the day, averaged over the previous 7 days. The statistical parameters are also calculated from the ensemble of the models, and the ensemble of bias-corrected models. The summary statistics for the models and their ensemble are shown in Fig. J1 for O3 and Figure J2 for PM2.5. Also shown are the RMSE for persistence and climatology forecasts, and correlation coefficients for persistence. Categorical statistics (measures of how well a model predicts an event) related to the predictions of daily maximum 8-hour-average O3 levels exceeding the 85 ppbv regulatory threshold are also summarized for the seven models, their mean ensemble, and with two bias-correction options (Figures J3 and J4). The four categorical statistics evaluated in this work were: Probability of Detection, False Alarm Rate, Critical Success Index, and Bias Ratio.

The forecasts of roughly 25 chemical, aerosol, meteorological, and radiation variables from the seven models have been compared with data collected on board the WP-3D aircraft for individual transects and vertical profiles on a flight-by-flight basis for the first 12 flight legs of the aircraft deployment. These comparisons are available for viewing at the publicly available NOAA web page, http://www.esrl.noaa.gov/csd/2006/modeleval/. Concentrations and fluxes of key atmospheric constituents upwind and downwind of Houston and Dallas have also been compared, allowing assessments of model reliability in terms of emissions inventories, and efficiencies of O3 and PM2.5 formation from these large urban sources.

The University of Houston also made nine AQ forecasts using three model resolutions and three emission scenarios with the MM5/CMAQ model. Detailed statistical summaries for O3 and

Air Quality Forecast Models

AURAMS – A Unified Regional Air-Quality Modeling System

MAQSIP-RT – Multiscale Air Quality Simulation Platform-Real Time

CHRONOS – Canadian Hemispheric and Regional Ozone and NOx System

CMAQ5x/WRF – Community Multi-scale Air Quality model/Weather Research Forecast model

University of Iowa STEM – Sulfur Transport and Emissions Model

WRF/Chem – Weather Research Forecast model/Chemistry version


31 August 2007 113

PM2.5, for each forecast, and for each of the CAMS surface monitors can be found at the University of Houston web page, http://www.imaqs.uh.edu/ftp/AQF_usa/ (password protected).

Figure J1. Summary statistics for 8-hour maximum ozone for nine air quality forecast models and their ensemble mean and bias-corrected ensemble mean.


31 August 2007 114

Figure J2. Summary statistics for 24-hour-average PM2.5 for seven air quality forecast models and their ensemble mean and bias-corrected ensemble mean.


31 August 2007 115

FINDINGS

Finding J1a: Most forecast models exhibit skill in predicting maximum 8-hr-average O3 but none of the models is better than persistence in predicting 24-hr-average PM2.5 levels.

Analysis: Wilczak et al.-NOAA. Model Data: McQueen et al.-NOAA/NWS, Grell et al. NOAA/GSD, Carmichael et al.-U. Iowa, McHenry et al.-Baron AMS, Gong et al.-Met. Srv. Canada, Bouchet et al.-Met. Srv. Canada. When all O3 data are considered:

– Nearly all models beat persistence RMSE and r-coefficients without bias correction. – The ensemble performed better than all individual models, and the bias-corrected

ensemble showed the best statistical performance. – Bias correction improved RMSE scores appreciably and r-coefficients to some degree.

When all PM2.5 data are considered: – In contrast to O3, no model or ensemble, or bias correction, beat persistence. – All but one model was biased low, and bias correction improved r-coefficients

significantly. – Since they are not included in the models, the occurrence of Saharan dust events during

August and September influenced the bulk PM2.5 statistics. Removal of these events from the statistical evaluations is another required step in model analysis.

Figure J3. Categorical statistics related to the 85 ppbv exceedance for 8-hour maximum ozone for seven air quality forecast models and their ensemble.


31 August 2007 116

Finding J1b: Seven air quality forecast models and their ensemble were generally unable to forecast 85 ppbv 8-hr-average O3 exceedances with any reliability.

Analysis: McKeen et al.-NOAA. Model Data: McQueen et al.-NOAA/NWS, Grell et al. NOAA/GSD, Carmichael et al.-U. Iowa, McHenry et al.-Baron AMS, Gong et al.-Met. Srv. Canada, Bouchet et al.-Met. Srv. Canada.

– Although the 7-model ensemble had the best overall bulk statistics, and was biased high by 5 ppbv, its ability to predict an exceedance event was significantly worse than persistence.

– The ensemble, like most of the models, predicts far fewer events, and missed most of the exceedance events in the Houston area (see Figure J4). This collective model under-prediction could be due to the lack of an important source component within all the models, such as highly reactive VOC emissions.

– Only one model beat persistence for both probability of detection and false alarm rate. – Since most models and the ensemble are biased high, bias correction tended to reduce a

model’s ability to predict an exceedance event even further. This was in sharp contrast to a similar categorical analysis for New England during the summer of 2004 (McKeen et al., 2005), where bias corrections nearly always improved the critical success index.

Figure J4. Number of times the 85 ppbv exceedance for 8-hour maximum ozone occurred within the reported AIRNow observations, and the 7-model ensemble forecast.


31 August 2007 117

Finding J2: Sophisticated data assimilation of meteorological and even chemical observations is essential for improving photochemical model forecasts.

Analysis: Nielsen-Gammon et al.-Texas A&M U., Byun et al.-U. Houston, McKeen et al.-NOAA. Model Data: Nielsen-Gammon et al.-Texas A&M U., Byun et al.-U. Houston, McQueen et al.-NOAA/NWS, Grell et al. NOAA/GSD, Carmichael et al.-U. Iowa, McHenry et al.-Baron AMS, Gong et al.-Met. Srv. Canada, Bouchet et al.-Met. Srv. Canada. Spatial and temporal accuracy of AQ forecasts are to a large degree limited by the accuracy of the underlying meteorological forecasts within the AQ models. Most models rely on the available NCEP/NAM model product for initialization and boundary conditions, which may contribute to model biases in AQ model wind fields and pressure patterns documented throughout the field study. Retrospective forecasts and sensitivity studies are needed to assess the impact of the NAM forecasts available in the summer of 2006 to AQFMs, particularly in light of recent upgrades to the WRF-NMM model used in the NAM forecasts. Assimilation of wind profiler data (Nielsen-Gammon et al., 2006) has been shown to improve forecast meteorology in the Houston area. Similar research related to the assimilation of photochemical and aerosol data within AQFMs should be encouraged, utilizing the comprehensive data sets from the TexAQS 2006 field study.

Daytime planetary boundary layer (PBL) heights from seven of the models were compared to those derived from wind profiler measurements at several sites in east Texas. Some models showed persistent bias in daytime PBL, and no model appeared to stand out in terms of significantly better comparisons. Deficiencies in AQ forecasts from several of the models appear to be related to the formulation of PBL height and vertical transport within the PBL. This is shown in Figure J5, which shows a wide spread in the models’ ability to simulate water vapor and NOy profiles compared to WP-3D aircraft data for afternoon conditions on a particular clear day. It is quite common for models to exhibit unrealistic high biases in O3 and PM2.5 precursors downwind of source regions when PBL heights are under-predicted. The analogous misrepresentation of the stable PBL often affects model-predicted offshore pollutant transport and pollution precursor buildup from emissions along the coastlines. Persistent errors in the forecast of the low-level nocturnal jet are also characteristic of many models. Preliminary sensitivity studies with the MM5/CMAQ model also have demonstrated a case of over-predicted O3 associated with a missed forecast of widespread precipitation (24, 25 August). The collection of available satellite, radar, and surface network data sets for comparing cloudiness, precipitation, and radiation with model output are needed to perform further evaluations of these parameters, and their relationships to O3 and PM2.5 forecasts.


31 August 2007 118

Figure J5. Profiles of water vapor, relative humidity, and NOy on 25 September 06, south and downwind of Dallas. Black dots indicate observations. The horizontal transect shows high biases of NOy for those models with the lowest PBL heights and reduced vertical transport.


31 August 2007 119

Finding J3: Model performance evaluations and intercomparisons require a comprehensive, best-guess emissions inventory for the TexAQS 2006 Field Intensive.

Analysis: Byun et al.-U. Houston; Data: TCEQ. Ozone and PM2.5 forecasts are highly dependent on the emissions inventories of precursor emissions, and PM2.5 forecasts are also dependent on primary emissions at many of the urban and suburban CAMS locations. A high priority in the model evaluation effort should be placed on using TexAQS 2006 field data to determine the accuracy of the inventories that drive AQ forecasts. Ozone forecasts are particularly sensitive to emissions estimates of HRVOC, such as ethylene and propylene, from large petrochemical facilities, especially in the Houston ship channel region. Quality-assured VOC measurements from the various platforms and comparisons with AQ model results are currently in progress in order to evaluate the model emissions inventories. Two preliminary sensitivity results from the University of Houston MM5/CMAQ model relate directly to emissions inventory validation. That model shows generally improved NO2 and O3 comparisons with CAMS site data using an emission inventory based on projections to 2005 as opposed to an inventory based on 2000 estimates. A narrow plume of extremely high O3 observed downtown and west of Houston (7 September) that was significantly under-predicted by the original forecast was found to be replicated accurately in retrospective runs that included a VOC source in the ship channel region much larger than specified in the base emissions inventory.

KEY CITATIONS AND INFORMATION AND DATA SOURCES McKeen, S., J. Wilczak, G. Grell, I. Djalalova, S. Peckham, E.-Y. Hsie, W. Gong, V. Bouchet, S.

Menard, R. Moffet, J. McHenry, J. McQueen, Y. Tang, G.R. Carmichael, M. Pagowski, A. Chan, T. Dye, G. Frost, P. Lee, and R. Mathur. 2005. Assessment of an ensemble of seven real-time ozone forecasts over eastern North America during the summer of 2004. J. Geophys. Res., vol. 110, D21307, doi:10.1029/2005JD005858.

Nielsen-Gammon, J. W., R.T. McNider, W.M. Angevine, A.B. White, and K. Knupp. 2007. Mesoscale model performance with assimilation of wind profiler data: Sensitivity to assimilation parameters and network configuration. J. Geophys. Res., vol. 112, D09119, doi:10.1029/2006JD007633.

Model Acronyms and Web Links AURAMS (A Unified Regional Air-Quality Modeling System)

http://www.msc-smc.ec.gc.ca/research/icartt/aurams_e.html BAMS (Baron Advanced Meteorological Systems, Inc.)

http://www.baronams.com/projects/SECMEP/index.html CHRONOS (Canadian Hemispheric and Regional Ozone and NOx System)

http://www.msc-smc.ec.gc.ca/aq_smog/chronos_e.cfm CMAQ5x/WRF (Community Multi-scale Air Quality model/Weather Research Forecast model)

http://wwwt.emc.ncep.noaa.gov/mmb/aq/ University of Iowa STEM (Sulfur Transport and Emissions Model)

http://nas.cgrer.uiowa.edu/MIRAGE/mirage-2k6.html WRF/Chem (Weather Research Forecast model/Chemistry version)

http://www.wrf-model.org/WG11


31 August 2007 120

Response to Question K

QUESTION K How can observation and modeling approaches be used for determining (i) the sensitivities of high ozone in the HGB non-attainment area to the precursor VOC and NOx emissions, and (ii) the spatial/temporal variation of these sensitivities?

BACKGROUND The accurate prediction of the relative response of ozone concentrations to future reductions in NOx and VOC emissions is a much sought, but very elusive goal. This prediction is central to the very important SIP-relevant question of “direction of control” – that is, should ozone control efforts in an ozone non-attainment area be focused on: a) decreasing emissions of NOx alone, b) decreasing emissions of VOC alone, or c) decreasing emissions of both NOx and VOC? In this report the closely related Questions F and K both deal with aspects of this issue. The Background Section of the Response to Question F discusses some issues relevant to this Question and Response; the interested reader should peruse that discussion as a prelude to this Response to Question K and also review the recently revised Guidance on the Use of Models and Other Analyses for Demonstrating Attainment of Air Quality Goals for Ozone, PM2.5, and Regional Haze (USEPA, 2007). One point made in the background section of the Response to Question F is that a definitive answer to these questions requires a thoroughly tested Eulerian emissions-based air quality model that treats the region of interest with sufficient accuracy. Another point is that this model testing should involve comparisons between model results and direct observations of the relationships between ozone, its precursors, and the radicals that drive these reactions. Unfortunately however, it has not yet been possible to develop such generally applicable models because of deficiencies and uncertainties in many critical areas, including emissions, meteorological modeling, and photochemical mechanisms.

In the absence of a “perfected” Eulerian air quality model for the HGB non-attainment area, it has been necessary to take a variety of heuristic approaches to provide guidance to air quality managers. Heuristic implies a model that is simplified, but that is designed and used to learn more about some specific but important aspect of the more complex air-quality system that is to be managed. The Response to Question F discusses the application of both observation-based and emissions-based modeling approaches for approximately determining the sensitivity of ozone production and concludes that both VOC and NOx controls are presently effective in reducing the maximum observed ozone concentrations. Continued incremental reductions in emissions of either VOC or NOx emissions can be expected to yield incremental improvements in ozone design values in the HGB region.

Here the response to Question K addresses two very important questions that are related, but importantly different, in that they address how compliance with the current NAAQS for ozone can be achieved in an attainment demonstration:

• What emission controls will ultimately be necessary for the HGB area to achieve compliance with the 8-hour-average, 80 ppbv ozone standard?

• What is the most effective strategy by which to reach this compliance?

To address these questions, a very simple, emissions-based model was applied to the HGB region.


31 August 2007 121

FINDINGS

Finding K1: A simple, heuristic model based upon the Empirical Kinetic Modeling Approach (EKMA) method suggests that the HGB region may ultimately require drastic NOx emissions controls to reach compliance with the NAAQS for ozone.

Finding K2: The same model suggests that in a projected future scenario with very strict VOC emission controls, but without drastic NOx emission controls, biogenic VOC emissions plus background concentrations of CO and methane may be sufficient to cause ozone exceedances.

Analysis: Dimitriades-NCSU; Luecken-USEPA. A simplified model of an inherently complex system can sometimes provide valuable insight into the behavior of the complex system (Dimitriades, 2006; Dimitriades and Luecken, 2007). As a case in point, here we investigate a very simple, Empirical Kinetic Modeling Approach (EKMA) that includes an Ozone Isopleth Plotting Research Package (OZIPR) [see e.g. Finlayson-Pitts and Pitts, 2000]. This approach comprises a photochemical box model, and was applied to the HGB area in order to illuminate one aspect of the complex chemical-transport system that defines the near-surface atmosphere in the HGB region. The model was formulated to provide an answer to the very critical question: To reach compliance with the NAAQS ozone standard in HGB, what emission controls are necessary and most effective?

The model utilized an advanced version of the EKMA method for translating early morning (8 AM) ambient concentrations of VOC and NOx into the maximum 8-hour-average ozone concentration that accumulates by the end of the day. The model incorporates a SAPRC99 chemical mechanism and requires model and observational inputs on VOC composition, sunlight intensity, diurnally varying VOC (including biogenic VOC) and NOx emissions, diurnally varying boundary layer height, and background ozone concentration. The results presented here were obtained for:

(i) A “typical” Harris County VOC mix, based on measurements at the LaPorte site during TexAQS 2000. (ii) A constant 30-ppbv concentration of background ozone. (iii) Time-varying post-8 AM VOC and NOx emissions based on average daily anthropogenic emissions reported for Harris county in 2002, apportioned as hourly emissions based on modeled profiles for total NOx and anthropogenic VOC emissions from EPA’s 2001 National Emissions Inventory processed using the SMOKE emissions processor (http://www.smoke-model.org). (iv) Time-varying post-8 AM isoprene emissions estimated from the area average isoprene emissions for Harris County from the HGB 2002 emissions inventory (http://www.tceq.state.tx.us/assets/public/implementation/air/sip/hgb/hgb_sip_2007/appendices/06027SIP_Appendix_F_PEI.pdf). These emissions totaled 29 kg km-2day-1. (v) a typical time-varying mixing height based on measurements made at the Moody Tower site.

The results of the model are presented in the form of a traditional EKMA diagram in Figure K1. For illustration purposes, the method was applied to 8 AM NOx and VOC concentrations measured at five sites (Clinton, Channelview, Wallisville, Lynchburg Ferry, and HRM-3) within


31 August 2007 122

Harris County during April-October of 2005 and 2006. A single ozone concentration isopleth diagram was constructed from the calculated maximum 8-hour-average ozone concentrations that accumulated in the model initiated with those 8 AM measured concentrations.

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 28 AM anthropogenic VOC (ppmC)

g pp p p

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

8 A

M N

Ox

(ppm

)

A

BC

C

Figure K1. EKMA diagram constructed for the HGB region. The contour lines give the maximum 8-hour-average ozone concentration (in ppbv) that accumulated starting with the morning VOC and NOx concentrations given on the respective axes. The critical, dashed line isopleth represents the maximum ozone concentration allowed in an area in compliance with the NAAQS.

Figure K1 suggests that the ozone sensitivity conditions within Harris County vary from VOC-sensitive to NOx-sensitive from site to site and from day to day. For example, consider the points labeled A, B, and C in the figure. The two points labeled C represent non-exceedance conditions – they lie below the critical (84 ppbv) dashed line isopleth. Point A represents NOx-limited exceedance conditions. The predicted maximum 8-hour average ozone concentration is above 84 ppbv, but by decreasing NOx emissions sufficiently, the 8 AM NOx concentrations will decrease so that Point A will cross the critical dashed line isopleth. Similarly, Point B represents VOC-limited exceedance conditions; by decreasing VOC emissions sufficiently, the 8 AM VOC concentrations will decrease so that Point B will cross the critical dashed line isopleth.


31 August 2007 123

There is a critical difference in the behavior that the EKMA model predicts for the NOx-limited Point A and the VOC-limited Point B. Point B suggests that attainment can be most easily reached by reducing VOC emissions, but attainment can also be reached by reducing NOx emissions if the reductions are drastic enough. Initially at point B, NOx emission reductions alone would first increase the predicted maximum ozone concentration, but after a peak is reached, further NOx emission reductions would lead to a decrease in the predicted maximum ozone concentration until Point B crosses the critical isopleth. In contrast, VOC reductions alone applied to Point A can never bring that point below the critical isopleth, because the maximum ozone concentration is predicted to be above 84 ppbv even when the 8 AM anthropogenic VOC concentrations are zero. This model behavior occurs because biogenic VOC emissions alone (with background CO and methane concentrations), in the presence of sufficient NOx, can produce ozone concentrations in exceedance of the ozone standard. This behavior of Point A is similar to the predictions from the Eulerian modeling presented in Figure F3, which show that even extremely stringent VOC controls cannot decrease the HGB ozone design values enough to achieve compliance. Importantly, the EKMA model suggests that NOx-sensitive conditions as exemplified by Point A are relatively common in HGB. Of the days investigated, 54% for Clinton, 17% for Channelview, 35% for Wallisville, 39% for Lynchburg Ferry, and 43% for HRM-3 exhibited Point A-like, NOx-limited behavior.

Clearly the EKMA model is highly simplified, but it is notable that the NOx-sensitive behavior exemplified by Point A may be reasonably well represented by the model. One of the most critical simplifications in the model is the assumption of a “typical” Harris County VOC mix. It is well known that the VOC composition in HGB changes markedly, from the typical urban mix of downtown Houston to the widely varying VOC emissions from the petrochemical facilities. However, the critical Point A behavior is focused at zero anthropogenic VOC concentration, so the assumed VOC composition will have no effect.

Given the EKMA model prediction that drastic NOx emission controls will be required to eventually reach ozone compliance in HGB, the extremely difficult but central question arises:

How should air quality managers in the HGB area divide ozone management resources between NOx and VOC emission controls?

Several considerations favor control of VOC emissions, including:

• First, the Response to Question F indicates that incremental VOC emission reductions presently will result in incremental decreases in ambient ozone concentrations. Any progress in decreasing the maximum ambient ozone concentrations will have immediate health and welfare benefits.

• Second, Figure K1 does indicate that there is a modest advantage to VOC control when used in conjunction with NOx control. For example, if the 8 AM VOC concentration is reduced from 0.4 to 0.1 ppm C, then NOx controls need only reduce the 8 am NOx concentration to about 25 rather than about 20 ppbv. Figure F3 also indicates a similar benefit of VOC control.

• Third, VOC emissions reductions can decrease ambient concentrations of VOC hazardous air pollutants. For example, the ambient concentrations of formaldehyde, which in the HGB area is primarily a secondary product of HRVOC oxidation, will respond to HRVOC emission controls.


31 August 2007 124

Similarly, several considerations favor control of NOx emissions, including:

• First, NOx emissions will reduce not only locally produced ozone, but also regional background ozone. Outside of central urban areas and point source plumes, the regional ozone production environment is generally strongly NOx-limited.

• Second, all of the modeling approaches agree that sufficiently drastic NOx control is certain to lead to ozone attainment. Further, to the extent that the EKMA results reflect reality, such drastic control may be required to ultimately reach compliance. However, there is uncertainty whether the high degree of NOx control required for ozone attainment is socio-economically acceptable or even technologically feasible.

• Third, NOx controls will reduce local and regional levels of both nitrate aerosol, a significant contributor to PM2.5 concentrations, and nitric acid, a significant component of acid deposition. Both of the nitrate contribution to PM2.5 and the nitric acid contribution to acid deposition are particularly important issues in the western U.S.

The most efficient emission control strategies will then depend upon practical and economic considerations associated with the alternatives. Given that the Eulerian and EKMA models upon which all the preceding discussions are based are still imperfect, it is prudent to implement what VOC controls and/or NOx controls are presently beneficial and practical, while modeling improvements are effected.

KEY CITATIONS AND INFORMATION AND DATA SOURCES Dimitriades, D. 2006. A Proposed Study for Determining the Optimum Direction of Control in

the Houston-Galveston-Brazoria Ozone Non-Attainment Area. NC State University Report to the Texas Commission on Environmental Quality. Submitted October 10, 2006. College of Natural Resources, NC State University, Raleigh, NC.

Dimitriades, B. and D. Luecken. 2007. An observational investigation of ozone sensitivity to VOC and NOx and it variation in the Houston-Galveston-Brazoria (HGB) ozone non-attainment area. Presentation at TexAQS II Principal Findings Data Analysis Workshop, 29 May-1 June 2007. Austin, TX. http://www.tceq.state.tx.us/assets/public/implementation/air/texaqs/workshop/20070529/2007.06.01/6-Spatial+Temporal_Patterns-Dimitriades.pdf

Finlayson-Pitts, B.J. and J.N. Pitts. 2000. Chemistry of the Upper and Lower Atmosphere. pp. 892-893. Academic Press, New York, NY.

US Environmental Protection Agency. 2007. Guidance on the Use of Models and Other Analyses for Demonstrating Attainment of Air Quality Goals for Ozone, PM2.5, and Regional Haze. EPA Report 454/B-07-002. April 2007. EPA Office of Air Quality Planning and Standards, Air Quality Analysis Division, Air Quality Modeling Group, Research Triangle Park, NC. 262 pp. http://www.epa.gov/ttn/scram/guidance/guide/final-03-pm-rh-guidance.pdf


31 August 2007 125

APPENDIX 1. GLOSSARY OF ACRONYMS (The meaning of some of the acronyms more commonly used in the report are listed here.) AIRS Atmospheric Infrared Sounder AIRS Aerometric Information Retrieval System AOD Aerosol Optical Depth AQ Air Quality AQFM Air Quality Forecast Model AURAMS A Unified Regional Air-Quality System BEIS Biogenic Emissions Inventory System BL Boundary Layer CALIPSO Cloud Aerosol Lidar and Infrared Pathfinder Satellite Observation CAMS Community Air Modeling System CAMx Comprehensive Air Quality Model, with extensions CB Carbon Bond CB05 Carbon Bond Photochemical Mechanism, version 5 CB-IV Carbon Bond Photochemical Mechanism, version IV CEMS Continuous Emissions Monitoring System CHRONOS Canadian Hemispheric and Regional Ozone and NOx System CMAQ Community Multi-scale Air Quality Model CMBO 1-chloro-3-methyl-3butene-2-one CRES Lumped group of chemical species in photochemical reactions

mechanisms that includes cresol CST Central Standard Time DFW Dallas-Fort Worth DIAL DIfferential Absorption Lidar EGU Electric power Generating Unit(s) eGRID Emissions and Generation Resource Integrated Database EPA US Environmental Protection Agency EPS2 Emissions Processing System version 2 EKMA Empirical Kinetic Modeling Approach HGB Houston-Galveston-Brazoria HRVOC Highly Reactive Volatile Organic Compound(s) HSC Houston Ship Channel HSRL High Spectral Resolution Lidar HYSPLIT HYbrid Single-Particle Lagrangian Integrated Trajectory model LPAS Laser Photo-Acoustic Spectroscopy LST Local Standard Time MAE Mean Absolute Error MAQSIP-RT Multiscale Air Quality SImulation Platform-Real Time MM5 Meteorological Model version 5 MOBILE6 EPA Mobile Source Emission Factor Model version 6 MODIS Moderate Resolution Imaging Spectroradiometer MSA Metropolitan Statistical Area MSD Medium Speed Diesel NAAQS National Ambient Air Quality Standards


31 August 2007 126

NASA National Aeronautics and Space Administration NCAR National Center for Atmospheric Research NCEP National Centers for Environmental Prediction NEI National Emissions Inventory NOAA National Oceanic and Atmospheric Administration NOAA/ESRL NOAA Earth System Research Laboratory NOAA/NESDIS NOAA National Environmental Satellite Data and Information Service OMI Ozone Monitoring Instrument OZIPR Ozone Isopleth Plotting Research Package PAN PeroxyAcetyl Nitrate PBL Planetary Boundary Layer PM2.5 Particulate Matter less than 2.5 micrometers in diameter PTR-MS Proton Transfer Reaction-Mass Spectrometer RAQMS Realtime Air Quality Modeling System ROG Reactive Organic Gases RHB research vessel Ronald H. Brown RMSE Root Mean Square Error RSST Rapid Science Synthesis Team SAPRC [California] Statewide Air Pollution Research Center SAPRC99 [California] Statewide Air Pollution Research Center (photochemical

mechanism 1999 version) SAPRC07 [California] Statewide Air Pollution Research Center (photochemical

mechanism 2007 version) SIP State Implementation Plan SMOKE Sparse Matrix Operator Kernel Emissions SOF Solar Occultation Flux SOS-OD Southern Oxidants Study-Office of the Director SSD Slow Speed Diesel STEM Sulfur Transport and Emissions Model TCEQ Texas Commission on Environmental Quality TES Tropospheric Emission Spectrometer TexAQS 2000 First Texas Air Quality Study, completed during the summer of 2000 TexAQS 2006 Intensive field measurement study conducted during 11 weeks of TexAQS

II TexAQS II Second Texas Air Quality Study, completed during 18 months in 2005 and

2006 TRI Toxics Release Inventory UHI Urban Heat Island UTC Coordinated Universal Time VOC Volatile Organic Compound(s) WAS Whole Air Samples WP-3D Instrumented aircraft operated by NOAA WRF/Chem Weather Research Forecast model/Chemistry version


31 August 2007 127

APPENDIX 2. FINAL REPORT FROM QUESTION L WORKING GROUP.


31 August 2007 128

Question L Final Report

QUESTION L What existing observational databases are suitable for evaluating and further developing meteorological models for application in the HGB area?

QUESTION L WORKING GROUP Leader: Lisa Darby; Participants: Robert Banta, John Nielsen-Gammon, Daewon Byun, Wayne Angevine, Mark Estes, Bryan Lambeth, Stuart McKeen.

BACKGROUND In order to address this question, databases that are potentially useful to individuals performing air quality modeling for Texas, including both permanent measurements and enhanced measurements from TexAQS II deployments, were compiled. The databases were evaluated based on several criteria, including quality control, accessibility, regional coverage, and time resolution. Web links to the databases are given, followed by a brief description and evaluation.

FINDINGS

Surface Meteorology and Chemistry Data

COOP observations http://www7.ncdc.noaa.gov/IPS/CDPubs?action=getstate Consists of monthly printed pages (available as PDFs online) containing station observations. No description of QA is provided on the web site.

Because these data are not in machine-readable format, they are unlikely to be useful for any systematic study. They are redundant, in the sense that the same observations should be in the normal NWS data streams.

Crop Weather Program, Texas A&M University http://cwp.tamu.edu/cgi-bin/htmlos.cgi/6742.2.1749378041063346623 The Crop Weather Program for South Texas (CWP) was developed to help farmers and consultants make management decisions conducive to profitable crop production. It replaces an earlier cotton monitoring system known as the Weather Station Network Program. The CWP is the gateway for access to weather data measured by a network of 21 automated weather stations spread across 10 South Texas counties and provides hourly measurements of air temperature, relative humidity, solar radiation, wind direction and speed, precipitation, and soil temperature at 1", 3", and 8" depths. The wind direction is reported based on a 16-point compass and the wind speed appears to be arithmetic (no vector average direction or speed). The wind also appears to be measured about 10 feet above ground level based on an example site photo provided (this could exacerbate exposure problems where buildings and/or trees are nearby).

Harris County Office of Homeland Security and Emergency Management http://www.hcoem.org/ The Harris county rainfall map site allows you to enter an amount of time (in days, hours, or minutes) before the current time, and it produces a map of accumulated rainfall amounts from each site, over the time requested. The data come from 163 automatic remote sensors (part of


31 August 2007 129

the flood alert system) across the metropolitan area, and they are “unofficial” (probably means not quality-controlled). The density of the network allows for detailed information regarding the horizontal distribution of the rainfall. Their locations can be found on a map link and a text link, which includes latitudes and longitudes. There is a link to an archive site where you can indicate a given amount of time before your date of interest to obtain a map of accumulated rainfall, but I could not get this part to work. If this does eventually work, this could be a useful site for modelers, although it looks like the only output would be a map (i.e., no text dump). I suggest a following up on this site to determine if there is a way to order the archived data.

Also on the main page for Harris County Office of Homeland Security and Emergency Management is a link to a real time Houston speed map. Along the outlines of the major highways the current speed of traffic is shown in color (indicating speeds <20, 20-29, 30-39, 40-49, and 50+ MPH, or no data). On this site is a link to the Houston speed map archives (http://traffic.houstontranstar.org/map_archive/map_archive.aspx). From this site you can select a date and time (down to 15-minute intervals) and you get a traffic speed map for that time. This could be useful to determine if gridlock was worse on some days compared to others.

Texas A&M data http://dallas.tamu.edu/Weather/index.html This web site has data from two sites near Dallas. The sites are run by the Texas A&M Dallas Agricultural Research and Extension Center (phone: 972.231.5362), and details are sketchy. The locations are not specified, although one is on a research farm (Prosper) and the other is just called “Dallas.” The “Dallas” site has, by date, max/min soil temperature, max/min air temperature, max/min RH, a single column labeled “wind” (no units indicated on any of the columns), max/min soil moisture, and total rain. Some years have a column labeled ET_o (evapotranspiration?). At the end of each month is a row for monthly medians for each column and another row with the max, min, or total for each column (depending on the variable). The Prosper site has the same variables, plus “RAD” (radiation?), wind speed, wind direction and battery voltage. The Dallas site has data archived from 2000 and the Prosper site has data archived from 1997. Given how important soil moisture measurements are for modelers, it may be useful to investigate this database further to determine the location of the sites and the robustness of the soil moisture data.

Lower Colorado River Authority network http://hydromet.lcra.org/index2.shtml Lower Colorado River Authority network. This web page has a wealth of information regarding measurements throughout the Colorado River watershed (which extends from NW to SE of Austin, becoming quite narrow at Matagorda Bay). The network is most dense around Austin. They have: rainfall (24-hr accumulation, accumulation since midnight, and the most recent measurement); stage, flow, lake level, air temperature, relative humidity, and conductivity data, shown on maps. You can download historic data for a single site (precipitation, air temperature, relative humidity, wind speed and direction), but this is not very practical for obtaining data from many sites. It is stated that real-time data are provisional, but there is no indication about the quality of the archived data. It may be worth investigating if it is possible to obtain archived data directly from the agency.


31 August 2007 130

Texas A&M agricultural weather site http://texaset.tamu.edu/weatherstns.php This one would be useful for Texas meteorological comparisons of precipitation (not many CAMS sites have precipitation), temp, RH, wind speed, wind direction, and solar radiation. Pretty good coverage in East Texas. Hourly data should be simple to download and compare with model results.

Soil Climate Analysis Network, US Agriculture Department http://www.wcc.nrcs.usda.gov/scan/ This site only gives soil parameters - no standard met. There is only one site (Prairie View) in the region of Texas that may be useful to TexAQS 2006 participants. Nonetheless, it may useful for comparison of soil models and parameterizations in meteorological models, since soil data is so sparse in the east Texas region.

Louisiana agricultural weather data network http://www.agctr.lsu.edu/subjects/weather/ This site is specific to the state of Louisiana. It gives meteorological and soil parameter data at about 20 sites evenly distributed throughout Louisiana. The data are not so convenient to download. But the soil parameters may be useful for meteorological model soil data comparisons.

Louisiana Universities Marine Consortium weather network. http://weather.lumcon.edu/ This web site includes measurements from 5 sites in Louisiana run by LUMCOM (Louisiana Universities Marine Consortium). Four of the sites are on platforms over water. The 5th site is somewhat inland, but looks like it’s in a marshy/wetlands type of area. The Lake Pontchartrain station is in the northern part of the lake. Meteorological data include: atmospheric pressure, humidity, temperature, winds, solar radiation, and precipitation. Hydrographic instrumentation include: chlorophyll probe, conductivity probe, and a sonde, 6600. Three other sites have the same instruments: Tambour Bay, Southwest Pass/Miss River and LUMCOM. The Tambour Bay and Southwest sites are off the LA coast, on platforms. LUMCOM is the slightly inland site and also has a co-located 915-mHz wind profiler with RASS. The Audubon/Miss River site is on a floating structure near the coast (from the picture is looks like it’s in a harbor). It only has atmospheric pressure, humidity, and temperature for meteorological measurements. It has the same hydrographic measurements as Lake Pontchartrain, plus a wet chemical in-situ nitrate analyzer.

The web site is comprehensive, with a map and much information for each station. There are records regarding calibrations and inspections, implying that these sites are well maintained. These appear to be good sites for modelers to obtain coastal meteorological data for Louisiana. The archived files are easy to access.


31 August 2007 131

CAMS (TCEQ organized surface meteorological and chemical data) http://www.tceq.state.tx.us/nav/eq/mon_sites.html This website includes information regarding the details of the TCEQ measurements sites. Many sites have both meteorology and chemistry measurements. Some have just one or the other. All chemistry sites appear to have ozone measurements, but some also include NO, NO2, and perhaps other important constituents. Those with meteorology tend to have temperature and winds, perhaps precipitation. This site has two links:

1) TCEQ’s Air Monitoring Sites (Regional Map) provides details about the TCEQ's air monitoring sites and air pollution, weather and other parameters measured at each site.

2) Air Monitoring Sites (Table)

Provides a user interface to view sortable list of locations and descriptions of monitoring sites operated by the TCEQ and other entities around the state as well as link to photos of sites, lists of parameters monitored, and current measurements.

This is useful for modelers who want to know the locations of monitoring stations, and what is monitored at each station.

http://www.tceq.state.tx.us/compliance/monitoring/air/monops/historical_data.html This page provides access to two sources of pollutant and weather data. The first source, the Texas Commission on Environmental Quality (TCEQ) and local monitoring networks, provides hourly pollutant and weather data from 1972 to 2004. The second source, the U.S. Environmental Protection Agency (EPA), provides data summaries and hourly data collected since 1982 on numerous pollutants and meteorological parameters in Texas and other states.

A useful site for modelers to download hourly surface data for model evaluation.

METARs (NWS surface data) http://www.nndc.noaa.gov/cgi-bin/nndc/buyOL-001.cgi The Unedited Surface Weather Observations product consists of unedited hourly observations from over 700 U.S. locations. There is a charge to access this data online, but not if your domain is .gov, .mil, or .edu. (More details on the web site.) The time range of the available data is from July 1, 1996 to two days ago. A useful site for obtaining surface observations for model evaluation.

Oklahoma air quality monitors http://www.deq.state.ok.us/AQDnew/monitoring/index.htm This site has the details of the air quality monitoring stations in Oklahoma. There are several monitoring sites north of the Texas-Oklahoma border that would be useful for southerly flow events (e.g., looking at transport from Dallas to Oklahoma). For a graphical display, the site links to the EPA site http://www.airnow.gov/index.cfm?action=airnow.currentconditions, where the user can click on the state of Oklahoma to see the Oklahoma observations. Text data includes real-time data (for today and yesterday). The user can sort by pollutant or by station. Archived data includes 8-hour averages of ozone and CO, organized by year. Within each year is the date and amount of the 4 highest readings for each station. One-hour ozone exceedances are also available in this format. It does not appear that data other than the 4 highest readings per year are available via the web. This site is probably somewhat useful for modelers.


31 August 2007 132

Upper Air Data

ESRL (formerly ETL) Profiler Network, South Central Texas http://www.etl.noaa.gov/et7/data/ The ESRL (formerly ETL) network page allows access to real-time and archived plots of profiler winds and other profiler data. Real-time plots are provided through a clickable map interface. Archived plots and ASCII data can be downloaded for single profilers. A trajectory tool allows the calculation of forward and backward trajectories using profiler data. The site includes all regular wind profilers from the NOAA and TCEQ network as well as all those installed for the TexAQS-II field program. The data include profiler winds and signal-to-noise ratio, RASS virtual temperature and virtual potential temperature, and surface meteorological observations from profiler sites. Data should remain available for several months after the experiment, as well as the profiler trajectory tool.

NOAA National Profiler Network graphical display http://www.profiler.noaa.gov/npn/ The NOAA site used to include all permanent profilers, but now it appears to contain only the profilers in the NOAA demonstration network, including Ledbetter, Palestine, and Jayton in Texas. Users can request real-time plots or generate plots using archived data. There is considerable flexibility in the online data plotting interface. Archived data are available from the web site hosts.

Rapid Update Cycle (RUC) soundings http://rucsoundings.noaa.gov/ This sounding page allows the user to generate plots or ASCII data dumps of soundings from rawinsondes, profilers, and RUC/MAPS forecasts. The output is Java-based, allowing mouse-over data information and animation/looping of soundings. The interface requires the user to know the name or site ID’s of the stations to plot. Most of the data are available only in real-time or near-real-time, except that an online rawinsonde archive was begun early in 2006. Perhaps the most useful aspect of the web site is the ability to plot forecast soundings from the RUC model. These forecasts are available for any arbitrary location and extend up to 12 hours into the future, so they provide detailed guidance for mixing heights, vertical wind shear, and convection.

University of Wyoming sounding page http://weather.uwyo.edu/upperair/sounding.html This web site allows the user to select a station using a clickable map and generate graphical soundings or ASCII data output from real-time or archived rawinsonde observations. The output format includes all common sounding diagram types and ASCII data formats. Large amounts of data would be difficult to obtain, but this site is the best available on the web for individual archived soundings.


31 August 2007 133

ACARS aircraft observations http://amdar.noaa.gov/ ACARS observations are in situ meteorological observations made by commercial aircraft. The data include temperature, wind, and often dew point. The wind precision is not very good, but the temperature and dew point data are useful for estimating mixing heights and their diurnal variation. Most ACARS observations in Texas come from the Dallas-Fort Worth area, usually about two dozen per day. Much less frequent observations are available from Houston and other major airports. The data are not freely available in real time on the web, but they are available for research purposes upon approval by NOAA. Texas A&M presently receives ACARS data but is not funded by TCEQ to process or use the data for analysis or forecasting during 2006.

Coastal and Buoy Data

Texas Coastal Ocean Observation Network, Texas A&M Corpus Christi, Conrad Blucher Institute http://lighthouse.tamucc.edu/TCOON/HomePage Large network of coastal stations. Some of the reported stations are regular NOAA or other agency stations, and these are not identified as such. The additional stations seem to primarily provide water level, water temperature, and air temperature. Machine-readable historical data are available. Some QA is apparently done, but specifications are not easily found on the web site.

Possibly useful for improving resolution of model validations for simple parameters.

NDBC (National buoy data) http://www.ndbc.noaa.gov/Maps/WestGulf.shtml Provides listings of hourly meteorological data (air and sea-surface temperature, winds, pressure, etc.) and wave data for each meteorological buoy in the Gulf of Mexico (and elsewhere around the U.S.). Meteorological data are archived back as far as 1990 for some sites. Also a section gives data on ocean currents as a function of depth. Buoy and other instrument locations are displayed on a map, and data are obtained by clicking on the site of interest. Recent ship observations are also listed at this site, and the tri-annual Mariners Weather Log.

Houston/Galveston Port Meteorological Office http://www.srh.noaa.gov/hgx/marine/pro.htm Houston/Galveston Port Meteorological Office

Site includes a description of needs for maritime meteorological data and the role of this office in facilitation of the Voluntary Observing Ship (VOS) Program. The office also “works within the framework of the Shipboard Environmental (Data) Acquisition System (SEAS), by which meteorological data are collected and transmitted to NCEP, for inclusion in the major data bases. Under Past Weather, this site has climatological data and daily information for several Texas land stations around the Gulf of Mexico, including daily high and low temperatures, wind, precipitation, and some other meteorological data. The monthly Texas Climatic Bulletins and other climatological products and information are available at this site. We were unable to locate any actual shipboard data from this site (however, some current data could be found on the NDBC site).


31 August 2007 134

Satellite Data

Space Science and Engineering Center, University of Wisconsin, Madison http://www.ssec.wisc.edu/ Comprehensive archive of data, products, and downloadable processing software for geosynchronous and polar-orbiting satellites, including GOES-11 and -12 and MODIS data from Terra and Aqua. A host of real-time satellite images and products are also available, some stored for 7 days. Routine meteorological data are also available for McIdas users.

NASA Aura TES step and stare observations http://tes.jpl.nasa.gov TES is an infrared, high resolution, Fourier Transform spectrometer covering the spectral range 650 - 3050 cm-1 (3.3 - 15.4 µm) at a spectral resolution of 0.1 cm-1 (nadir viewing) or 0.025 cm-1 (limb viewing). Launched into a polar sun-synchronous orbit (13:38 hrs local mean solar time ascending node) on July 15, 2004, the TES orbit repeats its ground track every 16 days (233 orbits), allowing global mapping of the vertical distribution of tropospheric ozone and carbon monoxide along with atmospheric temperature, water vapor, surface properties (nadir), and effective cloud properties (nadir). TES has a fixed array of 16 detectors, which in the nadir mode, have an individual footprint of approximately 5.3 x .5 km. In order to increase the signal-to-noise ratio, these detectors are averaged together to produce a combined footprint of 5.3x8.4 km. TES has two basic observational modes: the global survey mode, where observations are taken 1.3 degrees apart in latitude, and the "step-and-stare" mode, where the separation between observations is approximately 35 km along the orbit. This step-and-stare mode was used extensively throughout the TexAQS 2006 campaign.

Maps of these profiles can be found at http://tes.jpl.nasa.gov/TexAQS_2006/main_SS_TEXAQS_2006.html

Contact information: [email protected]

NASA CALIPSO observations http://www-calipso.larc.nasa.gov/ The Cloud-aerosol lidar and Infrared Pathfinder satellite Observations (CALIPSO) satellite mission was launched on April 28, 2006 for a planned 3-year mission. CALIPSO is flying in formation with Aqua, Aura, CloudSat, and Parasol satellites as part of the Afternoon Constellation or A-train. The CALIPSO payload consists of three instruments: the Cloud-Aerosol LIdar with Orthogonal Polarizaton (CALIOP); the Infrared Imaging Radiometer (IIR) and the Wide Field Camera (WFC). CALIPO is a nadir-pointing instrument which provides profile measurements of aerosol backscatter at 532 and 1064 nm and depolarization at 532 nm. The IIR provides calibrated radiances at 8.65 μm, 10.6 μm, and 12.05 μm over a 64 km swath centered on the lidar footprint and a pixel resolution of 1 km. The WFC consists of a single channel covering the 620 nm to 670 nm spectral region providing images of a 61 km swath with a spatial resolution of 125 m in a band 5 km about the nadir track and 1000 m elsewhere. For TexAQS 2006 campaign, CALIPSO quick turn-around browse images were produced to identify aerosol and cloud layers for flight planning activities. Observations were also synthesized into aerosol modeling systems to better understand the origin of surface and elevated aerosol features from regions outside the experiment domain.


31 August 2007 135

Examples of aerosol browse images used during TexAQS can be found at http://www-calipso.larc.nasa.gov/products/lidar/

Contact information: [email protected]

NOAA and DoD satellite images http://www.class.noaa.gov/nsaa/products/welcome;jsessionid=1C0E54F015C2813E5A9ACFC22C675F90 The National Oceanic and Atmospheric Administration (NOAA) Comprehensive Large Array-data Stewardship System (CLASS) is NOAA's premier on-line facility for the distribution of NOAA and US Department of Defense (DoD) Polar-orbiting Operational Environmental Satellite (POES) data and derived data products. CLASS is operated by the Information Processing Division (IPD) of the Office of Satellite Data Processing and Distribution (OSDPD), a branch of the National Environmental Satellite, Data and Information Service (NESDIS).

CLASS maintains an active partnership with NOAA's National Climatic Data Center (NCDC). NCDC, the permanent US Archive for POES data and derived data products, supports CLASS through a user-interactive Help Desk facility and through the provision of POES supporting documentation, including the NOAA Polar Orbiter Data (POD) User's Guide and the NOAA KLM User's Guide. Additionally, NCDC and CLASS share data distribution responsibilities for Defense Meteorological Satellite Program (DMSP) data under a Memorandum of Understanding with the National Aeronautics and Space Administration (NASA) for the Earth Observing System (EOS) Program.

CLASS provides data free of charge. Anyone can search the CLASS catalog and view search results through CLASS's World Wide Web (WWW) site. Users who wish to order data are required to register with their names and email addresses. CLASS distributes data to those users via FTP services.

CLASS (originally called Satellite Active Archive), was established as a demonstration prototype for electronic distribution of POES data in 1994, and became operational in July 1995. During that first month, 379 Advanced Very High Resolution Radiometer (AVHRR) Level 1b data sets were distributed to 27 customers via the emerging Internet. During the first five years of operation, the average monthly volume of data distribution increased to 65,000 data sets with a total size of 1.2 TB, and the SAA customer base grew to more than 10,000 registered customers. The active archive was expanded during that period to include TIROS Operational Vertical Sounder (TOVS) data, Defense Meteorological Satellite Program (DMSP) data, Radarsat Synthetic Aperture Radar (SAR) imagery, operational (near-term) satellite-derived products, and climatic (time-series) satellite-derived products.

NASA Earth Observatory natural hazards http://earthobservatory.nasa.gov/NaturalHazards/ This is a NASA site that has awesome satellite images due to the following natural phenomena: crops & drought; dust & smoke, fires, floods, severe storms, and volcanoes. The images are organized by event, and are free to all. They just ask for proper acknowledgment. This site is probably of limited value to modelers, but for certain events, such as the Saharan dust events that occurred during TexAQS II, the images may add some visual interest for a case study presentation.


31 August 2007 136

NASA MODIS Rapid Response System images http://rapidfire.sci.gsfc.nasa.gov/realtime/2006297/ This site has images from MODIS (Terra and Aqua). Images are archived by day, and can be downloaded. This site may be somewhat useful for modelers.

NASA Aqua AIRS retrieved CO profiles http://asl.umbc.edu/pub/mcmillan/www/index.html#calendar AIRS output for the TexAQS II field campaign. There is a clickable calendar for a view of the data. Please work with Dr. Wallace McMillan if interested in using the data (contact information is on the web site).

Solar Radiation Data

Texas Solar Radiation data, from a solar energy research group at UT http://www.me.utexas.edu/~solarlab/tsrdb/ This site has solar radiation data for 15 sites throughout Texas. The data intervals and times of coverage vary by station, ranging from 15 minute data to monthly averages. Data stops in 2003 or earlier for many of the stations. Data reported: Global horizontal, direct normal, and diffuse horizontal (W m-2). Monthly averages include temperature (degrees C). Data are easy to access. This site may be moderately useful for modelers.

National Renewable Energy Lab (solar radiation data) http://rredc.nrel.gov/solar/new_data/confrrm/ Cooperative Networks for Renewable Resource Measurements (CONFRRM). This network was designed to capture long-term solar radiation and wind measurements. There are 5 sites in Texas, however the last month showing data for all sites is March 2000. Therefore, data on this site are not useful for modelers working on summers 2000 – 2006.


31 August 2007 137

Large, Multi-field Data Sets

MADIS http://madis.noaa.gov/ The Meteorological Assimilation Data Ingest System (MADIS) is dedicated toward making value-added data available from the National Oceanic and Atmospheric Administration's (NOAA) Earth System Research Laboratory (ESRL) Global Systems Division (GSD) (formerly the Forecast Systems Laboratory (FSL)) for the purpose of improving weather forecasting, by providing support for data assimilation, numerical weather prediction, and other hydro-meteorological applications.

MADIS subscribers have access to an integrated, reliable and easy-to-use database containing the real-time and archived observational datasets described below. Also available are real-time gridded surface analyses that assimilate all of the MADIS surface datasets (including the very dense integrated mesonet data). The grids are produced by the Rapid Update Cycle (RUC) Surface Assimilation System (RSAS) that runs at ESRL/GSD, which incorporates a 15-km grid stretching from Alaska in the north to Central America in the south, and also covers significant oceanic areas. RSAS grids are valid at the top of each hour, and are updated every 15 minutes.

• Observations o Meteorological Surface

METAR SAO Maritime Modernized NWS Cooperative Observer Integrated Mesonet

Observations from local, state, and federal agencies and private mesonets (including GPSMET water vapor)

o Radiosonde o NOAA Profiler Network o Hydrological Surface o Automated Aircraft

Automated Aircraft Reports Profiles at Airports

o Multi-Agency Profiler o Radiometer o Satellite Wind

GOES Operational 3-Hour GOES Experimental 1-Hour

o Satellite Sounding NOAA POES

o Satellite Radiance NOAA POES

o Snow • Grids

o RSAS Surface Analyses

TCEQ Air Pollution Events http://www.tceq.state.tx.us/compliance/monitoring/air/monops/sigevents06.html The TCEQ Air Pollution Events web pages provide preliminary analyses of large-scale high ozone and/or particulate events in Texas. The analyses include satellite imagery, webcam imagery, ozone contour animations, ozone plume animations, backward air trajectories, upper air


31 August 2007 138

data graphs, and pollution data time series graphs. The discussions describe the intensity and geographic coverage of each event. The discussions also report any transport related aspects to the pollution, if appropriate, and provide an estimate of background levels and local add-on for ozone cases.

EDAS (NCEP grid reanalysis) http://www.cdc.noaa.gov/cdc/reanalysis/reanalysis.shtml It is a website for the NCEP/NCAR Reanalysis Project at the NOAA/ESRL Physical Sciences Division

This page points you to information on the NCEP/NCAR Reanalysis project and the implementation of a netCDF-based, internet-accessible, data service at NOAA/ESRL PSD for this set of data products. * The 6-hourly and daily data currently available on-line. * The monthly and other derived data currently available on-line.

This site also has links to other reanalysis project sites (e.g., ECMWF).

NCEP/NCAR Reanalysis http://dss.ucar.edu/pub/reanalysis/ This site includes the following NCEP/NCAR REANALYSIS databases.

* DSS Reanalysis archives. Project Overview * Project Description - The project motivation and objectives, cooperative arrangement between NCEP and NCAR, and other published documentation are outlined * Model Description * Project Status * Other Related Sites Data Product Description More than 20 different data products are output from the Reanalysis data assimilation, model run, and model forecast. These products are defined in terms of the NCAR archive names, physical variables, resolutions (temporal and spatial), and media storage size. CDROMS are also used to distribute selected reanalysis products.

This web site includes much detail on all of the data bases used, etc.

* 2006AUG10 --All 1948-2006JUL pgb.f00 and grb2d files are now available on line for registered users. * 2006AUG10 --JUL 2006 data files are released. All 1948-2006JUL reanalysis files are available. * 2006APR20 --Public (non-restricted) version of 200309-200602 prepqm files are released. * 2006APR11 --2005 annual cdrom is released. All 1950-2005 reanalysis annual cdroms are available. * 2006MAR28 --2005OCT-2006FEB reanalysis forecasts are released.


31 August 2007 139

* 2005Apr20 --2004OCT-2004DEC reruns to fix sea-ice problems are released. * 2005Apr19 --2004AUG and 2004SEP reruns to fix sea-ice problems are released. * 2005Apr08 --There will be a rerun from 2004080100 to 2005032212 due to sea-ice data problem. The 200501 and 200502 results are in. * 2003Aug04 --NCEP-NCAR Reanalysis Temperature Change Plots, 1948-2002 * DSS Reanalysis archives. Project Overview * Project Description - The project motivation and objectives, cooperative arrangement between NCEP and NCAR, and other published documentation are outlined * Model Description * Project Status * Other Related Sites Data Product Description More than 20 different data products are output from the Reanalysis data assimilation, model run, and model forecast. These products are defined in terms of the NCAR archive names, physical variables, resolutions (temporal and spatial), and media storage size. CDROMS are also used to distribute selected reanalysis products. This web site includes much detail on all of the data bases used, etc.


31 August 2007 140

APPENDIX 3. INSTITUTIONAL AFFILIATIONS AND E-MAIL ADDRESSES FOR MEMBERS OF THE RAPID SCIENCE SYNTHESIS WORKING GROUPS AND OTHER COOPERATORS

Dave Allen, University of Texas at Austin [email protected] Wayne Angevine, NOAA Earth System Research Laboratory [email protected] Elliot Atlas, University of Miami [email protected] Bob Banta, NOAA Earth System Research Laboratory [email protected] Tim Bates, NOAA Pacific Marine Environmental Laboratory [email protected] Carl Berkowitz, DOE Pacific Northwest Laboratory [email protected] Veronique Bouchet, Meteorological Service of Canada [email protected] Kevin Bowman, NASA Jet Propulsion Laboratory, Caltech [email protected] Pete Breitenbach, Texas Commission on Environmental Quality [email protected] Chuck Brock, NOAA Earth System Research Laboratory [email protected] Steve Brown, NOAA Earth System Research Laboratory [email protected] Daewon Byun, University of Houston [email protected] Greg Carmichael, University of Iowa [email protected] Bill Carter, University of California, Riverside [email protected] Ellis Cowling, North Carolina State University [email protected] Lisa Darby, NOAA Earth System Research Laboratory [email protected] Joost de Gouw, NOAA Earth System Research Laboratory [email protected] Basil Dimitriades, North Carolina State University [email protected] Bright Dornblaser, Texas Commission on Environmental Quality [email protected] Mark Estes, Texas Commission on Environmental Quality [email protected] Alan Fried, National Center for Atmospheric Research [email protected] Maedeh Faraji, University of Texas at Austin [email protected] Cari Furiness, North Carolina State University [email protected] Noor Gillani, University of Alabama in Huntsville [email protected] Wanmin Gong, Meteorological Service of Canada [email protected] Georg Grell, NOAA Forecast Systems Laboratory [email protected] Mike Hardesty, NOAA Earth System Research Laboratory [email protected] Scott Herndon, Aerodyne Inc. [email protected] John Holloway, NOAA Earth System Research Laboratory [email protected] Harvey Jeffries, University of NC at Chapel Hill [email protected] John Jolly, Texas Commission on Environmental Quality [email protected] Dick Karp, Texas Commission on Environmental Quality [email protected] Siwan Kim, NOAA Earth System Research Laboratory [email protected] Susan Kemball-Cook, Environ Corp [email protected] Bryan Lambeth, Texas Commission on Environmental Quality [email protected] Barry Lefer, University of Houston [email protected] Deborah Luecken, US Environmental Protection Agency [email protected] John McHenry, Baron Advanced Meteorological Systems [email protected] Stuart McKeen, NOAA Earth System Research Laboratory [email protected] Wallace McMillan, University of Maryland, Baltimore County [email protected] Jeffery McQueen, NOAA National Weather Service [email protected] Johan Mellqvist, University of Göteborg, Sweden [email protected]


31 August 2007 141

Ann Middlebrook, NOAA Earth System Research Laboratory [email protected] Gary Morris, Valparaiso University [email protected] John Nielsen-Gammon, Texas A&M University [email protected] Jonathan Andy Neuman, NOAA Earth System Research Laboratory [email protected] John Nowack, NOAA Earth System Research Laboratory [email protected] Hans Osthoff, NOAA Earth System Research Laboratory [email protected] David Parrish, NOAA Earth System Research Laboratory [email protected] Ryan Perna, University of Houston [email protected] Brad Pierce, NOAA Satellite and Information Service [email protected] Particia Quinn, NOAA Pacific Marine Environmental Laboratory [email protected] Bernhard Rappenglück, University of Houston [email protected] James Roberts, NOAA Earth System Research Laboratory [email protected] Ted Russell, Georgia Institute of Technology [email protected] Tom Ryerson, NOAA Earth System Research Laboratory [email protected] Ken Schere, US Environment Protection Agency [email protected] Christoph Senff, NOAA Earth System Research Laboratory [email protected] David Sullivan, University of Texas at Austin [email protected] Michael Trainer, NOAA Earth System Research Laboratory [email protected] Youhua Tang, University of Iowa youhua.tang@noaa .gov Sara Tucker, NOAA Earth System Research Laboratory [email protected] William Vizuete, University of North Carolina at Chapel Hill [email protected] Carsten Warneke, NOAA Earth System Laboratory [email protected] Allen White, NOAA Earth System Research Laboratory [email protected] Jim Wilczak, NOAA Earth System Research Laboratory [email protected] Eric Williams, NOAA Earth System Research Laboratory [email protected] David Winker, NASA Langley Research Center [email protected] Yulong Xie, DOE Pacific Northwest Laboratory [email protected] Greg Yarwood, Environ Corp [email protected]


31 August 2007 142

APPENDIX 4. GUIDELINES FOR FORMULATION OF SCIENTIFIC FINDINGS TO BE USED FOR POLICY PURPOSES

The following guidelines in the form of checklist questions were developed by the NAPAP Oversight Review Board to assist scientists in formulating presentations of research results to be used in policy decision processes. 1) IS THE STATEMENT SOUND? Have the central issues been clearly identified? Does

each statement contain the distilled essence of present scientific and technical understanding of the phenomenon or process to which it applies? Is the statement consistent with all relevant evidence – evidence developed either through NAPAP [or TexAQS] research or through analysis of research conducted outside of NAPAP [or TexAQS II]? Is the statement contradicted by any important evidence developed through research inside or outside of NAPAP [or TexAQS II] Have apparent contradictions or interpretations of available evidence been considered in formulating the statement of principal findings?

2) IS THE STATEMENT DIRECTIONAL AND, WHERE APPROPRIATE, QUANTITATIVE? Does the statement correctly quantify both the direction and magnitude of trends and relationships in the phenomenon or process to which the statement is relevant? When possible, is a range of uncertainty given for each quantitative result? Have various sources of uncertainty been identified and quantified, for example, does the statement include or acknowledge errors in actual measurements, standard errors of estimate, possible biases in the availability of data, extrapolation of results beyond the mathematical, geographical, or temporal relevancy of available information, etc. In short, are there numbers in the statement? Are the numbers correct? Are the numbers relevant to the general meaning of the statement?

3) IS THE DEGREE OF CERTAINTY OR UNCERTAINTY OF THE STATEMENT INDICATED CLEARLY? Have appropriate statistical tests been applied to the data used in drawing the conclusion set forth in the statement? If the statement is based on a mathematical or novel conceptual model, has the model or concept been validated? Does the statement describe the model or concept on which it is based and the degree of validity of that model or concept?

4) IS THE STATEMENT CORRECT WITHOUT QUALIFICATION? Are there limitations of time, space, or other special circumstances in which the statement is true? If the statement is true only in some circumstances, are these limitations described adequately and briefly?

5) IS THE STATEMENT CLEAR AND UNAMBIGUOUS? Are the words and phrases used in the statement understandable by the decision makers of our society? Is the statement free of specialized jargon? Will too many people misunderstand its meaning?

6) IS THE STATEMENT AS CONCISE AS IT CAN BE MADE WITHOUT RISK OF MISUNDERSTANDING? Are there any excess words, phrases, or ideas in the statement which are not necessary to communicate the meaning of the statement? Are there so many caveats in the statement that the statement itself is trivial, confusing, or ambiguous?

7) IS THE STATEMENT FREE OF SCIENTIFIC OR OTHER BIASES OR IMPLICATIONS OF SOCIETAL VALUE JUDGMENTS? Is the statement free of influence by specific schools of scientific thought? Is the statement also free of words, phrases, or concepts that have political, economic, ideological, religious, moral, or other personal-, agency-, or organization-specific values, overtones, or implications? Does the choice of how the statement is expressed rather than its specific words suggest underlying biases or value judgments? Is the tone impartial and free of special pleading? If societal value judgments have been discussed, have these judgments been identified as such and described both clearly and objectively?


31 August 2007 143

8) HAVE SOCIETAL IMPLICATIONS BEEN DESCRIBED OBJECTIVELY? Consideration of alternative courses of action and their consequences inherently involves judgments of their feasibility and the importance of effects. For this reason, it is important to ask if a reasonable range of alternative policies or courses of action have been evaluated? Have societal implications of alternative courses of action been stated in the following general form?:

"If this [particular option] were adopted then that [particular outcome] would be expected." 9) HAVE THE PROFESSIONAL BIASES OF AUTHORS AND REVIEWERS BEEN

DESCRIBED OPENLY? Acknowledgment of potential sources of bias is important so that readers can judge for themselves the credibility of reports and assessments.

Reference: QBN1701-TCEQ-2007 Page 1 of 86

Date of Issue: 28 November 2007 Signed:

Checked by: Name: Melanie Williams for Managing Director

Commercial in confidence Measurements of VOC emissions from petrochemical industry sites in the

Houston area using Differential Absorption Lidar (DIAL) during Summer 2007.

DRAFT FOR COMMENT

FOR: Texas Commission

on Environmental Quality

FOR THE ATTENTION OF: Russell Nettles DATE OF TEST PERIOD: 16/07/07 to 18/08/07 AUTHORS: Rod Robinson, Tom Gardiner,

Bob Lipscombe

Commercial in confidence

Reference QBN1701-TCEQ-2007 Page 2 of 85

Checked by:

Contents

Contents ................................................................................................................................... 2 Introduction.............................................................................................................................. 3 Section 1. Measurements of VOC emissions from a Bulk Terminal, in the Houston non-attainment area. ........................................................................................................................ 4

Measurement description. ................................................................................................. 4 Definition of VOCs............................................................................................................. 4 Measurements on the 16th July. ........................................................................................ 6 Measurements on the 17th July ......................................................................................... 8 Measurements on the 18th July and 19th July ................................................................ 10 Night-time measurements on the 20th-21st July. ............................................................ 11

Section 2 Measurements of VOC emissions from a Refinery in the Houston non-attainment area. ........................................................................................................................................ 12

Measurement description. ............................................................................................... 12 Measurements of product storage tanks........................................................................ 14 Measurements of crude oil tanks.................................................................................... 16 Measurements of VOC emissions from coker C ........................................................... 24 Measurements of VOC emissions from heated oil tanks.............................................. 27 Measurement of VOC emissions from elevated flares.................................................. 28

Section 3 Measurements of Benzene emissions from Refinery........................................... 31 Measurement description ................................................................................................ 31 Benzene concentration measurements on 17th August. ................................................ 31 Measurements of benzene fluxes from the coker .......................................................... 36

Section 4. Measurements of SO2 from the Sulphur Recovery Unit...................................... 38 Annex 1. Description of the DIAL technique. ....................................................................... 39

Overview of the Dial technique....................................................................................... 39 Description of the theory of Dial measurements ........................................................... 39 Description of facility operated by NPL ........................................................................ 40 Calibration and validation .............................................................................................. 45 NPL open-path calibration facility................................................................................. 47

Annex 2 Speciated VOC measurements and Flux Correction Factors ................................. 48 Flux Correction Factors .................................................................................................. 48 Sampling Procedure......................................................................................................... 53 Analysis of the ATD tubes ............................................................................................... 54

Annex 3. Meteorological Measurements .............................................................................. 65


Checked by:

Introduction This report presents the results of a measurement campaign carried out using the NPL differential absorption LIDAR (DIAL) system at two petrochemical facilities in the The Houston non-attainment area region of South East Texas, between 16th July to 18th August 2007. The project was carried out for the Texas Commission on Environmental Quality (TCEQ). This report covers the measurements undertaken by NPL, and does not cover additional measurements undertaken by other contractors. The primary objective of the study was to assess emission fluxes of VOCs and benzene from a number of identified potential sources, including storage tanks, a delayed coker unit, waste water treatment areas and flares. The report provides a summary of the VOC fluxes measured using the DIAL technique at a Bulk Terminal in the Houston non-attainment area (hereafter the Bulk Terminal) (Section 1) and a Refinery in the Houston non-attainment area (hereafter the Refinery) (Sections 2 and 3). Annex 1 provides an overview of the DIAL technique, and discusses the calibration and validation procedures. Annex 2 presents the results of speciation measurements of air samples, and Annex 3 presents a summary of the meteorological measurements undertaken during the campaign.


Checked by:

Section 1. Measurements of VOC emissions from a Bulk Terminal, in the Houston non-attainment area.

Measurement description. DIAL measurements of the emissions of VOCs from storage tanks at the Bulk Terminal facility were carried out on 16th, 17th , 18th, 19th July and on the night of the 20th-21st July. The tanks selected by TCEQ/Bulk Terminal for measurement were naphtha storage tanks, numbers 22,23, 27, 28, and 29. Tank 3792 was also included in some of the measurements, though no specific identifiable emissions were observed that could be attributed to this tank. Two measurement locations were used for the DIAL measurements; these were both on Avenue H, the southern road within the plant. The GPS coordinates for these measurement locations were: DIAL Location 1. North side of Ave H, due south of Tank 3770 29°21'23.64"N 94°55'2.82"W DIAL Location 2 South side of Ave H, south of Tank 3792. 29°21'23.47"N 94°54'52.79"W Measurements of VOC fluxes were carried out using the NPL DIAL as described in Annex 1. Fluxes were obtained by scanning the DIAL measurement line-of-sight in vertical planes, downwind of the target emissions sources. Measurements were also made upwind of the target sources to check for any upwind VOC fluxes. The measurement lines are shown in Figure 1.1.

Definition of VOCs The DIAL technique, as operated during these measurements, provides a measurement of mass emission in gasoline vapour equivalent. The DIAL measures VOCs by measuring the differential absorption of two wavelengths of light sensitive to hydrocarbons of C3 and above. This has been calculated for each air sample taken, and Annex 2 presents a set of correction factors for each measurement location. For the measurements at the Bulk Terminal the factors range from between 0.8 to 1.6.


Checked by:

Figure 1.1.DIAL locations and scan lines used for VOC measurements at the Bulk Terminal.


Checked by:

Measurements on the 16th July. From Location 1, scans were made which measured the emissions from tanks 23,27,28,29 and 3792. Tank 22 emissions would also have been observed if these had been significant. Table 1.1 presents these results. The wind direction was generally from the SE, with speeds of 3 to 6 m/s. A small flux was measured, at a range consistent with emissions from tanks 27, 28 and 23. This is shown in Figure 1.2. No observable emissions were seen from tank 22 or tank 3792. The average emission observed was 28 lbs/hr with a standard deviation of ± 5 lbs/hr. It should be noted the standard deviation of the results includes both the inherent uncertainty in the DIAL measurements and any real variability in the emissions (whether due to source variability or variable dispersion). Upwind scans (Scan IDs 16 and 21) gave results at or below the detection limit, which for the conditions prevailing at the Bulk Terminal was 1 lbs/hr.

Table 1.1. VOC fluxes measured on 16th July at Bulk Terminal. The DIAL provides a range resolved measurement of VOC concentrations. A DIAL scan can provide not only a measurement of the mass emissions flux, but also a visual indication of the location of the emission plume. Figure 1.2 shows an example of such a visualisation, for Scan ID 12 on the 16th July. The plot illustrates the vertical plane through which the DIAL scanned, and indicates the distribution of VOCs within this plume. The wind direction is indicated by the red arrow. The intensity of the colour plot is related to the concentration of the VOCs. The maximum VOC concentration in this plot is 300 ppb.


Checked by:

Figure 1.2. Example of the VOC distribution seen in a vertical scan at the Bulk Terminal (Scan 12, 16th July). The green arrow represents the average wind direction. The peak VOC concentration observed in this scan was 300 ppb.

DIAL


Checked by:

Measurements on the 17th July A further series of measurements were made of tanks 27,28,29 and 23 from Location 1. The wind was from the south, shifting slightly to south-westerly during the day with wind speeds ranging between 2.8 and 8.7 m/s. The results are presented in Table 1.2. The average flux observed for these tanks on the 17th July was 21 lbs/hr with a standard deviation of ± 5 lbs/hr. Upwind scans were below the detection limit, indicating no contribution from other sources. In order to separate out and directly measure fluxes from tank 22, the DIAL was moved to Location 2, to the south of tank 3792. Scans could then be made isolating the emissions from tank 22. It should be noted that this measurement configuration is not ideal for flux measurements, as the measurement line of sight is very close to tank 22. Other locations were not possible due to construction work on the road to the north of tank 22. Scans made from location 1 had indicated that the emissions observed from tank 22 were not significant. This was apparent because no significant concentrations were observed at ranges which would have been consistent with a plume from tank 22. Moving to a measurement line of sight further away from tank 22, to the north of Ave G, would have reduced the chance of measuring any small emissions from tank 22 due to dispersion. The fluxes observed, even close to tank 22, were very small, with an average flux of 3 lbs/hr and a standard deviation of 1.3 lbs/hr. The peak concentrations observed were less than 100 ppb, and these emissions would generally be considered insignificant in the context of a DIAL survey.

Table 1.2. VOC fluxes measured on 17th July at the Bulk Terminal.


Checked by:

Figure 1.3. Example of a horizontal scan showing the VOC distribution above naphtha tanks at the Bulk Terminal (Scan 17, 17th July). The green arrow represents the average wind direction. The peak VOC concentration observed in this scan was 288 ppb.


Checked by:

Measurements on the 18th July and 19th July Further scans were made of tank 22 on both the 18th and 19th July, tables 1.3 and 1.4 present these data. The average emissions fluxes measured were 5 lbs/hr with a standard deviation of 2.4 lbs/hr on the 18th and 3 lbs/hr with a standard deviation of 2.4 lbs/hr on the 19th. Two scans were made on the 18th July to measure the emissions from tanks 23 and 28 from location 2. These scans may also have included the contribution from tanks 29 and 22. However, the emissions from these tanks would be quite dispersed and are probably not included in the measured fluxes. The scans are not ideal for flux determination, as the wind field would be very disturbed by the proximity of the tanks. The scans gave flux results that are much lower than those measured on the previous days for these tanks, with an average value of 3.5 lbs/hr. This indicates either, that the larger emissions originated from tank 27 (included in earlier measurements), and are not included in these scans. A more likely explanation, given subsequent measurements on the 19th, is that the wind field was not ideal and the scans did not capture the emissions from the tanks. On the 19th July similar scans measured an average emission flux of 18 lbs/hr, which is consistent with the measurements made on the 16th and 17th July for these tanks.

Table 1.3. VOC fluxes measured on 18th July at the Bulk Terminal

Table 1.4. VOC fluxes measured on 19th July at the Bulk Terminal


Checked by:

Night-time measurements on the 20th-21st July. A series of measurements was made on the night of the 20th-21st July from Location 2, the results of these scans are given in Table 1.5. These repeated the scans made from this location during the daytime. The wind was initially from further to the east, which was not ideal for this location, however it subsequently shifted further south, providing useable measurements of tank 22 and tanks 23/28. The average emission flux measured from tank 22 was 1 lbs/hr, which is at the detection limit. The measurements of the emissions from tanks 23 and 28 gave an average emission flux of 14 lbs/hr, with a standard deviation of 7 lbs/hr. Although this emission is lower than that measured during the day. That, coupled with the variable wind pattern close to the tanks and the fact that the wind speed itself was lower during these measurements, suggests that conclusions about the relationship between day and night-time emissions cannot be drawn from these data, except to recognise that emissions are observed during the night-time, and therefore night time cannot be discounted from annualisation calculations based on day time measurements. A series of air samples, using canisters and pumped sorbent tubes was also taken, and the results of these are reported in Annex 2. Figure A1.1 indicates the locations of these samples at the Bulk Terminal.

Table 1.5. VOC fluxes measured on the night of 20th July at the Bulk Terminal


Checked by:

Section 2 Measurements of VOC emissions from a Refinery in the Houston non-attainment area.

Measurement description. This section of the report describes the results of measurements of VOC emissions from various sources within the Refinery. Measurements of VOC emissions were carried out from the 25th July to the 11th August. Table 2.1 lists the measurements days and the sources monitored on each day. The DIAL measurement locations are identified on individual maps in the following Sections. Date DIAL Location Sources monitored 25th July Refinery Location 1 Product tanks 26th July Refinery Location 1 Product tanks 28th July Refinery Location 2 Crude oil tanks, coker 29th July Refinery Location 2 Crude oil tanks 30th July Refinery Location 3 Gasoline tanks and flare 6 31st July Refinery Location 2, 2C Crude oil tanks and coker 1st August Refinery Location 2 Coker 2nd August Refinery Location 4 Waste water treatment and

crude oil tanks 3rd August Refinery Location 2 Coker 5th –6th August (Night) Refinery Location 2 Crude oil tanks 6th – 7th August (Night) Refinery Location 3 Gasoline tanks, coker and

flare 6 7th – 8th August (Night) Refinery Location 5,6,7 Product tanks

Hot oil tanks 9th August Refinery Location 8 Flares 10th August Refinery Location 8 Flares 11th August Refinery Location 8 Flares Table 2.1 Summary of DIAL VOC flux measurements at the Refinery The GPS positions for the DIAL locations were: Refinery Location 1 By Muster Point 8 29°22'1.48"N 94°56'23.11"W Refinery Location 2 E 4th street, next to tank 1020 29°22'4.56"N 94°55'13.56"W


Checked by:

Refinery Location 2C Further north on E 4th street 29°22'5.58"N 94°55'14.09"W Refinery Location 3 Avenue J, North of tank 1045 29°22'10.26"N 94°55'29.40"W Refinery Location 4 Avenue N, south of Surge Basin 2 29°21'52.38"N 94°55'17.22"W Refinery Location 5 Outside Gate 16 29°22'3.78"N 94°56'27.42"W Refinery Location 6 On W 4th St, by tank 17 29°22'21.36"N 94°56'10.98"W Refinery Location 7 Avenue G, south of ARU 29°22'23.52"N 94°56'15.36"W Refinery Location 8 Grant Avenue, South of tank 20. 29°22'35.45"N 94°56'29.47"W The measurement results are presented in the following sections, grouped by the sources measured. It is important to note that the potential emission sources measured were identified by TCEQ and the refinery, and measurements were focussed on these sources. This study did not include a full survey of all sources within the refinery, and was not aimed at determining total refinery emissions.


Checked by:

Measurements of product storage tanks. A small number of distillate product storage tanks were identified as being of interest. These included tanks 55 and 53. Scans were made of these tanks on the 25th 26th July and during the night of 7th August. These measurements were made from Location 1 on the 25th and 26th and from Location 5 on the 7th August. The results are presented in Table 2.2. The emissions from tank 55 were not significant, with average emission fluxes measured of 5 lbs/hr (not counting the negligible fluxes measured during very low and variable wind conditions). Upwind scans were in general below the detection limit, however one scan down Avenue K and upwind of the hydrogen plants No.1 and No.2 measured a small flux of 7 lbs/hr. Measurements of tank 55 during the evening of the 7th August showed similar emission fluxes, with an average of 6 lbs/hr. Tank 57 was too close to the DIAL to enable it to be measured on the 7th August. Measurements of tank 53 were made during the evening of the 7th August and an average flux of 24 lbs/hr with a standard deviation of 8 lbs/hr was measured. Upwind measurements were below detection limits.

Table 2.2 Measurements of VOC fluxes from product storage tanks


Checked by:

Figure 2.1. DIAL locations and scan lines used for DIAL measurements of VOC emission fluxes from the product storage tanks at the Refinery.


Checked by:

Measurements of crude oil tanks A series of measurements of the emissions from the crude oil storage tanks was made over several days. Measurements were made on the 28th 29th 31st July, 2nd August and on the night of the 6th August. The results of measurements of different tanks are summarised in Table 2.3, all flux results are presented in Table 2.4. Measurements were made from Location 2, apart from the measurements of tanks 1020 and 1021 on the 31st July, for which the DIAL was moved north to Location 2C. Tanks Average emission flux

Lbs/hr 1020 <2 1021 16 1024 5 1025 11 1052 22 28th July

39 2nd August 24 5th August, Night

1053 7 1055 <5 Table 2.3 Summary of VOC fluxes from crude storage tanks In order to assign emission fluxes to individual tanks, some interpretation of the flux results in Table 2.4 has had to be made. In some cases, flux measurements were made which were downwind of a number of tanks. Where these tanks were also measured separately on other occasions then the split of the flux between the possible sources has been made based on these individual measurements. For the measurements made on the 2nd August from Refinery Location 4, the measurements of tanks 1055 and 1053 probably include a component of the emissions observed from tank 1052. The flux is not significantly higher than that measured from tank 1052 alone, and so it is clear there is no significant emission flux from either tank 1055 or tank 1053. This is consistent with the previous measurements of these tanks, which did not include emissions upwind from tank 1052. Similarly measurements of tanks 1020 and 1021 on the 29th July may have included some emissions from tanks 1025 and 1024, however, the wind direction was such that this is unlikely. The wind direction, and location of the observed flux, also indicate that most of the observed emission flux is attributable to tank 1021, and so the average emissions have been assigned to tank 1021. This is supported by measurement of tank 1020 only, on the 31st July, which gave very low fluxes, many below 1 lbs/hr. No significant emissions were observed from tank 1055, which was in the near field during a number of scans. In summary, most of the tanks observed during these measurements did not have significant emission fluxes, and in general, the emission fluxes from individual tanks were < 10 lbs/hr.


Checked by:

A VOC emission was observed from tank 1052, with an average emission of 39 lbs/hr observed on the 2nd August.

Figure 2.2.DIAL locations and scan lines used for DIAL measurements of VOC emission fluxes from the crude oil tanks at the Refinery.


Checked by:

Table 2.4 Measurements of VOC fluxes from crude oil storage tanks


Checked by:

Table 2.4 (continued) Measurements of VOC fluxes from crude oil storage tanks


Checked by:

Measurements of gasoline tanks and flare 6. Measurements of gasoline storage tanks 501, 502, 503, and 504 were carried out from location 3 on the 30th July and on the night of the 6th August. The results are presented in Table 2.5. The wind directions for these measurements were not ideal for flux determination; however, any significant emissions would have been observed. Upwind measurements from location 3 included measurements of emissions from flare 6, which was operational during the measurement periods. As can be seen from the results presented in Table 2.5, there was no significant VOC flux observed from these tanks; the measurements are consistent with an emission of approximately 5 lbs/hr from individual tanks. Night time measurements found slightly lower emissions; however as these measurements were close to the detection limit, drawing any conclusion from these results about night versus day emissions is not recommended. Owing to the wind direction during the night time measurements, it is also likely these scans would not have fully captured emissions from tank 502. Measurements of emissions from flare 6 on the 30th July gave an average flux of 4 lbs/hr. During the night time measurements on the 6th August, an average flux of 22 lbs/hr was measured from flare 6. It should be noted that these emissions could act as an upwind source for the gasoline tank measurements. However, they were spatially separated from the emissions from the gasoline tanks and were not included in these measured fluxes. Therefore this upwind source did not have to be subtracted from the tank measurements, which can lead to a significant source of uncertainty and is something that is avoided if possible.

Table 2.5 Measurements of VOC fluxes from gasoline storage tanks


Checked by:

Figure 2.3. DIAL locations and scan lines used for DIAL measurements of VOC emission fluxes from the gasoline tanks at the Refinery.


Checked by:

Measurements of Water Treatment area. Measurements were made of the water treatment plant on the 2nd August. Measurements were also made of crude tanks on this day and these have been reported in the relevant Section above. The emissions fluxes observed from the waste treatment areas are given in Table 2.6. The average flux seen from the waste water treatment area, and specifically from the west side of the secondary and tertiary effluent treatment facilities, was 30 lbs/hr. Scans upwind of the water treatment plant measured less than 1 lbs/hr. Two scans were also made of the flux from the API separator located at the NE corner of Surge Basin 2, and these measured a small average emission flux of 7 lbs/hr.

Table 2.6 Measurements of VOC fluxes water treatment areas

Figure 2.4. DIAL locations and scan lines used for DIAL measurements of VOC emission fluxes from the waste water treatment areas at the Refinery.


Checked by:

Figure 2.5. Example of the VOC distribution seen in a vertical scan at the Refinery (Scan 285). This scan measured the VOC emissions from tank 1052. The green arrow represents the average wind direction. The peak VOC concentration observed in this scan was 495 ppb.

DIAL

Tank 1052


Checked by:

Measurements of VOC emissions from coker C Measurements were made of the VOC emissions from the operating coker, unit C, on the 28th July, 31st July, 1st 3rd August. It had been suggested that refinery coker units could be a significant source of emissions, and figures of several hundred lbs/hr were discussed. One of the primary aims of these measurements was therefore to determine if the emissions from this coker were of a similar level. Over a number of measurement days, with different conditions, and through several cycles of the coker, the highest emission flux observed during a DIAL scan from the coker was 44 lbs/hr The average emission fluxes observed from the coker are given in Table 2.7, and all individual flux results are presented in Table 2.8. Date Average emission flux from

coker Average upwind flux

28th July 10 lbs/hr 2 lbs/hr 31st July 31 lbs/hr 3 lbs/hr 1st August 12 lbs/hr 4 lbs/hr 3rd August 32 lbs/hr 8 lbs/hr 6th August - Night 4 lbs/hr N/A Table 2.7, Summary of VOC emissions fluxes measured from Coker ‘C’ Measurements upwind of the cokers identified some small fluxes intermittently on the 31st July and 3rd August. These presumably arise from process units close to the cokers, including potentially, the coker vapour recovery unit. These upwind concentrations have not been subtracted from the observed coker results, however, their potential impact on the uncertainty of the results should be considered. During the coker measurements on the 3rd August a small, localised emission was observed south of tank 1010. This was consistent with emissions from waste water vent pipes, located at the north end of E-4th street. The average emission flux observed from this source was 9 lbs/hr.


Checked by:

Figure 2.6.DIAL locations and scan lines used for DIAL measurements of VOC emission fluxes from coker C at the Refinery.

Figure 2.7. Example of the VOC distribution seen in a vertical scan at the Refinery (Scan 263), downwind of coker C. The green arrow represents the average wind direction. The peak VOC concentration observed in this scan was 234 ppb.


Checked by:

Table 2.8. Measurements of VOC emission fluxes from coker ‘c’.


Checked by:

Measurements of VOC emissions from heated oil tanks A small number of heated tanks containing fuel oil were measured. These tanks were monitored on the night of the 7th August. The emissions from two tanks 43, and 60 were measured, and the results are given in Table 2.9. The average emission flux from Tank 43 was 6 lbs/hr and from Tank 60 the average emission flux was 9 lbs/hr. Upwind fluxes were less than 1 lbs/hr. These results show the emissions from these tanks are very low and they are not significant sources of VOC emissions.

Table 2.9. Measurements of VOC emission fluxes from hot oil tanks.

Figure 2.8. DIAL locations and scan lines used for DIAL measurements of VOC emission fluxes from the hot oil tanks at the Refinery.


Checked by:

Measurement of VOC emissions from elevated flares. Measurements of a temporary flare stack located in the NW corner of the refinery were undertaken. This flare had a very visible flame, and the aim was to determine any emission due to incomplete combustion in the flare. The DIAL was located in Grant Avenue and measured downwind and upwind of the flare to identify any plume of un-burnt hydrocarbons. These measurements identified very low levels of hydrocarbon downwind of the flare, and the VOC emission flux measured from this flare was ~ 6 lbs/hr. This is not surprising given that the flare was reported to be mainly burning hydrogen. However, during the measurement of the temporary flare, a high concentration VOC plume was observed at a closer range than the temporary flare stack. This plume was consistent with emissions from another elevated flare, the ULC flare. This flare had an almost invisible flame during daylight, but was observed to be lit at night. Table 2.11 lists all the measured fluxes from these flares, and Table 2.10 provides a summary of the emissions from the ULC flare. Date Average VOC emission flux 9th August 147 lbs/hr 10th August 167 lbs/hr 11th August 263 lbs/hr Table 2.10 Summary of measurements of VOC emissions from ULC flare. The measurements of the ULC plume showed a fairly high degree of variability, and this will be due to a combination of variability in the flare emissions and the possibly to movement of the narrow plume during the period of each DIAL scan.


Checked by:

Figure 2.9. DIAL locations and scan lines used for DIAL measurements of VOC emission fluxes from flare stacks at the Refinery

Figure 2.10 Example of the VOC distribution seen in a vertical scan at the Refinery (Scan 432) downwind of the ULC flare. The green arrow represents the average wind direction. The peak VOC concentration observed in this scan was 3.86 ppm.


Checked by:

Table 2.11. Measurements of VOC flux from ULC and temporary flares.


Checked by:

Section 3 Measurements of Benzene emissions from Refinery

Measurement description Measurements of benzene were carried out on the 17th and 18th August 2007 at the Refinery. Measurements were made of the Aromatic Recovery Unit (ARU) on the 17th and the operational Coker (Coker Unit C) on the 18th. Measurements on the 17th were carried out in parallel with a UV Open Path DOAS spectrometer, operated by the EPA and were made for comparison purposes. Measurements on the 18th were carried out to obtain emissions flux measurements from the coker operations, and were timed to include the period during which the unit was cutting coke.

Benzene concentration measurements on 17th August. The measurements on the 17th August were carried out to compare the concentrations measured by DIAL with those obtained by the open-path UV-DOAS system operated by Cary Secrest from EPA. Both systems were set up to measure a 200m path close to the ground, downwind of the ARU. Figure 1 shows the DIAL and DOAS locations and also the locations of 2 air samples, which were taken during the tests. The DIAL was located in Avenue C, South of Tank 5, GPS location 29°22'35.28"N, 94°56'2.10"W. The wind was from directions between 130° to 170° from N, at speeds of between 6 mph to 16 mph.

Figure 3.1, Location for measurements of benzene concentrations from the ARU During these measurements the DIAL beam was kept at a fixed line of sight, in a similar configuration to the UV-DOAS, and not scanned to obtain concentration profiles. Therefore, benzene fluxes were not determined. The two measurement-paths for the DIAL and the UV-


Checked by:

DOAS were as close as was logistically possible. The DIAL data were processed to assess the total integrated concentration over the assigned path. The path length of the DOAS was stated by the EPA operators to be from 247m to 440m from the DIAL. The ARU unit was measured from the site map to be 100m long from 280 to 380 m from the DIAL. The DIAL and UV-DOAS results are plotted for comparison in Figures 2 and 3. The DIAL results are tabulated in Tables 1 and 2. The results plotted for the UV-DOAS are as supplied by Cary Secrest (US EPA). There is no difference between the UV-DOAS results in figures 2 and 3, only the DIAL pathlength has been varied, the UV-DOAS pathlength is a fixed parameter dependant on the instrument configuration. The DIAL results shown in the tables are the averages of a number of measurements taken during the periods indicated, and the standard deviations of the concentrations is also given. Where no standard deviation is given this is because the result is for a single measurement.

Table 3.1, DIAL results – ppb benzene determined over a 193m path comparable to the UV-DOAS.


Checked by:

Table 3.2, DIAL results – ppb benzene determined over a 100m path comparable to the extent of the ARU. Air samples were taken in canisters and with pumped sorbent tubes, at the locations shown in Figure 1. Table 3 shows the results of the analyses of these canisters for benzene. The samples were taken between 17:29 and 18:29 for the EPA line of sight and 17:53 and 18:53 for the DIAL line of sight.

Table 3.3, Air sample analyses for ARU measurements. The concentrations of benzene measured using the pumped sorbent tubes are similar to the levels measured with both the UV-DOAS and the DIAL. The concentrations measured by the DIAL are close to the limit of detection for the DIAL system and the standard deviations of the results reflects this.

Benzene Toluene Ethylbenzene m/p-Xylene o-Xylene Tubes Canisters Tubes Canisters Tubes Canisters Tubes Canisters Tubes CanistersARU ppb ppb Ppb ppb ppb Ppb ppb ppb ppb ppb DIAL LOS 12.87 <2.00 18.79 12.26 2.33 2.04 9.36 5.51 3.05 <2.00 DOAS LOS 6.71 <1.88 9.79 20.52 1.90 2.46 5.29 8.49 1.44 2.59


Checked by:

Figure 3.2, Comparison of DIAL and UV-DOAS benzene concentrations. DIAL integrated path corresponding to reported UV-DOAS 193m pathlength.


Checked by:

Figure 3.3, Comparison of DIAL and UV-DOAS benzene concentrations. DIAL results for 100m path corresponding to the location of the ARU unit.


Checked by:

Measurements of benzene fluxes from the coker Measurements were made of the benzene concentrations downwind of coker, unit C, during the operation of this coker on the 18th August. The measurements were carried out from 17:20 to 20:45. The coker was cutting coke between 19:50 and 20:38. The configuration of the measurements and location of the DIAL during these measurements is shown in Figure 4


Checked by:

Figure 3.4. Location of DIAL for measurements of benzene fluxes from coker Unit C. The DIAL was located on Avenue J, to the west of Tank 1010, at GPS coordinates 29°22'10.98"N, 94°55'19.50"W. The wind during the measurements was from directions between 125° to 143° from N, with speeds between 6 mph to 9 mph. Vertical scans were made downwind of the coker; scans were also made upwind of the coker to check for any other sources of benzene. The results are given in Table 4. During most of the period of measurements the DIAL values are at or below the detection limits. For these measurements, the detection limits, in terms of lbs per hr flux, were of the order of 0.5 to 1 lbs per hr. During the period before the coker commenced cutting, the fluxes measured were often below the detection limit and have been reported as 0 lbs per hr. In some cases, a figure of less than a given flux is reported. In these cases, a detectable concentration was recorded, but the uncertainty in the results is such that only an upper limit is reported. Detected concentrations of benzene were in general around 0.1 ppm. During the coker cut, a small but observable flux of benzene was measured, the two results giving fluxes of 1.5 lbs per hr and 2.1 lbs per hr. Upwind scans showed no significant source of benzene. The results of the analyses of the air sample taken downwind of the coker are shown in Table 5. This sample was taken during the period of the coker cut, and a low concentration of benzene ~ 1 to 2 ppb was measured.

Table 3.4. Measured fluxes during coker operation.

Table 3.5, results of air sample analyses downwind of coker.

Benzene Toluene Ethylbenzene m/p-Xylene o-Xylene Tubes Canisters Tubes Canisters Tubes Canisters Tubes Canisters Tubes Canisters ppb ppb ppb ppb ppb ppb ppb ppb ppb ppb Downwind Coker 1.33 <2.00 0.61 2.11 0.44 <2.00 0.49 <2.00 <0.3 <2.00


Checked by:

Section 4. Measurements of SO2 from the Sulphur Recovery Unit A series of measurements was carried out on the 14th August to measure the SO2 flux from the Sulphur Recovery Unit wet gas scrubber plume. Close to the stack this appeared as a visible plume, mainly consisting of water vapour. The measurements were made through the plume, roughly at the point at which it became invisible to the unaided eye. The enhanced backscatter from the plume was used to check that the scans did intercept the plume. The aim of the measurements was to compare the measured flux with an emission rate determined from the stack flow rate and measurements of the stack gas SO2 concentration determined from an installed continuous emissions monitor. However, prior calculations indicated that the likely concentration within the plume would be at or below the DIAL detection limit, and this was found to be the case. The average flux measured by the DIAL was 0.003 lbs/hr, with a standard deviation of 0.008 lbs/hr. Table 4.1 lists the measured SO2 fluxes.

Table 4.1. Measurements of SO2 flux from SRU. The DIAL location was: Refinery Location 13 Avenue F, north of flare 5. 29°22'28.42"N 94°54'54.56"W


Checked by:

Annex 1. Description of the DIAL technique.

Overview of the Dial technique The Differential Absorption Lidar (DIAL) technique is a laser-based remote monitoring technique which enables range-resolved concentration measurements to be made of a wide range of atmospheric species. This section explains the theory of the DIAL technique and describes the NPL system in detail.

Description of the theory of Dial measurements The atmospheric return signal measured by a DIAL system is given by the Light Detection and Ranging (Lidar) equation, a simplified form of which is given in Equation 1.

where Dx is a range independent constant, C(r) is the concentration of an absorber with absorption coefficient αx and Ax(r) is the absorption coefficient due to all other atmospheric absorption, Ex is the transmitted energy and Bx is the backscatter coefficient for the atmosphere at wavelength x. The equation has three basic components: - a backscatter term based on the strength of the signal scattering medium - parameters associated with the DIAL system - a term which is a measure of the amount of absorption of the signal which has occurred

due to the presence of the target species. In the DIAL technique, the laser is operated alternately at two adjacent wavelengths. One of these, the "on-resonant wavelength", is chosen to be at a wavelength which is absorbed by the target species. The other, the "off-resonant wavelength", is chosen to be at a wavelength which is not absorbed significantly by the target species. Pairs of on- and off-resonant signals are then acquired and averaged separately until the required signal to noise ratio is achieved. The two wavelengths used are close together, hence the atmospheric terms Ax(r) and Bx(r) in the lidar equation can be assumed to be the same for both wavelengths. These terms are then cancelled by taking the ratio of the two returned signals. The path-integrated concentration (CL) may be derived (Equation 2) by multiplying the logarithm of the ratio of the signals by the ratio of the absorption of the two wavelengths by the target species.

x xx

2 x0

r

x xP (r)= EDr

B (r) {-2 [ A (r )+ C(r )]d r}exp ∫ ′ ′ ′α (1)


Checked by:

where N is the number of pulse pairs averaged, Δα = αOFF-αON is the differential absorption coefficient and S represents the received power after normalisation for the on- and off-resonant signals respectively. This path-integrated concentration represents the total concentration of the target species in the atmosphere along the measured line-of-sight out to the range r. The range-resolved concentration can then be derived by differentiating the path-integrated concentration (Equation 3).

where C(r) is the point concentration at range r along the line-of-sight.

Description of facility operated by NPL The DIAL system operated by NPL is housed in a mobile laboratory. It can operate in the infrared and ultraviolet spectral regions allowing coverage of a large number of atmospheric species. A scanner system directs the output beam and detection optics, giving almost full coverage in both the horizontal and vertical planes. The system also contains ancillary equipment for meteorological measurements, including an integral 10 m meteorological mast with wind speed, direction, temperature and humidity measurements. The system is fully self contained, with power provided by an on board generator, and has full air conditioning to allow operation in a range of ambient conditions. The following sections describe the DIAL system in more detail. Source The source employs a combination of Nd-YAG and dye lasers together with various non-linear optical stages to generate the tuneable infrared and ultraviolet wavelengths. The source has a pulse repletion rate of 10 Hz and an output laser pulse duration of ~10 ns. A small fraction of the output beam in each channel is split off by a beam splitter and measured by a pyroelectric detector (PED) to provide a value for the transmitted energy with which to normalise the measured backscatter return. Detection The returned atmospheric backscatter signal is collected by the scanning telescope. This directs the collected light into separate paths for the infrared and ultraviolet lidar channels. The

CL(r)=1

21N

S (r)S (r)i=1

NON,i

OFF,iΔα ∑ log (2)

C(r)=dCL(r)

dr (3)


Checked by:

returned light passes through band pass filters relevant to each detection channel and is then focused onto the detection elements. Solid-state cryogenically-cooled detectors are used in the infrared channel and low-noise photomultipliers in the ultraviolet. After amplification the signals from these detectors are digitised using high speed digitisers. The digitisers are clocked using a clock generator triggered by an optical detector in the transmission chain. This ensures the range gating is correctly synchronised to the laser pulse transmission. The signals from the PED monitoring the transmitted energy are also digitised and stored. Data Analysis The data acquired are analysed, using the DIAL techniques described below, to give the range-resolved concentration along each line-of-sight. The data analysis process consists of the following steps : i) Background subtraction Any DC background value is subtracted from the signals. This measured background takes account of any DC signal offset which may be present due to electronic offsets and from incident background radiation. The background level is derived from the average value of the far field of the returned lidar signal where no backscattered light is present. ii) Normalisation for variation in transmitted energy The two signal returns are normalised using the monitored values of the transmitted energy. The mean transmitted energy is used to normalise the averaged return signal. For this application, this has been shown to be equivalent to normalising individual shots against transmitted energy and then averaging the normalised values. iii) Calculation of path-integrated concentration The path-integrated concentration of the target species, out to the range r, is calculated by multiplying the log of the ratio of the returned normalised signals by the differential absorption. The absorption coefficients used in this calculation are derived from high-resolution spectroscopy carried out using reference gas mixtures at NPL. iv) Derivation of range-resolved concentrations. In order to better visualise the data the integrated concentration profiles are piecewise differentiated to give the range-resolved concentration along the line-of-sight. v) Calculation of emission fluxes Range-resolved concentration measurements along different lines-of-sight are combined to generate a concentration profile. This is carried out using algorithms developed at NPL which reduce artefacts due to the difference in data density at different ranges, due to the polar scanning format of the data. The emission flux is then determined using the concentration profile together with meteorological data. The emitted flux is calculated using the following mathematical steps:-


Checked by:

(a) The product is formed of the gas concentration measured with the DIAL technique at a given point in space, and the component of the wind velocity perpendicular to the DIAL measurement plane at the same location.

(b) This product is computed at all points within the measured concentration profile, to

form a two-dimensional array of data. (c) This array of results is then integrated over the complete concentration profile to

produce a value for the total emitted flux. Considerable care is needed in applying the meteorological data, particularly when the concentration profile measured by the DIAL technique has large and complex spatial variations since, for example, errors in the wind speed in regions where large concentrations are present will significantly affect the accuracy of the results. In such cases, a more complex procedure is used which employs a further software package to combine the data from the set of anemometers with that of an additional meteorological model, to generate the complete wind field over the concentration profile. This model calculates the variation of wind speed with height, as a function of various parameters (such as the roughness of the terrain). The calculated wind field is then combined with the measured gas concentration profile using the procedure described above. A summary of the ultraviolet and infrared performance capabilities of the NPL DIAL facility are given in Tables A1.1 and A1.2. The values given in these tables are based on the actual levels of performance of the system obtained during field measurements, rather than calculations based on theoretical noise performances. For simplicity the numbers are presented as a single concentration sensitivity and maximum range values. However, the detailed performance behaviour of a DIAL system is much more complex and there are a number of key points that should be noted :

• The DIAL measurement is of concentration per unit length rather than just concentration. So the sensitivity applies for a specified pathlength – 50 metres in this case. Measurements over a shorter path would have a lower sensitivity, and would be more sensitive over a longer path.

• Since the backscattered lidar signal varies with range, generally following a (range)-2 function, the sensitivity is a function of range. The sensitivity values given in the table apply at a range of 200 metres, and these will get poorer at longer ranges.

• The maximum range of the system is generally determined by the energy of the emitted pulse and the sensitivity of the detection system, except in the case of nitric oxide where range is limited by oxygen absorption at the short ultraviolet wavelengths required for this species.

• In all cases the performance parameters are based on those obtained under typical meteorological conditions. For the ultraviolet measurements the meteorological conditions do not have a great effect on the measurements as the backscattered signal level is predominantly determined by molecular (Rayleigh) scattering, and this does not vary greatly. However, in the infrared the dominant scattering mechanism is from particulates (Mie scattering). So the signal level, and therefore the sensitivity, is dependant on the particular loading of the atmosphere, and this can vary dramatically over relatively short timescales.


Checked by:

The NPL DIAL has a theoretical range resolution of 7 metres along the measurement beam, and a vertical or horizontal scan resolution which can be less than 1 metre at 100 metres. However, the actual range resolution determined by the signal averaging used, will depend on atmospheric conditions and the concentration of the measured pollutant, and may be of the order of 30 m. The DIAL is able to make measurements of a wide range of compounds, including benzene and other aromatics, individual VOCs and total VOCs, see Tables 2a and 2b. The methodology for obtaining measurements of the total VOC content from C3 to C15 is provided below. It consists of the combination of DIAL measurements with air sampling and GC analysis. The system is able to monitor individual aromatic compounds and VOC species, which have absorption features in the IR and UV spectral regions covered by the DIAL system. NPL has the spectral expertise, access to spectral libraries and in-house spectroscopic capability to assess the DIAL sensitivity for additional individual species. The general hydrocarbon measurement listed in Table A1.2 uses an infrared absorption that is common to all hydrocarbons with three or more carbon atoms, linked to the stretch frequency of the carbon-hydrogen bond. As such it provides a measure of the mixture of volatile organic compounds (VOCs) that are present at an oil or petrochemical site. The pair of infrared wavelengths used for this DIAL measurement are selected so that the absorption per unit mass is relatively invariant with respect to the mix of different hydrocarbons that are present. However, the sensitivity of this measurement in terms of ppb of hydrocarbon depends on the mixture of species present, and the value given in the table reflects the typical mix of hydrocarbons found at oil refineries. Although the general hydrocarbon measurement provides a good estimate of the overall amount of hydrocarbons present, the accuracy of this measurement can be improved, and the total VOC concentration calculated, by combining the DIAL measurements with the results of gas chromatography (GC) analysis of the emitted gases. The standard procedure for this involves taking whole air samples around the site in locations where the DIAL measurements show the emitted plumes are present. The VOCs present in these samples are identified and quantified by GC analysis. The results provide the relative levels of all the VOCs present with a concentration of 0.1 ppb or higher. The results of this analysis are combined with NPL’s unique spectral library of quantified infrared absorptions of an extensive set of VOCs to calculate the combined absorption coefficient for the actual VOC mixture present at the site. Appling this absorption coefficient to the DIAL results enables the total VOC emission rates to be calculated.


Checked by:

Species Sensitivity(1) Maximum range(2)

Nitric oxide 5 ppb 500 m Sulphur dioxide 10 ppb 3 km

Ozone 5 ppb 2 km Benzene 10 ppb 800 m Toluene 10 ppb 800 m

Table A1.1 Ultraviolet capability of NPL DIAL Facility

Species Sensitivity(1) Maximum range(2)

Methane 50 ppb 1 km Ethane 20 ppb 800 m Ethene 10 ppb 800 m Ethyne 40 ppb 800 m

General hydrocarbons 40 ppb 800 m Hydrogen chloride 20 ppb 1 km

Methanol 200 ppb 500 m Nitrous oxide 100 ppb 800 m

Table A1.2 Infrared capability of NPL DIAL Facility Note 1. The concentration sensitivities apply for measurements of a 50 metre wide plume at a range of 200 metres, under typical meteorological conditions. Note 2. The range value represents the typical working maximum range for the NPL DIAL system.


Checked by:

Calibration and validation The NPL DIAL system has several in-built calibration techniques and procedures. The most important are the in-line gas calibration cells. The gas cells are filled with known concentrations of the target species, from NPL standard gas mixtures, which are directly traceable to national standards. A fraction of the transmitted beam is split off and directed through a gas cell to a PED, in the same way as the beam for the transmitted energy monitors. This provides a direct measurement of the differential absorption at the operating wavelengths by the target gas. The transmission through the gas cells is continuously monitored during the operation of the system to detect any possible drift in the laser wavelengths. The system also employs a wavemeter to monitor the wavelengths transmitted during operation. The calibration cells are also periodically placed in the output beam to show the concentration response of the whole system is as expected. A number of field comparisons have been undertaken to demonstrate the accuracy of the measurements obtained with DIAL. Examples of these carried out by NPL are detailed below: i) Intercomparisons have been carried out in the vicinity of chemical and petrochemical

plants where a large number of different volatile organic species are present. In these intercomparisons, the DIAL radiation was directed along the same line of sight as a line of point samplers. The point samplers were operated either by drawing air into internally-passivated, evacuated gas cylinders or by pumping air at a known rate, for a specified time, through a series of absorption tubes which efficiently absorb all hydrocarbon species in the range C2 - C8. The results obtained for the total concentrations of VOCs measured by the point samplers and those measured by the infrared DIAL technique agreed within ± 15%. The concentrations of atmospheric toluene measured by the ultraviolet DIAL system agreed with those obtained by the point samplers to within ± 20%.

ii) The ultraviolet DIAL system was used to monitor the fluxes and concentrations of

sulphur dioxide produced from combustion and emitted by industrial stacks. These stacks were instrumented with calibrated in-stack sampling instruments. The results of the two sets of measurements agreed to within ± 12%.

iii) DIAL Measurements of controlled releases of methane from a stack agreed with the

known emission fluxes to within ± 15%.


Checked by:

Measurements of gas cell concentrations during the Texas measurement campaign. A series of measurements was made of a 10 cm gas cell filled with a pentane standard gas mixture provided by TCEQ/REFINERY. The measurements were made by placing the gas cell in the output transmitted beam. This introduces an absorption into the DIAL signal at zero range, which has the effect of introducing an offset in the signals, equivalent to the total concentration-pathlength of the gas in the cell. The differential absorption for pentane at the wavelengths used for the VOC DIAL measurements was determined from the NPL quantified spectral database, and this was used to calculate the concentration of pentane in the gas cell. In addition a number of other gas cell measurements were made of propane and benzene. The results of these measurements are given in Table A1.3 below. These present the means of a series of measurements made of the gas cells on different days.

Date Scans Cell Measured column

Concentration (ppm)

Predicted Actual 27/07/2007 110-112 10 cm propane 0.57 ± 0.04 8180 ± 530 8413

27/07/2007 117-123 10 cm pentane 1.16 ± 0.12 7100 ± 700 7500

30/07/2007 222-223 10cm pentane 1.34 ± 0.20 8200 ± 1200 7500

07/08/2007 371-372 10cm pentane 1.24 ± 0.01 7600 ± 100 7500

17/08/2007 557-558 20cm benzene 0.09 900 ± 70 1000

Table A1.3, Measurements of gas cell concentrations. The uncertainty figures are based on the standard deviation of the individual measurements.


Checked by:

NPL open-path calibration facility NPL has also developed and operate a full-scale facility for the calibration of open path monitors, including DIAL. This consists of a 10 m long windowless cell able to maintain a uniform, independently-monitored concentration of a gaseous species along its length. This provides a known controlled section of the atmosphere with traceable concentration over a defined range (10m). The absence of windows removes reflections and other artefacts from measurements made using optical techniques, providing a direct way to validate and assess the calibration of DIAL instruments. The calibration facility is windowless with a 1 m diameter, to minimise any beam reflections from the cell walls and ends. At each end of the cell is an annular calibration-gas feed ring with multiple outlets injecting the calibration gas mixture into the cell. A ring of tangential fans around the centre of the cell extract gas and entrained air pulled in through the open ends of the cell. This ensures the backscatter in the cell approximates to the ambient air conditions. Each fan has a long exhaust tube to avoid recirculation of the gas into the cell. This facility has been employed to directly validate VOC measurements by the NPL DIAL facility [2]. Figure A1.1 The NPL 10m calibration cell. The facility provides the ability to generate a defined concentration path and so it also provides range-resolution validation for DIAL and lidar instruments. The system was used to validate the DIAL with a number of measurements of propane and methane, as a part of its acceptance tests for Siemens, Shell and British Gas. [1] Measurements of the Emissions to Atmosphere of Volatile Organic Compounds

from the Hellenic Aspropyrgos Oil Refinery; T D Gardiner, M.J.T. Milton, R.A. Robinson, P.T.Woods, A.S.Andrews, H. D’Souza, D Alfonso, N.R Swann; NPL Report QM S99, Sept 1996

[2] Calibration of DIAL and Open Path Systems Using External Gas Cells; M.J.T. Milton,

P.T. Woods, R.H. Partridge, B.A Goody; Proc. Europto, Munich 1995.


Checked by:

Annex 2 Speciated VOC measurements and Flux Correction Factors The DIAL measures VOCs by measuring the differential absorption of two wavelengths of light. The wavelengths used, at around 3000 cm-1, are chosen to measure, in effect, the C-H stretch in the hydrocarbons for C3 and above. The sensitivity of the DIAL is slightly different to different hydrocarbons, and for example an oxygenated hydrocarbon will give a different absorption per mass than a straight chain alkane. The differential absorption strength used was calibrated to give a mass emission rate for a gasoline vapour. A different ‘cocktail’ of hydrocarbons would give a slightly different response per unit mass. Air samples were taken at locations which would provide an indication of the actual speciation of the emission fluxes sampled by the DIAL. If the actual (relative) composition is known, from the air sample analyses, then by determining the absorption of each constituent it is possible to determine a correction factor which will provide the true ‘mass value’ assuming the emissions measured have a similar relative composition to the sampled air.

Flux Correction Factors The multi-component analysis of the ambient air sample results was used to determine the differential absorption coefficient (in m2g-1) for the particular cocktail of hydrocarbons in each sample. This was done by taking the differential absorption coefficient for each species from the NPL quantified database of the infrared absorption of industrial organic species, and scaling it by the mass-fraction of that species in the sample. The standard DIAL differential absorption coefficient for VOC measurements is 0.49 m2g-1, based on the infrared absorption of gasoline vapour. These results are summarised in Table A2.1, which show the differential absorption coefficients, and the factor by which the standard DIAL VOC flux results need to be multiplied by to take into account the particular VOC mix present.


Checked by:

Table A2.1. Flux correction factors determined from each of the air samples. Air samples were taken at a number of locations around the plants. Figures A2.1, A2.2, A2.3 and A2.4 show the locations of the samples. The Air samples were taken using SUMO canisters and pumped Perkin Elmer Automatic Thermal Desorption (ATD) tubes. The ATD tubes were sampled at a flow rate of 40ml/min, to enable a comparison between the SUMO canisters and the ATD tubes; a flow restrictor was fitted to canister, so each sample was taken over the same time period. The sampler tubes are approximately 6mm OD and length 90mm long. The sampler tubes contained approximately 200milligrams to 300 milligrams of sorbent. Two sorbent tubes were used in series containing two sorbent materials used, a porous polymer (Tenax TA) and a carbon black (Carbopack X). Different sorbents are needed to cover the diverse boiling point ranges and chemical functional groups of VOCs. The SUMO canister analysis gave results in the Carbon number range of C2-C10, whilst the linked ATD tubes containing Tenax TA and Carbopack-X gave a Carbon number range of C4-C22

Tube Sample diff abs flux factor

DTTS1 0.62 0.8 DTTS2 0.39 1.3 DTTS3 0.55 0.9 DTTS4 0.31 1.6 DTTS5 0.29 1.7 DTTS6 0.41 1.2 DTTS7 0.37 1.3 DTTS8 0.41 1.2 DTTS9 0.44 1.1 DTTS10 0.46 1.1 DTTS11 0.24 2.0 DTTS12 0.43 1.1 DTTS13 0.24 2.1 DTTS14 0.50 1.0 DTTS15 0.44 1.1 DTTS16 0.47 1.0


Checked by:

Figure A2.1 Location of air samples around the naphtha tanks at the Bulk Terminal


Checked by:

Figure A2.2 Location of air samples in the crude oil storage areas at the Refinery.

Figure A2.3 Location of air samples in the gasoline tanks and coker areas at the Refinery.

DTTS 5

DTTS 6

DTTS 7

DTTS 11

DTTS 8

DTTS 12

DTTS 9 & 10

DTTS 16


Checked by:

Figure A2.4 Location of air samples in the gasoline tanks and coker areas at the Refinery.

DTTS 13

DTTS

DTTS 15


Checked by:

Sampling Procedure The ATD tubes were attached to the flow restrictor so as the tubes were at the same height as the top of flow restrictors sample entry point.

The samples were taken at a height of approximately 2m by mounting the sampling array on a Tripod. The samples were taken by opening the valve on the SUMO canister and starting the sampling pumps. The flow restrictors had been set so as they sampled over a period of 1.25 hours.

ATD Tubes

Flow restrictor


Checked by:

Analysis of the ATD tubes The method of analysis was based on EN ISO 16017-2 [1] and was carried out using UKAS (United Kingdom Accreditation Service) accredited method QPDQM/B/526. This method combines Automatic Thermal Desorption with Gas Chromatography, with a Flame Ionisation Detector [1] The analysis instrument used is an Automated Thermal Desorber autosampler coupled to a Gas Chromatograph usually with a flame ionisation detector. The VOCs are released from the sampler tube using a heated oven in an inert gas stream of helium. The VOCs are refocused onto a small cold trap prior to transfer onto the gas chromatography column. Generally a coated fused silica gas chromatography column of diameter 320 micrometers and length 60 meters is used to separate the individual VOCs collected. Using VOC standard materials, the identification of the individual VOC components are compared to the column elution time (retention time) of the standard VOC materials. The mass of VOCs collected is quantified using the flame ionisation detector. A series of calibrations standards are used to calibrate the flame ionisation detector response. The concentration of the VOC in ambient air is then calculated using from the mass collected and the volume of air sampled. Tables A2.2 and A2.3 present the results from the canister and tube analyses respectively. The tube analyses were carried out by NPL’s in house accredited analysis laboratory, the canister samples were analysed by EAS Environmental Analytical Service, Inc, California.


Checked by:

Table A2.2 Full results of the speciation analyses on the Canister Samples. Sample No. DTTS1 DTTS2 DTTS3 DTTS4 DTTS5 DTTS6 DTTS7 DTTS8Date 17-Jul-

07 18-Jul-

07 19-Jul-

07 21-Jul-

07 28-Jul-

07 28-Jul-

07 29-Jul-

07 30-Jul-

07 Start time 17:17 16:20 12:31 03:57 14:25 14:40 16:56 16:39 End time 18:09 17:27 13:44 05:06 15:47 16:00 18:12 17:51 Average Wind Direction/deg 164 128 322 95 170 174 149 159 Average Wind Speed/mph 4.5 2.4 4.6 2.3 4.3 4.3 3.9 4.2 Cannister Number 169 610 708 881 190 996 762 606

Species ppb V ppb V ppb V ppb V ppb V ppb V ppb V ppb V Ethene <1.93 <1.91 4.76 <1.91 <1.96 <1.94 <2.00 20.24 Acetylene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 Ethane <1.93 1.98 15.93 3.14 2.83 10.86 7.01 223.95Propene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 Propane <1.93 6.32 10.65 2.15 <1.96 5.45 12.16 23.61 I-Butane 3.73 5.17 12.74 8.22 6.02 16.72 8.00 72.47 Methanol <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 1-Butene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 1,3-Butadiene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 4.88 n-Butane <1.93 <1.91 24.44 2.43 <1.96 3.97 6.69 11.16 t-2-Butene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 4.70 c-2-Butene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 Ethanol <1.93 <1.91 <2.09 7.52 <1.96 <1.94 <2.00 2402.73-Methyl-1-butene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 Acetone <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 i-Pentane 323.89 5.67 63.51 17.78 4.80 26.94 8.13 36.51 1-Pentene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 Isopropanol <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 2-Methyl-1-butene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 n-Pentane <1.93 <1.91 38.65 <1.91 <1.96 2.65 2.35 39.77 Isoprene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 t-2-Pentene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 c-2-Pentene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 Tert butyl alcohol <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 2-Methyl-2-butene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 2,2-Dimethylbutane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 Cyclopentene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 n-Propanol <1.93 <1.91 2.78 <1.91 <1.96 <1.94 <2.00 4.66 Cyclopentane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 Methyl tert butyl ether <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 2,3-Dimethylbutane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 2-Methylpentane <1.93 <1.91 8.13 <1.91 <1.96 <1.94 <2.00 <2.06 3-Methylpentane <1.93 <1.91 5.33 <1.91 <1.96 <1.94 <2.00 <2.06 1-Hexene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 n-Hexane <1.93 <1.91 10.26 <1.91 <1.96 <1.94 <2.00 12.04 Diisopropyl ether <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 3-Methylcyclopentene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 Ethyl tert butyl ether <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 Methylcyclopentane <1.93 <1.91 3.77 <1.91 <1.96 <1.94 <2.00 <2.06 2,4-Dimethylpentane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 Benzene <1.93 <1.91 3.17 <1.91 <1.96 <1.94 5.19 <2.06 Cyclohexane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 2-Methylhexane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 2,3-Dimethylpentane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 3-Methylhexane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 2-Methyl-1-hexene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06


Checked by:

Sample No. DTTS1 DTTS2 DTTS3 DTTS4 DTTS5 DTTS6 DTTS7 DTTS8Date 17-Jul-

07 18-Jul-

07 19-Jul-

07 21-Jul-

07 28-Jul-

07 28-Jul-

07 29-Jul-

07 30-Jul-

07 Start time 17:17 16:20 12:31 03:57 14:25 14:40 16:56 16:39 End time 18:09 17:27 13:44 05:06 15:47 16:00 18:12 17:51 Average Wind Direction/deg 164 128 322 95 170 174 149 159 Average Wind Speed/mph 4.5 2.4 4.6 2.3 4.3 4.3 3.9 4.2 Cannister Number 169 610 708 881 190 996 762 606

Tert amyl methyl ether <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 2,2,4-Trimethylpentane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 n-Heptane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 Methylcyclohexane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 2,5-Dimethylhexane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 2,4-Dimethylhexane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 2,3,4-Trimethylpentane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 Toluene <1.93 <1.91 3.03 <1.91 1.99 <1.94 3.35 28.65 2,3-Dimethylhexane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 2-Methylheptane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 4-Methylheptane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 2-Ethyl-3-methylpentane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 3-Methylheptane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 2-Methyl-1-heptane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 n-Octane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 2.12 Ethylbenzene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 m,p-xylene <1.93 2.74 <2.09 <1.91 <1.96 <1.94 <2.00 2.84 Styrene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 o-xylene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 1-Nonene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 n-Nonane <1.93 4.09 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 i-Propylbenzerne <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 n-Propylbenzerne <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 a-Pinene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 3-Ethyltoluene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 4-Ethyltoluene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 1,3,5-Trimethylbenzene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 2-Ethyltoluene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 b-Pinene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 1,2,4-Trimethylbenzene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 2.07 <2.06 n-Decane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 1,2,3-Trimethylbenzene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 2.28 Indan <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 d-Limonene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 1,3-Diethylbenzene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 1,4-Diethylbenzene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 n-Butylbenzene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 1,4-Dimethyl-2-ethylbenzene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 1,3-Dimethyl-4-ethylbenzene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 1,2-Dimethyl-4-ethylbenzene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 Undecane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 1,2,4,5-Tetramethylbenzene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 1,2,3,5-Tetramethylbenzene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 Napthalene <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06 Dodecane <1.93 <1.91 <2.09 <1.91 <1.96 <1.94 <2.00 <2.06


Checked by:

Sample No. DTTS9 DTTS10 DTTS11 DTTS12 DTTS13 DTTS14 DTTS15 DTTS16Date 3-Aug-

07 3-Aug-

07 5-Aug-

07 6-Aug-

07 7-Aug-

07 17-Aug-

07 17-Aug-

07 18-Aug-

07 Start time 13:30 15:37 04:48 04:30 04:55 17:53 17:29 20:05 End time 14:47 17:28 06:03 05:54 06:19 18:58 18:30 20:57 Average Wind Direction/deg 84 113 183 140 201 131 131 128 Average Wind Speed/mph 4.3 5 3.4 2.6 2 4.4 4.7 2.4 Cannister Number 2968 711 782 2962 2970 758 732 656

Species ppb V ppb V ppb V ppb V ppb V ppb V ppb V ppb V Ethene 9.65 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 Acetylene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 Ethane 24.58 27.98 5.36 117.36 3.35 9.17 6.96 15.92 Propene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 Propane 28.92 14.99 8.47 48.25 80.36 13.33 16.46 10.30 I-Butane 15.23 9.19 7.12 39.59 5.26 20.06 15.94 5.97 Methanol <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 1-Butene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 1,3-Butadiene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 3.09 <2.00 n-Butane 16.33 9.88 4.99 64.73 <1.93 11.47 29.96 5.95 t-2-Butene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 c-2-Butene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 Ethanol <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 3-Methyl-1-butene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 Acetone <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 i-Pentane 19.88 <1.97 <1.94 <1.96 8.87 53.05 110.80 16.86 1-Pentene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 Isopropanol <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 2-Methyl-1-butene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 n-Pentane <1.94 <1.97 <1.94 <1.96 <1.93 4.64 39.57 11.68 Isoprene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 t-2-Pentene <1.94 <1.97 <1.94 2.85 <1.93 <2.00 <1.88 <2.00 c-2-Pentene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 Tert butyl alcohol <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 2-Methyl-2-butene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 2,2-Dimethylbutane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 5.39 <2.00 Cyclopentene <1.94 2.91 <1.94 9.48 <1.93 <2.00 <1.88 <2.00 n-Propanol <1.94 2.11 <1.94 2.86 <1.93 <2.00 <1.88 <2.00 Cyclopentane <1.94 <1.97 <1.94 <1.96 <1.93 2.25 <1.88 <2.00 Methyl tert butyl ether <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 2,3-Dimethylbutane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 2-Methylpentane <1.94 <1.97 <1.94 <1.96 <1.93 5.21 <1.88 10.70 3-Methylpentane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 5.76 1-Hexene 2.88 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 n-Hexane <1.94 <1.97 <1.94 <1.96 6.72 <2.00 <1.88 <2.00 Diisopropyl ether <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 3-Methylcyclopentene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 Ethyl tert butyl ether <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 Methylcyclopentane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 2,4-Dimethylpentane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 Benzene <1.94 <1.97 <1.94 <1.96 4.58 <2.00 <1.88 <2.00 Cyclohexane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 2-Methylhexane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 2,3-Dimethylpentane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 7.20 3-Methylhexane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 2-Methyl-1-hexene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00


Checked by:

Sample No. DTTS9 DTTS10 DTTS11 DTTS12 DTTS13 DTTS14 DTTS15 DTTS16Date 3-Aug-

07 3-Aug-

07 5-Aug-

07 6-Aug-

07 7-Aug-

07 17-Aug-

07 17-Aug-

07 18-Aug-

07 Start time 13:30 15:37 04:48 04:30 04:55 17:53 17:29 20:05 End time 14:47 17:28 06:03 05:54 06:19 18:58 18:30 20:57 Average Wind Direction/deg 84 113 183 140 201 131 131 128 Average Wind Speed/mph 4.3 5 3.4 2.6 2 4.4 4.7 2.4 Cannister Number 2968 711 782 2962 2970 758 732 656

Tert amyl methyl ether <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 2,2,4-Trimethylpentane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 n-Heptane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 5.54 <2.00 Methylcyclohexane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 2,5-Dimethylhexane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 2,4-Dimethylhexane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 2,3,4-Trimethylpentane 3.92 <1.97 2.58 <1.96 <1.93 <2.00 <1.88 <2.00 Toluene <1.94 <1.97 <1.94 <1.96 1.96 12.26 20.52 2.11 2,3-Dimethylhexane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 2-Methylheptane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 4-Methylheptane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 2-Ethyl-3-methylpentane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 3-Methylheptane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 2-Methyl-1-heptane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 n-Octane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 2.09 <2.00 Ethylbenzene <1.94 <1.97 <1.94 <1.96 <1.93 2.04 2.46 <2.00 m,p-xylene <1.94 <1.97 <1.94 <1.96 <1.93 5.51 8.49 <2.00 Styrene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 o-xylene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 2.59 <2.00 1-Nonene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 n-Nonane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 i-Propylbenzerne <1.94 2.13 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 n-Propylbenzerne <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 a-Pinene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 3-Ethyltoluene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 4-Ethyltoluene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 1,3,5-Trimethylbenzene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 2-Ethyltoluene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 b-Pinene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 1,2,4-Trimethylbenzene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 2.06 <2.00 n-Decane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 1,2,3-Trimethylbenzene 2.54 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 Indan <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 d-Limonene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 1,3-Diethylbenzene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 1,4-Diethylbenzene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 n-Butylbenzene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 1,4-Dimethyl-2-ethylbenzene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 1,3-Dimethyl-4-ethylbenzene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 1,2-Dimethyl-4-ethylbenzene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 Undecane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 1,2,4,5-Tetramethylbenzene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 1,2,3,5-Tetramethylbenzene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 Napthalene <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00 Dodecane <1.94 <1.97 <1.94 <1.96 <1.93 <2.00 <1.88 <2.00


Checked by:

Table A2.3 Full results of the speciation analyses on the Sorbent Tube Samples.

Sample No. TRV BLK

A TRV BLK

B DTTS1 DTTS2 DTTS3 DTTS4 DTTS5

Date 17-Jul-07 18-Jul-07 19-Jul-07 21-Jul-07 28-Jul-07Start time 17:17 16:20 12:31 03:57 14:25 End time 18:09 17:27 13:44 05:06 15:47

Sample Volume/ml 2500.5 3230.8 2970.4 3297.5 3948.6 Average Wind Direction/deg 164 128 322 95 170 Average Wind Speed/mph 4.5 2.4 4.6 2.3 4.3 ppb v/v ppb v/v ppb v/v ppb v/v ppb v/v ppb v/v ppb v/v i-Butane 0.8 0.7 4.4 1.6 13.1 7.5 2.4 n-Butane <0.6 <0.6 <0.6 0.6 19.3 1.0 <0.6 trans-2-Butene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 1-Butene <0.6 <0.6 <0.6 <0.6 0.8 <0.6 <0.6 2-methylpropene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 cis-2-butene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 1,3 Butadiene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 i-pentane <0.6 <0.6 60.8 0.9 13.7 1.9 0.6 1-pentene <0.6 <0.6 2.8 <0.6 0.9 <0.6 <0.6 2 me-1-butene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 isoprene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-pentane <0.6 <0.6 <0.6 1.4 34.6 1.1 <0.6 cis-2-pentene (Z)- <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 trans-2-pentene (E)- <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 1,2 dime cyclopropane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 2,2-di me butane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 cyclo pentene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 4-me-1-pentene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 2,3-di me butane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 cyclo pentane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 2-me pentane <0.6 <0.6 <0.6 <0.6 7.1 <0.6 <0.6 4-me-cis-2-pentene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 4-me-trans-2-pentene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 3-me pentane <0.6 <0.6 <0.6 <0.6 3.6 <0.6 <0.6 1-hexene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-hexane <0.6 <0.6 6.5 <0.6 11.8 <0.6 <0.6 cis-2-hexene (Z)- <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 trans-3-hexene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 trans-2-hexene (E)- <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 methylcyclopentane <0.6 <0.6 <0.6 <0.6 3.0 <0.6 <0.6 2,4 di me pentane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 2,2,3-trimethyl-1-butene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 2,2,3-tri-me butane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 benzene <0.2 <0.2 0.2 0.2 5.0 1.8 0.2 cyclohexane <0.6 <0.6 <0.6 <0.6 1.5 <0.6 <0.6 2-methyl hexane <0.6 <0.6 <0.6 <0.6 1.2 <0.6 <0.6 2,3 di me pentane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 3-me hexane <0.6 <0.6 <0.6 <0.6 1.6 <0.6 <0.6 cyclohexene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 1-heptene <0.6 <0.6 <0.6 <0.6 0.9 <0.6 <0.6 2,2,4 tri me pentane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-heptane <0.6 <0.6 <0.6 <0.6 2.3 <0.6 <0.6 methylcyclohexane <0.6 <0.6 <0.6 <0.6 2.0 <0.6 <0.6 2,3,4 tri me pentane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 2,3,3-tri me pentane <0.6 <0.6 <0.6 <0.6 2.7 2.4 <0.6


Checked by:

Sample No. TRV BLK

A TRV BLK

B DTTS1 DTTS2 DTTS3 DTTS4 DTTS5

Date 17-Jul-07 18-Jul-07 19-Jul-07 21-Jul-07 28-Jul-07Start time 17:17 16:20 12:31 03:57 14:25 End time 18:09 17:27 13:44 05:06 15:47

Sample Volume/ml 2500.5 3230.8 2970.4 3297.5 3948.6 Average Wind Direction/deg 164 128 322 95 170 Average Wind Speed/mph 4.5 2.4 4.6 2.3 4.3

ppb v/v ppb v/v ppb v/v ppb v/v ppb v/v ppb v/v ppb v/v toluene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 2-methyl heptane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 4-methyl heptane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 cycloheptene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 cycloheptane <0.6 <0.6 <0.6 <0.6 1.0 <0.6 <0.6 n-octane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 ethylbenzene <0.6 <0.6 <0.6 <0.6 1.5 <0.6 <0.6 m/p-xylene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 2 me octane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 3 me octane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 styrene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 o-xylene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 nonane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 cis-cyclooctene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 isopropylbenzene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 cyclooctane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-propylbenzene <0.6 <0.6 <0.6 <0.6 <0.6 1.0 <0.6 3-ethyl toluene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 4-ethyl toluene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 1,3,5 tri me benzene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 2-ethyl toluene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 2-me nonene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 1,2,4 tri me benzene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-decane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 1,2,3-Trimethylbenzene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 1,2,3,5-tetra me benzene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-butyl benzene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-undecane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-pentylbenzene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-dodecane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 naphthalene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-tridecane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-tetradecane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-pentadecane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-hexadecane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-heptadecane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-octadecane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-nonadecane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-cosane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-docosane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6


Checked by:

Sample No. DTTS6 DTTS7 DTTS8 DTTS9 DTTS10 DTTS11 DTTS12

Date 28-Jul-07 29-Jul-07 30-Jul-07 3-Aug-07 3-Aug-07 5-Aug-07 6-Aug-07Start time 14:40 16:56 16:39 13:30 15:37 04:48 04:30 End time 16:00 18:12 17:51 14:47 17:28 06:03 05:54

Sample Volume/ml 3927.9 3637.8 3409.8 5289 3600.9 3980.4 Average Wind Direction/deg 174 149 159 84 113 183 140 Average Wind Speed/mph 4.3 3.9 4.2 4.3 5 3.4 2.6

ppb v/v ppb v/v ppb v/v ppb v/v ppb v/v ppb v/v ppb v/v i-Butane 7.5 9.6 24.8 6.5 9.4 3.2 11.8 n-Butane 8.7 4.2 9.8 <0.6 13.8 3.4 21.6 trans-2-Butene <0.6 <0.6 1.3 <0.6 <0.6 <0.6 1.1 1-Butene <0.6 <0.6 1.4 <0.6 <0.6 <0.6 1.0 2-methylpropene <0.6 <0.6 1.4 <0.6 <0.6 <0.6 0.9 cis-2-butene <0.6 <0.6 1.0 <0.6 <0.6 <0.6 <0.6 1,3 Butadiene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 i-pentane 5.4 2.6 16.5 0.7 7.3 2.2 21.4 1-pentene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 2 me-1-butene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 isoprene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-pentane 3.6 1.5 1.6 <0.6 6.4 0.8 4.0 cis-2-pentene (Z)- <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 trans-2-pentene (E)- <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 1,2 dime cyclopropane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 2,2-di me butane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 cyclo pentene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 4-me-1-pentene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 2,3-di me butane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 cyclo pentane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 2-me pentane 0.7 <0.6 <0.6 <0.6 0.8 <0.6 1.7 4-me-cis-2-pentene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 4-me-trans-2-pentene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 3-me pentane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 1.0 1-hexene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-hexane 0.9 <0.6 <0.6 <0.6 1.7 <0.6 1.9 cis-2-hexene (Z)- <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 trans-3-hexene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 trans-2-hexene (E)- <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 methylcyclopentane 0.6 <0.6 <0.6 <0.6 0.7 <0.6 1.1 2,4 di me pentane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 2,2,3-trimethyl-1-butene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 2,2,3-tri-me butane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 benzene 0.9 4.4 0.5 0.2 0.7 0.3 1.6 cyclohexane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 2-methyl hexane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 0.7 2,3 di me pentane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 3-me hexane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 0.7 cyclohexene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 1-heptene <0.6 <0.6 0.9 <0.6 <0.6 <0.6 1.1 2,2,4 tri me pentane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-heptane <0.6 <0.6 <0.6 <0.6 0.9 <0.6 <0.6 methylcyclohexane 0.6 4.8 <0.6 <0.6 0.8 <0.6 0.8 2,3,4 tri me pentane <0.6 1.7 <0.6 <0.6 <0.6 <0.6 <0.6 2,3,3-tri me pentane 1.6 0.9 1.4 <0.6 0.8 1.5 2.1


Checked by:

Sample No. DTTS6 DTTS7 DTTS8 DTTS9 DTTS10 DTTS11 DTTS12

Date 28-Jul-07 29-Jul-07 30-Jul-07 3-Aug-07 3-Aug-07 5-Aug-07 6-Aug-07Start time 14:40 16:56 16:39 13:30 15:37 04:48 04:30 End time 16:00 18:12 17:51 14:47 17:28 06:03 05:54

Sample Volume/ml 3927.9 3637.8 3409.8 5289 3600.9 3980.4 Average Wind Direction/deg 174 149 159 84 113 183 140 Average Wind Speed/mph 4.3 3.9 4.2 4.3 5 3.4 2.6

ppb v/v ppb v/v ppb v/v ppb v/v ppb v/v ppb v/v ppb v/v toluene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 2-methyl heptane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 4-methyl heptane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 cycloheptene <0.6 <0.6 <0.6 <0.6 <0.6 10.0 <0.6 cycloheptane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-octane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 ethylbenzene <0.6 <0.6 1.5 <0.6 <0.6 <0.6 1.4 m/p-xylene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 2 me octane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 3 me octane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 styrene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 o-xylene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 nonane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 cis-cyclooctene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 isopropylbenzene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 cyclooctane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-propylbenzene 0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 3-ethyl toluene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 4-ethyl toluene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 1,3,5 tri me benzene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 2-ethyl toluene <0.6 <0.6 <0.6 <0.6 <0.6 1.3 <0.6 2-me nonene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 1,2,4 tri me benzene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-decane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 1,2,3-Trimethylbenzene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 1,2,3,5-tetra me benzene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-butyl benzene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-undecane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-pentylbenzene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-dodecane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 naphthalene <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-tridecane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-tetradecane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-pentadecane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-hexadecane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-heptadecane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-octadecane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-nonadecane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-cosane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 n-docosane <0.6 <0.6 <0.6 <0.6 <0.6 <0.6 <0.6


Checked by:

Sample No. DTTS13 DTTS14 DTTS15 DTTS16

Date 7-Aug-07 17-Aug-07

17-Aug-07

18-Aug-07

Start time 04:55 17:53 17:29 20:05 End time 06:19 18:58 18:30 20:57

Sample Volume/ml 4000.8 3217.6 2833.9 2532.6 Average Wind Direction/deg 201 131 131 128 Average Wind Speed/mph 2 4.4 4.7 2.4

ppb v/v ppb v/v ppb v/v ppb v/v i-Butane 2.1 10.7 4.0 4.0 n-Butane 0.6 22.8 9.0 7.3 trans-2-Butene <0.6 5.2 2.7 <0.6 1-Butene <0.6 1.6 1.0 <0.6 2-methylpropene <0.6 1.5 0.6 <0.6 cis-2-butene <0.6 3.7 1.9 <0.6 1,3 Butadiene <0.6 <0.6 <0.6 <0.6 i-pentane 1.0 82.2 37.6 8.1 1-pentene 1.1 0.8 0.6 <0.6 2 me-1-butene <0.6 <0.6 <0.6 <0.6 isoprene <0.6 7.7 2.8 4.2 n-pentane <0.6 23.7 8.2 4.2 cis-2-pentene (Z)- <0.6 3.1 1.3 <0.6 trans-2-pentene (E)- <0.6 <0.6 <0.6 <0.6 1,2 dime cyclopropane <0.6 3.8 1.5 <0.6 2,2-di me butane <0.6 <0.6 <0.6 <0.6 cyclo pentene <0.6 <0.6 <0.6 <0.6 4-me-1-pentene <0.6 <0.6 <0.6 <0.6 2,3-di me butane <0.6 2.0 0.7 <0.6 cyclo pentane <0.6 <0.6 <0.6 <0.6 2-me pentane 2.7 11.4 4.2 1.6 4-me-cis-2-pentene <0.6 <0.6 <0.6 <0.6 4-me-trans-2-pentene 0.8 <0.6 <0.6 <0.6 3-me pentane <0.6 6.1 2.4 0.8 1-hexene <0.6 <0.6 <0.6 <0.6 n-hexane 5.3 13.0 4.0 1.6 cis-2-hexene (Z)- <0.6 <0.6 <0.6 <0.6 trans-3-hexene <0.6 0.6 <0.6 <0.6 trans-2-hexene (E)- <0.6 <0.6 <0.6 <0.6 methylcyclopentane <0.6 7.5 2.0 <0.6 2,4 di me pentane <0.6 <0.6 <0.6 <0.6 2,2,3-trimethyl-1-butene <0.6 <0.6 <0.6 <0.6 2,2,3-tri-me butane <0.6 <0.6 <0.6 <0.6 benzene 5.1 12.9 6.7 1.3 cyclohexane <0.6 4.0 2.0 1.5 2-methyl hexane <0.6 3.8 1.3 <0.6 2,3 di me pentane <0.6 0.7 <0.6 <0.6 3-me hexane 0.7 4.1 1.6 <0.6 cyclohexene <0.6 <0.6 <0.6 <0.6 1-heptene <0.6 1.3 <0.6 <0.6 2,2,4 tri me pentane <0.6 <0.6 <0.6 <0.6 n-heptane <0.6 5.7 1.6 <0.6 methylcyclohexane 0.6 6.6 1.7 0.6 2,3,4 tri me pentane <0.6 <0.6 <0.6 <0.6 2,3,3-tri me pentane 0.7 15.2 7.9 <0.6


Checked by:

Sample No. DTTS13 DTTS14 DTTS15 DTTS16

Date 7-Aug-07 17-Aug-07

17-Aug-07

18-Aug-07

Start time 04:55 17:53 17:29 20:05 End time 06:19 18:58 18:30 20:57

Sample Volume/ml 4000.8 3217.6 2833.9 2532.6 Average Wind Direction/deg 201 131 131 128 Average Wind Speed/mph 2 4.4 4.7 2.4

ppb v/v ppb v/v ppb v/v ppb v/v toluene <0.6 <0.6 <0.6 <0.6 2-methyl heptane <0.6 <0.6 <0.6 <0.6 4-methyl heptane <0.6 1.3 <0.6 <0.6 cycloheptene <0.6 3.3 <0.6 <0.6 cycloheptane <0.6 <0.6 0.9 <0.6 n-octane <0.6 2.2 1.8 <0.6 ethylbenzene 0.9 9.4 5.3 <0.6 m/p-xylene <0.6 <0.6 <0.6 <0.6 2 me octane <0.6 0.6 <0.6 <0.6 3 me octane <0.6 <0.6 <0.6 <0.6 styrene <0.6 3.1 1.5 <0.6 o-xylene <0.6 1.6 <0.6 <0.6 nonane <0.6 <0.6 <0.6 <0.6 cis-cyclooctene <0.6 <0.6 <0.6 <0.6 isopropylbenzene <0.6 <0.6 <0.6 <0.6 cyclooctane <0.6 <0.6 <0.6 <0.6 n-propylbenzene <0.6 0.6 <0.6 0.6 3-ethyl toluene <0.6 <0.6 <0.6 <0.6 4-ethyl toluene <0.6 0.7 <0.6 <0.6 1,3,5 tri me benzene <0.6 <0.6 <0.6 <0.6 2-ethyl toluene <0.6 0.6 <0.6 <0.6 2-me nonene <0.6 1.3 <0.6 <0.6 1,2,4 tri me benzene <0.6 0.9 <0.6 <0.6 n-decane <0.6 0.7 <0.6 <0.6 1,2,3-Trimethylbenzene <0.6 <0.6 <0.6 <0.6 1,2,3,5-tetra me benzene <0.6 <0.6 <0.6 <0.6 n-butyl benzene <0.6 0.8 <0.6 <0.6 n-undecane <0.6 <0.6 <0.6 <0.6 n-pentylbenzene <0.6 <0.6 <0.6 <0.6 n-dodecane <0.6 <0.6 <0.6 <0.6 naphthalene <0.6 <0.6 <0.6 <0.6 n-tridecane <0.6 <0.6 0.6 <0.6 n-tetradecane <0.6 <0.6 0.6 <0.6 n-pentadecane <0.6 <0.6 <0.6 <0.6 n-hexadecane <0.6 <0.6 <0.6 <0.6 n-heptadecane <0.6 <0.6 <0.6 <0.6 n-octadecane <0.6 <0.6 <0.6 <0.6 n-nonadecane <0.6 <0.6 <0.6 <0.6 n-cosane <0.6 <0.6 <0.6 <0.6 n-docosane <0.6 <0.6 <0.6 <0.6


Checked by:

Annex 3. Meteorological Measurements Wind direction and speed were recorded during the DIAL measurements at one fixed location, at the north east of the Bulk Terminal site, and using a mast located on the DIAL facility itself. Two further meteorological stations were operated by TCEQ, located to the north and south of the REFINERY site (one of these stations was located at the southern side of Bulk Terminal during the period measurements were being made at Bulk Terminal). The NPL fixed mast provided wind speed and direction at two heights, ~ 3m and 11m. In general the fixed mast data have been used to derive the wind field for the DIAL flux measurements, though in some cases the DIAL wind has been used where it was observed to be more representative of local conditions. The following series of plots present the wind roses, taken from the 11m wind vane mounted on the fixed meteorological mast.

Figure 3.1 Wind rose diagram – top mast – 16 July 2007


Checked by:



Checked by:

Figure 3.5 Wind rose diagram – top mast – 20 & 21 July 2007


Checked by:



Checked by:

Figure 3.10 Wind rose diagram – top mast – 1 August 2007


Checked by:



Checked by:

Figure 3.13 Wind rose diagram – top mast – 5 & 6 August 2007


Checked by:


Cost Analysis of HRVOC Controls on Polymer Plants and

FlaresProject 2008-104

Work Order 582-07-84005-FY08-12

Prepared for: Texas Commission on Environmental Quality

Austin, Texas

Prepared by: ENVIRON International Corporation

Houston, Texas

Date: August 2008

Project Number: 06-17477L

Cost Analysis of HRVOC Controls Project 2008-104

Project Number 06-17477L i

Contents Page

Executive Summary .................................................................................................................................... 1

1 Introduction .................................................................................................................................... 3 1.1 Project Purpose................................................................................................................................ 3 1.2 Project Scope................................................................................................................................... 3 1.3 Project Methodology ........................................................................................................................ 6

2 Background Information ............................................................................................................... 7 2.1 Emissions from Polymer Production Processes.............................................................................. 7 2.2 Special Inventory of HRVOC Emissions.......................................................................................... 8 2.2.1 HECT Allowance Allocations ......................................................................................................... 10 2.3 TCEQ Guidance for Controlling Emissions from Polymer Plants .................................................. 11

3 Catalog of Potential Control Technologies ............................................................................... 12 3.1 Overview ........................................................................................................................................ 12 3.2 Process Changes........................................................................................................................... 12 3.3 Changes in Operating Procedures ................................................................................................ 12 3.3.1 Enhanced Maintenance ................................................................................................................. 13 3.3.2 Dynamic Process Simulation ......................................................................................................... 14 3.4 Vent Stream Controls..................................................................................................................... 15 3.4.1 Flaring ............................................................................................................................................ 16 3.4.2 Thermal Oxidation.......................................................................................................................... 17 3.4.3 Catalytic Oxidation ......................................................................................................................... 19 3.4.4 Boilers and Process Heaters ......................................................................................................... 20 3.4.5 Bekaert Burners ............................................................................................................................. 22 3.4.6 Stripping ......................................................................................................................................... 23 3.4.7 Adsorption / Concentration ............................................................................................................ 23 3.4.8 Biofiltration / Bioscrubbing ............................................................................................................. 25 3.4.9 Refrigerated Condensers............................................................................................................... 25 3.4.10 Non-Thermal Plasma ..................................................................................................................... 25 3.5 Flare Minimization.......................................................................................................................... 26 3.5.1 Recycling/Reuse ............................................................................................................................ 26 3.5.2 Flare Gas Recovery ....................................................................................................................... 27

4 Facility Survey Findings.............................................................................................................. 29 4.1 Polymer Plant Questionnaire Results ............................................................................................ 29 4.1.1 Process Types Surveyed ............................................................................................................... 29 4.1.2 Process Vent Control Techniques ................................................................................................. 30 4.1.3 Finishing Operations ...................................................................................................................... 31 4.1.4 HRVOC Emission Reduction Projects ........................................................................................... 31


Project Number 06-17477L ii

4.1.5 Excess Monomer Removal ............................................................................................................ 38 4.1.6 Costs of Polymer Plant HRVOC Emission Reduction Projects ..................................................... 39 4.2 Flare Questionnaire Results .......................................................................................................... 40 4.2.1 Flare Inventory ............................................................................................................................... 40 4.2.2 Flare Reduction and Minimization Projects ................................................................................... 41 4.2.3 Flare Gas Recovery ....................................................................................................................... 44 4.2.4 Costs of HRVOC Flare Reduction and Minimization Projects ....................................................... 44 4.3 Analysis.......................................................................................................................................... 46 4.3.1 Types of Emission Reduction Projects .......................................................................................... 46 4.3.2 Cost Effectiveness of Emission Reduction Projects ...................................................................... 47 4.3.3 Projects Not Undertaken................................................................................................................ 49

5 Conclusions & Recommendations ............................................................................................ 50

6 Project Team Contact Information ............................................................................................. 52


Project Number 06-17477L iii

List of Tables Page

Table 1. Facilities that Received the Flare Questionnaire ..................................................... 4

Table 2. Facilities that Received the Polymer Production Questionnaire .............................. 5

Table 3. HRVOC Special Inventory Summary (Mass) ........................................................... 9

Table 4. HRVOC Special Inventory Summary (Percentage) ................................................. 9

Table 5. Comparison of HRVOC Emissions and HECT Allowance Allocations................... 10

Table 6. Examples of Thermal Oxidation Used to Control HRVOC Emissions ................... 19

Table 7. Examples of Boilers or Process Heaters Used to Control HRVOC Emissions ...... 21

Table 8. Examples of Stripping Used to Control HRVOC Emissions................................... 23

Table 9. Physical Properties of HRVOCs ............................................................................ 24

Table 10. Summary of Process Vent Control Techniques by Process Type ........................ 30

Table 11. Summary of HRVOC Emission Reduction Projects............................................... 39

Table 12. Cost Effectiveness of HRVOC Emission Reduction Projects ................................ 39

Table 13. Summary of HRVOC Flare Reduction and Minimization Projects ......................... 41

Table 14. Cost Effectiveness of HRVOC Flare Reduction and Minimization Projects........... 45

Table 15. Summary of Project Type by Industry Sector ........................................................ 46

Table 16. Cost and Emission Reduction Summary by Plant ................................................. 47


Project Number 06-17477L iv

List of Figures Page

Figure 1. Simplified Polymer Process Flow Diagram .............................................................. 8

Figure 2. Typical Configuration – Regenerative Thermal Oxidizer........................................ 18

Figure 3. Typical Configuration – Recuperative Thermal Oxidizer ........................................ 18

Figure 4. Typical Configuration – Catalytic Oxidizer.............................................................. 20

Figure 5. Bekaert Burners ..................................................................................................... 22

Figure 6. Type and Number of Operational Polymer Process Units...................................... 30

Figure 7. Percent Average Flow and Number of Flares in HRVOC Service ......................... 41

Figure 8. Cost and Emission Reduction Summary by Plant.................................................. 48

Figure 9. HRVOC Emission Reductions as a Function of Cost for Individual Projects ......... 49


Project Number 06-17477L v

List of Appendices Appendix A: Questionnaire Templates


Project Number 06-17477L vi

List of Abbreviations BACT: Best Available Control Technology Btu: British Thermal Unit DRE: Destruction and Removal Efficiency EPA: United States Environmental Protection Agency FGR: Flare Gas Recovery FTIR: Fourier Transform Infrared HARC: Houston Advanced Research Center HDPE: High Density Polyethylene HECT: Highly Reactive Volatile Organic Compound Emissions Cap and Trade HRVOC: Highly Reactive Volatile Organic Compound IR: Infrared LAER: Lowest Achievable Emission Rate LDPE: Low Density Polyethylene LLDPE: Linear Low Density Polyethylene M: Thousand MECT: Mass Emission Cap and Trade MM: Million NNSR: Nonattainment New Source Review NSR: New Source Review PE: Polyethylene PP: Polypropylene PSD: Prevention of Significant Deterioration ppm: Parts per Million ppmw: Parts per Million by Weight RACT: Reasonably Available Control Technology RN: Regulated Entity Number SBR: Styrene-Butadiene Rubber scf: Standard Cubic Feet scfm: Standard Cubic Feet per Minute TCEQ: Texas Commission on Environmental Quality tph: Tons per Hour tpy: Tons per Year VOC: Volatile Organic Compound


Project Number 06-17477L 1

Executive Summary In June 2007, the TCEQ conducted a special emissions inventory, requesting highly reactive volatile organic compound (HRVOC) emissions data from those sources in Harris County, Texas, that are subject to HRVOC emissions cap-and-trade (HECT) program requirements. The results of the survey seemed to indicate that polymer production plants (e.g. polyethylene or polypropylene) are more likely than petroleum refineries, olefins plants, chemical plants, or independent storage terminals to have emissions that exceed their HECT allowance allocation. The TCEQ is evaluating what changes to the State Implementation Plan (SIP), if any, should be considered in light of these findings.

To support the SIP evaluation process, the TCEQ directed ENVIRON to collect and analyze information on projects undertaken to reduce emissions of HRVOC. Included in the investigation were HRVOC emission reduction projects undertaken by polymer manufacturing plants and HRVOC flaring reduction projects undertaken by polymer plants, olefins plants and petroleum refineries. To acquire the desired information, ENVIRON submitted 12 polymer plant questionnaires and 11 flare questionnaires to 16 HECT-affected sites in Harris County. All of these sites are currently subject to the requirements of the HRVOC emissions cap and trade (HECT) program.

Responding to the survey were sites that collectively included nine polymer production plants, four olefins manufacturing plants, and one petroleum refinery. These facilities provided information on 38 HRVOC emission reduction projects in three broad categories:

Changes in operating procedures:................17 projects

Vent gas Control: ..........................................7 projects

Flare minimization:........................................14 projects

Key project findings are as follows:

1. The 30 projects implemented at the nine polymer plant respondents resulted in a collective reduction in HRVOC emissions of approximately 346 tons per year at an average cost of $14,774 per ton controlled.

2. The six flare reduction projects undertaken at the four respondent olefins plants resulted in a collective reduction in HRVOC emissions of approximately 243 tons/year at an average cost of $5,295 per ton controlled.

3. Without additional information, it cannot be determined if polymer plants undertook more or fewer projects or if the cost effectiveness of those projects was higher or lower than those projects undertaken by other types of facilities.

4. Determining whether additional controls to further reduce HRVOC emissions at polymer plants would be cost effective is most properly done on a case-by-case basis. No



control technologies were identified that could be universally applied to achieve further reductions in HRVOC emissions from polymer plants in a cost-effective manner.

5. Use of flares is by far the most common method for controlling emissions of HRVOC. Most of the flares used to control routine emissions of HRVOC are designed to handle the very large flow rates that may occur during MSS and emission events – “emergency flares.” For the facilities that responded to this survey, average flows were approximately 4.4% of flare design capacity. However, the range was from less than 0.0001% to 38%.

6. While not the focus of this investigation, it was determined during the conduct of this study that polymer production units employing different processes (e.g. gas-phase vs. liquid-phase slurry) may have very different emission profiles even though they are manufacturing a similar product (e.g. polyethylene). It was also determined that making changes to the type of process used to manufacture a product may require an investment on the order of replacing the entire production unit. No facility surveyed undertook a process change to reduce HRVOC emissions.



1 Introduction

1.1 Project Purpose The purpose of this project is for ENVIRON International Corporation (ENVIRON) to support the Texas Commission on Environmental Quality’s (TCEQ) State Implementation Plan (SIP) development process by performing the following tasks:

• Identify control technologies that may be available to further reduce emissions of highly reactive volatile organic compounds (HRVOC) from polymer plants and from flares;

• Estimate potential costs for further controlling emissions of HRVOC from polymer manufacturing plants and from flares; and

• Collect cost information for measures that facilities have already undertaken to reduce HRVOC emissions.

This information may be used by the TCEQ to determine: 1) what additional control measures, if any, polymer production facilities could potentially implement to reduce emissions of HRVOC; 2) the cost effectiveness of potential additional HRVOC control measures; and 3) whether the TCEQ should consider reallocating the HRVOC emissions cap and trade (HECT) program allowances.

1.2 Project Scope The scope of work includes the following tasks:

Task 1. Develop work plan.

Task 2. Prepare separate questionnaires for flare issues and polymer production issues to be sent to selected Harris County sites.

Task 3. Develop draft interim report on list of potential control technologies for polymer processing and flare minimization.

Task 4. Develop draft interim report on evaluation of potential control technologies for polymer processing and flare minimization.

Task 5. Prepare final report.

For Task 2, ENVIRON prepared questionnaires for flare issues and polymer production issues that were sent to Harris County industrial facilities identified by the TCEQ. The questionnaire templates are included in Appendix A. The selected facilities are presented in Tables 1 (Flare Issues) and 2 (Polymer Production Issues).



Table 1. Facilities that Received the Flare Questionnaire Company Name Site Name Account RN Products

Basell USA Inc. Basell USA Bayport Plant HG0323M RN100216761 Polypropylene

Chevron Phillips Chemical Company, L.P.

Chevron Cedar Bayou Chemical Plant HG0310V RN103919817

Ethylene, propylene, polyethylene, normal alpha olefins, poly alpha olefins


Chevron Phillips Pasadena Plastics Complex HG0566H RN102018322 Polyethylene, polypropylene, styrene-

butadiene copolymer

Equistar Chemicals LP Equistar Chemicals Channelview Complex HG0033B RN100542281 Ethylene, propylene, butadiene,

benzene

Equistar Chemicals LP Equistar Chemicals La Porte Complex HG0770G RN100210319 Ethylene, propylene, polyethylene,

acetic acid, vinyl acetate monomer

Exxon Mobil Corporation Exxon Mobil Baytown Facility HG0232Q RN102579307 Fuels, refined petroleum products,

chemical feedstocks

ExxonMobil Chemical Company ExxonMobil Chemical Baytown Chemical Plant HG0229F RN102574803

Polypropylene, isobutylene, benzene, xylenes, butyl polymers, normal paraffins

Exxon Mobil Corporation ExxonMobil Chemical Baytown Olefins Plant HG0228H RN102212925 Ethylene, propylene, 1,3-butadiene

Shell Deer Park Refining Company Shell Oil Deer Park HG0659W RN100211879

Ethylene, propylene, butylenes, isoprene, 1,3-butadiene, benzene, toluene, xylene

Total Petrochemicals USA Total Petrochemicals La Porte Plant HG0036S RN100212109 Polypropylene

Total Petrochemicals USA Total Petrochemicals Bayport Plant HG4662F RN100909373 Polyethylene



Table 2. Facilities that Received the Polymer Production Questionnaire Company Name Site Name Account RN Products

Basell USA Inc. Basell USA Bayport Plant HG0323M RN100216761 Polypropylene


Chevron Cedar Bayou Chemical Plant

HG0310V RN103919817

Ethylene, propylene, polyethylene, normal alpha olefins, poly alpha olefins


Chevron Phillips Pasadena Plastics Complex

HG0566H RN102018322 Polyethylene, polypropylene, styrene-butadiene copolymer

Equistar Chemicals LP Equistar Chemicals La Porte Complex HG0770G RN100210319 Ethylene, propylene, polyethylene,

acetic acid, vinyl acetate monomer

ExxonMobil Chemical Company ExxonMobil Chemical Baytown Chemical Plant HG0229F RN102574803

Polypropylene, isobutylene, benzene, xylenes, butyl polymers, normal paraffins

Innovene Polyethylene North America

BP Solvay Polyethylene NA HG0665E RN102537289 Polypropylene and olefins

Innovene Polymers, Inc. Battleground Polyethylene Plant

HX2897U RN100229905 Polyethylene and polypropylene

Sunoco Inc. (R&M) Bayport Polyethylene Plant RN103773206 Polyethylene

Sunoco Inc. (R&M) Sunoco R&M Bayport Polypropylene

RN100524008 Polypropylene

Sunoco Inc. (R&M) Sunoco La Porte Plant HG0825G RN102888328 Polypropylene

Total Petrochemicals USA Total Petrochemicals La Porte Plant

HG0036S RN100212109 Polypropylene

Total Petrochemicals USA Total Petrochemicals Bayport Plant

HG4662F RN100909373 Polyethylene



1.3 Project Methodology

This study is conducted in two phases. The first phase is to identify technologies that have been used, or may potentially be used, to control emissions of HRVOC from polymer plants. The findings from this phase of the work are presented in Section 3 of this report. The second phase is the identification of control options, and associated costs, that have been used to reduce:

1. HRVOC emissions from Harris County polymer plants, and

2. Flaring at polymer plants, olefins manufacturing plants, and petroleum refineries in Harris County.

The findings from this phase are presented in Section 4 of this report.



2 Background Information

2.1 Emissions from Polymer Production Processes HRVOC compounds are used extensively in the manufacture of polymers.1 Examples include:

• Use of ethylene in the manufacture of polyethylene (PE). PE is a thermoplastic (becomes soft when heated and hard when cooled) that is heavily used in consumer products (e.g. plastic shopping bags). PE is classified into a large number of categories based on its density and branching. Examples include high-density PE (HDPE), low-density PE (LDPE), and linear low-density PE (LLDPE).

• Use of propylene in the manufacture of polypropylene (PP). Like PE, PP is a thermoplastic that finds wide use in a variety of applications.

• Use of 1,3-butadiene in the manufacture of synthetic rubbers such as polybutadiene and styrene butadiene rubber (SBR). SBR is widely used in the manufacture of automobile tires.

• Use of butenes in the manufacture of synthetic rubbers such as polyisobutylene and as copolymers. A copolymer is a polymer derived from two or more monomers. An example is SBR which is derived from styrene and 1,3-butadiene.

Figure 1 presents a simplified process flow diagram for manufacturing polymer pellets.2 For polymer products other than pellets, downstream operations may vary. For example, when manufacturing a polymer flake or powder, there will not be an extruder. Similarly, SBR is typically not extruded, but sold as SBR crumb.

Sources of HRVOC from a polymer manufacturing process may include one or more of the following:3

• Monomer or comonomer storage

• Process fugitives

• Cooling tower heat exchange system losses

1 A polymer is a large molecule composed of repeating structural units. Examples include polyethylene and polypropylene. 2 Technical Guidance for Chemical Sources: Polyethylene and Polypropylene Manufacturing, Draft RG-244, TCEQ Air Permits Division, February 2001. 3 Ibid



• Process vents upstream of the extruder (e.g. reactor, resin degassing)

• Extruder

• Polymer storage and loading

• Wastewater treatment facilities

Figure 1. Simplified Polymer Process Flow Diagram

2.2 Special Inventory of HRVOC Emissions In June 2007, the TCEQ conducted a special emissions inventory, requesting HRVOC emissions data from those sources in Harris County, Texas, that are subject to HRVOC emissions cap-and-trade (HECT) program requirements. The reporting period for this special inventory was February 1, 2006, through January 31, 2007.

Special inventory responses were categorized based upon the primary activity at the site:

• Chemical manufacturing (non-olefin, non-polymer)

• Olefins manufacturing4

• Polymer manufacturing

• Petroleum Refining

• Independent storage terminals (not dedicated to an individual refinery, olefins, chemical or polymer manufacturing site)

4 An “olefin,” or alkene, is an unsaturated chemical compound containing at least one carbon-carbon double bond. The simplest olefin is ethylene which has the following chemical structure: H2C=CH2.



HRVOC special inventory responses are summarized in Tables 3 and 4. The values shown in Table 3 are total emissions for each industry sector. Emissions for the five industry sectors combined are shown in the far right column. The percentages presented in Table 4 are for total emissions reported by that industry sector. Percentages for the five combined sectors are presented in the far right column. Values presented in Tables 3 and 4 are taken from summaries provided by TCEQ personnel. ENVIRON has not reviewed the special inventory submittals nor validated the values provided.

Table 3. HRVOC Special Inventory Summary (Mass) Emissions by Industry Sector (tons)

Source Chemical2 Olefins Polymers Refining Terminal2 Combined2

Flares 278.3 376.7 460.0 308.7 45.8 1,469.5Cooling Towers 35.4 18.1 20.9 88.1 0.1 162.6Other Vents 97.5 157.1 221.1 125.0 1.7 602.4Fugitives 19.6 115.5 22.0 26.0 0 183.1Total2,3 443.2 667.4 724.1 547.8 50.9 2,433.4

Type MSS & Events 34.8 124.8 81.0 83.2 0.0 323.8Uncontrolled1 136.4 245.3 234.0 135.9 1.8 753.4Controlled1 259.5 297.3 409.0 328.8 45.9 1,340.51 Uncontrolled and controlled routine emissions. MSS and event emissions are accounted for separately. 2 Total includes emissions from sites that were not broken down by source or type of emissions. 3 Total includes fugitive emissions from equipments leaks which are not subject to HECT.

Table 4. HRVOC Special Inventory Summary (Percentage) Emissions by Industry Sector (%)

Source Chemical1 Olefins Polymers Refining Terminals1 Combined

Flares 64.6 56.4 63.5 56.3 96.3 60.8Cooling Towers 8.2 2.7 2.9 16.1 0.1 6.7Other Vents 22.6 23.5 30.5 22.8 3.6 24.9Fugitives 4.5 17.3 3.0 4.7 0.0 7.6

Type MSS & Events 8.1 18.7 11.2 15.2 0.0 13.4Uncontrolled2 31.7 36.8 32.3 24.8 3.8 31.2Controlled2 60.2 44.5 56.5 60.0 96.6 55.41 Emissions (%) were determined using 430.8 tons as the total emissions for the Chemical sector and 47.6 tons as the total emissions for the Terminals sectors.

2 Uncontrolled and controlled routine emissions. MSS and event emissions are accounted for separately.



As shown in Table 4, the information contained within the special inventory submittals indicate that approximately 61% of total reported HRVOC emissions are controlled using flares. This includes MSS and events that are controlled by flare. The 55.4% noted as “controlled” does not include MSS and events. Therefore, approximately 64% of routine emissions are controlled.

2.2.1 HECT Allowance Allocations Under the HECT, allowances are allocated by the TCEQ to affected sites in Harris County according to the procedures defined by rule in 30 TAC §101.394. The initial allocation occurred January 1, 2007, with subsequent allocations occurring January 1 of each year thereafter. Covered facilities at these sites include flares, cooling tower heat exchange systems and vent gas streams. Fugitive emissions are not covered by the HECT.

On August 18, 2006, the TCEQ published a list of HECT allowance allocations. A total of 51 sites in Harris County were allocated 3,451.5 tons of HRVOC. Table 5 presents a comparison of HRVOC emissions reported as part of the special inventory, by industry sector, with the HECT allowance allocation for that sector. The allowance allocation shown is only for those facilities that reported emissions as part of the special HRVOC emissions inventory. Since all facilities with HECT allowance allocations did not respond to the special inventory request, the summation of allowance allocations does not equal the total number of allowances allocated (3,451.5. tons) but does account for over 98% of the allocations.

There are several sites that could be included in more than one industry sector. In those cases, the site is placed into the industry sector that seems to best represent their primary business. For example, a petroleum refinery with a collocated chemical plant will be included in the Refining Industry sector.

Table 5. Comparison of HRVOC Emissions and HECT Allowance Allocations

Industry Sector HRVOC Emissions (tons)1

Annual HECT Allowance Allocation (tons)

Emissions as a % of Allowance Allocation

Chemical 411.2 718.6 57.2Olefins 551.9 1,123.8 49.1Polymers 702.0 496.1 141.5Refining 521.8 996.7 52.4Terminals 47.6 57.0 83.5Combined 2,234.5 3,392.2 65.91 Total emissions from emission points covered by the HECT: flares, cooling towers and other vents. Fugitive emissions are not covered by the HECT. Does not included uncharacterized emissions from Chemical and Terminals sectors.

As shown in Table 5, during the period covered by the special inventory, polymer production plants were more likely than petroleum refineries, olefins plants, chemical plants, or independent storage terminals to have emissions that exceed their HECT allowance allocation.



2.3 TCEQ Guidance for Controlling Emissions from Polymer Plants Prior to developing a catalog of potential control strategies that may be available for further reducing emissions of HRVOC from polymer plants, ENVIRON first reviewed previous TCEQ guidance on the subject. TCEQ guidance issued in February 2001 by the Air Permits Division provides the following regarding control of emissions from PE and PP manufacturing facilities undergoing New Source Review (NSR) permitting.5

• Best Available Control Technology (BACT) for all types of processes requires control of all waste gas streams upstream of the extruder.

• Control devices specified by guidance are as follows:

o For vent streams, combustion using a flare, incinerator, boiler, heater, etc.

o For fugitive emissions, a 28 VHP fugitive monitoring program.6

• Maximum allowable residual volatile organic compound (VOC) – which for PE and PP manufacturing facilities would be all HRVOC – in the polymer at the first uncontrolled vent should be less than 90 parts per million by weight (ppmw) for all manufacturing processes with the exception of high-pressure polyethylene manufacturing processes where guidance states a limit of 100 ppmw.

• Total non-fugitive VOC emissions, including HRVOC emissions, should generally be less than 200 pounds per million pounds (MM lb) of product.

The most recent published TCEQ guidance on BACT for PE and PP production facilities (October 17, 2006), is more restrictive than the 2001 technical guidance document. Specifically, it requires that total non-fugitive, uncontrolled VOC (including HRVOC) emissions are to be reduced to less than 80 pounds per MM lb of PE or PP produced.7

It should be noted that HRVOC fugitive monitoring requirements for affected facilities under 30 TAC Subpart H, Division 3, is more stringent than the requirements of 28 VHP.

5 Technical Guidance for Chemical Sources: Polyethylene and Polypropylene Manufacturing, Draft RG-244. 6 The revised requirements of 28 VHP are specified in TCEQ guidance issued in May 2008: http://www.tceq.state.tx.us/assets/public/permitting/air/Guidance/NewSourceReview/rev28vhp_508.pdf 7 TCEQ Chemical Sources, Current Best Available Control Technology (BACT) Requirements, Polyethylene and Polypropylene Facilities, October 2006. (http://www.tceq.state.tx.us/assets/public/permitting/air/Guidance/NewSourceReview/bact/bact_polys.pdf)



3 Catalog of Potential Control Technologies

3.1 Overview Methods identified for potentially reducing HRVOC emissions include:

• Process changes,

• Changes in operating procedures,

• Vent stream controls, and

• Flare minimization.

Each of these is discussed in turn within this section.

3.2 Process Changes For existing production facilities, short of replacing older technology with newer technology, there are limited opportunities for making process changes that reduce HRVOC emissions. One of these limited opportunities discussed in a 1997 EPA document on the polymer industry is changing catalysts.8 This document suggests that there are opportunities to replace an older catalyst with a newer, better catalyst, resulting in overall yield improvements, and, thus, reducing the amount of un-reacted monomer that remains in the polymer. This information is somewhat misleading. A polymer manufacturing facility must use a catalyst that is most appropriate for their process and the polymer properties sought.

For example, a gas phase PE reactor may use an older catalyst that has lower reactivity and selectivity than a newer, better catalyst. However, the newer, better catalyst may not be suitable for use in the gas phase reactor, but can only be used in a high pressure slurry PE process. Additionally, use of the old catalyst may be necessary to produce the desired PE properties.

3.3 Changes in Operating Procedures As with process changes, there may be opportunities to modify operating procedures to reduce emissions of HRVOC. Potential changes in operating procedures may range from enhanced maintenance activities to the use of sophisticated dynamic simulation algorithms to reduce losses during non-steady state operating conditions, such as those that occur during startup and shutdown.

8 Profile of the Plastic Resin and Manmade Fiber Industries, Sector Notebook Project, EPA-310/R-97-006, September 1997.



3.3.1 Enhanced Maintenance There may be opportunities to improve maintenance of existing equipment and reduce losses of HRVOC during normal operations and/or scheduled maintenance activities. Examples of enhanced maintenance activities may include more aggressive investigation and timely corrective action of leaking pressure safety valves (PSV), leaking compressor seals, leaking valves and flanges, leaking heat exchangers, etc. However, sites subject to the HECT are already required to monitor cooling tower return lines for leaking heat exchange systems and implement stringent fugitive monitoring and control requirements.9 [As noted previously, fugitive emissions are not included in the HECT.]

Enhanced maintenance may also include use of predictive and preventive maintenance processes. The predictive maintenance process involves review of the equipment types, the failure mechanisms associated with those equipment types and the associated mean time to failure. Preventive maintenance takes the next step and focuses maintenance on taking action before equipment fails. This preventive maintenance can lead to higher reliability, greater on-stream time, and fewer unscheduled maintenance shutdowns.

Required HRVOC monitoring of flare headers and cooling tower returns has resulted in some subject facilities implementing enhanced maintenance programs.10 For example, one site uses the HRVOC monitoring system as a feedback mechanism. When flaring, operators use the HRVOC monitoring system to help identify the source of the flows (e.g., PSV leaks, open valves). The monitors do not identify specific equipment, but the ability to speciate the compounds going to the flare helps to narrow troubleshooting efforts to a particular process area.

Some sites are also using passive infrared (IR) cameras to find and eliminate/reduce emissions of HRVOC. As part of the HARC H-76 project, ENVIRON found that, as of the summer of 2006, three of the nine surveyed sites were using the IR cameras to locate potential sources of HRVOC emissions.11 Uses of the IR cameras include:

• Integration of IR cameras into routine leak detection and repair (LDAR) programs.

• Use of IR cameras to monitor for emissions during startup and shutdown. Camera findings are used to direct corrective measures.

9 Any process unit or process within a petroleum refinery, synthetic organic chemical, polymer, resin, or methyl tert-butyl ether manufacturing process or natural gas/gasoline processing operation in the 8-county Houston/Galveston/ Brazoria area in which an HRVOC is a raw material, intermediate, final product, or in a waste stream is subject to the requirements of 30 TAC 115, Subchapter H, Division 3 which specifies monitoring and control requirements for fugitive emissions from equipment. 10 “How Chemical Manufacturing and Petroleum Refining Facilities in Harris County are Using Point Source Monitoring to Identify and Reduce HRVOC Emissions,” HARC Project H76, ENVIRON, prepared for the Houston Advanced Research Center, October 2006. 11 Ibid



As discussed in the Project H-76 report, those companies that have embraced use of the passive IR cameras are strong supporters of using this technology to find and fix sources of emissions.

As part of an agreement with the City of Houston, one facility in Harris County has installed a fence line Fourier Transform Infrared (FTIR) monitoring system to monitor for 1,3-butadiene.12 While not initially intended as a tool for improving maintenance or operating practices, the fence line FTIR system has allowed the facility to identify operations and activities that result in emissions of 1,3-butadiene and to use that information in taking corrective action.

3.3.2 Dynamic Process Simulation Understanding how operating conditions affect waste gas production can lead to improved operating techniques. Rather than relying solely on engineering trial-and-error and operator experience to modify operating procedures to minimize flaring, dynamic process simulation has been used to minimize HRVOC flaring during shutdown and startup at an ethylene production facility.13 In the referenced study performed by Lamar University in cooperation with LyondellBasell’s Equistar Channelview Plant, dynamic process simulation was used to critically evaluate potential process and procedural modifications prior to the actual shutdown/startup or upset event. Dynamic simulation was developed for the recovery section of the ethylene plant and used to examine the following process steps:

• Approaching shutdown,

• Startup with recycle ethane, and

• Starting the cracked feed and increasing the feed to normal production rates.

The researchers found that dynamic process simulation provides an insight into process behavior that is not readily apparent through steady state simulation and process engineering calculations. Process simulations are performed using standard software, such as Aspen Plus and Aspen Dynamics™. Operators can use the results of the dynamic process simulations to modify control settings during shutdown/startup and upset conditions to minimize flaring of off-specification (off-spec) streams.

Results of the Equistar Channelview Plant dynamic process simulation study are as follows:14

• Actual flaring associated with shutdown and startup of the ethylene plant was 75% less than a previous startup of a similar plant at the site.

12 Ibid 13 Flare Minimization via Dynamic Simulation. Singh, A., K. Li, H. H. Lou, J. R. Hopper, H. B. Golwala and S. Ghumare. Int. J. Environment and Pollution. Vol. 29, Nos. 1/2/3, pp. 19-29. 14 Ibid



• Flaring emissions from the shutdown and startup were 56% less than the estimates made prior to the turnaround.

Using dynamic process simulation, other ethylene production facilities have reduced HRVOC emissions from flaring. Examples include:

• Huntsman Petrochemical reduced flaring to less than 3.5 hours during a startup event, and

• BASF-TOTAL reduced flaring by 50% compared to a previous startup.15

This project has not determined if dynamic process simulation can be applied to polymer manufacturing plants.

3.4 Vent Stream Controls Based on our review of EPA’s RACT/BACT/LAER Clearinghouse (RBLC), ENVIRON identifies three technologies that have been used to control HRVOC emissions from process vents at polymer manufacturing plants: flares, thermal oxidizers and boilers. Additionally, while not an ultimate control device, air and steam stripping have been used to remove un-reacted monomer from SBR crumb prior to finishing. While not identified in the RBLC review, ENVIRON is aware that polymer plant waste gas streams containing HRVOC have also been managed using catalytic oxidizers. Each of these technologies is briefly described within this section. Also included are brief discussions of other control technologies considered: adsorption, biofiltration, Bekaert burners and refrigerated condensers.

It is important to keep in mind that the technical feasibility, including process safety considerations, and economic feasibility of any control system can only be determined on a case-by-case basis.

Costs are not included within these general discussions of control technologies. Costs of control are highly dependent upon a number of design and operating variables and often cannot be estimated accurately within even one or two orders of magnitude without specific information as to the application. For example, USEPA Air Pollution Control Technology Fact Sheet EPA-452/F-03-019 provides the following range of costs for flares.

Capital Cost: .....................................$13 to $21,000 per standard cubic foot per minute (scfm) of flow

Operation & Maintenance Cost:........$1 to $10 per scfm

Annualized Cost:...............................$3 to $300 per scfm 15 Near-Zero Flare for Chemical Process Industry via Plant-wide Optimization and Simulation. Xu, Q., K. Li and J. L. Gossage. TERC SAC Meeting, Houston, Texas. February 2008.



Cost Effectiveness: ...........................$15 to $5,000 cost per ton of pollutant controlled

It is questionable if even this very broad range brackets actual costs that might be incurred for flaring of one or more process vent streams. Meaningful control costs can only be determined on a case-by-case basis.

3.4.1 Flaring Flaring is a combustion control process in which the combustible material to be flared is piped to a remote, usually elevated location, and burned in an open flame in the air using a specially designed burner tip, auxiliary fuel, and steam or air to promote mixing for nearly complete (>98%) destruction efficiency. Completeness of combustion in a flare is governed by flame temperature, residence time in the combustion zone, turbulent mixing of the gas stream components to complete the oxidation reaction, and available oxygen for free radical formation.16

Flares that conform to the design requirements of 40 CFR 60.18 are assumed by rule to achieve 98% destruction efficiency for C4 HRVOCs (1,3-butadiene and butenes) and 99% destruction efficiency for C2-C3 HRVOCs (ethylene and propylene).17 The design requirements of 40 CFR 60.18 include the following:

• Flame present at all times.

• Minimum net heating value of the gas being burned of 300 Btu/scf (steam or air-assisted) or 200 Btu/scf (non-assisted).

• Maximum exit velocity of 60 feet per second for steam-assisted or non-assisted flares. Velocity limits for air-assisted flares are dependent upon the net heating value of the gas being combusted.

Flares are commonly used to control HRVOC emissions at petroleum refineries, chemical plants, olefins manufacturing plants, polymer manufacturing plants and for-hire storage terminals. Review of the RBLC identified use of flaring to control emissions from product storage and product loading at the Chevron Phillips Chemical Company Pasadena Plastics Plant. The permit for this application was dated December 15, 1998.

Flaring of HRVOC emissions from extruders and finishing operation vent streams has been limited for a number of reasons, including: cost of capture (e.g. installation of hooding on extruders, etc.), cost of supplemental fuel, and safety (air in flare header resulting in a potentially explosive waste stream). To illustrate the cost of supplement fuel, assume a 1,000 standard cubic feet per minute vent stream containing 200 ppm ethylene – approximately 4

16 USEPA Air Pollution Control Technology Fact Sheet EPA-452/F-03-019 17 30 TAC 115.725(d) contains several references to these control efficiencies.



tons/year – is routed to a flare for control. The heat value of the ethylene at this concentration is approximately 0.3 Btu/scf. To meet the 40 CFR 60.18 design requirement of 300 Btu/scf for a steam-assisted flare, this vent stream will either need to be combined with a higher heat content vent stream prior to flaring or supplemented with a fuel, such as natural gas. If a supplementary fuel is used, approximately 429 scfm of natural gas will need to be added to the vent stream to raise the heat content to 300 Btu/scf. On an annual basis, this equates to approximately 225.5 million scf (MMscf). Assuming a natural gas price of $9.00 per MMBtu, supplemental fuel for flaring this 1,000 scfm vent gas stream will cost approximately $2,000,000 per year. The incremental cost of control for this vent stream using a flare, just considering the cost of the supplemental fuel, is approximately $500,000 per ton. This cost of control would apply to any 200 ppm ethylene vent stream that is flared. Additionally, the use of supplemental fuel would result in additional CO and NOx emissions from the flare.

3.4.2 Thermal Oxidation There are two general types of thermal oxidizers (TO) in common use: regenerative and recuperative. A regenerative thermal oxidizer, or RTO, uses a high-density media such as a ceramic packed bed still hot from a previous cycle to preheat an incoming waste gas stream. The preheated waste gases then enter a combustion chamber where they are heated by auxiliary fuel combustion (e.g. natural gas) to a final oxidation temperature typically between 1,400 and 1,500°F. The hot exit gases are directed to one or more ceramic packed beds where the heat from the gases is absorbed before they are vented to the atmosphere. An RTO will typically achieve a control efficiency of 95 to 99%.18

Recuperative TOs are comprised of the combustion chamber, waste gas preheater and, if appropriate, a secondary energy recovery preheater. Recuperative TOs can recover up to 70% of the waste heat from the exhaust gases and achieve destruction efficiencies ranging from 98% to as high as 99.9999%.19

The typical design conditions required to achieve at least 98% destruction efficiency in a recuperative TO are:

• Minimum combustion temperature of 1,600°F,

• Combustion chamber residence time of 0.75 second, and

• Proper mixing.

18 USEPA Air Pollution Control Technology Fact Sheet EPA-452/F-03-021 19 USEPA Air Pollution Control Technology Fact Sheet EPA-452/F-03-020



Figures 2 and 3 present typical configurations for regenerative and recuperative TOs, respectively.20

Figure 2. Typical Configuration – Regenerative Thermal Oxidizer

Figure 3. Typical Configuration – Recuperative Thermal Oxidizer

Table 6 summarizes the findings of the RBLC review (last 10 years) with respect to control of HRVOC emissions from polymer plants using thermal oxidation. This listing is not comprehensive. Sources that did not undergo federal NSR, either Prevention of Significant Deterioration (PSD) or Nonattainment NSR (NNSR), will not be listed in the RBLC.

20 USEPA Air Pollution Cost Control Manual, 6th Edition, EPA-452/B-02-001, January 2002



Table 6. Examples of Thermal Oxidation Used to Control HRVOC Emissions Company: Fagerdala Pac-Lite Incorporated Location: St. Clair, Michigan Permit Date: 02/01/2001 Process Description: Expandable polypropylene bead production. Control Application: Emissions from the fluidized bead dryer and regrind extruder are

controlled by thermal oxidizer. Die area is hooded. Control Efficiency: 85% Company: Formosa Plastics Corporation Location: Point Comfort, Texas Permit Date: 03/09/1999 Process Description: Polypropylene plant with multiple trains with two reactors each. Control Application: Process off-gases are routed to incinerator. Reactor gases are

routed to flare header in case of an upset. Control Efficiency: Unknown. HRVOC emissions are limited to 31.5 lb/MMlb for Train 4,

133 lb/MMlb for Trains 1-3. Company: Total Petrochemicals USA (formerly Atofina Petrochemicals Inc.) Location: La Porte, Texas Permit Date: 11/05/2001 Process Description: Polypropylene production Control Application: Backup thermal oxidizer. Areas of process controlled are not

identified. Control Efficiency: 99.99% Company: Goodyear Tire and Rubber Company Location: Beaumont, Texas Permit Date: 02/19/2004 Process Description: Styrene butadiene rubber production Control Application: Regenerative thermal oxidizer. Application is unspecified. Control Efficiency: Not specified

3.4.3 Catalytic Oxidation Catalytic oxidizers operate similar to thermal oxidizers with the primary difference being that, after passing through the flame zone, the waste gases pass through a catalyst bed. The catalyst has the effect of increasing the reaction rate, enabling oxidation of the organics in the waste stream at a lower reaction temperature than would be required in a thermal oxidizer to achieve the same destruction efficiency. Catalysts also allow for a reduced residence time and, thus, allow for smaller oxidizers. The waste gas is typically heated to between 600°F and 800°F



Waste Gas Pre-heater

Combustion Chamber

Catalyst Chamber

Stack

Waste Gas

AuxiliaryFuel

Pre-Heated Waste Gas

before entering the catalyst. Control efficiencies as high as 95-99% can be achieved with catalytic oxidation. 21

Catalytic oxidizers are subject to plugging as well as catalyst deactivation or poisoning. Therefore, they are not as widely applicable as thermal oxidizers. As with any control approach, however, technical and economic feasibility can only be evaluated on a case-by-case basis.

Figure 4 presents a typical configuration for a catalytic oxidizer. 22

Figure 4. Typical Configuration –Catalytic Oxidizer

Review of the RBLC did not identify any application of catalytic oxidizers to control emissions of HRVOC from polymer plants in the last 10 years. However, as noted previously, this listing is not comprehensive. As discussed in Section 4 of this report, catalytic oxidizers have been used at sites in Harris County to control emissions of HRVOC.

3.4.4 Boilers and Process Heaters Boilers and process heaters, under certain process conditions, may be used to combust waste streams containing HRVOC. Considerations in burning HRVOC waste gas streams in boilers and process heaters include the following: 23

• Most chemical plants, olefins manufacturing plants and polymer plants do not have fuel gas headers that facilitate collection of waste gases for use as fuels. This may limit or eliminate consideration of this control option.

• Boilers designed specifically for HRVOC control use discrete or vortex burners.24

21 USEPA Air Pollution Control Technology Fact Sheet EPA-452/F-03-021 22 EPA Air Pollution Cost Manual, 6th Edition, EPA-452/B-02-001, January 2002 23 Some of the items listed are derived from discussions with industry personnel. 24 Polymer Manufacturing Industry, Background Information for Proposed Standards, EPA-450/3-83-019a, September



• Use of high concentration ethylene streams as a fuel tends to result in more nitrogen oxides (NOX) formation than the burning of natural gas.25

• The combustion characteristics of certain olefin derivatives, such as propylene oxide, make it unsuitable for combustion in boilers or process heaters because it combusts explosively. This characteristic could also limit the use of thermal or catalytic oxidizers.

• Olefins such as ethylene and propylene may polymerize in a fuel gas header, resulting in plugged lines with associated safety and performance implications.

While not common, there are examples of boilers being used to control emissions from polymer plants. In a 1985 document, EPA identifies two polypropylene plants and a high density polyethylene (HDPE) plant that route off-gases to a boiler for control. 26

In addition, ENVIRON identified one facility during review of the RBLC (last 10 years) that was permitted to use a boiler to control emissions of HRVOC from a polymer production facility (Table 7). Note that this listing is not comprehensive. Sources that did not undergo federal NSR, either PSD or NNSR, will not be listed in the RBLC.

Table 7. Examples of Boilers or Process Heaters Used to Control HRVOC Emissions

Company: Total Petrochemicals USA (formerly Atofina Petrochemicals Inc.) Location: La Porte, Texas Permit Date: 11/05/2001 Process Description: Polypropylene production Control Application: Waste heat boiler and regenerative gas heater. Areas of process

controlled are not identified. Control Efficiency: 99.99%

In addition to the application identified in Table 7, within the description for a project at the Chevron Phillips Chemical Company Pasadena, Texas, plant (permit date of 02/23/2000) is the following:

“The Phillips Chemical Company seeks authorization to use certain PE and PP process off gases as fuel at existing flares located within their Houston Chemical Complex. The process off gases are generated on-site at process units and were used as fuel at four on-site boilers;

1985. 25 Ethylene has a high heat content: approximately 1,600 Btu/scf compared with approximately 1,000 Btu/scf for methane (natural gas). Consequently, unless specifically designed for burning ethylene, the combustion device will burn hotter and have higher NOX emissions. 26 EPA-450/3-83-019a



however, the boilers are being permanently shutdown . . . Certain PE/PP off gases will be routed to the flare fuel gas system for use as flare fuel gas.”

It is ENVIRON’s understanding that as part of their strategy for complying with the NOX Mass Emission Cap and Trade (MECT) program, Chevron Phillips Chemical shut down their boilers and started purchasing steam from a nearby cogeneration facility. As documented in the RBLC, however, boilers were used for controlling PE and PP emissions prior to that time.

If technically and economically feasible, due to the high temperature and long residence times typical of boilers and process heaters, high destruction efficiencies (greater than 98%) can be achieved.27

The use of flares, thermal oxidizers, catalytic oxidizers and/or boilers to reduce HRVOC emissions from polymer plants is built upon the assumption that the uncontrolled emissions can be effectively captured. The effective, efficient and safe capture of reactive monomer streams must be evaluated on a case-by-case basis.

3.4.5 Bekaert Burners Another alternative to flaring is use of Bekaert CEB® burners28. Bekaert burners use a meshed fiber surface that divides the main flame into tiny flames thereby resulting in more complete combustion of the HRVOC in the waste gas stream. Bekaert reports that HRVOC destruction efficiencies as high as 99.99% have been achieved at operating temperatures of 2,000-2,200°F with NOX emissions less than 15 ppm. The capacity on the largest burner model Bekaert currently makes is approximately 2,500 scfm. As discussed in Section 4, with one exception, the annual average flare flowrates at the Harris County polymer production facilities surveyed is less than 2,500 scfm. However, to handle emergencies and other large releases, the flare design capacities are much greater than 2,500 scfm.

Based on information provided by Bekaert, there are six of their systems currently in use in the Houston area. Four of these are at petrochemical plants. The largest system in use is used to control emissions (non-HRVOC) from barge loading and unloading operations. Since the burner uses fine fiber mesh, presence 27 Ibid 28 Information on Bekaert burners is derived from discussions with Mr. Timothy F. Egan, Bekaert Corporation, and from their website: http://www.bekaert.com/flaring

Figure 5. Bekaert Burners



of particulate matter in the waste gas stream may clog the flame openings thereby impairing performance.

Figure 5 shows in-field installation of two Bekaert CEB® Model 4500 burners (Source: Bekaert website). Currently, these are Bekaert’s largest models.

3.4.6 Stripping Stripping is used in SBR production to remove un-reacted monomer from the rubber crumb, with the overheads routed to a combustion device for destruction. Stripping is common in SBR production facilities; however, the intent of the stripping is primarily to remove the un-reacted styrene. The un-reacted 1,3-butadiene is removed through flash distillation and reduction in system pressure.

Review of the RBLC identified two facilities where stripping is used to control emissions from SBR production facilities. Table 8 summarizes the findings.

Table 8. Examples of Stripping Used to Control HRVOC Emissions Company: Firestone Polymers LLC Location: Lake Charles, Louisiana Permit Date: 07/30/2003 Process Description: Styrene butadiene rubber production Control Application: Steam stripping of solvent from crumb rubber. Emissions stream from

stripping operation is collected and routed to flare Control Efficiency: Unknown Company: Goodyear Tire and Rubber Company Location: Beaumont, Texas Permit Date: 02/19/2004 Process Description: Styrene butadiene rubber production Control Application: Air stripping. Application is unspecified. Control Efficiency: Not specified

While search of the RBLC did not identify any instances of either air or steam stripping being used to control emissions from PE or PP production facilities, as discussed in Section 4, nitrogen stripping is used to remove un-reacted monomer in PE and PP production facilities.

3.4.7 Adsorption / Concentration Adsorption is the attachment of gaseous molecules to the surface of a solid. During adsorption, a gas molecule migrates from the gas stream to the surface of the solid where it is held by physical attraction. Adsorption in the form of a concentrator can be used to raise the concentration of an organic vapor to provide more economical treatment in downstream



combustion or condensation devices. Activated carbon is the most widely used adsorbent for VOCs. Other adsorbents include zeolites and certain synthetic polymers. 29

Factors that influence the performance of activated carbon in controlling gas phase VOC emissions include:30

• The type of compound to be removed. In general, compounds with a high molecular weight and higher boiling point are better adsorbed.

• Concentration. The higher the concentration the better the adsorption.

• Temperature. The lower the temperature the greater the adsorption capacity.

• Pressure. The higher the pressure the greater the adsorption capacity.

• Humidity. The lower the humidity the greater the adsorption capacity.

As shown in Table 9, HRVOCs are low molecular weight compounds with low boiling points.31

Table 9. Physical Properties of HRVOCs Compound Molecular Weight Boiling Point (°C)

Ethylene 28 -104

Propylene 42 -48

1,3-Butadiene 54 -4

1-Butene 56 -5

2-Butene (cis & trans) 56 +3

Isobutylene 56 -7

Vents from extruders and downstream operations will typically have low HRVOC concentrations and pressures close to ambient. Collectively, this information indicates that activated carbon would, most likely, be a poor choice for either direct control of HRVOC emissions or for use in a concentrator. The limitations affecting adsorption using activated carbon would also be expected to affect adsorption using zeolites or polymer adsorbents.

Concentrator suppliers have stated that, for a concentrator to be effective, the boiling point of the material to be adsorbed should be higher than the inlet gas phase temperature, but lower

29 Choosing an Adsorption System for VOC: Carbon, Zeolite, or Polymers? EPA Technical Bulletin, EPA 456/F-99-004, May 1999. 30 EPA Air Pollution Cost Manual. 6th Edition, EPA/452/B-02-001, January 2002. 31 Chemical Engineer’s Handbook, 5th Edition, edited by Robert H. Perry and Cecil H. Chilton, McGraw-Hill Book Company, 1973.



than the desorption temperature. For applications involving the HRVOCs listed in Table 9, the boiling points would be less than the desorption temperature; however, they would not be above the inlet temperature. This “rule of thumb” confirms that HRVOCs are not amenable to adsorption or concentration.

Review of the RBLC did not identify any application of adsorption for the control of HRVOC emissions from polymer plants in the last 10 years.

3.4.8 Biofiltration / Bioscrubbing Biofiltration involves routing a vent gas stream through a biologically active media, similar to compost, where the pollutants of interest are adsorbed and/or absorbed and biologically degraded into water and carbon dioxide. A bioscrubber is similar in function; however, the control system involves a tower packed with synthetic media that supports a biological culture. Biofiltration and bioscrubbing have been successfully applied in a number of full-scale applications to control odors, VOC and emissions of air toxics from a wide range of sources. Application of biofiltration and bioscrubbing are typically limited to relatively low concentration vent streams – approximately 1,000 ppm or less – and the pollutants need to be water soluble.32 HRVOC compounds are generally slightly soluble to insoluble in water, making them poor candidates for control through biofiltration.

3.4.9 Refrigerated Condensers A refrigerated condenser is a control device that is used to cool an emission stream containing organic vapors and to condense the organic vapors into a liquid that is then collected and either recycled or disposed of. Refrigerated condensers work best on emission streams containing high concentrations of VOC. To achieve any reduction, even on saturated streams, the condenser must achieve a temperature that is lower than the boiling point of the compound in question.

HRVOC emissions from extruders and downstream operations are poor candidates for control through use of refrigerated condensers because: 1) the boiling points are low (refer to Table 9) and 2) the concentration of HRVOC in the vents is expected to be low.

3.4.10 Non-Thermal Plasma Low-temperature, non-thermal plasma (NTP) is a developing technology that may, in the future, be an option for controlling emissions of HRVOC.33 The basic principle of NTP is the use of

32 EPA Survey of Control Technologies for Low Concentration Organic Vapor Gas Streams. EPA/456/R-95-003, May 1995. 33 Pulsed Corona Plasma Pilot Plant for VOC Abatement in Industrial Streams as Mobile and Educational Laboratory.

Tak, G., Gutsol, A. and Fridman, A., 2005. (http://plasma.mem.drexel.edu/publications/documents/ISPC-ID670.pdf)



electricity to create a plasma – an ionized gas containing free electrons.34 These energetic electrons excite, dissociate and ionize molecules to produce chemically active radicals and ions. In the laboratory, NTP has been shown to have ability to destroy a number of different VOCs and polycyclic aromatic hydrocarbons (PAHs).35 Pilot-scale experiments conducted at pulp mills and wood product plants have shown VOC destruction efficiencies greater than 98%.36 In theory, NTP could be used to treat industrial waste gas streams containing HRVOC across a wide range of flow rates.37 NTP, however, has not been demonstrated on a commercial scale nor has it been demonstrated to control emissions of HRVOC.

3.5 Flare Minimization Flare minimization refers to the reduction in the number of instances of flaring, both during routine operations and during startup, shutdown and malfunction, and to the reduction in the quantity of material flared. The concept of flare minimization applies to both stream recycling/reuse and flare gas recovery.

3.5.1 Recycling/Reuse As discussed in Section 2.2, a significant percentage of HRVOC emissions occur during maintenance, startup, and shutdown (MSS) activities and during emission events. Through certain capital investments (e.g. the addition of process loops and storage capacity) and changes in the way that the production units are managed during MSS activities, the amount of HRVOC released to the flare header can be significantly reduced. Following are two examples of recycling and reuse at Harris County facilities.

• One Harris County olefins producer implemented a flare minimization program that resulted in reduced flaring during the shutdown and startup of the unit. The shutdown and startup of the unit is a sequence of steps where each section of the process is shutdown or started before the next section. Past practice had been to vent to the flare during this sequence until the unit was gas-free during shutdown or until producing on-specification product when going through startup. The operator made modifications to the process that allowed for streams to be recycled within the unit that dramatically reduced the amount of material sent to the flare during shutdown and startup.

34 Application of Non-Thermal Plasma for Air Pollution Control

(http://miedept.mie.uic.edu/lab/kennedy/Application_Plasma_4.htm) 35 Pulsed Corona Plasma Pilot Plant for VOC Abatement in Industrial Streams as Mobile and Educational Laboratory.

Tak, G., Gutsol, A. and Fridman, A., 2005. (http://plasma.mem.drexel.edu/publications/documents/ISPC-ID670.pdf) 36 Pulsed Corona Plasma Technology for Treating VOC Emissions from Pulp Mills, July 28, 2004.

(http://www.osti.gov/energycitations/servlets/purl/826442-clriuJ/826442.PDF) 37 Destruction of Highly Diluted Volatile Organic Compounds (VOCs) in Air by Dielectric Barrier Discharge and Mineral

Bed Adsorption. Martin, L., Ognier, S., Gasthauer, E., Cavadias, S., Dresvin, S. and Amouroux, J. Energy & Fuels. Vol. 22, 576-582, 2008.



• As discussed in the HARC Project H-76 report, as of 2006, one Harris County olefins producer was planning on sending off-specification HRVOC to an off-site salt dome storage facility for later reprocessing. This would result in a reduction in HRVOC emissions from flaring by approximately 17 tons/year at a projected capital cost of approximately $700,000.38

3.5.2 Flare Gas Recovery Flare Gas Recovery (FGR) refers to taking low pressure waste gases in the flare header, compressing the gases, and then reprocessing them or using them as a fuel gas in the plant. When the flow is less than or equal to the capacity of the FGR system, the flare gas will be recovered. During these periods of normal operations, emissions from the flare will approach zero. When the flare gas flow rate is greater than the capacity of the FGR system, the excess flare gas will flow through a liquid seal drum and to the flare tip for combustion. FGR systems are designed for recovering waste gases during normal operations. During non-routine operating conditions (e.g. MSS and emission events), excess waste gas will flow to the flare for combustion.

Potential benefits of FGR include:

• Waste gas may have substantial heating value and could be used as a fuel source in the plant, thereby reducing fuel purchase costs;

• Waste gas could be used as feedstock or product in certain applications; and

• Emissions from flaring would be reduced.

FGR is widely used in petroleum refineries. As part of multi-facility, “global” consent agreements with the USEPA, a number of major petroleum refining companies have committed to installing FGR at one or more of their refineries. The Harris County refineries that are part of global consent agreements with EPA are:

• ExxonMobil Baytown Refinery,

• Shell Deer Park Refinery, and

• Valero Houston Refinery.39

38 How Chemical Manufacturing and Petroleum Refining Facilities in Harris County are Using Point Source Monitoring to Identify and Reduce HRVOC Emissions, HARC Project H-76, ENVIRON International Corporation, prepared for the Houston Advanced Research Center, October 2006. 39 Petroleum Refinery Consent Decree Emission Reduction Assessment for Ozone and Regional Haze SIPs, ENVIRON International Corporation, prepared for the Texas Commission on Environmental Quality, November 2007.



The ExxonMobil, Shell and Valero global consent agreements do not require the installation of FGR at these refineries. However, the ExxonMobil and Shell agreements specify compliance with the emission limits of New Source Performance Standard (NSPS) Subpart J (40 CFR 60.104(a)); specifically the provision that

“No owner or operator subject to the provisions of this subpart shall: (1) Burn in any fuel gas combustion device any fuel gas that contains hydrogen sulfide (H2S) in excess of 230 mg/dscm (0.10 gr/dscf).”

Compliance may require the installation of FGR on some flares. Assuming the FGR system is sized to handle worst-case flows during normal operations, emissions from these flares during normal operations should be limited to pilot gas combustion, or very close to zero.40

There are a number of potential limitations to use of FGR in olefins, derivatives (e.g. propylene oxide production) and polymer manufacturing facilities. These include:41

• Refineries have fuel gas headers that collect high heat content waste streams from around the refinery, compress it, and send it to boilers and/or process heaters for use as a fuel. Olefins, derivatives and polymer production facilities do not typically have this same infrastructure.

• The combustion characteristics of high concentration olefin or olefin derivative streams may not be conducive to use as a fuel. For example, propylene oxide burns explosively.

• Use of the olefin or olefin derivative as a fuel may result in undesirable environmental impacts. For example, ethylene burns very hot and results in the formation of excess NOX.

• The olefins may polymerize in the flare header, leading to plugging with associated safety and performance implications.

As with all other emission control approaches, the technical and economic feasibility of flare gas recovery must be evaluated on a case-by-case basis.

40 Ibid. 41 Developed during discussions with industry representatives during the survey portion of the project.



4 Facility Survey Findings As discussed in the Introduction, ENVIRON prepared and sent questionnaires to certain Harris County industrial sites for the purpose of gathering information on projects that have been implemented to reduce HRVOC emissions at those sites, the costs of those projects, and on other HRVOC emission reduction projects that have been considered but not implemented. ENVIRON developed separate questionnaires for flare issues and polymer production issues. The questionnaire templates are included in Appendix A.

The TCEQ identified 16 sites in Harris County to receive one or both of the surveys. Of the 16 sites, 11 participated by responding to the surveys and/or meeting with ENVIRON personnel to discuss their responses. Participating sites included 4 olefin plants and 9 polymer plants. Two facilities, out of the 11 that participated in the study, are involved in the manufacture of both polymers and olefins. Additionally, one facility that did not provide a response to the survey suggested that information obtained as part of HARC Project H76 be included in this investigation. That facility contains both petroleum refining and olefins production operations.

Survey findings are presented and discussed within this section.

4.1 Polymer Plant Questionnaire Results

4.1.1 Process Types Surveyed ENVIRON received survey responses from polymer plants that use the following processes:

• Polyethylene – low density, high pressure process

• Polyethylene – low density, low pressure process

• Polyethylene – high density, gas phase process

• Polyethylene – high density, liquid-phase slurry process

• Polyethylene – high density, liquid-phase solution process

• Polypropylene – liquid-phase slurry process

• Polypropylene – gas phase process

The number of process units in operation in the above-listed categories is presented in Figure 6. All polymer plants surveyed operate multiple process types and units at their sites.



Figure 6. Type and Number of Operational Polymer Process Units

4.1.2 Process Vent Control Techniques ENVIRON asked survey respondents whether process vents located upstream of the extruder were recycled back to the process or routed to a control device. Responses are summarized in Table 10. The number of process units utilizing each control technique is presented in the table. All polymer plants surveyed operate multiple process units.

Table 10. Summary of Process Vent Control Techniques by Process Type

Process Type Recycled to Process Flare Boiler Process

Heater Thermal Oxidizer FGR Catalytic

Oxidizer PE – low density, high pressure 2 1 1 0 1 0 1

PE – low density, low pressure process

1 1 0 0 0 0 1

PE – high density, gas phase process

1 2 0 0 1 0 1

PE – high density, liquid-phase slurry 11 12 8 0 0 0 0

PE – high density, liquid-phase solution

1 1 0 1 1 0 0

0 1 2 3 4 5 6 7 8 9 10 11 12 13

PE - LD HP

PE - LD LP

PE - HD Gas Phase

PE - HD Liq Phase Slurry

PE - HD Liq Phase Solution

PP - Liq Phase Slurry

PP - Gas Phase

Poly

mer

Pro

duct

ion

Proc

ess

Number of Process Units in Operation



Table 10. Summary of Process Vent Control Techniques by Process Type

Process Type Recycled to Process Flare Boiler Process

Heater Thermal Oxidizer FGR Catalytic

Oxidizer PP – liquid-phase slurry 6 6 1 0 0 3 0

PP – gas phase 5 5 1 0 0 2 0

Total 27 28 11 1 3 5 3

4.1.3 Finishing Operations ENVIRON asked survey respondents whether finishing vents (i.e., extruder and downstream to storage and loading) are routed to a control device. Nine facilities responded to this part of the survey.

• No Control. Of the nine facilities that responded to this part of the survey, five facilities have no control on the extruder, product storage or loading operations. Emissions from these facilities are from uncontrolled atmospheric vents.

• Thermal Oxidation. Two facilities use thermal oxidizers to control HRVOC emissions from the extruders and/or downstream operations. One facility routes the emissions from extruder vents to a thermal oxidizer. The other facility routes emissions from dryer vents and storage silos to a thermal oxidizer.

• Catalytic Oxidation. One facility routes intermediate storage emissions generated from the high-pressure LDPE manufacturing process to a catalytic oxidizer.

• Flare. One facility routes its dryer vents to a flare for control, but other finishing vents are uncontrolled.

4.1.4 HRVOC Emission Reduction Projects ENVIRON asked survey respondents whether any projects had been implemented to reduce emissions in response to the HRVOC rules. Projects could include, but were not limited to, process changes, changes in operating procedures, vent stream controls and/or flare minimization. Excluded from the survey were costs associated with installation of HRVOC monitoring equipment and any emission reductions that may have resulted from more robust monitoring of emissions. A detailed analysis of the costs of control and HRVOC emission reductions associated with the installation of HRVOC monitoring equipment is included in How Chemical Manufacturing and Petroleum Refining Facilities in Harris County Are Using Point Source Monitoring to Identify and Reduce HRVOC Emissions.42

42 How Chemical Manufacturing and Petroleum Refining Facilities in Harris County are Using Point Source Monitoring

to Identify and Reduce HRVOC Emissions, HARC Project H76, ENVIRON International Corporation, prepared for



Table 11 summarizes the projects identified by survey respondents.

Table 11. Summary of HRVOC Emission Reduction Projects

Project ID Project Name Process Type Capital Cost ($)

HRVOC Reduction

(tpy)

P1 Vent Gas Recovery Polyethylene – low density, high pressure process

650,000 40

P2 Ethylene Recovery Unit

Polypropylene – gas phase process

1,000,000 15

P3 De-inventory to propylene storage1

Polypropylene – liquid-phase slurry process

50,750 See note below

P4 Installation of Regenerative Thermal Oxidizer

Polyethylene – low density, high pressure process

11,500,000 143

P5 Changes to startup procedure2

Polyethylene – liquid-phase process

0 0.5

P6

Replacement of Catalytic Oxidizer with Thermal Oxidizer with higher DRE

Polyethylene – liquid-phase solution process

364,000 0.5

P7

Re-routing of extruder vents from carbon beds to thermal oxidizer

Polyethylene – liquid phase process

127,000 1

P8 Installation of PSA system

Polypropylene – liquid-phase slurry and gas-phase processes

7,000,000 42

P9 PSA system operability improvements


400,000 1

P10 Implementation of more efficient purge bin distributor design

Polypropylene – liquid-phase slurryand gas-phase processes

800,000 20

the Houston Advanced Research Center, October 2006.





HRVOC Reduction

(tpy)

P11 Routing of relief devices to flare3

Polypropylene –gas-phase process


P12 Installation of HRVOC caps and plugs3



P13 Implementation of leak detection probe3



P14 Installation of HRVOC awareness monitors3



P15

Installation of HRVOC sample points and caps and plugs4



P16 Installation of pump trap on compressor4



P17 Routing of pump off-gas flow to Flare Gas Recovery system4



P18 Installation of HRVOC monomer efficiency monitors4



P19 Implementation of leak detection probe4



P20 Implementation of atmospheric PSV monitoring4



P21 Enhancement of propylene storage sample system4







HRVOC Reduction

(tpy)

P22 Flare Gas Recovery system upgrade4



P23 Flare Gas Recovery system upgrade4



1 Estimated HRVOC reductions between 1 - 3 tph depending on frequency and duration of shutdown. Facility did not provide detailed information regarding the frequency and duration of shutdown; therefore, annualized HRVOC reductions were not estimated. 2 Startup performed once every two to three years. The estimated HRVOC reductions are annualized. 3 Facility provided total HRVOC emission reductions attributed to projects P11 – P14 of 27.14 tpy. However, HRVOC emission reductions attributed to each project were not available. 4 Facility provided total HRVOC emission reductions attributed to projects P15 – P23 of 21.67 tpy. However, HRVOC emission reductions attributed to each project were not available.

Additional details regarding the projects referenced in Table 11 are provided below.

• Project P1. The facility collected emissions from 8 continuous hourly production silos. Each storage silo stores approximately one hour’s worth of production. HRVOC emissions collected from these intermediate storage silos are routed to a catalytic oxidizer for control. Previously, emissions from these production silos were uncontrolled. HRVOC emission reductions due to the implementation of this project are estimated to be approximately 40 tpy.

• Project P2. The facility installed an ethylene recovery unit on an off-gas stream to recover up to 3,000,000 pounds of ethylene from the flare header system. Taking into account an assumed 99% destruction and removal efficiency (DRE), the recovery of 3,000,000 pounds of ethylene translates to post-flare HRVOC emission reductions of approximately 15 tpy.

• Project P3. This project consisted of piping modifications which allowed the facility to recycle liquid slurry back to monomer storage instead of flaring during shutdown. Depending on the duration of shutdown, HRVOC emission reductions are estimated to be between 1 and 3 tons per hour (tph) during the event.

• Project P4. The facility installed an RTO for MON/HRVOC compliance requirements. Post-extruder HRVOC emissions from the pellet dryer vent, compressor distance pieces and the pellet silo storage vents are routed to the thermal oxidizer. Previously, these emission sources were uncontrolled. HRVOC emission reductions due to the implementation of this project are estimated to be approximately 143 tpy.



• Project P5. The facility implemented procedural changes to pressure check the reactor before startup. Isobutane is used instead of ethylene as part of the reactor start-up procedure, thereby reducing the use of ethylene during this process. This operational procedure is repeated once every two to three years. Annualized HRVOC emission reductions are estimated to be approximately 0.5 tpy.

• Project P6. The facility replaced an existing catalytic oxidizer with a thermal oxidizer that achieves a higher DRE. HRVOC emission reductions due to this project are estimated to be approximately 0.5 tpy.

• Project P7. The facility rerouted the extruder vents from existing carbon adsorber beds to a thermal oxidizer. HRVOC emission reductions due to this rerouting are estimated to be approximately 1 tpy.

• Project P8. The facility installed a Pressure Swing Absorption (PSA) system to capture all continuous vent streams that were previously routed to the flare for control. These vent streams consist primarily of nitrogen (75%), with the remainder being propylene (25%). The PSA system recovers, condenses and recycles propylene back to the process in liquid form. Also, nitrogen is recycled back to the process. HRVOC emission reductions due to this project are estimated to be approximately 42 tpy.

• Project P9. The facility made improvements to the operability of its PSA system and routed additional small vents (e.g., analyzer vents, dry gas seals) to the PSA. HRVOC emission reductions due to this project are estimated to be approximately 1 tpy.

• Project P10. The facility implemented a more efficient distributor design on its purge bin. The redesigned distributor results in greater volatilization of the monomer from the polypropylene flake prior to the flake entering atmospheric storage vessels, thereby reducing atmospheric emissions. HRVOC emission reductions due to this project are estimated to be approximately 20 tpy.

• Projects P11 through P14. The facility implemented several projects related to reducing atmospheric emissions of HRVOC from atmospheric relief valves and fugitive components.43 Projects included the following:

- Routing of atmospheric relief devices to flare. Pressure relief devices on the polypropylene dryers, which were previously routed to the atmosphere, were tied in to the flare header.

- Installation of HRVOC caps and plugs. All the caps and plugs in VOC service were replaced with plugs painted fluorescent yellow for easy identification and were tethered to a cable so that they could be found easily if not in place.

- Implementation of leak detection probe.

43 Unlike other polymer facilities surveyed, this facility implemented projects primarily related to fugitive emission

reductions and monitoring.



- HRVOC awareness monitors. Project involved installation of a monitor in the control room to allow for continual monitoring of the monomer efficiency, HRVOC permit compliance and energy utilization at the unit.

Additionally, process improvements and improvements in operational reliability, mechanical integrity and monomer efficiency have contributed to HRVOC emission reductions at the facility. HRVOC emission reductions due to the capital projects and other improvements are estimated to be 27.14 tpy.

• Projects P15 through P23. The facility implemented several projects related to reducing atmospheric emissions of HRVOC from atmospheric relief valves and fugitive components and to upgrading the existing Flare Gas Recovery (FGR) system. Projects included the following:

- Installation of HRVOC sample points and caps and plugs. Project involved the installation of HRVOC sampling points on various process vent streams. The project also involved purchasing miscellaneous piping caps and plugs for lines in hydrocarbon service. All the caps and plugs in VOC service were replaced with plugs painted fluorescent yellow for easy identification and were tethered to a cable so that they could be found easily if not in place.

- Installation of pump trap on compressor. Project involved installing a skid mounted pump trap system on the sour oil/seal gas return line located in the propylene distillation section. This system allows for the separation of propylene entrained in the sour oil, allowing the propylene emissions to discharge to the flare header instead of the atmosphere.

- Routing of off-gas flow to FGR. Project involved installation of pipe with control valve to allow off-gas from compressor pump trap to be separately fed to FGR. Project was needed to optimize monomer efficiency.

- Installation of HRVOC monomer efficiency monitors. Project involved installation of monitors in control rooms to allow for continuous monitoring of HRVOC and monomer efficiency.


- Implementation of atmospheric PSV monitoring. Wireless pressure transmitters were installed on pressure relief valves, which are routed to the atmosphere, and tied into the DCS so that the time and duration of each pressure relief event could be monitored.

- Enhancement of propylene storage sample system. Sampling system on the propylene storage bullets was tied to the flare header to prevent atmospheric releases while sampling.

- Two FGR upgrades. One project improved the reliability of the FGR system from 95% to 99%. A second project improved the reliability of the FGR and improved the recovery of propylene and hexane.



Additionally, process improvements and improvements in operational reliability, mechanical integrity and monomer efficiency have contributed significantly to HRVOC emission reductions at the facility. HRVOC emission reductions due to the capital projects and other improvements are estimated to be 21.67 tpy.

ENVIRON also asked respondents about HRVOC emission reduction projects that had been considered, but not implemented. Details related to those projects are discussed below.

• Storage Silo Control. This project would have reduced monomer emissions from storage silos at a polyethylene manufacturing facility (low density, high pressure process). The project would have involved the installation of multiple transfer lines to different storage silos to route low heat value waste gas streams to a catalytic oxidizer. If implemented, the project would have required a capital investment of approximately $5MM for a catalytic oxidizer to reduce HRVOC emissions by up to 40 tpy.

• Propylene Nitrogen Recovery Unit (PNRU). The implementation of the PNRU project would have reduced HRVOC emissions by recovering the monomer (propylene) and reusing it in the polypropylene production process. For safety reasons, nitrogen, which is inert, would be used to recover propylene. In a typical PNRU application, the vent stream from the resin degassing bin is compressed and then cooled to condense the propylene. The gas leaving the condenser, which contains a significant amount of propylene, is fed to a membrane unit. The membrane unit separates the stream into a propylene-enriched permeate stream and a purified nitrogen stream. The permeate stream is recycled to the inlet of the compressor and then to the condenser, where the propylene is recovered. The purified nitrogen stream is recycled to the degassing bin. 44

Two different polymer manufacturing facilities have considered installation of a PNRU. Implementation of the two PNRU projects would involve capital expenditures of greater than $1 MM and $12 MM, respectively.45 Neither facility provided any information regarding estimated HRVOC emission reductions that would have been achieved by the implementation of PNRU.

• Isobutane Nitrogen Recovery Unit (INRU). The INRU is conceptually similar to PNRU except that it would be used to recover isobutane from the polyethylene production process. In addition to recovering isobutane, ethylene would potentially be recovered using INRU. Neither cost nor potential HRVOC emission reduction information was provided by the facility that considered this project.

44 http://www.mtrinc.com/polypropylene_production.html 45 The $12 MM estimate is considered more refined; the greater than $1 MM provided by one facility is considered a

rough estimate of the minimum amount of capital investment required.



• Off-gas Recycle. The off-gas recycle project would recycle the reactor off-gas back to the polypropylene production process. To make the necessary modifications and process changes would require a capital investment in excess of $1 MM.

• Route Vents to Flare Header. This project would involve routing reactor vents and atmospheric relief valves to a process flare header. While not yet implemented, the survey respondent indicated that the facility is proceeding with implementation in the near future. The flare header must be modified to handle the large potential flow during emergency shutdown. The facility plans to install a distributed control system (DCS) to program the shutdown sequence. The current estimated cost for this project is $8 MM. No estimate of the reduction in HRVOC emissions to be realized was provided. Routine emissions at this facility are routed to a thermal oxidizer. This project would only affect MSS and event emissions.

• Ethylene Recovery Unit. This project would involve routing intermediate flake tanks vents to an Ethylene Recovery Unit (ERU). Currently, these intermediate tanks are uncontrolled. Because the flake tanks are designed for atmospheric pressure, the facility would have to design a pressure control scheme, install piping and auxiliary equipment in addition to the ERU. The estimated cost for this project is $10 MM. No estimate of the reduction in HRVOC emissions to be realized was provided.

4.1.5 Excess Monomer Removal ENVIRON asked survey respondents whether excess monomer was removed from resins prior to finishing operations (i.e., extruder vents and vents downstream of the extruder). Nine facilities responded to this part of the survey. Responses are discussed below.

• No Control. Four of the nine facilities that responded to this part of the survey did not report any processes in place to recover raw materials prior to finishing operations.46

• Catalytic Oxidation. One facility routes emissions from a tertiary degasser for LLDPE and HDPE to a catalytic oxidizer.

• Hot Nitrogen Purge. One facility purges the polyethylene and polypropylene fluff with hot nitrogen into a closed loop system which also includes the extruder feed tank. Residual hydrocarbons recovered are eventually sent to the flare. In another facility, excess monomer is recovered from the polymer slurry prior to the production of pellets. In the liquid-phase slurry process, hot nitrogen stripping is used to recover the monomer in downstream units. For the liquid-phase solution process, the excess monomer is stripped using nitrogen and a system of centrifuges and flash dryers.

46 Although these four facilities did not report any processes in place to recover raw materials prior to finishing

operations, these facilities may operate a closed loop system with nitrogen purge.



• Low Pressure Recovery. One facility uses a low pressure recovery compressor to recover nitrogen, ethylene and isobutane from the facility’s intermediate polyethylene flake storage tanks. These intermediate tanks store polyethylene flake prior to extrusion. The recovery compressor routes the nitrogen, ethylene and isobutane to a flare. This stream is approximately 99% isobutane.

• Ethylene Recovery Unit. The same facility employing low pressure recovery also utilizes an Ethylene Recovery Unit (“ERU”). The ERU is a three step, cryogenic process, by which ethylene is recovered from the resins prior to finishing operations. Recovered ethylene is recycled to the process, routed to a boiler as fuel or routed to a flare if the boiler is not operational. Note that both the low pressure recovery system and the ERU were installed by the facility prior to the advent of the HRVOC rules.

• Pressure Swing Absorption. One facility utilizes a Pressure Swing Absorption (“PSA”) system to capture all continuous vent streams that were previously routed to the flare for control. These vent streams consist primarily of nitrogen (75%), with the remainder being propylene (25%). The PSA system recovers, condenses and recycles propylene back to the process in liquid form. Also, nitrogen is recycled back to the process.

4.1.6 Costs of Polymer Plant HRVOC Emission Reduction Projects Table 12 summarizes the cost effectiveness of HRVOC reduction projects implemented by polymer processing facilities as listed in Table 11. As noted in the discussion following Table 11, the largest project in terms of total capital investment, P4, was implemented for MON/HRVOC compliance, not just for controlling HRVOC emissions.

Table 12. Cost Effectiveness of HRVOC Emission Reduction Projects

Project ID Capital Cost ($)

Annualized Capital Cost 1

($)

Direct and Indirect

Annual Cost($)

Total Annual Cost2

($)

HRVOC Emission Reduction

(tpy)

Cost Effectiveness3

($/tpy)

P1 650,000 130,000 5,000 135,000 40 3,375

P2 1,000,000 200,000 Not Provided 200,000 15 13,333

P3 50,750 10,150 Not Provided 10,150 See Note4 N/A

P4 11,500,000 2,300,000 120,000 2,420,000 143 16,923

P5 0 0 0 0 0.5 N/A5

P6 364,000 72,800 25,000 97,800 0.5 195,600

P7 127,000 25,400 5,000 30,400 1 30,400

P8 7,000,000 1,400,000 Not Provided 1,400,000 42 33,333

P9 400,000 80,000 Not Provided 80,000 1 80,000

P10 800,000 160,000 Not Provided 160,000 20 8,000






($)

Direct and Indirect

Annual Cost($)

Total Annual Cost2

($)


(tpy)

Cost Effectiveness3

($/tpy)

P11 – P14 273,000 54,600 Not Provided 54,600 27.14 2,012

P15 – P23 755,600 151,120 Not Provided 151,120 21.67 6,974 1 Based on five-year project life and a discount rate of 0%. 2 Total Annual Cost = Annualized Capital Investment + Direct and Indirect Annual Cost 3 Cost Effectiveness = Total Annual Cost / Total HRVOC Emission Reduction 4 Estimated HRVOC reductions between 1 - 3 tph depending on frequency and duration of shutdown. Facility did not provide detailed information regarding the frequency and duration of shutdown; therefore, annualized HRVOC reductions were not estimated. 5 Startup performed once every two to three years. The estimated HRVOC reductions are annualized. Total capital investment and annual costs are negligible, so cost of control is effectively zero but HRVOC emission reductions are low.

ENVIRON asked survey participants to provide estimates of total capital investment as well as direct and indirect annual costs. As noted in Table 12, annual cost information was not made available for all projects. Some facilities indicated they do not track annual costs for these projects separately from other annual costs.

As shown in Table 12, there is a wide range in cost effectiveness, ranging from $2,012 to $195,600 per ton per year of HRVOC controlled.

4.2 Flare Questionnaire Results

4.2.1 Flare Inventory Figure 7 shows the number of flares in HRVOC service by industry sector. Figure 7 also shows the percentage of average flow to maximum design flow for the flares in service. As shown in Figure 7, under normal operating conditions, flares in HRVOC service operate at 0.6% - 7.1% of the maximum design flow. The percentage of design flow for each industry sector was determined by calculating the average value for the flares within a particular industry sector. For the facilities that responded to this survey, average flows across all sectors were approximately 4.4% of flare design capacity. The range was from less than 0.0001% to 38%. Most flares are designed to handle maximum flow rates expected during facility upsets. Therefore, instead of having a dedicated flare for HRVOC control, most facilities use one flare for both routine and emergency operations. All but three of the 31 flares at respondent facilities are designed to handle routine and emergency flows.



0 2 4 6 8 10 12 14 16 18

Polymer - PE

Polymer - PP

Olefins/Chemicals

Indu

stry

Sec

tor

Number of Flares in HRVOC Service

0% 2% 4% 6% 8%% Avgerage Flow to Design Flow

Number

Percentage

Figure 7. Percent Average Flow and Number of Flares in HRVOC Service

4.2.2 Flare Reduction and Minimization Projects ENVIRON asked survey respondents whether any projects had been implemented to reduce or minimize flaring. Responses are summarized in Table 13.

Table 13. Summary of HRVOC Flare Reduction and Minimization Projects


HRVOC Reduction

(tpy)

F1 Flare Gas Recovery

Polypropylene – gas phase process 608,400 12 – 14

F2 Flare Gas Recovery1

Polypropylene – liquid-phase slurry

and gas phase processes

970,000 5 – 10

F3 Vent Recycle Polyethylene – low

density, high pressure process

50,000 10 – 12

F4 Modifications to Shutdown Procedure

Polyethylene – low density, high

pressure process N/A N/A


Polyethylene - high density, gas phase

process N/A 2





HRVOC Reduction

(tpy)

F6 Ethylene Recovery Unit

Polypropylene – gas phase process 1,000,000 15

F7 De-inventory to propylene storage2


process 50,750 See note below

F8

Modifications to Startup and Shutdown Procedures

Polypropylene – liquid-phase and gas

phase processes N/A N/A

F9 Chiller Replacement Project

Olefins 3,500,000 85

F10 Flareless Startup and Shutdown Olefins 1,100,000 50 – 100

F11 Rerouting Degassing Vent Olefins 106,250 6.75

F12 Modifications to Startup Procedure

Polyethylene – liquid phase process 0 0.5

F13

Addition of Calorimeters to Vent Gas Monitoring System

Polyethylene – liquid phase process 183,000 1

F14 Flare Gas Recovery Refinery 17,900,000 9

F15 Flare Gas Recovery Refinery 34,500,000 45

F16 Heavy Ends Stream Recovery Olefins 220,000 50

F17 Olefins Flare Reduction Olefins 550,000 9

F18 Off-spec monomer to off-site storage Olefins 700,000 17

1 Project is installed, but not yet operating. 2 Estimated HRVOC reductions between 1 - 3 tph depending on frequency and duration of shutdown.

Projects P2, P3 and P5 from Table 11 also appear as Projects F6, F7 and F12 in Table 13. These projects cannot be classified solely as polymer projects as their implementation also affected HRVOC emissions through flaring. Therefore, these projects are included in the discussion of both polymer plant and flare projects.



Additional details concerning the projects referenced in Table 13 are provided below.

• Project F1. The facility is using an existing compressor to remove the waste gas from the flare header using suction pressure. The waste gas is the reactor off-gas from the polypropylene (gas phase) process. This waste gas is sold to a neighboring facility for use as boiler fuel. HRVOC emission reductions are estimated to be between 12 and 14 tpy.

• Project F2. The facility installed a compressor to remove the waste gas from the flare header using suction pressure. The waste gas is the reactor off-gas from the polypropylene (liquid-phase slurry and gas phase) processes. However, the project is not operational yet. When operational, HRVOC emission reductions are estimated to be between 5 and 10 tpy.

• Project F3. The facility added a recycle stream from the purge gas vessel. These emissions are routed back to the process (polyethylene – low density, high pressure). Reductions in HRVOC emissions due to reduced flaring as a result of vent recycle are estimated to be 10 to 12 tpy.

• Projects F4, F5, F8, and F12. Several facilities have modified their startup/shutdown procedures to minimize flaring emissions. One such modification to shutdown procedures (Project F4) is to reduce the reactor pressure before flaring. Reactor off-gas is recompressed and sent to purification step to recover ethylene. In another project (Project F5), the facility changed the product transition procedure to route the initial purge from the flare to the site’s ethylene unit fuel gas system. This reduced 19,000 lbs of ethylene per product transition or approximately 300,000 lbs/yr out of the flare. Assuming 99% flare destruction efficiency, the estimated HRVOC emission reductions are 2 tpy. In Project F12, modifications to startup procedures include pressure checking the reactor with isobutane instead of ethylene, resulting in annualized HRVOC emission reductions of approximately 0.5 tpy.

• Project F6. The facility installed an ethylene recovery unit on an off-gas stream to recover up to 3,000,000 pounds of ethylene from the flare header system. Taking into account an assumed 99% DRE, the recovery of 3,000,000 pounds of ethylene translates to post-flare HRVOC emission reductions of approximately 15 tpy.

• Project F7. This project consisted of piping modifications which allowed the facility to recycle liquid slurry back to monomer storage instead of flaring during shutdown. Depending on the duration of shutdown, HRVOC emission reductions are estimated to be between 1 and 3 tph during the event.

• Project F9. The facility implemented this project to condense and recover C4 compounds (mostly 1,3-butadiene) from vent streams in the olefins plant. This project has reduced the load on vent recovery compressors, thereby minimizing or eliminating the bypass of HRVOCs to the flare. HRVOC emission reductions due to minimized flaring are estimated to be approximately 85 tpy.

• Project F10. The facility was able to achieve 75 – 80% reduction in HRVOC emissions due to flaring with the implementation of “flareless” startup/shutdown procedures. These reductions in flaring translate into HRVOC emission reductions of 50 to 100 tpy.



• Project F11. The facility rerouted the propylene compressor degassing pot vent back to the process. The vent was previously routed to the flare. Reductions in HRVOC emissions due to flaring as a result of vent recycle into the process are estimated to be approximately 6.75 tpy.

• Project F13. The facility added calorimeters to the already existing vent gas monitoring system. The addition of calorimeters allows operations personnel to identify any spike in the heat value of the waste gas stream in a timely manner. This is also helpful in identifying a source that may be inadvertently venting to the flare. This project has resulted in a reduction of 0.5 to 1 tpy of HRVOC emissions due to flaring.

• Project F14 and F15. The refinery implemented two flare gas recovery projects. The flare gas recovery systems have resulted in HRVOC reductions of approximately 9 and 45 tpy, respectively. Additional details regarding these projects are not available; however, it is known that the projects were not implemented for purpose of achieving HRVOC emission reductions.

• Project F16 and F17. The olefins plant implemented two flare minimization projects. The flare minimization projects have resulted in HRVOC reductions of approximately 50 and 9 tpy, respectively. Additional details regarding these projects are not available.

• Project F18. The facility sends off-specification monomer from the process to off-site salt dome storage well during start-up and shutdown, resulting in HRVOC reductions of approximately 17 tpy.

4.2.3 Flare Gas Recovery Six facilities responded to this part of the survey. Two facilities, both polypropylene manufacturers, have installed FGR systems for flare minimization. Four facilities have not implemented FGR. Responses from two facilities discussing why they have not implemented FGR are as follows:

• At one facility, the flare controls only relief devices and emergency emissions. The flare header normally handles low flow rates. Therefore, a compressor installed for FGR would only run for short periods of time.

• At an olefins facility, tie-ins do not exist in the existing plant layout; therefore, any modification would require a major turnaround. Because it is an olefins plant, any oxygen ingress into the process resulting from operation of the FGR system could result in a dangerous situation. Another challenge for the olefins plant is the storage of off-spec material for future reuse. Currently, the plant is not equipped to store off-spec material or reuse it.

4.2.4 Costs of HRVOC Flare Reduction and Minimization Projects Table 14 summarizes the cost effectiveness of HRVOC flare reduction and minimization projects implemented by polymer plants, olefins plants and refineries listed in Table 13.



Table 14. Cost Effectiveness of HRVOC Flare Reduction and Minimization Projects



($)

Direct and Indirect

Annual Cost($)

Total Annual Cost2

($)


(tpy)

Cost Effectiveness3

($/tpy)

Polymer Plants:

F1 608,400 121,680 N/A 121,680 13 9,360

F2 970,000 194,000 N/A 194,000 7.5 25,867

F3 50,000 10,000 N/A 10,000 11 909

F4 N/A N/A N/A N/A N/A N/A

F5 N/A N/A N/A N/A 2 N/A

F6 1,000,000 200,000 N/A 200,000 15 13,333

F7 50,750 10,150 N/A 10,150 See Note5 N/A

F8 N/A N/A N/A N/A N/A N/A

F12 0 0 0 0 0.5 N/A

F13 183,000 36,600 15,000 51,600 1 51,600

Olefins Plants:

F9 3,500,000 700,000 50,000 750,000 85 8,824

F10 1,100,000 220,000 N/A 220,000 75 2,933

F11 106,250 21,250 N/A 21,250 6.75 3,148

F16 220,000 44,000 N/A 44,000 50 880

F17 550,000 110,000 N/A 110,000 9 12,222

F18 700,000 140,000 N/A 140,000 17 8,235

Refineries:

F14 17,900,000 3,580,000 N/A 3,580,000 9 397,778

F15 34,500,000 6,900,000 N/A 6,900,000 45 153,333

Total:

w/ Refinery 61,438,400 12,287,680 65,000 12,352,680 346.75 35,624

w/o Refinery 9,038,400 1,807,680 65,000 1,872,680 292.75 6,397

1 Based on five-year project life and a discount rate of 0% 2 Total Annual Cost = Annualized Capital Investment + Direct and Indirect Annual Cost 3 If the facility provided a range of HRVOC emission reductions, the median is used. 4 Cost Effectiveness = Total Annual Cost / Total HRVOC Emission Reduction 5 Estimated HRVOC reduction between 1 - 3 tph depending on frequency and duration of shutdown.



As shown in Table 14, there is a wide range in cost effectiveness, ranging from $880 to nearly $400,000 per ton of HRVOC controlled. As mentioned during discussion of the projects, the two refinery HRVOC abatement projects were not implemented for the purpose of reducing emissions of HRVOC. Excluding these two projects, cost effectiveness ranges from $880 to $51,600 per ton of HRVOC controlled.

4.3 Analysis

4.3.1 Types of Emission Reduction Projects Survey participants provided information on 38 HRVOC emission reduction projects. A summary of these projects by industry sector and type is presented in Table 15. The types of projects are as follows:

Process Change: ..............................Change in how the product is made. For example, changing from a gas phase process to a liquid phase slurry process.

Change in Operating Procedures: ....Change in operating procedures such as enhanced maintenance or use of more robust process simulation to reduce emissions during startup and shutdown.

Vent Gas Control: ............................. Installation of controls on vent streams where none existed previously or upgrading to control systems with higher control efficiencies, such as routing vent streams to a thermal oxidizer instead of a flare.

Flare Minimization: ............................Recovery of material for reuse instead of sending it to the flare header and/or recovering material in the flare header for beneficial reuse.

Table 15. Summary of Project Type by Industry Sector

Type of HRVOC Emission Reduction Project

Industry Sector Number of

Plants Surveyed1

Process Change

Change in Operating Procedure

Vent Gas Control

Flare Minimization

Polymers 9 0 16 7 8

Olefins 4 0 1 0 4

Refining 1 0 0 0 2

Total 14 0 17 7 14 1Two of the survey respondents manufacture both polymers and olefins at the site. One site contains both refining and olefins manufacturing operations.



As shown, no facility implemented a change in process for the purpose of reducing emissions of HRVOC.

4.3.2 Cost Effectiveness of Emission Reduction Projects Table 16 presents a plant-by-plant summary of projects that have been undertaken to reduce emissions of HRVOC. As previously noted, the refinery projects were not undertaken for the sole purpose of reducing emissions of HRVOC. In many cases, the annual direct and indirect costs (e.g. natural gas to fuel a thermal oxidizer) have not been provided by survey respondents. In those cases, the actual annual costs and the cost of controlling HRVOC emissions will be greater than what is reported herein.

Table 16. Cost and Emission Reduction Summary by Plant

Site1 Number of Projects Total Annual

Cost2 ($)

HRVOC Emission

Reduction (tpy)

Cost Effectiveness3

($/tpy)

Polymer 1 4 145,000 53 2,736

Polymer 2 1 200,000 15 13,333

Polymer 3 4 325,830 20.5 15,894

Polymer 4 1 2,420,000 143 16,923

Polymer 5 4 179,800 3 59,933

Polymer 7 3 1,640,000 63 26,032

Polymer 8 4 54,600 27.14 2,012

Polymer 9 9 151,120 21.67 6,974

Polymer Plant Subtotal 30 5,116,350 346.31 14,774

Olefins 1 1 140,000 17 8,235

Olefins 2 1 21,250 6.75 3,148

Olefins 3 2 970,000 160 6,036

Olefins 4 2 154,000 59 2,610

Olefins Plant Subtotal 6 1,285,250 242.75 5,295

Refinery 1 2 10,480,000 54 194,074

Refinery Subtotal 2 10,480,000 54 194,074

Total (w/ Refinery 1) 38 16,881,600 643.06 26,252

Total (w/o Refinery 1) 36 6,401,600 589.06 10,867 1 Polymer Plant 6 did not implement any projects in response to the HRVOC rules. 2 Total Annual Cost = Annualized Capital Investment + Direct and Indirect Annual Costs. Annualized Capital Investment assumes a 5-year project life and a discount rate of 0%. As indicated in Chapter 4, in many cases direct and indirect annual costs are not provided. 3 Cost Effectiveness = Total Annual Cost / Total HRVOC Emission Reduction



0 20 40 60 80 100 120 140 160 180

Polymer 9Polymer 8Polymer 7Polymer 5Polymer 4Polymer 3Polymer 2Polymer 1

Olefins 4Olefins 3Olefins 2Olefins 1

Estimated HRVOC Reduction (tpy)

0 2,000,000 4,000,000 6,000,000 8,000,000 10,000,000

Total Annual Cost ($)

Est. HRVOC ReductionTotal Annual Cost

Figure 8 graphically presents plant-by-plant annual costs for HRVOC emission reduction projects and the resulting reduction in annual emissions. As shown, there is wide variation in:

• The number of HRVOC emission reduction projects undertaken at polymer plants (1-9),

• The amount of money spent on HRVOC emission reduction projects (annual costs of $54,600 to $2,420,000), and

• The reduction in HRVOC emissions achieved (3 tpy to 143 tpy)

Since the projects were not undertaken for purposes of reducing HRVOC emissions, Refinery 1 is not included in Figure 8. Also, because Polymer 6 did not implement any projects in response to the HRVOC rules, Polymer 6 is not included in Figure 8.

Figure 8. Cost and Emission Reduction Summary by Plant

Figure 9 presents a scatter plot of individual HRVOC emission control projects showing the reduction in HRVOC emissions as a function of annual cost. Excluded are projects where HRVOC emission reductions are not known (Projects P3, P11 - P14, P15 – P23, F4, F7, and F8). Since the projects were not undertaken for the purpose of reducing HRVOC emissions, Projects F14 and 15 are also excluded.

As shown, of the 20 projects included in Figure 9, the annual cost of 17 of the projects is less than $250,000. Of the 20 projects, 14 resulted in HRVOC emission reductions of no more than 20 tpy.



0.0

20.0

40.0

60.0

80.0

100.0

120.0

140.0

160.0

0 500,000 1,000,000 1,500,000 2,000,000 2,500,000 3,000,000


Estim

ated

HR

VOC

Red

uctio

n (tp

y)

Figure 9. HRVOC Emission Reductions as a Function of Cost for Individual Projects

4.3.3 Projects Not Undertaken As discussed in Section 4.1.4, there are a number of HRVOC emission reduction projects that were not undertaken by polymer plants. Based on discussions with industry personnel, the varied reasons include:

• They didn’t need additional reductions in HRVOC emissions. This was because 1) emissions were already less than their HECT allowance allocation; and/or 2) the company manages their HRVOC emissions as a portfolio across a number of sites that may include polymer plants, olefins and chemical manufacturing sites and/or petroleum refineries and the portfolio as a whole may be sufficient.

• The cost of additional reductions in HRVOC emissions exceeded internal financial thresholds.

None of the companies surveyed have undertaken projects for the purpose of selling excess allowances. Reasons include:

• With little or no market activity to-date, there are not good pricing signals as to what HECT allowance vintages may be worth in the future. Therefore, there is insufficient information available to make sound investment decisions.

• Since MSS and event emissions (up to 1,200 pounds per hour) count toward the HECT annual cap, companies are unwilling to sell allowance stream ownership.



5 Conclusions & Recommendations Conclusions and recommendations that ENVIRON is able to make from the information collected, compiled and analyzed as part of this project are as follows:

1. This study was designed to focus on HRVOC emission reduction projects implemented by polymer plants, and on determining whether additional cost effective control technologies were available to further reduce HRVOC emissions from polymer plants. A total of 30 projects were identified at nine plant sites with cost effectiveness ranging from $2,012 to $59,933 per ton of HRVOC controlled. Without additional information, it cannot be determined if polymer plants undertook more or fewer projects or if the cost effectiveness of those projects was higher or lower than those projects undertaken by other types of facilities. Determining whether additional controls to further reduce HRVOC emissions at polymer plants would be cost effective is most properly done on a case-by-case basis. No control technologies were identified that could be universally applied to achieve further reductions in HRVOC emissions from polymer plants in a cost-effective manner. ENVIRON recommends that the TCEQ build upon this investigation by including other HECT-affected facilities.

2. Use of flares is by far the most common method for controlling emissions of HRVOC. Most of the flares used to control routine emissions of HRVOC are designed to handle the very large flow rates that may occur during MSS and emission events – “emergency flares.” For the facilities that responded to this survey, average flows were approximately 4.4% of flare design capacity. However, the range was from less than 0.0001% to 38%. ENVIRON recommends that the TCEQ investigate: 1) the amount of HRVOC that is currently controlled by emergency flares and, 2) HRVOC destruction efficiency as a function of flare throughput (actual flow rate vs. design flow rate).

3. While not the focus of this investigation, it was determined during the conduct of this study that polymer production units employing different processes (e.g. gas-phase vs. liquid-phase slurry) may have very different emission profiles even though they are manufacturing a similar product (e.g. polyethylene). Although there are a number of different types of reactors that make polyethylene, there are differences in the polyethylene physical properties among the various technologies. The difference in physical properties determines the end use of the polyethylene. It was also determined that making changes to the type of process used to manufacture a product may require an investment on the order of replacing the entire production unit. This constitutes a significant barrier to change, given that some technologies are proprietary and/or licensed. No facility surveyed undertook a process change to reduce HRVOC emissions.



ENVIRON recommends further investigation into potential differences in emission profiles for the different types of polymer production processes. This information may be of value to the TCEQ should they consider revising the method for allocating HECT allowances.



6 Project Team Contact Information TCEQ Project Representative: Teresa S. Hurley, P.E.

+1 512.239.5316 [email protected]

ENVIRON Principal in Charge: Dr. Gregory Yarwood +1 415.899.0704 [email protected]

ENVIRON Principal Investigator and Project Manager:

Steven H. Ramsey, P.E., BCEE +1 713.470.6657 [email protected]

ENVIRON Co-Investigators: Christopher J. Colville, P.E. +1 713.470.2647 [email protected]

Dr. Shagun Bhat +1 713.470.2648 [email protected]


Project Number 06-17477L

Appendix A:Questionnaire Templates

Polymer Plant QuestionnaireCost Analysis of HRVOC Controls on Polymer Plants and Flares

TCEQ Work Order No. 582-07-84005-FY08-12

1. What type of process is used? Please select all that apply.Polyethylene - low density, high pressure processPolyethylene - low density, low pressure processPolyethylene - high density, gas phase processPolyethylene - high density, liquid-phase slurry processPolyethylene - high density, liquid-phase solution processPolypropylene - liquid-phase slurry processPolypropylene - gas phase processOther

YesNo

Project 1Description:Capital Cost ($):Annual Cost ($):Estimated HRVOC Reduction (tpy):


3. Have any HRVOC emission reduction projects been considered, but not implemented?YesNo

High Cost (please describe what constitutes high cost)Technical Infeasibility (please describe what constitutes technical infeasibility)Other (please describe)

YesNo

If yes, how do they vary by type of production process?

2. Have any projects been implemented to reduce emissions in response to HRVOC rules? Projects could include process changes, pollution prevention techniques as well as add-on control technology.

If yes, please provide the following for each project: description, capital and annual cost, and estimated HRVOC reduction.

If yes, please select from among the following potential reasons.

4. Are there emission reduction methods that would be technically feasible for new plants but not for existing plants?

Questionnaire Instructions

1. Please add responses to the yellow highlighted fields only.

2. For those questions requesting a "yes" or "no" response, please enter an "a" in the yellow highlighted field next to the appropriate response.

3. Please feel free to add additional projects for question 2, as necessary.

Polymer Plant Questions 1 of 2



What is unique to your facility that would make emission reductions more difficult or more expensive?

Estimated costs and potential emission reduction?Method 1Description:Capital Cost ($):Annual Cost ($):Estimated HRVOC Reduction (tpy):Method 2Description:Capital Cost ($):Annual Cost ($):Estimated HRVOC Reduction (tpy):

Recycled to ProcessFlareBoilerProcess HeaterThermal OxidizerOther Add-on Control Device (please describe)

6. Are there VOC control devices on any finishing vents (i.e., extruder and downstream to storage and loading)?YesNo

If yes, please describe.

7. Is excess monomer removed from the pellets prior to the finishing operations (e.g., steam stripping)?YesNo

If yes, please describe how.

5. Are process vents (i.e., upstream of extruder) recycled back to the process or routed to a control device (e.g., flare, thermal oxidizer, boiler, process heater)? Please select all that apply.


Flare QuestionnaireCost Analysis of HRVOC Controls on Polymer Plants and Flares


1. How many flares are in HRVOC service?Please provide the average or typical flow and design capacity for each flare in HRVOC service.

EPN Average Flow Design Flow(MMscf/hr) (MMscf/hr)

2. Have any projects been implemented to reduce or minimize flaring?YesNo



3. Have any flare minimization/reduction projects been considered, but not implemented?YesNo


4. What factors affect the cost and feasibility of flare minimization projects at your facility?






3. Please feel free to add additional projects for questions 2 and 5, as necessary.

Flare Questions 1 of 2



5. Have any projects been conducted to route additional uncontrolled streams to flare?YesNo



6. Is a flare gas recovery (FGR) system installed at the plant?YesNo

If no, has any consideration been given to FGR as a potential flare minimization project?YesNo

7. Are any HRVOC process vents routed to a thermal oxidizer for control?YesNo



Cost Analysis of Highly Reactive Volatile Organic Compound Controls

on Refineries and Chemical Plants – Project 2009-52

Control of HRVOC Emissions in Flares at Low Flow Conditions –

Project 2009-53

Prepared for: Texas Commission on Environmental Quality

Austin, Texas

Prepared by: ENVIRON International Corporation

Houston, Texas

Date: June 2009

Project Number: 06-17477O

Final Report for TCEQ Projects 2009-52 and 2009-53 Work Order 582-07-84005-FY09-15

ENVIRON Project Number 06-17477O i

Contents Page

Executive Summary.................................................................................................................................... 1 Overview ....................................................................................................................................................... 1 HRVOC Emission Reduction Projects Survey.............................................................................................. 1 Flare Survey.................................................................................................................................................. 2

1 Introduction .................................................................................................................................... 4 1.1 Project Purpose................................................................................................................................ 4 1.2 Project Scope................................................................................................................................... 4 1.3 Project Methodology ...................................................................................................................... 15 1.4 Data Quality Assurance ................................................................................................................. 16

2 Background .................................................................................................................................. 17 2.1 Emissions from Polymer Production Processes ............................................................................ 17 2.2 Special Inventory of HRVOC Emissions........................................................................................ 18 2.2.1 HECT Allowance Allocations 20 2.3 TCEQ Guidance for Controlling Emissions from Polymer Plants .................................................. 20 2.4 Flare Issues Under Evaluation....................................................................................................... 21

3 Catalog of Potential Control Technologies ............................................................................... 24 3.1 Overview ........................................................................................................................................ 24 3.2 Process Changes........................................................................................................................... 24 3.3 Changes in Operating Procedures................................................................................................. 24 3.3.1 Enhanced Maintenance ................................................................................................................. 24 3.3.2 Dynamic Process Simulation ......................................................................................................... 26 3.4 Vent Stream Controls..................................................................................................................... 27 3.4.1 Flaring ............................................................................................................................................ 28 3.4.2 Thermal Oxidation.......................................................................................................................... 29 3.4.3 Catalytic Oxidation ......................................................................................................................... 31 3.4.4 Boilers and Process Heaters ......................................................................................................... 32 3.4.5 Bekaert Burners ............................................................................................................................. 34 3.4.6 Stripping ......................................................................................................................................... 35 3.4.7 Adsorption / Concentration ............................................................................................................ 35 3.4.8 Biofiltration / Bioscrubbing ............................................................................................................. 37 3.4.9 Refrigerated Condensers............................................................................................................... 37 3.4.10 Non-Thermal Plasma ..................................................................................................................... 37 3.5 Flare Minimization .......................................................................................................................... 38 3.5.1 Recycling/Reuse ............................................................................................................................ 38 3.5.2 Flare Gas Recovery ....................................................................................................................... 39

4 HRVOC Emission Reduction Projects Survey Findings .......................................................... 41 4.1 Facility Questionnaire Results ....................................................................................................... 41 4.1.1 Industry Types Surveyed ............................................................................................................... 41 4.1.2 Process Vent Control Techniques ................................................................................................. 42


ENVIRON Project Number 06-17477O ii

4.1.3 Finishing Operations ...................................................................................................................... 43 4.1.4 HRVOC Emission Reduction Projects ........................................................................................... 47 4.1.5 Excess Monomer Removal ............................................................................................................ 53 4.1.6 Costs of HRVOC Emission Reduction Projects............................................................................. 54 4.1.7 Flare Reduction and Minimization Projects.................................................................................... 57 4.1.8 Flare Gas Recovery ....................................................................................................................... 59 4.1.9 Costs of HRVOC Flare Reduction and Minimization Projects ....................................................... 60 4.2 Analysis.......................................................................................................................................... 63 4.2.1 Types of HRVOC Emission Reduction Projects ............................................................................ 63 4.2.2 Cost Effectiveness of Emission Reduction Projects ...................................................................... 64 4.2.3 Statistical Analysis of HRVOC Survey Data .................................................................................. 66 4.2.4 Projects Not Undertaken................................................................................................................ 69

5 Flare Survey Findings.................................................................................................................. 70 5.1 Collected Information ..................................................................................................................... 70 5.2 Flare Data Quality .......................................................................................................................... 80 5.3 Analysis.......................................................................................................................................... 81 5.3.1 Flare Design Capacity.................................................................................................................... 81 5.3.2 Flare Loading ................................................................................................................................. 83 5.3.3 Statistical Analysis of Flare Survey Data ....................................................................................... 88 5.4 Summary of Flare Survey Findings................................................................................................ 90

6 Conclusions.................................................................................................................................. 92 6.1 HRVOC Emission Reduction Projects ........................................................................................... 92 6.2 Flare Usage.................................................................................................................................... 92

7 References.................................................................................................................................... 94

8 Project Team Contact Information ............................................................................................. 95


ENVIRON Project Number 06-17477O iii

Tables Page

Table 1: TCEQ List of HECT Sites to Survey ................................................................................... 7

Table 2: TCEQ List of Flares to Survey. ......................................................................................... 11

Table 3: List of Additional Sites Providing Flare Data..................................................................... 14

Table 4: HRVOC Special Inventory Summary (Mass).................................................................... 19

Table 5: HRVOC Special Inventory Summary (Percentage) .......................................................... 19

Table 6: Comparison of HRVOC Emissions and HECT Allowance Allocations ............................. 20

Table 7: Service Type Reported in 2006 TCEQ Emission Inventory for Flares in HGB Area ........ 22

Table 8: Examples of Thermal Oxidation Used to Control HRVOC Emissions .............................. 31

Table 9: Examples of Boilers or Process Heaters used to Control RRVOC Emissions ................. 33

Table 10: Examples of Stripping Used to Control HRVOC Emissions ............................................. 35

Table 11: Physical Properties of HRVOCs ....................................................................................... 36

Table 12: Summary of Process Vent Control Techniques by Process Type for Polymers............... 42

Table 13: Summary of HRVOC Emission Reduction Projects.......................................................... 44

Table 14: Cost Effectiveness of HRVOC Emission Reduction Projects ........................................... 55

Table 15: Summary of HRVOC Flare Reduction and Minimization Projects ................................... 57

Table 16: Cost Effectiveness of HRVOC Flare Reduction and Minimization Projects ..................... 61

Table 17: Summary of Project Type by Industry Sector ................................................................... 63

Table 18: Cost and Emission Reduction Summary by Plant ............................................................ 64

Table 19: Flare Survey Data Summary – Routine Flaring ................................................................ 71

Table 20: Flare Survey Data Summary – Upset/MSS Flaring .......................................................... 76

Table 21: Flare Survey Data Analysis............................................................................................... 89


ENVIRON Project Number 06-17477O iv

Figures Page

Figure 1: Simplified Polymer Process Flow Diagram....................................................................... 18

Figure 2: HGB Industrial Flares by Service Type............................................................................. 23

Figure 3: Typical Configuration – Regenerative Thermal Oxidizer .................................................. 30

Figure 4: Typical Configuration – Recuperative Thermal Oxidizer .................................................. 30

Figure 5: Typical Configuration – Catalytic Oxidizer ........................................................................ 32

Figure 6: Bekaert Burners ................................................................................................................ 34

Figure 7: Type and Number of Facilities .......................................................................................... 42

Figure 8: Cost and Emission Reduction Summary – Facility ........................................................... 65

Figure 9: Total Annual Cost – Frequency and Cumulative Percentage........................................... 66

Figure 10: HRVOC Emission Reduction – Frequency and Cumulative Percentage ......................... 67

Figure 11: HRVOC Emission Reduction as Compared to HECT Allowance Allocations................... 67

Figure 12: Percent HRVOC Emission Reduction as Compared to HECT Allowance Allocations by Industry Sector with 95% Confidence Interval ................................................................. 68

Figure 13(a): Small Flare Maximum Design Capacity – Frequency and Cumulative Percentage ........ 81

Figure 13(b): Medium Flare Maximum Design Capacity – Frequency and Cumulative Percentage .... 82

Figure 13(c): Large Flare Maximum Design Capacity – Frequency and Cumulative Percentage......... 82

Figure 14: Average Routine Loading – Frequency and Cumulative Percentage............................... 83

Figure 15: Average Upset/MSS Loading – Frequency and Cumulative Percentage......................... 83

Figure 16(a): Small Flare Average Routine Loading to Design Capacity - Frequency and Cumulative Percentage ....................................................................................................................... 84

Figure 16(b): Medium Flare Average Routine Loading to Design Capacity - Frequency and Cumulative Percentage........................................................................................................................ 84

Figure 16(c): Large Flare Average Routine Loading to Design Capacity - Frequency and Cumulative Percentage........................................................................................................................ 85

Figure 17(a): Percent Average Flow and Number of Flares Surveyed – Industry Sector ...................... 86

Figure 17(b): Percent Average Flow and Number of Flares Surveyed – Type of Service ..................... 87

Figure 18: Percent Routine Hours of Operation – Frequency and Cumulative Percentage.............. 87

Figure 19: Percent Upset/MSS Hours of Operation – Frequency and Cumulative Percentage ........ 88

Figure 20: Routine Flare Loading as a Percent of Design Capacity by Industry Sector with 95% Confidence Interval ........................................................................................................... 90


ENVIRON Project Number 06-17477O v

Appendices Appendix A: Questionnaire Template


ENVIRON Project Number 06-17477O vi

Abbreviations ANOVA: Analysis of Variance BACT: Best Available Control Technology Btu: British Thermal Unit DRE: Destruction and Removal Efficiency EPA: United States Environmental Protection Agency EI: Emission Inventory FGR: Flare Gas Recovery FTIR: Fourier Transform Infrared HARC: Houston Advanced Research Center HGB: Houston-Galveston-Brazoria HDPE: High Density Polyethylene HECT: Highly Reactive Volatile Organic Compound Emissions Cap and Trade HRVOC: Highly Reactive Volatile Organic Compound IR: Infrared LAER: Lowest Achievable Emission Rate LDPE: Low Density Polyethylene LLDPE: Linear Low Density Polyethylene M: Thousand MECT: Mass Emission Cap and Trade MM: Million MMscf/hr: Million Standard Cubic Feet per Hour MSS: Maintenance Startup and Shutdown NNSR: Nonattainment New Source Review NSR: New Source Review PE: Polyethylene PP: Polypropylene PSD: Prevention of Significant Deterioration ppm: Parts per Million ppmw: Parts per Million by Weight RACT: Reasonably Available Control Technology RN: Regulated Entity Number SBR: Styrene-Butadiene Rubber scf: Standard Cubic Feet scfm: Standard Cubic Feet per Minute TCEQ: Texas Commission on Environmental Quality tph: Tons per Hour tpy: Tons per Year VOC: Volatile Organic Compound


ENVIRON Project Number 06-17477O 1

Executive Summary Overview In continuation of the information collected as part of Project 2008-104 (Cost Analysis of HRVOC Controls on Polymer Plants and Flares), for Projects 2009-52 and 2009-53 the TCEQ directed ENVIRON to gather and analyze certain additional data on projects undertaken by facilities in Harris County to reduce the emissions of highly reactive volatile organic compounds (HRVOC) and on flares, respectively. Specifically, the purpose of Project 2009-52 is to collect additional information on HRVOC emission reduction projects at refineries and chemical plants and to use this information to perform an analysis of the costs of controlling HRVOC emissions from different types of facilities. The purpose of Project 2009-53 is to gather information comparing the maximum design capacity and the average routine loading for flares in HRVOC service at various facilities.

The list of HECT-affected facilities was provided by the TCEQ and included chemical manufacturing facilities, polymer facilities, refineries and terminals. In order to collect this information, ENVIRON prepared a single questionnaire, pertaining to both HRVOC emission reduction projects and flares, that was sent to selected Harris County facilities.

HRVOC Emission Reduction Projects Survey The 35 sites that participated in the Phase 1 and Phase 2 surveys provided information on 51 HRVOC emission reduction projects. Key findings are as follows:

• The 9 polymer plants that participated in the study implemented 30 projects resulting in a collective reduction in HRVOC emissions of approximately 346 tons per year at an average cost of $14,774 per ton controlled.

• The 16 chemical plants – including olefins-producing facilities – that participated in the study implemented 17 projects resulting in a collective reduction in HRVOC emissions of approximately 361 tons per year at an average cost of $4,175 per ton controlled.

• The 3 petroleum refineries that participated in the survey implemented four projects that resulted in HRVOC emission reductions. However, only one of these projects was implemented for the purpose of reducing long-term HRVOC emissions. That one project resulted in HRVOC emission reductions of 13.5 tons per year at an average cost of $17,778 per ton controlled.

• Of the 8 terminals that participated in the study, none indicated that they had implemented projects for the purpose of reducing HRVOC emissions.

• Of the 51 HRVOC emission reduction projects identified by this survey, 27 involved changes in operating procedures, 17 involved flare minimization, and 7 involved installation of vent gas controls. No facility implemented a change in process to reduce HRVOC emissions.



• 76% of the total HRVOC emission reduction projects had a total annual cost less than or equal to $250,000.

• A majority of the HRVOC emission reduction projects, as reported by survey respondents, resulted in emission reductions of 20 tpy or less.

• Out of the four industrial sectors surveyed, polymer and chemical plants implemented the highest number of HRVOC emission reduction projects and achieved the largest reductions in HRVOC emissions.

• Polymer plants achieved the greatest reduction in HRVOC emission when compared to their HECT allowance allocation – approximately 55%. Chemical plants achieved HRVOC emission reductions equal to approximately 50% of their HECT allowance allocation and petroleum refineries achieved HRVOC emission reduction equal to approximately 16% of their HECT allowance allocation.

• Statistical analysis indicates that the differences in observed mean HRVOC emission reductions as a percent of HECT allowance allocations across different industry sectors are not statistically significant. The differences in observed means between chemical and polymer plants could be due to chance.

Flare Survey The sites that participated in the studies provided information on 82 flares in HRVOC service. The following key statistics on flare loading are derived from the information provided:

• Under routine operating conditions:

− For small flares with a design capacity of less than 1.0 MMscf/hr, 28% of flares operate at less than or equal to 5% of design capacity and 68% of flares operate at less than or equal to 25% of design capacity, on average.

− For medium flares with a design capacity between 1.0 and less than 10 MMscf/hr, 53% of flares operate at less than or equal to 1% of design capacity and 88% of flares operate at less than or equal to 5% of design capacity, on average.

− For large flares with a design capacity greater than 10 MMscf/hr, 92% of flares operate at less than or equal to 0.5% of design capacity and 100% of flares operate at less than 2.75% of the design capacity, on average.

• On average, flares in HRVOC service at:

− Terminals operate at approximately 20% of design capacity;

− Chemical plants operate at approximately 11% of design capacity; and

− Polymer plants operate at approximately 4% of design capacity.



Insufficient information was provided by the petroleum refineries that responded to the survey to estimate routine flare loading as a percentage of design capacity.

• Most of the flares surveyed, approximately 82%, are designed to handle routine as well as emergency flows.

• Out of all flares surveyed, 32% of flares operate at less than or equal to 0.5% of the maximum design capacity under routine operations; 62% of flares operate at less than or equal to 5% of the maximum design capacity; Almost 85% flares operate at less than or equal to 25% of the maximum design capacity. Of the flares surveyed, only one flare operates at 50% of the design capacity. This is the maximum among all flares.

• Statistical analysis indicates that variations across industry sectors in the observed means of routine flare loading as a percent of design capacity are statistically significant at the 5% level. A pair-wise comparison of industry sectors shows that the difference between polymer plants and terminals is the only statistically significant difference at the 95% confidence level although the difference between chemical plants and terminals is almost statistically significant (significant at the 94% confidence level).



1 Introduction 1.1 Project Purpose During 2008, ENVIRON conducted a study to evaluate the costs to reduce highly reactive volatile organic compound (HRVOC) emissions from polymer (primarily polyethylene and polypropylene) and olefins (primarily ethylene and propylene) production facilities in Harris County. This study also included an analysis of costs to further reduce HRVOC emissions from flaring and identification of projects, including costs that facilities have already undertaken to reduce HRVOC emissions. The final report for Project 2008-104, entitled Cost Analysis of HRVOC Controls on Polymer Plants and Flares (Work Order 582-07-84005-FY08-12) was submitted to the TCEQ in August 2008. As noted in the conclusions of that report, insufficient information was available to determine whether polymer plants undertook more or fewer projects or if the cost effectiveness of those projects was higher or lower than those projects undertaken by other types of facilities.

As follow-on to the 2008 or “Phase 1” investigations, the TCEQ directed ENVIRON to gather information on HRVOC emission sources in Harris County that were not included in the 2008 investigations; specifically, petroleum refineries, olefins and non-olefins producing chemical plants, and storage terminals. The information collected as part of the 2009 or “Phase 2” investigations entitled Cost Analysis of HRVOC Controls on Refineries and Chemical Plants (TCEQ Project 2009-52), is to be combined with the information collected during Phase 1 and a more thorough and analysis of HRVOC emission control measures and costs performed. This information and analysis may be used by TCEQ to evaluate whether the initial HRVOC allowance allocations were equitable and in identifying areas where HRVOC emissions might be further reduced at the lowest costs.

As part of 2008 studies, ENVIRON investigated the feasibility and cost of reducing flows to flares at selected polymer production and olefin production facilities. The study results indicated that many of the flares used to control routine HRVOC emissions are actually designed to handle much larger flows. For Project 2009-53, entitled Control of HRVOC in Flares at Low Flow Conditions, ENVIRON has been directed by the TCEQ to gather additional information on typical or routine flow rates and compare those values to flare design capacities. Information collected and analyzed as part of Project 2009-53 may be used by the TCEQ to support their on-going investigations of flare performance.

Since the analysis conducted for Project 2009-52 requires use of information collected during Project 2008-104, information collected during the previous study is included within the Facility Survey Findings section (Section 4) of this report. Also included from the previous study are relevant sections of background information (Section 2) and the catalog of potential control measures (Section 3).

1.2 Project Scope TCEQ Projects 2009-52 and 2009-53 are being conducted concurrently and in an integrated fashion. Therefore, the following scope of work is inclusive of both projects. The scope of work includes the following tasks:



Task 1. Develop Work Plan describing the work to be performed for the TCEQ.

Task 2. Prepare questionnaire to be sent to selected Harris County sites for gathering information on HRVOC emission reduction projects and flare data.

Task 3. Gather and compile information from facilities.

Task 4. Develop draft interim report on Control of HRVOC Emissions in Flares at Low Flow Conditions.

Task 5. Prepare final report comparing the economic feasibility of reducing HRVOC emissions from polymer processing to the costs that have been incurred by other industry sectors to reduce HRVOC emissions and the cost to further reduce HRVOC emissions by reducing flaring.

For Task 2, ENVIRON prepared a single questionnaire for gathering the following:

• Information on HRVOC emission reduction projects and their costs at refineries, chemical plants and other HRVOC Emissions Cap and Trade (HECT)-affected facilities to be used for performing an analysis of the costs of controlling HRVOC emissions from different types of facilities.

• Information to better understand HRVOC emissions from flares operating at low flow conditions, including design flow rates for flares used to abate routine HRVOC emissions along with the typical or average flow rates to the flares as determined using the flow monitors required by the HRVOC rules in Chapter 115, Subchapter H, Division 1.

TCEQ Projects 2009-52 and 2009-53 are a continuation of TCEQ Project 2008-104 (Cost Analysis of HRVOC Controls on Polymer Plants and Flares). For TCEQ Project 2008-104 (hereafter referred to as “Phase 1”), ENVIRON prepared questionnaires for flare issues and polymer production issues that were sent to Harris County industrial facilities identified by the TCEQ. For TCEQ Projects 2009-52 and 53 (hereafter referred to as “Phase 2”), ENVIRON prepared a single questionnaire pertaining to both HRVOC emission reduction projects and flares that was sent to another set of Harris County facilities selected by the TCEQ. Both the Phase 1 polymer production issues and flare questionnaire templates and the Phase 2 questionnaire are included in Appendix A.

For both Phase 1 and Phase 2, the TCEQ provided ENVIRON a list of sites to survey. For Phase 2, the TCEQ provided separate lists: one for sites with flares to survey and another for sites subject to the HECT Program that were not surveyed as part of Phase 1 but were to be surveyed as part of Phase 2. The former are sites that operate flow-monitored flares subject to HRVOC rules and the HECT Program. The latter are petroleum refineries, chemical manufacturing plants and storage terminals that are subject to the HRVOC rules and the HECT Program. ENVIRON contacted each company listed in Tables 1 (TCEQ List of HECT Sites to Survey) and 2 (TCEQ List of Flares to Survey) and provided participating sites with the questionnaire. Tables 1 and 2 include both Phase 1 and Phase 2 sites that received the questionnaire.



Of the 16 Phase 1 sites that were contacted:

• 11 participated by responding to the survey and/or meeting with ENVIRON personnel to discuss their responses.

• Five sites declined to participate.

Of the 38 Phase 2 sites that were contacted:

• 24 provided the requested information.

• Six sites declined to participate.

• Three sites that initially committed to participate did, in fact, not provide the requested information.

• One site was sold in 2007 and did not participate.

• One site is permanently shutdown.

• One site claimed to have no HRVOC emissions.

• Two sites that participated in the Phase 1 investigations requested that ENVIRON rely upon that information for Phase 2 efforts.

Although not specifically identified by the TCEQ for participation in the flare survey portion of Phase 2, the companies listed in Table 2 provided flare data in response to Questions 9 and 10 of the survey. Data provided by these companies is included in the findings section of this final report.



Table 1. TCEQ List of HECT Sites to Survey

Company Name Site Name TCEQ

Account Number

TCEQ RN Industry Category Project Phase

Albemarle Corporation Albemarle Houston Plant HG0225N RN100218247 Chemicals Phase 2

American Acryl LP American Acryl LP HX1772C RN101379287 Chemicals Phase 2

Basell USA Inc. Basell USA Bayport Plant HG0323M RN100216761 Polymers Phase 1

BASF Corporation BASF Corporation Pasadena HG1249P RN100225689 Chemicals Phase 2

Bigler Land, L.L.C. Bigler Pasadena Plant HX0055V RN102528197 Chemicals Phase 2

Celanese LTD Celanese Clear Lake Plant HGA058F RN100227016 Chemicals Phase 2


Chevron Cedar Bayou Chemical Plant HG0310V RN103919817 Polymers/Chemicals Phase 1


Chevron Phillips Pasadena Plastics Complex HG0566H RN102018322 Polymers Phase 1

Dow Chemical Company Clear Lake Operations HGA005E RN104150123 Chemicals Phase 2

Dow Chemical Company Dow Chemical La Porte Site HG0769O RN102414232 Chemicals Phase 2

EI Dupont de Nemours and Company

EI Dupont de Nemours La Porte Plant HG0218K RN100225085 Chemicals Phase 2

Enterprise Products Operating LP

Almeda LPG Facility HG0157F RN102940103 Terminal Phase 2


EPOLP Houston Ship Channel Marine Loading Facility

HX1182G RN102580834 Terminal Phase 2


Morgans Point Plant HG0714Q RN100210665 Terminal Phase 2

Equistar Chemicals LP Equistar Chemicals Channelview Complex HG0033B RN100542281 Chemicals Phase 2





Account Number


Equistar Chemicals LP Equistar Chemicals La Porte Complex HG0770G RN100210319 Polymers/Chemicals Phase 1

Exxon Mobil Corporation Exxon Mobil Baytown Facility HG0232Q RN102579307 Refinery Phase 2

Exxon Mobil Corporation Exxon Mobil Chemical Baytown Olefins Plant HG0228H RN102212925 Chemicals Phase 2

ExxonMobil Chemical Company

Mobil Chemical Houston Olefins Plant HG0035U RN102576063 Chemicals Phase 2


ExxonMobil Chemical Baytown Chemical Plant HG0229F RN102574803 Polymers/Chemicals Phase 1

Georgia Gulf Chemicals & Vinyls LLC

Georgia Gulf Chemicals & Vinyls HG0276T RN100213958 Chemicals Phase 2

Ineos Nova, L.L.C. Nova Chemicals Bayport Site HG3307M RN100542224 Chemicals Phase 2


BP Solvay Polyethylene NA HG0665E RN102537289 Polymers/Chemicals Phase 1

Innovene Polymers, Inc. Battleground Polyethylene Plant HX2897U RN100229905 Polymers Phase 1

Intercontinental Terminals Company

Intercontinental Terminals Deer Park Terminal HG0403N RN100210806 Terminal Phase 2

Johann Haltermann LTD Haltermann Plant 1 HG0319D RN100219237 Chemicals Phase 2

Johann Haltermann LTD Haltermann Plant 2 Channelview HG0929Q RN102610912 Chemicals Phase 2

Kaneka Texas Corporation Kaneka Texas Corporation HG1065E RN100218841 Terminal Phase 2

Kirby Inland Marine, LP Barge Cleaning Facility HG9625W RN102204211 Terminal Phase 2





Account Number


LBC Houston LP LBC Houston Bayport Terminal HG0029P RN101041598 Terminal Phase 2

Lyondell Chemical Worldwide Co.

Lyondell Chemical Bayport Plant HG0537O RN102523107 Chemicals Phase 2

Lyondell-Citgo Refining LP Lyondell-Citgo Refining HG0048L RN100218130 Refinery Phase 2

Millennium Petrochemicals Inc Millennium Petrochemicals La Porte Plant HX1726J RN100224450 Chemicals Phase 2

Natural Gas Odorizing Inc Natural Gas Odorizing HG0512H RN100683952 Terminal Phase 2

Nisseki Chemical Texas Inc Nisseki Chemical Texas Inc HG3626Q RN102887270 Chemicals Phase 2

Odfjell Terminals Houston, LP Odfjell Terminals HG1006U RN100218411 Terminal Phase 2

Pasadena Refining Systems, Inc.

Pasadena Refinery HG0175D RN100716661 Refinery Phase 2

Shell Deer Park Refining Company

Shell Oil Deer Park HG0659W RN100211879 Refinery Phase 2

Sunoco Inc. (R&M) Bayport Polyethylene Plant N/A RN103773206 Polymers Phase 1

Sunoco Inc. (R&M) Sunoco R&M Bayport Polypropylene HGA009I RN100524008 Polymers Phase 1

Sunoco Inc. (R&M) Sunoco La Porte Plant HG0825G RN102888328 Polymers Phase 1

Targa Midstream Services LP Galena Park Terminal HG0786O RN100214212 Terminals Phase 2

Texas Petrochemicals LP Texas Petrochemicals Houston Facility HG0562P RN100219526 Chemicals Phase 2

The Lubrizol Corporation Lubrizol Bayport Plant HG0460B RN101058410 Chemicals Phase 2

The Lubrizol Corporation Lubrizol Deer Park Plant HG0459J RN100221589 Chemicals Phase 2





Account Number


Total Petrochemicals USA Total Petrochemicals La Porte Plant HG0036S RN100212109 Polymers Phase 1

Total Petrochemicals USA Total Petrochemicals Bayport Plant HG4662F RN100909373 Polymers Phase 1

Valero Refining-Texas LP Valero Refining Houston Refinery HG0130C RN100219310 Polymers Phase 2



Table 2. TCEQ List of Flares to Survey


Account Number

TCEQ RN Industry Category Phase 1 or Phase 2 Number of Flares

Albemarle Corporation Albemarle Houston Plant HG0225N RN100218247 Chemicals Phase 2 2

BASF Corporation BASF Pasadena HG1249P RN100225689 Chemicals Phase 2 2

Basell USA Inc. Basell USA Bayport Plant HG0323M RN100216761 Polymers Phase 1 3

Celanese Ltd. Celanese Clear Lake Plant HG0126Q RN100227016 Chemicals Phase 2 1


Chevron Cedar Bayou Chemical Plant

HG0310V RN103919817Polymers/ Chemicals

Phase 1 3 (Polymers) 5 (Chemicals)

Chevron Phillips Chemical Company LP

Chevron Phillips Chemical Company HG0566H RN102018322 Polymers Phase 1 & Phase 2 4

EI Dupont de Nemours and Company

EI Dupont de Nemours La Porte Plant

HG0218K RN100225085 Chemicals Phase 2 1

Enterprise Products Operating LLC Almeda LPG Facility HG0157F RN102940103 Terminal Phase 2 1

Equistar Chemicals LP Channelview Complex HG0033B RN100542281 Chemicals Phase 1 & Phase 2 9

Equistar Chemicals LP Equistar Chemicals La Porte Complex HG0770G RN100210319

Polymers/ Chemicals

Phase 1 2 (Polymers) 2 (Chemicals)


Exxon Mobil Chemicals Baytown Olefins Plant

HG0228H RN102212925 Chemicals Phase 1 & Phase 2 3





Account Number



ExxonMobil Chemical Baytown Chemical Plant

HG0229F RN102574803 Polymers Phase 1 & Phase 2 2

ExxonMobil Oil Corporation

ExxonMobil Baytown Facility HG1276M RN102579307 Refinery Phase 1 9

Intercontinental Terminals Company LLC

Intercontinental Terminals Deer Park Terminal

HG0403N RN100210806 Terminal Phase 2 6

Lyondell Chemical Worldwide Inc

LCC Bayport Lyondell HG0537O RN102523107 Chemicals Phase 2 1

Lyondell Citgo Refining LP Houston Refining HG0048L RN100218130 Refinery Phase 2 6

Millennium Petrochemicals Inc

Millennium Petrochemicals La Porte Plant

HX1726J RN100224450 Chemicals Phase 2 1

Natural Gas Odorizing Inc

Natural Gas Odorizing HG0512H RN100683952 Chemicals Phase 2 1

Nisseki Chemical Texas Inc

Nisseki Chemical Texas HG3626Q RN102887270 Chemicals Phase 2 2

Odfjell Terminals Houston Inc Odfjell Terminals HG1006U RN100218411 Terminals Phase 2 1

Pasadena Refining System Inc

Pasadena Refining System HG0175D RN100716661 Refinery Phase 2 2

Shell Oil Company Shell Oil Deer Park HG0659W RN100211879 Refinery Phase 1 & Phase 2 9

Texas Petrochemicals LP Houston Plant HG0562P RN100219526 Chemicals Phase 2 1





Account Number


The Dow Chemical Company

The Dow Chemical Company Clear Lake Operations

HGA005E RN104150123 Chemicals Phase 2 1

Total Petrochemicals USA

Total Petrochemicals Bayport Plant

HG4662F RN100909373 Polymers Phase 1 2

Total Petrochemicals USA

Total Petrochemicals La Porte Plant

HG0036S RN100212109 Polymers Phase 1 Not Specified



Table 3. List of Additional Sites Providing Flare Data


Account Number

TCEQ RN Industry Category Phase 1 or Phase 2 Number o Flares

Enterprise Products Operating LLC

Morgans Point Complex HG0714Q RN100210665 Terminal Phase 2 2


Almeda LPG Terminal HG0157F RN102940103 Terminal Phase 2 1


EPOLP Houston Ship Channel Marine Loading Facility

HX1182G RN102580834 Terminal Phase 2 1

Georgia Gulf Chemicals & Vinyls LLC

Georgia Gulf Chemicals & Vinyls HG0276T RN100213958 Chemicals Phase 2 1

Ineos Nova LLC Nova Chemicals Bayport Site HG3307M RN100542224 Chemicals Phase 2 1

Innovene Polymers, Inc. Battleground Polyethylene Plant HX2897U RN100229905 Polymers Phase 1 1


BP Solvay Polyethylene NA HG0665E RN102537289 Polymers Phase 1 1

Johann Halterman LTD Johann Halterman HG0139D RN100219237 Chemicals Phase 2 1

Targa Midstream Services LP

Galena Park Terminal HG0786O RN100214212 Chemicals Phase 2 1



1.3 Project Methodology Following approval of the Phase 2 questionnaire by the TCEQ, ENVIRON contacted each Phase 2 site listed in Tables 1 and 2. ENVIRON sent the Phase 2 questionnaire to the sites that agreed to participate in the survey. ENVIRON requested information on projects that have been implemented to reduce HRVOC emissions at the site and their costs, and on any other HRVOC emission reducing projects that have been considered but not implemented along with the reason for not implementing the project (e.g., cost, technical feasibility). Specifically, ENVIRON requested the following information related to HRVOC emission reduction projects from each surveyed site:

• List of products manufactured at site.

• Identification of source types (e.g., cooling towers, process vents, and flares) in HRVOC service and included in HECT Program cap.

• Description of HRVOC emission reduction projects - including estimated HRVOC reduction, capital cost and annual cost.

• Description of HRVOC emission reduction projects considered, but not implemented.

• Details regarding HRVOC emission reduction methods that would be technically feasible for new facilities that manufacture the same product, but not for existing facilities.

• Details regarding anything unique that would make HRVOC emission reduction more difficult or expensive.

• Description of investments made to reduce the fugitive release of HRVOCs – including estimated HRVOC reduction, capital cost and annual cost.

• Details regarding any process units that have been shutdown or derated as a strategy for complying with the HECT Program cap.

• Description of any Flare Gas Recovery (FGR) projects as well as FGR projects considered, but not implemented.

• Identification of process vents in HRVOC service routed to control devices other than flares (e.g., thermal oxidizers, process heaters).

• Description of other projects not specifically carried out to reduce HRVOC emissions, but resulted in reducing HRVOC emissions.

ENVIRON used the information gathered from the questionnaires along with results previously obtained in the 2008 ENVIRON study to compare the economic feasibility of reducing HRVOC emissions from polymer processing to the costs that have been incurred by other industry sectors to reduce HRVOC emissions and the cost to further reduce HRVOC emissions by reducing flaring.



ENVIRON requested the following information related to flare operation from each surveyed site:

• Number of flares in HRVOC service.

• Flare service type (routine, upset/maintenance, or both).

• Flare design capacity.

• Average routine flow rate (excluding upset/maintenance) for calendar year 20081.

• Average upset/maintenance flow rate for calendar year 20082.

• Hours of operation (routine and upset/maintenance) for calendar year 20083.

Using this data, ENVIRON determined the average routine flow as a percentage of design capacity for each flare, performed additional statistical analyses, and made comparisons within and across industry sectors.

1.4 Data Quality Assurance Information contained within this report is presented as it was reported by participating industrial sites to ENVIRON. Data validation activities performed by ENVIRON were limited to email correspondence and telephone conversations with participants regarding perceived anomalies and/or clarification of projects performed. If confirmed by survey respondents as accurate, data is included in the findings and analysis presented within this report.

1 Phase 1 flare survey did not specify calendar year 2008. Phase 1 respondents provided data for a variety of years. 2 Phase 1 flare survey did not request upset/maintenance flow rates. 3 Phase 1 flare survey did not request hours of operation data.



2 Background 2.1 Emissions from Polymer Production Processes HRVOC compounds are used extensively in the manufacture of polymers.4 Examples include:

• Use of ethylene in the manufacture of polyethylene (PE). PE is a thermoplastic (becomes soft when heated and hard when cooled) that is heavily used in consumer products (e.g. plastic shopping bags). PE is classified into a large number of categories based on its density and branching. Examples include high-density PE (HDPE), low-density PE (LDPE), and linear low-density PE (LLDPE).

• Use of propylene in the manufacture of polypropylene (PP). Like PE, PP is a thermoplastic that finds wide use in a variety of applications.

• Use of 1,3-butadiene in the manufacture of synthetic rubbers such as polybutadiene and styrene butadiene rubber (SBR). SBR is widely used in the manufacture of automobile tires.

• Use of butenes in the manufacture of synthetic rubbers such as polyisobutylene and as copolymers. A copolymer is a polymer derived from two or more monomers. An example is SBR which is derived from styrene and 1,3-butadiene.

Figure 1 presents a simplified process flow diagram for manufacturing polymer pellets.5 For polymer products other than pellets, downstream operations may vary. For example, when manufacturing a polymer flake or powder, there will not be an extruder. Similarly, SBR is typically not extruded but sold as SBR crumb.

Sources of HRVOC from a polymer manufacturing process may include one or more of the following:6

• Monomer or comonomer storage

• Process fugitives

• Cooling tower heat exchange system losses

• Process vents upstream of the extruder (e.g. reactor, resin degassing)

4 A polymer is a large molecule composed of repeating structural units. Examples include polyethylene and polypropylene. 5 Technical Guidance for Chemical Sources: Polyethylene and Polypropylene Manufacturing, Draft RG-244, TCEQ Air Permits Division, February 2001. 6 Ibid



• Extruder

• Polymer storage and loading

• Wastewater treatment facilities

Figure 1. Simplified Polymer Process Flow Diagram

2.2 Special Inventory of HRVOC Emissions In June 2007, the TCEQ conducted a special emissions inventory, requesting HRVOC emissions data from those sources in Harris County, Texas, that are subject to HRVOC emissions cap-and-trade (HECT) program requirements. The reporting period for this special inventory was February 1, 2006, through January 31, 2007.

Special inventory responses were categorized based upon the primary activity at the site:

• Chemical manufacturing (non-olefin, non-polymer)

• Olefins manufacturing7

• Polymer manufacturing

• Petroleum Refining

• Independent storage terminals (not dedicated to an individual refinery, olefins, chemical or polymer manufacturing site)

HRVOC special inventory responses are summarized in Tables 4 and 5. The values shown in Table 4 are total emissions for each industry sector. Emissions for the five industry sectors combined are shown in the far right column. The percentages presented in Table 5 are for total

7 An “olefin,” or alkene, is an unsaturated chemical compound containing at least one carbon-carbon double bond. The simplest olefin is ethylene which has the following chemical structure: H2C=CH2.



emissions reported by that industry sector. Percentages for the five combined sectors are presented in the far right column. Values presented in Tables 4 and 5 are taken from summaries provided by TCEQ personnel. ENVIRON has not reviewed the special inventory submittals nor validated the values provided.

Table 4. HRVOC Special Inventory Summary (Mass)

Emissions by Industry Sector (tons) Source

Chemical2 Olefins Polymers Refining Terminal2 Combined2 Flares 278.3 376.7 460.0 308.7 45.8 1,469.5 Cooling Towers 35.4 18.1 20.9 88.1 0.1 162.6 Other Vents 97.5 157.1 221.1 125.0 1.7 602.4 Fugitives 19.6 115.5 22.0 26.0 0 183.1 Total2,3 443.2 667.4 724.1 547.8 50.9 2,433.4

Type MSS & Events 34.8 124.8 81.0 83.2 0.0 323.8 Uncontrolled1 136.4 245.3 234.0 135.9 1.8 753.4 Controlled1 259.5 297.3 409.0 328.8 45.9 1,340.5 1 Uncontrolled and controlled routine emissions. MSS and event emissions are accounted for separately. 2 Total includes emissions from sites that were not broken down by source or type of emissions. 3 Total includes fugitive emissions from equipments leaks which are not subject to HECT.

Table 5. HRVOC Special Inventory Summary (Percentage)

Emissions by Industry Sector (%) Source

Chemical1 Olefins Polymers Refining Terminals1 Combined Flares 64.6 56.4 63.5 56.3 96.3 60.8 Cooling Towers 8.2 2.7 2.9 16.1 0.1 6.7 Other Vents 22.6 23.5 30.5 22.8 3.6 24.9 Fugitives 4.5 17.3 3.0 4.7 0.0 7.6

Type MSS & Events 8.1 18.7 11.2 15.2 0.0 13.4 Uncontrolled2 31.7 36.8 32.3 24.8 3.8 31.2 Controlled2 60.2 44.5 56.5 60.0 96.6 55.4 1 Emissions (%) were determined using 430.8 tons as the total emissions for the Chemical sector and 47.6 tons as the total emissions for the Terminals sectors. 2 Uncontrolled and controlled routine emissions. MSS and event emissions are accounted for separately.

As shown in Table 5, the information contained within the special inventory submittals indicate that approximately 61% of total reported HRVOC emissions are controlled using flares. This



includes MSS and events that are controlled by flare. The 55.4% noted as “controlled” does not include MSS and events.

2.2.1 HECT Allowance Allocations Under the HECT, allowances are allocated by the TCEQ to affected sites in Harris County according to the procedures defined by rule in 30 TAC §101.394. The initial allocation occurred January 1, 2007, with subsequent allocations occurring January 1 of each year thereafter. Covered facilities at these sites include flares, cooling tower heat exchange systems and vent gas streams. Fugitive emissions are not covered by the HECT.

On August 18, 2006, the TCEQ published a list of HECT allowance allocations. A total of 51 sites in Harris County were allocated 3,451.5 tons of HRVOC. Table 6 presents a comparison of HRVOC emissions reported as part of the special inventory, by industry sector, with the HECT allowance allocation for that sector. The allowance allocation shown is only for those facilities that reported emissions as part of the special HRVOC emissions inventory. Since all facilities with HECT allowance allocations did not respond to the special inventory request, the summation of allowance allocations does not equal the total number of allowances allocated (3,451.5. tons) but does account for over 98% of the allocations.

There are several sites that could be included in more than one industry sector. In those cases, the site is placed into the industry sector that seems to best represent their primary business. For example, a petroleum refinery with a collocated chemical plant will be included in the Refining Industry sector.

Table 6. Comparison of HRVOC Emissions and HECT Allowance Allocations

Industry Sector HRVOC Emissions (tons)1

Annual HECT Allowance Allocation (tons)

Emissions as a % of Allowance Allocation

Chemical 411.2 718.6 57.2 Olefins 551.9 1,123.8 49.1 Polymers 702.0 496.1 141.5 Refining 521.8 996.7 52.4 Terminals 47.6 57.0 83.5 Combined 2,234.5 3,392.2 65.9 1 Total emissions from emission points covered by the HECT: flares, cooling towers and other vents. Fugitive emissions are not covered by the HECT. Does not included uncharacterized emissions from Chemical and Terminals sectors.

As shown in Table 6, during the period covered by the special inventory, polymer production plants were more likely than petroleum refineries, olefins plants, chemical plants, or independent storage terminals to have emissions that exceed their HECT allowance allocation.

2.3 TCEQ Guidance for Controlling Emissions from Polymer Plants Prior to developing a catalog of potential control strategies that may be available for further reducing emissions of HRVOC from polymer plants, ENVIRON first reviewed previous TCEQ



guidance on the subject. TCEQ guidance issued in February 2001 by the Air Permits Division provides the following regarding control of emissions from PE and PP manufacturing facilities undergoing New Source Review (NSR) permitting.8

• Best Available Control Technology (BACT) for all types of processes requires control of all waste gas streams upstream of the extruder.

• Control devices specified by guidance are as follows:

− For vent streams, combustion using a flare, incinerator, boiler, heater, etc.

− For fugitive emissions, a 28 VHP fugitive monitoring program.9

• Maximum allowable residual volatile organic compound (VOC) – which for PE and PP manufacturing facilities would be all HRVOC – in the polymer at the first uncontrolled vent should be less than 90 parts per million by weight (ppmw) for all manufacturing processes with the exception of high-pressure polyethylene manufacturing processes where guidance states a limit of 100 ppmw.

• Total non-fugitive VOC emissions, including HRVOC emissions, should generally be less than 200 pounds per million pounds (MM lb) of product.

The most recent published TCEQ guidance on BACT for PE and PP production facilities (October 17, 2006), is more restrictive than the 2001 technical guidance document. Specifically, it requires that total non-fugitive, uncontrolled VOC (including HRVOC) emissions are to be reduced to less than 80 pounds per MM lb of PE or PP produced.10

It should be noted that HRVOC fugitive monitoring requirements for affected facilities under 30 TAC Subpart H, Division 3, is more stringent than the requirements of 28 VHP.

2.4 Flare Issues Under Evaluation The TCEQ recently formed a Flare Task Force Stakeholder Group, whose stated goal is to make a comprehensive evaluation of all aspects of flares, including:

• Interaction of flares and air quality issues, such as ozone and air toxics;

• Understanding of flare use and efficiency; and

• Adequacy of state regulations for flares.

8 Technical Guidance for Chemical Sources: Polyethylene and Polypropylene Manufacturing, Draft RG-244. 9 The revised requirements of 28 VHP are specified in TCEQ guidance issued in May 2008: http://www.tceq.state.tx.us/assets/public/permitting/air/Guidance/NewSourceReview/rev28vhp_508.pdf 10 TCEQ Chemical Sources, Current Best Available Control Technology (BACT) Requirements, Polyethylene and Polypropylene Facilities, October 2006. (http://www.tceq.state.tx.us/assets/public/permitting/air/Guidance/NewSourceReview/bact/bact_polys.pdf)



The aspect most related to this current project is the understanding of flare use and efficiency. In a recent Flare Task Force presentation (TCEQ 2009), the TCEQ identified the following factors as potentially impacting flare performance:

• Meteorology

• Flare waste gas stream flow rate

• Flare waste gas stream composition

• Physical design characteristics and maintenance

• Assist flow rates

The focus of Project 2009-53 is on determining whether the low flows identified during the 2008 ENVIRON study are typical of flares used for HRVOC controls at other facilities. Flares that operate in both routine and upset/MSS service operate typically with a high turndown ratio, or conversely, at a low percentage of design capacity. Turndown ratio is the total design capacity compared to the actual flare waste gas stream flow rate. As presented in the 2008 ENVIRON study, flare waste gas flow rates during routine operations are often less than one percent of total design capacity.

In a recent Flare Task Force presentation (TCEQ 2009), the TCEQ quantified the number of flares in service as reported in the 2006 Emission Inventory (EI). EI data indicate that there are 1,132 flares in service statewide with 521 flares in the Houston-Galveston-Brazoria (HGB) area. A summary of the service types reported in the 2006 EI for flares in the HGB area is presented in Table 7.

Table 7. Service Type Reported in 2006 TCEQ Emission Inventory for Flares in HGB Area

Service Type Number Percentage of Total (%) Routine Only 110 21 Emergency Only 63 12 Both Routine and Emergency 280 54 Not Specified 68 13 Total 521 100

More than half of the flares in the HGB area are in both routine and upset/MSS service. Approximately 13 percent of the flares in HGB are not specified; therefore, it is possible that the number of flares in the HGB area in both routine and upset/MSS service is higher. Figure 1 presents the spatial distribution of the flares within the HGB area.



Figure 2. HGB Industrial Flares by Service Type11

11 Figure courtesy of Adam Bullock, Emissions Inventory Team, Texas Commission on Environmental Quality



3 Catalog of Potential Control Technologies 3.1 Overview Methods identified for potentially reducing HRVOC emissions include:

• Process changes,

• Changes in operating procedures,

• Vent stream controls, and

• Flare minimization.

Each of these is discussed in turn within this section.

3.2 Process Changes For existing production facilities, short of replacing older technology with newer technology, there are limited opportunities for making process changes that reduce HRVOC emissions. One of these limited opportunities discussed in a 1997 EPA document on the polymer industry is changing catalysts.12 This document suggests that there are opportunities to replace an older catalyst with a newer, better catalyst, resulting in overall yield improvements, and, thus, reducing the amount of un-reacted monomer that remains in the polymer. This information is somewhat misleading. A polymer manufacturing facility must use a catalyst that is most appropriate for their process and the polymer properties sought.

For example, a gas phase PE reactor may use an older catalyst that has lower reactivity and selectivity than a newer, better catalyst. However, the newer, better catalyst may not be suitable for use in the gas phase reactor, but can only be used in a high pressure slurry PE process. Additionally, use of the old catalyst may be necessary to produce the desired PE properties.

3.3 Changes in Operating Procedures As with process changes, there may be opportunities to modify operating procedures to reduce emissions of HRVOC. Potential changes in operating procedures may range from enhanced maintenance activities to the use of sophisticated dynamic simulation algorithms to reduce losses during non-steady state operating conditions, such as those that occur during startup and shutdown.

3.3.1 Enhanced Maintenance There may be opportunities to improve maintenance of existing equipment and reduce losses of HRVOC during normal operations and/or scheduled maintenance activities. Examples of enhanced maintenance activities may include more aggressive investigation and timely

12 Profile of the Plastic Resin and Manmade Fiber Industries, Sector Notebook Project, EPA-310/R-97-006, September 1997.



corrective action of leaking pressure safety valves (PSV), leaking compressor seals, leaking valves and flanges, leaking heat exchangers, etc. However, sites subject to the HECT are already required to monitor cooling tower return lines for leaking heat exchange systems and implement stringent fugitive monitoring and control requirements.13 [As noted previously, fugitive emissions are not included in the HECT.]

Enhanced maintenance may also include use of predictive and preventive maintenance processes. The predictive maintenance process involves review of the equipment types, the failure mechanisms associated with those equipment types and the associated mean time to failure. Preventive maintenance takes the next step and focuses maintenance on taking action before equipment fails. This preventive maintenance can lead to higher reliability, greater on-stream time, and fewer unscheduled maintenance shutdowns.

Required HRVOC monitoring of flare headers and cooling tower returns has resulted in some subject facilities implementing enhanced maintenance programs.14 For example, one site uses the HRVOC monitoring system as a feedback mechanism. When flaring, operators use the HRVOC monitoring system to help identify the source of the flows (e.g., PSV leaks, open valves). The monitors do not identify specific equipment, but the ability to speciate the compounds going to the flare helps to narrow troubleshooting efforts to a particular process area.

Some sites are also using passive infrared (IR) cameras to find and eliminate/reduce emissions of HRVOC. As part of the HARC H-76 project, ENVIRON found that, as of the summer of 2006, three of the nine surveyed sites were using the IR cameras to locate potential sources of HRVOC emissions.15 Uses of the IR cameras include:

• Integration of IR cameras into routine leak detection and repair (LDAR) programs.

• Use of IR cameras to monitor for emissions during startup and shutdown. Camera findings are used to direct corrective measures.

As discussed in the Project H-76 report, those companies that have embraced use of the passive IR cameras are strong supporters of using this technology to find and fix sources of emissions.

13 Any process unit or process within a petroleum refinery, synthetic organic chemical, polymer, resin, or methyl tert-butyl ether manufacturing process or natural gas/gasoline processing operation in the 8-county Houston/Galveston/ Brazoria area in which an HRVOC is a raw material, intermediate, final product, or in a waste stream is subject to the requirements of 30 TAC 115, Subchapter H, Division 3 which specifies monitoring and control requirements for fugitive emissions from equipment. 14 “How Chemical Manufacturing and Petroleum Refining Facilities in Harris County are Using Point Source

Monitoring to Identify and Reduce HRVOC Emissions,” HARC Project H76, ENVIRON, prepared for the Houston Advanced Research Center, October 2006. 15 Ibid



As part of an agreement with the City of Houston, one facility in Harris County has installed a fence line Fourier Transform Infrared (FTIR) monitoring system to monitor for 1,3-butadiene.16 While not initially intended as a tool for improving maintenance or operating practices, the fence line FTIR system has allowed the facility to identify operations and activities that result in emissions of 1,3-butadiene and to use that information in taking corrective action..

3.3.2 Dynamic Process Simulation Understanding how operating conditions affect waste gas production can lead to improved operating techniques. Rather than relying solely on engineering trial-and-error and operator experience to modify operating procedures to minimize flaring, dynamic process simulation has been used to minimize HRVOC flaring during shutdown and startup at an ethylene production facility.17 In the referenced study performed by Lamar University in cooperation with LyondellBasell’s Equistar Channelview Plant, dynamic process simulation was used to critically evaluate potential process and procedural modifications prior to the actual shutdown/startup or upset event. Dynamic simulation was developed for the recovery section of the ethylene plant and used to examine the following process steps:

• Approaching shutdown,

• Startup with recycle ethane, and

• Starting the cracked feed and increasing the feed to normal production rates.

The researchers found that dynamic process simulation provides an insight into process behavior that is not readily apparent through steady state simulation and process engineering calculations. Process simulations are performed using standard software, such as Aspen Plus and Aspen Dynamics™. Operators can use the results of the dynamic process simulations to modify control settings during shutdown/startup and upset conditions to minimize flaring of off-specification (off-spec) streams.

Results of the Equistar Channelview Plant dynamic process simulation study are as follows:18

• Actual flaring associated with shutdown and startup of the ethylene plant was 75% less than a previous startup of a similar plant at the site.

• Flaring emissions from the shutdown and startup were 56% less than the estimates made prior to the turnaround.

Using dynamic process simulation, other ethylene production facilities have reduced HRVOC emissions from flaring. Examples include:

16 Ibid 17 Flare Minimization via Dynamic Simulation. Singh, A., K. Li, H. H. Lou, J. R. Hopper, H. B. Golwala and S. Ghumare. Int. J. Environment and Pollution. Vol. 29, Nos. 1/2/3, pp. 19-29. 18 Ibid



• Huntsman Petrochemical reduced flaring to less than 3.5 hours during a startup event, and

• BASF-TOTAL reduced flaring by 50% compared to a previous startup.19

This project has not determined if dynamic process simulation can be applied to polymer manufacturing plants.

3.4 Vent Stream Controls Based on our review of EPA’s RACT/BACT/LAER Clearinghouse (RBLC), ENVIRON identifies three technologies that have been used to control HRVOC emissions from process vents at polymer manufacturing plants: flares, thermal oxidizers and boilers. Additionally, while not an ultimate control device, air and steam stripping have been used to remove un-reacted monomer from SBR crumb prior to finishing. While not identified in the RBLC review, ENVIRON is aware that polymer plant waste gas streams containing HRVOC have also been managed using catalytic oxidizers. Each of these technologies is briefly described within this section. Also included are brief discussions of other control technologies considered: adsorption, biofiltration, Bekaert burners and refrigerated condensers.

It is important to keep in mind that the technical feasibility, including process safety considerations, and economic feasibility of any control system can only be determined on a case-by-case basis.

Costs are not included within these general discussions of control technologies. Costs of control are highly dependent upon a number of design and operating variables and often cannot be estimated accurately within even one or two orders of magnitude without specific information as to the application. For example, USEPA Air Pollution Control Technology Fact Sheet EPA-452/F-03-019 provides the following range of costs for flares.

Capital Cost: .....................................$13 to $21,000 per standard cubic foot per minute (scfm) of flow

Operation & Maintenance Cost:........$1 to $10 per scfm

Annualized Cost:...............................$3 to $300 per scfm

Cost Effectiveness: ...........................$15 to $5,000 cost per ton of pollutant controlled

It is questionable if even this very broad range brackets actual costs that might be incurred for flaring of one or more process vent streams. Meaningful control costs can only be determined on a case-by-case basis.

19 Near-Zero Flare for Chemical Process Industry via Plant-wide Optimization and Simulation. Xu, Q., K. Li and J. L. Gossage. TERC SAC Meeting, Houston, Texas. February 2008.



3.4.1 Flaring Flaring is a combustion control process in which the combustible material to be flared is piped to a remote, usually elevated location, and burned in an open flame in the air using a specially designed burner tip, auxiliary fuel, and steam or air to promote mixing for nearly complete (>98%) destruction efficiency. Completeness of combustion in a flare is governed by flame temperature, residence time in the combustion zone, turbulent mixing of the gas stream components to complete the oxidation reaction, and available oxygen for free radical formation.20

Flares that conform to the design requirements of 40 CFR 60.18 are assumed by rule to achieve 98% destruction efficiency for C4 HRVOCs (1,3-butadiene and butenes) and 99% destruction efficiency for C2-C3 HRVOCs (ethylene and propylene).21 The design requirements of 40 CFR 60.18 include the following:

• Flame present at all times.

• Minimum net heating value of the gas being burned of 300 Btu/scf (steam or air-assisted) or 200 Btu/scf (non-assisted).

• Maximum exit velocity of 60 feet per second for steam-assisted or non-assisted flares. Velocity limits for air-assisted flares are dependent upon the net heating value of the gas being combusted.

Flares are commonly used to control HRVOC emissions at petroleum refineries, chemical plants, olefins manufacturing plants, polymer manufacturing plants and for-hire storage terminals. Review of the RBLC identified use of flaring to control emissions from product storage and product loading at the Chevron Phillips Chemical Company Pasadena Plastics Plant. The permit for this application was dated December 15, 1998.

Flaring of HRVOC emissions from extruders and finishing operation vent streams has been limited for a number of reasons, including: cost of capture (e.g. installation of hooding on extruders, etc.), cost of supplemental fuel, and safety (air in flare header resulting in a potentially explosive waste stream). To illustrate the cost of supplement fuel, assume a 1,000 standard cubic feet per minute vent stream containing 200 ppm ethylene – approximately 4 tons/year – is routed to a flare for control. The heat value of the ethylene at this concentration is approximately 0.3 Btu/scf. To meet the 40 CFR 60.18 design requirement of 300 Btu/scf for a steam-assisted flare, this vent stream will either need to be combined with a higher heat content vent stream prior to flaring or supplemented with a fuel, such as natural gas. If a supplementary fuel is used, approximately 429 scfm of natural gas will need to be added to the vent stream to raise the heat content to 300 Btu/scf. On an annual basis, this equates to approximately 225.5 million scf (MMscf). Assuming a natural gas price of $9.00 per MMBtu, supplemental fuel for flaring this 1,000 scfm vent gas stream will cost approximately $2,000,000 per year. The

20 USEPA Air Pollution Control Technology Fact Sheet EPA-452/F-03-019 21 30 TAC 115.725(d) contains several references to these control efficiencies.



incremental cost of control for this vent stream using a flare, just considering the cost of the supplemental fuel, is approximately $500,000 per ton. This cost of control would apply to any 200 ppm ethylene vent stream that is flared. Additionally, the use of supplemental fuel would result in additional CO and NOx emissions from the flare.

3.4.2 Thermal Oxidation There are two general types of thermal oxidizers (TO) in common use: regenerative and recuperative. A regenerative thermal oxidizer, or RTO, uses a high-density media such as a ceramic packed bed still hot from a previous cycle to preheat an incoming waste gas stream. The preheated waste gases then enter a combustion chamber where they are heated by auxiliary fuel combustion (e.g. natural gas) to a final oxidation temperature typically between 1,400 and 1,500°F. The hot exit gases are directed to one or more ceramic packed beds where the heat from the gases is absorbed before they are vented to the atmosphere. An RTO will typically achieve a control efficiency of 95 to 99%.22

Recuperative TOs are comprised of the combustion chamber, waste gas preheater and, if appropriate, a secondary energy recovery preheater. Recuperative TOs can recover up to 70% of the waste heat from the exhaust gases and achieve destruction efficiencies ranging from 98% to as high as 99.9999%.23

The typical design conditions required to achieve at least 98% destruction efficiency in a recuperative TO are:

• Minimum combustion temperature of 1,600°F,

• Combustion chamber residence time of 0.75 second, and

• Proper mixing.

Figures 3 and 4 present typical configurations for regenerative and recuperative TOs, respectively.24

22 USEPA Air Pollution Control Technology Fact Sheet EPA-452/F-03-021 23 USEPA Air Pollution Control Technology Fact Sheet EPA-452/F-03-020 24 USEPA Air Pollution Cost Control Manual, 6th Edition, EPA-452/B-02-001, January 2002



Figure 3. Typical Configuration – Regenerative Thermal Oxidizer

Figure 4. Typical Configuration – Recuperative Thermal Oxidizer

Table 8 summarizes the findings of the RBLC review (last 10 years) with respect to control of HRVOC emissions from polymer plants using thermal oxidation. This listing is not comprehensive. Sources that did not undergo federal NSR, either Prevention of Significant Deterioration (PSD) or Nonattainment NSR (NNSR), will not be listed in the RBLC.



Table 8. Examples of Thermal Oxidation Used to Control HRVOC Emissions

Company: Fagerdala Pac-Lite Incorporated Location: St. Clair, Michigan Permit Date: 02/01/2001 Process Description: Expandable polypropylene bead production. Control Application: Emissions from the fluidized bead dryer and regrind extruder are

controlled by thermal oxidizer. Die area is hooded. Control Efficiency: 85% Company: Formosa Plastics Corporation Location: Point Comfort, Texas Permit Date: 03/09/1999 Process Description: Polypropylene plant with multiple trains with two reactors each. Control Application: Process off-gases are routed to incinerator. Reactor gases are

routed to flare header in case of an upset. Control Efficiency: Unknown. HRVOC emissions are limited to 31.5 lb/MMlb for Train 4,

133 lb/MMlb for Trains 1-3. Company: Total Petrochemicals USA (formerly Atofina Petrochemicals Inc.) Location: La Porte, Texas Permit Date: 11/05/2001 Process Description: Polypropylene production Control Application: Backup thermal oxidizer. Areas of process controlled are not

identified. Control Efficiency: 99.99% Company: Goodyear Tire and Rubber Company Location: Beaumont, Texas Permit Date: 02/19/2004 Process Description: Styrene butadiene rubber production Control Application: Regenerative thermal oxidizer. Application is unspecified. Control Efficiency: Not specified

3.4.3 Catalytic Oxidation Catalytic oxidizers operate similar to thermal oxidizers with the primary difference being that, after passing through the flame zone, the waste gases pass through a catalyst bed. The catalyst has the effect of increasing the reaction rate, enabling oxidation of the organics in the waste stream at a lower reaction temperature than would be required in a thermal oxidizer to achieve the same destruction efficiency. Catalysts also allow for a reduced residence time and, thus, allow for smaller oxidizers. The waste gas is typically heated to between 600°F and 800°F



Waste Gas Pre-heater

Combustion Chamber

Catalyst Chamber

Stack

Waste Gas

AuxiliaryFuel

Pre-Heated Waste Gas

before entering the catalyst. Control efficiencies as high as 95-99% can be achieved with catalytic oxidation. 25

Catalytic oxidizers are subject to plugging as well as catalyst deactivation or poisoning. Therefore, they are not as widely applicable as thermal oxidizers. As with any control approach, however, technical and economic feasibility can only be evaluated on a case-by-case basis.

Figure 5 presents a typical configuration for a catalytic oxidizer. 26

Figure 5. Typical Configuration – Catalytic Oxidizer

Review of the RBLC did not identify any application of catalytic oxidizers to control emissions of HRVOC from polymer plants in the last 10 years. However, as noted previously, this listing is not comprehensive. As discussed in Section 4 of this report, catalytic oxidizers have been used at sites in Harris County to control emissions of HRVOC.

3.4.4 Boilers and Process Heaters Boilers and process heaters, under certain process conditions, may be used to combust waste streams containing HRVOC. Considerations in burning HRVOC waste gas streams in boilers and process heaters include the following: 27

• Most chemical plants, olefins manufacturing plants and polymer plants do not have fuel gas headers that facilitate collection of waste gases for use as fuels. This may limit or eliminate consideration of this control option.

• Boilers designed specifically for HRVOC control use discrete or vortex burners.28

25 USEPA Air Pollution Control Technology Fact Sheet EPA-452/F-03-021 26 EPA Air Pollution Cost Manual, 6th Edition, EPA-452/B-02-001, January 2002 27 Some of the items listed are derived from discussions with industry personnel. 28 Polymer Manufacturing Industry, Background Information for Proposed Standards, EPA-450/3-83-019a, September 1985.



• Use of high concentration ethylene streams as a fuel tends to result in more nitrogen oxides (NOX) formation than the burning of natural gas.29

• The combustion characteristics of certain olefin derivatives, such as propylene oxide, make it unsuitable for combustion in boilers or process heaters because it combusts explosively. This characteristic could also limit the use of thermal or catalytic oxidizers.

• Olefins such as ethylene and propylene may polymerize in a fuel gas header, resulting in plugged lines with associated safety and performance implications.

While not common, there are examples of boilers being used to control emissions from polymer plants. In a 1985 document, EPA identifies two polypropylene plants and a high density polyethylene (HDPE) plant that route off-gases to a boiler for control. 30

In addition, ENVIRON identified one facility during review of the RBLC (last 10 years) that was permitted to use a boiler to control emissions of HRVOC from a polymer production facility (Table 9). Note that this listing is not comprehensive. Sources that did not undergo federal NSR, either PSD or NNSR, will not be listed in the RBLC.

Table 9. Examples of Boilers or Process Heaters Used to Control HRVOC Emissions Company: Total Petrochemicals USA (formerly Atofina Petrochemicals Inc.) Location: La Porte, Texas Permit Date: 11/05/2001 Process Description: Polypropylene production Control Application: Waste heat boiler and regenerative gas heater. Areas of process

controlled are not identified. Control Efficiency: 99.99%

In addition to the application identified in Table 9, within the description for a project at the Chevron Phillips Chemical Company Pasadena, Texas, plant (permit date of 02/23/2000) is the following:

“The Phillips Chemical Company seeks authorization to use certain PE and PP process off gases as fuel at existing flares located within their Houston Chemical Complex. The process off gases are generated on-site at process units and were used as fuel at four on-site boilers; however, the boilers are being permanently shutdown . . . Certain PE/PP off gases will be routed to the flare fuel gas system for use as flare fuel gas.”

29 Ethylene has a high heat content: approximately 1,600 Btu/scf compared with approximately 1,000 Btu/scf for methane (natural gas). Consequently, unless specifically designed for burning ethylene, the combustion device will burn hotter and have higher NOX emissions. 30 EPA-450/3-83-019a



It is ENVIRON’s understanding that as part of their strategy for complying with the NOX Mass Emission Cap and Trade (MECT) program, Chevron Phillips Chemical shut down their boilers and started purchasing steam from a nearby cogeneration facility. As documented in the RBLC, however, boilers were used for controlling PE and PP emissions prior to that time.

If technically and economically feasible, due to the high temperature and long residence times typical of boilers and process heaters, high destruction efficiencies (greater than 98%) can be achieved.31

The use of flares, thermal oxidizers, catalytic oxidizers and/or boilers to reduce HRVOC emissions from polymer plants is built upon the assumption that the uncontrolled emissions can be effectively captured. The effective, efficient and safe capture of reactive monomer streams must be evaluated on a case-by-case basis.

3.4.5 Bekaert Burners Another alternative to flaring is use of Bekaert CEB® burners32. Bekaert burners use a meshed fiber surface that divides the main flame into tiny flames thereby resulting in more complete combustion of the HRVOC in the waste gas stream. Bekaert reports that HRVOC destruction efficiencies as high as 99.99% have been achieved at operating temperatures of 2,000-2,200°F with NOX emissions less than 15 ppm. The capacity on the largest burner model Bekaert currently makes is approximately 2,500 scfm. As discussed in Section 4, with one exception, the annual average flare flowrates at the Harris County polymer production facilities surveyed is less than 2,500 scfm. However, to handle emergencies and other large releases, the flare design capacities are much greater than 2,500 scfm.

Based on information provided by Bekaert, there are six of their systems currently in use in the Houston area. Four of these are at petrochemical plants. The largest system in use is used to control emissions (non-HRVOC) from barge loading and unloading operations. Since the burner uses fine fiber mesh, presence of particulate matter in the waste gas stream may clog the flame openings thereby impairing performance.

Figure 6 shows in-field installation of two Bekaert CEB® Model 4500 burners (Source: Bekaert website). Currently, these are Bekaert’s largest models.

31 Ibid 32 Information on Bekaert burners is derived from discussions with Mr. Timothy F. Egan, Bekaert Corporation, and from their website: http://www.bekaert.com/flaring

Figure 6. Bekaert Burners



3.4.6 Stripping Stripping is used in SBR production to remove un-reacted monomer from the rubber crumb, with the overheads routed to a combustion device for destruction. Stripping is common in SBR production facilities; however, the intent of the stripping is primarily to remove the un-reacted styrene. The un-reacted 1,3-butadiene is removed through flash distillation and reduction in system pressure.

Review of the RBLC identified two facilities where stripping is used to control emissions from SBR production facilities. Table 10 summarizes the findings.

Table 10. Examples of Stripping Used to Control HRVOC Emissions Company: Firestone Polymers LLC Location: Lake Charles, Louisiana Permit Date: 07/30/2003 Process Description: Styrene butadiene rubber production Control Application: Steam stripping of solvent from crumb rubber. Emissions stream

from stripping operation is collected and routed to flare Control Efficiency: Unknown Company: Goodyear Tire and Rubber Company Location: Beaumont, Texas Permit Date: 02/19/2004 Process Description: Styrene butadiene rubber production Control Application: Air stripping. Application is unspecified. Control Efficiency: Not specified

While search of the RBLC did not identify any instances of either air or steam stripping being used to control emissions from PE or PP production facilities, as discussed in Section 4, nitrogen stripping is used to remove un-reacted monomer in PE and PP production facilities.

3.4.7 Adsorption / Concentration Adsorption is the attachment of gaseous molecules to the surface of a solid. During adsorption, a gas molecule migrates from the gas stream to the surface of the solid where it is held by physical attraction. Adsorption in the form of a concentrator can be used to raise the concentration of an organic vapor to provide more economical treatment in downstream combustion or condensation devices. Activated carbon is the most widely used adsorbent for VOCs. Other adsorbents include zeolites and certain synthetic polymers. 33

33 Choosing an Adsorption System for VOC: Carbon, Zeolite, or Polymers? EPA Technical Bulletin, EPA 456/F-99-004, May 1999.



Factors that influence the performance of activated carbon in controlling gas phase VOC emissions include:34

• The type of compound to be removed. In general, compounds with a high molecular weight and higher boiling point are better adsorbed.

• Concentration. The higher the concentration the better the adsorption.

• Temperature. The lower the temperature the greater the adsorption capacity.

• Pressure. The higher the pressure the greater the adsorption capacity.

• Humidity. The lower the humidity the greater the adsorption capacity.

As shown in Table 11, HRVOCs are low molecular weight compounds with low boiling points.35

Table 11. Physical Properties of HRVOCs

Compound Molecular Weight Boiling Point (°C) Ethylene 28 -104

Propylene 42 -48

1,3-Butadiene 54 -4

1-Butene 56 -5

2-Butene (cis & trans) 56 +3

Isobutylene 56 -7

Vents from extruders and downstream operations will typically have low HRVOC concentrations and pressures close to ambient. Collectively, this information indicates that activated carbon would, most likely, be a poor choice for either direct control of HRVOC emissions or for use in a concentrator. The limitations affecting adsorption using activated carbon would also be expected to affect adsorption using zeolites or polymer adsorbents.

Concentrator suppliers have stated that, for a concentrator to be effective, the boiling point of the material to be adsorbed should be higher than the inlet gas phase temperature, but lower than the desorption temperature. For applications involving the HRVOCs listed in Table 11, the boiling points would be less than the desorption temperature; however, they would not be above the inlet temperature. This “rule of thumb” confirms that HRVOCs are not amenable to adsorption or concentration.

34 EPA Air Pollution Cost Manual. 6th Edition, EPA/452/B-02-001, January 2002. 35 Chemical Engineer’s Handbook, 5th Edition, edited by Robert H. Perry and Cecil H. Chilton, McGraw-Hill Book Company, 1973.



Review of the RBLC did not identify any application of adsorption for the control of HRVOC emissions from polymer plants in the last 10 years.

3.4.8 Biofiltration / Bioscrubbing Biofiltration involves routing a vent gas stream through a biologically active media, similar to compost, where the pollutants of interest are adsorbed and/or absorbed and biologically degraded into water and carbon dioxide. A bioscrubber is similar in function; however, the control system involves a tower packed with synthetic media that supports a biological culture. Biofiltration and bioscrubbing have been successfully applied in a number of full-scale applications to control odors, VOC and emissions of air toxics from a wide range of sources. Application of biofiltration and bioscrubbing are typically limited to relatively low concentration vent streams – approximately 1,000 ppm or less – and the pollutants need to be water soluble.36 HRVOC compounds are generally slightly soluble to insoluble in water, making them poor candidates for control through biofiltration.

3.4.9 Refrigerated Condensers A refrigerated condenser is a control device that is used to cool an emission stream containing organic vapors and to condense the organic vapors into a liquid that is then collected and either recycled or disposed of. Refrigerated condensers work best on emission streams containing high concentrations of VOC. To achieve any reduction, even on saturated streams, the condenser must achieve a temperature that is lower than the boiling point of the compound in question.

HRVOC emissions from extruders and downstream operations are poor candidates for control through use of refrigerated condensers because: 1) the boiling points are low (refer to Table 11) and 2) the concentration of HRVOC in the vents is expected to be low.

3.4.10 Non-Thermal Plasma Low-temperature, non-thermal plasma (NTP) is a developing technology that may, in the future, be an option for controlling emissions of HRVOC.37 The basic principle of NTP is the use of electricity to create a plasma – an ionized gas containing free electrons.38 These energetic electrons excite, dissociate and ionize molecules to produce chemically active radicals and ions. In the laboratory, NTP has been shown to have ability to destroy a number of different VOCs and polycyclic aromatic hydrocarbons (PAHs).39 Pilot-scale experiments conducted at pulp mills

36 EPA Survey of Control Technologies for Low Concentration Organic Vapor Gas Streams. EPA/456/R-95-003, May 1995. 37 Pulsed Corona Plasma Pilot Plant for VOC Abatement in Industrial Streams as Mobile and Educational Laboratory.

Tak, G., Gutsol, A. and Fridman, A., 2005. (http://plasma.mem.drexel.edu/publications/documents/ISPC-ID670.pdf) 38 Application of Non-Thermal Plasma for Air Pollution Control

(http://miedept.mie.uic.edu/lab/kennedy/Application_Plasma_4.htm) 39 Pulsed Corona Plasma Pilot Plant for VOC Abatement in Industrial Streams as Mobile and Educational Laboratory.

Tak, G., Gutsol, A. and Fridman, A., 2005. (http://plasma.mem.drexel.edu/publications/documents/ISPC-ID670.pdf)



and wood product plants have shown VOC destruction efficiencies greater than 98%.40 In theory, NTP could be used to treat industrial waste gas streams containing HRVOC across a wide range of flow rates.41 NTP, however, has not been demonstrated on a commercial scale nor has it been demonstrated to control emissions of HRVOC.

3.5 Flare Minimization Flare minimization refers to the reduction in the number of instances of flaring, both during routine operations and during startup, shutdown and malfunction, and to the reduction in the quantity of material flared. The concept of flare minimization applies to both stream recycling/reuse and flare gas recovery.

3.5.1 Recycling/Reuse As discussed in Section 2.2, a significant percentage of HRVOC emissions occur during maintenance, startup, and shutdown (MSS) activities and during emission events. Through certain capital investments (e.g. the addition of process loops and storage capacity) and changes in the way that the production units are managed during MSS activities, the amount of HRVOC released to the flare header can be significantly reduced. Following are two examples of recycling and reuse at Harris County facilities.

• One Harris County olefins producer implemented a flare minimization program that resulted in reduced flaring during the shutdown and startup of the unit. The shutdown and startup of the unit is a sequence of steps where each section of the process is shutdown or started before the next section. Past practice had been to vent to the flare during this sequence until the unit was gas-free during shutdown or until producing on-specification product when going through startup. The operator made modifications to the process that allowed for streams to be recycled within the unit that dramatically reduced the amount of material sent to the flare during shutdown and startup.

• As discussed in the HARC Project H-76 report, as of 2006, one Harris County olefins producer was planning on sending off-specification HRVOC to an off-site salt dome storage facility for later reprocessing. This would result in a reduction in HRVOC emissions from flaring by approximately 17 tons/year at a projected capital cost of approximately $700,000.42

40 Pulsed Corona Plasma Technology for Treating VOC Emissions from Pulp Mills, July 28, 2004.

(http://www.osti.gov/energycitations/servlets/purl/826442-clriuJ/826442.PDF) 41 Destruction of Highly Diluted Volatile Organic Compounds (VOCs) in Air by Dielectric Barrier Discharge and

Mineral Bed Adsorption. Martin, L., Ognier, S., Gasthauer, E., Cavadias, S., Dresvin, S. and Amouroux, J. Energy & Fuels. Vol. 22, 576-582, 2008.

42 How Chemical Manufacturing and Petroleum Refining Facilities in Harris County are Using Point Source Monitoring to Identify and Reduce HRVOC Emissions, HARC Project H-76, ENVIRON International Corporation, prepared for the Houston Advanced Research Center, October 2006.



3.5.2 Flare Gas Recovery Flare Gas Recovery (FGR) refers to taking low pressure waste gases in the flare header, compressing the gases, and then reprocessing them or using them as a fuel gas in the plant. When the flow is less than or equal to the capacity of the FGR system, the flare gas will be recovered. During these periods of normal operations, emissions from the flare will approach zero. When the flare gas flow rate is greater than the capacity of the FGR system, the excess flare gas will flow through a liquid seal drum and to the flare tip for combustion. FGR systems are designed for recovering waste gases during normal operations. During non-routine operating conditions (e.g. MSS and emission events), excess waste gas will flow to the flare for combustion.

Potential benefits of FGR include:

• Waste gas may have substantial heating value and could be used as a fuel source in the plant, thereby reducing fuel purchase costs;

• Waste gas could be used as feedstock or product in certain applications; and

• Emissions from flaring would be reduced.

FGR is widely used in petroleum refineries. As part of multi-facility, “global” consent agreements with the USEPA, a number of major petroleum refining companies have committed to installing FGR at one or more of their refineries. The Harris County refineries that are part of global consent agreements with EPA are:

• ExxonMobil Baytown Refinery,

• Shell Deer Park Refinery, and

• Valero Houston Refinery.43

The ExxonMobil, Shell and Valero global consent agreements do not require the installation of FGR at these refineries. However, the ExxonMobil and Shell agreements specify compliance with the emission limits of New Source Performance Standard (NSPS) Subpart J (40 CFR 60.104(a)); specifically the provision that

“No owner or operator subject to the provisions of this subpart shall: (1) Burn in any fuel gas combustion device any fuel gas that contains hydrogen sulfide (H2S) in excess of 230 mg/dscm (0.10 gr/dscf).”

Compliance may require the installation of FGR on some flares. Assuming the FGR system is sized to handle worst-case flows during normal operations, emissions from these flares during normal operations should be limited to pilot gas combustion, or very close to zero.44

43 Petroleum Refinery Consent Decree Emission Reduction Assessment for Ozone and Regional Haze SIPs, ENVIRON International Corporation, prepared for the Texas Commission on Environmental Quality, November 2007.



There are a number of potential limitations to use of FGR in olefins, derivatives (e.g. propylene oxide production) and polymer manufacturing facilities. These include:45

• Refineries have fuel gas headers that collect high heat content waste streams from around the refinery, compress it, and send it to boilers and/or process heaters for use as a fuel. Olefins, derivatives and polymer production facilities do not typically have this same infrastructure.

• The combustion characteristics of high concentration olefin or olefin derivative streams may not be conducive to use as a fuel. For example, propylene oxide burns explosively.

• Use of the olefin or olefin derivative as a fuel may result in undesirable environmental impacts. For example, ethylene burns very hot and results in the formation of excess NOX.

• The olefins may polymerize in the flare header, leading to plugging with associated safety and performance implications.

As with all other emission control approaches, the technical and economic feasibility of flare gas recovery must be evaluated on a case-by-case basis.

44 Ibid. 45 Developed during discussions with industry representatives during the survey portion of the project.



4 HRVOC Emission Reduction Projects Survey Findings As discussed in the Introduction, ENVIRON prepared and sent questionnaires to certain Harris County industrial sites for the purpose of gathering information on projects that have been implemented to reduce HRVOC emissions at those sites, the costs of those projects, and on other HRVOC emission reduction projects that have been considered but not implemented. For Phase 1, ENVIRON developed separate questionnaires for flare issues and polymer production issues. For Phase 2, ENVIRON prepared a single questionnaire pertaining to both HRVOC emission reduction projects and flares that was sent to another set of Harris County facilities selected by the TCEQ. Both the Phase 1 polymer production issues and flare questionnaire templates and the Phase 2 questionnaire are included in Appendix A.

For Phase 1, the TCEQ identified 16 sites in Harris County to receive one or both of the surveys. Of the 16 sites, 11 participated by responding to the surveys and/or meeting with ENVIRON personnel to discuss their responses. Participating sites included 4 olefin plants and 9 polymer plants. Two sites, out of the 11 that participated in Phase 1, are involved in the manufacture of both polymers and olefins. Additionally, one site that did not provide a response to the survey suggested that information obtained as part of HARC Project H76 be included in this investigation. That site contains both petroleum refining and olefins production operations.

For Phase 2, the TCEQ identified 38 sites in Harris County to receive the survey. Of the 38 sites, 24 participated by responding to the survey. Participating sites included 12 chemical manufacturing facilities; 8 terminals; 2 polymer facilities (flare survey only); and 2 refineries.

Synthesized survey findings from Phases 1 and 2 are presented and discussed within this section.

4.1 Facility Questionnaire Results

4.1.1 Industry Types Surveyed ENVIRON received survey responses from the following types of facilities:

• Polymer manufacturing facilities

- Polyethylene – low density, high pressure process

- Polyethylene – low density, low pressure process

- Polyethylene – high density, gas phase process

- Polyethylene – high density, liquid-phase slurry process

- Polyethylene – high density, liquid-phase solution process

- Polypropylene – liquid-phase slurry process

- Polypropylene – gas phase process

• Chemical, including olefins and non-olefins, manufacturing facilities



• Petroleum refineries

• Storage terminals

The number of facilities represented by the above-listed categories is presented in Figure 7.

Figure 7. Type and Number of Facilities

4.1.2 Process Vent Control Techniques ENVIRON asked survey respondents whether process vents were recycled back to the process or routed to a control device. For Phase 1 polymer plants, process vents include those located upstream of the extruder.

Responses are summarized in Table 12. The number of process units utilizing each control technique is presented in the table. All polymer plants surveyed operate multiple process units.

Table 12. Summary of Process Vent Control Techniques by Process Type for Polymers

Industry Sector Recycled Flare Boiler Process Heater

Thermal Ox FGR CatOx

PE – low density, high pressure 2 1 1 0 1 0 1

PE – low density, low pressure process 1 1 0 0 0 0 1

PE – high density, gas phase process 1 2 0 0 1 0 1

PE – high density, liquid-phase slurry 11 12 8 0 0 0 0

PE – high density, liquid-phase solution 1 1 0 1 1 0 0

0 2 4 6 8 10 12 14 16 18

Chemicals

Polymers

Refinery

Terminal

Indu

stry

Sec

tor

Number of Facilities



Table 12. Summary of Process Vent Control Techniques by Process Type for Polymers

Industry Sector Recycled Flare Boiler Process Heater

Thermal Ox FGR CatOx

PP – liquid-phase slurry 6 6 1 0 0 3 0

PP – gas phase 5 5 1 0 0 2 0

For Phase 2 facilities, 11 process vents are recycled to process; 369 process vents are routed to a control device; and 258 process vents are uncontrolled.

4.1.3 Finishing Operations ENVIRON asked Phase 1 polymer plant survey respondents whether finishing vents (i.e., extruder and downstream to storage and loading) are routed to a control device. Nine polymer facilities responded to this part of the survey.

• No Control. Of the nine facilities that responded to this part of the survey, five facilities have no control on the extruder, product storage or loading operations. Emissions from these facilities are from uncontrolled atmospheric vents.

• Thermal Oxidation. Two facilities use thermal oxidizers to control HRVOC emissions from the extruders and/or downstream operations. One facility routes the emissions from extruder vents to a thermal oxidizer. The other facility routes emissions from dryer vents and storage silos to a thermal oxidizer.

• Catalytic Oxidation. One facility routes intermediate storage emissions generated from the high-pressure LDPE manufacturing process to a catalytic oxidizer.

• Flare. One facility routes its dryer vents to a flare for control, but other finishing vents are uncontrolled.


Project Number 06-17477O 44


Project ID Project Name Industry Sector Capital Cost($)

HRVOC Reduction

(tpy) Phase 1 or

Phase 2

P1 Vent Gas Recovery Polymers (Polyethylene – low density, high pressure process) 650,000 40 Phase 1

P2 Ethylene Recovery Unit Polymers (Polypropylene – gas phase process) 1,000,000 15 Phase 1

P3 De-inventory to propylene storage1 Polymers (Polypropylene – liquid-phase slurry process) 50,750 See note 1 below Phase 1

P4 Installation of Regenerative Thermal Oxidizer

Polymers (Polyethylene – low density, high pressure process) 11,500,000 143 Phase 1

P5 Changes to startup procedure2 Polymers (Polyethylene – liquid-phase process) 0 0.5 Phase 1

P6 Replacement of Catalytic Oxidizer with Thermal Oxidizer with higher DRE

Polymers (Polyethylene – liquid-phase solution process) 364,000 0.5 Phase 1

P7 Re-routing of extruder vents from carbon beds to thermal oxidizer

Polymers (Polyethylene – liquid phase process) 127,000 1 Phase 1

P8 Installation of PSA system Polymers (Polypropylene – liquid-phase slurry and gas-phase processes) 7,000,000 42 Phase 1

P9 PSA system operability improvements Polymers (Polypropylene – liquid-phase slurry and gas-phase processes) 400,000 1 Phase 1

P10 Implementation of more efficient purge bin distributor design

Polymers (Polypropylene – liquid-phase slurry and gas-phase processes) 800,000 20 Phase 1

P11 Routing of relief devices to flare3 Polymers (Polypropylene – gas phase process) 175,000 See note 3 below Phase 1

P12 Installation of HRVOC caps and plugs3 Polymers (Polypropylene –gas-phase process) 75,000 See note 3 below Phase 1

P13 Implementation of leak detection probe3 Polymers (Polypropylene – gas phase process) 8,000 See note 3 below Phase 1





HRVOC Reduction

(tpy) Phase 1 or

Phase 2

P14 Installation of HRVOC awareness monitors3

Polymers (Polypropylene – gas phase process) 15,000 See note 3 below Phase 1

P15 Installation of HRVOC sample points and caps and plugs4

Polymers (Polypropylene – liquid-phase slurryand gas-phase processes) 93,500 See note 4 below Phase 1

P16 Installation of pump trap on compressor4 Polymers (Polypropylene – liquid-phase slurry and gas-phase processes) 130,000 See note 4 below Phase 1

P17 Routing of pump off-gas flow to Flare Gas Recovery system4

Polymers (Polypropylene – liquid-phase slurry and gas-phase processes) 47,100 See note 4 below Phase 1

P18 Installation of HRVOC monomer efficiency monitors4


P19 Implementation of leak detection probe4 Polymers (Polypropylene – liquid-phase slurry and gas-phase processes) 6,000 See note 4 below Phase 1

P20 Implementation of atmospheric PSV monitoring4


P21 Enhancement of propylene storage sample system4


P22 Flare Gas Recovery system upgrade4 Polymers (Polypropylene – liquid-phase slurry and gas-phase processes) 321,000 See note 4 below Phase 1

P23 Flare Gas Recovery system upgrade4 Polymers (Polypropylene – liquid-phase slurry

and gas-phase processes)

20,000 See note 4 below Phase 1

P24 Improve Butene Utilization Chemicals 122,000 4.9 Phase 2

P25 Reduce Ethylene Emissions from Atmospheric Tanks Chemicals 149,000 12.6 Phase 2

P26 Install a Pressure Controller Chemicals 0 NS5 Phase 2

P27 Optimize Vent Gas Compressor Timing Chemicals 0 0.3 Phase 2





HRVOC Reduction

(tpy) Phase 1 or

Phase 2

P28 Convert Atmospheric Vent to ‘Emergency Only’ Vent Chemicals 31,000 0.25 Phase 2

P29 Modify Emergency Vent Control Logic Chemicals 28,000 NS5 Phase 2

P30 Install a Compressor Chemicals 500,000 20 Phase 2

P34 Install a Flare Stack Refinery 40,700,000 NS5 Phase 2

P35 Coker Blow-down Routed to Flare Refinery 1,200,000 13.5 Phase 2

P36 Online Monitoring Chemicals (Olefins) NS 5 Phase 2

P37 Bundle Upgrades Chemicals (Olefins) NS 15 Phase 2

P38 Vent Gas Recovery Chemicals (Olefins) NS 45 Phase 2

P39 Fenceline Monitoring Chemicals (Olefins) NS 15 Phase 2 1 Estimated HRVOC reductions between 1 - 3 tph depending on frequency and duration of shutdown. Facility did not provide detailed information regarding the frequency and duration of shutdown; therefore, annualized HRVOC reductions were not estimated. 2 Startup performed once every two to three years. The estimated HRVOC reductions are annualized. 3 Facility provided total HRVOC emission reductions attributed to projects P11 – P14 of 27.14 tpy. However, HRVOC emission reductions attributed to each project were not available. 4 Facility provided total HRVOC emission reductions attributed to projects P15 – P23 of 21.67 tpy. However, HRVOC emission reductions attributed to each project were not available. 5 Facility was unable to estimate HRVOC emission reductions due to the project. Therefore, HRVOC emission reductions are not specified.



4.1.4 HRVOC Emission Reduction Projects ENVIRON asked Phase 1 and Phase 2 survey respondents whether any projects had been implemented to reduce emissions in response to the HRVOC rules. Projects could include, but were not limited to, process changes, changes in operating procedures, vent stream controls and/or flare minimization. Excluded from the survey were costs associated with installation of HRVOC monitoring equipment and any emission reductions that may have resulted from more robust monitoring of emissions. A detailed analysis of the costs of control and HRVOC emission reductions associated with the installation of HRVOC monitoring equipment is included in How Chemical Manufacturing and Petroleum Refining Facilities in Harris County Are Using Point Source Monitoring to Identify and Reduce HRVOC Emissions.46

Table 13 summarizes the projects identified by Phase 1 and Phase 2 survey respondents. Additional details regarding the projects referenced in Table 13 are provided below.

• Project P1. The facility collected emissions from 8 continuous hourly production silos. Each storage silo stores approximately one hour’s worth of production. HRVOC emissions collected from these intermediate storage silos are routed to a catalytic oxidizer for control. Previously, emissions from these production silos were uncontrolled. HRVOC emission reductions due to the implementation of this project are estimated to be approximately 40 tpy.

• Project P2. The facility installed an ethylene recovery unit on an off-gas stream to recover up to 3,000,000 pounds of ethylene from the flare header system. Taking into account an assumed 99% destruction and removal efficiency (DRE), the recovery of 3,000,000 pounds of ethylene translates to post-flare HRVOC emission reductions of approximately 15 tpy.

• Project P3. This project consisted of piping modifications which allowed the facility to recycle liquid slurry back to monomer storage instead of flaring during shutdown. Depending on the duration of shutdown, HRVOC emission reductions are estimated to be between 1 and 3 tons per hour (tph) during the event.

• Project P4. The facility installed an RTO for MON/HRVOC compliance requirements. Post-extruder HRVOC emissions from the pellet dryer vent, compressor distance pieces and the pellet silo storage vents are routed to the thermal oxidizer. Previously, these emission sources were uncontrolled. HRVOC emission reductions due to the implementation of this project are estimated to be approximately 143 tpy.

• Project P5. The facility implemented procedural changes to pressure check the reactor before startup. Isobutane is used instead of ethylene as part of the reactor start-up procedure, thereby reducing the use of ethylene during this process. This operational

46 How Chemical Manufacturing and Petroleum Refining Facilities in Harris County are Using Point Source Monitoring

to Identify and Reduce HRVOC Emissions, HARC Project H76, ENVIRON International Corporation, prepared for the Houston Advanced Research Center, October 2006.



procedure is repeated once every two to three years. Annualized HRVOC emission reductions are estimated to be approximately 0.5 tpy.

• Project P6. The facility replaced an existing catalytic oxidizer with a thermal oxidizer that achieves a higher DRE. HRVOC emission reductions due to this project are estimated to be approximately 0.5 tpy.

• Project P7. The facility rerouted the extruder vents from existing carbon adsorber beds to a thermal oxidizer. HRVOC emission reductions due to this rerouting are estimated to be approximately 1 tpy.

• Project P8. The facility installed a Pressure Swing Absorption (PSA) system to capture all continuous vent streams that were previously routed to the flare for control. These vent streams consist primarily of nitrogen (75%), with the remainder being propylene (25%). The PSA system recovers, condenses and recycles propylene back to the process in liquid form. Also, nitrogen is recycled back to the process. HRVOC emission reductions due to this project are estimated to be approximately 42 tpy.

• Project P9. The facility made improvements to the operability of its PSA system and routed additional small vents (e.g., analyzer vents, dry gas seals) to the PSA. HRVOC emission reductions due to this project are estimated to be approximately 1 tpy.

• Project P10. The facility implemented a more efficient distributor design on its purge bin. The redesigned distributor results in greater volatilization of the monomer from the polypropylene flake prior to the flake entering atmospheric storage vessels, thereby reducing atmospheric emissions. HRVOC emission reductions due to this project are estimated to be approximately 20 tpy.

• Projects P11 through P14. The facility implemented several projects related to reducing atmospheric emissions of HRVOC from atmospheric relief valves and fugitive components.47 Projects included the following:

- Routing of atmospheric relief devices to flare. Pressure relief devices on the polypropylene dryers, which were previously routed to the atmosphere, were tied in to the flare header.

- Installation of HRVOC caps and plugs. All the caps and plugs in VOC service were replaced with plugs painted fluorescent yellow for easy identification and were tethered to a cable so that they could be found easily if not in place.


47 Unlike other polymer facilities surveyed, this facility implemented projects primarily related to fugitive emission

reductions and monitoring.



- HRVOC awareness monitors. Project involved installation of a monitor in the control room to allow for continual monitoring of the monomer efficiency, HRVOC permit compliance and energy utilization at the unit.

Additionally, process improvements and improvements in operational reliability, mechanical integrity and monomer efficiency have contributed to HRVOC emission reductions at the facility. HRVOC emission reductions due to the capital projects and other improvements are estimated to be 27.14 tpy.

• Projects P15 through P23. The facility implemented several projects related to reducing atmospheric emissions of HRVOC from atmospheric relief valves and fugitive components and to upgrading the existing Flare Gas Recovery (FGR) system. Projects included the following:

- Installation of HRVOC sample points and caps and plugs. Project involved the installation of HRVOC sampling points on various process vent streams. The project also involved purchasing miscellaneous piping caps and plugs for lines in hydrocarbon service. All the caps and plugs in VOC service were replaced with plugs painted fluorescent yellow for easy identification and were tethered to a cable so that they could be found easily if not in place.

- Installation of pump trap on compressor. Project involved installing a skid mounted pump trap system on the sour oil/seal gas return line located in the propylene distillation section. This system allows for the separation of propylene entrained in the sour oil, allowing the propylene emissions to discharge to the flare header instead of the atmosphere.

- Routing of off-gas flow to FGR. Project involved installation of pipe with control valve to allow off-gas from compressor pump trap to be separately fed to FGR. Project was needed to optimize monomer efficiency.

- Installation of HRVOC monomer efficiency monitors. Project involved installation of monitors in control rooms to allow for continuous monitoring of HRVOC and monomer efficiency.


- Implementation of atmospheric PSV monitoring. Wireless pressure transmitters were installed on pressure relief valves, which are routed to the atmosphere, and tied into the DCS so that the time and duration of each pressure relief event could be monitored.

- Enhancement of propylene storage sample system. Sampling system on the propylene storage bullets was tied to the flare header to prevent atmospheric releases while sampling.

- Two FGR upgrades. One project improved the reliability of the FGR system from 95% to 99%. A second project improved the reliability of the FGR and improved the recovery of propylene and hexane.



Additionally, process improvements and improvements in operational reliability, mechanical integrity and monomer efficiency have contributed significantly to HRVOC emission reductions at the facility. HRVOC emission reductions due to the capital projects and other improvements are estimated to be 21.67 tpy.

• Project P24. The facility uses HRVOC feed to manufacture monomer. The project involved control of reactor vents during HRVOC feed to improve HRVOC usage and utilization and minimize losses. Previously, emissions from reactor loading operations were uncontrolled. HRVOC emission reductions due to the implementation of this project are estimated to be 4.9 tpy.

• Project P25. Due to low pressure in atmospheric pressure tanks, the carryover material (ethylene) from the reactor escapes through the atmospheric vents. In an effort to reduce these ethylene emissions, the reactor pressure is lowered when the reactor is flared and then purged with nitrogen. Low pressure helps to reduce HRVOC emissions through atmospheric vents and nitrogen purge helps to bubble-out ethylene in the reactor solution. HRVOC emission reductions of approximately 12.6 tpy were realized as a result of implementing this change in operational procedure.

• Project P26. Maintaining the cooling water pressure above the process pressure prevents ethylene emissions in case of a condenser tube leak. This project involved the installation of a pressure control device and upgrading the operating procedures in order to maintain the cooling water pressure above the process pressure. HRVOC emission reductions due to the implementation of this project were not specified by the respondent. Site personnel told ENVIRON that the site does not have HRVOC emission reduction data for this project.

• Project P27. This project was a procedural change to optimize the timing of vent gas compressor startup in order to reduce ethylene emissions to flare during process startups. Annualized HRVOC emission reductions are estimated to be approximately 0.3 tpy.

• Project P28. The facility converted an uncontrolled atmospheric vent from operation as a startup/shutdown vent to an ‘emergency only’ vent and diverted the vent stream to a flare. HRVOC emission reductions due to this project are estimated to be approximately 0.25 tpy.

• Project P29. The facility modified vent control logic and hardware to reduce the probability of unintended activation of the release vent. HRVOC emission reductions due to the implementation of this project were not specified by the respondent. Site personnel told ENVIRON that the site does not have HRVOC emission reduction data for this project.

• Project P30. The process at the facility consists of a reaction between isobutylene and various other raw materials. However, the reaction is not 100% efficient. The unreacted isobutylene used to be stripped out of the product and recycled back to the process. However, a mass balance identified that 10-20% of unreacted isobutylene was lost to the flare. The project consisted of installation of a compressor in place of the scrubbers to reduce the amount of isobutylene flared. It is estimated that only 1% of unreacted



HRVOC is currently flared. HRVOC emission reductions due to this project are estimated to be approximately 20 tpy.

• Project 31. This project consisted of installation of a flare stack to reduce short-term emissions from episodic events. The flare was primarily installed due to reliability concerns for a compressor in pure ethylene service. However, the facility anticipated limited reduction in HRVOC emissions due to the project. HRVOC emission reductions due to the implementation of this project were not specified by the respondent. Site personnel told ENVIRON that the site does not have HRVOC emission reduction data for this project.

• Project 32. This project consisted of routing coker blow-down emissions to the flare. HRVOC emission reductions due to this project are estimated to be approximately 13.5 tpy.

• Project 33 through Project 36. The facility implemented several projects related to reducing HRVOC emissions. As shown in Table 13, these projects were related to installation of online HRVOC monitors, bundle upgrades, vent gas recovery and fenceline monitoring of certain HRVOCs. However, additional details of these projects were not provided by the survey respondent. The total HRVOC emission reductions due to these projects are estimated to be approximately 80 tpy.

ENVIRON also asked respondents about HRVOC emission reduction projects that had been considered, but not implemented. Details related to those projects are discussed below.

• Storage Silo Control. This project would have reduced monomer emissions from storage silos at a polyethylene manufacturing facility (low density, high pressure process). The project would have involved the installation of multiple transfer lines to different storage silos to route low heat value waste gas streams to a catalytic oxidizer. If implemented, the project would have required a capital investment of approximately $5MM for a catalytic oxidizer to reduce HRVOC emissions by up to 40 tpy.

• Propylene Nitrogen Recovery Unit (PNRU). The implementation of the PNRU project would have reduced HRVOC emissions by recovering the monomer (propylene) and reusing it in the polypropylene production process. For safety reasons, nitrogen, which is inert, would be used to recover propylene. In a typical PNRU application, the vent stream from the resin degassing bin is compressed and then cooled to condense the propylene. The gas leaving the condenser, which contains a significant amount of propylene, is fed to a membrane unit. The membrane unit separates the stream into a propylene-enriched permeate stream and a purified nitrogen stream. The permeate stream is recycled to the inlet of the compressor and then to the condenser, where the propylene is recovered. The purified nitrogen stream is recycled to the degassing bin.48 Two different polymer manufacturing facilities have considered installation of a PNRU. Implementation of the two PNRU projects would involve capital expenditures of greater

48 http://www.mtrinc.com/polypropylene_production.html



than $1 MM and $12 MM, respectively.49 Neither facility provided any information regarding estimated HRVOC emission reductions that would have been achieved by the implementation of PNRU.

• Isobutane Nitrogen Recovery Unit (INRU). The INRU is conceptually similar to PNRU except that it would be used to recover isobutane from the polyethylene production process. In addition to recovering isobutane, ethylene would potentially be recovered using INRU. Neither cost nor potential HRVOC emission reduction information was provided by the facility that considered this project.

• Off-gas Recycle. The off-gas recycle project would recycle the reactor off-gas back to the polypropylene production process. To make the necessary modifications and process changes would require a capital investment in excess of $1 MM.

• Route Vents to Flare Header. This project would involve routing reactor vents and atmospheric relief valves to a process flare header. While not yet implemented, the survey respondent indicated that the facility is proceeding with implementation in the near future. The flare header must be modified to handle the large potential flow during emergency shutdown. The facility plans to install a distributed control system (DCS) to program the shutdown sequence. The current estimated cost for this project is $8 MM. No estimate of the reduction in HRVOC emissions to be realized was provided. Routine emissions at this facility are routed to a thermal oxidizer. This project would only affect MSS and event emissions.

• Ethylene Recovery Unit. This project would involve routing intermediate flake tanks vents to an Ethylene Recovery Unit (ERU). Currently, these intermediate tanks are uncontrolled. Because the flake tanks are designed for atmospheric pressure, the facility would have to design a pressure control scheme, install piping and auxiliary equipment in addition to the ERU. The estimated cost for this project is $10 MM. No estimate of the reduction in HRVOC emissions to be realized was provided.

• Process to Process Heat Exchangers. This project would involve rerouting the process stream containing HRVOCs from the cooling tower to a heat exchanger, thereby reducing the HRVOC emissions. However, this project would have been cost prohibitive, as defined by the site, costing in excess of $500,000 for each heat exchanger, resulting in a minimal HRVOC emission reduction.

• Recycle Raw Material. This project would recycle process stream HRVOC raw materials, at a chemicals manufacturing facility, back to the production process instead of venting to a flare. However, the project would not be feasible due to the sensitivity of the batch production process to inert materials, e.g. nitrogen, in the recycle stream.

49 The $12 MM estimate is considered more refined; the greater than $1 MM provided by one facility is considered a

rough estimate of the minimum amount of capital investment required.



• Ethylene Recovery System. This project would use membrane separation and absorption to recover ethylene from the primary purge stream. However, extensive retrofitting to incorporate this technology into the existing plant design would be cost prohibitive, as defined by the site.

• Analyzer Emissions Control. This project would involve installation of a control device to reduce propylene emissions from a Gas Chromatogram (GC) analyzer instead of venting the injected stream to the atmosphere. The project was not implemented due to the limited emission reduction potential of the project, estimated to be 0.5 tpy of HRVOC.

• Vapor Balancing System. A vapor balancing system would recover propylene and isobutylene emissions from truck loading and unloading operations instead of routing them to a flare system. However, due to space requirements, the existing 40+ year-old equipment in the loading area would have to be relocated. This would significantly increase the incremental costs associated with such a system and make this project infeasible.

• Vapor Recovery and Controlled Process Vent. The implementation of this project would result in a reduction of 4 tpy of HRVOC emissions from storage operations. The project would involve installation of a chiller/vapor recovery unit on a storage tank to recover vapors that would otherwise be lost as breathing or working losses to the flare system. The emission reduction estimates are based on actual flare loading and assumptions on the chiller recovery efficiency. A high cost-benefit ratio, as defined by the facility, makes this project infeasible. However, the facility is trying to gather funds for the project and is hopeful that the project will be implemented.

• Flare Gas Recovery. The facility intended to install a flare gas recovery system. The facility currently has 4 such systems installed. However, due to questionable emission reductions data and an unrefined cost estimate, a cost-benefit analysis was not conducted for this project.

• Heat Exchanger Upgrade. The facility intended to upgrade heat exchangers in HRVOC service. However, due to questionable data and unrefined cost estimate, a cost to benefit analysis was not conducted for this project.

4.1.5 Excess Monomer Removal ENVIRON asked Phase 1 polymer plant survey respondents whether excess monomer was removed from resins prior to finishing operations (i.e., extruder vents and vents downstream of the extruder). Nine polymer facilities responded to this part of the survey. Responses are discussed below.

• No Control. Four of the nine facilities that responded to this part of the survey did not report any processes in place to recover raw materials prior to finishing operations.50

50 Although these four facilities did not report any processes in place to recover raw materials prior to finishing



• Catalytic Oxidation. One facility routes emissions from a tertiary degasser for LLDPE and HDPE to a catalytic oxidizer.

• Hot Nitrogen Purge. One facility purges the polyethylene and polypropylene fluff with hot nitrogen into a closed loop system which also includes the extruder feed tank. Residual hydrocarbons recovered are eventually sent to the flare. In another facility, excess monomer is recovered from the polymer slurry prior to the production of pellets. In the liquid-phase slurry process, hot nitrogen stripping is used to recover the monomer in downstream units. For the liquid-phase solution process, the excess monomer is stripped using nitrogen and a system of centrifuges and flash dryers.

• Low Pressure Recovery. One facility uses a low pressure recovery compressor to recover nitrogen, ethylene and isobutane from the facility’s intermediate polyethylene flake storage tanks. These intermediate tanks store polyethylene flake prior to extrusion. The recovery compressor routes the nitrogen, ethylene and isobutane to a flare. This stream is approximately 99% isobutane.

• Ethylene Recovery Unit. The same facility employing low pressure recovery also utilizes an Ethylene Recovery Unit (“ERU”). The ERU is a three step, cryogenic process, by which ethylene is recovered from the resins prior to finishing operations. Recovered ethylene is recycled to the process, routed to a boiler as fuel or routed to a flare if the boiler is not operational. Note that both the low pressure recovery system and the ERU were installed by the facility prior to the advent of the HRVOC rules.

• Pressure Swing Absorption. One facility utilizes a Pressure Swing Absorption (“PSA”) system to capture all continuous vent streams that were previously routed to the flare for control. These vent streams consist primarily of nitrogen (75%), with the remainder being propylene (25%). The PSA system recovers, condenses and recycles propylene back to the process in liquid form. Also, nitrogen is recycled back to the process.

4.1.6 Costs of HRVOC Emission Reduction Projects Table 14 summarizes the cost effectiveness of HRVOC reduction projects implemented by various facilities as listed in Table 13. The largest project in terms of total capital investment, P31, was implemented specifically to reduce HRVOC emissions during episodic events, not for controlling routine HRVOC emissions. Also, P4 was implemented for MON/HRVOC compliance, not just for controlling HRVOC emissions.

ENVIRON asked survey participants to provide estimates of total capital investment as well as direct and indirect annual costs. As noted in Table 14, annual cost information was not made available for all projects. Some facilities indicated they do not track annual costs for these projects separately from other annual costs. As shown in Table 14, there is a wide range in cost effectiveness, ranging from $2,012 to $195,600 per ton per year of HRVOC controlled. Project P27 is excluded from the range because of zero capital and annual costs.

operations, these facilities may operate a closed loop system with nitrogen purge.






($)

Direct and Indirect Annual

Cost ($) Total Annual

Cost2 ($) HRVOC Emission Reduction (tpy)

Cost Effectiveness3

($/tpy) Phase

P1 650,000 130,000 5,000 135,000 40 3,375 Phase 1

P2 1,000,000 200,000 NS 200,000 15 13,333 Phase 1

P3 50,750 10,150 NS 10,150 See Note4 N/A Phase 1

P4 11,500,000 2,300,000 120,000 2,420,000 143 16,923 Phase 1

P5 0 0 0 0 0.5 N/A5 Phase 1

P6 364,000 72,800 25,000 97,800 0.5 195,600 Phase 1

P7 127,000 25,400 5,000 30,400 1 30,400 Phase 1

P8 7,000,000 1,400,000 NS 1,400,000 42 33,333 Phase 1

P9 400,000 80,000 NS 80,000 1 80,000 Phase 1

P10 800,000 160,000 NS 160,000 20 8,000 Phase 1

P11 – P14 273,000 54,600 NS 54,600 27.14 2,012 Phase 1

P15 – P23 755,600 151,120 NS 151,120 21.67 6,974 Phase 1

P24 122,000 24,400 0 24,400 4.9 4,980 Phase 2 P25 149,000 29,800 0 29,800 12.6 2,365 Phase 2 P26 0 0 5,000 5,000 NS NA Phase 2 P27 0 0 0 0 0.3 0 Phase 2 P28 31,000 6,200 0 6,200 0.25 24,800 Phase 2 P29 28,000 5,600 0 5,600 NS NA Phase 2 P30 500,000 100,000 50,000 150,000 20 7,500 Phase 2 P31 40,700,000 8,140,000 890,000 9,030,000 NS NA Phase 2






($)

Direct and Indirect Annual

Cost ($) Total Annual

Cost2 ($) HRVOC Emission Reduction (tpy)

Cost Effectiveness3

($/tpy) Phase

P32 1,200,000 240,000 0 240,000 13.5 17,778 Phase 2 P33 NS NS NS NS 5 NA Phase 2 P34 NS NS NS NS 15 NA Phase 2 P35 NS NS NS NS 45 NA Phase 2 P36 NS NS NS NS 15 NA Phase 2

1 Based on five-year project life and a discount rate of 0%. 2 Total Annual Cost = Annualized Capital Investment + Direct and Indirect Annual Cost 3 Cost Effectiveness = Total Annual Cost / Total HRVOC Emission Reduction 4 Estimated HRVOC reductions between 1 - 3 tph depending on frequency and duration of shutdown. Facility did not provide detailed information regarding the frequency and duration of shutdown; therefore, annualized HRVOC reductions were not estimated. 5 Startup performed once every two to three years. The estimated HRVOC reductions are annualized. Total capital investment and annual costs are negligible, so cost of control is effectively zero but HRVOC emission reductions are low.



4.1.7 Flare Reduction and Minimization Projects ENVIRON asked survey respondents whether any projects had been implemented to reduce or minimize flaring. Responses are summarized in Table 15.



HRVOC Reduction

(tpy) Phase

F1 Flare Gas Recovery Polypropylene – gas phase process 608,400 12 – 14 Phase 1

F2 Flare Gas Recovery1


and gas phase processes

970,000 5 – 10 Phase 1

F3 Vent Recycle Polyethylene – low

density, high pressure process

50,000 10 – 12 Phase 1


Polyethylene – low density, high

pressure process N/A N/A Phase 1


Polyethylene - high density, gas phase

process N/A 2 Phase 1

F6 Ethylene Recovery Unit Polypropylene – gas phase process 1,000,000 15 Phase 1

F7 De-inventory to propylene storage2


process 50,750 See note

below Phase 1

F8 Modifications to Startup and Shutdown Procedures

Polypropylene – liquid-phase and

gas phase processes

N/A N/A Phase 1

F9 Chiller Replacement Project Olefins 3,500,000 85 Phase 1

F10 Flareless Startup and Shutdown Olefins 1,100,000 50 – 100 Phase 1

F11 Rerouting Degassing Vent Olefins 106,250 6.75 Phase 1

F12 Modifications to Startup Procedure

Polyethylene – liquid phase

process 0 0.5 Phase 1

F13 Addition of Calorimeters to Vent Gas Monitoring System

Polyethylene – liquid phase

process 183,000 1 Phase 1

F14 Flare Gas Recovery Refinery 17,900,000 9 Phase 1

F15 Flare Gas Recovery Refinery 34,500,000 45 Phase 1





HRVOC Reduction

(tpy) Phase

F16 Heavy Ends Stream Recovery Olefins 220,000 50 Phase 1

F17 Olefins Flare Reduction Olefins 550,000 9 Phase 1

F18 Off-spec monomer to off-site storage Olefins 700,000 17 Phase 1

1 Project is installed, but not yet operating. 2 Estimated HRVOC reductions between 1 - 3 tph depending on frequency and duration of shutdown.

Projects P2, P3 and P5 from Table 13 also appear as Projects F6, F7 and F12 in Table 15. These projects cannot be classified solely as polymer projects as their implementation also affected HRVOC emissions through flaring. Therefore, these projects are included in the discussion of both polymer plant and flare projects.

Additional details concerning the projects referenced in Table 15 are provided below.

• Project F1. The facility is using an existing compressor to remove the waste gas from the flare header using suction pressure. The waste gas is the reactor off-gas from the polypropylene (gas phase) process. This waste gas is sold to a neighboring facility for use as boiler fuel. HRVOC emission reductions are estimated to be between 12 and 14 tpy.

• Project F2. The facility installed a compressor to remove the waste gas from the flare header using suction pressure. The waste gas is the reactor off-gas from the polypropylene (liquid-phase slurry and gas phase) processes. When operational, HRVOC emission reductions are estimated to be between 5 and 10 tpy.

• Project F3. The facility added a recycle stream from the purge gas vessel. These emissions are routed back to the process (polyethylene – low density, high pressure). Reductions in HRVOC emissions due to reduced flaring as a result of vent recycle are estimated to be 10 to 12 tpy.

• Projects F4, F5, F8, and F12. Several facilities have modified their startup/shutdown procedures to minimize flaring emissions. One such modification to shutdown procedures (Project F4) is to reduce the reactor pressure before flaring. Reactor off-gas is recompressed and sent to purification step to recover ethylene. In another project (Project F5), the facility changed the product transition procedure to route the initial purge from the flare to the site’s ethylene unit fuel gas system. This reduced 19,000 lbs of ethylene per product transition or approximately 300,000 lbs/yr out of the flare. Assuming 99% flare destruction efficiency, the estimated HRVOC emission reductions are 2 tpy. In Project F12, modifications to startup procedures include pressure checking



the reactor with isobutane instead of ethylene, resulting in annualized HRVOC emission reductions of approximately 0.5 tpy.

• Project F6. The facility installed an ethylene recovery unit on an off-gas stream to recover up to 3,000,000 pounds of ethylene from the flare header system. Taking into account an assumed 99% DRE, the recovery of 3,000,000 pounds of ethylene translates to post-flare HRVOC emission reductions of approximately 15 tpy.

• Project F7. This project consisted of piping modifications which allowed the facility to recycle liquid slurry back to monomer storage instead of flaring during shutdown. Depending on the duration of shutdown, HRVOC emission reductions are estimated to be between 1 and 3 tph during the event.

• Project F9. The facility implemented this project to condense and recover C4 compounds (mostly 1,3-butadiene) from vent streams in the olefins plant. This project has reduced the load on vent recovery compressors, thereby minimizing or eliminating the bypass of HRVOCs to the flare. HRVOC emission reductions due to minimized flaring are estimated to be approximately 85 tpy.

• Project F10. The facility was able to achieve 75 – 80% reduction in HRVOC emissions due to flaring with the implementation of “flareless” startup/shutdown procedures. These reductions in flaring translate into HRVOC emission reductions of 50 to 100 tpy.

• Project F11. The facility rerouted the propylene compressor degassing pot vent back to the process. The vent was previously routed to the flare. Reductions in HRVOC emissions due to flaring as a result of vent recycle into the process are estimated to be approximately 6.75 tpy.

• Project F13. The facility added calorimeters to the already existing vent gas monitoring system. The addition of calorimeters allows operations personnel to identify any spike in the heat value of the waste gas stream in a timely manner. This is also helpful in identifying a source that may be inadvertently venting to the flare. This project has resulted in a reduction of 0.5 to 1 tpy of HRVOC emissions due to flaring.

• Project F14 and F15. The refinery implemented two flare gas recovery projects. The flare gas recovery systems have resulted in HRVOC reductions of approximately 9 and 45 tpy, respectively. Additional details regarding these projects are not available; however, it is known that the projects were not implemented for purpose of achieving HRVOC emission reductions.

• Project F16 and F17. The olefins plant implemented two flare minimization projects. The flare minimization projects have resulted in HRVOC reductions of approximately 50 and 9 tpy, respectively. Additional details regarding these projects are not available.

• Project F18. The facility sends off-specification monomer from the process to off-site salt dome storage well during start-up and shutdown, resulting in HRVOC reductions of approximately 17 tpy.

4.1.8 Flare Gas Recovery Six polymer manufacturing facilities and one refinery responded to this part of the survey. Two facilities, both polypropylene manufacturers, have installed FGR systems for flare minimization.



Four facilities have not implemented FGR. Responses from two facilities discussing why they have not implemented FGR are as follows:

• At one facility, the flare controls only relief devices and emergency emissions. The flare header normally handles low flow rates. Therefore, a compressor installed for FGR would only run for short periods of time.

• At an olefins facility, tie-ins do not exist in the existing plant layout; therefore, any modification would require a major turnaround. Because it is an olefins plant, any oxygen ingress into the process resulting from operation of the FGR system could result in a dangerous situation. Another challenge for the olefins plant is the storage of off-spec material for future reuse. Currently, the plant is not equipped to store off-spec material or reuse it.

• A chemicals facility has an existing FGR system that utilizes compressors to supply fuel gas to a cogeneration facility. However, the system was installed prior to HRVOC rules and does not reduce reported HRVOC emissions.

• At a refinery, FGR systems exist on 4 out of 8 flares. These flares were installed, almost 20 years ago, before the HRVOC rules were formulated.

4.1.9 Costs of HRVOC Flare Reduction and Minimization Projects Table 16 summarizes the cost effectiveness of HRVOC flare reduction and minimization projects implemented by polymer plants, olefins plants and refineries listed in Table 15. Only Phase 1 survey respondents have implemented HRVOC flare reduction and minimization projects. No Phase 2 survey respondents identified any projects categorized as HRVOC flare reduction and minimization projects.

As shown in Table 16, there is a wide range in cost effectiveness, ranging from $880 to nearly $400,000 per ton of HRVOC controlled. As mentioned during discussion of the projects, the two refinery HRVOC abatement projects were not implemented for the purpose of reducing emissions of HRVOC. Excluding these two projects, cost effectiveness ranges from $880 to $51,600 per ton of HRVOC controlled.






($)

Direct and Indirect

Annual Cost ($)

Total Annual Cost2

($)

HRVOC Emission

Reduction (tpy)

Cost Effectiveness3

($/tpy) Phase

F1 608,400 121,680 N/A 121,680 13 9,360 Phase 1

F2 970,000 194,000 N/A 194,000 7.5 25,867 Phase 1

F3 50,000 10,000 N/A 10,000 11 909 Phase 1

F4 N/A N/A N/A N/A N/A N/A Phase 1

F5 N/A N/A N/A N/A 2 N/A Phase 1

F6 1,000,000 200,000 N/A 200,000 15 13,333 Phase 1

F7 50,750 10,150 N/A 10,150 See Note5 N/A Phase 1

F8 N/A N/A N/A N/A N/A N/A Phase 1

F12 0 0 0 0 0.5 N/A Phase 1

F13 183,000 36,600 15,000 51,600 1 51,600 Phase 1

F9 3,500,000 700,000 50,000 750,000 85 8,824 Phase 1

F10 1,100,000 220,000 N/A 220,000 75 2,933 Phase 1

F11 106,250 21,250 N/A 21,250 6.75 3,148 Phase 1

F16 220,000 44,000 N/A 44,000 50 880 Phase 1

F17 550,000 110,000 N/A 110,000 9 12,222 Phase 1

F18 700,000 140,000 N/A 140,000 17 8,235 Phase 1

F14 17,900,000 3,580,000 N/A 3,580,000 9 397,778 Phase 1

F15 34,500,000 6,900,000 N/A 6,900,000 45 153,333 Phase 1

w/ Refinery 61,438,400 12,287,680 65,000 12,352,680 346.75 35,624 Phase 1






($)

Direct and Indirect

Annual Cost ($)

Total Annual Cost2

($)

HRVOC Emission

Reduction (tpy)

Cost Effectiveness3

($/tpy) Phase

w/o Refinery 9,038,400 1,807,680 65,000 1,872,680 292.75 6,397 Phase 1

1 Based on five-year project life and a discount rate of 0% 2 Total Annual Cost = Annualized Capital Investment + Direct and Indirect Annual Cost 3 If the facility provided a range of HRVOC emission reductions, the median is used. 4 Cost Effectiveness = Total Annual Cost / Total HRVOC Emission Reduction 5 Estimated HRVOC reduction between 1 - 3 tph depending on frequency and duration of shutdown.



4.2 Analysis

4.2.1 Types of HRVOC Emission Reduction Projects Survey participants provided information on 54 HRVOC emission reduction projects. A summary of these projects by industry sector and type is presented in Table 17. The types of projects are defined as follows:

Process Change: ..............................Change in how the product is made. For example, changing from a gas phase process to a liquid phase slurry process.

Change in Operating Procedures: ....Change in operating procedures such as enhanced maintenance or use of more robust process simulation to reduce emissions during startup and shutdown.

Vent Gas Control: ............................. Installation of controls on vent streams where none existed previously or upgrading to control systems with higher control efficiencies, such as routing vent streams to a thermal oxidizer instead of a flare.

Flare Minimization:............................Recovery of material for reuse instead of sending it to the flare header and/or recovering material in the flare header for beneficial reuse.

Table 17. Summary of Project Type by Industry Sector

Type of HRVOC Emission Reduction Project

Industry Sector Number of

Plants Participating1

Process Change

Change in Operating Procedure

Vent Gas Control

Flare Minimization

Polymers 9 0 16 7 7

Chemicals2 16 0 11 0 6

Refinery 3 0 0 0 4

Terminals 8 0 0 0 0

Total 36 0 27 7 17 1Two of the survey respondents manufacture both polymers and olefins at the site. One site contains both refining and olefins manufacturing operations. 2 Chemicals sector includes olefin plants.

As shown, no facility implemented a change in process for the purpose of reducing emissions of HRVOC.



4.2.2 Cost Effectiveness of Emission Reduction Projects Table 18 presents a plant-by-plant summary of projects that have been undertaken to reduce emissions of HRVOC. Only sites that implemented HRVOC emission reduction projects are listed in Table 18.

Table 18. Cost and Emission Reduction Summary by Plant

Site1 Number of Projects Total Annual

Cost2 ($)

HRVOC Emission

Reduction (tpy)

Cost Effectiveness3

($/tpy)

Polymer 1 4 145,000 53 2,736 Polymer 2 1 200,000 15 13,333 Polymer 3 4 325,830 20.5 15,894 Polymer 4 1 2,420,000 143 16,923 Polymer 5 4 179,800 3 59,933 Polymer 7 3 1,640,000 63 26,032 Polymer 8 4 54,600 27.14 2,012 Polymer 9 9 151,120 21.67 6,974

Polymer Plant Subtotal 30 5,116,350 346.31 14,774 Chemicals 1 1 140,000 17 8,235 Chemicals 2 1 21,250 6.75 3,148 Chemicals 3 2 970,000 160 6,063 Chemicals 4 2 154,000 59 2,610 Chemicals 5 2 54,200 17.5 3,097 Chemicals 6 4 16,800 0.55 30,545 Chemicals 7 1 150,000 20 7,500 Chemicals 9 4 NS 80 NA

Chemicals Plant Subtotal 17 1,506,250 361 4,175 Refinery 1 2 10,480,000 54 194,074 Refinery 2 1 9,030,000 NS NA Refinery 3 1 240,000 13.5 17,778

Refinery Subtotal (without Refinery 1 and 2) 1 240,000 13.5 17,778

Total (w/ Refinery 3) 48 6,862,600 720.81 9,521 1 Polymer Plant 6 and Chemical Plant 8 did not implement any projects in response to the HRVOC rules. 2 Total Annual Cost = Annualized Capital Investment + Direct and Indirect Annual Costs. Annualized Capital Investment assumes a 5-year project life and a discount rate of 0%. In many cases direct and indirect annual costs are not provided. 3 Cost Effectiveness = Total Annual Cost / Total HRVOC Emission Reduction

As previously noted, the refinery projects at Refinery 1 and 2 were not undertaken for the sole purpose of reducing emissions of HRVOC. In many cases, the annual direct and indirect costs (e.g. natural gas to fuel a thermal oxidizer) have not been provided by survey respondents. In



those cases, the actual annual costs and the cost of controlling HRVOC emissions will be greater than what is reported herein.

Figure 8 graphically presents the annual costs for HRVOC emission reduction projects and the resulting reduction in annual emissions for surveyed facilities from all industry sectors. Only facilities that provided the total annual cost and HRVOC emission reduction information have been included in the figure.

Figure 8. Cost and Emission Reduction Summary – Facility

As can be seen from Figure 8:

• There is large variation in the amount of HRVOC emission reductions achieved at polymer plants, chemical plants and refineries;

• The amount of money spent on HRVOC emission reduction projects annually varies from $16,800 for a chemical plant (Chemicals 6) to $2,420,000 for a polymer plant (Polymer 4); and

• The reduction in HRVOC emissions achieved varies from 0.55 tpy to 160 tpy, both for a chemical plant (Chemicals 6 and Chemicals 3, respectively).

0 20 40 60 80 100 120 140 160 180

Refinery 3

Chemicals 7

Chemicals 6

Chemicals 5

Chemicals 4

Chemicals 3

Chemicals 2

Chemicals 1

Polymer 9

Polymer 8

Polymer 7

Polymer 5

Polymer 4

Polymer 3

Polymer 2

Polymer 1

Faci

lity

HRVOC Emission Reduction (tpy)

0 500,000 1,000,000 1,500,000 2,000,000 2,500,000 3,000,000



Total Annual Cost



Since the projects were not undertaken for purposes of reducing HRVOC emissions, Refinery 1 and Refinery 2 are not included in Figure 8.

4.2.3 Statistical Analysis of HRVOC Survey Data Figure 9 presents a frequency histogram of the total annual cost of HRVOC emission reduction projects. Excluded are projects where HRVOC emission reductions are not known and/or the capital cost for the project is zero or unknown (Projects P3, P11 - P14, P15 – P23, P26, P27, P29, P31, P33 - P36, F4, F7, and F8). Since the projects were not undertaken for the purpose of reducing HRVOC emissions, Projects F14 and 15 are also excluded.

Figure 9. Total Annual Cost – Frequency and Cumulative Percentage

As shown in Figure 9, of the 17 projects included in the figure, the total annual cost of 13 projects (76% of the total projects) is less than or equal to $250,000.

Figure 10 presents a frequency histogram for the HRVOC emission reductions achieved. As shown in Figure 10, of the 17 projects included in the figure, 9 projects (53% of the total projects) resulted in HRVOC emission reductions of 20 tpy or less.

Figure 11 is a graphical representation of total HRVOC emission reduction achieved by surveyed facilities that undertook HRVOC emission reduction projects as a percentage of annual HECT allowance allocations.51 Figure 11 shows that the polymer plants surveyed have a total HECT allowance allocation of approximately 275 tons and achieved a total HRVOC emission reduction of 150.3 tons, or 55% of their allocation. The chemical facilities surveyed have a total HECT allowance allocation of approximately 1,035 tons and achieved an HRVOC

51 It is important to note that three facilities included in this analysis had operations in multiple sectors, namely, polymers, chemicals and refining. The facilities were grouped into the polymer, chemical or refinery sector on the basis of the primary Standard Industrial Classification (SIC) code, solely for the purposes of this analysis.

0%6% 6% 6%

24%29%

82%88%

94%100%

24%

76%

0123456789

5,000

7,500

10,00

0

20,00

0

40,00

0

50,00

0

100,0

00

250,0

00

500,0

00

1,000

,000

2,000

,000

2,500

,000


Freq

uenc

y

0%

20%

40%

60%

80%

100%

120%

Cum

ulat

ive

Perc

ent

Frequency

Cumulative %



emission reduction of approximately 498 tons, or 48% of their allocation. Refineries had a total HECT allowance allocation of 770 tons and reduced their HRVOC emissions by approximately 126.5 tons or 16% of their allocation. None of the terminals surveyed undertook HRVOC emission reduction projects.

Figure 10. HRVOC Emission Reduction – Frequency and Cumulative Percentage

Figure 11. HRVOC Emission Reduction as Compared to HECT Allowance Allocations

6% 6%

24%

71% 71%

88% 88% 88%94%

100%

6%

18%

53%

0

1

2

3

4

5

6

0.25 0.5 1 5 10 20 40 50 75 100 125 150 170

HRVOC Emission Reduction (tpy)

Freq

uenc

y

0%

20%

40%

60%

80%

100%

120%

Cum

ulat

ive

Per

cent

Frequency

Cumulative %

0% 10% 20% 30% 40% 50% 60%

Polymer

Chemicals

Refinery

Terminals

Indu

stry

Sec

tor

Total Percent Reduction to Allowance

0 200 400 600 800 1000 1200

Total Annual HECT Allowances (tons)

Percent HRVOC Reduction

Annual HECT Allow ances



Figure 12 presents the 95% confidence interval for the average HRVOC emission reduction as percent HECT allowance for chemical plants and polymer plants. Terminals are not included in this analysis since they did not undertake any HRVOC emission reduction projects. Refineries are also not included since the percent HECT allowance could only be calculated for one refinery.52

Figure 12. HRVOC Emission Reduction as Compared to HECT Allowance Allocations by Industry Sector with 95% Confidence Interval

As shown in Figure 12, the Chemicals interval overlaps with the Polymer interval, which indicates that the differences in the observed mean HRVOC emission reductions as a percentage of HECT allowance allocations between Chemicals and Polymer plants are not statistically significant. For the sake of completeness, a Student’s t-test was performed to confirm that the difference in mean percent of HECT allowance between chemical plants and polymer plants is statistically significant. A two-tail test is used as the differences could be positive or negative. A t-test allowing for unequal variances in the mean percent of HECT allowance between these two industry sectors results in a p-value of 0.48, well above the value of 0.05 (corresponding to a 95% confidence level) typically used to determine statistical significance. A more powerful t-test based on the assumption of equal variances results in a slightly higher p value (0.67) which is still not statistically significant. These results show that the available data are not sufficient to identify a statistically significant difference in the percent of HECT allowance between these two industry sectors. We note that this conclusion is based on the assumptions that the observed percentage of HECT allowance allocations represent random, independently-drawn samples from each industrial sector and that the percentages are at least approximately normally distributed.

52 One of the refineries included in the survey that undertook HRVOC emission reduction projects did not receive an allocation of allowances.

Chemicals

Polymer-100.00%

-50.00%

0.00%

50.00%

100.00%

150.00%

200.00%

250.00%

300.00%

350.00%

% A

vera

ge H

RVO

C E

mis

sion

Red

uctio

ns C

ompa

red

with

HEC

T A

llow

ance

Allo

catio

ns



4.2.4 Projects Not Undertaken As discussed in Section 4.1.4, there are a number of HRVOC emission reduction projects that were not undertaken by the facilities. Based on discussions with industry personnel, the reasons include:

• They did not need additional reductions in HRVOC emissions. This was because 1) emissions were already less than their HECT allowance allocation; and/or 2) the company manages their HRVOC emissions as a portfolio across a number of sites that may include polymer plants, olefins and chemical manufacturing sites and/or petroleum refineries and the portfolio as a whole may be sufficient.

• The cost of additional reductions in HRVOC emissions exceeded internal financial thresholds.

• These are significant space limitations that would impact the ability to add new equipment.

• The facility is large, complex or old which would make upgrades or modifications difficult.

None of the companies surveyed have undertaken projects for the purpose of selling excess allowances. Reasons include:

• With little or no market activity to-date, there are not good pricing signals as to what HECT allowance vintages may be worth in the future. Therefore, there is insufficient information available to make sound investment decisions.

• Since MSS and event emissions (up to 1,200 pounds per hour) count toward the HECT annual cap, companies are unwilling to sell allowance stream ownership.



5 Flare Survey Findings 5.1 Collected Information Tables 19 and 20 present the flare data summary for routine loading and emergency loading, respectively. The tables present a merged summary of the data collected during Phase 1 and Phase 2 of the study. It is important to note that the data is presented as reported by facilities in response to the facility and flare questionnaires sent out by ENVIRON. Data fields in Table 19 and Table 20 are defined as follows.

• Flare ID is the ENVIRON-assigned identification number for the facility flare.

• Industry Sector refers to one of the 4 manufacturing sectors targeted for this study: polymer, chemicals, refinery and terminals.

• Service refers to the type of operation associated with the flare. The flare can be in routine, emergency (upset/MSS) or both routine and emergency service.

• Flare Design Capacity is the maximum flow rate, in MMscf/hr, that can be handled by the flare. [This is not the smokeless capacity, which may be considerably less than the maximum design capacity.]

• Average Routine Flow Rate is the average flow rate, in MMscf/hr, to the flare when operating under normal (non-emergency) service. For single flares operating in both routine and emergency service, facilities may or may not separate emergency loading when reporting routine flows to the flare.

• Average Upset/MSS Flow Rate is the average flow rate, in MMscf/hr, to the flare when operating under emergency service.

• Percentage of Design Capacity is the ratio of the average routine or emergency loading to the flare to its maximum design flow rate, expressed as a percentage.

• As specified above, the reported average routine flow rates may contain loading during emergency events also. This has been specified in Table 19 and Table 20, if provided by the respondent.

• Routine or Emergency Hours of Operation refers to the total number of hours in 2008 the flare was in routine service or events/MSS service, respectively.

• Percentage of Total Hours is the ratio of the routine or emergency service hours to the total hours of operation for the flare, expressed as a percentage.

• Average Routine Flow Rate and Average Upset/Maintenance Flow Rate Data Source refers to the source of flare loading data at the facility. Sources of flare data may include flow meter data, estimates based on year-round operation and HRVOC database.

• Not Specified (NS) means that at the time of writing this report the facility had not provided the data. However, this data may be included in the final report.



Table 19. Flare Survey Data Summary – Routine Flaring

Flare ID Industry Sector

Service Flare Design

Capacity (MMscf/hr)

Average Routine

Flow Rate (MMscf/hr)

Average Routine Flow Rate as a % of

Design Capacity

Routine Hours of

Operation (hr)

Percentage of Total Hours

Average Routine Flow Rate Contains

Upset/MSS Emissions?

Average Routine Flow Rate Data

Source

1 Polymer Both 12.66 0.010 0.08% NS NA NS NS



4 Chemicals Both NS NS NA NS NA NS NS



7 Chemicals Upset/MSS NS NA NA NA NA NA NA


9 Polymer Both 8.50 0.013 0.15% NS NA Yes Flow meter data – 2008 YTD



12 Chemicals Both 2.66 0.007 0.25% NS NA NS NS











Capacity (MMscf/hr)

Average Routine



Design Capacity

Routine Hours of

Operation (hr)





Source

18 Chemicals Routine 0.63 0.058 9.06% NS NA No NS

19 Chemicals Routine 0.20 0.077 37.78% NS NA No NS


21 Chemicals Both 8.34 0.051 0.61% NS NA No Flow meter data


23 Polymer Both NS 0.004 NA NS NA No Flow meter data


25 Polymer Both NS 0.000 NA NS NA No Flow meter data

26 Polymer Both 3.06 0.024 0.80% NS NA Yes Flow meter data – 2007

27 Polymer Both 5.34 0.031 0.58% NS NA Yes Flow meter data – 2007

28 Polymer NS NS NS NA NS NA NS NA

29 Chemicals Both 10 0.03 0.30% 8,500 96.77% No Flow meter data

30 Chemicals Routine1 3.6 0.51 14.17% 8,470 96.69% No Estimate based on year-round operation

31 Chemicals Routine1 3.3 0.79 23.94% 8,450 96.46% No Estimate based on year-round operation

32 Chemicals Both 0.325 0.043 13.23% 6,824 78.85% No Flow meter data

33 Chemicals Both 0.5 0.12 24.00% 8,544 97.27% Yes Flow meter data

34 Terminal Upset/MSS 0.1 NA NA NA NA NA NA






Capacity (MMscf/hr)

Average Routine



Design Capacity

Routine Hours of

Operation (hr)





Source

35 Chemicals Both2 0.7 0.2 28.57% 8,618 100.00% No Flow meter data

36 Chemicals Upset/MSS 7.9 NA NA NA NA NA NA






42 Chemicals Both 0.0024 0.00049 20.42% 8,774 99.89% Yes Estimate based on year-round operation


44 Terminal Upset/MSS NS NA NA NA NA NA NA


46 Polymer Both 4.7 0.13 2.77% 7,462 87.49% Yes Estimate based on year-round operation




50 Terminal Both 0.3 0.02 5.00% 8,760 99.85% No Flow meter data






Capacity (MMscf/hr)

Average Routine



Design Capacity

Routine Hours of

Operation (hr)





Source

51 Polymer Both2 NS 0.0002 NA 3 100.00% No Flow meter data

52 Polymer Both NS 0.1678 NA 8,410 95.76% No Flow meter data




56 Chemicals Both 31.3 NS NA NS NA No NA


58 Refinery Both NS 0.0808 NA 7 0.48% No Flow meter data

59 Refinery Both NS 0.386 NA 7,731 90.09% No Flow meter data







66 Refinery Both NS 0.0312 NA 8,444 96.13% NS Flow meter data

67 Chemicals Both 0.45 0.02 4.44% 8,426 96.19% No NS

68 Chemicals Both2 0.04 0.02 50.00% 8,760 100.00% No NS








Capacity (MMscf/hr)

Average Routine



Design Capacity

Routine Hours of

Operation (hr)





Source




74 Terminal Both2 0.021 0.005 23.81% 8,760 100.00% No Flow meter data


76 Terminal Both 0.013 4.80E-06 0.04% 8,740 99.77% Yes Flow meter data

77 Terminal Both 0.084 0.022 26.19% 8,757 99.97% Yes Flow meter data

78 Terminal Both 0.035 0.014 40.00% 8,680 99.88% Yes Flow meter data

79 Refinery NS NS NS NA 8760 NA NS NS

80 Refinery NS NS NS NA 8760 NA NS NS

81 Chemicals Both NS NS NA 8717 99.5% NA NS

82 Terminal Both 1.3 1.50E-3 0.12% 8760 100% NS NS 1 Although site indicated flare service as routine only, the site categorized a number of hours as upset/MSS. 2 Site reported zero hours of upset/MSS operation for 2008 calendar year.



Table 20. Flare Survey Data Summary - Upset/MSS Flaring

Flare ID

Industry Sector


Capacity (MMscf/hr)

Average Upset/MSS Flow Rate (MMscf/hr)


Design Capacity

Upset/MSS Hours of

Operation (hr)


Averaging Period

(Upset/MSS period or Annual)

Average Upset/MSS Flow

Rate Data Source

7 Chemicals Upset/MSS NS NS NA NS NA NS NS

29 Chemicals Both 10 0.1 1.00% 284 3.23% Upset/MSS Flow meter data

32 Chemicals Both 0.325 0.049 15.08% 1,830 21.15% NS NS

33 Chemicals Both1 0.5 0.020 4.00% 240 2.73% Annual Flow meter data

34 Terminal Upset/MSS 0.1 0.1 100.00% 1 100.00% Upset/MSS Estimate - based on approximate

Upset/MSS hours

35 Chemicals Both2 0.7 NS NA 0 0.00% Upset/MSS NA

36 Chemicals Upset/MSS 7.9 0.3 3.80% 174 100.00% Upset/MSS Estimate - based on approximate

Upset/MSS hours

37 Chemicals Upset/MSS 16.8 NS NA 5 100.00% Upset/MSS Estimate - based on approximate

Upset/MSS hours

38 Chemicals Upset/MSS 16.8 NS NA 1 100.00% Upset/MSS NA

39 Chemicals Both 0.1173 0.021 17.90% 48 0.59% Upset/MSS Flow meter data



42 Chemicals Both 0.0024 0.00001 0.42% 10 0.11% Upset/MSS Estimate - based on approximate

Upset/MSS hours

43 Terminal Upset/MSS 0.12 0.0001 0.08% 8,760 100.00% NS Flow meter data




Flare ID

Industry Sector


Capacity (MMscf/hr)



Design Capacity

Upset/MSS Hours of

Operation (hr)


Averaging Period



Rate Data Source

44 Terminal Upset/MSS NS 0.0004 NA NS NA NS Flow meter data

45 Terminal Upset/MSS 0.14 0.003 2.14% NS NA NS Flow meter data

46 Polymers Both 4.7 0.004 0.09% 1,067 12.51% Upset/MSS Flow meter data



49 Polymers Both 0.2 0.002 1.00% 45 0.51% Upset/MSS Flow meter data

50 Terminal Both 0.3 0.05 16.67% 13 0.15% Upset/MSS Flow meter data

51 Polymers Both NS NS NA NS NA Upset/MSS NA

52 Polymers Both NS 0.1125 NA 372 4.24% Upset/MSS Flow meter data




56 Chemicals Both1 31.3 0.6732 2.15% 186 100.00% Upset/MSS Flow meter data


58 Refinery Both NS 0.2653 NA 1,444 99.52% Upset/MSS Flow meter data

59 Refinery Both NS 0.3619 NA 850 9.91% Upset/MSS Flow meter data








Flare ID

Industry Sector


Capacity (MMscf/hr)



Design Capacity

Upset/MSS Hours of

Operation (hr)


Averaging Period



Rate Data Source




67 Chemicals Both 0.45 0.1 22.22% 334 3.81% NS NS

68 Chemicals Both2 0.04 0.02 50.00% 0 0.00% NS NS

69 Chemicals Upset/MSS 0.42 0.0002 0.05% 10 100.00% Upset/MSS Flow meter data




73 Terminal Upset/MSS 0.118 0.029 24.58% 96 100.00% Upset/MSS Flow meter data



76 Terminal Both 0.013 9.80E-06 0.08% 20 0.23% Upset/MSS

Estimate – based on volume in

loading arms for the event

77 Terminal Both 0.084 0.022 26.19% 3 0.03% Upset/MSS

Estimate – based on volume in

loading arms for the event

78 Terminal Both 0.035 0.014 40.00% 10 0.12% Upset/MSS Estimate – based

on volume in loading arms for




Flare ID

Industry Sector


Capacity (MMscf/hr)



Design Capacity

Upset/MSS Hours of

Operation (hr)


Averaging Period



Rate Data Source

the event

81 Chemicals NS NS NS NA 43 0.5% NS NS

82 Terminal Both 1.3 2.09E-5 0.002% NS NA NS NS 1 Although site indicated flare service as both routine and upset/MSS, the site reported only upset/MSS hours of operation for 2008 calendar year. 2 Although site indicated flare service as both routine and upset/MSS, the site reported only routine hours of operation for 2008 calendar year.



5.2 Flare Data Quality

As noted in the Introduction, the information contained within this report is presented as it was reported by participating industrial sites to ENVIRON. Data validation activities performed by ENVIRON were limited to email correspondence and telephone conversations with participants regarding perceived anomalies and/or clarification of projects performed. If confirmed by survey respondents as accurate, data is included in the findings and analysis presented within this report.

In a number of instances, the routine flare loading rates reported by facilities seems high. As an example, the reported “average routine” flow rate for Flare 31 is 0.79 MMscf/hour. Using the ideal gas law and assuming that all of the material flared is ethylene, the flare loading is estimated as follows:

M = (PV/RT) • MW (Equation 1)

Where:

M = Mass P = Pressure = 1.0 atmosphere V = Volume = 790,000 ft3/hr R = Gas Constant = 0.7302 atm•ft3/lbmole•°R T = Temperature = 520°R MW = Molecular Weight = 28 lbs/lbmole (ethylene)

Solving:

M = {[(1 atm)•(790,000 ft3/hr)]/[( 0.7302 atm•ft3/lbmole•°R)•( 520°R)]}•(28 lbs/lbmole)

M = 58,256 lbs/hr

M = 29.1 tons/hr

Assuming a 99% destruction efficiency (standard assumption for flaring of ethylene), emissions from this flare would be 0.291 tons/hr or 2,461 tons/year (8,450 hours of annual operation). This value seems excessive for actual emissions from a single flare.. However, the following should be kept in mind while reviewing this report:

• Main flares at petroleum refineries, chemical plants, polymer plants and terminals can be very large sources of emissions. These sources can be permitted for routine and scheduled MSS emissions in excess of 1,000 tons/year of volatile organic compounds (VOC).

• Survey recipients were only asked to report average routine total flow to the flare. They were not asked to speciate the flare flows. Therefore, the reported flows will include not only HRVOC and other VOCs, but methane or refinery gas sweep, nitrogen purge gas, carbon monoxide, carbon dioxide, water vapor, and any other gases that are in the flare header. If, for purposes of illustration, the average heat content of material being routed



to Flare 31 is assumed to be 400 Btu/scf and the combustible hydrocarbon has the characteristics of ethylene, the ratio of flare gas heat content to ethylene heat content (approximately 1,600 Btu/ft3) can be used to estimate flare hydrocarbon emissions.

M = [0.291 tons/hr (@ 1,600 Btu/ft3)] x [(400 Btu/ft3)/(1,600 Btu/ft3)]

M = 0.0728 tons/hr

M = 615 tons/year (8,450 hours of annual operation)

Total hydrocarbon emissions of this amount, while still large, seem within the realm of possibilities.

• The purpose of the survey was to obtain an estimate of routine flow rates in relation to design capacity, not to allow estimation of emissions. The information provided by survey respondents is insufficient to allow for an accurate estimate of emissions.

5.3 Analysis

5.3.1 Flare Design Capacity Figures 13(a), 13(b) and 13(c) present histograms that represent the frequency of occurrence for flares of a certain design capacity range and the corresponding cumulative percentage. The flare design capacities have been divided into three bins:

• “Small” flares with design capacities less than 1 MMscf/hr;

• “Medium” flares with design capacities ranging from 1 to 10 MMscf/hr; and

• “Large” flares with design capacities greater than 10 MMScf/hr.

Figure 13(a). Small Flare Maximum Design Capacity - Frequency and Cumulative Percentage

0.00%6.67% 6.67%

30.00%

90.00%100.00%

80.00%

23.33%

0

2

4

6

8

10

12

14

16

0.001

0.005 0.0

10.0

5 0.1 0.5 0.75 0.8

Maximum Design Capacity (MMscf/hr)

Freq

uenc

y

0%

20%

40%

60%

80%

100%

120%C

umul

ativ

e Pe

rcen

tFrequency

Cumulative %



Figure 13(a) shows that almost 80% of small flares have a design capacity equal to or less than 0.5 MMscf/hr and that approximately 20% of small flares have a design capacity greater than 0.5 MMscf/hr.

Figure 13(b). Medium Flare Maximum Design Capacity - Frequency and Cumulative Percentage

Figure 13(b) shows that almost 67% of medium flares have a design capacity equal to or less than 5 MMscf/hr and that approximately 33% of the flares have a design capacity greater than 5 MMscf/hr.

Figure 13(c). Large Flare Maximum Design Capacity - Frequency and Cumulative Percentage

0%

22%28%

50%56%

67%72%

78% 78% 78% 78%83%

94% 94%100%

11%22%

44%

0

0.5

1

1.5

2

2.5

3

3.5

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5


Freq

uenc

y

0%

20%

40%

60%

80%

100%

120%

Cum

ulat

ive

Perc

ent

Frequency

Cumulative %

11% 11% 11% 11%

56%

89%

100%

11%

22% 33%

78%

56%

0

0.5

1

1.5

2

2.5

10 10.5 11 11

.5 12 12.5 15 20 25 30 35 37

.5


Freq

uenc

y

0%

20%

40%

60%

80%

100%

120%C

umul

ativ

e Pe

rcen

t

Frequency

Cumulative %



Figure 13(c) shows that almost 56% of large flares have a design capacity equal to or less than 20 MMscf/hr.

5.3.2 Flare Loading Figure 14 presents a histogram that represents the frequency of occurrence for flares of a certain routine loading and the corresponding cumulative percentage.

Figure 14. Average Routine Loading - Frequency and Cumulative Percentage

Figure 14 shows that 92% of flares have a routine loading equal to or less than 0.25 MMscf/hr.

Figure 15 presents a histogram that represents the frequency of occurrence for flares in emergency service and the corresponding cumulative percentage.

Figure 15. Average Upset/MSS Loading - Frequency and Cumulative Percentage

2% 3% 6% 10%

76%

90% 92% 94% 95% 95% 97% 98% 100%

16%24%

51%

70%

02468

1012141618

1E-05

0.000

1

0.000

50.0

010.0

05 0.01

0.05 0.1 0.1

5 0.2 0.25

0.35

0.45 0.5 0.5

5 0.6 0.8

Average Routine Loading (MMscf/hr)

Freq

uenc

y

0%

20%

40%

60%

80%

100%

120%

Cum

ulat

ive

Perc

ent

Frequency

Cumulative %

9%13% 13%

64%69% 71%

78%82%

87%93% 96% 98% 98% 100%

51%

29%

7%

0

2

4

6

8

10

12

5E-05

0.000

1

0.000

50.0

010.0

05 0.05 0.1

0.125 0.1

5 0.2 0.3 0.35 0.4 0.4

50.7

5 11.1

5

Average Upset/MSS Loading (MMscf/hr)

Freq

uenc

y

0%

20%

40%

60%

80%

100%

120%C

umul

ativ

e Pe

rcen

tFrequency

Cumulative %



Figure 15 shows that 78% of the flares have an upset/MSS loading equal to or less than 0.2 MMscf/hr.

Figures 16(a), (b) and (c), present frequency histograms that represent the percentage of average routine loading to maximum design capacity for small, medium and large flares, respectively, in either routine only service or both routine and emergency service.

Figure 16(a). Small Flare Average Routine Loading to Design Capacity - Frequency and Cumulative Percentage

Figure 16(b). Medium Flare Average Routine Loading to Design Capacity - Frequency and Cumulative Percentage

0% 4% 8% 8%

32% 36% 40%

84% 84%

96% 96% 100%

4%

28%

68%

0

1

2

3

4

5

6

7

8

0.010.05 0.1 0.5 1 5 10 15 20 25 30 35 40 45 50

Percent Design Flow (%)

Freq

uenc

y

0%

20%

40%

60%

80%

100%

120%

Cum

ulat

ive

Perc

ent

Frequency

Cumulative %

0%

88%94% 94%

100%

29%

53%

88%

0

1

2

3

4

5

6

7

0.1 0.5 1 5 10 15 20 25


Freq

uenc

y

0%

20%

40%

60%

80%

100%

120%

Cum

ulat

ive

Perc

ent

Frequency

Cumulative %



Figure 16(c). Large Flare Average Routine Loading to Design Capacity - Frequency and Cumulative Percentage

Key statistics from Figures 16(a), 16(b) and 16(c) are as follows:

• For small flares, almost 68% of the flares operate at less than or equal to 25% of the maximum design capacity. This amounts to 17 out of a total of 25 flares in this design capacity range;

• For medium flares, approximately 88% of the flares operate at less than or equal to 5% of the maximum design capacity. This amounts to 15 out of a total of 17 flares in this design capacity range;

• For large flares, 92% of the flares operate at less than or equal to 0.5% of the maximum design capacity. This amounts to 11 out of a total of 12 flares in this design capacity range;

• Out of a total 54 of flares, in either routine only or both routine and emergency service for which design capacity is known, only 1 flare operates routinely at 50% of design capacity. This is the maximum among all flares surveyed.

Figure 17(a) represents the number of flares in HRVOC service and the corresponding percent average flow, by industry sector.

8%17%

92% 92% 92%100%

8%

92%

42%

0

1

2

3

4

5

6

7

0.005 0.0

10.0

5 0.1 0.5 0.75 1.5 2.5 2.7

5


Freq

uenc

y

0%

20%

40%

60%

80%

100%

120%

Cum

ulat

ive

Perc

ent

Frequency

Cumulative %



Figure 17(a). Percent Average Flow and Number of Flares Surveyed – Industry Sector

As shown in Figure 17(a), among the flares surveyed, 19 flares are in HRVOC service for polymer manufacturing facilities, 11 flares are in HRVOC service for refineries, 15 flares are in HRVOC service at terminals, and 37 flares are in HRVOC service at chemical plants. Figure 17(a) also shows the percentage of average flow to maximum design capacity for the flares in HRVOC service at different facilities. Under routine operating conditions, flares in HRVOC service at polymer plants operate at 0.08% - 40% of the maximum design capacity; flares in HRVOC service at chemical plants operate at 0.03% - 50% of the maximum design flow rate; and flares in HRVOC service at terminals operate at 0.04% - 40% of the design flow rate. The percent average flow to design flow could not be calculated for refinery flares as the maximum design flow rate for these flares was not provided by the respondents. The percentage of design flow for each industry sector was determined by calculating the average value for the flares within a particular industry sector. For the facilities that responded to this survey, the average loading across all sectors was approximately 12% of flare design capacity; the range was from less than 0.03% to 50%.

Most flares are designed to handle maximum flow rates expected during facility upsets and MSS events. Therefore, instead of having a dedicated flare for HRVOC control during emission events, most facilities use a single flare for both routine and upset/MSS operations. Figure 17(b) breaks down the flare data based on the type of service. Out of a total of 82 flares in HRVOC service, 66 flares are designed to handle routine as well as emergency flows; 4 flares are designed for routine service only; and 10 flares are dedicated to handle emergency flows only. Type of flare service was not specified for 2 flares. Figure 17(b) also shows the percentage of average flow to maximum design capacity for these flares.

0 10 20 30 40

Chemicals

Polymer

Refinery

Terminal

Indu

stry

Sec

tor


0 5 10 15 20 25

Percent Average Flow to Design Flow

Number

Percent Average



Figure 17(b). Percent Average Flow and Number of Flares Surveyed – Type of Service

Flares in both routine and emergency service operate at 0.04% - 50% of the maximum design flow rate; flares in routine only service operate at 0.03% - 37.8% of the maximum design flow rate; and flares in upset/maintenance service operate at 0.05% - 24.58% of the maximum design flow rate.

Figure 18 is a frequency histogram representing the percentage of routine hours of operation to the total hours of operation for flares in both routine and upset/MSS service.

Figure 18. Percent Routine Hours of Operation – Frequency and Cumulative Percentage

0 10 20 30 40 50 60 70

Routine

Both

Upset/MSS

Type

of S

ervi

ce


0 5 10 15 20 25

Percent Average Flow to Design Flow

Number

Percent Average

2%7% 7% 10%

15% 15% 15% 17%22% 24%

39%

100%

0

5

10

15

20

25

30

0 5 15 25 35 50 75 80 85 90 95 100

Percent Routine Hours of Operation

Freq

uenc

y

0%

20%

40%

60%

80%

100%

120%

Cum

ulat

ive

Perc

ent

Frequency

Cumulative %



Figure 18 shows that out of a total of 41 flares in both routine and upset/MSS service, 24% of flares operated in routine service less than or equal to 90% of the total hours of operation during the calendar year 2008.

Figure 19 is a frequency histogram representing the percentage of upset/MSS hours of operation to the total hours of operation for flares in both routine and upset/MSS service.

Figure 19. Percent Upset/MSS Hours of Operation – Frequency and Cumulative Percentage

Figure 19 shows that out of a total of 41 flares in both routine and upset/MSS service, 61% of the flares operated in upset/MSS service less than or equal to approximately 5% of the total hours of operation during the calendar year 2008.

It is important to note that only flares in both routine and upset/MSS service, with operating hours data provided by the respondent, were chosen for this analysis.

5.3.3 Statistical Analysis of Flare Survey Data Table 21 lists the descriptive data analysis for the maximum flare design capacity, average routine loading and average upset/MSS loading.

76% 78%83% 85% 85%

90% 93% 93%100%

61%

59%

12%

0

5

10

15

0 1 5 10 15 20 25 50 75 85 95 100

Percent Upset/MSS Hours of Operation

Freq

uenc

y

0%

20%

40%

60%

80%

100%

120%

Cum

ulat

ive

Perc

ent

Frequency

Cumulative %



From Table 21, the average maximum design capacity is 4.92 MMscf/hr; the average loading to the flares is 0.01 MMscf/hr and 0.14 MMscf/hr, for routine loading and emergency loading, respectively. Half of the flares have design capacities less than 0.78 MMscf/hr and half have design capacities greater than 0.78 MMscf/hr. Similarly, half of the flares have routine loading greater than 0.04 MMscf/hr and half have routine loading less than 0.04 MMscf/hr. The median loading value for flares in emergency service is 0.05 MMscf/hr. The standard deviation indicates that almost 68% of the flare design capacities fall within 1 standard deviation of the mean (4.92+/-8.29) and about 95% fall within 2 standard deviations of the mean (4.92+/-16.58). The minimum and maximum values indicate a large variability in the design, routine and emergency flows to the flares. The design capacities vary from 0.002 MMscf/hr to 37.13 MMscf/hr; the routine loading varies from less than 0.0001 MMscf/hr to 0.79 MMscf/hr; and the emergency loading varies from less than 0.0001 MMscf/hr to 1.10 MMscf/hr.

An analysis of variance (ANOVA) statistical test was performed to determine whether variations across industry sectors in the average routine flare loading as a percent of design capacity for flares in HRVOC service are statistically significant. The average routine flare loading as a percent of design capacity for flares in different industry sectors are 11% for chemicals; 4% for polymers; and 20% for terminals. The average routine flare loading as a percent of design capacity could not be calculated for refinery flares as the design flow rate for these flares was not provided. Results from the ANOVA show that the variations are statistically significant at the 95% confidence level (p value of 0.045; see discussion in Section 4.2.3).

Comparisons of the average routine flare loading as a percent of design capacity between pairs of industry sectors were also made using a two-tail Student’s t-test with the unequal variance assumption as described in Section 4.2.3. Resulting p-values between Chemicals and Polymer (p = 0.10) and between Chemicals and Terminals (p = 0.06) are both greater than 0.05, thus indicating differences that are not significant at the 95% confidence level (although they are significant at the 90% confidence level) However, the p-value between Polymer and Terminals (p < 0.01) indicates a statistically significant difference between these two sectors. Figure 18 illustrates these results by showing the mean percent design capacity for each industry sector with corresponding 95% confidence intervals.

Table 21. Flare Survey Data Analysis

Descriptive Statistics


Average Routine Loading (MMscf/hr)

Average Upset/MSS Loading (MMscf/hr)

Mean 4.92 0.01 0.14 Median 0.78 0.04 0.05 Mode 0.70 0.01 0.003 Standard Deviation 8.29 0.15 0.21 Minimum 0.002 4.8E-06 9.8E-06 Maximum 37.13 0.79 1.10



Chemicals

Polymer

Terminal

-5.00%

0.00%

5.00%

10.00%

15.00%

20.00%

25.00%

30.00%

35.00%%

of D

esig

n C

apac

ity

Figure 20. Routine Flare Loading as a Percent of Design Capacity by Industry Sector with 95% Confidence Interval

5.4 Summary of Flare Survey Findings This section provides an summary of flare data that was collected and compiled as part of the Phase 1 and Phase 2 studies. Key findings are as follows:

• A total of 82 flares were analyzed. Of these, 37 were in service at chemical plants; 19 were in service at polymer plants; 11 were in service at refineries; and 15 were in HRVOC service at terminals.

• Of the flares surveyed, 66 flares (82%) are designed to handle routine as well as emergency flows; 4 flares (5%) are designed for routine service only; and 10 flares (13%) are dedicated to handle emergency flows only.

• The most frequent routine loading among all flares surveyed, represented by the mode, was 0.01 MMscf/hr.

• Flares in HRVOC service at polymer plants operate at 0.08% - 40% of design capacity; flares in HRVOC service at chemical plants operate at 0.01% - 50% of design capacity; and flares in HRVOC service at terminals operate at 0.04% - 40% of design capacity.

• On an average, flares in HRVOC service at terminals operate at approximately 20% of design capacity, followed by flares at chemical plants at approximately 11% of the design capacity.

• Flares in HRVOC service at polymer plants have the largest disparity between design capacity and actual routine loading and operate, on average, at 4% of design capacity.



• The average routine loading for all flares surveyed was approximately 11% of design capacity. Half of the flares have routine loading less than 0.043 MMscf/hr; half of flares have emergency loading less than 0.049 MMscf/hr.

• Out of all flares surveyed, 32% of flares operate at less than or equal to 0.5% of the maximum design capacity under routine operations; 62% of flares operate at less than or equal to 5% of the maximum design capacity; Almost 85% flares operate at less than or equal to 25% of the maximum design capacity.

• Only 1 flare included in the survey responses operates at 50% or more of design capacity.

• The observed differences in average routine flare loading as a percent of design capacity between all industry sectors is statistically significant.

• The observed differences in average routine flare loading as a percent of design capacity between polymers and terminals are statistically significant.

• The observed differences in the average routine flare loading as a percent of design capacity between chemicals and polymers and between chemicals and terminals are not statistically significant and could be due to chance.

• 32 flares operated in routine service for 90% or more of the total hours of operation during the calendar year 2008.

• 16 flares operated in upset/MSS service for 60% or more total hours of operation during the calendar year 2008.



6 Conclusions 6.1 HRVOC Emission Reduction Projects Conclusions on HRVOC emission reduction projects implemented by polymer plants, chemical plants, petroleum refineries and storage terminals are as follows:

• Out of the four industrial sectors surveyed, polymer and chemical plants implemented the highest number of HRVOC emission reduction projects and achieved maximum HRVOC emission reductions. Out of the 3 refineries surveyed, only one refinery implemented one or more HRVOC emission reduction projects in response to HRVOC rules. Out of the 8 terminals that participated in the study, none indicated that they had implemented HRVOC emission reduction projects.

• Of the 51 HRVOC emission reduction projects identified by this survey, 27 involved changes in operating procedures, 17 involved flare minimization, and 7 involved installation of vent gas controls. No facility implemented a change in process to reduce HRVOC emissions.

• 76% of the total HRVOC emission reduction projects had a total annual cost less than or equal to $250,000.

• A majority of the HRVOC emission reduction projects resulted in emission reductions of 20 tpy or less.

• Polymer plants achieved the greatest reduction in HRVOC emission when compared to their HECT allowance allocation – approximately 55%. Chemical plants achieved HRVOC emission reductions equal to approximately 48% of their HECT allowance allocation and petroleum refineries achieved HRVOC emission reduction equal to approximately 16% of their HECT allowance allocation.

• Statistical analysis indicates that the differences in observed mean HRVOC emission reductions as a percent of HECT allowance allocations across different industry sectors are not statistically significant. The differences in observed means between chemical and polymer plants could be due to chance.

6.2 Flare Usage Conclusions on flare use are as follows:

• Flares in service at polymer plants, on average, operate at approximately 4% of design capacity; flares in HRVOC service at chemical plants, on average, operate at 11% of design capacity; and flares in HRVOC service at terminals, on average, operate at 20% of the design capacity. Insufficient information was provided by survey respondents to estimate flare capacity utilization at petroleum refineries.

• Most of the flares surveyed, approximately 82%, are designed to handle routine as well as emergency flows.



• Out of all flares surveyed, 32% of flares operate at less than or equal to 0.5% of the maximum design capacity under routine operations; 62% of flares operate at less than or equal to 5% of the maximum design capacity; Almost 85% flares operate at less than or equal to 25% of the maximum design capacity. Of the flares surveyed, only one flare operates at 50% of the design capacity. This is the maximum among all flares.

• Statistical analysis indicates that the difference in the observed means of average routine flare loading as a percent of design capacity for flares across all industry sectors is significant.



7 References Texas Commission on Environmental Quality. 2009. Flare Task Force Stakeholder Group. Presented at the March 30 and April 2, 2009, stakeholder meetings, Austin.



8 Project Team Contact Information TCEQ Project Representative: Dr. Bogdan Slomka

+1 512.239.1709 [email protected]

ENVIRON Principal in Charge: Dr. Gregory Yarwood +1 415.899.0704 [email protected]

ENVIRON Principal Investigator and Project Manager:

Steven H. Ramsey, P.E., BCEE +1 713.470.6657 [email protected]

ENVIRON Co-Investigators: Christopher J. Colville, P.E. +1 713.470.2647 [email protected]

Dr. Shagun Bhat +1 713.470.2648 [email protected]



Appendix A:Questionnaire Templates



1. What type of process is used? Please select all that apply.Polyethylene - low density, high pressure processPolyethylene - low density, low pressure processPolyethylene - high density, gas phase processPolyethylene - high density, liquid-phase slurry processPolyethylene - high density, liquid-phase solution processPolypropylene - liquid-phase slurry processPolypropylene - gas phase processOther

YesNo



3. Have any HRVOC emission reduction projects been considered, but not implemented?YesNo


YesNo

If yes, how do they vary by type of production process?

2. Have any projects been implemented to reduce emissions in response to HRVOC rules? Projects could include process changes, pollution prevention techniques as well as add-on control technology.



4. Are there emission reduction methods that would be technically feasible for new plants but not for existing plants?




3. Please feel free to add additional projects for question 2, as necessary.




What is unique to your facility that would make emission reductions more difficult or more expensive?

Estimated costs and potential emission reduction?Method 1Description:Capital Cost ($):Annual Cost ($):Estimated HRVOC Reduction (tpy):Method 2Description:Capital Cost ($):Annual Cost ($):Estimated HRVOC Reduction (tpy):

Recycled to ProcessFlareBoilerProcess HeaterThermal OxidizerOther Add-on Control Device (please describe)

6. Are there VOC control devices on any finishing vents (i.e., extruder and downstream to storage and loading)?YesNo

If yes, please describe.

7. Is excess monomer removed from the pellets prior to the finishing operations (e.g., steam stripping)?YesNo

If yes, please describe how.

5. Are process vents (i.e., upstream of extruder) recycled back to the process or routed to a control device (e.g., flare, thermal oxidizer, boiler, process heater)? Please select all that apply.




1. How many flares are in HRVOC service?Please provide the average or typical flow and design capacity for each flare in HRVOC service.

EPN Average Flow Design Flow(MMscf/hr) (MMscf/hr)

2. Have any projects been implemented to reduce or minimize flaring?YesNo



3. Have any flare minimization/reduction projects been considered, but not implemented?YesNo


4. What factors affect the cost and feasibility of flare minimization projects at your facility?





2. For those questions requesting a "yes" or "no" response, please enter an "a" in the yellow highlighted field next to the appropriateresponse.

3. Please feel free to add additional projects for questions 2 and 5, as necessary.




5. Have any projects been conducted to route additional uncontrolled streams to flare?YesNo



6. Is a flare gas recovery (FGR) system installed at the plant?YesNo

If no, has any consideration been given to FGR as a potential flare minimization project?YesNo

7. Are any HRVOC process vents routed to a thermal oxidizer for control?YesNo



TCEQ Work OrderNo. 582-07-84005-FY09-15 1

Facility QuestionnaireCost Analysis of HRVOC Controls on Refineries and Chemical Plants and

Control of Highly Reactive Volatile Organic Compound (HRVOC) Emissionsin Flares at Low Flow Conditions


For this project, ENVIRON will gather information on HRVOC emission reductionprojects at refineries, chemical plants and other HECT-affected facilities and use theinformation to perform an analysis of the costs of controlling HRVOC emissions fromdifferent types of facilities. ENVIRON will also gather certain information to betterunderstand how emergency flares are being used to control HRVOC emissions.

This interactive questionnaire contains 14 questions. We ask that you complete thisquestionnaire to the best of your ability and return it to ENVIRON by March 31,2009 . We realize this request may coincide with other reporting deadlines; therefore, wegreatly appreciate your efforts to complete this questionnaire by the end of March. Thequestionnaire is compatible with Adobe Acrobat Professional or Reader. You mayperiodically save your responses to the questionnaire to your desktop/computer so thatyou do not have to complete the questionnaire all at once. When you have completedthe questionnaire and are satisfied with your responses, please click the blue "Submit"button located at the top of the first page of the questionnaire. Next, select the optionthat best describes how you send email and click OK. Then, click "Send Data File" tosend the form's data file. When the email message is generated, please send the emailmessage. Should you have any questions regarding the questionnaire or the informationsought, please contact any of the following ENVIRON personnel.

• Chris Colville, +1 713.470.2647, [email protected]• Dr. Shagun Bhat, +1 713.470.2648, [email protected]• Steven Ramsey, +1 713.470.6657, [email protected]

1) What products are manufactured at your facility?

2) Please identify which of the following source types at your facility are in HRVOCservice and included in your HECT program cap:

Flares?

Cooling Towers 8,000 gpm?

Cooling Towers < 8,000 gpm?

Uncontrolled Process Vents?

Process Vents Routed to Control Device?

Process Vents Recycled to Process?

How Many?

How Many?

How Many?

How Many?

How Many?

How Many?

NOYES

NOYES

YES NO

YES NO

YES NO

YES NO

Submit by Email


3) Have any HRVOC emission reduction projects been implemented at your facility for the following units for the purpose of reducing routine, MSS or event emissions?

Cooling Towers?

If YES, please provide details in Table 3 (a).

Controlled Process Vents?

If YES, please provide details in Table 3 (b).

Uncontrolled Process Vents?

If YES, please provide details in Table 3 (c).

Flares?

If YES, please provide details in Table 3 (d).

These projects could include installation of emission controls, such as routing anuncontrolled process vent to a flare header, changes in the manufacturing process,change in operating procedures, elimination of HRVOC sources, etc.

Table 3 (a). HRVOC Emission Reduction Projects - Cooling TowersRequestedInformation Project 1 Project 2 Project 3

ProjectDescription:EstimatedHRVOCReduction (tpy):

Capital Cost ($):

Annual Cost($/year):

Table 3 (b). HRVOC Emission Reduction Projects - Controlled Process VentsRequestedInformation Project 1 Project 2 Project 3


Capital Cost ($):


NOYES

NOYES

NOYES

NOYES


Table 3 (c). HRVOC Emission Reduction Projects - Uncontrolled Process VentsRequestedInformation Project 1 Project 2 Project 3


Capital Cost ($):


Table 3 (d). HRVOC Emission Reduction Projects - FlaresRequestedInformation Project 1 Project 2 Project 3


Capital Cost ($):


4) Have any HRVOC emission reduction projects been considered, but not implemented?

If YES, please provide details in Table 4.

Table 4. HRVOC Emission Reduction Projects Not ImplementedRequestedInformation Project 1 Project 2 Project 3

ProjectDescription:Source Type(e.g. processvent):ProjectedHRVOCReduction (tpy):Reasons (e.g.,high cost,technicalinfeasibility, etc)- Please explain.

NOYES


5) Are there any HRVOC emission reduction methods that would be technically feasible for new facilities that manufacture the same product as your facility, but not for existing facilities?

If YES, please provide details.

6) Is there anything unique to your facility that would make HRVOC emission reduction more difficult or expensive?


7) While not a part of the HECT program, control of fugitive sources is a key component of reducing HRVOC emissions. Have investments been made to reduce the fugitive release of HRVOCs at the facility?


Table 7. HRVOC Emission Reduction Projects - Fugitive ReleasesRequestedInformation Project 1 Project 2 Project 3


Capital Cost ($):


8) Have any process units been shutdown or derated as a strategy for complying with the HECT cap?


NOYES

NOYES

NOYES

NOYES


9) For Flares in HRVOC service, please provide details in Table 9.

Table 9. Flare Operation During Calendar Year 2008Flare Service

Flare EPNDesign

Capacity(MMscf/

hr)

NormalProcess

Only

Event& MSSOnly

NormalProcess+ Event& MSS

AverageLoading -ExcludingEvent &

MSS(MMscf/hr)

AverageLoading -Event &

MSS only(MMscf/hr)


10) For the calendar year 2008, how long (in hours) did each flare operate under normal process conditions and Emission Event and MSS conditions? Please provide details in Table 10.

Table 10. Flare Operating Hours During CalendarYear 2008

Flare Operation (Hours)Flare EPN Normal Process

Only Event & MSS Only


11) Have any Flare Gas Recovery (FGR) projects been implemented at the facility?

If YES, please provide details in Table 11 (a).

If NO, please provide reasons in Table 11 (b).

Table 11 (a). Flare Gas Recovery Projects ImplementedRequestedInformation Project 1 Project 2 Project 3


Capital Cost ($):


Table 11 (b). Flare Gas Recovery Projects Considered But Not ImplementedRequestedInformation Project 1 Project 2 Project 3

ProjectDescription:

ProjectedHRVOCReduction (tpy):Reasons (e.g.,high cost,technicalinfeasibility, etc)- Please explain.

NOYES


12) Are any process vents in HRVOC service routed to control devices other than flares (e.g., thermal oxidizers, process heaters, etc.)?


Table 12. Controlled Process VentsProcess Vent Controlled Control Device HRVOC Emission

Reduction (tpy)

13) Have projects, that were not specifically carried out to reduce HRVOC emissions but resulted in reducing HRVOC emissions, been implemented at the facility (e.g., construction or enhancement of a Flare Gas Recovery system by a refinery under the Global Petroleum Refinery Consent Decree)?


Table 13. Other ProjectsRequestedInformation Project 1 Project 2 Project 3


Capital Cost ($):


14) Please provide any additional comments/information related to HRVOC control atyour facility that you would like to share:

NOYES

NOYES

TEXAS COMMISSION ON ENVIRONMENTAL QUALITY

WORK ASSIGNMENT 5

DRAFT FLARE WASTE GAS FLOW RATE AND COMPOSITION MEASUREMENT METHODOLOGIES EVALUATION DOCUMENT

Prepared By:

Shell Global Solutions

Shell Global Solutions (US) Inc. 3333 Highway 6 South Houston, Texas 77082

Lee K. Gilmer - Principal Investigator

Christopher A. Caico Jacob J. Sherrick Gary R. Mueller

Karl R. Loos

TCEQ Work Assignment 5 Flare Waste Gas Flow Rate and Composition

Measurement Methodologies Evaluation DRAFT REPORT

ii

TABLE OF CONTENTS

Page

Executive Summary...................................................................................................................E-1

1.0 INTRODUCTION................................................................................................................ 1-1 1.3 Typical Flare Header/Tip Sizes ........................................................................... 1-4

1.4 Typical Operating Conditions.............................................................................. 1-4

1.5 Typical Flow Rates/Ranges ................................................................................. 1-4

1.5.1 Typical Velocity Ranges in Flare Systems....................................................... 1-5

1.5.2 NSPS J Flow Rate Limitations.......................................................................... 1-5 1.6 Typical Ranges/Variability of Waste Gas Composition...................................... 1-6

1.7 Inherent flow meter precision and accuracy ........................................................ 1-9

1.8 Secondary selection criteria ............................................................................... 1-10

1.9 Summary ............................................................................................................ 1-10

2.0 FLARE WASTE GAS FLOW MEASUREMENT ................................................................... 2-1 2.1 Flow Measurement Fundamentals ....................................................................... 2-1

2.2 Flow Meter Evaluations....................................................................................... 2-1

2.2.1 Pitot Tubes.......................................................................................................... 2-4

2.2.2 Thermal Mass Meters........................................................................................ 2-8

2.2.3 Ultrasonic Time-of-Flight Meters................................................................... 2-13

2.2.4 Combination Systems ...................................................................................... 2-18 2.3 Factors Influencing Measurement Performance ................................................ 2-19

2.3.1 Measurement Uncertainties ............................................................................ 2-19

2.3.2 Combination Systems ...................................................................................... 2-22 2.4 Summary Overall Results of Flow Meter Evaluation........................................ 2-33

3.0 FLARE WASTE GAS COMPOSITION MEASUREMENT...................................................... 3-1 3.1 Fundamentals ....................................................................................................... 3-1

3.2 Analytical Performance Criteria Applicable to Flare Systems ............................ 3-1

3.2.1 Representative Sampling................................................................................... 3-1

3.2.2 Constituents of Interest ..................................................................................... 3-1



iii

3.2.3 Dynamic Range .................................................................................................. 3-2

3.2.4 Separation........................................................................................................... 3-3

3.2.5 Quantification..................................................................................................... 3-3 3.3 Basics of Analytical Instrumentation Applicable to Flare Systems..................... 3-4

3.3.1 Gas-Solid Chromatography .............................................................................. 3-4

3.3.2 Non Dispersive Infrared (NDIR) ...................................................................... 3-5

3.3.3 Photoacoustic Sensor ......................................................................................... 3-6 3.4 Offline Techniques for Measurement of Flare Gas Compositions ...................... 3-8

3.4.1 EPA Method 18 .................................................................................................. 3-8

3.4.2 ASTM D-1945 and D-1946 .............................................................................. 3-10

3.4.3 GPA 2261 and 2177.......................................................................................... 3-12

3.4.4 Multichannel Micro GCs................................................................................. 3-14

3.4.5 Summary of Offline Methods ......................................................................... 3-19 3.5 Online Techniques for Measurement of Flare Gas Compositions..................... 3-20

3.5.1 Total VOC Analyzers ...................................................................................... 3-23

3.5.2 NDIR Analyzers ............................................................................................... 3-23

3.5.3 Permanent Online Monitoring Gas Chromatographs.................................. 3-27

3.5.4 Permanent Online Monitors: Commercially Available Systems................. 3-28

3.5.5 Photoacoustic Spectroscopy ............................................................................ 3-32

3.5.6 Summary of Online Methods.......................................................................... 3-33

4.0 FLARE GAS EMISSION CALCULATION ERROR ANALYSIS.............................................. 4-1 4.1 Overall Error in Calculated Emissions ................................................................ 4-1

5.0 CONTROL OF GAS ASSIST RATIO ON OPERATING FLARES............................................ 5-1 5.1 Introduction.......................................................................................................... 5-1

5.2 Examples of Assist Gas Ratio Control Techniques ............................................. 5-2

5.3 Assist Gas Ratio Technique Performance............................................................ 5-4

5.4 Summary Results of Assist Gas Control Techniques Evaluation........................ 5-5

5.5 Developmental Recommendations ...................................................................... 5-6

6.0 CONCLUSIONS AND RECOMMENDATIONS....................................................................... 6-1

7.0 REFERENCES.................................................................................................................... 7-1

8.0 BIBLIOGRAPHY ................................................................................................................ 8-1



iv

LIST OF FIGURES

Page

Figure 1. Elevated Flare 1-3 Figure 2. John Zink Thermal Oxidizer Flare “Ground Flare” 1-3 Figure 3. Range of Observed Waste Gas Velocities in Flare Systems 1-6 Figure 4. Molecular Weight Variations of Waste Gas 1-7 Figure 5. Generic Model of Pitot Tube 2-5 Figure 6. Generic Example of Thermal Mass Flow Meter 2-10 Figure 7. Comparison of Various Flow Measurement Techniques 2-13 Figure 8. Generic Example of Ultrasonic Flow Meter 2-14 Figure 9. Example Gas Flow Meter Calibration Curve 2-26 Figure 10. Effect of Averaging Time on Measured Flows 2-33 Figure 11. Typical Gas Chromatographic Schematic 3-5 Figure 12. Non-Dispersive Infrared Gas Analyzer 3-6 Figure 13. The Principal Components of a PAS Detector 3-7 Figure 14. Effect of Assist Gas on Combustion Efficiency 5-7 Figure 15. Influence on Steam on Global Efficiency 5-8 Figure 16. Manual Assist Gas Ratio Control Example 5-9 Figure 17. Automated Assist Gas Ratio Control Example 5-9 Figure 18. Smoke Monitor Assist Gas Ratio Control Example 5-10

LIST OF TABLES

Page

Table 1. Flare Waste Gas Compositions – Constituents and Variability 1-9 Table 2. Flow Meter Selection Table 2-3 Table 3. Typical Micro GC Configuration 3-15 Table 4. Summary of Offline Waste Gas Composition Measurement Techniques 3-17 Table 5. Online Non-Speciated Waste Gas Composition Measurement Techniques

Evaluation 3-21 Table 6. Online Speciated Waste Gas Composition Measurement Techniques Evaluation 3-25 Table 7. Emerging Waste Gas Composition Measurement Techniques 3-35



E-1

Executive Summary

The Texas Commission on Environmental Quality (TCEQ) has determined that Volatile Organic Compound (VOC) emissions may be underestimated in air shed emission inventories. It is expected that emissions from flares would be better estimated if they were based on waste gas flow rate and composition measurements. The TCEQ has established data quality objectives for conducting such measurements. This study evaluates currently available flow measurement and gas composition measurement techniques; their applicability to the measurement of waste gas flow and composition in flare systems and their ability to meet the data quality objectives established by TCEQ. Additionally, this study evaluates current practices for controlling assist gas to waste gas ratios in flares. The overall objective of the Texas Commission on Environmental Quality (TCEQ) studies on flare emissions is to obtain performance specifications that ensure quality assured sampling, testing, monitoring, measurement, and monitoring systems for waste gas flow rate, waste gas composition, and assist gas flow rate. Specific objectives of this report are shown below. To evaluate existing methods of monitoring or measuring waste gas flow rate and composition to determine the mass rate of volatile organic compounds (VOC) being routed to the flare within a known certainty. The overall desired accuracy, precision, and sensitivity is + 20 percent. Specifically for flow rate, composition measurements and the resulting calculated mass rate to the flare the desired accuracy, precision and sensitivity is + 20% (never to exceed + 5000 lb/hr) of total VOC emitted from the flare, whichever is greater. An additional consideration is the desire that the waste gas composition measurement methods must be able to provide 90% speciation by mass of the individual VOC in the waste gas with specific focus on propylene, ethylene, formaldehyde, acetaldehyde, isoprene, all the butenes/butylenes, 1,3-butadiene, toluene, all pentenes, all trimethylbenzenes, all xylenes, and all ethyltoluenes. To evaluate methods of controlling the assist gas (steam or air) to waste gas ratio to assure the proper ratio is maintained within a known certainty. The overall desired accuracy, precision, and sensitivity is + 20 percent.



2

Flare systems are typically designed to burn large quantities of gas resulting from de-inventorying process units during emergencies. However, frequently they are also used to dispose of smaller quantities of waste gas, which are uneconomical to recover for beneficial use or as fuel. This dual mode of operation results in typical flows that are quite low, but sporadic high flow events. By design the number of compounds of interest and the dynamic range of these compounds can be quite large and variable. As a result of a detailed review of operating data from over 40 flares systems, current flow measurement techniques, and gas composition measurement techniques the following conclusions and recommendations have been developed. Waste Gas Flow Measurement Range Due to the unique requirements of flare systems (low pressures, large pipe sizes, and 1000:1 turndown on flows) only multi-ranged pitot tubes, thermal mass meters, and ultrasonic time-of-flight meters are broadly applicable to the measurement of flows in flare systems. Ultrasonic time-of-flight meters and Thermal mass meters can measure a greater range of gas velocities, and therefore are easier to apply to flare systems than pitot tubes. For the flares examined, all flows were below velocities of 300 fps, the high end measurable by ultrasonic and thermal mass meters. However, many flares are designed for worst-case events that will exceed this measurement capability. Flow measurement systems that combine multiple measurement techniques can measure this design worst case. Using multiple measurement techniques increases the costs associated with purchase, installation, calibration, and maintenance of dual meters. Design worst-case events are infrequent and rarely occur. Typically, design worst-case events are associated with emergency, unplanned shutdowns of a complete process unit or the shutdown of several process units serving a flare due to a power failure. Other process data usually exist which can be used to estimate the mass flow during such events. Thus, the additional expense of installing multiple measurement systems may not be warranted due to the relative infrequency of such events and the availability of process data that can be used to estimate mass flow during worst-case design events. Effects of Waste Gas Composition on Flow Measurements Ultrasonic time-of-flight meters have greater accuracy and precision for measuring flows in flare gas systems, because they are less sensitive to changes in gas composition and can correct for variable gas composition using only flow meter inputs. Thermal mass meters and pitot tubes



3

need significant correction for changing gas compositions. Currently, real time correction for gas composition changes has not been demonstrated in field applications with thermal mass meters or pitot tubes, although vendors claim such meters are available. Correcting with periodic composition data adds to the uncertainty of the flow measurement of thermal mass meters and pitot tubes used in flare applications. Measurement of flows in flare systems can be made with an uncertainty in the range of + 5-10 percent. However, obtaining this accuracy in flare systems with highly variable compositions or where the meter cannot be located in a section of pipe with a representative flow profile will be a challenge. Overall Flow Measurement Assessment Currently, ultrasonic time of flight meters show the most promise of meeting the stated data quality objectives of this study in waste gas service. However, their cost probably limits their application to flow measurements at the terminal end of flare systems . Thermal mass meters offer a suitable alternative in services where the waste gas composition does not vary greatly, and as a sturdy portable instrument that can be used to troubleshoot flare systems (e.g. identify stray flows in complex flare headers). Flare Waste Gas Composition Measurement Measurement of the gas composition in flare systems on a continuous basis will be a challenge due to the number of compounds of interest and the dynamic range of concentrations, that may be present for the compounds of interest. To design and implement a continuous monitoring system capable of detecting the full range of compounds of interest across all concentrations of interest using conventional GCs will require considerable up-front data collection, and fairly sophisticated combinations of columns and detectors. If aldehydes happen to be present at quantities that would impact achieving the data quality objectives, the GC system required to analyze for them will become even more complex. Portable multi-channel, micro GCs have proven extremely capable of detecting most compounds of interest in flare systems across an extremely large dynamic range (105). Application of such instruments in online applications will significantly reduce the complexity of the GC instrumentation needed to achieve the data quality objectives of this study. However, online versions of these instruments have yet to be demonstrated in flare applications. While online GC



4

systems can be designed to meet the data quality objectives of this study a high percentage of the time, the actual percent of time the data quality objectives can be met will be a strong function of the specific flare system, in particular the variability of the waste gas composition. There is insufficient data available to make a fact-based estimate on this percentage. Application of US EPA Performance Specification 9 to online GCs in flare service is problematic given the potential number of compounds of interest and the dynamic range potentially encountered for these compounds. The linearity requirement will be extremely difficult to meet and the calibration procedures impractical. Picking a few representative compounds with which to calibrate and using relative response factors for other compounds of interest seems a practical solution. Due to the large dynamic range, the linearity requirement of PS9 may need to be relaxed in favor of a multipoint calibration curve. Assist Gas Rate Automatic controlling of the assist gas to waste gas ratio based on waste gas flow to the flare appears to be more effective at controlling this parameter at a desired value than manual control, but there was very little data available from which to make this conclusion. Additional data collection on the effectiveness of manual control would be warranted. Waste Gas Flow and Composition Measurement Data Quality The author’s assessment is that current flow and measurement technologies are capable of meeting the stated data quality objectives a high percentage of the time in most flares applications. However, the scarcity of data on real world applications and the high degree of variability in flare applications make it impossible to quantify what percentage of the time these data quality objectives can be met, and on what percent of flare systems they may be met. Both flow and gas composition measurement vendors have responded to the added interest in measurement of waste gas flow and composition. Significant recent developments have been made in both flow and composition measurement techniques. Demonstration of these advancements could significantly enhance the ability of sources to meet these data quality objectives over a wider range of operations in the near future.



5

Under ideal conditions the mass flow in flare systems can be measured with an uncertainty in the + 5-10% range. However, problematic applications with widely varying gas compositions and poor meter installation geometries will have difficulty meeting the + 20% data quality objectives of this study. Obtaining measurement of flare emissions within an absolute uncertainty of + 5,000 lb/hr appears doable for all but the very largest flare. However, this assessment is strongly influenced by the 98% destruction efficiency assumption used in this study.



1-1

1.0 INTRODUCTION

The Texas Commission on Environmental Quality (TCEQ) has determined that flares may be a significant source of emissions of volatile organic compounds (VOC). Of particular interest are highly reactive volatile organic compounds (HRVOC), which may contribute to rapid ozone formation under certain atmospheric conditions. TCEQ is concerned that flare emissions may be one source of VOC and HRVOC emissions that are being underestimated in current air emission inventories and wants to investigate methods to improve emission estimates from this source. This study evaluates current flow measurement and gas composition measurement technologies, and their applicability to measuring waste gas flows and compositions to flares. In the past, other than requiring pilot lights to be monitored and lit at all times as well as limiting flare tip exit velocities and flare waste gas heating values, flares have not been subject to environmental regulations mandating accurate measurement of waste gas flows and compositions. Typically, flare waste gas flow rate and compositional measurements have been used to assist in designing flare gas recovery compressors, to identify sources upstream of the flare contributing to flare flows, to quantify the change in flare waste gas flow rates and compositions associated with process changes upstream of the flare header, and to assess normal flows versus design flows. The level of accuracy required for these purposes is not as stringent as that associated with regulatory requirements. Thus, existing instruments for measuring flare waste gas flows and compositions were not designed to meet stringent regulatory requirements. TCEQ has established preliminary data quality objectives for measurements of waste gas flows to flares and wants to assess the ability of current technology to obtain these objectives. Further, as assist gas addition to flares is a primary means of improving combustion efficiency in flares at high load conditions, the TCEQ would like to evaluate methods used to control assist gas addition to flares, and assess the ability of these methods to control the assist gas ratio within known certainties. This study assesses the ability of three common methods to control the assist gas to waste gas ratio.

1.1 Objectives

The overall objective of the TCEQ studies on flare emissions is to obtain performance specifications that ensure quality assured sampling, testing, monitoring, measurement, and monitoring systems for waste gas flow rate, waste gas composition, and assist gas flow rate. The



1-2

purpose of this project is to provide data and methodologies to improve emission estimates from flares. Specific objectives of this report are: To evaluate existing methods of monitoring or measuring waste gas flow rate and composition to determine the mass rate of volatile organic compounds (VOC) being routed to the flare within a known certainty. The overall desired accuracy, precision, and sensitivity is + 20 percent. Specifically for flow rate, composition measurements and the resulting calculated mass rate to the flare the desired accuracy, precision and sensitivity is + 20% (never to exceed + 5000 lb/hr) of total VOC emitted from the flare, whichever is greater. An additional consideration is the desire that the waste gas composition measurement methods must be able to provide 90% speciation by mass of the individual VOC in the waste gas with specific focus on propylene, ethylene, formaldehyde, acetaldehyde, isoprene, all the butenes/butylenes, 1,3-butadiene, toluene, all pentenes, all trimethylbenzenes, all xylenes, and all ethyltoluenes. To evaluate methods of controlling the assist gas (steam or air) to waste gas ratio to assure the proper ratio is maintained within a known certainty. The overall desired accuracy, precision, and sensitivity is + 20 percent.

1.2 Flare System Design and Operations

1.2.1 Purpose

Process flares are used as safety devices to combust large quantities of waste gases during emergency conditions, such as power failures or process upsets. Additionally, many facilities routinely flare smaller quantities of hydrocarbon gases that are exhausted at pressures too low to allow them to be routed to the facility fuel gas system or hydrocarbon gases that are uneconomical to recover. Smaller ground flares are sometimes used to handle the smaller, more routine flows while higher capacity elevated flares are used for emergency condition flaring. Ground flares are designed for lower flow rates than elevated flares and typically the burners are enclosed in a stack. Also, their design (in the case of thermal oxidizers such as the one shown in Figure 2) results in a wider, less elevated, less visible, and less noisy flame. In simple terms, a thermal oxidizer ground flare flame is similar to a burner on a gas stove (several small flames from several points on a circle) while a traditional elevated flare flame is similar to a lit match (a larger, more defined flame emitted from a point). Figure(s) 1 and 2 provide schematics of typical ground and elevated flare systems.



1-3

1.2.2 Schematics

Figure 1. Elevated Flare1

Figure 2. John Zink Thermal Oxidizer Flare “Ground Flare” 2

1 Figure 1 is courtesy of John Zink Company. 2 Figure 2 is courtesy of John Zink Company.



1-4

1.3 Typical Flare Header/Tip Sizes

The authors have worked with or knowledge of the design and operation of approximately 40 flare systems in numerous U.S. refineries and chemical plants. The range in size of the terminal flare header, just prior to the flare, in these flare systems is from 24 to 60 inches in diameter with about 75% having a terminal diameter in the 36 to 48 inch range. Flare tip diameters utilized in these 40 flares range from 24 to 36 inches with over 80% in the 30 to 36 inch range. Therefore, to be broadly applicable flare waste gas flow measurement devices should be capable of being applied in line sizes from 4” up to 72”.

1.4 Typical Operating Conditions

The operating temperature in flare headers is typically at or very near ambient temperature due to the long lengths of pipe from the process unit boundary to the flare. Flare headers normally operate at very low pressures, typically at 1 to 2 psig. Some upstream relief valves relieve at pressures as low as 5 psig, necessitating the flare header operating pressure be designed as low as possible to insure flow from all required process relief systems to the flare system. The low operating pressures of most flare systems requires that a maximum pressure drop specification across any potential flare waste gas flow measurement device be set at less than 0.5 psig. Assist gas (usually steam or air) may be routed to the flare flame at the flare tip to assist mixing of combustion air, minimize smoking, and to maximize combustion efficiency. Since assist gas is added at the flare tip or into the flare riser, this gas will have no impact on the measurement of waste gas flows and compositions to the flare.

1.5 Typical Flow Rates/Ranges

Flare waste gas flows can range from nominally small flow rates of a few standard cubic feet per hour (SCFH) to several million SCFH. The U.S. EPA New Source Performance Standard for Flares (40CFR60.18) limits the maximum tip velocity for all flares constructed since 1973 to 400 feet per second, if the flare gas has a BTU content of at least 1,000 BTU/SCF and 60 feet per second for flares combusting gas with a BTU content of 300-1000 BTU/SCF. For the tip and header sizes in the flare systems the authors have worked with this velocity limitation equates to a maximum allowable volumetric flow rate of 4.5 million to 10 million SCFH or 100 to over 240



1-5

million standard cubic feet per day (for 24-36” flare tips). The authors’ discussions with knowledgeable individuals from other companies would indicate that the above flow rate ranges are representative of flares at many of the petrochemical facilities in the Houston-Galveston airshed. 1.5.1 Typical Velocity Ranges in Flare Systems

Most flow measurement devices that will be applicable to flare systems work on the principle of sensing the velocity of the waste gas stream, which is then converted into a volumetric or mass flow measurement. The large difference in flows between design maximums and normal operating flows for most flare systems leads to a requirement that any flow measurement device applied in flare systems to be capable of sensing a wide range of velocities. While there is no such thing as a typical flare system, hourly average data from 5 flares were examined over an eight-month period to demonstrate the range of velocities encountered in flare systems. These data are presented in Figure 3. A cursory analysis of the data would indicate that anywhere from 90-98% of the hourly averages are in the 0.1-20 fps range, and the balances of the hourly averages (2-10%) are in the 20-100 fps range. While the design case for some of these flares is in the 275-400 fps range, velocities in this range are rare and seldom, if ever, seen. In the event that a design case scenario does occur, there frequently will be other process data available that can be used to estimate the quantity of material that is sent to the flare, such as which portions of the unit were isolated and de-inventoried.

1.5.2 NSPS J Flow Rate Limitations

Given the NSPS Subpart J regulatory limit of 400 feet per second (fps) tip velocity limit applicable to many flares, a working range of 0-400 fps was established as the desired range for flow measurement devices applicable to flare systems. The limited data presented suggest that for many applications a range of velocities from 0-100 fps will cover the bulk of the flaring events. In addition to the ability to detect velocities up to this maximum velocity, the flow measurement device must also have the ability to quantify velocities across a three decades range of velocities (e.g. 1000:1 turndown ratio) to give sufficient accuracy to cover the range of encountered flows.



1-6

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

80.0

90.0

0 30 60 90 120 150 180 210 240 270

Run Time, Days

Was

te G

as V

eloc

ity, F

eet P

er S

econ

d

Flare AFlare BFlare CFlare DFlare E

Figure 3. Range of Observed Waste Gas Velocities in Flare Systems 1.6 Typical Ranges/Variability of Waste Gas Composition

Compositions of waste gases routed to flares can vary significantly depending on the nature of the event(s) that are causing the need to flare waste gas. Sweep or “purge” gas is maintained in the flare header at all times to ensure clear lines and positive flow to the flare tip. Sweep gas normally is natural gas or fuel gas. At low flow rates the sweep gas flow may have a significant impact on the flow and composition measurements depending on the actual locations of the meters and the sweep gas injection point. Figure 4 illustrates the variability in the molecular weight of waste gas observed at two process unit flares, as measured by online flow measurement instrumentation.



1-7

Figure 4. Molecular Weight Variations of Waste Gas

0

5

10

15

20

25

30

35

40

45

50

0 30 60 90 120

150

180

210

240

270

Run Time, Days

Was

te G

as M

olec

ular

Wei

ght

Flare A FlareC



1-8

These two examples clearly illustrate significantly different types of flare applications. A flare that is has a relatively consistent composition, and one that sees great changes in composition, in terms of molecular weight and perhaps even in components of interest. Table 1 gives additional examples of waste gas compositions in terms of both constituent of interest and the degree of variation of these components. Table 1 shows 2 cases as measured at different times for each of three flares. Note there is significant variation between the first and 2nd case for each flare. Also note that the recoveries ranged from 89.86 to 100.02 depending on the flare, the components analyzed for, and the components identified. The composition of the waste gas being measured can have a significant impact on the accuracy of many flow measurement technologies. Correcting measured flow rates for these compositional changes will be discussed later.



1-9

Table 1. Flare Waste Gas Compositions – Constituents and Variability

Table 1. Examples of Flare Waste Gas Compositons-Constituents of Interest and Variability

Component mole % mole % mole % mole % mole % mole %hydrogen 86.18 48.77 0.02 0.04 37.11 0.00methane 5.93 3.52 86.22 47.09 14.73 0.70ethane 0.81 0.26 1.06 0.41 0.44 0.21ethylene 0.02 0.01 0.13 0.22 0.18 16.37propane 0.34 0.14 0.10 43.30 0.07 0.89propylene 0.00 0.01 0.01 0.23 0.12 24.29n-butane 0.11 0.05 0.02 0.21 0.35 0.03i-butane 0.11 0.06 0.03 0.05 0.02 0.00cis, 2-butylene 0.16 0.06 ND ND 0.24 0.01trans, 2-butylene 0.17 0.06 ND ND 0.26 0.00isobutylene 0.12 ND ND ND 0.211,3-butadiene ND ND ND ND 0.74 0.01n-pentane 0.03 0.08 0.02 NDi-pentane 0.05 0.05 0.03 NDpentenes ND ND ND NDC6

+ 0.01 0.01 ND NDCO 0.02 0.04 ND NDN2 4.99 45.80 0.38 1.69 31.01 50.40O2CO2 0.06 0.04 1.11 0.07 0.15hydrogen sulfide 0.24 0.35 0.00 0.00 0.03water vapor 0.68 0.70 0.73 0.87Totals 100.02 100.00 89.86 94.18 85.63 92.94

Flare 1 Flare 2 Flare 3

1.7 Inherent Flow Meter Precision and Accuracy

Each flow measurement device will have an inherent precision and accuracy associated with that technology and model. These metrics will be a function of the fundamental property being measured (e.g. pressure, temperature, frequency, etc.) and the components used in the meter. These metrics are determined under ideal conditions in the laboratory. While the inherent precision and accuracy of the meter is but one of the components that goes into determining the precision and accuracy of the flow measurement in a field application, it is an important specification to qualify flow measurement technologies. Due to the extreme variations and conditions encountered in flare applications, it is important to have relatively high specifications for the inherent precision and accuracy of flow measurement devices. For this reason an inherent precision specification of +1% and an inherent accuracy of +5% was selected for this study. The reason for selecting such tight inherent flow meter precision and accuracy specifications will be



1-10

addressed later in the discussion of the roll up of errors in the overall uncertainty of the mass flow of waste gas to flares.

1.8 Secondary selection criteria

There are a number of other selection criteria, which while important in the selection of a flow measurement device can to some degree be engineered around. These include such characteristics as applicability to corrosive environments and environments with aerosols and particulate matter. As well as what temperatures and pressures the flow measurement device can be applied. Other factors such as the length of straight pipe required for installation, and the ability to detect backflow in the line can also be of importance. While not primary selection criteria, these characteristics where also assessed for flow measurement technologies which satisfied all of the primary selection criteria.

1.9 Summary

Based upon the above description of flare design and operating parameters the following list of primary design specifications were selected to assess whether available flow measurement technology was applicable to measuring flows in flare systems.

• Applicable in lines from 4”-72” in diameter.

• Maximum pressure drop across measurement device of <0.5 psig.

• Applicable to measurement of velocities from 0-400 ft/sec.

• Quantification of flow (velocities) across 3 decades (1000:1 turndown).

• Inherent flow meter accuracy of + 5 percent.

• Inherent flow meter precision of + 1 percent.

There are a number of secondary criteria such as applicability to corrosive service, specific temperature and pressure constraints, length of straight pipe required, and the ability to sense backflow that are important, but can potentially be engineered around or into a particular



1-11

application. These criteria were reviewed and used to evaluate the performance of flow measurement technologies for application in flare service.



2-1

2.0 FLARE WASTE GAS FLOW MEASUREMENT

2.1 Flow Measurement Fundamentals

Virtually all flow measurement devices calculate volumetric flow by sensing a velocity in the pipe and multiplying that velocity by the cross sectional area of the pipe in which the velocity is being sensed. This velocity is assumed to be uniform across the cross section. Applications where this may not be a good assumption may have reduced precision and accuracy. In fact, the consensus of the flow measurement vendor representatives at a recent API meeting to begin developing an API Standard for measuring flow rates in flares was that the flow profile is the most important consideration in the choice, location, and operation of any flare flow meter.28

Inherent in all of these velocity measurements is an assumption that the gas is of a known composition. Deviation from this assumed composition must be corrected to maintain the overall precision and accuracy of the flow measurement. 2.2 Flow Meter Evaluations

A variety of flow meters were evaluated as a part of this report. The following types of flow meters were evaluated: Differential Pressure

• Orifice Run with Differential Pressure Taps

• Venturi Run with Differential Pressure Taps

• Pitot Tubes

• Rotameters

Mechanical

• Turbines

• Pistons/Disks/Impellers

• Dry and Wet Test Meters



2-2

Mass

• Thermal

• Coriolis

Electronic

• Magnetic Meters

• Vortex Meters

• Ultrasonic Meters

A brief description of the principles of operation for each of these meters and their applicability to flare applications is given in Appendix A. The performance characteristics of each meter type versus the evaluation criteria developed previously for flare systems is summarized in Table 2. If a flow measurement technique failed one or more of the primary evaluation criteria developed for flare systems, it was not considered further. Of the technologies evaluated only constant temperature anemometers (thermal mass meters), pitot tubes, ultrasonic time of flight meters, and possibly a recently developed Doppler shredding vortex meter did not fail one or more of the primary evaluation criteria. However, all of these meters had at least one or more concerns when evaluated against the complete list of flare flow measurement selection criteria. Application of these four techniques to flare applications will be discussed in more detail in the following sections.



2-3

Table 2. Flow Meter Selection Table • See Note At Bottom of Table

Flare Waste Gas

Flow Meter Vapor

Service Pipe ID Range

Backpressure @ Max Design Flow

Obstructs Flare Header

Flow

Corrosive Service

Backflow Measurement

Capability

Aerosols and Particulates

Transmitter Signal Output

Straight Pipe

Required

Principle of Operation

Process Velocity Range

Process Temperature

Range

Process Pressure

Range

Stated Model Accuracy

TD Ratio

Table I Flow Meter Evaluation Results Evaluation Criteria Yes 4"-72" <0.5 psig No Yes Yes Yes < 100' 0-1000 FPS -20F to 300F 0 to 60 psig

minimum + 5% 1000:1

minimum FLOWMETER Constant Temperature Anemometer (Thermal Mass Meter)

Yes all < 0.1 psig N notrecommended

No Yes analog 10 Feet Thermal - Mass 0-300 FPS -40C to 200C or 500C model variation

150 psig 1-3% varying with Temperature

High

Pitot Tubes and Annubars

Yes 0.5 to 72" < 0.5 psig No not recommended

Yes Yes transducer(mA)

48 Feet Pressure - Velocity

10 to 1000 FPS

up to 546C 2600 psig 1%, depending onapplication (sensor)

Medium 30:1

Ultrasonic Time Of Flight Flowmeters

Yes 3-120" < 0.1 psig No notrecommended

Yes Yes 100 Feet time transit 0.1-275 FPS -166 to 300F or 500F optional

1500psig 1 path 3 - 7%, 2 path 2.4 - 5%

High 2750:1

Vortex Meters Yes 1-5 or more psig Yes not recommended

Magnetic Meters No not recommended

Coriolis Meters Yes < 6" 10 or more psig not recommended

Dry Test Meters Yes < 6" 1-5 or more psig Yes not recommended

Wet test Meters Yes < 6" 1-5 or more psig Yes not recommended

Pistons/Discs/Impellers

Yes 1-5 or more psig Yes not recommended

Turbines Yes 1-5 or more psig Possibly not recommended

Rotameters Yes < 4" 1-5 or more psig Yes not recommended

3-10:1

Venturis Yes 1-5 or more psig Yes not recommended

3-10:1

Orifice Plates Yes 1-5 or more psig Yes Not recommended

3-10:1

* Traditional Vortex Meters were ruled out here due to their bluff body design creating excessive pressure drop. A new, novel design, Doppler Shredding Vortex meter that may be able to overcome this limitation was recently developed. According to the Manufacturer and their representative, several of these meters were recently installed in flare headers. The authors have no data or experience with these meters. However, they seem to have potential. Additional information on these meters is available from the manufacturer’s representative.30



2-4

2.2.1 Pitot Tubes

Principle of Operation A pitot tube calculates a point velocity by measuring the static pressure and the impact or velocity pressure. The static pressure is the operating pressure of the pipe, duct, or stack at a specific point. The impact pressure is a combination of the static pressure and the pressure created by the flowing fluid. Pitot tubes generally measure the impact velocity by situating the opening directly into the oncoming fluid stream. Static ports are often located perpendicular to the fluid flow either on the tube itself or on the pipe or duct. The two pressures create a differential pressure that is directly related to the velocity through the following equation: Equation 1

( )( )21/PPCV ST fp ℘−×=

Where: Vp = Point Velocity C = Dimensional constant ρf = Density of the fluid PT = Impact Pressure (a) PS = Static Pressure (b) This equation gives a linear velocity, not the mass flow rate. The mass flow rate is calculated from the flare header’s cross-sectional area and the molecular weight of the fluid being measured assuming ideal gas behavior of the flared waste gas. In the petrochemical industry, pitot tubes are primarily used to conduct isokinetic sampling traverses in boiler and process heater stacks. The measured flow rate is corrected for the actual fluid using empirically derived K-factors. These measurements are traditionally post-combustion measurements in internal or external combustion device stacks where the gas is primarily air, nitrogen, and water vapor with trace concentrations (<0.1% by volume) of other constituents. Some vendors of averaging pitot tubes are currently advertising them as capable of accurately (within + 2-3%) measuring flows from 10-1000 feet per second in flare headers when used in conjunction with an online gas chromatograph to determine the actual composition of the gas being measured.



2-5

Design Variations There are several variations of pitot tubes: single-port, averaging, area averaging, and s-type. Single-port pitot tubes similar to the one shown in Figure 5, below measure velocity at one point in the flow stream with one static port and one impact port. They are typically used for turbulent streams with well-defined velocity profiles.

Figure 5. Generic Model of Pitot Tube3 To overcome the limitations of a single-port pitot tube, averaging pitot tubes were developed. Averaging pitot tubes have multiple impact pressure and static pressure ports that measure pressure across the pipe diameter. The pressures are tabulated and averaged, thereby negating the issue of placement of a single tube. Area averaging pitot tubes measure multiple impact and static pressures over the pipe diameter and also over a length of pipe. Area averaging pitot tubes usually are comprised of several sets of pitot tubes connected to a manifold. These are typically used for large stacks or ducts. S-type pitot tubes are specially designed to withstand dirty flows with a high level of particulate matter. Any of the above instruments could employ the s-type design for use in flows with high levels of particulate matter. For flare applications averaging pitot tubes are typically used. For the remainder of this section the applicability of pitot tubes to measuring flare flows will be discussed.

3 http://www.svce.ac.in/~msubbu/FM-WebBook/Unit-III/PitotTube.htm

http://www.svce.ac.in/~msubbu/FM-WebBook/Unit-III/PitotTube.htm



2-6

Applicability to Flares

Pitot tubes have been used for years to measure flare flows. This is primarily because until fairly recently pitot tubes were the only viable flow measurement option. Although pitot tubes are fully capable of measuring flare flows, they may not be the best solution for all applications. Using an averaging pitot tube to measure flare flow has several positives; but there are several limitations to the design and operation of the device as compared to some newer waste gas flow measurement technologies. Pitot tubes are capable of meeting the data quality objectives of this study across a limited range of flows (10-1000 feet/second). However, the stipulation that the flow meters measure through the entire range of the potential flow may cause problems for pitot tubes. As stated previously, flares are capable of having wide ranging flare flows, upwards of 100-1000:1. A typical averaging pitot tube turndown ratio is 30:12. A particular problem with pitot tubes is their ability to measure low flows in large diameter pipes, as the pressure differentials that must be sensed are quite low. To accurately measure over the entire possible design window for flares would require multiple pitot tubes calibrated over several ranges of flow. While this could be considered an acceptable approach to measuring flare waste gas flow rates, this has to the authors’ knowledge not been done. Several flow meter suppliers are now offering multiple ranges averaging pitots, but there is limited data available on their use in flare applications. Under ideal conditions the precision and accuracy of pitot tubes is quite good. For example, the quoted accuracy of the Dieterich Standard Diamond II Annubar is +1% with a precision of 0.1% for the sensor3. Ideal conditions for pitot tubes are stable flows, operating temperatures and pressures, and constant density of the stream being measured. These unfortunately are not common conditions in most flare applications. Correcting for the actual gas composition in most cases will induce as much or more uncertainty into the flow measurement than the inherent error of the flow meter being used. A more detailed discussion of the roll-up of error in flow measurements is presented in Section 4.0, but in most field applications of pitot tubes in waste gas service the total uncertainty of the flow measurement will be in the range of 5-20 percent. The exact value being a function not only of the inherent uncertainty of the flow meter, but also of the variability of the gas composition and one’s ability to measure and compensate for the actual gas composition.



2-7

Pitot tubes require a rather large amount of straight run piping prior to the device. If straightening vanes are not used, a straight run of between 10-25 pipe diameters is often quoted. Typical downstream straight pipe run requirements are 5-10 diameters. In many flare applications this may not be an issue due to the length and amount of piping needed to transport the flare gas to the remote areas usually used for flare locations. Where space is limited and the flare stacks are very near the flare knockout drum, this 10-25-pipe diameter requirement will definitely be problematic. Straightening vanes or other flow correcting devices can be used to minimize the impact on observed meter accuracy of not having the desired runs of straight pipe for installing pitot tubes. However, the installation of straightening vanes and/or other flow correcting devices in flare headers can pose a serious safety hazard, due to the potential to collect debris and inhibit flow, and may not be allowed in many flare installations. Other than applications requiring straightening vanes, the application of pitot tubes to flare systems should not pose any significant safety risk. The actual instrument is a slim, aerodynamic tube that does not obstruct the flow enough to be of significant concern for collecting hydrocarbons. Additionally there are no mechanical or moving parts that could be affected by electrical problems. Pitot tubes also induce very low-pressure loss across the device due to the unobtrusive design, typically less than 0.5 psig. One potential problem area for pitot tubes is measuring flows in corrosive service or applications with high levels of particulate matter. The pressure ports are prone to plug when used in dirty flows. Some designs attempt to alleviate this problem by using larger size portholes or flushing the holes with an inert gas. Periodically cleaning the ports and checking the differential pressure cell ensures device accuracy and usually is a routine task. Annual calibrations may also be required to ensure accurate performance. Pitot tubes can be used to measure flows in virtually any size pipe. Pitot tubes can be installed in piping in excess of 48" without incurring additional costs for installation. For example, the Diamond II+ Annubar is capable of being installed in piping from 0.5” to 72". 3 Additionally, most pitot tubes can function across a wide range of temperatures and pressures, up to 546oC and 2600 psi. This is well within the operating range for almost all flares in terms of temperature and pressure. In the authors’ experience, pitot tubes were originally installed in many flare headers to measure waste gas flow rates for informational purposes rather than for critical process controls or meeting regulatory requirements. Typically, the pitot tubes ports were found to have “plugged



2-8

up” from salts, polysulfides, ammonium sulfide, and other particulates that may form in flare headers from time to time. Also, many of the originally installed pitot tubes were found to be bent or broken as a result of maintenance activities on the flare headers. Typically, since the flow signal was not a critical control variable or a compliance parameter, the level of maintenance was not adequate to keep the pitot tube functioning. Also, the readings from these meters became more and more suspect as additional sources were tied into the flare line thereby increasing the potential for the waste gas composition to vary considerably. 2.2.2 Thermal Mass Meters


Thermal mass flow meters operate on the principle of measuring temperature differences between two sensors. Generally, a unit contains two temperature sensors, such as thermocouples, and a heating element. The temperature difference between the two sensors is a function of the flow past the sensor. There are two methods to achieve the temperature differential, constant temperature and constant power. In the constant power approach, one of the sensors receives a constant supply of energy, or heat, from an electric heater. The temperature rise of the sensor is measured as well as the heat supplied. The other sensor measures the operating temperature of the fluid. The constant temperature approach maintains one of the sensors at a specific temperature above operating conditions. The amount of energy needed to maintain the temperature is measured and used to compute a mass flow rate. As in the previous method the second sensor measures operating temperature conditions. One recently developed constant temperature thermal mass meter has two sets of sensors with one set in the traditional waste gas flow area and the other set very near the flare header wall. According to the vendor, this allows the meter to continually measure both a waste gas flow and a near-zero flow in real time. 29 Having two pairs of sensors, one in the main waste gas flow stream and one near the walls, allows the meter to obtain a flowing stream waste gas velocity and a near zero waste gas velocity. One pair of sensors could conceivably give the same velocity reading for two or more widely varying gas mixtures. According to the vendor29, no two gas compositions will have identical velocity readings at both the in stream velocity and at a near



2-9

zero velocity. Thus, the meter is able to identify the actual composition of the waste gas by comparing the two readings and correct the velocity measurement for the actual composition in real time. The underlying equation governing both methods of thermal mass flow meters is essentially the same: Equation 2:

( )[ ]12p TTCKQ/ M −=

Where: M = Mass flow rate K = Meter coefficient Q = amount of energy supplied Cp = specific heat of the flow T2 = Temperature of the constant power sensor T1 = Temperature at operating conditions. Per Equation 2, one can see that the observed flow from a thermal mass meter requires knowledge of the gas composition. Any significant deviation in composition of the waste gas being measured from the composition of the gas on which the meter was calibrated will impact the accuracy of the observed flow measurement. While real time correction of thermal mass meters for changes in gas composition are possible (correction for variable gas composition will be discussed further in Compositional impacts on calibration), there are a limited number of applications where this has been done in the petrochemical industry. Design Variations

There are several variations of thermal mass flow meters that use one of the methods mentioned above to measure mass flow rate including heated tube designs, bypass, insertion probes, and hot wire anemometers. Heated tube mass meters are designed to protect the heater and temperature sensor elements from corrosive or dirty flows. The heater is mounted on the outside of the pipe and directly heats a section of the pipe wall. One temperature sensor is situated far enough downstream of the heater to measure operating temperature conditions. A second sensor is upstream in close proximity with the electric-driven heater. The mass flow rate is a nonlinear



2-10

function of the heat transfer of the pipe walls and the sensor. Bypass thermal mass meters are generally used in high flow, high-pressure applications. Essentially a portion of the flow is bypassed into a tube that is heated. A portion of the energy is transferred to the gas, which is measured by temperature sensors. Due to the bypassing of flow, a large pressure drop is incurred across this type of meter, and generally is not suitable for flare applications.

Figure 6. Generic Example of Thermal Mass Flow Meter 4

Insertion probes use the constant temperature method stated above except the sensors or probes are inserted into the flow stream. One sensor is maintained at a constant temperature above ambient conditions. The energy necessary to maintain the temperature is recorded and used to calculate mass flow. Hot wire anemometers use the same principles as constant-temperature or constant-power mass flow meters. Anemometers use thin filaments to measure the heat transfer between a reference and a heated sensor. Thermal mass meters are capable of obtaining and relaying continuous measurements. There are several current petrochemical industry examples of this process. Turndown ratios are a minor issue for thermal meters as most can quantify flow velocities across a two-decade range (100:1) or more. Flares can have turndown ratios of 1000:1; therefore the use of two or more thermal meters may be necessary to cover the whole range of flow seen in these applications. The quoted accuracy from the K-bar 2000 from Kurz Instruments is listed as +1-3%, depending on operating temperature.69 Because thermal mass meters require the composition of the gas to be known to compute the flow, the uncertainty of the flow will also be a function of the



2-11

uncertainty of the gas composition. In the authors’ experience, total flow measurement error can be in the 5-10% range and could be as high as 20% in applications with widely varying compositions. Correcting measured linear velocities to actual mass flow rates can be problematic if the molecular weight of the waste gas varies by more than 20% from the molecular weight of the meter’s calibration gas. [More detail on this problem is presented Section 2.3 “Factors Influencing Measurement Performance”.] The repeatability of the instrument is given as 0.25%4. Many source-testing protocols dictate that flow meters have a minimum of 8 pipe diameters of straight pipe upstream and 2 pipe diameters of straight pipe downstream of the meter to ensure consistent velocity and velocity profiles. For most circumstances, thermal meters do not require additional space beyond these requirements. It is possible to install thermal meters by inserting them through existing gate valves on flare lines or permanently installing by hot-tapping into the flare system. Thermal meters do not present safety concerns as they induce very little pressure drop and do no not act as an obstruction in most circumstances. The instrument has low-pressure drops (< 0.1 psig) due to the unobtrusive design. Typical thermal meters consist of a metal probe that houses two thermal transducers. Corrosive or hazardous flows may be a problem for carbon steel instruments, but most instruments are typically constructed of stainless steel or other materials that resist corrosion or rusting. Thermal meters can be used across a wide range of temperatures and pressures, and for the most part are unaffected by density, viscosity, pressure, or temperature fluctuations. Therefore, process conditions tend to not be an issue when evaluating the applicability of a thermal mass meter. For example, the operating limits for the K-bar 2000 can range from -40 to 500oC over a pressure range of 150 psig. The flow meter is designed to measure flows from 0-18000 standard cubic feet per minute (SCFM). Thermal meters can be installed in pipe sizes well in excess of 48 inches. Typical costs of thermal meters are between $3,000-$5,000. Permanent installment increases the cost as does harsh operating conditions and maintenance due to operator error. Flows with high levels of particulate matter may induce error by coating the transducers and limiting transducer life. Typical maintenance for thermal flow meters consists of periodic cleaning of the transducers. In severe applications additional particulate removal equipment or frequent meter change outs may 4 http://kurzinstruments.com/K-BAR1.pdf

http://kurzinstruments.com/K-BAR1.pdf



2-12

be required to keep the meter functional. When thermal mass meters are removed from the flare header for cleaning or being moved to new locations (temporary installations) care must be taken when removing and reinserting to avoid damaging the transducers. Additionally, annual calibrations of the probe may be required in order to ensure accurate measurement. One drawback of thermal meters is the inability to measure bi-directional flow and compensate for backflow. Unlike most other flow meters, thermal mass meters are not unidirectional, meaning that they sense flow in either direction as a positive signal. While this would be of little concern when detecting flows at the terminal end of a flare system, it may be a confounding factor when using a thermal mass meter to detect the source of waste gas flows in a complex flare piping system. Applicability to Flares Both constant-power and constant-temperature thermal meters have been used to measure petrochemical industry waste gas flows to flares. John Zink engineers use thermal mass meters exclusively in measuring flare header waste gas flows for designing flare gas recovery compressors. Several petrochemical facilities use thermal mass meters to isolate potential sources of flows to flare headers. One petrochemical facility has over 50 thermal mass meters installed for this purpose. The authors’ experience is that thermal mass meters are more reliable than pitots since they aren’t as prone to the mechanical integrity problems mentioned previously concerning pitot tubes. Figure 7 shows a comparison of flare waste gas measurements between single point pitot tubes and thermal mass meters. Note that both the high-range and low-range pitot tubes only provide measurements over a narrow range at each end of the measured flows while the thermal mass meter covers the entire range. Also, note that the flows in Figure 7 are in typical flare waste gas ranges all of which both the thermal mass meter and the pitot tube vendors claim they can handle accurately. Using the pitot tubes only, one could conclude that either the low or high flow pitot is not giving accurate readings and base the measurements on either a very high or a very low flow compared to the oscillating flow rates as measured by the thermal mass meter and the tracer injections. This can be particularly misleading for flare systems that are subject to surge or oscillating flows, as was the case for this example.



2-13

Flare J Flow Meter Comparison

0

10000

20000

30000

40000

50000

60000

70000

80000

90000

100000

110000

120000

130000

140000

150000

160000

170000

180000

190000

200000

0.00

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.10

Run Time, Days

Was

te G

as F

low

Rat

e, S

CFH

Thermal Mass MeterPrandtl# Correction

Radioactive KryptonTracer Time Of FlightInjectionsLow-Range DP FlowMeter

High Range DP FlowMeter

Figure 7. Comparison of Various Flow Measurement Techniques 2.2.3 Ultrasonic Time-of-Flight Meters

Principle of Operation There are two types of ultrasonic flow meters, Doppler shift and transit time. Flow meters that use Doppler shift technology rely on the principle that wavelengths are perceived as shorter as the source approaches and longer as the source retreats. In fluid applications, Doppler shift flow meters are anchored or clamped onto the pipe and house a transmitter and a receiver. The transmitter emits signals that are reflected back to the receiver by discontinuities in the flowing fluid such as air bubbles, particles, or other turbulent phenomena. The receiver detects the change in wavelength of the signal reflected from these discontinuities in the flowing fluid, rather than the fluid itself. For this reason Doppler shift only works on fluids with appreciable levels of discontinuities. The underlying equation governing flow meters using Doppler shift is:



2-14

Equation 3:

( )Kff V 1o −= Where: V = Velocity of the fluid fo = transmitter frequency f1 = receiver frequency K = constant of the transmitter and system

Figure 8. Generic Example of Ultrasonic Flow Meter5

5 http://www.gepower.com/dhtml/panametrics/en_us/products/ultrasonic_gas_flow meters/gf868_flare_gas_flow meter.jsp

http://www.gepower.com/dhtml/panametrics/en_us/products/ultrasonic_gas_flowmeters/gf868_flare_gas_flowmeter.jsp

http://www.gepower.com/dhtml/panametrics/en_us/products/ultrasonic_gas_flowmeters/gf868_flare_gas_flowmeter.jsp



2-15

Flow meters that operate on transit time measurement operate on the principle that ultrasonic signals travel faster with the flowing fluid and slower when traveling against the flowing fluid. Typically, one transducer is mounted upstream while another is mounted downstream at a fixed separation. The transducers act as transmitters and receivers alternatively, and emit and receive sound waves simultaneously. Waves that travel with the fluid flow exhibit a faster velocity than waves that travel against the flow, giving a differential time. In the absence of flow, sound waves travel from both transducers to each other in the same amount of time, therefore the differential is zero. The basic equation used to calculate the velocity of the fluid is37: Equation 4:

⋅−

=1221

12212

2 TTTT

XLV 1

Where: V = Velocity of the flow L = direct distance between transducers X = lateral distance between transducers T21 = travel time from downstream transducer to upstream transducer T12 = travel time from upstream transducer to downstream transducer An additional advantage of ultrasonic time of flight meters is that the differential times may also be use to solve for the speed of sound in the flowing gas. For an ideal gas the speed of sound, c, can be related to the molecular weight of the gas by the following,39 Equation 5:

MWRTc /γ=

Where: c = speed of sound in gas R =universal gas constant T = absolute temperature γ = specific heat ratio MW= molecular weight of gas



2-16

Thus, in addition to the velocity of the gas, ultrasonic time of flight meters can be used to estimate the molecular weight of the gas. This allows one to estimate the mass flow rate without doing any other composition measurements. Design Variations Aside from the design variations involved with Doppler shift and transit time, ultrasonic flow meters come in portable clamp-on models and installed tap-in models. While clamp-on Doppler shift meters have been used successfully to measure liquid flows, where substantial flow disturbances or particulate matter may be present, the authors are aware of no clamp-on Doppler shift meters used in gas service. The current generation of portable, clamp-on, transit time flow meter requires pressures approaching 80 psig or more when measuring gaseous flow. Thus, portable clamp-on transit time meters are not applicable to measuring waste gas flows in near atmospheric pressure flare headers. For petrochemical industry flare applications, permanently installed transit time devices are used. Available literature indicates that well over 2000 time of flight ultrasonic flow meters have been install in flare applications worldwide.76 The remainder of this section will focus primarily on permanently installed transit time models. Transit time flow meters are capable of providing continuous flow measurements and there are many industry applications. Ultrasonic flow meters are capable of handling very large ranges of flows, as is prevalent in flare systems. One supplier gives a typical turndown ratio of 2750:1, over a flow range of 0.1 to 275 ft/s.5 As with most flow meters, the accuracy declines at very low and very high flow rates. For ultrasonic time-of-flight meters, the high level of sound at high flow rates (above 275 feet per second) makes it impossible to obtain a reading forcing their inaccuracy to approach infinity. At very low flow rates (<0.1 feet per second); the very low level of sound makes the sound very difficult to detect. Panametrics quotes the accuracy of the Digital Flow GF868 Mass Flow meter at + 2-5% at a velocity range of 0.1-275 ft/s and measuring gas of the same composition as that used to calibrate the meter.5 However, actual accuracies depend on transducer design, electronics installed, flow rate, the actual composition of the waste gas, and the profile of the actual flow. Accuracies may range from 5-10% or higher with additional error incorporated for additional pieces of equipment and operating conditions (see Section C Factors influencing measurement performance). The quoted precision for the instrument is given as 1 percent.5 Transit time



2-17

models are claimed by the vendors to be the most accurate of all the flow meters currently used to measure flare or waste gas flows. Manufacturers of transit time meters suggest installing the meter in a straight run of pipe with upstream to down stream ratios of straight pipe of 20 pipe diameters upstream and 10 pipe diameters downstream to ensure a smooth velocity profile and accurate performance. This may be an issue in flares that do not contain the necessary piping. As mentioned for pitot tubes, straighten vanes or other flow correction devices may be used in some applications to correct for not having enough straight pipe run, but due to safety concerns they may not be utilized in many applications. Ultrasonic flow meters do not obstruct the flare lines and are not indirect safety hazards. Most meters consist of transducers that are slim in design and do not inhibit the flow of material. Due to the unobtrusive design, ultrasonic flow meters also offer negligible pressure drop (<0.1 psig) across the device. This, as stated in previous sections, is an important consideration due to the low operating pressures of most flares. Ultrasonic time-of-flight meters can also handle corrosive or hazardous flows with additional material expense incurred. Typically, all the parts of the meter exposed to the waste gas are made of stainless steel or other specialty materials making them appropriate for most petrochemical industry applications. For very harsh applications, it is possible to design the parts of the meter that come in contact with the waste gas using more expensive specialty materials. Ultrasonic transit time meters are available for application in piping from approximately 3” up to 120",5 which make them applicable to the large pipe sizes frequently found in flare systems. These meters may be used across temperature ranges of -110 to 150° C and pressures up to 1500 psig. Ultrasonic meters can accurately measure bi-directional flow due to the design of the meter. One transducer acts as an accelerated sound wave, while the second acts as a retarded wave. In the event of bi-directional flow the transducers simply measure this as a negative flow. The instrument can calculate the actual positive flow by adding the negative number from one transducer to the positive number from the other transducer. Additionally, ultrasonic meters can operate in flows that have a fairly high level of particulate matter. Past designs often had induced error associated with flows with high levels of particulate matter, however current models have been specifically designed to overcome these limitations. Changes to the design to reduce waste gas particulate limitations include orienting the two transducers at a 45 degree



2-18

angle, decreasing the diameter of the transducer probes, and reducing the distance the transducer probes intrude into the flare header line. Additionally, process changes driven by regulatory requirements and economics reduced the particulate levels in flare waste gas considerably from those originally encountered. Ultrasonic meters are typically permanently installed into the flare line and require a unit shutdown for installation. Some models can be hot-tapped into the process and are portable. However, some petrochemical facilities do not allow hot tapping of flare lines under any circumstances while many others only allow hot tapping under very specific process conditions due to safety concerns around the flare. In these cases, it may be necessary to take the flare out of service to install the meters. For the most part ultrasonic transit time meters require little maintenance in order to ensure accurate performance well within the stated TCEQ data quality objectives. Panametrics states in the brochure for the GF868 mass flow meter that the unit "does not require regular maintenance", "has no moving parts to clog or wear", and " is constructed of titanium or other metals that withstand the corrosive environment usually found in flare gas applications". 5 Annual calibrations of the meter may be necessary to ensure accurate performance. Applicability to Flares Ultrasonic transit time meters have successfully measured flare flows in refineries and in flare gas applications. Since ultrasonic time of flight meters do not have to be corrected for varying gas composition, and in fact can be used to estimate the molecular weight from the observed speed of sound, they are in many instances the preferred flow meter for flare applications. One drawback to the time transit flow meter is the high purchase and installation cost. Typical costs for an ultrasonic time-of-flight meter are in the $20,000-30,000 range. The author’s experience is that design, site preparation, installation, field calibration, and connecting the meter to the distributed control system typically make the overall up-front cost for an ultrasonic time-of-flight meter approach $100,000.

2.2.4 Combination Systems

Because of the potential large dynamic range of velocities that can be encountered in flare systems, there has been some interest in flow measurement systems that use a combination of



2-19

flow measurement technologies to cover different ranges of the flow spectrum (multiple range averaging pitots, time of flight meters used in conjunction with high flow pitots, etc.). While there are several vendors whose product literature mentions such combination systems, the authors were not able to identify any such systems that had been installed in flare systems to date. In theory extending the range of velocities that can be measured should be possible by combining flow measurement technologies to cover different ranges. The relative strength and weaknesses of the individual technologies is largely unchanged by combining them into an integrated flow measurement system. However, there are some inherent problems with matching the flows measured in the region where their measurement capability overlap and it remains to be demonstrated in a full scale application. In most cases the advantage of such a combined system may be to extend the range to cover large design case release scenarios that may seldom or never be seen. It is unclear to the authors whether this seldom-used extended range will warrant the additional complication of a combined flow measurement system.

2.3 Factors Influencing Measurement Performance

2.3.1 Measurement Uncertainties

The types of flow meters acceptable for flare measurement service have some differences in the standard accuracy and precision responses. The data for these metrics are shown in Table 2, along with other comparative data that will be discussed later. One vendor of ultrasonic flare gas meters quotes velocity measurement accuracy within + 1.4 to + 3.5% 5. However, using the vendor-supplied parameters in the International Organization for Standardization (ISO) TR 5168 root sum square model 6 results in a higher measurement error. This model, given in Equation 6, can be used to estimate flow measurement inaccuracy with a 95% confidence interval based on the principal sources of uncertainty in the measurement,



2-20

Equation 6

Overall Error = (Do)2 (Po)2 (To)2 (So)2 (Vo)2 ++++ Where: Vo = Gas velocity uncertainty (Per vendor + 1.4 to 3.5 % for a 2 path flow meter and velocities between 1 and 275 feet/second) So = Uncertainty in speed of sound for gas being measured (+ 1.0% per vendor) To = Gas temperature uncertainty (+ 2% per Rule 115 requirements) Po = Gas pressure uncertainty (+ 5mm Hg per Rule 115 requirements is ~0.66%) Do = Pipe inside diameter uncertainty (+ 0.5% per Shell Flow Meter Engineering Handbook 14) Insertion of the above values into the equation results in an overall uncertainty of + 2.75% to + 4.25% for ultrasonic flow meters, depending on the gas velocity uncertainty. For thermal mass meters, vendors quote 0.25% repeatability, + 0.5% of temperature reading for velocities above 100 standard feet per minute (SFPM, velocity of air at standard temperature and pressure conditions) and with optional velocity/temperature/mapping an overall accuracy of + [3% of the reading +(20 SFPM +0.25 SFPM/°C], above or below 25°C. Using the same values above for errors in temperature, pressure, and diameter together with the minimum + 3% accuracy quoted for thermal mass meters, and neglecting the speed of sound term (does not enter in the uncertainty of thermal mass meters), results in a minimum overall error of + 3.8% for thermal mass meters. As mentioned earlier one vendor of pitot tubes claims overall accuracy of + 1%. Using the same values above for errors in temperature, pressure, and diameter together with the quoted + 1% accuracy and neglecting the speed of sound term (does not enter into the uncertainty of pitot tubes) results in a minimum overall error of + 2.6%. These uncertainties only represent the inherent uncertainty of the input variables measured to derive the flow measurement. For all the meters there is also an uncertainty associated with the composition of the gas being measured. For ultrasonic meters the uncertainty of the gas composition can be estimated directly from the input variables measured by the flow meter by using equation 6 and the speed of sound derived from the flow meter. However, for pitot tubes and thermal mass meters an independent gas composition measurement must be made, in order



2-21

to correct these meters for varying gas composition. Correcting pitot tubes and thermal mass meters for varying gas composition will be addressed in more detail in the next section. (Additional uncertainties from flare destruction efficiency measurements would add to this, but estimates of this uncertainty are beyond the scope of this report.) The Measurement Data Quality Objectives in the TCEQ Work Assignment 5 Work Order state that the overall desired accuracy, precision, and sensitivity is + 20 percent. Specifically for flow rate, composition measurements, and the resulting calculated mass rate to the flare the desired accuracy, precision, and sensitivity should be + 20% (never to exceed + 5,000 pounds per hour of total VOC emitted from the flare). More experience is needed using these devices to ensure that this criteria can be met under all conditions. Currently, there is only limited operating data that can be used to evaluate compliance with these data quality objectives. While the + 20% would seem doable for a well engineered system within the limitations of the individual meter, this may not be achievable in all cases for the following reasons:

• No single analyzer is capable of achieving the turndown ratios required to measure all possible waste gas flow rates. This may require the installation of multiple monitors to achieve this goal.

• Both ultrasonic time-of-flight meters and thermal mass meters have maximum flow rate limitations significantly below the design maximum flare waste gas flow rates.

• Using widely accepted root sum square error models and vendor-supplied parameters, calculation of overall measurement errors range between + 2.6 to + 4.25 percent.

• Pitots are not capable of accurately measuring flows below 10-15 feet per second. Figure 3 shows that typical flare waste gas flow rates are consistently in this range.

To illustrate the first and second points, ultrasonic flow meter vendors quote maximum velocity measurement capacities of 150-275 feet per second (fps). Thermal mass meter vendors quote maximum velocity measurement capacities of 200-300 fps. Flares can have waste gas flow rates ranging from almost zero up to design maximums of over 1000 fps.



2-22

For ethylene gas in a 36” diameter flare header at TCEQ Standard conditions, 300 fps equates to a flow rate of ~7,500,000 standard cubic feet per hour (SCFH) or ~555,000 pounds per hour. Large flares can be designed for flows as high as 2,000,000 pounds per hour during major emergencies. Thus, for maximum flows during emergency conditions, a thermal mass meter or ultrasonic time-of-flight meter will only measure ~28% of the potential waste gas flow rate for a large flare. However, many major emergency flows can be reasonably approximated based on process knowledge. For example, most flares are designed for a worst-case, maximum flow assuming a total power failure requiring complete shutdown of the facility. For example, by determining which parts of the process unit have been isolated, and de-inventoried the quantity and composition of the material sent to the flare can be estimated during these rare design case events. While instantaneous numbers may not have the desired degree of accuracy, event totals should be reasonably approximated.

2.3.2 Calibration Systems

Initial and Scheduled Calibration Flow measurement quality assurance depends on the instrument and the desired service. Service considerations such as the characteristics of the waste gas stream and materials compatibility are essential considerations. For all the preferred flow measurement devices, thermal mass flow meters, pitot tubes, ultrasonic time of flight analyzers, accuracy is directly affected by flow profiles, gas stream temperature, gas stream pressure, and gas stream composition. Although ultrasonic time of flight analyzers are less affected by gas stream composition than pitots or thermal mass meters. In order to compensate for the effects of service, it is important to select the appropriate calibration methods and frequencies. The portable thermal mass flow meters and pitot tubes used frequently for flare application generally are calibrated in methane or air in a wind tunnel or controlled chamber. A twelve to sixteen point calibration curve is generated and calibrations are certified and traceable to U.S. National Institute of Standards and Technology (NIST) standards. For quality assurance purposes, thermal mass meters and pitot tubes are sent out for an annual recalibration certification and a log of certificates and maintenance work is maintained to assure long-term accuracy and reliable performance relative to manufacturers specifications. Temporarily positioning two thermal mass flow meters in series is an option considered as a QA check. This can be especially useful if the certified factory calibrations are offset in time to



2-23

determine relative drift and bias. Reproducibility under equivalent conditions is 0.5 to 1.0 % relative standard deviation (RSD). Ultrasonic time of flight analyzers being less dependent on stream composition and thus not factory calibrated to a given composition normally do not require being returned to the factory for calibration checks. One additional test that can be performed to check both ultrasonic and thermal mass flow meters is the zero flow test5. The zero test can be performed by removing the sensors and covering them to isolate the sensor from any air currents. A zero flow reading should be obtained. Compositional Impacts on Calibration

Particulates For long duration measurements in flares that contain significant particulate matter, the probe should be periodically pulled and checked for deposits on the sensor elements. These deposits should be removed at a frequency that ensures minimal particulate buildup. By positioning the probe downstream of a wet gas knock out vessel, most deposition problems can be eliminated or minimized. Thermal Mass Meters and Pitot Tubes The vendor reported accuracies in flow measurement for thermal mass meters and pitot tubes are for situations where the calibration gas has the same composition as the gas being measured. Due to the wide potential variation in flare gas composition (from mostly ethylene or propylene to mixtures containing varying amounts of C1-C5s, hydrogen, nitrogen, hydrogen sulfide, and possibly other inert gases), correction of the observed flow meter reading to account for the actual gas composition versus that which the meter was calibrated on will be required to meet the data quality objectives of this study. Figure 9 shows the impact of gas composition and gas velocity on the correction factors for a thermal mass meter. As shown, the correction factors for converting measurements of a waste gas significantly different from the calibration gas are nonlinear with respect to observed velocity. For example, Figure 9 shows that at a 1000 fpm velocity reading the correction factor for air is ~1.5 while the correction factor for hydrogen is ~2.0. Thus, at 1000 fpm for a meter calibrated on air the actual conversion factor for hydrogen is then 2.0/1.5 = 1.33. Figure 9 shows that at 5000 fpm the correction factor for air is ~2.0 while the correction factor for hydrogen is ~2.7. Thus, at 5000 fpm for a meter calibrated on air, the conversion factor for hydrogen is 2.7/2.0 =1.35. Thus at 1000 fpm using the direct reading for a meter calibrated on air as the hydrogen flow rate will give a 33% error and at 5000 fpm using the direct reading for a meter calibrated on



2-24

air will give a 35% error, a 2% increase in overall error. Note that as the flow rate increases the difference between hydrogen and air also increases. Thermal mass meter vendors typically recommend using simple density and heat capacity ratios to estimate the correction factor for varying gas compositions. Since this method also does not vary with velocity, the inaccuracy associated with varying gas compositions also increases as the flow rate increases. In recent discussions with technical experts from a thermal mass meter supplier, the supplier has stated that thermal mass meters are not recommended for flare waste gas flow applications where highly variable, high volume % (> 5 volume %) concentrations of hydrogen may be present.30 Although the vendor’s technical experts now state this, the author’s experience has been that most local representatives are unaware of this recommendation. In the past, some vendor representatives told the authors that hydrogen, in particular, was a problem for thermal mass meters, and that hydrogen measurements would be off by a factor of 5 to 20. The value of this factor varied depending upon the individual representative. The dual sensor pair thermal mass meter mentioned earlier in this report is an attempt to resolve the potential measurement inaccuracies associated with variable gas composition. Extensive laboratory studies using wind tunnels and flow chambers have been conducted measuring pure gases as well as various gas mixtures over a wide range of flow rates. The measurements have been conducted at velocities approaching zero up to 275-300 feet per second and the entire range in between. In essence, curves such as Figure 9 have been developed for an extensive array of gas compositions and these data used to develop exponential equations to describe the functional relationship. In recently developed thermal mass meters these laboratory-developed equations have been programmed into a minicomputer that is an integral part of the flow meter. Since every compound has a distinctive thermal signature, the vendor claims that the dual sensor pair thermal mass meter is capable of correcting the measured flow rate for the actual gas composition in real-time. Having two pairs of sensors, one in the main waste gas flow stream and one near the walls, allows the meter to obtain a flowing stream waste gas velocity and a near zero waste gas velocity. One pair of sensors could conceivably give the same velocity reading for two or more widely varying gas mixtures. According to the vendor29, no two gas compositions will have identical velocity readings at both the in stream velocity and at a near zero velocity. Thus, the meter is able to identify the actual composition of the waste gas by



2-25

comparing the two readings and correct the velocity measurement for the actual composition in real time. Thermal mass meters with two pairs of sensors have only recently been developed. The authors have no experience with these systems. According to the vendor several of these dual sensor instruments have recently (within the last year) been installed in flare headers.29 If the gas composition varies from the extensive compositions in the on-board computer, the meter will still give a reading that may be suspect. The vendor of these meters (as well as the vendors of most other flow meters) asks for a list of the possible waste constituents and the potential range of constituents up front to ensure the accuracy of the measurements. Good process knowledge or field data will be required to design and implement thermal mass flow meters capable of making real time corrections for variable gas compositions. While these meters have the potential to be more accurate than traditional single sensor pair thermal mass meters, the user pays for this improved accuracy. A single sensor pair thermal mass meter costs on the order of $5,000 while the dual sensor pair thermal mass meters cost between $15,000-$20,000.



2-26

0

0.5

1

1.5

2

2.5

3

3.5

40

1000

2000

3000

4000

5000

6000

7000

8000

9000

1000

0

1100

0

1200

0

1300

0

1400

0

1500

0

1600

0

1700

0

1800

0

1900

0

2000

0

Linear Velocity, SFPM

Velo

city

Cor

ectio

n Fa

ctor

H2

HE

CH4

70%CH430%CO2N2

Air

CO2

Argon (Chlorine)

Figure 9. Example Gas Flow Meter Calibration Curve6

6 Data Courtesy of Kurz Instruments, Irvine, California



2-27

At the suggestion of a thermal mass meter supplier, the authors have started using the ratio of the Prandtl numbers for the gases being measured to the Prandtl number of the gas the meter was calibrated on to correct observed flow meter readings (see Figure 7). Using gas chromatography data to obtain composition data and to compute mole weighted average physical properties, the author’s have observed Prandtl numbers ranging from 0.13 to 1.3 on the same flare during a 1-week timeframe. Therefore, depending upon the application, uncorrected thermal mass meter readings could be off by a factor of 0.13 to 1.3. However, even using a gas chromatograph to collect composition measurements is subject to inaccuracy simply due to time lag necessary to get the GC result (GC cycle times of 6, 12, or 15-minutes are commonly required to obtain one composition result). The Prandtl number is a dimensionless number defined by the following equation: Equation 7

( )/kc N pPR µ=

Where:

NPR = Prandtl Number cp = specific heat of the gas or gas mixture at constant pressure (BTU/lb F) F = absolute viscosity lb/ft-second k = thermal conductivity BTU/ft-hr F

However, this method also calculates a constant correction factor independent of flow rate contrary to the data in Figure 9. Whether the waste gas is in plug flow, pulsed flow, turbulent flow, or laminar flow also impacts the correction factor for gases different than the gas the analyzer is calibrated on. The authors believe that getting accurate correction factors to correct for variable gas composition for pitot tubes and thermal mass meters currently requires laboratory or in-field calibrations. More information on calibration of gas flow meters is presented in this section.



2-28

The above information on calibration and composition correction is specific to pitot tubes and thermal mass meters. Ultrasonic Time of Flight Meters GE Panametrics, one of the primary suppliers of ultrasonic time-of-flight flow meters, states the following in their instrument literature; “The GF868 is the only meter on the market that uses Panametrics’ patented method for calculating the average molecular weight of hydrocarbon mixtures. This proprietary algorithm extends the range for measuring average molecular weight, while improving accuracy and compensating for non-hydrocarbon gases better process control problems, and accurate plant balance.” 5 The vendor’s literature states that molecular weight accuracy is + 1.8% for molecular weights ranging from 2 to 120 pounds per pound mole and should be optimized for other gas compositions. 5 For mass flow of hydrocarbon mixtures the vendor’s literature states that typical accuracies are + 3 to 7% for a 1-path system and + 2.4 to 5% for a 2-path system. 5 The vendor’s literature further states that the quoted accuracies are dependent on the accuracy of the temperature and pressure measurements needed for the proprietary algorithm.5 Given that the algorithm is proprietary, it is not possible to objectively and independently assess ultrasonic time-of-flight meters overall accuracy when considering actual waste gas composition’s impact on overall accuracy. Multi Service Flares For flares with multiple process units and multiple process streams tied into them, the accuracy of the calculated flow rate declines significantly for situations where the actual waste gas composition is significantly different than the gas composition the meter was originally calibrated on. Field calibration of these meters is difficult and it is not feasible to calibrate these meters for all the waste gas compositions they may see. Initial and periodic in-field calibrations of these flow measurement meters using tracer injection techniques can improve the overall accuracy of the results. Only ultrasonic flow meters have been demonstrated to correct for varying gas composition in real time, although recent developments in thermal mass meters may soon allow this capability for these meters. Tracers for Field Calibration/Verification In addition to laboratory calibrations traceable to certified standards, field measurements using tracer injection techniques can be used to field calibrate ultrasonic time-of-flight meters and



2-29

thermal mass meters. Calibration here means comparison of the flow meter output to an independent reference flow measurement. Verification means observation of the flow meters operating characteristics and independent calculation of flow. Metered Injected Gas Tracers A chemical tracer gas may be used to calibrate /verify the flow rate. Helium is the preferred tracer gas for flare gas measurements. Because it is an inert gas, helium will not react chemically with any other waste gas constituent. In addition, it is non-toxic, mixes well, is readily available, and relatively inexpensive. Argon and sulfur hexafluoride are not as well suited principally because they do not act consistently as ideal gases when mixed into light gas streams, in particular hydrogen, and along unobstructed lateral sections of pipe. Cost and ease of analysis in the field are also considerations. The amount of Helium metered and injected will largely depend on total flare gas flow and lower detection limit of the downstream measurement instrument. An example of a typical test is a metered injection of one to ten liters of helium per minute through a rotameter. The rotameter is calibrated with the helium through a wet or dry test meter and pressure corrected with an in line pressure gauge. The injection point is selected upstream at a point that promotes mixing. Usually a bank or six-pack of full size cylinder is incorporated with a regulator set for a delivery pressure as low as reasonably possible to maintain a steady flow. Downstream gas is drawn into Tedlar bags and analyzed or a sample loop and on line micro GC may be used. Helium samples should be drawn at a frequency that permits plotting a concentration gradient tracked over a period of several hours. Helium lower detection limit (LDL) on a Micro GC using an argon carrier gas is below 100 parts per million by volume (ppmv). Calculation of flow is by simple ratio:

(Metered Helium rate injected * 100%He) / [He] downstream = flare gas rate. Radioactive Tracers Radioactive tracer flow rate measurement involves injecting a gas radioisotope into the waste gas stream. Testing involves monitoring the “time-of-flight” of the isotopes between two or more detectors placed a known distance apart. The results of the measurement are most accurate in “plug flow” applications. The measured velocity is used to calculate mass and volumetric flow rates, and to characterize different flow regimes (laminar or plug flow). Tracer flow rate measurements are commonly used to check existing flow measurement devices and to measure flow distribution between main headers and branch headers.



2-30

The radioactive tracers used have very short half-lives. Tests involving radioactive tracers are complex, and therefore require detailed planning and coordination. This measurement is contracted to licensed and certified specialty companies. Prior to using this technique, comprehensive safety reviews and permission from the facility radiation safety officer are required. 7 Using radioactive Krypton 85 (a gamma emitter) and three external receivers equally spaced apart starting at 10 meters downstream of the injection point, measurements time the flow as a time-of-flight across the scintillation detectors (gamma ray collectors). By factoring the time-of-flight with a specific gravity and a cross-sectional area of the pipe at the receivers, a mass flow is calculated. By interpretation of the data (ratio of multiple areas), the degree of pulsed flow and back flow may also be determined. 8 Figure 7 shows a specific example where a radioactive tracer was used to verify measurements from both pitot tubes and thermal mass meters. Each of the red squares in Figure 7 represents one radioactive tracer injection. Note that each injection yields a single point measurement, thus several measurements are required to accurately determine the flow rate. In the case of Flare J, the radioactive tracer injection confirmed that the flow was in pulse flow as measured by the thermal mass meter rather than the plug flow indicated by either of the single-point pitot tubes. Other field validation measurements Another test is to take a sample of flare gas at the same time the ultrasonic meter is recorded for its values of mass weight and/or sound speed. The calculated MW value from the chromatographic analysis is compared with the readings of the ultrasonic meter. Manufacturer specifications relating to this comparison verify proper spacing of the sensors, programming, and proper ultrasonic operation. An oscilloscope can be used to monitor actual transit times and difference in transit times from test points on the ultrasonic meter. Calculation of the flow rate is then done by hand and compared to calculation of the rate by the meter. 9 Online measurements to correct for composition changes Given the potential variation in stream composition, online instruments may be added to improve overall measurement accuracy. The cost of using such equipment must be balanced against the



2-31

gain in accuracy actually achieved. The two most common additions are densitometers and direct online composition measurement devices. The latter is discussed in detail in the next section. Densitometers may be used as a means to collect real time data on the gas composition to correct thermal mass meters, or as an independent verification of the molecular weight determined by time-of-flight ultrasonic meters. Insertion probe gas specific gravity transducers are designed to give an output proportional to the molecular weight and consequently specific gravity of the gas. The most common type of densitometer is a resonating element, which references a fixed volume of reference gas at a constant temperature with the measurement gas. A diaphragm ensures that the pressure on the reference gas and measured gas are at the same pressure. Under conditions of equalized temperature and pressure and with super compressibility effects considered, the specific gravity and relative density are equivalent. Response time may be an applicability consideration when there are resonating pressure effects with a frequency faster than the diaphragm can respond to. Absolute reference pressure range must be able to track flare header pressure range, which can be below 14.7 psia. If a mass flow is desired without regard to composition, a densitometer may be used in conjunction with a thermal mass flow meter to determine a mass flow to the flare. 10,11 While additional instruments to measure the waste gas density can improve the accuracy of the measured flow rate, the importance of these measurements being paired in time with the velocity measurements cannot be overemphasized. For, example the most accurate method of measuring the actual waste gas composition, and thus the physical properties of the waste gas may be online gas chromatographs. However, these traditionally have 6,12, or even15 minute cycle times, while the flow meters measure the velocity every minute. Figure 10 graphically illustrates the importance of pairing the flow and compositional measurements in time. Note that the hourly average flow measurements in Figure 10 tends to smooth out the short-term peak flows and would indicate that the flow is between 2000 and 3000 pounds per hour, while the actual peak flows vary from 1500 to as high as 7000 pounds per hour. By looking at the wide range of potential compositions shown in Figure 4 and the instantaneous vs. hourly averaged flow rates shown in Figure 10, it is apparent that using GC data collected on 6, 12, or 15 minute cycles to correct velocities measured at 1-minute intervals has the potential to miss concentrations shifts, which would then lead to the application of inadequate correction factors to adjust for these composition changes, resulting in potential flow rate measurement uncertainties that dwarf other



2-32

potential measurement errors. More discussion of this issue is presented in the following Sections.



2-33

0

1

2

3

4

5

6

7

8

0.00

0.50

1.00

1.50

2.00

2.50

Run Time, Days

Was

te G

as F

low

Rat

e, 1

000

Poun

ds p

er H

our

Flare A 1-hour Average DataFlare A 1-minute Average Data

Figure 10. Effect of Averaging Time on Measured Flows

2.4 Summary Overall Results of Flow Meter Evaluation

Of all the flow measurement technologies evaluated, only pitot tubes, thermal mass meters, and ultrasonic time of flight meters satisfied all of the primary criteria for application to measurement of waste gases in flare systems. Pitot tubes are applicable across a much narrower range of flow velocities than thermal mass meters or ultrasonic meters, but this can be mitigated to some degree by having multiple meters to cover the full range of interest. While thermal mass meters and ultrasonic time of flight meters can quantify a considerably wider range of flow velocities, in flare applications they may not be able to sense the velocities that present during the extremely rare maximum design case event. While sensing the flow during a maximum design case event



2-34

may be possible by using these meters in tandem with a high flow pitot tube or similar meter, the authors are not certain the added complexity and cost are warranted given the extremely rare occurrence of such an event, and the presence of other process data that would allow the quantity of material being flared to be estimated. As shown in Figure 3, flare waste gas velocities are infrequently seen above 40 feet per second. Additionally, infrequent upset events for the flares analyzed are in the 60-90 feet per second range, well within the optimal range for ultrasonic time of flight analyzers, thermal mass meters, and pitot tubes. The inherent error of all three applicable measurement technologies in a standard gas service are similar (+ 3-7%), however, the actual observed uncertainty of pitot tubes and thermal mass meters in flare service will be greater than that of ultrasonic time of flight meters, as the former technologies depend upon accurately assessing and correcting the observed flow for variation in waste gas composition. Ultrasonic meters are less sensitive to changes in gas composition, and by operating principle can assess changes in gas composition from the primary signals processed by the meter. Real time correction of pitot tubes and thermal mass meters for changes in gas composition, while possible in concept, have to only a limited degree been field demonstrated in real world applications on flare system. This is an area where additional development of thermal mass meters and pitot tubes would be warranted to demonstrate similar precision and accuracy to ultrasonic time of flight meters in waste gas service. While ultrasonic time of flight meters currently are the most accurate meter for use in waste gas service, they are also the most costly to install, being an order of magnitude more expensive than annubars and thermal mass meters. Currently, ultrasonic time of flight meters show the most promise of meeting the stated data quality objectives of this study in waste gas service. However, their cost probably limits their application to flow measurements at the terminal end of flare systems. Thermal mass meters offer a suitable alternative in services where the waste gas composition does not vary greatly, and as a sturdy portable instrument that can be used to troubleshoot flare systems (e.g. identify stray flows in complex flare headers).



3-1

3.0 FLARE WASTE GAS COMPOSITION MEASUREMENT

3.1 Fundamentals

To meet the data quality criteria laid out in the objectives of this study for the measurement of waste gas mass flows, it is obvious that reasonably accurate and precise waste gas flow composition measurements are required. This section of the report lays out basic considerations for obtaining representative samples and performance criteria necessary to achieve the data quality objectives, discusses the basics of operation for analyzers of interest in waste gas systems, presents the key performance criteria for both offline and online analytical methods and techniques deemed to be appropriate for flare service, summarizes the current state-of-the art of available methods/techniques, briefly discusses emerging techniques, and suggests areas where technical improvements could have a significant impact on the cost effective measurement of waste gas compositions in flare systems.

3.2 Analytical Performance Criteria Applicable to Flare Systems

3.2.1 Representative Sampling

The most accurate and precise analytical measurement is of absolutely no value in meeting the stated data quality objectives, if it is not done on a representative sample. Given the potential for significant variability in flare gas composition and the fact that many flare headers are elevated, making accessibility more difficult, at grade flow loops are frequently advisable. Since flare systems are operated at minimal pressure (1-5 psig), sample pumps or ejectors are usually required. A detailed discussion of sample systems is beyond the scope of this report. However, the interested reader is referred to Appendix B, which discusses in more detail key technical considerations in the design and operation of sample systems for the delivery of representative samples.

3.2.2 Constituents of Interest

One of the key data quality objectives of this study is to achieve 90% speciation of individual VOC, with special interest in being able to characterize highly photochemical reactive VOCs, including but not limited to: ethylene, propylene, all butenes (butylenes), all pentenes, 1,3-butadiene, isoprene, toluene, xylenes, ethyltoluenes, trimethylbenzene, formaldehyde, and



3-2

acetaldehyde. As a practical consideration in many flare systems, in order to assess whether this 90% VOC speciation requirement has be met will require also analyzing for or knowledge of fixed gases (hydrogen, methane, carbon dioxide, carbon monoxide, oxygen, and nitrogen), as well as light alkanes (C2-C7). Analysis of compounds such as hydrogen sulfide, mercaptans, ammonia and light amines, while perhaps not needed to meet the stated data quality objectives, may be of interest as they may have an impact on safe, representative sample handling, conditioning, and disposal. This list is not comprehensive. Constituents of interest required to meet the 90% speciation of VOC in chemical plant flare applications could be quite varied and may have to be assessed on an application-by-application basis. For the purpose of this report, analytes have been grouped by classes for assessment of applicable analytical techniques as follows: hydrocarbons (alkenes, alkanes, and aromatics), fixed gases (hydrogen, methane, carbon monoxide, carbon dioxide, nitrogen, and oxygen), aldehydes, and others. The analysis of aldehydes presents some unique challenges not encountered with the other compounds specifically identified in the data quality objectives. Formaldehyde and acetaldehyde are not expected to be present in reduced waste gas streams from refineries or olefins plants. They may be present in trace quantities in waste gas streams from a few chemical process units that employ oxidation such as the production of phenol, acetone, methyl ethyl ketone, ethylene oxide, propylene oxide, and urea-formaldehyde resins, to name a few. Because aldehydes are not expected to be encountered in significant quantities in most flare gas applications, the special analytical techniques and considerations required for this class of compounds have been placed in Appendix C, and will not be discussed explicitly in the body of this report.

3.2.3 Dynamic Range

Analytical techniques used to analyze waste gas steams in flare system must possess a wide dynamic concentration range due to the variable nature of waste gas sources. Generally, these analytical techniques must be able to detect analytes at concentrations from 100 ppmv to 100% in waste gas streams (105 dynamic range). The lower concentration limit is based on both achieving the data quality objectives and practical lower detection limit for most thermal conductivity detectors. Use of online dilution equipment or multiple sample loops (e.g. vary sample size introduced to analyzer on the order of 10x to 50x) can be used with some analyzers to effectively extend the dynamic range of the instrumentation. Parallel systems or programmed logic loops may be required to make these alternatives effective options for continuously



3-3

speciating waste gas streams. The more complex the system needed to achieve the required dynamic concentration range, the more likely there will be additional expense to install and maintain the system, as well as the potential for more system reliability issues. 3.2.4 Separation

Because speciation is part of the data quality objectives, the ability to cleanly separate the constituents of interest is an important performance criterion for analytical systems used in flare service. Co-elution and poor separation of constituents can be a significant problem in flare systems due to the large number of possible components and the dynamic range of those components. In order to meet cycle time requirements and simultaneously achieve adequate component separation, hardware options must be critiqued not only for their ability to achieve the desired separation in a given service, but also for their adaptability for changing conditions and sources. While not explicitly addressed in this study, software (e.g. advanced GC column temperature programming, assessment of peak retention time shifting, etc.) can also be used to enhance the separation of constituents of interest. While component separation is clearly a performance criterion that must be evaluated for analytical systems applied to the measurement of components in flare systems, there is no absolute measure for this performance criteria. Rather it is a relative scale related to flexibility and adaptability to field applications.

3.2.5 Quantification

Lower detection limits as defined by signal to noise ratio, linearity of response, electronic gain, detector overload, carryover or cross contamination, detector drift, bias, and loss of detector sensitivity are all quantification performance criteria that need to be considered. Column selection, contaminants, retention time shifting, peak distribution, integration width, and sensitivity parameters all effect quantification and consequently impact performance specifications. The quantification standards set forth in EPA Performance Specification 8 for continuous VOC analyzers,24 Performance Specification 9 for continuous speciated analyzers,25and the performance specification in EPA method 1818 for offline or grab analyses were used to evaluate the level of quantification achievable by each of the analytical techniques assessed for use in flare systems.



3-4

3.3 Basics of Analytical Instrumentation Applicable to Flare Systems

3.3.1 Gas-Solid Chromatography

Gas chromatography is the most common method for analysis of industrial waste gas streams. It is essentially a physical method of separation where the components to be separated are distributed between two phases, a stationary phase contain within the column and a mobile fluid which travels through the stationary bed. The chromatographic separation of the gas components occurs as a result of the varying affinity of each component for the stationary phase selected. Due to the difference in retention time of the various components on the stationary phase, the sample is separated into discrete peaks of the constituents of interest. The mobile phase then conveys these peaks to a detector for quantification. A gas chromatograph consists of essentially three integrated functions, sample introduction (injector), separation (column), and detection. The detector signal is sent to a recording instrument where identification is determined by retention time or elution time, and quantitative values are obtained by signal strength over time (peak height or area counts). The relationship of these three functions is depicted in Figure 11. Many types of columns are available to achieve the desired separation based on composition of the waste gas stream. Often, multiple columns are required for complex separations and temperature control is often used to enhance separations. Following separation in the column, individual components emerge from the column and are conveyed to the detector by the carrier gas. There are many detectors available, however, the most common associated with waste gas streams are thermal conductivity detectors (TCD), flame ionization detectors (FID), and photoionization detectors (PID). The TCD measures the difference in thermal conductivity of the eluting component with the pure carrier gas. It is considered a universal detector since it measures all components relative to the carrier gas. The FID combusts the sample components in hydrogen flame, producing ions that are collected and converted into a signal. The FID measures only ionizable components. The PID measures ionization radiation based on the photoionization potential of the components. Excitation energy is supplied by an ultraviolet lamp. The PID measures only components that have an ionizing activation energy equivalent to or below that supplied by the lamp employed. It has a good response to many compounds containing double bonds. 2 While the above detectors are the most broadly applicable to the class of compounds normally found in flare systems, there are a variety



3-5

of other chemical or class specific detectors (helium ionization, thermal ionization, flame photometric, nitrogen phosphorous, electron capture, catalytic combustion, etc.) that may be used in conjunction with gas chromatography in particular flare applications. The interested reader is referred to Appendix D for a more detailed description of these detectors.

Figure 11. Typical Gas Chromatographic Schematic

3.3.2 Non-Dispersive Infrared (NDIR)

Non-dispersive infrared spectroscopy is based upon the principle that individual compounds and classes of compounds adsorb infrared light at characteristic wavelengths. The attenuation of the IR source at this characteristic wavelength within the sample cell versus the attenuation in the



3-6

reference cell can be related to the concentration of the compound or class of compounds in the sample cell. In NDIR spectroscopy the source emits light at all IR wavelength, and thus is a non-dispersed source. An optical filter is used to focus the IR light to the sample to the desired wavelength. The detector then senses the attenuation of the characteristic wavelength in the sample cell versus the reference cell. The infrared detector that is combined with the filter is sensitive only to the infrared rays that have wavelengths within a narrow range depending on the wavelength and filter selected. Historically, NDIR techniques have been used for the measurement of CO, CO2, NOx NO2, and SO2. Within the past several years several portable NDIR instruments for the analysis of hydrocarbons have been marketed. Several companies now offer a custom configurable multi-channel instrument for the detection of hydrocarbons. There are commercially available NDIR systems available for the analysis of methane, propane, C1-C6, CO, CO2, NOx, NO2, and SO2. Since these analyzers can not speciate, individual compounds they can not meet the data quality objectives of this study, but may be useful in troubleshooting flare systems or collecting preliminary data prior to the design of online GC systems.

Figure 12. Non-Dispersive Infrared Gas Analyzer 3.3.3 Photoacoustic Sensor

When a gas is irradiated with IR light it absorbs incident radiation within its own characteristic absorption spectrum. The amount of absorbed radiation, which follows the Beer-Lambert absorption law, is a function of the gas concentration, the path length and the specific absorption coefficient of the gas. This absorbed radiation, which for a very short period of time is stored as intramolecular vibrational-rotational energy, is quickly released by relaxation to translational energy. Translational energy is equivalent with heat and, when the absorption chamber is sealed,



3-7

this will causes the pressure to rise. Each gas has a unique IR spectrum, and strong absorption takes place only at certain wavelengths. When the incident light is modulated at a given frequency, a periodic pressure change is generated in the absorption chamber. This photoacoustic pressure signal can be measured with a sensitive pressure sensor, usually a microphone. In a conventional photoacoustic sensor, the gas to be analyzed is pumped into an absorption chamber that is sealed with mechanical valves during the measurement. The cell is irradiated with modulated IR light filtered at the wavelengths at which the gases of interest absorb strongly. 13,14 Photoacoustic instruments are an emerging technology. Instruments that can speciate specific organic compounds are just starting to enter the marketplace. They are mentioned here for completeness, but at this juncture they are not mature enough to be considered as being able to meet the data quality objectives of this study.

Figure 13. The Principal Components of a PAS Detector



3-8

3.4 Offline Techniques for Measurement of Flare Gas Compositions

This section reviews methods and analytical techniques applicable for the offline analysis of waste gas samples from flare systems. Each method or technique is reviewed based upon its ability to analyze the constituents of interest, ability to analyze over the dynamic range encountered in flare systems, flexibility and adaptability to deal with separation issues, and against applicable quantification performance standards.

3.4.1 EPA Method 18

Constituent of Interest Since EPA Method 18 is not a unique analytical method, but rather a collection of techniques designed to meet specific performance test objectives, this method has the flexibility to analyze for all of the constituents of interest in the data quality objectives of this study. However, depending upon the specific flare waste gas application, multiple analytical configurations and/or instruments may be needed to speciate all compounds of interest. Dynamic Range For EPA Method 18 the low end of the dynamic range is usually determined by the sampling system. An adsorbent trap can extend the lower detection limit to below 1 ppmv for most VOCs. Direct interfacing or bag samples can typically achieve 1-ppmv lower detection levels. GC detector saturation or column overloading governs the upper detection limit. The upper range can be extended by dilution of sample with an inert gas or by using smaller volume gas sampling loops. The upper limit can also be negatively impacted by condensation of high molecular weight compounds. For VOC above the molecular weight of benzene, one can extend the upper detection limit range by heating the entire sample line and injection system to a temperature where condensation does not occur for the analytes of interest. Heating of grab bag samples may be done in a temperature-controlled environment to vaporize condensates simultaneous with sample introduction. The following are generally accepted dynamic ranges for detectors commonly used with EPA Method 18: TCD, 103; FID, 107; PID, 104-5. With miniaturization technology and three gain settings, the dynamic range using a micro GC TCD detector can be extended to 106. As previously stated, for waste gas applications the dynamic range encountered can be from below 1



3-9

ppmv to 100% for any of the VOCs listed in the data quality objectives. Since adsorbent traps are generally not applicable in waste gas service, due to the high potential for overloading of traps and poor retention of lighter molecular weight components, the dynamic range is then a function of the detector employed and any sample dilutions or variable sample loops employed. Separation Separation using EPA Method 18 criteria are based on analysis development, selection of GC parameters including temperature programming, and column choice. Preliminary GC adjustment is done using the standards and selected column(s) to perform initial tests to determine appropriate GC conditions that provide good resolution and minimum analysis time for the compounds of interest. Based upon the data quality objectives set forth in this study, separation of the VOCs of interest can be performed on a 30 meter non-polar methyl silicone type PLOT (Porous Layer Open Tubular) columns with a resolution down to baseline at a signal to noise ratio of 3. Quantification EPA Method 18 identifies compounds based upon peaks in the samples eluting at the same retention time as known standards. Any unidentified peaks that have areas larger than 5% of the total area are identified using GC/MS, or by matching the peak with additional known standards, with confirmation by further GC analysis. Secondly EPA Method 18 relies on a three-point calibration of each target analyte to generate a response factor. If this response factor deviates from linear by more than 5% over the expected concentration range, then a calibration curve is required. Method 18 establishes precision and accuracy criteria for field audit samples. Precision of triplicate analyses of calibration standards must fall within 5% of their mean value. The accuracy of the prepared audit samples must fall within 10% of their prepared values. Recovery is checked to meet a + 30% recovery of the prepared values for the sampling techniques employed, typically direct online sample loops or bags for waste gas steams. Based upon the recovery tests, a recovery correction is applied for each sampling system used. Summary Statement Method 18 is a compilation of many gas chromatographic methods used for the measurement of gaseous organic compounds emitted from industrial sources. This method is most applicable to low level organic emissions from vents, scrubbers, or post combustion emission points. The method is also suited for emissions from individual sources that have relatively steady operation



3-10

and subsequently steady flow and composition relative to what is typically encountered in waste gas streams in flares. While the adaptability and flexibility provided in the method is sufficient to say that Method 18 can analyze for the compounds of interest, across the dynamic range encountered in flare systems, with an acceptable level of precision and accuracy to achieve the data objectives of this study, this can only be achieved by tailoring the specific method to the specific application. For many flare applications this may require multiple methods and analyzers to achieve the desired result. Method 18 is not broadly applicable to waste gas streams due to the component and compositional variability that can be encountered during monitoring of uncontrolled (per-combustion) waste gas streams. Offline sampling methods, as detailed in EPA Method 18, with subsequent analysis are not well suited for waste gas measurements because of cycle time constraints to achieve adequate characterization of the waste gas stream. The nature of flaring events is such that episodic events can occur over the order of minutes with sampling and subsequent analysis estimated to require on the order of hours. Due to the range of constituents and concentrations that can be encountered in flare systems, achieving the data quality objectives may require multiple analyses on multiple instruments of the same sample. This potential time lag is even more problematic if the composition data is to also be used to correct the flow meter for composition changes from that on which the meter was calibrated.

3.4.2 ASTM D-1945 and D-1946

Constituent of Interests ASTM D1945 (Standard Test Method for Analysis of Natural Gas by Gas Chromatography) is a standard method used for natural gas applications. ASTM D1946 (Standard Practice for Analysis of Reformed Gas by Gas Chromatography) is essentially similar but with the addition of carbon monoxide and has the ability to quantify light olefinic compounds. These methods are not applicable to waste gas streams containing significant quantities of aldehydes, ketones, or alcohols. These methods are not specifically configured for detection and resolution of olefinic components or higher molecular weight components, including but not limited to all pentanes, trimethylbenzenes, xylenes, and ethyltoluenes.



3-11

Dynamic Range This method is useful to characterize simple natural gas streams relative to possible waste gas compositions and ranges. Natural gas and reformer gas have by comparison a relatively narrow range of compositions relative to many waste gas streams. The TCD used in these methods offers adequate linearity of response for hydrocarbons over a 103 range, however this range may not be large enough for many waste gas applications, which can have a dynamic range of 105. The sensitivity of the traditional TCD used in this method is questionable for applications where the minimum detection requirements are at or below 100 ppmv and linearity is desired over a large dynamic range. Separation The permanent gases hydrogen, methane, oxygen, nitrogen, carbon dioxide and carbon monoxide are measured using a molecular sieve (5A or 13X column) with thermal conductivity detection. Simultaneously, light hydrocarbons methane through hexane are measured using a Porplot Q column with a thermal conductivity detector and back flush of all higher mass components is performed to give a C6

+ value. This separation does not give adequate resolution of C4 and higher mass components as required by the data quality objectives of this study. These methods are unable to adequately separate aldehydes, ketones, or alcohols. Quantification In the ASTM 1945 Method, reference standards are prepared from pure components. Precision is checked by confirmation of two consecutive checks of reference standards that agree within 1% for each component. Linearity of response is checked by injection of a series of standards of increasing mass by way of partial pressure increments through a fixed loop. Linearity is checked with major components. Repeatability and reproducibility are set on sliding scales found in the comparison table. Relative accuracy is checked by normalization of the molar values obtained in the analyses and comparison with the original values. The sum of the original values should not differ from 100% by more than 1 percent. Summary ASTM methods D1945 and D1946 do not have the ability to analyze all the constituents of interest, nor do they have the dynamic range required for many waste gas applications. While



3-12

these methods may be applicable to flare applications with relatively constant compositions and constituents, they generally are time intensive (estimate >10x factor), less accurate, and do not offer the degree of speciation and resolution other methods offer. Also, the method does not address determination of hydrogen sulfide. The significance and use of ASTM D 1945 as stated in section 4.1 of the method is that it provides data for calculating physical properties of the sample, such as heating value and relative density, and for monitoring the concentrations of one or more of the components in a mixture.

3.4.3 GPA 2261 and 2177

Constituent of Interest Gas Processors Association Method 2261 is a natural gas method used primarily to determine the calorific value of fuel gas. It measures C1–C6

+ hydrocarbons, inert gases oxygen, nitrogen, carbon dioxide, and hydrogen sulfide. It typically uses a two-column system with one 10-port valve or three valves to allow analysis from both columns with back flushing to get a C6

+ value and utilizes a single TCD detector. These methods are not applicable to aldehydes, ketones, or alcohols. Similar to the ASTM methods, these methods are not specifically configured for detection and resolution of olefinic components or high molecular weight components, including but not limited to pentenes, trimethylbenzenes, xylenes, and ethyltoluenes. Separation of aldehydes, ketones, and alcohols are not achieved with this method. While these methods may be applicable to some waste gas streams, they do no have the flexibility to analyze for all the constituents of interest for this study. Dynamic Range The dynamic range for GPA Methods 2261 and 2177 covers the typical range encountered in natural gases. These molar volume ranges are defined as 0.01-100% methane, 0.01-100% ethane, 0.01-100% propane, 0.01-10% N-butane, 0.01-10% isobutane, 0.01-2% isopentane, 0.01-2% N-pentane, 0.01-2% hexanes and heavier, 0.01-10% helium, 0.01-20% oxygen, 0.01-100% nitrogen, hydrogen sulfide 3.0-100%, and 0.01-20% carbon dioxide. The dynamic range encountered in waste gas streams is much wider than the range demonstrated by these methods. GPA 2261 uses varying partial pressures to determine calibration and linearity. With the wider dynamic range encountered in waste gas streams, errors can be introduced due to unknown degrees of compressibility with different compositions.



3-13

Separation Separation of natural gas components in GPA 2261 is achieved with an adsorption molecular sieve column and one or two partition columns (Porpack Q and/or silicone). This serial configuration separates out non-condensable gases, including CO and natural gas components, with a reverse flow to group hexane and heavier components. For waste gas streams, separation of alkanes from alkenes may not be adequately resolved with this configuration. Resolution and determination of hydrogen sulfide is included. Quantification In the GPA methods, calibration is achieved with response factors derived from reference standards of known composition or from duplicate injections of pure components that agree within + 0.5% of the average areas. Precision data is specific for each component in terms of both reproducibility and repeatability and specifications are presented in the table below. Linearity is checked by sequentially increasing the partial pressure of a pure component standard gas of the major component of interest. Any linearity deviations indicate the sample loop is too large. For waste gas streams, this linearity method has the potential to increase the lower minimum detection level, as it may result in a sample loop size too small to detect some low concentration VOCs of interest. Additionally, hydrogen linearity and subsequently quantification is unacceptable using helium as a carrier without modifications and adding an alternative carrier gas. Multiple injections and use of special supplementary procedures with additional columns and valve switching would be required to meet the data quality objectives of this study on many water gas applications. Such serial type column configurations are not well suited for quantification of waste gas streams, as large concentration variations could lead to column overloading, resolution, and long cycle times relative to parallel configurations. Summary The GPA methods cover the determination of natural gas and similar mixtures within a limited range. The GPA methods do not offer the versatility and resolution required to meet the data quality objectives of this study for all waste gas stream applications. These methods may be applicable in a few waste gas applications with limited constituents and relatively narrow composition ranges.



3-14

3.4.4 Multichannel Micro GCs.

Transportable, multi-channel, micro GCs are well suited for waste gas flare measurements because of the adaptability and dynamic range allowed for an extended range of fixed gases and hydrocarbons, as well as many other components of interest in waste gas streams. These micro GCs typically contain four channels. Each channel contains a micro-machined injector, a high resolution GC capillary column, and a micro-thermal conductivity detector, thereby making each channel essentially a self-contained GC. As a result, simultaneous analyses of samples on up to four different columns using four different sets of operating conditions can be used. Digitally controlled pneumatics and an extended detector range enable analysis of gas streams containing multiple components over a wide range of concentrations with a good degree of precision and accuracy.15,16 Agilents’ 3000 Micro GC and Varian’s CP-4900 are examples of transportable, multi-channel, micro gas chromatographs. Constituents of Interest The micro-machined thermal detectors allow for near universal detection of gases, which in many cases makes obtaining a compositional material balance possible. Four channels are used to analyze complex samples quickly and easily. Pre-configured models are available for refinery gas analyses that are directly applicable with minimal modification to waste gas applications. Table 3 illustrates the constituents commonly analyzed on each channel. Of the constituents of interest listed in the data quality objectives, only formaldehyde and acetaldehyde cannot be measured with these instruments.



3-15

Table 3. Typical Micro GC Configuration

Suggested Micro GC Instrument Settings for Initial Testing are: Channel A Channel B Channel C Channel D Column 12 M molecular sieve 5A Porplot U 8M alumina 10M OV-1 !0M 2um Column Temp C 100 105 100 125 Run Time (sec) 160 160 160 160 Sample Time (sec) 40.0 40.0 40.0 40.0 Inject Time (sec) 30 30 30 30 Detector filament on on on on Detector auto zero on on on on Detector sensitivity low medium medium high Peak Attribute area area area area Calibration fit (3 minimum)

point to point point to point point to point point to point

Calibration Type external std external std external std external std Column head pressure 20-30psig 20-30psig 20-30psig 20-30psig Components hydrogen carbon dioxide ethylene n-pentane helium ethane propylene 1-pentene oxygen acetylene n-butane collective C5 s

nitrogen propane isobutene benzene methane hydrogen sulfide 1-butene toluene carbon monoxide carbonyl sulfide isobutylene ethylbenzene methyl mercaptan cis-2-butene xylenes

(collective), trans-2-butene n-heptane 1,3 butadiene ethyl acetylene Dynamic Range Typical lower detection limits are 10-20 ppmv for most components. A low, medium and high electronic gain may be used as a 10x to 100x step change in signal and increases the dynamic range of the detectors from 103 to 106, and is selective by channel. Dynamic range with sensitivity gain and linearity are superior to traditional GC Thermal Capture Detectors (TCD). Due to this extended detector range, there is no need for dilution interfacing or complex valving configurations, and the risk of detector saturation is low. All components can be measured up to 100% molar concentrations, with the assumption that heated injectors are used and no vapor condenses. 15,16



3-16

Separation Separation can be achieved for all components listed in the data quality objectives, with the exception of formaldehyde and acetaldehyde. Resolution is best for compounds with a carbon number of five and lighter. Software controlled alternate conditions may be required for applications where toluene, xylenes, ethyltoluenes, or trimethylbenzenes are present at the percent level. Quantification A full multipoint calibration using three or more points over the expected range of concentrations can be performed in the laboratory or field. Multi-channel micro GCs are capable of meeting US EPA Performance Specification 9 calibration criteria for continuous monitors with one exception, the linearity criteria. Calibration type is user selected, based on area count, with the options to use point to point or linear regression, without forcing the intercept through zero. Relative reproducibility and drift of all components in standard gases using the micro GC has been superior to traditional methods when compared with GPA 2261 specifications. This can be attributed to digital control of the injectors and detectors; as well as precision silicon based micro machining. Refer to Table 4 for detailed performance specifications, relative accuracy, and precision specifications.



3-17

Table 4. Summary of Offline Waste Gas Composition Measurement Techniques Waste Gas Composition Measurement Technique

Selection

Oven Temperature

Range C Cycle Time Flow

Range cc/m Calibrations Standards

QA/QC Columns Injectors Detectors Performance Specifications

Manual Techniques Portable Micro GCs, Agilent 3000 Micro GC , Varian CP-4900 *

4 independent set from 35-180, isothermal

~ 3 minutes 3 to 12, column specific External reference stds, multiple stds point to point or best linear fit daily checks single point for each module

4 parallel Single pump through with 4 electronic timed injectors, fixed volume injectors, backflushed timed mode or fixed mode injectors, splitless sample, direct interface with flowing gas or collected bags attached manually

(4) TCD thermistor type

3 point linear fit or >3 point point-to-point calibration. Less than 5 % variance from average of 3 injections, 106 dynamic linearity +/- 10%, repeatability is ~1% RSD

EPA method 18 Not specified, generally temperature programmed for appropriate resolution

Not specified, generally 3 analyses /hour possible

30-100 for appropriate resolution

External reference stds for each target analyte, multiple gases and points, recovery tests, multipoint calibration with regression analysis to obtain a least square line,

1 with loop inject Single fixed volume loop, direct interface, bags, dilution interface, single loop volume

TCD, FID, (not specified)

Collection efficiency, Recovery of direct or dilution, recovery for bags and tubes, standards certified to +/- 2% accuracy, calibration curves based on at least three points with dilution as an option, regression analysis and least square line.

ASTM D-1945 Not specified, generally temperature programmed for appropriate resolution

Not specified, generally 3 analyses /hour possible, longer if speciation of C6 or C7 components

Variable, as appropriate for resolution

External pure components or reference stds., relative response factors, linearity, reproducibility and repeatability checks

2 in parallel, 13X molecular sieve and silicone 200/500

Single fixed volume loop Single.TCD Filament type

Known linearity for each component over the range of interest, dry reference standards or blended pure components, repeatability with two consecutive samples within 1%,

ASTM D-1946 Temperature programmed as feasible

Not specified, generally 3 analyses /hour possible



2 columns, an adsorption column for H2, O2, N2, CH4, CO and a partition column for C2H6, CO2, C2H4

Single fixed volume loop Single or Dual TCD Filament type

Known linearity for each component over the range of interest, dry reference standards or blended pure components, repeatability with two consecutive samples within 1%.

GPA 2261and GPA 2177 40 isothermal Not specified, generally 3-4 analyses /hour possible



2 in parallel, silicone and molecular sieve and pre-cut for C6+ group

Single fixed volume loop, pressure dependent

Single.TCD Filament type

Linearity check with methane based on partial pressure and fixed loop,. Response factors or response curves from gas reference standards, , unnormalized value total should not vary more than +/-1% from 100%, relative response factors OK., retention times.



3-18

Table 4. Summary of Offline Waste Gas Composition Measurement Techniques (Continued)

Waste Gas Composition Measurement Technique

Selection

Detection limits/Ranges

(volume) Accuracy / Precision

Expected +/- % Potential

Interferences Sample

Conditioning Data

Management Advantages Limitations Additional Comments

Recommendations and Summary Statement

Manual Techniques Portable Micro GCs, Agilent 3000 Micro GC, Varian CP-4900 *

100 ppm-100% 5% for GC precision of 3 injections, 10% relative accuracy for sample train over dynamic range (dual standards, from same supplier), manufacturer specs <1% RSD precision. RSD daily over 2 months 2 components 3%. Preceeding based on authors testing over 1,600 hours and 20 flares. A * Manufacturer specs <1% RSD repeatability, linear range 106 +/- 10%, .

NH3, amines Membrane, filters Windows EZChrom, stand alone

160 second cycle time, portable, compact

Transferring data time consuming, not for classified areas, weather, sieve column needs periodic regeneration

Versatile, field adjustments easy, laptop required, duration of testing must be aligned with waste gas composition variability

Good for testing to generate design data or data for online analyzer variables, Easily portable, easy calibration and setup, good dynamic range for surveys and testing to acquire exemption or CEMS design data

EPA Method 18 For ppm sources, generally 1 ppmv- 1%vol

+/- 5% for GC precision for triplicate injections, +/- 10% accuracy for prepared audit samples, +/-30% for tube recovery, +/- 30% for bag sampling. Audit analysis +/- 10% of certified concentration. Based on Method 18 criteria

High % components, detector saturation

Externally collected, Tedlar bags, sorbant tubes, pumps, water needs managed

Variable, recorder and integrator, cumbersome

EPA Protocol Limited at % levels, principally used for post combustion vents, many alternatives and multiple variables impact accuracy and precision, long cycle times because of manual sampling

Time consuming, not enough chromatography details, duration of testing must be aligned with waste gas composition variability

Traditional stack source sampling, not recommended for waste gas streams without modifications. Criteria referenced in a variety of regulations, however not easily or practically transcribed for speciated waste gas measurements

ASTM D-1945 100 ppm-100%, except hydrogen sulfide 300 ppmv-30%, typical calibration follows natural gas ranges

1% for all components basis deviation from 100% normalized results, relative response factors acceptable. Sliding repeatability scale: mole % ( 0.1-1) R</= 0.04, (1-5) R</=0.07, (5-10) R</= 0.08, over 10 R</= 0.1. S Basis re-approved method literature 2001.

High water content and natural gas liquids

% a column for interfering hydrocarbons, drier in front of sample valve

Variable, recorder and integrator, cumbersome, manual calculations of physical properties

Good for natural gas samples, acid gas samples, supplemental methods for heavier components

Limited application for complex waste gas streams. Resolution and run time concerns Groups all C5+, C6+, or C7+ components, long cycle times

Supplemental procedures for heavy ends

Needs updating, antiquated, too long cycle times, not recommended because of single detector and configuration. Some learnings but not generally directly applicable for waste gas streams

ASTM D-1946 100 ppm-100%, including carbon monoxide, typical calibration follows reformer gas ranges

Sliding repeatability scale: mole % ( 0-1) R</= 0.05, (1-5) R</=0.1, (5-25) R</= 0.3, over 25 R</= 0.5. Sliding reproducability scale: mole % (0-1) r</=0.1, (1-5) r</=0.2, (5-25) r</=0.5, over 25 r</= 1.0 Basis re-approved method literature 2000,



Suited for reformer gas containing carbon monoxide

Limited application for complex waste gas streams. Resolution and run time concerns

Supplemental procedures for heavy ends


GPA 2261and GPA 2177 100 ppm-100%, typical calibration follows natural gas ranges

Based on published Methods revised 2000. Relative precision over expected molar ranges are: Nitrogen:(1.0-7.7) repeatability 2% reproducability 7%, CO2 (0.14-7.9) repeat 3% reproducability 12%, Methane (71.6-86.4) repeat 0.2% reproducability 0.7%, ethane (4.6-9.7) repeat 1% re


Phosphorous pentoxide or magnesium perchlorate tube dryers


Good for natural gas samples, acid gas samples

Groups all C6+ components, long cycle times

Good relative response factor table




3-19

The authors have used portable, multi-channel micro GCs extensively to do diagnostic studies on flare systems. When using a portable micro GC, it may be restarted (rebooted) daily and a GC verification program is typically run to ensure there are no error conditions in the embedded computer system firmware. Column and detector performance is checked daily and logged by running a standard that contains a particular surrogate component on each channel. If the relative standard deviation varies on any channel by more than 5%, a bake-out conditioning method is run for 1 hour, after which the operating medium is opened and the standard is rerun. If the standard deviation remains above 5 %, a complete calibration is performed and logged. 17,18,19,20

Retention time shifting is checked daily but has not required frequent corrections (e.g. less than monthly based upon continuous use 24 hours a day, 5 days a week), with the exception of the molecular sieve column. Due to the accumulation of water on the molecular sieve, the molecular sieve column contains a feature to update the retention time after each analysis. This feature reduces elution time as a function of water load. The molecular sieve needs to be regenerated at a frequency determined by the water content in the waste gas and resulting loss of peak separation. This typically takes two hours every two days. In the event retention time shifting does occur, flows are rechecked and adjusted as necessary. Formation of ammonium salts is one common cause for retention time shifting. These salts can generally be removed by running frequent short duration runs with hot columns and injectors using humidified air. Summary Transportable multi-channel, micro gas chromatographs are well suited for waste gas flare measurements because of the adaptability and dynamic range available for an extensive list of fixed gases and hydrocarbons, as well as many other components typically found in waste gas streams. The micro-machined thermal detectors allow for near universal detection of gases across a 106 dynamic range that makes a compositional material balance possible on many waste gas streams. Precision and accuracy performance in most cases is sufficient to meet typical performance specifications for continuous analyzers, such as USEPA Performance Specification 9.

3.4.5 Summary of Offline Methods

EPA Method 18, ASTM 1945 and ASTM 1946, as well as GPA Methods 2261 and GPA 2177 all have limited applicability for waste gas measurements. EPA Method 18 requires detailed



3-20

interpretation and selection of various options to be successfully deployed in waste gas service. The ASTM and GPA methods cover a narrow range of concentrations and a limited number of components pertinent to waste gas streams. Consequently, these methods require extensive interpretation and modification to become broadly applicable for waste gas stream analyses. All of the above referenced portable methods have longer cycle times than is desirable for waste gas stream measurements and have the potential to be labor intensive for field deployment. Transportable, multichannel, micro gas chromatographs are well suited for waste gas flare measurements. Due to the extended range of the micro TCD and the universal nature of this detector for gas species, most of the fixed gases and hydrocarbons of interest in waste gas streams may be analyzed. In many cases this allows a fairly complete compositional material balance to be completed. Field deployment and implementation is relatively easy and may be less labor intensive than the other traditional methods reviewed. Table 4 summarizes the performance statistics for each of these methods. While the standard methods can meet the data quality objectives for some applications across narrow concentration ranges, only the multichannel, micro GCs appear to be broadly applicable to most waste gas flare applications. All of these methods would require considerable modification if formaldehyde and acetaldehyde are present to any significant degree.

3.5 Online Techniques for Measurement of Flare Gas Compositions

This section outlines some general considerations that must be taken into account for any permanent online analyzer in flare gas service. Commercially available online analyzer systems are evaluated as to their ability to meet the data quality objective of this study on a continuous basis (per EPA definition, must be able to analyze the flare gas at least once every 15 minutes). Online VOC analyzers were included for completeness even though their inability to speciate VOCs precludes them from being able to achieve the data quality objectives of this study. At the current time the only available online analyzers with the potential to meet the data quality objectives laid out in this study are online GC analyzers. We have summarized the currently available online techniques, along with recommendations where improvements could significantly impact the ability to consistently achieve the data quality objectives set out for this study. In addition, a brief description of several emerging analyzer systems is given. However, at this time insufficient information is currently available on these emerging analyzers to assess their ability to continuously meet the data quality objectives of this study.



3-21

Table 5. Online Non-Speciated Waste Gas Composition Measurement Techniques Evaluation Waste Gas Composition Measurement Technique

Selection

Oven Temperature

Range C Cycle Time Flow

Range cc/m Calibrations Standards

QA/QC Columns Injectors Detectors Performance Specifications

Manual Techniques Continuous VOCs (FID - Generic) Temperature

controlled detector block

Continuous, response time < 2 seconds for 90% full scale

Variable Nominally 0.5 lpm

Serial dilutions of a single VOC or mixture that tracks waste gas stream composition averages. Single point for each range.

None Continuous flow through FID VOCs reported as relative to a methane standard or a mixed gas proportional to a source average as per PS8, Hydrogen not counted, so not reliable for BTU if H2 present. Instruments follow EPA Method 25A Criteria.

Continuous Hydrogen (TCD - Generic) Temperature controlled detector block 60 C

Continuous, response time < 10 seconds for full scale

Variable, 0.5-1.5lpm with reference gas flow also

Pure hydrogen /pure nitrogen composites

None Continuous flow through TCD Potential use for waste gas streams with routinely high hydrogen content >90% molar and noble gases, up to 5% error and zero offset of 2% with 5% residual hydrocarbon interferent. Specific scenario will not meet PS8 criteria.

Continuous VOCs (NDIR Generic) Temperature controlled detector block

Continuous, response time < 2 seconds for 90% full scale

Variable, generally 0.5 lpm or less through drier, 0.6-1.5bar absolute with pressure switch

Dependent on wavelength range of 2-9 um. Propane used for conditions that track EPA method 25B

None Continuous flow through IR dual beam double layer detector / optical coupler

Performance effected by density of gas relative to reference gas density.



3-22

Table 5. Online Non-speciated Waste Gas Composition Measurement Techniques Evaluation (Continued)

Waste Gas Composition Measurement

Technique Selection




Interferences Sample

Conditioning Data Management Advantages Limitations Additional Comments

Recommendations and Summary

Statement

Online techniques (non speciated) Continuous VOCs (FID - Generic)

0.5ppmv to 10% or 50ppmv to 100%

Basis compiled product literature, Linearity of 10 7 as per FID characteristics, Carbon atom counter, drift and accuracy tested required as per PS8. Basis compiled product literature, Beckman, Siemens, California Analytical. (All EPA Method 25A compliant)

None known Filters Windows Ezchrom, ethernet RS485 tie in to distributed control system for recording and record keeping

Reliable, simple, industry standard

Unable to detect hydrogen.

Can likely meet PS8 criteria

Several good simple systems to select from.

Continuous Hydrogen (TCD - Generic)

0.1-10% as hydrogenover 4 ranges

Thermal conductivity measured for binary gas systems only, multi point hydrogen calibration, four ranges all linear, auto ranging, drift < 1% /week of smallest span, repeatability <1% of respective span, linear error <1% of respective span. Basis Siemens product literature

None known Filters, Cooler, pump, RS485 Ethernet TCP/IP May be useful for some hydrocracker or hydrotrater units when hydrogen is >90% of waste gas.

Significant errors with more than binary waste gas or periodic/random venting of VOCs.

Not a measure of VOCs

Specialty use only.

Continuous VOCs (NDIR Generic)

Single / multiple ranges with autoswitching possible for 1 ppmv-100% range as propane

Zero drift < +/- 1% of measurement range weekly, Span drift < +/- 1% of measurement range weekly, Repeatability < 1% of respective measuring range, linear deviation < 0.5% of full-scale value. Basis Siemens Product literature, other vendors near equivalent specifications.

Filter and wavelength dependent

Filters, RH < 90% RS485 Ethernet TCP/IP Low maintenance, only 4 gases max. Can be stacked

Specific limited number of chemicals or groups of chemicals.

PS8 criteria may apply, but may be selection dependent (single components vs. groups of components)

Possible configuration is: C1-C6 on one channel, Methane wavelength on another channel. Less common that FID configuration.



3-23

3.5.1 Total VOC Analyzers

Since TOC analyzers do not speciate individual compounds, they by definition cannot meet the data quality objectives of this study. However, a brief description of available VOC analyzers is given here, as they may be useful in providing load variability and range finding capability for more refined methods. There are many suppliers of continuous VOC analyzers with flame ionization detectors. Specifications for the following commercial analyzers were reviewed: VIG Ratfisch, Signal 300HM, Siemens Fidamat 5E, Beckman 400, Moncon 8800 H, and others. Based on Performance Specification 8 (Performance Specifications for Volatile Organic Compound Continuous Emission Monitoring Systems in Stationary Sources) and specifications literature, it appears that most instruments of this type meet the drift specification of 2.5% of the span value and relative accuracy tests. The analyzers also meet criteria set in EPA 204 (Volatile Organic Compounds Emissions in Captured Streams). Those specifications are: zero drift of less than + 3 % of the span value, calibration drift of less than + 3% of the span value, calibration error of less than + 5% of the calibration gas value, and response time of less than 30 seconds. Factors affecting performance are relative response to various hydrocarbon species as a function of carbon content and structure, as well as the presence and arrangement of other heteroatoms in the molecule. Response factors express the variance from the carbon count and are usually given in terms of a reference species such as propane or methane. In some applications, the composition of the waste gas may differ significantly from that of gas used to calibrate the analyzer. Because of the difference in relative response factors, variation in actual gas composition from that the instrument was calibrated on may lead to greater uncertainty in the observed VOC value.

3.5.2 NDIR Analyzers

As mentioned earlier, non-dispersive infrared detection has until most recently only been used for trace hydrocarbon analysis. These analyzers are generally more expensive and complicated than FID detectors, but may offer reduced maintenance and lower drift than FID continuous VOC instruments. However, this analyzer has an extremely narrow response range for the range of components anticipated in waste gas streams, and consequently is not expected to be useful in measurements on systems containing aromatics, olefins, oxygenated species in addition to saturated hydrocarbons.



3-24

The authors have used an NDIR hydrocarbon analyzer which is as part of a commercially available portable emissions monitoring analyzer (Enerac 3000E). The application was not on a waste gas stream but rather on vents from 2 and 4 cycle internal combustion engines. Drift, zero and span were less than + 3% when tested against calibration gas (0.5% volume propane) in the field and hydrocarbon in instrument air. The detector meets EPA's Reference Method 25B "Determination of Total Gaseous Organic Concentration Using a Non-dispersive Infrared Analyzer" and measures hydrocarbons as propane using the latest state of the art pulsed infrared LED emitter and dual lead selenide detector technology. Methane and ethane are not measured. The manufacturer’s recommendation is for a single point calibration with propane. A second point can be run and linearity checked for EPA compliance as a conditional test method. The manufacturers performance specifications are: concentration range: 0-5% or 0-1% by volume accuracy: +5% of reading or 0.01% volume (whichever is greater); repeatability: + -2% of reading or 0.01% volume; maximum zero drift (per year): + 0.02% volume; response Time: <10 sec; warm up time (@22 deg. C): 60 seconds; operating temperature range: -10 to 40 oC. Another instrument with similar capabilities is the Greenline 8000 from E Instruments Group. 79 Several companies offer a configure-your-own multi-channel instrument with a wide bandwidth of 9�m. This will generally detect all C1-C6 straight chain hydrocarbons, but results may be biased low on streams with significant aromatic content. 78

Pertinent information for online non-speciated VOC analyzers reviewed is summarized in Table 5.



3-25

Table 6. Online Speciated Waste Gas Composition Measurement Techniques Evaluation

Waste Gas Composition

Measurement Technique Selection

Oven Temperature

Range C Cycle Time

Flow Range cc/m

Calibrations Standards QA/QC Columns Injectors Detectors Performance Specifications

Online Techniques Speciated

Siemens Maxum 11 (HRVOC flare base package configuration)*

Isothermal, not specified

Stream dependent, either 7.5 minutes or 15 minutes.

10 to 40, column specific

External reference stds initially, internal use of methane and relative response factors on line, spec 9 performance quarterly

4 in parallel Four 10-port valves with backflush, heated, direct interface

Four TCD thermistor type LDL 0.5% mole H2, O2, N2, CH4, CO, CO2, C2H6, C2H4 all 0.01% mole, all others 0.02% mole, range 0.01-100% mole all components, linearity +/- 2% for hydrogen up to 50%, linearity for other components to +/-2% multipoint calibration likely required.

Process Analytical Applications Inc PAAI - (flare base package configuration) *

Isothermal 140 Generally under 12 minutes

Not specified External reference stds initially, linearity checks on ethene and propene, spec 9 performance quarterly not fully verified at this time, uses 4 separate calibration standards and employs single point calibration

9 columns with three trap columns

Two 6-port valves, two 10-port valves, direct interface, splitless injection

Single TCD Filament type Linearity tested for ethene and propene and meets 0.5% repeatability. Multipoint calibration likely required for PS9 r2 requirements. Calculated BTUs, not tested with all target analytes,

Star Instruments VOC Analyzer

Unknown Under 15minutes based on literature.

Not specified Star claim is not specific but states QA/Validation protocols are EPA QA / R-2 & R-5, 40CFR Chapter 1 part 60, Appendix F, and TCEQ compliant. No supporting tests data.

Single column Single injector, splitless PID with FiD option. PID has one-year life with auto lens cleaning capability.

Basis Star Instrument TCEQ Monitoring Guide (HGA) . Inboard QA and QC requirements maintained automatically through onboard computer control and data archives. Star developed a comprehensive QA plan and Validation Protocols, with defensible reporting on a continuous basis.



3-26

Table 6. Online Speciated Waste Gas Composition Measurement Evaluation (Continued) Waste Gas Composition Measurement Technique

Selection




InterferencesSample

Conditioning Data Management Advantages Limitations Additional Comments Recommendations and Summary Statement

Online Techniques Speciated

Siemens Maxum 11 (HRVOC flare base package configuration) *

100ppmv-100% for permanent gases, 200 ppmv to 100% for Carbon 3 or > #C components, 0.5%vol-50%vol hydrogen

From Siemens Draft Product Bulletin HR-VOC using Maxum ll, overall accuracy 5% for all components except 10% for hydrogen, repeatability +/-1% all components except Hydrogen at 5%.

NH3, amines Filters, pre-columns, and backflushing, non-condensing

Windows Ezchrom, ethernet RS485 tie in to distributed control system for recording and record keeping

Excellent range and applicability, 7.5 minute cycle times

Hydrogen looses linearity above 50% volume

BTU calculation imbedded, FID as an additional option may help with performance testing linearity requirement

Good option for continuous monitoring, dual system has redundant for two sources or backup system. Multiple parallel “appellets” allow for analysis of complex waste gas streams over a wide dynamic range and with a relatively fast cycle time. Considered a versatile system.

Process Analytical Applications Inc PAAI - (flare base package configuration) *

130ppmv - to 100% Analyzer repeated within 0.5% for 3 of 5 runs with ethene. Analyzer repeated within 0.75% 3 of 5 runs for propene. LDL demonstrated at 130 ppmv for ethene and propene. Based on PAAI presentation data.

Unknown Filters, pre-columns, and backflushing

ABB Vista, Emerson-Rousmount, or Yokagawa platform using their preferred software

Single injection and detector configuration with isothermal conditions, 11.5 minute cycle time, EPA Method 301 Field Validation service offered

QA protocols are not well established, not field proven, no internal calibration capabilities in base package.

New instrument, first environmental application, maintenance and retention time shifting due to complexity yet to be determined.

Expandability and versatility questionable because of complexity of valve switching and timing. Uses a single TCD detector and very complex valve and column arrangement.

Star Instruments VOC Analyzer Unknown for flares No information for flares. Unknown Windows CE Computer with touch screen, RS-232 and RS485 for network connectivity.

PID specific and selective for olefins and aromatics at low ppm levels, however not essential for flare waste gas measurements

Unable to detect hydrogen, or fixed gases such as CO2, CO, nitrogen, or oxygen with these detectors. Limited range with PID.

Needs further development for flare applications. All application literature is focused on cooling water towers. Literature states “Identical hardware and software may be used for flare gas analysis, less the water sampling components”.



3-27

3.5.3 Permanent Online Monitoring Gas Chromatographs

Development of an unattended analytical method to achieve 95% on-stream time and meet reporting requirements including cycle times of less than 15 minutes requires these additional considerations: Validation of Components and Ranges – Representative data must be obtained during the pre-design survey to encompass the full range of waste gas stream components and compositions. Representative pre-design survey data may be collected for as little as little as a few hours to as long as several weeks. Statistical methods may be applied to aid in assessing the required duration of pre-design survey testing. Pre-design surveys should also include review of process feeds, process upsets, and flaring event reports. Online Calibration – While perhaps being the closest continuous online analyzer performance guideline for analyzers on flare systems, Performance Specification 9 (Specifications and Test Procedures for Gas Chromatographic Continuous Emission Monitoring Systems In Stationary Sources) was written for emission monitoring applications. This fact leads to several problems when applied to waste gas applications. Due to the potential large number of analytes of interest and the wide range of “expected” concentrations, running calibration standards at the low, medium, and high concentration ranges could be impractical from both a time and feasibility of obtaining the required calibration gases. A more practical approach might be to pick a manageable number of surrogate HRVOCs to calibrate and assess performance against PS 9 for these compounds, and use relative response factors for other compounds. Including gases such as light alkanes, nitrogen, oxygen, hydrogen, carbon monoxide, carbon dioxide, and other significant gaseous components based upon their likelihood of occurrence in units discharging to the flare system being monitored is advantageous to assess whether the 90% VOC speciation target is met. Data Management and Reporting - Significant quantities of data will be generated by these online systems. Depending upon applicable regulatory requirements, significant amounts of data manipulation, retention and reporting may be required and appropriate data management systems can make these data much more usable.



3-28

Bias and Drift – These are addressed in the calibration protocols and linearity requirements and often present no problems with properly functioning flame ionization detectors, however signal degradation should be considered when using a TCD in corrosive waste gas service and when using a PID. The PID lamp signal should be checked on a weekly basis because of lamp fogging and intensity decay.

3.5.4 Permanent Online Monitors: Commercially Available Systems

In this section we review three online GC analyzers currently being marketed for analysis of HRVOC in flare systems. The strengths, weaknesses, and problems cited are deemed to be representative of those that will be present when trying to apply virtually all available GC methods to flare systems. These suppliers sell generic configurations that are evolving and potentially can be custom configured to meet the 90% speciation of VOCs requirement in the data quality objectives of this study. As configured these instruments have limitations, and all rely on front-end engineering design and services to build a system for a specific flare application. Consequently, the generic versions presented here are considered basic platforms and work in progress. Each specific flare application will most likely require additional modifications to meet the present PS9 requirements, or require approval of specific variances. Siemens, a major industrial supplier of GC analyzers, markets a basic packaged system that measures ethylene, propylene, butylenes, and 1,3-butadiene. The cycle time to measure the above referenced HRVOC and other flare gas constituents necessary to calculate the BTU content of the gas is 450 seconds. Therefore each analyzer can analyze up to 2 sample streams and still satisfy the continuous analysis requirement of one analysis every 15 minutes. Sample lag time target should be in the 60-second range. The separation of flare gas components utilizes 4 independent “appelets” configured in parallel. Each appelet consists of a 10-port valve for injection and backflush, a pre-column, a separation column, and a TCD detector. One single stream from the flare feeds the separation system and all components are analyzed simultaneously. One appelet measures hydrogen, oxygen, nitrogen, carbon monoxide, and methane. Water is removed in the pre-column, which effectively eliminates peak shifting associated with separations using molecular sieve columns. Since helium is used as the carrier, there are some range limitations with hydrogen analysis that may or may not be acceptable, basis actual or expected waste gas hydrogen content. Minimum detection



3-29

of hydrogen is 0.5% volume and linearity is 2% up to a maximum of 50% volume hydrogen. A separate additional carrier gas can be custom configured for higher hydrogen levels and linearity response, but it is not part of the basic platform configuration. A second appelet measures carbon dioxide, ethane, acetylene, and ethylene. A third appelet measures propane, propylene, trans- and cis– 2 –butylene, and C5

+. This generic platform does not speciate C5

+ components. While custom configuration to speciate beyond C5 is possible, such customization may have a negative impact on cycle times. The fourth appelet measures isobutylene and 1- butylene. Electronic pressure control is part of a reliable sample delivery system when used in conjunction with a pump, since it is common for pressures to fluctuate below ambient in some flare waste gas streams. The system using TCD detectors offers good reliability and about 2% linear deviation over a range from below 100 ppmv to high percent levels. The manufacturer suggests using only a bottle of pure methane for calibration combined with the use of relative response factors, which is a well-established practice for flame ionization detection. However, it is clear this calibration method does not meet Performance Specification 9 requirements, (Section 10.1 of 40 CFR Part 60 Appendix B) as a calibration for “each target analyte” is not conducted. Full multipoint calibrations must be run monthly using representative gases containing all “target analytes” as stated in PS9 Section 7.1. Based on the manufacturer’s literature, calibration error meets the 10% response variance of calibration gas at each level after 24 hours, and meets the linearity precision of 5% for each triplicate injection at each concentration level. The linear regression specification for each target compound at all three levels must have a r2 value ≥ 0.995. This may not be achievable for all applications; given the linear dynamic range of a TCD detector (103) is not as wide as for an FID detector (107). Combined with the potential for 105 swings in actual concentration for some flare applications, the linear regression curve fit requirement of PS9 may be difficult to meet for many applications. It is possible that equivalency testing using a multipoint calibration and corresponding Method 301 protocols can be performed to satisfy the calibration requirements. Additionally, meeting the linearity requirement for any flaring application where hydrogen can be a major contributor to the waste gas composition at greater than 50% volume will be an issue.



3-30

In order to meet the performance requirements of PS-9, additional hardware may be needed if linearity over the tested range does not meet PS9 and alternate Method 301 validation protocols are not performed or not met. Other means of addressing this potential problem are a FID detector in parallel with the (TCD) using a molecular sieve column or a parallel TCD train are two possible configurations. The additional detector may also require additional data handling and reporting considerations for calculation of rolling HRVOCs hourly average mass emission rates. 2

Process Analytical Applications, Inc., who has principally been a supplier of process analytical instruments, has developed a basic packaged system for HRVOC flare applications. Their system uses one of three gas chromatographs (ABB Vista, Emerson-Rosemount, or Yokagawa) equipped with a single TCD and a complex valving arrangement using two10-port valves, and two 6-port valves, with nine columns and three traps to detect 15 peaks with a cycle time under 12 minutes. The basic configuration for this application detects 15 peaks (nitrogen, oxygen, carbon monoxide, carbon dioxide, hydrogen, methane, ethane, ethylene, propane, propylene, n-butane, isobutane, n-pentane, isopentane, and C6+). Another application uses two TCD detectors in parallel to detect 25 components including all the butylenes and 1,3- butadiene. Hydrogen is handled as a stand-alone component using nitrogen as a carrier gas. Helium is used for all other components as a carrier. The exact valving, trap, column selection, and overall separation configuration is proprietary. The vendor claims a range from 130 ppmv to 100% for the components analyzed. Their application uses an isothermal oven temperature of 140oF. Their linearity testing protocol used four validation gases that were customer specified and an accuracy of + 2% relative standard deviation (rsd) was achieved over the range tested. The basic configuration does not include an automated calibration protocol and complete Performance Specification 9 testing has not been conducted. Performance data as presented used only a single or dual point calibration and low-end bias was 34% for ethylene and 12% for propylene. These performance statistics may be improved significantly with more calibration effort (more points and more repetitions of multiple standards). Further concerns with this system are that the butylenes are not separated, and the C6

+ peaks are lumped and reported together. The vendor offers a service to conduct Method 301 validation as an alternative test method. As with the previous online analyzer discussed, PS9 requirements would not likely be met because of the TCD detector linear range capability, combined with the expected range of compounds encountered in many waste gas applications. With the complexity of the column switching



3-31

employed, retention time shifting of components is also a concern. This is a newly developed system and has yet to be proven under field conditions. 14

Another gas chromatograph instrument from Star Instruments Inc. is marketed as an HRVOC online instrument. While the design and quality assurance approach focuses on cooling water tower applications, the literature provided by Star and conversations with a Star officer indicate that this system is also applicable to waste gas streams. It is essentially a dual platform with a continuous VOC analyzer monitoring VOC levels, and when a predetermined VOC threshold is exceeded, a single GC column and GC switching valve are automatically activated for speciation of the sample. The vendor’s literature states that identical hardware and software may be used for flare gas analysis as is used for cooling water analysis. The principal detector is a photoionizaton detector (PID) with an extended life design and an automated lamp lens cleaner. FID and TCD detectors are available depending on the specific application. No performance data is available with reference to waste gas applications, although TCEQ operating permit compliance program documentation for one installation in the Houston-Galveston ship channel area lists only ethylene, propylene, butylenes, and 1,3 butadiene for that facility. Based on the data quality objectives for this study, this combination of a VOC analyzer and a single column, single switching valve, single detector GC would not meet the objectives without significant modifications, consisting of additional detectors, columns, injectors, and valve switching. A single PID may pass PS9 criteria for the listed olefins however because of poor response to alkanes it would likely not pass range or linearity criteria. Also, the PID detector generally has an upper detection limit that is likely below the range that will be encountered in many waste gas applications without additional dilution equipment or variable sample loop configurations. Pertinent information on the online GC instruments reviewed for use in waste gas service in flare systems has been summarized in Table 6. Emerging Technology Siemens has recently developed a process multichannel micro GC that has six modular configurations, and capabilities as versatile as the portable multichannel micro GC described previously for offline testing. The same micro-machining technology and a combination of serial and parallel TCD micro detector arrays are employed. The chromatograph uses a single valveless injector and contains valving for backflushing of pre-columns and heart-cut valving. Three columns are used for separation, and an array of up to eight TCD detectors is used in a



3-32

smart mode to range samples and speciate the HRVOC of interest. A BTU calculation is integrated into the data analysis. The same EZChrom software is employed as with the portable micro GCs and the instrument is packaged for permanent service in classified areas. Additionally, since the instrument is small (“size of a soccer ball”), the manufacturer claims it can be installed at the sample tap with protection from direct sun and rain. This online GC potentially has the same advantages as those previously cited for the portable micro GCs. Precision, accuracy, resolution, and detection limits are similar. Cycle times are also fast (under 240 seconds) with the standard configuration. Offline calibration is the anticipated mode of calibration with methane used for each TCD micro array to adjust relative response factors after initial multipoint calibration. Interfacing to distributed control systems and data management for reporting and recording may present some challenges, however communications are standard Ethernet or RS485 using TCP/IP or MODBUS RTU protocols. This instrument is a new product and has not been field tested or marketed for HRVOC flare waste gas applications. There are a few process installations in Europe and the first one is now being installed in the USA. It warrants further consideration for flare waste gas applications as more systems are installed in the field. 2

3.5.5 Photoacoustic Spectroscopy

A portable, photoacoustic multi-gas instrument is available that can measure up to five components and water vapor in gas mixtures. The measurement principal is based on a photoacoustic infrared detection method. Most gases that can absorb infrared light are measurable. Five specific optical filters selectively measure the five components. Response is linear. Lower detection limits are molecule dependent and are typically in the 10 to 100-ppbv range. The dynamic range is 105 times the detection limit. Applicability for flare waste gas streams is problematic since the dynamic range, although adjustable, may be too low and there are potential interference considerations with matrix effects in complex mixtures. This instrument is probably more applicable to cooling water tower emissions in conjunction with a permeable membrane than to flare applications. There may be potential applications for this technology for waste gas streams where five or fewer HRVOCs may be measured using a single instrument. Applications associated with flares from a single source on a simple process unit may be good candidates for this technology. Also, if the waste stream contains more than five components of interest, the instruments can potentially be stacked with different target analytes filtered for each instrument.13 This instrument is too early in its development to be realistically



3-33

considered as a candidate to meet the data quality objectives of this study. Pertinent information on emerging online instrumentation is summarized in Table 7.

All the systems considered for permanent installations have inherent limitations with regard to complete speciation as configured in the basic generic HRVOC packaged system. Both the Siemens and the Process Analytical Applications systems appear to have the capability of modification to include additional detectors and column configurations. Since both use conventional TCD detectors, questions remain as to whether they possess the dynamic range sufficient to handle many waste gas applications. Both systems using TCD appear to adequately address separation of analytes, interferences, GC hardware, sampling system, seven-day calibration error tests of <10%, and data acquisition and reporting criteria with the basic configured system. The Star Instrument Inc. online VOC analyzer as presently configured appears will have significant problems consistently meeting the speciation requirements of the data quality objectives of this study.

3.5.6 Summary of Online Methods

The potential number of components present and the large dynamic range of concentrations potentially encountered make continuous online analysis of waste gas composition in flare systems an extremely difficult challenge. While online VOC instrumentation is available, it will have limited use in achieving the data quality objectives of this study, as not speciation can be done for specific HRVOC. Existing GC techniques have the ability to analyze for the majority of the compounds specifically listed in the data quality objectives and other key components of interest needed to obtain a 90% closure on VOC. The notable exceptions to this are formaldehyde and acetaldehyde, which will require significant additional hardware to be able to quantify. However, it is doubtful that aldehydes will be present in most waste gas systems at significant quantities that would compromise achieving the data quality objectives. One can custom design a multi-detector, multi-column gas chromatography analyzer using commercially available systems that can measure many of the components that can be in waste gas streams to flares with the precision and accuracy required to meet the data quality objectives. In fact, there are some vendors developing single-detector, multi-column gas chromatograph-based instruments that may be able to sequentially measure most flare waste gas constituents. However, the use of a single detector complicates the ability to achieve speciation of all possible



3-34

components with adequate resolution and accuracy within the required analysis cycle time. The complexity of the GC system required will be a function of both the number of compounds of interest in each flare application and the dynamic concentration range of these compounds encountered in the specific application. Considerable data collection will be required prior to design of the online analyzer system to be able to meet the data quality objectives for many flare applications. Online micro-machined, multi-channel, micro GCs, similar to those discussed in the offline section, offer great promise of analyzing the greatest number of compounds of interest over the largest dynamic range with the simplest GC configuration. However, these systems have yet to be applied in a permanent online flare application. The only performance specification available for continuous speciating analyzers is Performance Specification 9. This specification was developed for emission monitors with a limited number of target analytes over a limited concentration range. The number of “compounds of interest” and the dynamic range of these compounds in flare systems will make achieving the calibration and linearity requirements of PS9 a formidable challenge for many flare applications. In particular, the requirement to achieve + 2% accuracy from predicted values using calibration gases and re-calibration once a year using NIST traceable primary flow standards with an uncertainty of < 0.25% will be difficult to meet, given the number of compound of interest in flare system. It is not clear to the authors that achieving compliance with PS9 for online GCs is a necessary requirement to meet the data quality objectives of this study. Some suggestions to meet PS9 linearity requirements or to demonstrate equivalency to PS9 linearity requirements include: smart injection systems that incorporate multiple loops of different volumes and online dilution sampling equipment that meets EPA Method 205 verification requirements for gas dilution systems. Multiple detectors specific for different ranges of gases can also be incorporated to narrow the range for each detector to meet linearity requirements. Additionally, another approach might be to generate a multipoint calibration curve for each target analyte, generate a quadratic regression equation and have a requirement that all points fall within an error band relative to that calibration curve. Then EPA Method 301 criteria can be applied to demonstrate equivalency.



3-35

Table 7. Emerging Waste Gas Composition Measurement Techniques


Selection Oven Temperature

Range C Cycle Time Flow Range cc/m

Calibrations Standards QA/QC Columns Injectors Detectors Performance

Specifications

Emerging Technologies Siemens MicroSAM - Micro Single Analytical Module *

50-200 Under 15 minutes based on literature.

Not specified External reference stds, multiple stds point-to-point or best linear, spec 9 performance tests quarterly

2 or 3 columns with a pre-column and backflushing

One injector port, smart switching based on detector array and heart cut valve, and backflush

Four TCD thermistor type, logic loops

Unknown for waste gas applications

Innova 1314 Photoacoustic Multi-gas Monitor *

Isothermal 40 Continuous, response time 75 seconds

100-140 Zero point, humidity -interference, humidity-span, and gas calibration performed quarterly, water vapor compensation, self checks

Five chemically specific optical filters and a compensating water filter

Flow through photoacousticcell, normal 75 second response

Photo Acoustic Unknown for waste gas applications

Table 7. Emerging Waste Gas Composition Measurement Techniques (Continued)


Selection




InterferencesSample

Conditioning Data Management Advantages Limitations Additional Comments

Recommendations and Summary

Statement

Emerging Technologies Siemens MicroSAM - Micro Single Analytical Module *

2ppmv-100% over variable settings

Overall accuracy 5% for all components except 10% for hydrogen, repeatability +/-1% all components except Hydrogen at 5%. Basis Siemens Product Literature.

any liquids membrane, filters, backflushing

Windows Ezchrom, ethernet RS485 tie in to distributed control system for recording and record keeping

Fast cycle times, good range, small and good for installation at sample tap, low relative cost, BTU in 140 seconds

Not marketed or proven for waste gas applications although no apparent obstacles

Few installations in Europe, only one new installed in USA, something to watch

May have potential for low cost simple waste gas applications, observe performance Not proven yet. Only one US installation, concept good, outstanding concerns about use with complex streams and accuracy /precision.

Innova 1314 Photoacoustic Multi-gas Monitor *

Variable basis LDL, nominally~10ppmv- 10% volume

~2% on detector, repeatability 1% measured value, drift 2.5% measured value/3 months. Based on Innova Air Tech Instruments Model 1314 specifications.

Variable density, condensables, strong vibrations at 20Hz

coarse and fine particle filters

Windows 95,98, 200, NT stand alone display, RS232 to ASCII files, or IEEE 488 or CAN interfaces, 7300 application software for sequential monitoring

Fast cycle times, low maintenance, no carrier gases

limited to 5 HRVOCs per instrument, accuracy density dependent, few proven field applications - none on flares, not for classified areas

Selectivity unproven, portable

Possible application on vents or single source flares May work well for specific HR-VOCs, not good for >90% speciation<BTU, or compositional closure



4-1

4.0 FLARE GAS EMISSION CALCULATION ERROR ANALYSIS

4.1 Overall Error in Calculated Emissions

The Measurement Data Quality Objectives for this study were to assess the waste gas flow to flares with an overall accuracy, precision, and sensitivity of + 20 percent. Specifically for flow rate, composition measurements, and the resulting calculated mass rate to the flare the desired accuracy, precision, and sensitivity is + 20% (never to exceed + 5,000 pounds per hour of total VOC emissions). Sources of error in computing this mass flow include the inherent error of the flow meter, the error in the flow measurement associated with measuring a gas different from that used to calibrate the meter, the error introduced by measuring a non-representative velocity or flow distribution (e.g. errors introduced by placement of the meter), and the error associated with determining the composition of the waste gas. The relative contribution of each of these sources of error will be discussed in regards to their impact on meeting the stated data quality objectives. As stated in the flow measurement section, the inherent flow meter error associated with all three applicable flow measurement techniques (pitot tubes, thermal mass meters, and ultrasonic time of flight meters) will typically be in the range of 2-5%, depending upon the type and quality of the meter. The impact of varying gas composition on the accuracy of the flow measurement is a function of the type of meter. As discussed previously, ultrasonic meters can assess gas composition changes in real time and adjust for changing gas compositions, and therefore this source of error for ultrasonic meters is typically less than 1% of the flow measurement.37 Pitot tubes and thermal mass meters have yet to demonstrate that they can be corrected in real time for variable gas compositions. If not corrected for changing compositions, these flow measurements could be off by 50-100% in applications with large variation in gas composition. Based upon limited field data, the authors estimate that correcting thermal mass meters with independently collected, periodic composition data can reduce this source of error down to 5%-10% of the measured flow. The level of error introduced to the measurement by incorrectly sensing the real flow profile (e.g. flow channeling, stratified flow, poor sensor placement, etc.) in the pipe is not easily estimated. However, recent petroleum industry discussions with technical experts from the major flow measurement vendors indicate that this can be a significant source of error that is often



4-2

overlooked.28 With proper straight runs of pipe upstream and downstream of the meter, or with the use of flow conditioners, this source of error can be kept under 1% for custody transfer flow meter applications.37 However, since flow meters added to flare systems will generally be added to existing piping systems, and because safety concerns may limit the use of flow conditioners, this source of error will undoubtedly be higher. For illustration purposes, we assume this source of error to be ~2 percent. Getting a single error estimate for the gas composition to be used in the mass flow calculation is also somewhat difficult. Based upon the analyzer specification given in the flow composition section of this report, the accuracy for any given component should be approximately + 5 percent. The relative error for components at or near their detection limit could be higher, perhaps approaching 30-35 percent. However, it is error associated with the major components that will have the largest impact on the calculated mass flow. The errors for individual components will be somewhat self-correcting, as the data quality objectives requiring a 90% closure on VOC will force a mass closure. Therefore, while the relative error for individual components may be higher, the error associated with the total mass flow should be within + 5 percent. Assuming that these sources of error are independent, one can use the root sum square model presented previously Equation 6 to estimate the sum of these errors on the mass flow of waste gas to the flare. For illustration assume that the inherent flow meter error is + 3%, the impact of variable gas composition on the flow meter is + 2%, that the non-ideal flow profile error is +2%, and the gas composition error is + 5 percent. The resulting root sum error for the mass flow would be + 6.5 percent. In the authors view this represent about the minimum error that can currently be achieved under ideal conditions. However, this is just illustrative of the sum of errors that may be present in flare measurement systems. The rollup of error in non-ideal flow meter installation applications and with highly variable gas compositions could result in considerably larger error estimates. The illustrative calculations above would seem to indicate that the + 20% data quality objective should be achievable with existing technologies. Assessment of the + 5000 lb/hr absolute limit on emissions requires one to estimate a destruction efficiency across the flare. A detailed study of factors affecting flare destruction efficiency is beyond the scope of this work. For our assessment, we have assumed a constant destruction efficiency of 98 percent. A 3-inch diameter



4-3

flare header flowing at 200 fps with waste gas of specific gravity 0.44 would result in 168,621 lb/hr of gas to the flare. A + 20% uncertainty in this flow results in + 33,724 lb/hr uncertainty on the flow to the flare, and at 98% destruction efficiency + 675 lb/hr uncertainty on the emissions from the flare. Even at a design case event that might be at 1,000 fps, the absolute uncertainty in the emissions from the flare should still be under + 5,000 lb/hr. Except for all but the very largest flares achieving the + 5,000 lb/hr absolute uncertainty should be achievable. As stated previously, the issue with meeting the uncertainty at the design case will be the inability to measure the high flow velocity, probably not the ability to meet the + 5,000 lb/hr criteria. Obviously, this conclusion is a strong function of the assumed destruction efficiency. Assuming a lower destruction efficiencies would make obtaining this absolute mass uncertainty on emissions from flares much more challenging.



5-1

5.0 CONTROL OF GAS ASSIST RATIO ON OPERATING FLARES

5.1 Introduction

Several pioneering studies in the 1980s demonstrated that when operated properly, flares should have destruction efficiencies between 98 and 99 percent. Several studies have shown that entrained liquid in the waste gas being flared can significantly reduce flare combustion efficiency to levels approaching 70-80%, thus increasing emissions.21, 22, 23, 24, 25 Liquid knockout pots and water seals prior to the flare stack are designed to minimize the presence of entrained liquids from either the process or condensation of the heavier waste gas components. To maximize flare combustion efficiency as well as waste gas destruction efficiency, proper operation of these liquid collection systems cannot be overemphasized. Besides significantly reducing the flare combustion efficiency, entrained liquids in flared waste gas can also cause rain out of hydrocarbon materials, which can lead to unsafe situations from both a safety and an environmental perspective. 1, 26, 27 Also, with respect to the flared waste gas composition, significant quantities of low-BTU (<300 BTU/SCF) materials can result in reduced combustion efficiency, thereby increasing waste gas emissions. For these situations, auxiliary fuel gas with a substantially higher heating value (~1,000BTU/SCF) is combined with the waste gas to ensure complete combustion. The US EPA New Source Performance Standard addresses the burning of low-BTU waste gas. This regulation (40CFR60.18) requires specified flare tip velocities, monitors to ensure the pilot light is lit at all times, and other aspects of flare operations to ensure minimized air emissions during flaring conditions. Assist gas is supplied to the flare tips of many large, high capacity flares to minimize flare noise, smoking, soot formation, and to further maximize combustion efficiency. Normally, either steam or air is used as the assist gas. Injection of assist gas at the flare tip increases the mixing of waste gas with air, as well as the residence time of the waste gas constituents into the flame zone, thereby increasing combustion efficiency. The choice of steam or air is primarily dictated by economics, such as the availability of excess steam at the facility, the costs associated with producing steam vs. air, etc. Using steam as the assist gas gives some improvement in combustion efficiency versus air, as steam has a lower activation energy for the production of H and OH radical needed in combustion reactions versus that required to produce O radicals from



5-2

air. 1, 26 In fact, one of John Zink’s many technical papers on the basics of flare design and operation1 specifically states, “Steam injection is the most common technique for adding momentum to low pressure gases. Steam itself provides an additional smoke suppression benefit as it interacts in combustion chemistry. Low-pressure air can be utilized in cases where gas pressure is low and steam is not available.” Limited data indicate that steam injection rates that yield a steam to waste gas mass ratio of 0.15 to 0.5 and higher are typically required to achieve smokeless operation and maximum combustion efficiency. 1, 21, 22, 23, 24, 25, 26 Since air does not enter into the flame zone chemical reactions as easily as steam, higher ratios of air to waste gas are typical. Steam or air injection does reduce the smoking nature of flaring events by significantly reducing soot formation. However, assist gas injection may also reduce the overall combustion efficiency by cooling the flame to below optimum temperatures for destruction of some waste gas constituents, and in severe cases may even snuff out the flame, thus significantly reducing combustion efficiency and significantly increasing flare exhaust gas emissions. Figures 14 and 15 graphically illustrate this point. Figure 14 shows the optimum steam to waste gas mass ratios to maximize combustion efficiency to be in the 0.5 to 1.5 range. Slight reductions in combustion efficiency occurred at assist to waste gas ratios of 3-4, and significant reductions in combustion efficiency occurred at assist to waste gas ratios above 5.5. Figure 15 shows similar results with optimum soot destruction at assist gas to waste gas ratios from 0.25 to 1.0. However, Figure 15 shows reductions in combustion efficiency for HC, CO, and total waste gas below assist gas to waste gas ratios of 0.3 and above assist to waste gas ratios of 0.8. There are insufficient data currently available for assist gas to flare waste gas ratios between 1.0 and 5.5 to definitively establish the critical ratio where the combustion efficiency declines sharply. The above discussion indicates that the effect of assist gas to waste gas ratio on flare combustion efficiency, as well as destruction efficiency, requires further investigation. Over aerating or over steaming of the flare flame has the potential to significantly reduce combustion efficiency.

5.2 Examples of Assist Gas Ratio Control Techniques

As noted previously, one of the objectives of this report is “to evaluate current methods of controlling assist gas (steam or air) to waste gas ratios to assure that ratios are maintained within



5-3

a known certainty.” The authors reviewed current practices for controlling assist gas to waste gas ratios during flaring events. There are three commonly used techniques for controlling assist gas to waste gas ratios; manual control, distributed control system (DCS) auto set point control, and infrared sensor feedback control. In the authors’ experience, a comprehensive review of approximately 50 flares in U.S. refineries and chemical plants, informal discussions with representatives of several other petrochemical companies, and some published references 5, 6, 42 indicate the following: Manual Control The first and most widely practiced method of controlling the assist gas injection rate is manual adjustment by the operator based on flame observations (either directly or on a video monitor). Where manual adjustment is used, some facilities have flow meters on the assist gas and the operators are trained to minimize the use of the assist gas for economic reasons. Of the approximately 50 flares in refineries and chemical plants of which the authors have knowledge, roughly 90% use this method. Discussions with representatives of other petrochemical companies indicate this high percentage of manual assist gas control is representative of the industry as a whole, with one or two companies being exceptions, and using a lower percentage of manual vs. automated controls. Distributed Control System (DCS) Auto Set Point Control The next most practiced method of controlling the assist gas injection rate consists of continuous assist gas and waste gas flow meters tied into a distributed process control system with automatic control to a set ratio. Of the approximately 50 flares investigated, roughly 10% use this method. Discussions with representatives of other petrochemical companies indicate this low percentage of DCS auto set point control assist gas controls is representative of the industry as a whole, with one or two companies being exceptions using a higher percentage of automated vs. manual controls. The waste gas flow meters used for these measurements vary from flare to flare and company to company. The assist gas meters used for these measurements also vary from flare to flare and company to company. However, there should be more commonality in steam/air meters since the assist gas is at high pressures (100-600 psig), and is of constant, known composition. The control systems used to adjust the assist gas flow rate based on the waste gas flow rate are a very small part of a much larger process control system that varies from company to company and facility to facility. In addition to controlling the assist gas ratio for the flare(s), these control



5-4

systems also control all the process temperatures, pressures, and flows. A discussion of the design and operation of distributed process control systems is outside the scope of this report. Infrared Sensor Feedback Control The least practiced method of controlling assist gas injection rate consists of infrared sensors that sense the flare flame characteristics and automatically adjust the assist gas flow rate to maintain smokeless operation. Of the flares the authors have knowledge, less than 2% of the flares use this method. Discussions with representatives of other petrochemical companies indicate this very low percentage of flares using infrared sensors to control the assist gas ratio is representative of the petrochemical industry as a whole. Where these systems are used, the sensors are primarily infrared radiation cameras. There are numerous vendors of infrared radiation cameras for measuring flare smoke, including Raytheon®, Mikron®, and many others. As with all infrared systems, the accuracy of these systems is affected by interferences (primarily CO2 and water vapor), emissivity, and luminosity. 5.3 Assist Gas Ratio Technique Performance

With respect to the stated data quality objective of assuring that assist gas to waste gas ratios are maintained within a known certainty, the authors reviewed available data from several flares. Figures 16, 17, and 18 show data from 4 flares each using one of the above three assist gas control techniques. Figure 16 indicates that for the flares in question, manual control appears very similar to an on/off switch. The majority of data in Figure 16, and the median value for all the points in this figure are essentially zero meaning that very little steam is actually flowing to the flare tip. Figure 16 also shows that when the steam valve is opened to the flare, the actual assist gas to waste gas ratio can vary widely from around 2 to above 50. Note that the average deviation from the mean for all the points in Figure 16 is 13. The authors were unable to obtain sufficient data to assess whether this performance is indicative of all manually controlled assist gas systems, or merely an outlier. However, given the previous discussion on the potential impact of assist gas ratio on combustion efficiency it does warrant further investigation. Figure 17 shows that DCS auto set point control, with the steam flow controlled by the actual waste gas flow, has the potential to maintain the assist gas to waste gas ratio with greater certainty. Figure 17 shows data from 2 different flares using DCS auto set point assist gas to



5-5

waste gas ratio control. Figure 17 shows that the median for Flare C is 0.7 with an average deviation from the mean of 1.4 while the median value for Flare E is 0.7 with an average deviation from the mean of 0.9. Figure 17 also shows that the assist gas ratios were consistently maintained around 1 for eight of the nine months, with Flare E maintaining an assist gas ratio around 4.5 for the first month prior to being reset to a lower set point. The outlier points shown in Figure 17 are thought to be due to the time lag between the waste gas flow meter sending a signal to the steam valve and the steam valve opening. Since the points in Figure 17 are 1-hour averages, short-term spikes in the waste gas flow rate accompanied with a time lag until the steam valve opens can result in some hourly averages appearing much higher than the mean. Figure 18 shows that infrared sensor feedback automated assist gas ratio control using a smoke monitor can also result in effectively controlling the assist gas ratio. Figure 18 shows that prior to the associated process unit being down for a process turnaround, no attempt was being made to control the % smoke to a value by adjusting the assist gas. After the turnaround, assist gas adjustments were used to control the observed % smoke. As a result the average deviation from the mean dropped from 8.6 % smoke to 1.8 % smoke, with the smoke consistently being maintained in the 0 to10% range.

5.4 Summary Results of Assist Gas Control Techniques Evaluation

Assist gas is used to improve mixing in the flare flame zone, thereby reducing visible smoke and improving overall combustion efficiency and destruction. Air or steam is traditionally used as the assist gas with the majority of facilities using steam. Earlier 1980’s vintage studies demonstrated that assist gas to waste gas mass ratios between 0.4 and 4 were effective in reducing soot while ratios between 0.2 and 0.6 achieved the highest hydrocarbon destruction efficiency. Manual control consisting of adjusting the steam flow based on visual observations, DCS auto set point automated ratio control consisting of automatically adjusting the assist gas flow based on the waste gas flow, and infrared sensor feedback automated control consisting of automatically adjusting the steam flow based on the [% smoke] measured by a smoke monitor were identified as techniques currently used for controlling the assist gas ratio. Of the 50 flares reviewed for this report, < 2% are using the smoke-based automated control technique, < 10%



5-6

are using the DCS auto set point waste gas flow based automated control technique, and approximately 90% are using manual control based on visual observations. Both the infrared sensor feedback smoke-based automated control and the DCS auto set point waste gas flow-based automated control techniques appear capable of meeting the TCEQ Data Quality Objective of maintaining the assist gas ratio within a known certainty. The results for manually controlled assist gas systems exhibited considerably more uncertainty, but the amount of data available were not sufficient to make strong conclusions on all manually controlled assist gas system. However, the degree of uncertainty that was observed warrants further investigation of manually controlled assist gas systems.

5.5 Developmental Recommendations

The limited data sets examined for this study would indicate that controlling assist gas rates using DCS auto set point control based upon waste gas flows or infrared sensor feedback control based on smoke observations of the flare is significantly more effective than manual control for controlling the assist gas to waste gas ratio within a known certainty. While previous investigators’ data on destruction efficiency versus assist gas ratio obtained under controlled conditions would suggest that poor assist gas control might negatively impact destruction efficiencies, there are little to no data available on the impact of assist gas ratio control on destruction efficiency of operating flares. The effect of assist gas to waste gas ratio on flare combustion efficiency as well as destruction efficiency of operating flare systems requires further investigation, since over aerating or over steaming of the flare flame could significantly impact combustion efficiency.



5-7

60616263646566676869707172737475767778798081828384858687888990919293949596979899

100101102103104105

0 1 2 3 4 5 6 7 8

Steam to Waste Gas Ratio (lb/lb)

Com

bust

ion

Effic

ienc

y (%

)

High Flow Rate Tests

Low Flow Rate Tests

EPA-600/2-83-052, Flare Efficiency Study, M. McDaniel, Engineering-Science, Inc., July, 1983

Figure 14. Effect of Assist Gas on Combustion Efficiency



5-8

98

99

100

0 0.2 0.4 0.6 0.8 1 1.2

Steam to Waste Gas Ratio (lb/lb)

Com

bust

ion

Effic

ienc

y (%

)

Soot

HC

CO

Total

Figure 15. Influence on Steam on Global Efficiency7

7 US EPA/600/2-84-095, Evaluation of the Efficiency of Industrial Flares: Test Results, J. H. Pohl, R. Payne, and J. Lee, Energy and Environmental Research Corporation, May, 1984



5-9

0.0

10.0

20.0

30.0

40.0

50.0

0 30 60 90 120 150 180 210 240Run Time, Days

Ass

ist G

as R

atio

1-H

our A

vera

ge D

ata

Flare B Median = 0 Average Deviation = 13

Figure 16. Manual Assist Gas Ratio Control Example

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

8.00

9.00

10.00

0 30 60 90 120 150 180 210 240Run Time, Days

Ass

ist G

as R

atio

1-H

our A

vera

ge D

ata

Flare C Median = 0.7 Avg. Deviation = 1.4Flare E Median = 0.7 Avg. Deviation = 0.9

Figure 17. DCS Auto Set Point Assist Gas Ratio Control Example



5-10

0

10

20

30

40

50

60

0.00

30.0

0

60.0

0

90.0

0

120.

00

150.

00

180.

00

210.

00

240.

00

270.

00

Run Time, Days

Com

bust

ed F

lare

Was

te G

as, %

Sm

oke

Flare A Days 0-60 Median = 0.4 Avg. Deviation = 8.6 Days 160-240 Median = 0.1 Avg. Deviation = 1.8

Unit Down for Inspection and Turnaround

Figure 18. Infrared Sensor Feedback Assist Gas Ratio Control Example



6-1

6.0 CONCLUSIONS AND RECOMMENDATIONS

Due to the unique requirements of flare systems (low pressures, large pipe sizes, and 1000:1 turndown on flows) only multi-ranged pitot tubes, thermal mass meters, and ultrasonic time-of-flight meters are broadly applicable to the measurement of flows in flare systems. Ultrasonic meters and Thermal mass meters can measure a greater range of gas velocities, and therefore are easier to apply to flare systems than pitot tubes. For the flares examined, all flows were below velocities of 300 fps, the high end measurable by ultrasonic and thermal mass meters. However, many flares are designed for worst-case events that will exceed this measurement capability. Flow measurement systems that combine multiple measurement techniques can measure this design worst case. Using multiple measurement techniques increases the costs associated with purchase, installation, calibration, and maintenance of dual meters. Design worst-case events are infrequent and rarely occur. Typically, design worst-case events are associated with emergency, unplanned shutdowns of a complete process unit or the shutdown of several process units serving a flare due to a power failure. Other process data usually exist which can be used to estimate the mass flow during such events. Thus, the additional expense of installing multiple measurement systems may not be warranted due to the relative infrequency of such events and the availability of process data that can be used to estimate mass flow during worst-case design events. Ultrasonic time-of-flight meters have greater accuracy and precision for measuring flows in flare gas systems, because they are less sensitive to changes in gas composition and can correct for variable gas composition using only flow meter inputs. Thermal mass meters and pitot tubes need significant correction for changing gas compositions. Currently, real time correction for gas composition changes has not been demonstrated in field applications with thermal mass meters or pitot tubes, although vendors claim such meters are available. Correcting with periodic composition data adds to the uncertainty of the flow measurement of thermal mass meters and pitot tubes used in flare applications. Measurement of flows in flare systems can be made with an uncertainty in the range of + 5-10 percent. However, obtaining this accuracy in flare systems with highly variable compositions or



6-2

where the meter cannot be located in a section of pipe with a representative flow profile will be a challenge. Measurement of the gas composition in flare systems on a continuous basis will be a challenge due to the number of compounds of interest and the dynamic range of concentrations, that may be present for the compounds of interest. To design and implement a continuous monitoring system capable of detecting the full range of compounds of interest across all concentrations of interest using conventional GCs will require considerable up-front data collection, and fairly sophisticated combinations of columns and detectors. If aldehydes happen to be present at quantities that would impact achieving the data quality objectives, the GC system required to analyze for them will become even more complex. Portable multichannel, micro GCs have proven extremely capable of detecting most compounds of interest in flare systems across an extremely large dynamic range (105). Application of such instruments in online applications will significantly reduce the complexity of the GC instrumentation needed to achieve the data quality objectives of this study. However, online versions of these instruments have yet to be demonstrated in flare applications. While online GC systems can be designed to meet the data quality objectives of this study a high percentage of the time, the actual percent of time the data quality objectives can be met will be a strong function of the specific flare system, in particular the variability of the waste gas composition. There is insufficient data available to make a fact-based estimate on this percentage. Application of US EPA Performance Specification 9 to online GCs in flare service is problematic given the potential number of compounds of interest and the dynamic range potentially encountered for these compounds. The linearity requirement will be extremely difficult to meet and the calibration procedures impractical. Picking a few representative compounds with which to calibrate and using relative response factors for other compounds of interest seems a practical solution. Due to the large dynamic range, the linearity requirement of PS9 may need to be relaxed in favor of a multipoint calibration curve.



6-3

Under ideal conditions the mass flow in flare systems can be measured with an uncertainty in the + 5-10% range. However, problematic applications with widely varying gas compositions and poor meter installation geometries will have difficulty meeting the + 20% data quality objectives of this study. Obtaining measurement of flare emissions within an absolute uncertainty of + 5,000 lb/hr appears doable for all but the very largest flare. However, this assessment is strongly influenced by the 98% destruction efficiency assumption used in this study. Controlling the assist gas to waste gas ratio by measuring waste gas flow to the flare appears to be more effective at controlling this parameter at a desired value than manual control, but there was very little data available from which to make this conclusion. Additional data collection on the effectiveness of manual control would be warranted. The authors assessment is that current flow and measurement technologies are capable of meeting the stated data quality objectives a high percentage of the time in most flare applications. However, the extremely limited amount of data on real world applications and the high degree of variability in flare applications make it impossible to assess what percentage of the time these data quality objectives can be met, and on what percent of flare systems they may be met. Both flow and gas composition measurement vendors have responded to the added interest in measurement of waste gas flow and composition. Significant recent developments have been made in both flow and composition measurement techniques. Demonstration of these advancements could significantly enhance the ability of sources to meet these data quality objectives in the near future.



7-1

7.0 REFERENCES

1. Schwartz, R.E. and Dr. S. Kang. “Effective Design of Emergency Flaring Systems.” Hydrocarbon Engineering, pp. 57-62, February 1998.

2. Golker, U., Analytical Configurations for HR-VOC Measurement on Flares and Cooling Towers. Draft Product Bulletin, HRVOC Description IV, Siemens Applied Automation, Houston, Texas, June 25, 2003.

3. Dieterich Standard Brochure, “Diamond II+Annubar Primary Flow Element,” 1998 http://www.rosemount.com/document/pds/annubar.pdf 11/17/03.

4. Kurzinstruments Brochure, “K-Bar 2000 Multi Point Insertion Mass Flow Elements” http://kurzinstruments.com/K-BAR.pdf 11/17/03.

5. Panametrics Brochure, “Digital Flow GF868 Mass Flow Meter,” 04/02 http://www.gepower.com/prod_serv/products/flowmeters/en/downloads/gf868.pdf 11/17/03.

6. International Organization for Standardization (ISO), Technical Report ISO/TR 5168 1998-04-01 “Measurement of Fluid Flow – Evaluation of Uncertainties,” Geneva, Switzerland.

7. Gilmer, L.K., Shell Global Solutions US. Confidential Report to Confidential Petrochemical Facility Client, “Flare Gas Recovery – Performance Data for Design,” August 2000.

8. Tru-Tec Services, Inc., Tru-News Vol. 11, Edition 1 and Vol. 6, Edition 1. http://www.tru-tec.com/ 08/05/03.

9. GE Power Systems, “GE Panametrics - Model 7100/7168/GF868 Flare Gas Flow Meter On-Site Calibration And/Or Verification Techniques.”

10. OMEGA Handbook Series, “Transactions in Measurements and Controls,” Vol. 4, “Flow and Level Measurement,” OMEGA Engineering, Inc., Stamford, Connecticut, 2000.

11. Solartron Mobrey, “Gas Density and Specific Gravity Products Data Sheet B1253” http://www.solartronusa.com/downloads/datasheets/DS_b1253.pdf 06/16/03.

12. Janen, D.W.A. Shell Flow Meter Engineering Handbook. McGraw Hill Publishing Company, 2nd Edition, The Hague, The Netherlands, January 1986.

http://www.rosemount.com/document/pds/annubar.pdf

http://kurzinstruments.com/K-BAR.pdf

http://www.gepower.com/prod_serv/products/flowmeters/en/downloads/gf868.pdf

http://www.tru-tec.com/



7-2

13. Neti, Dr. M., California Analytical Instruments, Inc. Technical Presentation on “Innova 1314 Photoacoustic Multi-Gas Monitor” at Westhollow Technology Center in Houston, Texas, June 2003.

14. Smith, S., Process Analytical Applications, Inc., HR-VOC Applications Seminar at Clear Lake Conference Center in League City, Texas, July 2003.

15. Agilent Technologies, “Agilent 3000 Micro GC Data Sheet” http://www.chem.agilent.com/temp/radC7E18/00037995.pdf 08/06/03.

16. Schumacher, M. and B. Thompson, Analysis of Natural Gas and Natural Gas Liquids. Varian Application Note Number 29, Varian Chromatographic Systems, August 11, 2003.

17. US EPA 40 CFR Part 60, Appendix A, Method 18, Measurement of Gaseous Organic Compound Emissions by Gas Chromatography, 10/17/00.

18. US EPA 40 CFR Part 60, Appendix A, Method 25, Determination of Total Gaseous Nonmethane Organic Emissions as Carbon, 10/17/00; Appendix A, Method 25A, Determination of Total Gaseous Nonmethane Organic Emissions Using a Flame Ionization Analyzer, 10/17/00; Appendix A, Method 25B, Determination of Total Gaseous Nonmethane Organic Emissions Using a Non-Dispersive Infrared Analyzer, 10/17/00.

19. US EPA 40 CFR Part 60, Appendix B, Performance Specification 9, Specification and Test Procedures for Gas Chromatographic Continuous Emission Monitoring Systems in Stationary Sources, 10/17/00.

20. US EPA 40 CFR Part 60, Appendix A, Test Methods. Method 301, Field Validation of Pollution Measurement Methods from Various Waste Media, 10/17/00.

21. M. McDaniel, US EPA/600-2-83-052, Flare Efficiency Study. Engineering-Science, Inc., July 1983.

22. D. Joseph, J. Lee, C. McKinnon, R. Payne, and J. Pohl, US EPA/600/2-83-07, Evaluation of the Efficiency of Industrial Flares: Background – Experimental Design – Facility, Energy and Environmental Research Corporation, August 1983.

23. Pohl, J.H., R. Payne and J. Lee, US EPA/600/2-84-095, Evaluation of the Efficiency of Industrial Flares: Test Results, Energy and Environmental Research Corporation, May 1984.

http://www.chem.agilent.com/temp/radC7E18/00037995.pdf



7-3

24. Pohl, J.H. and N.R. Soelberg, US EPA/600/2-85-106, Evaluation of the Efficiency of Industrial Flares: Flare Head Design and Gas Composition; Energy and Environmental Research Corporation, September 1985.

25. Pohl, J.H. and N.R. Soelberg, US EPA/600/2-86-080, Evaluation of the Efficiency of Industrial Flares: H2S Gas Mixtures and Pilot Assisted Flares; Energy and Environmental Research Corporation, September 1986.

26. Baukal, C.E. and R.E. Schwartz. The John Zink Combustion Handbook. CRC Press, 2001.

27. Evans, L.B. and W.M. Vatavuk, EPA/452/B-02-001, Section 3 VOC Controls, Chapter 1, “Flares,” US EPA, September 2000.

28. American Petroleum Institute (API) Committee meeting on “Flare Gas Flow Measurement Standard Development” in Houston, Texas, October 28-29, 2003.

29. Wyble, E., Fluid Components International (FCI), San Marcos, California. Presentation to API Committee on “Flare Gas Flow Measurement Standard Development” in Houston, Texas, October 29, 2003.

30. Sullivan, B., Instrument Tags, Houston, Texas. Presentation to API Committee on “Flare Gas Flow Measurement Standard Development” in Houston, Texas, October 29, 2003.

31. US EPA. Compilation of Air Pollutant Emission Factors (AP-42), 1993, and Miscellaneous Sources, Industrial Flares, September 1991. Research Triangle Park, North Carolina, US EPA.

32. ExxonMobil. HRVOC Reduction Project, Flare Gas Flow Metering, Houston, Texas, 2003.

33. ExxonMobil. “Composition Analysis for Flare Gas and Cooling Water Leak Detection and HRVOC Speciation.” ExxonMobil Baytown HRVOC Reduction Project. Baytown, Texas, 2003.

34. Deleted

35. Kurz Instruments Inc., Theory and Application of Kurz Thermal Convection Mass Flow Meters, 1997 http://www.kurz-instruments.com/a12.pdf 11/17/03.



7-4

36. Yoder, Jesse, “Flowmeter Shootout Part III,” Control for the Process Engineer, 03/09/01 http://www.flowresearch.com/articles/Control%20Shootout%20III%203_9_01.-PDF 11/17/03.

37. Lansing, John, “Principle of Operation for Ultrasonic Gas Flow Meters,” http://www.daniel.com/Products/Gas/usonic/Seniorsonic/AppNotes/WEI%202003,%20Basics%20of%20USMs,%20JL%20109KB.pdf 11/17/03.

38. Flowmeter Directory, “Why and How to Measure Flare Gas,” 02/14/02 http://www.flowmeterdirectory.com/flowmeter_artc/flowmeter_artc_02021401.-html 11/17/03.

39. Smalling, J.W. and L.D. Braswell, Exxon Company, Baytown, Texas; and Lynnworth, L.C. and D.R. Wallace, Panametrics, Inc., Waltham, Massachusetts. “Flare Gas Ultrasonic Flow Meter” reprint from Proceedings - Thirty-Ninth Annual Symposium on Instrumentation for the Process Industries, copyright 1984 by the Department of Chemical Engineering of Texas A&M University at College Station, Texas.

40. Deleted

41. Forbes, A., Texas Commission on Environmental Quality, HRVOC Rules Overview, Presentation to AWMA Gulf Coast Chapter, May 13, 2003.

42. Texas Commission on Environmental Quality. Chapter 115, Control of Air Pollution from Volatile Organic Compounds, Sections 115.125 – 115.126, Testing Requirements.

43. Texas Commission on Environmental Quality. Chapter 115, Control of Air Pollution from Volatile Organic Compounds, Subchapter H: Highly Reactive Volatile Organic Compounds Division 1 Vent Gas Controls Sections 115.720, 115.722, 115.725 - 115.727, and 115.729.

44. US EPA 40 CFR Part 60, Appendix A, Stack Velocity and Flow Measurement and Correction Methods 1-4, 07/01/89 – 10/11/00.

45. US EPA 40, CFR Part 60, Appendix A, Method 22, Visual Determination of Fugitive Emissions from Material Sources and Smoke Emissions from Flares, 10/17/00.

46. US EPA 40 CFR, Part 51, Appendix M, Amended, Method 205, Verification of Gas Dilution Systems for Field I Instrument Calibrations, 06/16/97.

http://www.flowresearch.com/articles/Control Shootout III 3_9_01.-PDF

http://www.daniel.com/Products/Gas/usonic/Seniorsonic/AppNotes/WEI 2003, Basics of USMs, JL 109KB.pdf

http://www.daniel.com/Products/Gas/usonic/Seniorsonic/AppNotes/WEI 2003, Basics of USMs, JL 109KB.pdf

http://www.flowmeterdirectory.com/flowmeter_artc/flowmeter_artc_02021401.-html



7-5

47. US EPA 40 CFR, Part 60, Appendix B, Performance Specification 2, Specification and Test Procedures for SO2 and NOx Continuous Emission Monitoring Systems in Stationary Sources, 10/17/00.

48. US EPA 40 CFR, Part 60, Appendix B, Performance Specification 8, Performance Specifications for Volatile Organic Compound Continuous Emission Monitoring Systems in Stationary Sources, 10/17/00; and US EPA 40 CFR Part 60, Appendix B, Performance Specification 8A, Specifications and Test Procedures for Total Hydrocarbon Continuous Monitoring Systems in Stationary Sources, 10/17/00.

49. US EPA 40 CFR Part 60, Appendix B, Performance Specification 15, Performance Specification for Extractive FTIR Continuous Emission Monitoring Systems in Stationary Sources, 10/17/00.

50. US EPA 40 CFR Part 60, Appendix B, Performance Specification 6, Performance Specification and Test Procedures for Continuous Emission Rate Monitoring Systems in Stationary Sources, 10/17/00.

51. US EPA 40 CFR Part 60, Appendix F, Quality Assurance Procedures, 02/11/91.

52. ASTM D 1945-96, Standard Test Method for Analysis of Natural Gas-by-Gas Chromatography, reapproved 2001.

53. ASTM D1946-90, Standard Practice for Analysis of Reformed Gas-by-Gas Chromatography, reapproved 2000.

54. ASTM D 2588-98, Standard Practice for Calculating Heat Value, Compressibility Factor, and Relative Density of Gaseous Fuels.

55. Gas Processors Association, GPA STD 2261-00. Analysis of Natural Gas and Similar Gaseous Mixtures by Gas Chromatography.

56. Gas Processors Association, GPA STD 2145-03. Table of Physical Constants for Hydrocarbons and Other Compounds of Interest to the Natural Gas Industry.

57. Gas Processors Association, GPA STD 2172-96. Calculation of Gross Heating Value, Relative Density and Compressibility Factor for Natural Gas Mixtures from Compositional Analysis.



7-6

58. Gas Processors Association, GPA STD 2172-96. Calculation of Gross Heating Value, Relative Density and Compressibility Factor for Natural Gas Mixtures from Compositional Analysis.

59. International Standard ISO 6974-1:2000(E) Through 6974-6:2000(E) Natural Gas – Determination of Composition with Defined Uncertainty by Gas Chromatography, 2000.

60. International Standard ISO 6976:1995 Technical Corrigendum 3, Natural Gas – Calculation of Calorific Values, Density, Relative Density and Wobbe Index from Composition, 1995.

61. Gilmer, L.K., Shell Global Solutions (US). Confidential Report to Confidential Petrochemical Facility Client, “Flare Gas Recovery – Performance Data for Design,” April 2002.

62. Agilent Technologies, “Agilent 3000 Micro GC Data Sheet” http://www.chem.agilent.com/temp/radC7E18/00037995.pdf 08/06/03.

63. Thermo ONIX, “Calorific Value Measurement of Flare Gas” 2001 http://www.thermo.com/eThermo/CDA/Applications/Application_Detail/0,1210,1000000010043-11214-136-136-1000000010043,00.html 08/07/03.

64. International Standard ISO 10723:1995(E) Natural Gas – Performance Evaluation for Online Analytical Systems, 1995.

65. de Jonge, G., AC Analytical Controls BV and Frederick, B., AC Analytical Controls Inc., Hewlett Packard Application Note 228-352 Separation and Reporting Hydrogen Sulfide, Inert Gases and C1-C6+ Hydrocarbons in Natural Gas According to GPA 2261, March 1996.

66. Durana, N., et al. “Online Hourly Determination of 62 VOCs in Ambient Air: System Evaluation and Comparison with Another Two Analytical Techniques,” Journal of Air and Waste Management Association 52:1176-1185, 2002.

67. Fashimpaur, D., BP. Petroleum Environmental Research Forum Presentation on “Overview of Conventional Devices for Continuous Emissions Monitoring” in Paris, France, June 3, 2003.


http://www.thermo.com/eThermo/CDA/Applications/Application_Detail/0,1210,1000000010043-11214-136-136-1000000010043,00.html

http://www.thermo.com/eThermo/CDA/Applications/Application_Detail/0,1210,1000000010043-11214-136-136-1000000010043,00.html



7-7

68. Loos, K.R., Shell Global Solutions US. Petroleum Environmental Research Forum Presentation on “Refinery Flaring Revisited or my old Flame” in Paris, France, June 3-4, 2003.

69. Blackwood, T.R. “An Evaluation of Flare Combustion Efficiency Using Open-Path Fourier Transform Infrared Technology.” Journal of Air and Waste Management Association 50:1714-1722, 2000.

70. SRI Instruments, “GC Detector Overview” http://www.srigc.com/index.htm 08/06/03.

71. Flow Dynamics, “Gas Calibration Systems” Primary Standard Flow Calibration Laboratory http://www.flow-dynamics.com/ 08/03.

72. Deleted

73. SKC Inc., “Sampling Guides by Chemical” http://www.skcinc.com/guides.asp 08/06/03.

74. Deleted

75. Rosemount, Inc., “Fundamentals of Flow Metering,” Technical Data Sheet 00816-0100-3031, 03/01 http://www.rosemount.com/document/pds/flowfund.pdf 11/17/03.

76. Walters, J., Procon Systems, Edmonton, Alberta, Canada; Smalling, J.W., Consultant, Baytown, Texas; and Ao, S., J. Hill, and L. Lynnworth, Panametrics, Inc., Waltham, Massachusetts. “Transit-time Ultrasonic Flowmeters for Gases,” presented at and published in part in the Proc. 41st Annual CGA (Canadian Gas Association) Gas Measurement School, Grand Okanagan, Kelowna BC, Canada, June 4-6, 2002.

http://www.srigc.com/index.htm

http://www.flow-dynamics.com/

http://www.skcinc.com/guides.asp

http://www.rosemount.com/document/pds/flowfund.pdf



8-1

8.0 BIBLIOGRAPHY

Golker, U., F. Mueller and U. Gellert, “Application Benefits of a Versatile Compact Process Gas Chromatograph,” Proceedings from Instrument Society of America, Siemens Applied Automation, Houston, Texas, Copyright 2001.

Yoder, Jesse, “Flowmeter Shootout Part II,” Control for the Process Engineer, 02/22/01 http://www.flowresearch.com/articles/Control%20Shootout%20II%202_22_01.-PDF 11/17/03.

Yoder, Jesse, “Flowmeter Shootout Part I,” Control for the Process Engineer, 02/14/01 http://www.flowresearch.com/articles/Control%20Shootout%20I%202_014.PDF 11/17/03.

http://www.flowresearch.com/articles/Control Shootout II 2_22_01.-PDF

http://www.flowresearch.com/articles/Control Shootout I 2_014.PDF

Appendices

Appendix A

Description of Flow Measurement Devices not Applicable to Flare Systems



A-1

A.1 Differential Pressure Techniques Orifice Plates Orifice Run with Differential Pressure Taps Principle of Operation

Orifice plates operate on the principle that the pressure drop caused by constricting flow is proportional to the velocity. In a typical orifice plate application, fluid flows through a pipe and nears the constriction in the plate. The pressure begins to drop and the velocity increases as the fluid flows through the smaller opening. As the fluid exits, the velocity decreases and the pressure increases, but not back to the original pressure. Pressure loss occurs due to friction from the orifice plate and turbulence losses from the flow. Pressure taps located upstream and downstream of the plate measure the pressure differential and using orifice plate equations provide a flow rate. The pressure differential is proportional to the velocity squared. Applicability to Flares

Although orifice plates are explicitly referenced in 40 CFR 60.18 Method 2D, they are not used in flare or waste gas service. They may be used to measure stack flows in situations in which pressure drop and the limitations in the flow meters range are not an issue. In flares, orifice plates have the potential to serve as an obstruction to the flow and often induce a large the plate pressure drop due to. Both of these situations are potential problems for flare lines due to safety considerations. Additionally, the turndown ratio of orifice plates is typically a 3:1 ratio but may be extended in some circumstances to 10:1. This range of turndown ratios is not suited for flare flow measurements that may fluctuate considerably. Venturis: Venturi Run with Differential Pressure Taps Principle of Operation

Venturi flow tubes operate on the same principles as orifice plates, constricting the flow through a known opening and equating the pressure drop to a velocity. Instead of using a plate and an orifice, venturi flow tubes use a flow channel that has a tapered inlet and a flared outlet. The basic underlying principle remains the same: as the area decreases through the venturi opening, the pressure of the fluid decreases. As with the orifice plate, there is a head loss associated with



A-2

the venturi models due to friction and turbulence. However, overall pressure loss is less than in orifice plate applications. Applicability to Flares

Venturi meters are not mentioned outright in 40CFR60.18 Method 2D, but it is eluded to in being a pressure drop measuring device. While they may be used to measure stack gases in similar situations as orifice plates, Venturis are seldom used in flare applications. Reasons include obstruction concerns, large pressure drops, and low turndown ratios. The biggest driver being the pressure drop, as flare lines are generally pressured slightly above atmospheric and therefore cannot have a large pressure drop across the flow measurement device. Rotameters Variable Area Rotameters Principle of Operation

The operation of a rotameter is quite simple, as fluid flows through the device a float is raised that indicates fluid velocity. The basic components consist of a tapered tube, a float that is the same size as the inner diameter of the smaller end of the tapered tube, and a scale that measures the flow rate in proportion with the float height. Rotameters are often mounted vertically, with the fluid passing upward through the smaller end of the tapered tube to the larger opening. As the fluid enters the tube the float is raised, increasing the size of the area around which the fluid can flow. As the float raises and the area increases, the differential pressure around the float decreases. The float equilibrates when the force exerted by the fluid is equal to the weight of the float. For this reason, floats and the tubes are often designed specifically for the fluid involved in the process. Applicability to Flares

Rotameters are referred to in 40CFR60.18 Method 2D, but have never been used in flare applications. Flares often have a tendency to have a wide range of flows while rotameters are ideal for small volumes of known flows. In most flare applications the sheer size needed to capably handle the potential flare flows rules out this technology.



A-3

Mechanical Techniques Turbines Principle of Operation

Turbine flow meters operate simply by counting the rotations of a bladed rotor. Generally the rotor is suspended into the fluid flow and the rotor is slightly smaller than the actual metering chamber. As fluid enters the chamber it rotates the blades of the rotor. Each rotor pass corresponds to a calibrated amount of fluid. Rotor passes are counted through the use of pulses generated from magnets in the coils or in the rotor blades themselves. Applicability to Flares

Turbine meters are referred to in 40CFR60.18 Method 2A, but are not generally used to measure flare flows. In the past, some forms of insertion turbine meters have been used to measure flare flows, but recorded accuracies were very poor1. They are not considered a reliable, accurate option for measuring waste gas flare flow. In the slim chance of mechanical failure, the turbine itself could act as an obstruction in the flare line and therefore would pose a safety risk. Turbines generally also incur a moderate to high-pressure drop in the line once installed, which is unacceptable in flare line flows. Pistons/Discs/Impellers


Piston, disk, and impeller meters operate similarly to turbines in that they equate the device’s rotations or oscillations to a velocity. All three rely on the force of the fluid to operate the flow meter. Piston meters use the fluid flow to power an oscillating or reciprocating piston. The piston's oscillations or strokes are counted by a magnetic transmitter or, in less sophisticated models, a mechanical register. Disk meters work very similarly to turbines except a disk-like plate is used instead of bladed rotors. The disk may have fins to help catch the fluid used to power the plate. A magnetic transmitter counts the disk’s rotations similarly to a turbine or by mechanical means. In an impeller meter, the fluid rotates an impeller that is linked to mechanical registers.



A-4

Applicability to Flares

Pistons, disks, and impellers are not explicitly mentioned in 40CFR60.18 Method 2A, but they are devices capable of measuring volume. However, pistons, disks, and impellers are not used in flare or waste gas services. For all three of these flow meters, the energy of the fluid is used to operate the device by rotating the piston, disk, or impeller. This is unattractive for measuring waste gas flows to flares since impedances in the gas flow can lead to process upsets or safety concerns. Using the force of the fluid to power the flow meter often leads to pressure loss, which is unacceptable in flares. Dry and Wet Test Meters Principle of Operation

Dry and Wet test meters operate similarly to other mechanical meters in that the fluid powers some form of driver that is counted by a magnetic transmitter or mechanical register. Dry test meters are used in many stack-sampling applications. They consist of chambers that are alternatively compressed and expanded by the gas. The compression/expansion action drives the device and also turns a mechanical register that measures gas usage. Wet test meters are also used in many stack-sampling applications. A typical wet test meter consists of a mutating disk that wobbles as water flows through the device. As the disk wobbles it rotates a spindle that is connected to the mechanical register. The velocity of the fluid is proportional to the rotations of the spindle. Applicability to Flares

Dry and wet test meters are not mentioned explicitly in 40CFR60.18 Method 2A, but they are devices capable of measuring volume. Dry and wet test meters are used in many stack sampling applications but are not used in flare applications. As with all mechanical meters, dry and wet test meters have the possibility to act as obstructions in the pipeline, which could lead to safety issues. Additionally, these meters are not designed to handle the large flow rates or the wide ranges of flow rates associated with waste gas flows to flares.



A-5

Direct Mass Measurement Techniques Coriolis Mass Meters Principle of Operation

Coriolis meters operate on the principle of the Coriolis effect and Coriolis acceleration. The instrument achieves this by using one or two tubes that are anchored and equipped with a coil and a magnet. Current is sent through the coil, which causes the tubes to vibrate at a set frequency. At zero flow, magnetic sensors pick up the vibrations and compute to an even sine wave. When fluid is flowing through the pipe, the vibrations cause a force to act on the fluid due to Coriolis acceleration and an equal force is imparted on the tube. This force causes the tubes to slightly twist, which is measured by the magnetic sensor. The resulting sine wave becomes slightly out of phase, which is an indication of the mass flow. Additionally, the resonant frequency of the vibrations, or period of the sine wave, is an indication of the density of the fluid. Applicability to Flares

In flare applications, Coriolis meters are not a viable option due to the size constraints and pressure loss through the instrument. As previously stated, flare lines can be sized up to 48". Coriolis meters typically are not available in sizes over 6", creating an obvious problem in measuring such large flows. Additionally, the cost of Coriolis meters increases dramatically the larger they are sized. The expected pressure loss through the device is also a concern, as Coriolis meters can have a head loss of 10psig. This will be an issue in many flare applications since flare header lines are usually pressurized slightly above atmospheric conditions. Electronic Flow Meters Magnetic Meters Principle of Operation

The major principle of operation for magnetic flow meters is Faraday's Law. Essentially, coils located in the flow meter produce a magnetic field across electrodes. As the fluid, which must be slightly conductive, flows through the magnetic field, a voltage is created. Transmitters relay the voltage, which has been converted into a current or a frequency output, to a control system that infers a mass flow rate.



A-6

The underlying Faraday's equation is: Equation 1

kBDV E = Where: k = Faraday Constant E = Induced Voltage Generated by the Coils B = Magnetic Field Strength D = Distance Between Electrodes V = Velocity of Conductive Process Fluid Applicability to Flares

Magnetic Flow meters are not recommended for use in flare or waste gas streams. Magnetic flow meters must always be full of liquid while in service, consequently this is impossible to achieve in flare lines. Flare flows, aside from not being liquid, also are not inherently conductive, thus negating the instrument and technology. Vortex Meters


Vortex flow meters incorporate a bluff body, which is an obstruction with a flat, broad front facing into the fluid stream. As the fluid flow impacts the bluff body, a vortex is formed on one side of the bluff body. On the side of the vortex formation, velocity increases and pressure decreases. As the vortex grows downstream it eventually detaches, the same pattern is then repeated on the opposite side of the bluff body. Sensors determine the flow velocity by measuring pressure oscillations or through temperature differences. Additionally, external sensors are also used to determine the flow rate by measuring the force impacted on the bluff body.



A-7

The flow velocity is proportional to the frequency of the vortex formations. The relationship used to calculate velocity is: Equation 2

( )StD /f V =

Where: V = Fluid velocity f = vortex shedding frequency D = width of bluff body St = Strouhal Number (a dimensionless number normally approximated as 0.2

for vortex shedding purposes) Applicability to Flares

Ideal conditions for Vortex meters are clean, high-speed flows that are relatively free of turbulence. Steam flow lines are one ideal service for Vortex use. Flare flows, on the other hand are not clean flows and can have very low flow conditions as well as very high flow conditions. Corrosive services and those with particulates present problems to vortex meters. Additionally, vortex meters do not have the required turndown ratio to handle the range of flows present in flares. Additional issues are the pressure drop and obstruction that the bluff body presents. Both are capable of presenting enough concern to make vortex meters unacceptable for measuring waste gas flow rates to flares. (Authors’ note: A new generation of vortex meter, which does not use a bluff body, is in the final stages of development and may soon enter the market. Vendors have indicated that this new type of vortex meter may be applicable to flare applications. Insufficient data on this new meter was available at the time of this writing to include a meaningful evaluation of this type meter.)



A-8

Reference

1. Flowmeter Directory, “Why and How to Measure Flare Gas,” 02/14/02 http://www.flowmeterdirectory.com/flowmeter_artc/flowmeter_artc_02021401.-html 11/17/03.


Appendix B

Technical Considerations for Design and Operation of Sample Systems for Flare Service



B-1

Real time representative compositions are necessary for three reasons. First factors are calculated based on composition to generate a pseudo specific heat capacity, thermal conductivity, and dynamic viscosity. These factors are used to adjust measured values from thermal mass flow meters or time-of-flight meters for improved accuracy and precision. Second, the compositions may be required to estimate HRVOC emissions using flow rate, composition, and destruction efficiency. Third, the composition may be used to determine heating value that factors into destruction efficiency. 1,2,3,4 Flow Loops Continuous flow loops to obtain representative waste gas samples at grade or easily accessible points are advisable, especially for waste gas streams where variable compositions are anticipated such as may be encountered where multiple unit operations are manifold into one flare header or where unit batch operations exist. The selection of a sample draw location that minimizes liquid or condensable entrainment as well as promotes collection of a time weighed sample is preferred. Avoid draw locations where there is a potential to sample stagnant gas or entrained liquids. An example of a suitable location is a high point gate valve on top of the final wet-gas knockout drum downstream of any baffles or weirs. Ideally, the sample loop configuration runs over to the measuring instrumentation where a low volume tee allows sample introduction and then the loop is completed by return into the flare header, preferably but not essentially downstream of where the sample is drawn off. Pumps Since most elevated flare systems operate very near atmospheric pressure, a pump is necessary to drive the gas through the loop. A small, portable instrument air-driven impeller pump with four vanes is one option successfully demonstrated. Since there is only very small pressure differential under most conditions, a low powered, moderate velocity pump works well for this application. The pump is positioned upstream of the filters and upstream of the sample introduction tee. This insures the necessary positive pressure required for sample introduction. The drive for the pump can be a regulated facility instrument air in the 30-100 psig range.



B-2

A high flow back pressure regulator with a regulating valve may be installed to drive the flow in many ground flare designs, however it is important to determine if there is sufficient pressure drop at the return line for an adequate and reliable measured flow. Ejectors For permanent installations, a steam or nitrogen ejector positioned downstream of the sample introduction may be a good alternative to pumps. Jet ejectors are gas or steam-driven venturi jet devices that use the energy in compressed gas or steam to create a vacuum that can entrain or mix another gas stream. Ejectors offer less maintenance and less potential for contaminants but may also prevent an easy return of the stream back into the waste gas header. Aspirators or ejectors using air for the motive fluid should never be sent back into the flare header. Ejected waste gas should not be vented to atmosphere because of flammability, toxicity, and environmental impact. Tubing and Valves Nominal bore one-quarter inch outside diameter (OD) 316 stainless steel tubing is typically used for loop runs of less than 100 feet. Three-eighths inch to one-half inch tubing can be used for longer loops if required. This addresses most safety concerns associated with flammable gas and toxic gas. Polyethylene, Teflon, Tygon or other plastic tubing is not suitable. Additionally, small diameter tubing minimizes volume, resulting in more representative real time sampling. Some locations may also require installation of a backflow check valve. If this is necessary, a one- third pounds per square inch gage (PSIG) cracking pressure valve is usually sufficient. Plugging with wet sulfur can cause check valves to become obstructed. Block valves at the sample draw and return is a good safety measure. Block valves and by-pass tubing around the instrumentation, the pump, the filters, and the rotameter are used for isolation to perform any necessary maintenance. Sample Conditioning Pre-analysis sample conditioning typically includes dual series in-line coalescing filters placed downstream of the pump and before the rotameter. A protective gas selective membrane filter on the instrument inlet and a sintered metal particulate filter on the instrument inlet help prevent any fouling or plugging of the analyzers.



B-3

Rates an Synchronizing Cycles Rates should be set so that loop throughput is a minimum of 10 volumes of waste gas per sampling cycle. For most applications a rotameter is placed in line to gauge the flow settings. Typically 10 Standard Liters Per Minute (SLPM) is sufficient for most conditions. The flow meter also acts as a flow indicator to alert the operator to any plugging issues or liquid entrainment. A small in-line coalescing type filter is also advisable to trap sour water, sulfur or other condensable hydrocarbons. A low point drain can be used to periodically collect fractions of condensed liquids for laboratory analysis or to manually drain any collected material into a waste collection container for tracking volume or simply proper disposal. Materials Compatibility, Conditioning, and Sample Integrity Assurance All wetted materials should be compatible for wet corrosive service and rated for the design pressure of the flare, which is typically from 30 to 50 psig. The loop system should be run for several hours to condition any new tubing for hydrogen sulfide service and deactivate any surface effects associated with new tubing. 304 or 316 stainless steel works well for most applications. A sample loop recovery test may be performed to insure acceptable reactive or sorbable species losses. This can be done by injecting a flowing gas standard of a trace amount of a chemical of interest such as 1-butene at the point where the sample is drawn and then running samples over time and comparing those values with samples directly taken from the standard and injected into the instrument. Fraction recovery criteria as per EPA Method 18 are considered validation, 0.70 < R < 1.30. Starting concentrations should be low enough such that small losses can be recognized; a suggested working standard of 100 ppmv to 1000 ppmv is an appropriate range for levels of concern. Because the temperatures of the waste gases at the knock out drums are generally near ambient temperatures, no heat tracing of the sampling lines is required under most conditions. Grab Samples Grab samples may be taken when the duration of testing is short, such as verification testing done to determine an exemption or the need for continuous compliance monitoring when it is well established by analysis or other means that the composition of the waste gas does not significantly change. Multiple samples may be collected over several hours to demonstrate performance or compliance. This would typically be used in a scenario where steady state conditions are expected by design or historical documentation. Generally, three types of grab samples are available: Tedlar Bags, evacuated stainless steel Department of Transportation



B-4

(DOT) sample cylinders, and evacuated Summa or silanized canisters for more reactive chemicals such as H2S and 1,3-butadiene. Some considerations when using these containers are: exposure potential and safety considerations of a rupture using Tedlar bags with toxic concentrations of hydrogen sulfide, evacuation of dead space from the sampling ports, driving force needed to fill the containers, driving force needed to introduce the collected sample into the instrument for analysis, hold times, proximity and reliability of the laboratory, and reactivity. Since the waste gas streams will be very near atmospheric pressure a Vac-U-Tube or Vac-U-Chamber may be required to fill an evacuated bag. These devices act by applying a negative pressure around the outside of a Tedlar bag and help eliminate the risk of contamination associated with dirty pumps. 5 Tedlar bags should be clean, evacuated, and analyzed within 48-hours. Canisters and cylinders should be clean and evacuated. Canisters and cylinders generally have longer hold time than bags, however it is good practice to have the samples analyzed as soon as reasonably possible. Use a chain of custody form to record sample position, start and stop times, beginning and ending pressures, and cylinder or canister identification. Samples should be taken in duplicate during each sampling event as a minimum and analyzed in duplicate. Using a travel blank of a standard in the same collection media and treating it as a sample can obtain a degree of quality assurance. A travel blank containing air can identify any dirty or contaminated sample media. Generally follow evacuated container sampling procedures are listed in EPA Method 18 for details and variants. Pressure Considerations Many elevated flares are operated at or very near atmospheric pressure. Care should be taken to expect pressure fluctuations and increases up to design pressures due to unexpected releases. Many ground flares operate at some positive pressure, typically 2-15 psig. Additionally many ground flares are staged such that secondary and tertiary burners are activated when a set flow threshold is reached. This can cause pressure spikes and dips. In some designs of flares used solely for emergency service, it is important to recognize that under routine operating process conditions there may be rupture disks or water seals that prevent any waste gas flow while only supplemental fuel or inert sweep gas may be present.



B-5

The delivery pressure at the injector port to multi channel micro gas chromatograph (GC) instruments can range from 0 psig to 7.5 psig without affecting mass injected into the analyzer. An internal pressure transducer equalizes the pressure with ambient barometric pressure before the sample is introduced on the columns. The sample is drawn through a filtered and heated inlet system by a timed vacuum pump. By minimizing the tee volume using 0.125” OD tubing and a <4” long tube connector and with consideration of the filter housing volume, a 40-second sample pump time is sufficient to obtain representative samples. It is important to maintain a fixed pressure in sample loops that may be used with continuous emission monitoring systems or periodic emissions monitoring systems High flow backpressure regulation may be required. 6,7 Performing an analysis to measure oxygen with the sample loop system open to the waste gas can serve as a good check for leak integrity in a nearly atmospheric system, since the waste gas should not have any oxygen and any oxygen detected would consequently come from a breach in the sampling system or a breach internal to the analyzer. Traditional static leak checking of the sample loop using pressurized nitrogen and soap bubble solution is recommended before startup. Frequency and Cycle Considerations Some pressure surge or resonance is common on many flare systems. It is not uncommon for waste gas streams to cycle rapidly from several inches of water vacuum to several inches of water pressure. Sources of pressure fluctuations can include: sizing and seal design of wet knock out vessel internals (sloshing); 3 mismatched manifold sources from different unit operations; lack of pulsation dampening from compressor or plunger pump operations; extent of, size, and position of dry knock out drums; and internals design with water and level interface considerations in the wet gas knock out drums. Frequencies of pressure cycles have been observed in the range of less than 2-second to 30-second periods. 8,9 Condensate Condensable water can be minimized in the sample loop by selecting an elevated point off the knock out vessel. Most permanent continuous installations will be heat traced to minimize condensate. An in-line vessel with inert demister material such as glass wool can also be used in extreme conditions.



B-6

The presence of ammonia gas in the waste gas can be problematic, as salting out can occur. This can result in plugging of fine capillary tubing in instrumentation, or fouling of pumps and rotameters. Larger-bore tubing, lower gas rates through the loop, back flushing with nitrogen, and heated instrument regeneration are remedies used with varying degrees of success in conditions favorable to salting out phenomena. Under extreme salting conditions, addition of an ammonia scrubbing device may be required. Ammonia and low-mass amines are usually checked prior to continuous sampling using an electrochemical cell type instrument or colorimetric indicator tube from a collected grab sample. 8 Particulates Particulates can be a plugging and fouling issue especially when tying into lines that have not been in service for years. Pulsed low-pressure nitrogen from a regulated cylinder can help free lose iron sulfide and other metals from the sampling loop. Sticky sulfur buildup can also be a cause of plugging. 8 The loop sampling systems will not collect any significant condensable hydrocarbon since any slugs of liquid hydrocarbon are trapped at the unit boundary knockout drum or condensed in the wet knockout drum. Any condensate trapped in the coalescing filter drop out can be collected, weighed and analyzed by the refinery lab or third party lab if desired. 8,9



B-7

References Deleted by Jake Kurz Instruments Inc., Theory and Application of Kurz Thermal Convection Mass Flow Meters, 1997 http://www.kurz-instruments.com/a12.pdf 11/17/03. Lansing, John, “Principle of Operation for Ultrasonic Gas Flow Meters,” http://www.daniel.com/Products/Gas/usonic/Seniorsonic/AppNotes/WEI%-202003,%20Basics%20of%20USMs,%20JL%20109KB.pdf 11/17/03. Flowmeter Directory, “Why and How to Measure Flare Gas,” 02/14/02 http://www.flowmeterdirectory.com/flowmeter_artc/flowmeter_artc_02021401.-html 11/17/03. SKC Inc., “Sampling Guides by Chemical” http://www.skcinc.com/guides.asp 08/06/03. Agilent Technologies, “Agilent 3000 Micro GC Data Sheet” http://www.chem.agilent.com/temp/radC7E18/00037995.pdf 08/06/03. Schumacher, M. and B. Thompson, Analysis of Natural Gas and Natural Gas Liquids. Varian Application Note Number 29, Varian Chromatographic Systems, August 11, 2003. Gilmer, L.K., Shell Global Solutions US. Confidential Report to Confidential Petrochemical Facility Client, “Flare Gas Recovery – Performance Data for Design,” August 2000. Gilmer, L.K., Shell Global Solutions US. Confidential Report to Confidential Petrochemical Facility Client, “Flare Gas Recovery – Performance Data for Design,” Apri1 2002.


http://www.skcinc.com/guides.asp


Appendix C

Analytical Techniques for Aldehydes



C-1

Formaldehyde and acetaldehyde are not expected to be present in reduced waste gas streams from refineries or olefins plants. They may be present in waste gas streams from a few process units that employ oxidation such as production of the solvents acetone or methyl ethyl ketone, intermediates such as ethylene oxide, propylene oxide, and phenol, or secondary processes such as urea-formaldehyde based resin productions. Manual Testing Methods for Aldehydes The authors experience with aldehydes in waste gas streams is limited. The authors have done screening on five refinery waste gas streams using a Kitagawa colorimetric detector tube #171SB and have not detected any formaldehyde or acetaldehyde in the 1-35 ppmv measurement range using this technique. These tubes use a pretreatment tube to neutralize interfering reactants. In each of these tests, none of the listed interfering compounds were expected to be present basis the process. These tests were performed on a composite three-liter Tedlar bag sample using a fixed volume plunger pump. A modified EPA TO-11A method1 was used to determine the presence of formaldehyde and acetaldehyde on the inlet and outlet to a catalytic oxidizer on the vent gas from a Phenol Acetone Unit. Modifications involved shorter duration sample times of 1, 10 and 20 minutes since concentrations from a process vent were unknown but were expected to be greater than ambient air concentrations.2 With this distributed sampling technique employed, validity of data is better ensured in terms of matrix effects, retention volumes, and breakthrough volumes. Results achieved using this technique were +/- 30% for duplicate samples and replicate analyses. All results were in the low ppmv range. Metered active sampling with before and after flow checks using primary flow calibrators was used as well as additional tubes in series to assure validity by demonstrating no breakthrough. Silica gel coated 2,4- dinitrophenylhydrazine (DNPH) tubes from SKC were used as well as an ozone scrubber tube to form stable hydrozone derivatives. Blank DNPH cartridges were analyzed but no spiked sample or round robin testing was performed as the testing was not for compliance demonstration. Analysis was performed by HPLC with UV detection. TO-11A quality assurance and quality control criteria limits were met for flow calibration, leak check, sampler blank, back up cartridges, trip blanks, field blanks, continuing calibration standard, and replicate injections.



C-2

Use of EPA Method TO-53 that employs midget impinges containing the DNPH derivitizing reagent is applicable but not recommended. We are in agreement with the EPA Method 11A summary statement which outlines the drawbacks to EPA TO-5 as labor intensive, uses acidic and hazardous reagents, lacks sensitivity, prone to interference’s, and poor reproducibility at ambient concentration levels. Use of OSHA Method 52 which employs a 2-(hydroxymethly) piperidine (2-HMP) derivitizing agent on solid XAD-2 resin may be applicable to flare applications. However, it is generally not recommended because it specifies use of a relatively uncommon nitrogen phosphorus detector (NPD). It is an industrial hygiene ambient air / workplace method. We have no experience with this method. Other essentially similar specific ambient air method for aldehydes include EPA Method IP-6A (Formaldehyde and Other Aldehydes in Indoor Air), NIOSH Methods 2016 and 2532, and ASTM method D5197. These methods all use Silica gel treated with 2, 4-dinitrophenylhydrazine (DNPH). There is a hand held electrochemical cell based aldehyde meter available, the “Formaldemeter 400” introduced by PPM Technology in 2000.5 It is battery operated with a 10 second response time, and with a range of 0.05-10 ppmv with an option to extend the range to 100 ppmv. This instrument is unproven, and interferences in waste gas streams are unknown. The product is focused on applications for industrial hygiene in medical and industrial environments, but may be a potential technology for waste gas stream screening. It will non-selectively detect formaldehyde, acetaldehyde, propylaldehyde and glutaraldehyde. Online Monitoring of Aldehydes

Formaldehyde and acetaldehyde are not expected to be present in reduced waste gas streams from refineries or olefins plants. They may be present in waste gas streams from a few process units that employ oxidation such as production of the solvents acetone or methyl ethyl ketone, intermediates such as ethylene oxide, propylene oxide, and phenol, or secondary processes such as urea-formaldehyde based resin productions.



C-3

Traditional chromatographic techniques have proven less than desirable for formaldehyde with direct injection. Poor recoveries and detector response are generally encountered. For practical use online, a separate stand-alone gas chromatograph with FID may be used in conjunction with a DB-5 column, heart cut valve or two if both formaldehyde and acetaldehyde are present and a methanizer for conversion to optimize FID response. This may be included in a single instrument with a parallel FID detector, however it may also add complexity such that cycle times are increased to an unacceptable level. “Heart cutting” can be characterized as use of a mulitiport valve to directly remove selective components based on retention time after they have been resolved from the other components and before they reach the detector. The components are then either held in a sample loop, held on a trap or placed on another column for further resolution and speciation. In this scenario, formaldehyde and acetaldehyde can be sent to a “methanizer” FID which reduces the aldehydes to methane so they can be detected over a range comparable to methane. A Siemens Maxum ll GC/FID was recently brought online in a mobile van used for EO process gas monitoring. This configuration uses a methanizer for formaldehyde and acetaldehyde. The Methanizer option enables the FID to detect low levels of aldehydes or other combustible oxygen containing chemicals. It is installed as the removable jet in a special FID detector assembly. The Methanizer is packed with a nickel catalyst powder. During analysis, the Methanizer is heated to 380oC with the FID detector body. When the column effluent mixes with the FID hydrogen supply and passes through the methanizer, aldehydes are converted to methane. Due to the chemical relationship between nickel and sulfur, the methanizer can be poisoned by large quantities of sulfur gas. Recovery, linearity and, response was determined to be adequate for levels of acetaldehyde normally encountered, however, for the smaller amounts of formaldehyde normally generated (<10ppmv), there were significant problems encountered in reproducibility of standards on the order of +/- 30% volume. Since this application does not involve emissions, no comparative PS9 testing data is available.6 After speaking with at least two vendors who are promoting their online HRVOC analyzers for waste gas streams, it appears that a requirement for analysis of aldehydes or ketones is not a primary consideration in the HRVOC first tier design, but rather is viewed as a potential secondary ancillary add-on.



C-4

References

1. US EPA Compendium Method TO-11A ( Determination of Formaldehyde in Ambient Air Using Adsorbent Cartridge Followed by High Performance Liquid Chromatography (HPLC)) [Active Sampling Methodology] January 1999.

2. Caico, C.A., K.O. Moore and R. Szentirmay, Shell Global Solutions US - Internal Work Product, Modified TO-11A Testing Results. July 8-21, 2002.

3. US EPA Compendium Method TO-5 (Method for the Determination of Aldehydes and Ketones in Ambient Air Using High Performance Liquid Chromatography,Rev.1 April, 1984.

4. US EPA Monitoring and Quality Assurance Group Photochemical Assessment Monitoring Stations (PAMS) Issue Summary Section 5, Rev. 1 September 30, 1998 Methodology for Determination of Carbonyl Compounds in Ambient Air.

5. PPM Technology, “Formaldemeter 400 Technical Specifications” http://www.ppm-technology.com/ 08/06/03.


http://www.ppm-technology.com/

Appendix D

Gas Chromatography Detectors



D-1

D.1 Thermal Conductivity Detector

Because it detects all molecules, the Thermal Conductivity Detector is commonly used for fixed gas analysis (O2, N2, CO, CO2, H2S, CH4, etc.) where the target analytes do not respond well on other, more sensitive detectors. The TCD can detect concentrations from 100 ppmv up to 100 % volume on a flat baseline with sharp peaks. Where the peak is broad or the baseline is not flat, detection limits of 300ppmv are more realistic. The TCD consists of four tungsten-rhenium filaments in a Wheatstone bridge configuration. Electric current flows through the filaments, causing them to heat up. Carrier gas (usually helium, which has very high thermal conductivity) flows across the filaments, removing heat at a constant rate. Two of the filaments are exposed only to carrier gas (reference), and two are exposed to the carrier/sample flow. When a sample molecule with lower thermal conductivity than the carrier gas exits the column and flows across the two sample filaments, the temperature of the filaments increases. This temperature increase unbalances the Wheatstone bridge and generates a peak as sample molecules transit through the detector. Sensitivity may be adjusted by the filament temperature, the detector temperature, and the signal amplification. The TCD is widely employed in many standard methods of analysis for gas streams, and is widely available on portable and permanent gas chromatographs. 1 Helium Ionization detectors

The helium ionization detector or HID is a universal type detector that responds to all small molecules except neon. It is not widely employed on continuous or portable analyzers. It is especially useful for volatile inorganics like NOx, carbon monoxide and ammonia. The HID needs no hydrogen or air and requires only helium and makeup gas. Sensitivity is in the low ppmv range. The HID incorporates electrodes, which support a low current arc through the helium make-up gas flow. This elevates the helium to a metastable state. When the metastable helium molecules collide with sample molecules as they elute from the column, the sample molecules are ionized and attracted to a collector electrode, amplified, and output to the data system. A useful application of HID in analysis of waste gas streams is in conjunction with a TCD for determination of ammonia. The detector becomes saturated in the low % ranges. 1



D-2

Photoionization Detectors The Photoionization detector or PID is a non-destructive detector used to quantify gases that have excitation energy above a constant voltage supplied by an ionizing UV lamp. The PID responds to molecules with carbon double bonds and is principally used for unsaturated olefinic or aromatic compounds. The response is poor for saturated hydrocarbons, especially lighter alkanes that would be commonly found in waste gas streams. This detector has applicability as an ancillary detector for specific determination of HRVOCs with customized chromatographic configurations. There are maintenance and quality assurance issues related to the PID lamp. The lamp has a limited life of typically two months, the lamp sensitivity decays non-uniformly over time, and the window becomes fogged over time. 1 Flame Ionization Detectors Flame ionization detectors are commonly used in Method 18 and for many hydrocarbon applications. The FID has good linear response over low ppmv to high % ranges and can act as a carbon counter for unknowns. The FID responds to molecules with a C-H bond. One accessory that can be used with an FID is a “methanizer”. The methanizer reduces CO and CO2 to methane using a special catalyst jet. The FID has traditional applicability for waste gas streams, but is unable to detect inert gases, hydrogen, oxygen, carbon monoxide, and carbon dioxide. The FID is widely employed in many standard methods of analysis for gas streams, and is widely available on portable and permanent gas chromatographs. 1 Catalytic Combustion Detectors

The Catalytic Combustion Detector or CCD

The CCD detector consists of a coil of platinum wire embedded in a catalytic ceramic bead. A small electric current flows through the platinum coil, heating the ceramic bead to around 500oC. The CCD is maintained in an oxidative environment by using air carrier gas. When a hydrogen or hydrocarbon molecule impacts the hot bead, it combusts on the surface and raises the temperature and resistance of the platinum wire. This resistance change causes the detector output signal to change, thus producing a peak. The CCD is selective as an FID but has about a 50x less sensitive lower detection limit. The CCD is operated on air alone as a required gas source. The CCD is generally not applicable to HRVOCs in waste gas streams and is not widely employed. 1



D-3

Thermal ionization Detector TID Thermionic Ionization Detector – An electrically heated thermionic bead causes a catalytic surface reaction with nitro functional groups and the resulting ions are attracted to a collector electrode and amplified. Since the TID is extremely selective, having little or no response to most aromatic and aliphatic hydrocarbons, it has low applicability for waste gas flare measurements. 2 Flame Photometric, NP, and Electron Capture Detectors Flame photometric, nitrogen phosphorous (NP), and electron capture detectors are chemical class specific detectors and not directly applicable to waste gas measurements. The electron capture detector offers extreme sensitivity for electronegative compounds (parts per trillion for SF6) and is a good detector of tracer testing when performed with sulfur hexafluoride.



D-4

References 1. Loos, K.R., Shell Global Solutions US. Petroleum Environmental Research Forum

Presentation on “Refinery Flaring Revisited or my old Flame” in Paris, France, June 3-4, 2003.


http://www.srigc.com/index.htm

Findings from the Second Texas Air Quality Study (TexAQS II)

Documents

Transcript of Findings from the Second Texas Air Quality Study (TexAQS II)