Fluid Catalytic Cracking Unit Emissions and Their Impact

Wael H. Yateem & Vahid Nassehi &Abdul Rehman Khan

Received: 12 June 2010 /Accepted: 18 August 2010 /Published online: 1 September 2010# Springer Science+Business Media B.V. 2010

Abstract Fluid catalytic cracking unit is of greatimportance in petroleum refining industries as it treatsheavy fractions from various process units to producelight ends (valuable products). FCC unit feedstockconsists of heavy hydrocarbon with high sulfurcontents, and the catalyst in use is zeolite impregnatedwith rare earth metals, i.e., lanthanum and cerium.Catalytic cracking reaction takes place at elevatedtemperature in fluidized bed reactor generating sulfur-contaminated coke on the catalyst with large quantityof attrited catalyst fines. In the regenerator, coke iscompletely burnt producing SO2, PM emissions. Theimpact of the FCC unit is assessed in the immediateneighborhood of the refinery. Year-long emissioninventories for both SO2 and PM have been preparedfor one of the major petroleum refining industry inKuwait. The corresponding comprehensive meteoro-logical data are obtained and preprocessed usingAermet (Aermod preprocessor). US EPA approveddispersion model, Aermod, is used to predict groundlevel concentrations of both pollutants in the selected

study area. Model output is validated with measuredvalues at discrete receptors, and an extensive para-metric study has been conducted using three scenar-ios, stack diameter, stack height, and emission rate. Itis noticed that stack diameter has no effect on groundlevel concentration, as stack exit velocity is a functionof stack diameter. With the increase in stack height,the predicted concentrations decrease showing aninverse relation. The influence of the emission rateis linearly related to the computed ground levelconcentrations.

Keywords Dispersion model . Aermod . Emissions .

FCC . Pollutants exceedance

1 Introduction

Fluid catalytic cracking (FCC) of heavy ends intohigh value liquid fuels is a common practice in the oilrefining industry. In this process the heavy feedstockcontaining sulfur as a major contaminant is cracked tolight products. Sulfur is redistributed in the liquid andgaseous products and coke on the catalyst. In theregenerator coke with sulfur contamination is com-pletely burnt and flue gas containing SO2 is dis-charged. In the present work, a comprehensiveemission inventories from FCC unit in an oil refineryhave been prepared. These inventories are calculatedbased on complete combustion of sulfur and cokeimpregnated on the catalyst in the regenerator. Mainly

Water Air Soil Pollut (2011) 218:37–47DOI 10.1007/s11270-010-0622-7

W. H. Yateem (*) :V. NassehiDepartment of Chemical Engineering,Loughborough University,Leicestershire LE11 3TU, UKe-mail: w.yateem@lboro.ac.uk

A. R. KhanDepartment of Environment Technology and Management,College for Women, Kuwait University,Kuwait, Kuwait

for both SO2 and particulate matters (PM) emissionrates are calculated accurately using material balancefor a year-long period considering seasonal variationsin the operation of the process unit in one of the mainrefining industry in Kuwait (Yateem et al. 2010).These results reflect the variation of sulfur infeedstock that comes from various refinery units.SO2 and PM emission inventories are completed andused in dispersion model to assess their impact on theimmediate surroundings of the refinery.

The most advanced dispersion model Aermod hasbeen selected for prediction ground level concentra-tion of SO2 and PM based on comprehensive year-long emission inventories of FCC unit.

Aermod is a dispersion model that uses Gaussiandistribution for the stable conditions and non-Gaussian probabilities density function for the unsta-ble conditions. Aermod has two preprocessors;Aermet that provides planetary boundary layerparameters over a high altitude to yield accuratepredicted concentration values for a given meteoro-logical conditions. It can accommodate large meteo-rological data (multiple years). Aermap generatesregular receptors over a given terrain for the evalua-tion of pollutants ground level concentrations.

The meteorological data for year 2008 are obtainedand are used in preprocessor Aermet to generateplanetary boundary layer parameters. These generateddata are used in Aermod for fixed emission rate toassess the influence of prevailing meteorologicalconditions at this particular site. Aermod has beenused for actual year-long inventories to predict groundlevel concentrations and validate the model bycomparing the results against the recorded valuesfrom Kuwait Environmental Public Authority (K-EPA) monitoring stations.

2 Background

Heavy fractions from different refining units arecracked in FCC unit to useful products, gener-ating SO2 and PM emissions. SO2 emissioninventory is prepared from elemental sulphurbalance over the unit, Yateem et al. (2010).

Whitcombe et al. (2003) showed the formation offines in a fluidized catalytic cracker unit due tocatalyst attrition and fracture as a major source of

catalyst loss. The petroleum industry employs fluidcatalytic cracker units (FCCUs) as the major tool toproduce gasoline from crude oil. At the center of thisunit is regenerator which is used to burn coke fromthe surface of the spend catalyst. As the regenerationprocess is very turbulent, a large amount of catalystmaterial is discharge to the atmosphere. In addition tothe fine particles present in the catalyst, the turbulentconditions inside the FCC alter the particle sizedistribution of the catalyst generating fine particlesand significant amount of aerosols, which has beenidentified in the stack emission of FCCUs.

Caputo et al. (2003) conducted an inter-comparisonbetween Gaussian, Gaussian-segmented plumes, andLagrangian codes. Gaseous emissions are simulatedunder real meteorological conditions for dispersionmodels Aermod, HPDM, PCCOSYMA, and HYS-PLIT. The Aermod and HPDM meteorological pre-processors results are analyzed, and the maindifferences found are in the sensible heat flux (SHTF)and u* (friction velocity) computation, which havedirect effect on the Monin–Obukov length and mixingheight calculation. Gaussian models (Aermod, HPDM)computed the dispersion parameters by using thesimilarity relationships, whereas Gaussian-segmentedmodel (PCCOSYMA) used P–G stability class toevaluate these parameters. Lagrangian transport model(HYSPLIT) advected the puff and calculated its growthrate with local mixing coefficients. Meteorologicalparameters have great effect on the performance ofair dispersion models. Therefore, Aermod and HPDMhave developed effective and sophisticated meteoro-logical parameters preprocessors. It is noticed thatHPDM computed the most stable condition and thelowest mixing height. The comparison also showed asignificant discrepancy between HPDM and otherGaussian models. The maximum ground level concen-tration predicted by Aermod, HPDM, and PCCO-SYMA are similar.

Rama Krishna et al. (2004) examined the assimi-lative capacity and the dispersion of pollutantsresulted from various industrial sources in theVisakhapatnam bowl area, which is situated in coastalAndhra Pradesh, India. Two different air dispersionmodels (Gaussian plume model, GPM, and ISCST-3)are used to predict ground level concentrations ofsulfur dioxide and oxides of nitrogen and assimilativecapacity of the Visakhapatnam bowl area’s atmo-sphere for two seasons, namely, summer and winter.

38 Water Air Soil Pollut (2011) 218:37–47

The computed 8-h-averaged concentrations of the twopollutants obtained from the GPM and ISCST-3 arecompared with those monitored concentrations atdifferent receptors in both seasons and the validationcarried out through Q–Q plots. Both model outputsshowed similar trend with the observed values from themonitoring stations. The GPM output showed over-prediction, whereas the ISCST-3 showed underpredic-tion in comparison with the observed concentrations.Terrain features and land/sea breeze influences are notconsidered in this study, which strongly affected themodels outputs.

Venkatram et al. (2004) evaluated dispersionmodels for estimating ground level concentrations inthe vicinity of emission sources in the urban area ofuniversity of California, Riverside. Aermod-PRIMEand ISC-PRIME dispersion models are used to predictSF6 at different receptors, where SF6 is used as tracerin a simulated non-buoyant release from a smallsource in urban area. Both models output arecompared with hourly observed concentrations. Thecomparison showed that both models overestimate thehighest concentrations, whereas lower range of con-centrations is underestimated. It is concluded thatAermod can predict reliable concentrations if turbu-lent velocity measurements are used to estimateplume dispersion.

Lopez and Mandujano (2005) assessed the impactof natural gas and fuel oil consumption on the airquality in an Industrial Corridor, Mexico to determinethe optimal NG and fuel oil required to reduce SO2

concentration. Air dispersion model Aermod is usedto compute ground level concentration of SO2. Modeloutput is then validated against SO2 field measure-ments. Different hypothetical emission scenarios areperformed to examine the impact of NG and fuel oilmixture. The obtained results in this work indicatethat dispersion model Aermod presented good corre-lation with the measured concentrations. It is alsoconcluded that increasing 40% of NG consumptionwill reduce SO2 concentration by 90%.

Kesarkar et al. (2007) studied the spatial variationof PM10 concentration from various sources overPune, India. Gaussian air pollutant dispersion modelAermod is used to predict the concentration of PM10.Weather research and forecasting model is used tofurnish Aermod with planetary boundary layer andsurface layer parameters required for simulation.Emission inventory has been developed and field-

monitoring campaign is conducted under Pune airquality management program of the ministry ofEnvironment and Forests. This inventory is used inAermod to predict PM10. A comparison betweensimulated and observed PM10 concentration showedthat the model underestimated the PM10 concentrationover Pune. However, this work is conducted over ashort period of time, which is not sufficient toconclude on adequacy of regionally averaged meteo-rological parameters for driving Gaussian modelssuch as Aermod.

Isakov et al. (2007) examined the usefulness ofprognostic models output for meteorological observa-tions. These model outputs are used for dispersionapplications to construct model inputs. Dispersionmodel Aermod is used to simulated observed tracerconcentrations from Tracer Field Study conducted inWilmington, California in 2004. Different meteoro-logical observation sources are used, i.e., onsitemeasurements, National Weather Services (NWS),forecast model output from ETA model, and readilyavailable and more spatially resolved forecast modelfrom MM5 prognostic model. It is noted that MM5with higher grid resolution than ETA performed betterin describing sea breeze related to flow patternsobserved and provided adequate estimates of maxi-mum mixed layer heights observed at the site. It isconcluded that MM5 and ETA prognostic modelsprovided reliable meteorological inputs for dispersionmodels such as Aermod because wind directionestimates from forecast models are not reliable incoastal areas and complex terrain. Therefore, com-prehensive prognostic meteorological models canreplace onsite observations or NWS observations.

Abdul Wahab et al. (2002) studied the impact ofSO2 emissions from a petroleum refinery on theambient air quality in Mina Al-Fahal, Oman. Disper-sion model ISCST is used to predict SO2 ground levelconcentration. The study is performed over a periodof 21 days. Computed SO2 concentrations arecompared with the measured values of SO2 formaximum hourly average concentration, maximumdaily concentration, and total period average con-centration. It is noted that the model output under-predicted the SO2 concentration for all the threecases due to unavailability of background concen-trations and the presence of more dominant sources.Based on the maximum daily average concentrationand the total period maximum concentration, the

Water Air Soil Pollut (2011) 218:37–47 39

model underpredicted the average measured concen-tration by 31.77% and 41.8%, respectively. Themodel performed slightly better based on maximumhourly average concentration and underpredicted by10.5%.

Zou et al. (2010) evaluated the performance ofAermod in predicting SO2 ground level concentrationin Dallas and Ellis counties in Texas as these twocounties are populous and air pollution has been aconcern. Two emission sources are considered in thisstudy, i.e., point sources and on-road mobile sources.Aermet is used to calculate the hourly planetaryboundary layer parameters such as Monin–Obikhovlength, convective scale, temperature scale, mixingheight, and surface heat flux. Dispersion modelAermod is used to simulate SO2 ground levelconcentration at different time scale, i.e., 1, 3, and8 h, daily, monthly, and annually for both countiesseparately. The results are validated with the observedconcentrations. The results showed that Aermodperformed well at the 8 h, daily, monthly, and annualtime scale when combined point, and mobile emissionsources are used in the simulation as model input. It isalso noticed that Aermod is performed much better insimulating the high end of the spectrum of SO2

concentrations at monthly scale than at time scales of1, 3, and 8 h and daily.

Alrashidi et al. (2005) studied the locations of K-EPA monitoring station, which measure SO2 concen-trations emitted from the power stations in the state ofKuwait. The major sources of SO2 emissions inKuwait are from west Doha, east Doha, Shuwaikh,Shuaiba, and Az-Zour power stations. The IndustrialSource Complex Short Term dispersion model is usedto predict SO2 ground level concentrations overresidential areas. Year-long meteorological data areobtained from Kuwait International Airport and usedin the simulation of the dispersion model. Differentdiscrete receptors in the residential areas are selected.It is observed that the weather pattern in Kuwait,specially the prevailing wind direction, has stronginfluence on the ground level concentration of SO2 inthe residential areas located downwind of the botheast and west Doha stations. The comparison betweenthe predicted and the measured concentrations of SO2

from the monitoring stations located at the majorpopulated areas showed that most of these monitoringstations locations are not adequate to measure SO2

concentrations emitted from the power stations.

Therefore, relocation of the monitoring stations ishighly recommended to accurately record the highestground level concentrations of SO2 emitted from thepower stations in Kuwait.

3 Model Application

3.1 Input Data

Aermod dispersion model implementation requiresthree main input data. These are:

1. Source information: This includes pollutant emis-sion rate (g/s), location coordinates in UniversalTransverse Mercator (UTM; m), base elevationfrom the sea level (m), stack height (m), exit stackinner diameter (m), exit stack gas velocity (m/s),and exit stack gas temperature (°K).

2. Meteorological information for the region ofinterest: This includes anemometer height (m),wind speed (m/s), wind direction (flow vectorfrom which the wind is blowing; in degreesclockwise from the north), ambient air temper-ature (°C), stability class at the hour ofmeasurement (dimensionless), and hourly mix-ing height (m).

3. Receptor information: This can be specified orgenerated by the program to predict the pollu-tants’ concentrations at the selected receptors.

The entire required source input data areobtained from FCC unit in the refinery. A stackof 80 m height, an inner diameter of 2.3 m, with anaverage exit gas velocity of 20 m/s and exit gastemperature of 550°K are fed into the model.Monthly emission variation is considered with totalSO2 emission rate of 6,089.2 g/s and total PMemission rate of 302 g/s as presented in detail(Yateem et al. 2010).

3.2 Area of Study

The area of study in this work covers portion ofAhmadi governorate in the state of Kuwait. Fahaheelarea is adjacent to the petroleum refinery has one ofthe Kuwait EPA air quality monitoring station locatedat a polyclinic. Both areas Fahaheel and Ahmadi aresurrounded by arid desert in the west side andbordered by the Gulf from the east.

40 Water Air Soil Pollut (2011) 218:37–47

Two different types of receptor coordinates areused as input to the Aermod model to predict theground level concentration of SO2, these are:

1. Discrete Cartesian receptors specified at thesensitive areas viz., a school, a shopping area,and EPA monitoring stations in Fahaheel. Ahospital and petroleum services companies’ offi-ces are selected in Ahmadi.

2. Uniform Cartesian grid receptors covering theentire area of study where the FCC stack(emissions source) is located almost in the centerof the mesh grid.

The receptors selected are based on the actual sitesin a UTM location coordinate of the area of interestmap. Table 1 shows the selected discrete receptorsinformation.

The uniform grid receptors of a total 1,764 (42×42) were divided into (Δx=300 m×Δy=250 m) tocover about 12×10 km area of study. The optimumselection of the mesh size is based on the computa-tional accuracy and time.

4 Results and Discussion

A year-long comprehensive meteorological data areprocessed by Aermet to generate boundary layerparameters and to pass all meteorological observa-tions to Aermod.

Figure 1 shows wind direction and magnitude for aperiod of year 2008. It is observed that most of thetime, the prevailing wind direction is from north west.There is strong influence from the neighboring Gulfas the refinery is located at the coast, resulting intostrong sea breeze blowing from east direction. Windclass frequency distribution for the entire year con-firming 2% calm conditions, while 39.8% is between3.6 and 5.7 m/s. The highest wind class 8.8–11.1 m/sis less than 1%.

A model run is performed for actual monthlyemission variation with total annual SO2 emissionrate of 6,089.2 g/s and total PM emission rate of302 g/s independently. Monthly emission factors forboth pollutants are tabulated in Tables 2 and 3,respectively.

A discrete receptor is selected at Kuwait Environ-mental Public Authority monitoring station located atpolyclinic in Fahaheel area. Concentrations of SO2,NOx, H2S, O3, CO, CO2, methane, non-methanehydrocarbon, benzene, toluene, xylenes, ethylben-zene, total suspended particulates, and meteorologicalparameters are continuously recorded on hourly basis.

Hourly predicted ground level concentrations atspecified discrete receptor showed large scatter due tovariation in meteorological conditions and therecorded values influenced by the contribution ofvarious emission sources, resulting into specificbackground concentration that has made the compar-ison impracticable. There is large fluctuation in thebackground concentration, which is difficult to quan-tify. Hence, zero background concentration has beenassumed to resolve this uncertainty. Therefore, dailyaverage measured concentrations of SO2 were com-pared with the daily predicted concentrations tovalidate the model output.

Figure 2 shows the plot between the measured top20 daily average values vs. the daily predicted top 20values at the discrete receptor, Kuwait EPA monitor-ing station.

The slope is equal to 0.72, reflecting highmeasured values compared to predicted values,depicting the contribution of other emission sources.The correlation coefficient is equal to 0.91 reflectingan acceptable validation of the model output withmeasured average daily SO2 concentrations.

The predicted hourly average ground level con-centrations of SO2 are compared with Kuwait EPAAmbient Air Quality Standards at all of the selectedreceptors.

ID number Discrete receptor identity X coordinate Y coordinate

1 Fahaheel Polyclinic 219,854.25 3,219,765.79

2 Petroleum Services Offices in Ahmadi 216,666.87 3,220,105.63

3 School in Fahaheel 220,300.00 3,219,820.85

4 Ahmadi Hospital 213,458.86 3,221,523.64

5 Shopping area in Fahaheel 219,274.32 3,219,554.21

Table 1 The selecteddiscrete receptorsinformation

Water Air Soil Pollut (2011) 218:37–47 41

The maximum allowable level for the hourlyaverage concentration of SO2, specified by KuwaitEPA, is 444 μg/m3. Figure 3 shows the isopleths ofthe predicted hourly average ground level concentra-tion of SO2 calculated at the selected uniform gridreceptors.

The isopleths indicate the predicted spatial varia-tions of the ground level concentrations of SO2. Themaximum predicted hourly average ground levelconcentration of SO2 in the vicinity of the refineryexceeded by as much as 300 μg/m3. The highestpredicted concentration is equal to 769 μg/m3, ob-served on the 8 March 2008 at 0800 hours and about1.713 km in the NW direction from the FCC stack, andnot far from the Fahaheel and Ahmadi areas at thereceptor coordinates of X=218,557.94, Y=3,219,169.This high value of the predicted SO2 concentration isexpected due to the elevated SO2 emission rate, whichresulted from the high sulfur content in the FCCfeedstock and other operational conditions and theprevailing meteorological conditions (temperature,humidity, wind speed, wind direction, stability class,and planetary boundary layer).

A thorough inspection on Fig. 3 indicates thatpredicted concentrations of SO2 exceed the allowablehourly limit at 5.3% of the study area from north westand south west direction from the stack.

Similarly, the predicted daily average groundlevel concentration of SO2 is compared with KuwaitEPA ambient air quality standards at all receptors.The allowable level for the daily average concentra-tion of SO2 is 157 μg/m3. Figure 4 shows theisopleths of the predicted daily average ground levelconcentration of SO2 computed at the selecteduniform grid receptors.

The isopleths indicate the daily predicted spatialvariations of the ground level concentrations of SO2

in the area of study. The highest daily predictedconcentration is equal to 335 μg/m3, observed on the9 November 2008 and about 0.75 km in the SEdirection from the stack, at a receptor coordinates ofX=220,357.94, Y=3,217,419 affecting the neighbor-ing Shuaiba industrial area, Kuwait main industrialcomplex. This high value of the daily predicted SO2

concentration is exceeded the allowable level by157 μg/m3 and obviously influenced by the prevailing

Fig. 1 Wind rose for aperiod of year 2008

Table 2 SO2 monthly emission factors

January February March April May June July August September October November December

0.077 0.083 0.096 0.1 0.077 0.088 0.067 0.067 0.088 0.077 0.1 0.075

42 Water Air Soil Pollut (2011) 218:37–47

meteorological conditions, especially the predominantnorth west wind and other meteorological factors.

Discrete receptor 2, located at Petroleum servicesoffices, has shown the highest SO2 hourly concentra-tion equal to 544 μg/m3 on 27 February at 0800hours. The hourly exceedance is occurred four timesat this location throughout the study period. Thehighest daily concentration at the same receptor isequal to 39 μg/m3 on 8 March.

Discrete receptor 3, located at school, hasshown the highest SO2 hourly concentration equalto 279 μg/m3 on 2 March at 0400 hours. Thisconcentration is below the Kuwait EPA hourly stand-ards. The daily highest concentration is equal to57 μg/m3 on 2 March.

Discrete receptor 4, located at Ahmadi hospital,has shown the highest SO2 hourly ground levelconcentration equal to 288 μg/m3 on 27 February at0800 hours. This value is also below the specifiedhourly limit set by Kuwait EPA. The daily predictedconcentration is equal to 23 μg/m3 on 30 April.

Discrete receptor 5, located at shopping area, hasshown the highest SO2 hourly ground level concen-tration is equal to 336 μg/m3 on 23 October at 0800hours. The daily predicted concentration is equal to45 μg/m3 on 22 April. Both hourly and dailypredicted values are below Kuwait EPA hourly anddaily ambient air quality standards.

Kulkarni et al. (2009) have reported that lanthanumand lanthanides are used as markers for particulatematters pollution as PM2.5 in petroleum refineries,mainly from FCC units.

US EPA daily PM2.5 standard is 35 μg/m3. In thepresent work, the application of Aermod to predictground level concentration of PM is considered asPM2.5 for rare earth metals, i.e., lanthanum andcerium. PM2.5 is inhalable and has adverse impacton public health causing cardiovascular diseases.Kuwait EPA has no standard for PM2.5 and has onlyspecified daily and yearly standard for PM10. Figure 5shows the isopleths of the predicted hourly averageground level concentration of PM calculated at theselected uniform grid receptors.

The isopleths indicate the hourly predicted spatialvariations of the ground level concentrations of PM.The maximum hourly predicted average ground levelconcentration of PM is equal to 45 μg/m3, observed onthe 27 February 2008 at 0800 hours and about 1.56 kmin the NW direction from the FCC stack, and atreceptor coordinates of X=218,557.94, Y=3,218,919.

Similarly, the predicted daily average ground levelconcentration of PM is compared with US EPAambient air quality standards for PM2.5 at all receptors.Figure 6 shows the isopleths of the predicted dailyaverage ground level concentration of PM computed atthe selected uniform grid receptors.

The isopleths indicate the daily average predictedspatial variations of the ground level concentrations ofPM in the area of study. The highest daily predictedconcentration is equal to 16 μg/m3, observed on the29 December 2008 and about 0.75 km in the SEdirection from the stack, at a receptor coordinates ofX=220,657.94, Y=3,217,419 due to the influence ofthe prevailing meteorological conditions, especiallythe predominant north west wind and other meteoro-logical factors.

To observe the computational model sensitivity,another scenario run is performed adding two finermeshes consisting of 21×21 uniform receptor points,the first one covering hourly highest ground levelconcentration area, the other one covering daily

Table 3 PM monthly emission factors

January February March April May June July August September October November December

0.093 0.097 0.091 0.079 0.079 0.083 0.064 0.063 0.085 0.079 0.079 0.1

Fig. 2 Daily predicted SO2 concentrations vs. measured SO2

concentrations

Water Air Soil Pollut (2011) 218:37–47 43

Fig. 3 Isopleths plot of the predicted hourly average ground level concentration of SO2 (Google Inc. 2010)

Fig. 4 Isopleths plot of the predicted daily average ground level concentration of SO2 (Google Inc. 2010)

44 Water Air Soil Pollut (2011) 218:37–47

Fig. 5 Isopleths plot of the predicted hourly average ground level concentration of PM (Google Inc. 2010)

Fig. 6 Isopleths plot of the predicted daily average ground level concentration of PM (Google Inc. 2010)

Water Air Soil Pollut (2011) 218:37–47 45

highest predicted ground level concentration area.The output accuracy has improved for both pollutantsdue to application of interpolation using small valuesof Δx=150 m, Δy=110 m for the first mesh and Δx=100 m, Δy=100 m for the second mesh. There is0.65% increase in the hourly highest ground levelconcentration and 2.8% increase in the daily highestground level concentration, which are insignificant.Therefore, the only parent mesh is used in thecomputational process for all the other scenariosconsidered in the parametric studies.

FCC stack sensitivity analysis is performed onthree scenarios (stack height, SO2 emission rate, andstack diameter).

In scenario 1, analysis for stack heights 50, 80,120, 160, and 200 m is conducted while keeping theemission rate, exit flue gas velocity, exit temperature,and stack diameter constant.

The influence of stack height is shown in Fig. 7. Itis obvious from the figure that the highest predictedhourly and daily ground level concentrations of SO2

are reduced substantially as stack height is increased.The reduction in the highest computed hourly groundlevel concentration of SO2 is almost 50% when stackheight is doubled. The decrease in evaluated hourlySO2 concentration as a function of stack height isgiven as an exponential expression C mg=m3ð Þ ¼1; 600:7e�9:071�10�3h and r2 is 0.999, where h is thestack height (m). The hourly gradient dC=dh ¼14:52e�9:071�10�3h becomes insignificant at higherstack elevations. The highest daily predicted groundlevel concentration as a function of stack height isgiven as C mg=m3ð Þ ¼ 1; 409:8e�1:732�10�2h and r2 is0.984. The daily highest predicted concentration

gradient is dC=dh ¼ 24:42e�1:732�10�2h. The locationsof hourly highest predicted concentrations of SO2

from the stack, as a function of stack height is shownin Fig. 7 and related as D kmð Þ ¼ 0:597e1:16�10�2h andr2 is 0. 978.

In scenario 2, SO2 emission rate effect from FCCstack is tested at stack height of 80 m for differenttotal monthly emission rates of 3,000, 4,000, 5,000,6,000, 7,000, and 8000 g/s, taking into considerationthe monthly emission variations (by using emissionfactors, Table 2) and fixing other stack parameters,i.e., exit temperature, exit flue gas velocity, and stackdiameter.

It is noticed from Fig. 8 that the highest predictedhourly and daily ground level concentrations of SO2

is substantially decreased as SO2 emission rate isreduced. At 50% reduction in the emission rate, thehighest hourly and daily ground level concentrationsdecreased by 50%.

In scenario 3, FCC stack diameter effect is examinedat stack height of 80 m for different diameters of 1.5,2.3, 3, and 4 m. The exit flue gas velocity is alsochanged as directly related to the square of the diameterfor a fixed exit flue gas flow rate. It is observed that thedispersion and rise of the plume are not affected bydiameter variation and the predicted ground levelconcentration of SO2 remained almost unaltered. Thehourly and daily predicted concentrations of SO2 arealmost identical for all the cases.

5 Conclusion

FCC unit in a refinery is the major contributor of SO2,and PM emissions are those responsible for adverse

Fig. 7 Stack height vs. hourly and daily predicted ground levelconcentrations of SO2

Fig. 8 SO2 emission rate vs. hourly and daily predicted SO2

ground level concentrations

46 Water Air Soil Pollut (2011) 218:37–47

impact on the immediate neighborhood of therefinery. A complete comprehensive emission inven-tories for a year-long period have been prepared forboth SO2 and particulate matters.

A model run performed for actual monthly emissionvariation with total SO2 emission rate of 6,089.2 g/sand total PM emission rate of 302 g/s independently,taking into consideration monthly emission factors forboth pollutants. The daily predicted ground levelconcentrations of SO2 are compared with KuwaitEPA monitoring station daily measured SO2 concen-trations at the same discrete receptor and showedacceptable validation of the model output.

The highest hourly predicted concentration of SO2

is equal to 769 μg/m3. It is observed on the 8 March2008 at 0800 hours, due to elevated SO2 emission ratein this month and the prevailing meteorologicalconditions, especially sea breeze effect in the earlymorning hours. The highest daily predicted concen-tration is equal to 335 μg/m3. It is observed on the 9November 2008 and obviously influenced by thepredominant north west wind and high SO2 emissionrate in the month of November.

The maximum hourly predicted average groundlevel concentration of PM is equal to 45 μg/m3. It isobserved on the 27 February 2008 at 0800 hours. Thehighest daily predicted concentration is equal to16 μg/m3, observed on the 29 December 2008.

The stack sensitivity is explored by changing stackheight, total emission rate, and stack diameterindependently. It is observed that the higher stackfacilitated good dispersion, thus lowering the groundlevel average concentration of the pollutant up to 50%when the stack height doubled.

It is notice that the highest predicted hourly anddaily ground level concentrations of SO2 are substan-tially decreased as SO2 emission rate is reduced. At50% reduction in the emission rate, the highest hourlyand daily ground level concentrations decreased byalmost 48%.

The influence of stack diameter inherentlychanged the exit flue gas velocity due to invariableflue gas flow rate. The plume rise and dispersion arerelated to the exit flue gas velocity, which decreasedwith the increase of stack diameter because of

proportionality to the square of diameter. For a fixedload, there is no noticeable change in the averagehourly and daily predicted ground level concentra-tions of SO2.

References

Abdul Wahab, S. A., Al-Alawi, S. M., & El-Zawahri, A.(2002). Patterns of SO2 emissions: A refinery case study.Environmental Modeling and Software, 17, 563–570.

Alrashidi, M. S., Nassehi, V., & Wakeman, R. J. (2005).Investigation of the efficiency of the existing air pollutionmonitoring sites in the state of Kuwait. EnvironmentalPollution, 138, 219–229.

Caputo, M., Gimenez, M., & Schlamp, M. (2003). Inter-comparison of atmospheric dispersion models. Atmospher-ic Environment, 37, 2435–2449.

Google Inc. (2010). Google earth. Version 5.1, http://www.earth.google.com.

Isakov, V., Venkatram, A., Touma, S. J., Koracin, D., & Otte, L.T. (2007). Evaluating the use of outputs from comprehen-sive meteorological models in air quality modelingapplications. Atmospheric Environment, 41, 1689–1705.

Kesarkar, A. P., Dalvi, M., Kaginalkar, A., & Ojha, A. (2007).Coupling of the weather research and forecasting modelwith AERMOD for pollutant dispersion modeling. A casestudy for PM10 dispersion over Pune, India. AtmosphericEnvironment, 41, 1976–1988.

Kulkarni, P., Chellam, S., & Fraser, M. P. (2009). Trackingpetroleum refinery emission events using lanthanum andlanthanides as elemental markers for PM2.5. Environmen-tal Science & Technology, 43(8), 2990–2991.

Lopez, J. L., & Mandujano, C. (2005). Estimation of the impactin the air quality by the use of clean fuels (fuel oil versusnatural gas). Catalysis Today, 106, 176–179.

Rama Krishna, T. V. B. P. S., Reddy, M. K., Reddy, R. C., &Singh, R. N. (2004). Assimilative capacity and disper-sion of pollutants due to industrial sources in Visakha-patnam bowl area. Atmospheric Environment, 38, 6775–6787.

Venkatram, A., Isakov, V., Yuana, J., & Pankratza, D. (2004).Modeling dispersion at distances of meters from urbansources. Atmospheric Environment, 38, 4633–4641.

Whitcombe, J. M., Agranovski, I. E., & Braddock, R. D.(2003). Attrition due to mixing of hot and cold FCCcatalyst particles. Powder Technology, 137, 120–130.

Yateem, W., Nassehi, V., & Khan, A. R. (2010). Inventories ofSO2 and PM emissions from fluid catalytic cracking(FCC) units in petroleum refineries. Water, Air, and SoilPollution. doi:10.1007/s11270-010-0423-z.

Zou, B., Zhan, F. B., Wilson, J. G., & Zeng, Y. (2010).Performance of aermod at different time scales. SimulationModeling Practices and Theory, 18, 612–623.

Water Air Soil Pollut (2011) 218:37–47 47