202

The open-channel method allows the measurement ofstream denitrification under natural conditions withoutexpensive 15N-tracers with greater ease than traditional totalnitrogen mass-balance techniques and without limitationsassociated with incubations and inhibitor-based methods(Laursen and Seitzinger 2002 Groffman et al 2006 Table 1)The open-channel method applies a simple N2 mass balanceN2 is gained due to denitrification and groundwater inputsand lost (or gained) via air-water gas exchange Nitrogen fixa-tion anaerobic ammonium oxidation (anammox) N2 loss to

groundwater and nitrous oxide (N2O) production via denitrifi-cation are typically assumed to be negligible Similar mea-surements based on excess N2 have been applied in ground-water riparian zones lakes and oceans (Groffman et al 2006)

The open-channel method has been applied across a rela-tively broad suite of streams and rivers (Web Appendix 1mdashTable 1) and modifications to the method have allowed appli-cation in deep rivers and estuaries (Yan et al 2004 Kana et al2006) Open-channel estimates have been compared with totalnitrogen mass balance-based estimates of denitrification yield-ing broadly similar results for the South Platte and ConnecticutRivers (Pribyl et al 2005 Smith et al 2008) If streams have lowgas exchange coefficients and well-characterized groundwaterN2 inputs the open-channel method may be preferred overmass balances due to easier implementation and reduced uncer-tainty (Pribyl et al 2005) although the error associated with themethod is sometimes high (Web Appendix 1mdashTable 1)

The open-channel method is analogous to oxygen mass bal-ances used to determine rates of aquatic photosynthesis and res-

Revisiting the application of open-channel estimates ofdenitrificationH M Baulch1 J J Venkiteswaran2 P J Dillon3 and R Maranger 4

1Environmental and Life Sciences Graduate Program Trent University 1600 West Bank Drive Peterborough Ontario K9J 7B82Department of Earth and Environmental Sciences University of Waterloo 200 University Avenue West Waterloo OntarioN2L 3G13Department of Environment and Resources Studies Trent University 1600 West Bank Drive Peterborough Ontario K9J 7B84Deacutepartement de sciences biologiques Universiteacute de Montreacuteal Pavillon Marie-Victorin 90 Vincent drsquoIndy Montreacuteal QueacutebecH3C 3J7

AbstractDevelopment of an open-channel method for measurement of denitrification without the use of expensive

isotopic tracers has generated considerable interest among researchers attempting to quantify N loss from loticsystems Membrane inlet mass spectrometry allows measurement of small changes in N2 concentrations facili-tating calculation of whole reach denitrification rates using an N2 mass balance corrected for gas exchange Themethod has been applied successfully within numerous rivers ranging widely in size and denitrification ratePrevious model-based analyses suggest that the method can be applied in a broader suite of ecosystems andspecifically that it is well suited to shallow streams where denitrification rates as low as 30-100 micromol N mndash2 hndash1

may be measurable This coupled with increasing availability of necessary equipment relatively low cost ofmeasurements and the ability to measure denitrification at environmentally relevant spatial scales suggests thatbroad adoption of the method is likely In this paper we revisit this model-based analysis using alternate mod-els of gas exchange and demonstrate that benthic turbulence-induced gas exchange will restrict the suite of suit-able study streams Specifically we note that within shallow streams and fast-flowing systems denitrificationmay be measurable only at moderate or high rates To help facilitate further application of the method weextend our discussion beyond site selection to discuss assumptions of the open-channel method options forestimating the error in denitrification rates and recommended practices for future studies

Corresponding author E-mail helenbaulchtrentuca

AcknowledgmentsThis work was funded by a NSERC Canada Graduate Scholarship to

HMB Ontario Graduate Scholarship to JJV and an NSERC Discoverygrant to PJD This work benefited from discussions at a DenitrificationResearch Coordination Network workshop and comments of anony-mous reviewers

DOI 104319lom20108202

Limnol Oceanogr Methods 8 2010 202ndash215copy 2010 by the American Society of Limnology and Oceanography Inc

LIMNOLOGYand

OCEANOGRAPHY METHODS

Baulch et al Denitrification Open-channel method

203

piration and as in oxygen mass balances two differentapproachesmay be applied A two-stationmethod follows a singleparcel of water through time and space and to date has been themore common approach in open-channel denitrification mea-surements Many studies extend beyond two stations into three ormore stations with time of travel and gas transfer known betweeneach series of stations This allows better spatial representation ofdenitrification rates and could also be used to assess error associ-ated with the method at broader spacing of study stations Theone station or diel approach (eg McCutchan et al 2003)employs a time for space substitutionWhile themethods are con-

ceptually similar andmathematically equivalent the assumptionsinherent in each differ (see ldquoAssessment and Discussionrdquo)

The most significant limitation of the open-channelmethod relates to detection limits and uncertainty of themethod Whereas N2 concentrations can be measured withhigh precision (coefficient of variation 05 for N2 005 forN2Ar) using membrane-inlet mass spectrometry (MIMS Kanaet al 1994) our ability to measure denitrification rates usingthe open-channel method depends not only on instrumentalprecision and accuracy but also on rates of N2 loss and gainMethod uncertainty is dependent on a suite of parameters

Table 1 Major methods of measuring denitrification in streams

Stream or reach-scale measurements

Major advantages These measurements are typically performed under in-stream conditions at relatively large spatial scales with direct relevance tobroad scale N modelsMajor limitations and assumptions Cannot be directly replicated therefore uncertainty estimates must be model-based Tend to obscure importance ofmicroscale processes

Open-channel N2 accumulation Method and general principle Accumulation of N2 is measured over time and model-based analyses are performed topartition biological from physical processes and estimate denitrification ratesMajor advantages Where equipment is available costs are relatively low Environmental perturbation is minimalMajor limitations and assumptions Suite of suitable streams is restricted Requires detailed modeling and carefulmeasurements to partition physical and biological controls on N2 accumulation Measures ldquonet denitrificationrdquoOptimal conditions See Web Appendix 1mdashTable 2

Natural abundance isotopicanalysis of nitrate

Method and general principle Denitrification leads to 15N and 18O enrichment of the residual NO3 pool Measurementsdepend upon concurrent measurements of nitrate concentration and δ15N as well as known fractionations (Kellmanand Hillaire-Marcel 1998)Major advantages Represents in situ conditions (no environmental perturbations)Major limitations and assumptions High analytical costs Careful characterization of all nitrate sources is requiredPresence of multiple nitrate sources may confound results Oxygen-exchange reactions may complicate interpretationof δ18O-NO3 Ineffective if denitrification proceeds to completion Magnitude of fractionation may varyOptimal conditions Low groundwater inputs low lateral inflows Residual NO3 pool (denitrification does not proceedto completion)

Mass balance Method and general principle Denitrification is measured as the difference between N inputs and outputs (Sjodin et al1997)Major advantages Straightforward widely used Analytical equipment widely availableMajor limitations and assumptions Unmeasured N inputs (N fixation groundwater) and outputs (seasonal N uptakeby autotrophs) transformations (mineralization nitrification) and changes in storage will affect N balance andestimation of denitrification ratesOptimal conditions Systems where all parameters can be precisely constrained (see Groffman et al 2006 for warningsregarding application of this method)

Whole-stream addition of15N-enriched nitrate

Method and general principle 15N-enriched NO3 is added to a stream Denitrification is calculated using measurementsof 15N-enriched N2 corrected for gas exchangeMajor advantages Allows denitrification measurement with minimal environmental perturbation Even relatively lowrates of denitrification can be measuredMajor limitations amp assumptions High cost of tracer and analyses Excludes denitrification that is coupled tomineralization and nitrification Surface loading does not allow mixing into deeper sediment layers and mayunderestimate denitrification (Mulholland et al 2004) Nitrate addition may stimulate denitrificationOptimal conditions High cost of tracers restricts this approach to relatively low-flow systems with low-moderatebackground nitrate concentrations

Baulch et al Denitrification Open-channel method

204

including gas exchange rates (and scaling of these rates to thetemperatures and gases of interest) gas solubility (thereforewater temperature pressure and uncertainty in solubilityequations) reach depth N2 concentrations in stream waterand groundwater and rates of groundwater gain and loss (seeMcCutchan et al 2003 Laursen and Seitzinger 2005 Smith etal 2008) The ratio of areal denitrification rates to overlyingwater volume affects the maximum rate of increase in N2 con-centrations The rate of air-water gas exchange dictates howquickly excess N2 is lost to the atmosphere Groundwaterinputs and outputs nitrogen fixation ebullition diffusionmixing and temperature also affect N2 accumulation andmeasurement of denitrification rates

Laursen and Seitzinger (2005) performed amodel-based analy-sis of factors affecting method sensitivity and detection limits

Air-water gas exchange was estimated using a wind-based modelof gas exchange illustrating that the method is most sensitive atlow wind speeds when gas exchange rates are low They foundthat stream depth was a critical factor affecting the detectionlimit At constant denitrification rates N2 accumulates more rap-idly within shallow streams contributing to an ability tomeasurelower rates of denitrification Temperature also affected detectionlimits via effects on solubility and rates of gas transfer This analy-sis focused on systems with wind-driven gas exchange Howeverwithin many fluvial systems and particularly in shallow systemsfriction caused by water flow over the bottom substrate is a majorcause of turbulent mixing and an important control on rates ofgas transfer (Raymond and Cole 2001)

In the first part of this article we revisit site-selection crite-ria for application of the open-channel method Based on the

Table 1 Continued

Patch-scale measurements in chambersMajor advantages These measurements are typically performed in flow-through chambers or static chambers of varying size They allow relatively easyexperimental manipulation can be highly replicated and allow isolated study of communities of interestMajor limitations and assumptions Scaling to reach or ecosystem levels can incorporate significant error (see Kellman 2004) Incubations can lead tochanges in oxygen carbon and nitrogen concentrations flow patterns physical structure of substratum boundary layer thickness and microbialpopulations which may induce changes in denitrification ratesOptimal conditions Broadly applicable across most or all conditions

N2 or N2Ar Method and general principle Production of N2 is measured in chambers with N2 accumulation representing ldquonetdenitrificationrdquo (Denitrification minus nitrogen fixation) N2 accumulation is often modeled based on measurementsof N2Ar and N2 is calculated assuming Ar is at saturationMajor advantages InexpensiveMajor limitations amp assumptions Can require long periods of incubation Necessary equipment is only beginning tobecome widely available Measures net rather than gross denitrification Water temperature must be carefullycontrolled to prevent misinterpretation of physical gas flux as denitrification or N2 fixationOptimal conditions Production of bubbles is extremely problematic As a result low rates of benthic ebullition andlow rates of photosynthetic O2 production are necessary

Acetylene inhibition Method and general principle Acetylene inhibits the reduction of N2O to N2 in the final step of denitrificationDenitrification is measured as the accumulation of N2OMajor advantages Inexpensive straightforward and widely used in the pastMajor limitations amp assumptions Numerous analytical problems have been identified including incomplete inhibitionof the N2O-N2 step of denitrification (eg due to slow diffusion sulphide interferences) inhibition of nitrification(leading to underestimation of denitrification in systems with highly coupled nitrification-denitrification) andstimulation of denitrification via addition of a carbon substrateOptimal conditions Can be applied in a broad suite of systems

15N labeling Method and general principle 15N labeled NO3 is added to a chamber and production of 15N labeled gases ismeasured (Nielsen 1992)Major advantages Can measure gross denitrification (excludes N2 fixation) and importance of water column NO3

versus nitrification derived NO3 as the substrate for denitrificationMajor limitations and assumptions Nitrate pool must be evenly labeled Nitrate addition can stimulate denitrificationThere is potential for significant error in systems where nitrification and denitrification are tightly coupled or whereanammox is important (Thamdrup and Dalsgaard 2002)Assumptions Complete label mixingOptimal conditions Homogenous substrates where assumption of complete label mixing is a reasonable assumptionSee warnings in Groffman et al (2006) regarding application in complex aquatic systems

Baulch et al Denitrification Open-channel method

205

expectation that bottom turbulence is an important determi-nant of gas exchange in many small streams we applied alter-nate models of water-air gas exchange and predicted that bothstream depth and stream velocity would be important controlsupon rates of gas exchange N2 accumulation and as a resultthe utility of the open-channel method We anticipated thatwhereas dilution effects would be lower in shallow streamsgas exchange coefficients were likely to be higher and shallowstreams might not constitute ideal systems for application ofthis method Because the open-channel method is extremelysensitive to rates of gas transfer and existing models of water-air gas transfer incorporate significant error measurements ofdenitrification require direct concurrent measurements of gasexchange rates However our model analysis in conjunctionwith the Laursen and Seitzinger (2005) analysis will helpresearchers identify candidate streams for application of theopen-channel method to measure denitrification This is par-ticularly important at preliminary stages of project planningand in the grant application and review process

In the second part of this paper best practices for imple-mentation of the method are reviewed We discuss assump-tions of both the one-station and two-station approaches andthe major sources of uncertainty in open-channel mea-surements of denitrification We review considerations forfield and laboratory measurements and subsequent modelingof uncertainty to help guide researchers in the use of the open-channel method Finally we discuss alternative methods formeasuring denitrification in systems where open-channelmethods may not be appropriate or where comparisonsamong methods are to be performed

Materials and proceduresA model-based approach was applied to determine the suit-

ability of shallow (le 1 m) streams for application of the open-channel method Briefly we modeled N2 accumulation using asimple N2 mass balance model using Matlab (Matlab R2008b770 The MathWorks) N2 accumulation was modeled within aparcel of water with a 1-min time step according to Eq 1 (seebelow) The mathematical formulation of this model is identi-cal for one-station and two-station approaches N2(t) is the con-centration of dinitrogen gas (micromol N2 L

ndash1) at time t N2sat is thesaturation concentration of gas (micromol N2 Lndash1) Denitrificationis a calculation of the volumetric N2 input resulting from den-itrification (micromol N2 L

ndash1 minndash1) K is the gas transfer coefficientfor N2 (minndash1) The length of the time step (t) is 1 minK modelsmdashSix alternative models of gas transfer coefficient

(K) were incorporated into the N2 model three models based onwind-induced aeration (Wanninkhof 1992 Cole and Caraco1998 Laursen and Seitzinger 2005) and three models wherebenthic turbulence-induced gas exchange is estimated based on

stream depth and velocity (OrsquoConnor and Dobbins 1958Churchill et al 1962 Owens et al 1964 seeWeb Appendix 2 fordetails including model formulae) Because no single modelappears adequate for describing rates of gas transfer across thediverse conditions within streams (Melching and Flores 1999Gualtieri et al 2002) we apply this range of model outputs asan indicator of whether the open-channel method may beapplicable in a single study stream These equations wereselected based on their limited data requirements and resultingusefulness in making first assessments of gas exchange how-ever these are just a few examples of a vast literature describingrates of gas transfer More complex models that incorporate fac-tors such as slope and roughness may provide more accurateresults in some systems If the criterion of Schwarzenbach et al(1993 seeWeb Appendix 2) is applied the majority of our studyscenarios (Fig 1) reflect systems where benthic turbulence islikely to drive gas exchange The only exception is at a depth of1 m and the minimum stream velocity tested of 002 m sndash1

where wind models are expected to be more appropriate (seeWeb Appendix 2) Revisiting method sensitivity is clearly nec-essary as previous analyses (Laursen and Seitzinger 2005) werebased on the assumption of wind-induced gas exchange Weprovide the wind model analysis to highlight differences fromthe previous analysis In subsequent discussion we refer to allgas exchange models by using the name of only the first authorModel-based analysismdashThe N2 model was run for denitrifica-

tion rates ranging from 100 to 2000 micromol N mndash2 hndash1 (in100 micromol N mndash2 hndash1 increments) for each K model under arange of stream depths velocities and wind speeds (Fig 1) Werecorded the time after the model start (at saturation condi-tions) that N2 accumulation due to denitrification could be dis-tinguished reliably from background N2 concentrations that isthe time at which N2 accumulation is greater than or equal toN2 accumulation in the absence of denitrification plus anerror term N2 modeled denitrification = times gt N2 modeled denitrification = 0 +error term

To facilitate comparison with the work of Laursen andSeitzinger (2005) we use an error term of 1 micromol N2 L

ndash1 This fig-ure is based on analytical error of 022 to 041 micromol N2 Lndash1Whereas this figure reflects the precision of the method (iereproducibility of multiple replicates) method accuracy (ie dif-ference from the true value) can be somewhat poorer Analysesof Kana et al (1994) indicate that differences between measuredvalues and expected values across a range of concentrations from326 to 473 micromol N2 L

ndash1 were typically between 024 and 165depending on temperature and whether N2 or N2Ar was con-sidered Accuracy of analyses is likely to vary among labs andover time depending on factors such as instrument stability Forcomparative purposes we reran the samemodel analysis assum-ing a considerably higher error term of 4 micromol N2 L

ndash1

(1)N N + denitrificationumolL

umol N

m

2 12 2

2

( )( )

tt

+

times

=hh

m

L depth m

htimes times times

times

1

1000

1 1

60

3

( ) min

(mint ))( )

min(mindash ( ndash )N Numol

Lumol

Lsat2 21

tK ttimes

times nn)

Baulch et al Denitrification Open-channel method

205

expectation that bottom turbulence is an important determi-nant of gas exchange in many small streams we applied alter-nate models of water-air gas exchange and predicted that bothstream depth and stream velocity would be important controlsupon rates of gas exchange N2 accumulation and as a resultthe utility of the open-channel method We anticipated thatwhereas dilution effects would be lower in shallow streamsgas exchange coefficients were likely to be higher and shallowstreams might not constitute ideal systems for application ofthis method Because the open-channel method is extremelysensitive to rates of gas transfer and existing models of water-air gas transfer incorporate significant error measurements ofdenitrification require direct concurrent measurements of gasexchange rates However our model analysis in conjunctionwith the Laursen and Seitzinger (2005) analysis will helpresearchers identify candidate streams for application of theopen-channel method to measure denitrification This is par-ticularly important at preliminary stages of project planningand in the grant application and review process

In the second part of this paper best practices for imple-mentation of the method are reviewed We discuss assump-tions of both the one-station and two-station approaches andthe major sources of uncertainty in open-channel mea-surements of denitrification We review considerations forfield and laboratory measurements and subsequent modelingof uncertainty to help guide researchers in the use of the open-channel method Finally we discuss alternative methods formeasuring denitrification in systems where open-channelmethods may not be appropriate or where comparisonsamong methods are to be performed

Materials and proceduresA model-based approach was applied to determine the suit-

ability of shallow (le 1 m) streams for application of the open-channel method Briefly we modeled N2 accumulation using asimple N2 mass balance model using Matlab (Matlab R2008b770 The MathWorks) N2 accumulation was modeled within aparcel of water with a 1-min time step according to Eq 1 (seebelow) The mathematical formulation of this model is identi-cal for one-station and two-station approaches N2(t) is the con-centration of dinitrogen gas (micromol N2 L

ndash1) at time t N2sat is thesaturation concentration of gas (micromol N2 Lndash1) Denitrificationis a calculation of the volumetric N2 input resulting from den-itrification (micromol N2 L

ndash1 minndash1) K is the gas transfer coefficientfor N2 (minndash1) The length of the time step (t) is 1 minK modelsmdashSix alternative models of gas transfer coefficient

(K) were incorporated into the N2 model three models based onwind-induced aeration (Wanninkhof 1992 Cole and Caraco1998 Laursen and Seitzinger 2005) and three models wherebenthic turbulence-induced gas exchange is estimated based on

stream depth and velocity (OrsquoConnor and Dobbins 1958Churchill et al 1962 Owens et al 1964 seeWeb Appendix 2 fordetails including model formulae) Because no single modelappears adequate for describing rates of gas transfer across thediverse conditions within streams (Melching and Flores 1999Gualtieri et al 2002) we apply this range of model outputs asan indicator of whether the open-channel method may beapplicable in a single study stream These equations wereselected based on their limited data requirements and resultingusefulness in making first assessments of gas exchange how-ever these are just a few examples of a vast literature describingrates of gas transfer More complex models that incorporate fac-tors such as slope and roughness may provide more accurateresults in some systems If the criterion of Schwarzenbach et al(1993 seeWeb Appendix 2) is applied the majority of our studyscenarios (Fig 1) reflect systems where benthic turbulence islikely to drive gas exchange The only exception is at a depth of1 m and the minimum stream velocity tested of 002 m sndash1

where wind models are expected to be more appropriate (seeWeb Appendix 2) Revisiting method sensitivity is clearly nec-essary as previous analyses (Laursen and Seitzinger 2005) werebased on the assumption of wind-induced gas exchange Weprovide the wind model analysis to highlight differences fromthe previous analysis In subsequent discussion we refer to allgas exchange models by using the name of only the first authorModel-based analysismdashThe N2 model was run for denitrifica-

tion rates ranging from 100 to 2000 micromol N mndash2 hndash1 (in100 micromol N mndash2 hndash1 increments) for each K model under arange of stream depths velocities and wind speeds (Fig 1) Werecorded the time after the model start (at saturation condi-tions) that N2 accumulation due to denitrification could be dis-tinguished reliably from background N2 concentrations that isthe time at which N2 accumulation is greater than or equal toN2 accumulation in the absence of denitrification plus anerror term N2 modeled denitrification = times gt N2 modeled denitrification = 0 +error term

To facilitate comparison with the work of Laursen andSeitzinger (2005) we use an error term of 1 micromol N2 L

ndash1 This fig-ure is based on analytical error of 022 to 041 micromol N2 Lndash1Whereas this figure reflects the precision of the method (iereproducibility of multiple replicates) method accuracy (ie dif-ference from the true value) can be somewhat poorer Analysesof Kana et al (1994) indicate that differences between measuredvalues and expected values across a range of concentrations from326 to 473 micromol N2 L

ndash1 were typically between 024 and 165depending on temperature and whether N2 or N2Ar was con-sidered Accuracy of analyses is likely to vary among labs andover time depending on factors such as instrument stability Forcomparative purposes we reran the samemodel analysis assum-ing a considerably higher error term of 4 micromol N2 L

ndash1

(1)N N + denitrificationumolL

umol N

m

2 12 2

2

( )( )

tt

+

times

=hh

m

L depth m

htimes times times

times

1

1000

1 1

60

3

( ) min

(mint ))( )

min(mindash ( ndash )N Numol

Lumol

Lsat2 21

tK ttimes

times nn)

Baulch et al Denitrification Open-channel method

206

The model was run from saturation conditions using a 1 mintime step for up to 12 d We plot the time after model start thatdenitrification becomes measurable based on these two errorterms in Figs 2 and 3 We also note (at the top of the figures)scenarios where even after 12 d denitrification rates were notmeasurable All model runs were performed at a temperature of20degC pressure of 10083 hPa and wind speed of 2 m sndash1 (windspeeds are at 10 m height) These conditions were selected tofacilitate easy comparison with the model analyses of Laursenand Seitzinger (2005) If temperature and pressure are fixedthey have minor effects on model results while wind speed canhave a more important influence (Laursen and Seitzinger 2005)In studies that have applied the open-channel method to datenitrogen fixation anammox loss of N2 in bubbles and loss ofN2 to groundwater have been assumed to be negligible and areexcluded from this model Likewise N2 gain from groundwateris also excluded from the model-based analysis However theimportance of these error terms is discussed

This analysis only determines that biotic N2 production canbe distinguished from physical N2 fluxes and does not constrainthe uncertainty in the measurements Given that measurementuncertainty can be quite significant (Web Appendix 1mdashTable 1)this is a broad first approximation of suitable streams

Assessment and discussionSite selection based on modeled gas transfermdashThe initial time

point in our model-based analysis is at saturation conditionsWithin shallow systems with high gas transfer and a highdegree of diel temperature variation measured N2 concentra-tions may bisect saturation concentrations twice per day sug-gesting that researchers studying systems such as these wouldbe restricted to sites that show measurable N2 accumulation intime periods lt 8-12 h In contrast in shallow systems with lowgas exchange rates and sufficiently high denitrification rates N2

may remain above saturation throughout diel sampling periods(McCutchan et al 2003 Pribyl et al 2005) Where longer-termN2 accumulation is likely (eg thermally stable systems withlow gas exchange) the symbols at the top of Figs 2 and 3 maybe used to estimate detection limits These symbols identify sce-

narios under which N2 accumulation was not measurable Thatis measured concentrations did not exceed saturation concen-trations by more than 1 micromol N2 Lndash1 (Fig 2) or 4 micromol N2 Lndash1

(Fig 3) after a period of 12 d In the vast majority of casessteady-state conditions were established within this timeperiod In some cases (largely in systems where the method ishighly sensitive) N2 accumulation may continue beyond 12 dHowever we use 12 d as an upper limit because longer termthermal stability and stability in discharge of shallow (lt 1 mdepth) streams seem unlikely As well the error associated withmeasuring factors such as gas transfer over this time periodcould be considerable Interpretations based on 40 h accumula-tion or 12 d accumulation yield broadly similar trendsmdashthemethod is not suited to measurement of low denitrificationrates in shallow systems with moderate-high flow In deeper sys-tems the method is more sensitive but sensitivity declines asstream velocity and associated turbulence increases Compari-son of the two error terms indicates that very good analyticalaccuracy is critical to the ability to measure low or even mod-erate rates of denitrification

The model-based analyses indicate that shallow streams(02 m) may represent systems in which low denitrificationrates may be measured but only at extremely low flow ratesand with good analytical accuracy Gas exchange rates atstream velocities of 002 m sndash1 are predicted to be low enoughthat N2 produced at low rates can be detected in the shallowwater column relatively quickly and denitrification ratesle100 micromol N mndash2 hndash1 may be measurable (Fig 2) Howevercareful sampling design may be required to account for poormixing within such streams At higher stream velocities pre-dictions of the wind and benthic reaeration models divergequite markedly and minimum measurable denitrificationrates are increased Rates of denitrification below 1000 micromol Nmndash2 hndash1 should be measurable in shallow (02 m) streams withmoderate flows (016-032 m sndash1 Fig 2) if analytical accuracyis high (within 1 micromol N2 Lndash1) Clearly the method is notsuited to measuring low denitrification rates in very shallowstreams with very high velocity (064 m sndash1) where turbulence-induced gas exchange leads to rapid re-equilibration of water

Fig 1 K values resulting from 6 reaeration models under simulation conditions These K values are used to generate results in Figs 2 and 3 The first threemodels are indicative of benthic turbulence induced gas exchange The latter three models estimate gas exchange driven by wind speeds of 2 m sndash1 SeeWeb Appendix 2 for further details

Baulch et al Denitrification Open-channel method

207

masses with the atmosphere Within streams with depths of05-1 m denitrification rates le 500 micromol N mndash2 hndash1 should bemeasurable even at the highest stream velocity tested (064 msndash1) if analytical accuracy is high In deeper systems with low

denitrification rates the time until denitrification becomesmeasurable can be rather long (Figs 2 3) although such sys-tems are likely to have greater thermal stability than shallowstreams (Caissie 2006) If analytical accuracy is poor this has

Fig 2 Time required from model start (at saturation conditions) before denitrification is measurable under a suite of stream depths (columns) andstream velocities (rows) assuming analytical uncertainty of 1 microM N2 Base conditions of stable wind speed (2 m sndash1) temperature (20degC) and pressure(10083 hPa) were used NM indicates that denitrification rates were not measurable after 12 d

Baulch et al Denitrification Open-channel method

208

Fig 3 Time required from model start (at saturation conditions) before denitrification is measurable under a suite of stream depths (columns) andstream velocities (rows) assuming analytical uncertainty of 4 microM N2 Base conditions of stable wind speed (2 m sndash1) temperature (20degC) and pressure(10083 hPa) were used NM indicates that denitrification rates were not measurable after 12 d

Baulch et al Denitrification Open-channel method

209

considerable implications for application of the method par-ticularly in systems with moderate to high stream velocity andin streams lt 1 m in depth (Figs 2 and 3)

The selection of benthic turbulence versus wind-basedmodels (sensu Laursen and Seitzinger 2005) to estimatemethod sensitivity in candidate study streams is clearly impor-tant The selected gas exchange models were developed for dif-ferent environmental conditions and researchers may wish toreview these conditions (Web Appendix 2) Alternative modelsof benthic turbulence-induced gas exchange that incorporatefactors such as surface roughness and slope may help furtherguide researchers in site selection Likewise models incorpo-rating both benthic and wind-based gas exchange may be suit-able in some systems (Chu and Jirka 2003) However nomodel-based approach is adequate for in-field application ofthe open-channel method except in very deep systems wherewind-based models may be sufficiently accurate (Yan et al2004) and direct measurements are logistically difficultAlthough the model analysis of Laursen and Seitzinger (2005)in conjunction with our findings can guide researchers in siteselection we suggest that the next step in ensuring appropri-ate site selection is detailed site surveys to assess spatial vari-ability and in particular to reveal the presence of high gasexchange reaches such as reaches with high slopes or riffleswhich may limit method sensitivity

Finally we note that many researchers will lack a prioriknowledge of denitrification rates to allow them to makeapproximations of whether the method is likely to succeed inindividual study systems Published denitrification rates fromdifferent systems may be used as a preliminary guide (egMulholland et al 2004 Laursen and Seitzinger 2005) Surveysof N2Ar ratios may also be used to determine whether theratio is considerably above the saturation value Alternativelypreliminary measurements using other techniques or a con-servative approach to site selection is sensibleOne station versus two station approachesmdashWhile the one-

station and two-station approaches are mathematicallyequivalent important differences in the assumptions of thesemethods as well as logistical concerns associated with theirimplementation may dictate which approach is more suit-able Importantly the one-station method incorporates atime-for-space substitution where temporal patterns of con-centration and solubility at an (unsampled) upstream loca-tion are assumed to be the same as those measured in thesame parcel of water as it passes through the downstreamsampling station

Determining where to add tracers when measuring gastransfer coefficient is less straightforward for the one-stationapproach than the two-station approach Gas exchange is afirst-order reaction (ie exponential loss) hence the length ofstream that is integrated by a concentration measurement canbe calculated Researchers may choose to use a metric such asEq 2 as a starting point in selecting reach length Assuming astream is supersaturated Eq 2 represents the length of reach

over which 95 of upstream gases will be released and 5 willremain in solution

(2)

where U is stream velocity K is the gas transfer coefficient andDistance5 is the distance at which 5 of upstream gasesremain in solution (Beaulieu et al 2008 Chapra and Di Toro1991) However reach lengths calculated using this methodare often quite long and may be impractical for tracer additionexperiments

There are likely to be important differences in the accuracyof modeled saturation conditions between one-station andtwo-station approaches Systems with high diel temperaturevariation are not ideal for application of any open-channeldenitrification measurements (Laursen and Seitzinger 2005)although as in dissolved oxygen studies with appropriatemodeling and careful measurements the method may still beused (Butcher and Covington 1995) Using the one-stationapproach installation of temperature and pressure loggers issimple but as mentioned incorporates the assumption thatan upstream site has similar conditions Within two-stationstudies even with relatively low diel temperature variationdifficulty in measuring the temperature of a parcel of watermoving downstream may lead to error in the estimation ofsaturation concentrations Ideally real-time measurements oftemperature as the parcel of water passes downstream shouldbe obtained (eg Smith et al 2008) however interpolationmay also provide adequate results

Although one station versus two station comparisons haveyielded similar results for measurements of photosynthesisand respiration these comparisons have not been made forthe open-channel denitrification method and may yield newinsights into confidence in the open-channel methodMeasurement of gas transfer coefficientmdashOne of the most crit-

ical parameters in determination of denitrification rates usingthis method is the measurement of rates of air-water gas trans-fer Modeling this parameter is not adequate for accuratedetermination of denitrification rates in most systems Effortsshould be made to ensure the measurement of gas transfercoefficient reflects the period of sampling and study reach asclosely as possible (Web Appendix 1mdashTable 2) Whereas sta-bility in the gas transfer coefficient over time has been shownin some streams and rivers (eg Pribyl et al 2005) othersshow very high seasonal and spatial variability (Hope et al2001) Using the two-station approach the gas transfer coeffi-cient can be measured to exactly align with N2 sampling inspace and time Using the one-station approach the mea-surement should be designed to reflect rates of gas transfer forthe stream length over which gas concentrations are inte-grated The importance of diel changes in wind speed andpotential for resulting bias in measured gas transfer rateshould also be assessed

The measurement of gas transfer coefficient using directtracer additions is a widely used relatively straightforward

Distance5

3 =

U

K

Baulch et al Denitrification Open-channel method

208

Fig 3 Time required from model start (at saturation conditions) before denitrification is measurable under a suite of stream depths (columns) andstream velocities (rows) assuming analytical uncertainty of 4 microM N2 Base conditions of stable wind speed (2 m sndash1) temperature (20degC) and pressure(10083 hPa) were used NM indicates that denitrification rates were not measurable after 12 d

Baulch et al Denitrification Open-channel method

209

considerable implications for application of the method par-ticularly in systems with moderate to high stream velocity andin streams lt 1 m in depth (Figs 2 and 3)

The selection of benthic turbulence versus wind-basedmodels (sensu Laursen and Seitzinger 2005) to estimatemethod sensitivity in candidate study streams is clearly impor-tant The selected gas exchange models were developed for dif-ferent environmental conditions and researchers may wish toreview these conditions (Web Appendix 2) Alternative modelsof benthic turbulence-induced gas exchange that incorporatefactors such as surface roughness and slope may help furtherguide researchers in site selection Likewise models incorpo-rating both benthic and wind-based gas exchange may be suit-able in some systems (Chu and Jirka 2003) However nomodel-based approach is adequate for in-field application ofthe open-channel method except in very deep systems wherewind-based models may be sufficiently accurate (Yan et al2004) and direct measurements are logistically difficultAlthough the model analysis of Laursen and Seitzinger (2005)in conjunction with our findings can guide researchers in siteselection we suggest that the next step in ensuring appropri-ate site selection is detailed site surveys to assess spatial vari-ability and in particular to reveal the presence of high gasexchange reaches such as reaches with high slopes or riffleswhich may limit method sensitivity

Finally we note that many researchers will lack a prioriknowledge of denitrification rates to allow them to makeapproximations of whether the method is likely to succeed inindividual study systems Published denitrification rates fromdifferent systems may be used as a preliminary guide (egMulholland et al 2004 Laursen and Seitzinger 2005) Surveysof N2Ar ratios may also be used to determine whether theratio is considerably above the saturation value Alternativelypreliminary measurements using other techniques or a con-servative approach to site selection is sensibleOne station versus two station approachesmdashWhile the one-

station and two-station approaches are mathematicallyequivalent important differences in the assumptions of thesemethods as well as logistical concerns associated with theirimplementation may dictate which approach is more suit-able Importantly the one-station method incorporates atime-for-space substitution where temporal patterns of con-centration and solubility at an (unsampled) upstream loca-tion are assumed to be the same as those measured in thesame parcel of water as it passes through the downstreamsampling station

Determining where to add tracers when measuring gastransfer coefficient is less straightforward for the one-stationapproach than the two-station approach Gas exchange is afirst-order reaction (ie exponential loss) hence the length ofstream that is integrated by a concentration measurement canbe calculated Researchers may choose to use a metric such asEq 2 as a starting point in selecting reach length Assuming astream is supersaturated Eq 2 represents the length of reach

over which 95 of upstream gases will be released and 5 willremain in solution

(2)

where U is stream velocity K is the gas transfer coefficient andDistance5 is the distance at which 5 of upstream gasesremain in solution (Beaulieu et al 2008 Chapra and Di Toro1991) However reach lengths calculated using this methodare often quite long and may be impractical for tracer additionexperiments

There are likely to be important differences in the accuracyof modeled saturation conditions between one-station andtwo-station approaches Systems with high diel temperaturevariation are not ideal for application of any open-channeldenitrification measurements (Laursen and Seitzinger 2005)although as in dissolved oxygen studies with appropriatemodeling and careful measurements the method may still beused (Butcher and Covington 1995) Using the one-stationapproach installation of temperature and pressure loggers issimple but as mentioned incorporates the assumption thatan upstream site has similar conditions Within two-stationstudies even with relatively low diel temperature variationdifficulty in measuring the temperature of a parcel of watermoving downstream may lead to error in the estimation ofsaturation concentrations Ideally real-time measurements oftemperature as the parcel of water passes downstream shouldbe obtained (eg Smith et al 2008) however interpolationmay also provide adequate results

Although one station versus two station comparisons haveyielded similar results for measurements of photosynthesisand respiration these comparisons have not been made forthe open-channel denitrification method and may yield newinsights into confidence in the open-channel methodMeasurement of gas transfer coefficientmdashOne of the most crit-

ical parameters in determination of denitrification rates usingthis method is the measurement of rates of air-water gas trans-fer Modeling this parameter is not adequate for accuratedetermination of denitrification rates in most systems Effortsshould be made to ensure the measurement of gas transfercoefficient reflects the period of sampling and study reach asclosely as possible (Web Appendix 1mdashTable 2) Whereas sta-bility in the gas transfer coefficient over time has been shownin some streams and rivers (eg Pribyl et al 2005) othersshow very high seasonal and spatial variability (Hope et al2001) Using the two-station approach the gas transfer coeffi-cient can be measured to exactly align with N2 sampling inspace and time Using the one-station approach the mea-surement should be designed to reflect rates of gas transfer forthe stream length over which gas concentrations are inte-grated The importance of diel changes in wind speed andpotential for resulting bias in measured gas transfer rateshould also be assessed

The measurement of gas transfer coefficient using directtracer additions is a widely used relatively straightforward

Distance5

3 =

U

K

Baulch et al Denitrification Open-channel method

209

considerable implications for application of the method par-ticularly in systems with moderate to high stream velocity andin streams lt 1 m in depth (Figs 2 and 3)

The selection of benthic turbulence versus wind-basedmodels (sensu Laursen and Seitzinger 2005) to estimatemethod sensitivity in candidate study streams is clearly impor-tant The selected gas exchange models were developed for dif-ferent environmental conditions and researchers may wish toreview these conditions (Web Appendix 2) Alternative modelsof benthic turbulence-induced gas exchange that incorporatefactors such as surface roughness and slope may help furtherguide researchers in site selection Likewise models incorpo-rating both benthic and wind-based gas exchange may be suit-able in some systems (Chu and Jirka 2003) However nomodel-based approach is adequate for in-field application ofthe open-channel method except in very deep systems wherewind-based models may be sufficiently accurate (Yan et al2004) and direct measurements are logistically difficultAlthough the model analysis of Laursen and Seitzinger (2005)in conjunction with our findings can guide researchers in siteselection we suggest that the next step in ensuring appropri-ate site selection is detailed site surveys to assess spatial vari-ability and in particular to reveal the presence of high gasexchange reaches such as reaches with high slopes or riffleswhich may limit method sensitivity

Finally we note that many researchers will lack a prioriknowledge of denitrification rates to allow them to makeapproximations of whether the method is likely to succeed inindividual study systems Published denitrification rates fromdifferent systems may be used as a preliminary guide (egMulholland et al 2004 Laursen and Seitzinger 2005) Surveysof N2Ar ratios may also be used to determine whether theratio is considerably above the saturation value Alternativelypreliminary measurements using other techniques or a con-servative approach to site selection is sensibleOne station versus two station approachesmdashWhile the one-

station and two-station approaches are mathematicallyequivalent important differences in the assumptions of thesemethods as well as logistical concerns associated with theirimplementation may dictate which approach is more suit-able Importantly the one-station method incorporates atime-for-space substitution where temporal patterns of con-centration and solubility at an (unsampled) upstream loca-tion are assumed to be the same as those measured in thesame parcel of water as it passes through the downstreamsampling station

Determining where to add tracers when measuring gastransfer coefficient is less straightforward for the one-stationapproach than the two-station approach Gas exchange is afirst-order reaction (ie exponential loss) hence the length ofstream that is integrated by a concentration measurement canbe calculated Researchers may choose to use a metric such asEq 2 as a starting point in selecting reach length Assuming astream is supersaturated Eq 2 represents the length of reach

over which 95 of upstream gases will be released and 5 willremain in solution

(2)

where U is stream velocity K is the gas transfer coefficient andDistance5 is the distance at which 5 of upstream gasesremain in solution (Beaulieu et al 2008 Chapra and Di Toro1991) However reach lengths calculated using this methodare often quite long and may be impractical for tracer additionexperiments

There are likely to be important differences in the accuracyof modeled saturation conditions between one-station andtwo-station approaches Systems with high diel temperaturevariation are not ideal for application of any open-channeldenitrification measurements (Laursen and Seitzinger 2005)although as in dissolved oxygen studies with appropriatemodeling and careful measurements the method may still beused (Butcher and Covington 1995) Using the one-stationapproach installation of temperature and pressure loggers issimple but as mentioned incorporates the assumption thatan upstream site has similar conditions Within two-stationstudies even with relatively low diel temperature variationdifficulty in measuring the temperature of a parcel of watermoving downstream may lead to error in the estimation ofsaturation concentrations Ideally real-time measurements oftemperature as the parcel of water passes downstream shouldbe obtained (eg Smith et al 2008) however interpolationmay also provide adequate results

Although one station versus two station comparisons haveyielded similar results for measurements of photosynthesisand respiration these comparisons have not been made forthe open-channel denitrification method and may yield newinsights into confidence in the open-channel methodMeasurement of gas transfer coefficientmdashOne of the most crit-

ical parameters in determination of denitrification rates usingthis method is the measurement of rates of air-water gas trans-fer Modeling this parameter is not adequate for accuratedetermination of denitrification rates in most systems Effortsshould be made to ensure the measurement of gas transfercoefficient reflects the period of sampling and study reach asclosely as possible (Web Appendix 1mdashTable 2) Whereas sta-bility in the gas transfer coefficient over time has been shownin some streams and rivers (eg Pribyl et al 2005) othersshow very high seasonal and spatial variability (Hope et al2001) Using the two-station approach the gas transfer coeffi-cient can be measured to exactly align with N2 sampling inspace and time Using the one-station approach the mea-surement should be designed to reflect rates of gas transfer forthe stream length over which gas concentrations are inte-grated The importance of diel changes in wind speed andpotential for resulting bias in measured gas transfer rateshould also be assessed

The measurement of gas transfer coefficient using directtracer additions is a widely used relatively straightforward

Distance5

3 =

U

K

Baulch et al Denitrification Open-channel method

210

technique within systems of low water mass (see Hibbs et al1998 Kilpatrick et al 1989) Various other methods exist forestimating the gas transfer coefficient each with their owninherent assumptions and errors Methods include diel O2

change (Odum 1956 Streeter and Phelps 1925 Venkiteswaranet al 2007) controlled flux techniques (see Jaumlhne andHauszligecker 1998) or the use of Ar anomalies (Laursen andSeitzinger 2002 2004) where gas transfer coefficient is esti-mated based on the deviation of Ar from saturation values TheAr anomaly method has been used in open-channel denitrifi-cation measurements (Laursen and Seitzinger 2002) and incor-porates many of the same assumptions Whereas this appearsto be an effective means of estimating gas transfer coefficientin some ecosystems it may be less accurate than direct tracermeasurements (Web Appendix 1mdashTable 2) Irrespective of themethod selected to measure gas transfer coefficient some esti-mate of the error associated with the measurement is requiredDiel variationmdashSignificant diel changes in water tempera-

ture occur in most aquatic systems with low thermal massesleading to changes in gas solubility air-water gas transfer coef-ficient and rates of biological processes The importance ofthese diel temperature changes in open-channel denitrifica-tion measurements was noted by Laursen and Seitzinger(2005) who recommended selection of sites or times whenthermal stability was maximal (Web Appendix 1mdashTable 2) Ifall parameters are adequately constrained the physical factorsassociated with diel temperature changes will not significantlyaffect measurement of denitrification rates However whererates of temperature variation are high more frequent sam-pling is required to account for changes in physical drivers ofgas flux As well changes in biological rates driven by changesin temperature are more likely to occur As a result where tem-perature changes are significant mathematical considerationof how temperature may affect denitrification rate is advisedrather than model-fitting assuming a constant denitrificationrate (Butcher and Covington 1995 Web Appendix 1mdashTable 2)

The role of physical gas fluxes from sediments is also wor-thy of consideration In systems with strong seasonal temper-ature variation but relatively stable daily thermal conditionsphysical N2 fluxes between sediments and overlying water canbe sufficiently high to lead to error in denitrification mea-surements (Lamontagne and Valiela 1995) In lower order sys-tems this effect could be exhibited to some extent over a dielperiod depending on sediment permeability and stream heatbudgets

Monitoring wind speed is recommended (Web Appendix 1mdashTable 2) due to potential effects of diel variation on gas trans-fer coefficient Diel variation in discharge is relatively com-mon in rivers (Lundquist and Cayan 2002) and is often drivenby variation in evapotranspiration and infiltration Whilevariation in discharge is likely to have some effect on reaera-tion rates the more important implication is that groundwa-ter dynamics may be temporally variable during open-channelmeasurements complicating N2 mass balances

GroundwatermdashGroundwater inputs if not adequately char-acterized could lead to significant error in denitrificationmeasurements (see Laursen and Seitzinger 2005) becausegroundwater is frequently supersaturated with N2 and becausetemperature often differs between ground and surface watersWhere groundwater or tile drain inputs are known to occuralong a stream Eq 2 (or an alternate first order loss calcula-tion with a different endpoint such as 1 or 10) can be usedto help constrain the region in which gas concentrations areaffected by the inflow Sites that lack significant groundwaterN2 inputs are best for application of this method Applicationof the method to sites with groundwater inputs relies uponadequate spatial and temporal characterization of the inputsand ensuring the inputs are well mixed at sampling points(Web Appendix 1mdashTable 2) Groundwater outputs are lessproblematic as they represent the loss of a mass of water witha N2 concentration equivalent to that of the stream

A final scenario is situations where groundwater is lost andgained along a reach This could effectively disguise ground-water inputs if researchers adopt the commonly used incre-mental stream flow method where upstream and downstreamdischarge are compared to identify net groundwater inputs orlosses In cases where groundwater dynamics are not wellknown groundwater inflows along a reach can be made con-currently with measurements of gas transfer coefficient bymeasuring dilution of the conservative tracer or using alterna-tive approaches (Kalbus et al 2006) Direct measurement ofgross groundwater inflow would significantly reduce uncer-tainty in the application of this method Alternatively in-stream piezometers could be used to indicate the direction ofgroundwater flow

A final note regarding groundwater inputs relates to themixing of two water masses with different temperaturesBecause saturation conditions vary nonlinearly if two watermasses at equilibrium mix the resulting water mass will havea N2 concentration that deviates at least slightly from satura-tion concentrations For example if 10 of stream flow isderived from N2-saturated groundwater at 10degC but thestream is at saturation at 20degC instantaneous mixing of thesewater masses would result in a temperature of 19degC but a N2

concentration of 5401 microM This is 19 microM above saturationfor that water temperature and could be misinterpreted as adenitrification signal Although a correction for the excess N2

in groundwater is relatively straightforward determining howto handle mixing of water masses with different temperaturesis more difficult and will depend on how quickly the mixedwater mass is subsequently heated or cooled If the excess N2

source can be identified and corrected for the spatial effect ontemperature measurements can also be constrained and maybe present only over a limited areaMixingmdashBoth the one-station and two-station approaches

assume that the water column is well mixed This may be adifficult assumption to fulfill in broad slow moving riverswhere lateral variation in gas concentrations or temperature

Baulch et al Denitrification Open-channel method

211

may occur or in systems with significant groundwater inflowsor macrophytes Problems of incomplete mixing may beaddressed by assessing this heterogeneity in sampling designhowever uncertainty in such studies would increase as thevolume N2 concentration and temperature of water massesacross stream transects would have to be characterized Themixing of water masses with different temperatures also hasimplications for estimation of N2 saturationEbullition and N2 fixation as loss mechanisms for N2mdashNitro-

gen fixation is a sink for dissolved N2 hence estimates of den-itrification using the open-channel method are often referredto as net denitrification (ie gross denitrification minus N2

fixation) Within fluvial systems the importance of N2 fixa-tion to N budgets is very much an open question Very fewmeasurements of this process have been made using anymethod and even fewer measurements have been made con-currently with measurements of other N uptake processes(Marcarelli et al 2008) The difference between net and grossdenitrification will depend on the study system and approxi-mations of the differences among systems cannot be madewithout further study Nonetheless net denitrification is animportant measurement for studies focused on constraining Nbudgets although understanding environmental controls onthis composite measurement is more difficult In contrast toN2 fixation anammox can lead to the production of N2although the importance of this process in fluvial systems isnot known

The open-channel method as described here does notincorporate bubble- or plant-mediated N2 fluxes Bubble-mediated gas fluxes are poorly constrained in lotic systemsbut a recent study in the South Platte River showed that ebul-lition may transport up to 044 g N mndash2 dndash1 from sediments tothe atmosphere (Higgins et al 2008) By comparing these esti-mates to past measurements of diffusive flux in the systemHiggins et al (2008) estimate 9 to 16 of N2 flux may bedriven by bubble-mediated transport The importance of bub-bles as an error term will depend on rates of bubble releasebubble N2 concentration and whether bubble N2 was ofatmospheric origin or produced via denitrification Systemswith high bubble fluxes are best avoided for the application ofthis method The importance of macrophytes as conduits ofN2 merits further investigationMeasurement of N2 concentrationsmdashMembrane-inlet mass

spectrometry allows high precision measurement of dissolvedN2 concentrations via either direct measurement of the N2 sig-nal or the use of N2 to argon (N2Ar ratios) Ratio measurementshave greater precision (coefficient of variation of lt 005Kana et al 1994) than measurement of the N2 signal (coeffi-cient of variation lt 05 Kana et al 1994) and are well-suitedto measurement of denitrification in chambers where Ar con-centrations can generally be assumed to equal saturation con-centrations Based on knowledge of Ar N2 concentrations canthen be calculated In either case (measuring N2 directly orN2Ar) obtaining three to four replicate samples is typically suf-

ficient to obtain good precision (Web Appendix 1mdashTable 2)Within open systems particularly where groundwater inputsare significant or where diel temperature changes occur Arconcentrations frequently deviate from saturation necessitat-ing their direct measurement (Laursen and Seitzinger 2002)

Calibration of MIMS is typically performed using one ormore air-equilibrated standards at known temperature pres-sure and salinity Method accuracy is typically good (Kana etal 1994) but very high accuracy is required for application ofthe open reach method (Figs 2 and 3) The majority of MIMS-based denitrification measurements are from closed chambersso slight errors in instrument calibration will have lesser con-sequences than in the open-system method Multiple stan-dards should be used to ensure adequate calibration whenapplying the open-channel method and analytical uncer-tainty terms for uncertainty analysis should be chosen care-fully (Web Appendix 1mdashTable 3) Sample handling is an addi-tional concern as changes in concentration during storagemay occur (McCutchan et al 2003) Ideally samples should beobtained in vials with ground glass stoppers and stored under-water at 1 to 2degC below ambient stream temperatures FinallyN2 may be scavenged by O2 in some mass spectrometersresulting in NO+ formation in the ion source and a decreasein the measured N2 signal (Eyre et al 2002) Researchersshould test for this effect and determine whether the installa-tion of a heated copper reduction column is necessary toremove O2 (Eyre et al 2002 Kana and Weiss 2004)Time scale of samplingmdashTime step can be of critical impor-

tance both in modeling and ensuring the sampling program isadequately capturing in situ N2 dynamics Data loggers can beemployed to characterize saturation conditions at very shorttime intervals however N2 sampling is performed at longertime intervals due to the effort involved in field sampling andlab analyses In one-station or diel approaches the ability ofthe model to reproduce the tops and the bottoms of concen-tration curves is a good test of model accuracy Samplesobtained near daily maximum and minimum temperaturesare therefore particularly valuable in assessing model fitAssumption of constant denitrification ratemdashThe model is

applied to generate an average denitrification rate thatresults in the best fit between measured and predicted N2

concentrations over the study period That is the methodassumes a single constant time and space-integrated denitri-fication rate Because denitrification rates can vary withnumerous factors including temperature oxygen status andsubstrate type (Risgaard-Petersen et al 1994 Kemp andDodds 2002ab) this assumption may not hold true in sys-tems with significant temporal or spatial variability in thesefactors For example diel variation in denitrification ratesmay occur (An and Joye 2001) In four systems where theopen-channel method has been applied short-term varia-tion in denitrification rates was significant (Laursen andSeitzinger 2004 Harrison et al 2005) but it was not signifi-cant in a fifth (McCutchan et al 2003)

Baulch et al Denitrification Open-channel method

212

Modeling errormdashThe open-channel method calculates den-itrification based on the following measured input termsmorphometry (depth width) stream velocity time mea-sured concentrations equilibrium concentrations measuredtemperature and gas transfer coefficient (Laursen andSeitzinger 2002) Equilibrium concentrations are determinedfrom solubility equations using measured temperature andsalinity Equilibrium concentrations should also be correctedfor atmospheric pressure Method uncertainty is a functionof uncertainty in the measured variables and the parametersderived from these measurements Method uncertainty isalso affected by assumptions of the method including theuse of a single (temporally and spatially integrated) denitrifi-cation rate

Many researchers have wisely attempted to constrain theerror in reported denitrification rates However the handlingof error terms has varied among research groups In severalstudies error was assumed to be additive and analytical errorerrors in gas transfer measurements Schmidt number expo-nent and morphometry have been incorporated into esti-mates (Laursen and Seitzinger 2002 Smith et al 2008) In adeeper river error in gas transfer and analytical uncertaintywere considered (Yan et al 2004) In other cases Monte Carloanalyses have been used incorporating analytical error calcu-lated solubility (due to temperature and pressure mea-surements) channel depth and groundwater inputs(McCutchan et al 2003) Where error within the method wasnot explored mathematically comparisons among methods ortimes have been used instead (Pribyl et al 2005)

Monte Carlondashbased analyses are likely to provide the mostrobust estimate of error (Beck 1987) In Monte Carlo uncer-tainty analysis the first step is the identification of uncertainmodel parameters In Web Appendix 1 (Table 3) we present asuite of error terms for the method and note which model

input parameters will be affected by these measurements Thenext step in Monte Carlo uncertainty analysis is assigningranges and distributions for parameters Where a single valueis most likely to be truemdashfor example in the measurement oftemperaturemdashthe measured value may be used to define themean of a distribution and specifications reported by themanufacturer can be used to place bounds on the distributionIf a parameter is highly insensitive (eg salinity in a freshwa-ter river) it may be set as fixed however the findings of theuncertainty analysis will then depend upon the assumption ofabsolute certainty in that parameter Assigning parameteruncertainty ranges and distributions is subjective but theimplications of different decisions can be easily explored byrerunning the analysis to reflect different assumptions Ifparameters covary covariances structures should be accountedfor or inflated estimates of uncertainty may result (Beck 1987Skeffington 2006)

Once the parameter distributions are assigned softwarepackages can then be used to run the model iteratively repeat-edly taking random (or stratified random) samples from theassigned parameter distributions The model may be run hun-dreds or thousands of times to generate a range of denitrifica-tion rates which can be used to estimate the uncertainty inopen-channel measurementsAlternative approaches to measuring denitrificationmdashThe

open-channel method is a powerful means of measuringdenitrification in some systems but its use is technicallychallenging and not suited to all ecosystems or researchquestions (Fig 4 Table 1) Reach scale denitrification esti-mates via open-channel measurements whole-stream 15Naddition experiments natural abundance nitrate isotopes orclassical mass balances are extremely useful in obtainingdenitrification rate measurements at scales relevant toecosystem models However each is limited in its applica-

Fig 4 Generalized suitability of different denitrification methods based on stream order (increasing from left to right) The diel temperature curve isbased on Caissie (2006) Chamber-based methods refer to in-stream or in-lab techniques

Baulch et al Denitrification Open-channel method

213

tion the open-channel method largely by detection limitsand groundwater inflow the 15N addition method by highcost of isotopes in high discharge systems or systems withhigh background nitrate natural abundance isotopes by theneed to characterize all nitrate sources and ensure all frac-tionating processes are adequately modeled and mass bal-ances by the need to measure all N fluxes other than deni-trification as well as changes in storage (Fig 4 Table 1) Incontrast chamber-based methods tend to be more flexibleand easier to apply across a range of ecosystem types thanmass balance natural abundance isotope or open-channelapproaches Chamber-based methods allow assessment ofheterogeneity at a fine spatial scale However scaling cham-ber-based estimates to reach and ecosystem scales can beerror prone and chambers can lead to the creation of unnat-ural conditions or allow shifts in microbial communities thatmay bias results (Table 1)

Methodological limitations have slowed progress in ourunderstanding of denitrification (Groffman et al 2006 David-son and Seitzinger 2006) While we agree with the assessmentof Groffman and others that the open-channel method is avaluable tool to help constrain denitrification rates and it islikely to be widely applied due to low costs and increasingavailability of analytical techniques it is clear that successfulapplication requires careful site selection highly accurate N2

measurements attention to sensitive parameters such as mea-surement of gas exchange rate and modeling that compre-hensively addresses methodological uncertainty (Web Appen-dix 1mdashTable 3)

RecommendationsThe open-channel method has contributed to improved

understanding of denitrification as a N removal process atenvironmentally relevant spatial scales Broader application ofthe method is feasible as analytical equipment becomes morewidely available However successful application of themethod is limited by our ability to reliably distinguish physi-cal and biotic N fluxes As a result the uncertainty associatedwith measurements can be large We recommend a conserva-tive approach to site selection and emphasize that themethod may not be suitable for shallow streams or fast flow-ing systems unless denitrification rates are high We concurwith Laursen and Seitzinger (2005) that deep rivers and peri-ods of high wind may also be unsuitable for application of thismethod However measurements in deep rivers may be feasi-ble via modifications to the method that rely upon longer-term gas accumulation and estimates of gas transfer (sensuYan et al 2004) All methods of measuring denitrificationhave inherent bias either in the types of systems in whichthey may be applied (Fig 4) or in potential artefacts associatedwith the methods themselves (eg acetylene block) For thisreason comparisons among methods should be applied whenpossible to foster a better understanding of the limitations ofeach method

References

An S and S B Joye 2001 Enhancement of coupled nitrifi-cation-denitrification by benthic photosynthesis in shallowestuarine sediments Limnol Oceanogr 4662-74

Beaulieu J J C P Arango S K Hamilton and J L Tank 2008The production and emission of nitrous oxide from headwa-ter streams in the Midwestern United States Glob ChangeBiol 14878-94 [doi101111j1365-2486200701485x]

Beck M B 1987 Water quality modeling a review of theanalysis of uncertainty Water Resour Res 231393-1441[doi101029WR023i008p01393]

Butcher J B and S Covington 1995 Dissolved-oxygen analy-sis with temperature dependence J Env Eng 121756-759[doi101061(ASCE)0733-9372(1995)12110(756)]

Caissie D 2006 The thermal regime of rivers a review Fresh-wat Biol 511389-1406 [doi101111j1365-2427200601597x]

Chapra S C and D M Di Toro 1991 Delta method for esti-mating primary production respiration and reaerationin streams J Env Eng 117640-655 [doi101061(ASCE)0733-9372(1991)1175(640)]

Chu C R and G H Jirka 2003 Wind and stream flowinduced reaeration J Env Eng 1291129-1136[doi101061(ASCE)0733-9372(2003)12912(1129)]

Churchill M A H L Elmore and R A Buchingham 1962The prediction of stream reaeration rates J Sanit Eng Div881-46

Cole J J and N F Caraco 1998 Atmospheric exchange ofcarbon dioxide in a low-wind oligotrophic lake measuredby the addition of SF6 Limnol Oceanogr 43647-656

Cox B A 2003 A review of dissolved oxygen modelling tech-niques for lowland rivers Sci Tot Environ 314303-334[doi101016S0048-9697(03)00062-7]

Davidson E A and S Seitzinger 2006 The enigma ofprogress in denitrification research Ecol Appl 162057-2063 [doi1018901051-0761(2006)016[2057TEOPID]20CO2]

Eyre B D S Rysgaard T Dalsgaard and P B Christensen2002 Comparison of isotope pairing and N2Ar methodsfor measuring sediment denitrification ndash assumptionsmodifications and implications Estuaries 251077-1087[doi101007BF02692205]

Groffman P M and others 2006 Methods for measuringdenitrification Diverse approaches to a difficult problemEcol Appl 162091-2122 [doi1018901051-0761(2006)016[2091MFMDDA]20CO2]

Gualtieri C P Gualtieri and G P Doria 2002 Dimensionalanalysis of reaeration rate in streams J Environ Eng12812-18 [doi101061(ASCE)0733-9372(2002)1281(12)]

Hamme R C and S R Emerson 2004 The solubility ofneon nitrogen and argon in distilled water and seawaterDeep-Sea Res 511517-1528

Harrison J A P A Matson and S E Fendorf 2005 Effects of

Baulch et al Denitrification Open-channel method

214

a diel oxygen cycle on nitrogen transformations and green-house gas emissions in a eutrophied subtropical streamAquat Sci 67308-315 [doi101007s00027-005-0776-3]

Hibbs D E K L Parkhill and J S Gulliver 1998 Sulfur hexa-fluoride gas tracer studies in streams J Env Eng 124752-760 [doi101061(ASCE)0733-9372(1998)1248 (752)]

Higgins T M J H McCutchan Jr and W M Lewis Jr 2008Nitrogen ebullition in a Colorado plains river Biogeo-chemistry 89367-377 [doi101007s10533-008-9225-4]

Ho D T L F Bliven R Wanninkhof and P Schlosser 1997The effect of rain on air-water gas exchange Tellus Ser B49149-158 [doi101034j1600-088949issue23x]

Hope D S M Palmer M F Billett and J J C Dawson 2001Carbon dioxide and methane evasion from a temperatepeatland stream Limnol Oceanogr 46847-857

Jaumlhne B 1987 On the parameters influencing air-water gasexchange J Geophys Res 921937-1950 [doi101029JC092iC02p01937]

Jaumlhne B and H Hauszligecker 1998 Air-water gas exchangeAnnu Rev Fluid Mech 30443-468 [doi101146annurevfluid301443]

Jha R C S P Ojha and K K S Bhatia 2001 Refinement ofpredictive reaeration equations for a typical Indian riverHydrol Proc 151047-1060 [doi101002hyp177]

Kalbus E F Reinstorf and M Schirmer 2006 Measuringmethods for groundwater surface water and their interac-tions a review Hydrol Earth Syst Sci 10873-887[doi105194hess-10-873-2006]

Kana T M C Darkangelo M D Hunt J B Oldham G EBennett and J C Cornwell 1994 Membrane inlet mass-spectrometer for rapid high-precision determination of N2O2 and Ar in environmental water samples Anal Chem664166-4170 [doi101021ac00095a009]

mdashmdashmdash and D L Weiss 2004 Comment on ldquoComparison ofisotope pairing and N2 Ar methods for measuring sedi-ment denitrificationrdquo Estuaries Coasts 27173-176[doi101007BF02803571]

mdashmdashmdash J C Cornwell and L Zhong 2006 Determination ofdenitrification in the Chesapeake Bay from measurementsof N2 accumulation in bottom water Estuaries Coasts29222-231 [doi101007BF02781991]

Kellman L 2004 Nitrate removal in a first-order stream rec-onciling laboratory and field measurements Biogeochem-istry 7189-105 [doi101007s10533-004-4318-1]

mdashmdashmdash and C Hillaire-Marcel 1998 Nitrate cycling instreams Using natural abundances of NO3

ndash-δ15N to measurein-situ denitrification Biogeochemistry 43273-292[doi101023A1006036706522]

Kemp M J and W K Dodds 2002a Comparisons of nitrifi-cation and denitrification in prairie and agriculturallyinfluenced streams Ecol Appl 12998-1009 [doi1018901051-0761(2002)012[0998CONADI]20CO2]

mdashmdashmdash and W K Dodds 2002b The influence of ammoniumnitrate and dissolved oxygen concentrations on uptake

nitrification and denitrification rates associated withprairie stream substrata Limnol Oceanogr 471380-1393

Kilpatrick F A R E Rathbun N Yotsukura G W Parker andL L DeLong 1989 Chapter A18 Determination of streamreaeration coefficients by use of tracers US Geological Sur-vey (Techniques of water-resources investigations of theUnited States Geological Survey)

Lamontagne M G and I Valiela 1995 Denitrification mea-sured by a direct N-2 flux method in sediments of WaquoitBay MA Biogeochemistry 3163-83 [doi101007BF00000939]

Laursen A and S Seitzinger 2005 Limitations to measuringriverine denitrification at the whole reach scale effects ofchannel geometry wind velocity sampling interval andtemperature inputs of N2-enriched groundwater Hydrobiol545225-236 [doi101007s10750-005-2743-3]

Laursen A E and S P Seitzinger 2002 Measurement of den-itrification in rivers an integrated whole reach approachHydrobiologia 48567-81 [doi101023A1021398431995]

mdashmdashmdash and mdashmdashmdash 2004 Diurnal patterns of denitrificationoxygen consumption and nitrous oxide production inrivers measured at the whole-reach scale Freshwat Biol491488-1458 [doi101111j1365-2427200401280x]

Lundquist J D and D R Cayan 2002 Seasonal and spatialpatterns in diurnal cycles in streamflow in the westernUnited States J Hydrometeor 3591-603 [doi1011751525-7541(2002)003lt0591SASPIDgt20 CO2]

Marcarelli A M M A Baker and W A Wurtsbaugh 2008 Isin-stream N2 fixation an important N source for benthiccommunities and stream ecosystems J N Am BentholSoc 27186-211 [doi10189907-0271]

Mackay D and A T K Yeun 1983 Mass-transfer coefficientcorrelations for volatilization of organic solutes from waterEnviron Sci Tech 17211-217 [doi101021es00110a006]

McCutchan J H J F Saunders A L Pribyl and W M Lewis2003 Open-channel estimation of denitrification LimnolOceanogr Methods 174-81

Melching C S and H E Flores 1999 Reaeration equationsderived from US geological survey database J Environ Eng125407-414 [doi101061(ASCE)0733-9372(1999)1255(407)]

Moog D B and G H Jirka 1998 Analysis of reaeration equa-tions using mean multiplicative error J Env Eng 124104-110 [doi101061(ASCE)0733-9372(1998)1242(104)]

Mulholland P J H M Valett J R Webster S A Thomas LW Cooper and S K Hamilton 2004 Stream denitrifica-tion and total nitrate uptake rates measured using a field15N tracer addition approach Limnol Oceanogr 49809-820

Nielsen L P 1992 Denitrification in sediment determinedfrom nitrogen isotope pairing FEMS Microb Ecol 86357-362 [doi101111j1574-69681992tb04828x]

OrsquoConnor D J and W E Dobbins 1958 Mechanism ofreaeration in natural streams Amer Soc Civil Eng Trans123641-684

Baulch et al Denitrification Open-channel method

215

Odum H T 1956 Primary production in flowing waters Lim-nol Oceanogr 1102-117 [doi104319lo1956120102]

Owens M R W Edwards and J W Gibbs 1964 Some reaer-ation studies in streams Int J Air Water Poll 8469-486

Pribyl A L J H McCutchan W M Lewis and J F Saunders2005 Whole-system estimation of denitrification in aplains river a comparison of two methods Biogeochem-istry 73439-455 [doi101007s10533-004-0565-4]

Raymond P A and J J Cole 2001 Gas exchange in riversand estuaries Choosing a gas transfer velocity Estuaries24312-317 [doi1023071352954]

Risgaard-Petersen N S Rysgaard L P Nielsen and N P Revs-bech 1994 Diurnal variation of denitrification and nitrifi-cation in sediments colonized by benthic microphytesLimnol Oceanogr 39573-579

Schwarzenbach R P P M Gschwend and D M Imboden1993 Environmental organic chemistry Wiley See p 235

Sjodin A L W M Lewis Jr and J F Saunders III 1997Denitrification as a component of the nitrogen budget fora large plains river Biogeochemistry 39327-342[doi101023A1005884117467]

Skeffington R A 2006 Quantifying uncertainty in criticalloads (A) literature review Water Air Soil Pollut 1693-24[doi101007s11270-006-0382-6]

Smith T E A E Laursen and J R Deacon 2008 Nitrogenattenuation in the Connecticut River northeastern USA acomparison of mass balance and N2 production modelingapproaches Biogeochemistry 87311-323 [doi101007s10533-008-9186-7]

Streeter H W and E B Phelps 1925 A study of the pollutionand natural purification of the Ohio River III Factors con-cerned in the phenomena of oxidation and reaerationUnited States Public Health Service US Department ofHealth Education and Welfare Public health bulletin no146 February 1925 Reprinted by USDHEW PHA 1958

Thamdrup B and T Dalsgaard 2002 Production of N2

through anaerobic ammonium oxidation coupled tonitrate reduction in marine sediments Appl Env Micro-biol 681312-1318

Thomann R V and J A Mueller 1987 Principles of surfacewater quality modeling and control quality modeling andcontrol Harper amp Row

Venkiteswaran J J L I Wassenaar and S L Schiff 2007Dynamics of dissolved oxygen isotopic ratios a transientmodel to quantify primary production community respira-tion and air-water exchange in aquatic ecosystems Oecolo-gia 153385-398 [doi101007s00442-007-0744-9]

Wanninkhof R 1992 Relationship between wind-speed andgas-exchange over the ocean J Geophys Res Oceans977373-7382 [doi10102992JC00188]

Yan W A E Laursen F Wang P Sun and S P Seitzinger2004 Measurement of denitrification in the ChangjiangRiver Environ Chem 195-98 [doi101071EN04031]

Submitted 18 March 2009Revised 16 September 2009Accepted 18 February 2010