Sep 20, 1999 - south of that creek, G6teborg University, G6teborg, Sweden .... tors Inc., Clinton Township, Michigan) to the pastry cover as a bottom plate.
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 104, NO. D17, PAGES 21,859-21,871, SEPTEMBER 20, 1999
Factors affecting the measurement of mercury emissions
from
soils with
flux
chambers
Dirk Wallschl/iger,Ralph R. Turner, and JacquelineLondon FrontierGeosciences Inc., Seattle,Washington
Ralf Ebinghausand Hans H. Kock Instituteof PhysicalandChemicalAnalytics,GKSS ResearchCenter,Geesthacht,Germany
Jonas Sommar and Zifan Xiao Departmentof InorganicChemistry,G0teborgUniversity,G0teborg,Sweden
Abstract. Air-surfaceexchangeof mercury(Hg) abovean aridgeothermalareawasmeasuredwith threeparallelflux chamberexperiments. The differentexperimental designs wereintercompared with eachother,with regardto themagnitudeof themeasured Hg fluxesandtheirresponse to environmentalchanges. Qualitatively,the measured Hg fluxesagreedwell throughout the diurnal cycle,andin theirresponse to environmental eventsandexperimentalmanipulations, but quanti-
tatively,thereweresignificant discrepancies between theindividua ! fluxresults.On average, the threedesignsyieldedHg fluxesagreeingwithina factorof 2, but evenmorepronounced differences were observedduringmiddayhigh emissionperiodsandduringapparentnighttimedeposition events.The chamberflushingrateappearsto havea very significantimpacton the measuredfluxes and on the responsebehaviorto environmentalchange.This studydemonstrates thatbothexperimentaldifferences andsmall-scale regionalvariabilityintroducelargeuncertaintyin the estimationof naturalHg air-surfaceexchangeby differentflux chambertechniques. Also, the impactof environmental parameters on Hg air-surfaceexchange was studied.Rain eventsled to a strongincreasein the Hg emissions, evenwhenthe coveredsoilremaineddry, suggesting thatthe apparent chamberfootprintis largerthanthe actuallycoveredarea.Exclusionof sunlightled to decreases in Hg emissions. Statisticalanalysisrevealedthe strongest correlations betweenthe measuredHg fluxesandradiationandwind speed.Weakercorrelations wereobservedwith air andsoiltemperature andwind direction(probablydueto localHg sources).Fluxeswere alsoinverselycorrelated with relative humidity.
1.
Introduction
The measurement of air-surfaceexchangeis emergingas one of the most important,but also one of the most challenging topics in studyingthe global biogeochemicalcycling of mercury. One unresolvedkey question is the relative importance of natural atmosphericHg emissionsin comparisonto direct anthropogenic discharges into the air, and to "indirect" anthropogenicHg emissions,which originate from historic industrial emissions,but are currently being (re)emitted from environmentalsurfaces[Ebinghaus et al., 1998]. Also, the biogeochemicalmechanismsdriving the exchangeof gaseous mercurybetween the atmosphereand waters or soils are not
and Casimir, 1998], soil moisture [Advokaat and Lindberg, 1996; D. Wallschl/iger et al., Mechanism and significance of mercuryvolatilization from contaminatedfloodplains of the Germanriver Elbe, submitted to Atmospheric Environment,
1999a,hereinafter referredto as Wallschl/igeret al., submitted manuscript,1999a]and radiation[Carpi and Lindberg, 1998; Amyot et al., 1994], and with decreasingrelativehumidity [Poissant and Casimir, 1998]. However, these individual observationscan currently not yet be linked to form a coherentpictureof the Hg air-surfaceexchangeprocessand the underlying biogeochemicalpathways. We feel that this problemis partially causedby a lack of studiesthat identify the individual Hg species contributing to the air-surface well understood. Previous studies have established correlaexchangeprocess. On the basis of older atmospheric tions between Hg fluxes and environmental parameters,such speciationdata [Schroederand Jackson,1987], it is generally as increasing emissions with increasing water temperature assumedthat total gaseousHg (TGM) in the environment [Xiao et al., 1991], soil temperature[Poissant and Casimir, 1998; Carpi and Lindberg, 1998], air temperature[Poissant
Copyright1999by theAmericanGeophysical Union. Papernumber1999JD900314. 0148-0227/99/1999 JD900314 $09.00
consists exclusively of Hgø, but somerecentreportshave demonstratedthe existenceof significanceamounts of dimethylmercury (DMM) in floodplainsoil air [Wallschl•t'geret al., 1995], and the emissionof volatile monomethylmercury compounds (MMM) from sewagesludgeamendedsoils [Carpi et al., 1997]. Establishingthe speciationof Hg fluxeswill be of fundamentalimportancefor understandingthe entire volatilization process,becauseeach specieswill be formedvia its
21,859
21,860
WALLSCHL•GER ET AL.: MEASUREMENT OF MERCURY EMISSIONS FROM SOILS
own characteristicbiogeochemicalpathways, which are, in turn, impacteddifferently by environmentalparameters. As there currently is no method that measuresHg airsurfaceexchangedirectly,we alsoneedto confinethe limits of experimentalerror associated with any of the competing indirect approachesused instead, which are mostly either micrometeorologicalmethods[Lindberg et al., 1995] or flux chambertechniques[Xiao et al., 1991]. Both approacheshave significant method-inherentadvantages and disadvantages [D. Wallschl•igeret al., Intercomparisonof different experimentalapproaches to quantifygaseousmercuryemissionsfrom contaminatedfloodplain soils to the atmosphere,submittedto Environmental Science and Technology, 1999b, hereinafter referred to as Wallschl•iger et al., submitted manuscript, 1999b], and even though there are somestudies that indicate fairly good agreement when both techniques are used simultaneously [Carpi and Lindberg, 1998; Poissant and Casimir, 1998], it has yet to be shown that both approaches are equivalent (or: which one is correct). In support of this doubt,a recent study of methanefluxes frompeatlandswith a static flux chamber technique and a micrometeorological methodreported similar averageand total fluxes,but showed large variations (up to a factor of 10) between individual simultaneousflux measurements[Chan et al., 1998]. In dynamicflux chambermeasurements, a defined area of the investigated surfaceis covered with a container of defined volume,andambientair is drawn through the chamberwith a known flushing rate. The concentration of the volatile compoundof interest (here Hg) is measuredat the chamber inlet and outlet, and the measuredHg flux J (typically
expressed in [ngm'2h'l])isthendetermined asJ= Acß Q / A, where
Ac
is
the
TGM
concentration
difference
between
chamberoutletand inlet (correctedfor the positiveor negative chamberblank, if necessary),Q is the chamberflushing rate, and A is the surfaceareacoveredby the chamber.Besidesthis generalprinciple,thereare not many commonfeaturesbetween individual flux chamberdesigns. Some investigations have
factorsdriving environmentalHg air-surfaceexchange.The field measurements were conductedat SteamboatSprings,an arid geothermalarea 15 km south of Reno, where surfacesoils are naturally enrichedin Hg due to the geothermalactivity [Gustin et al., this issue]. For the description of individual studies,we use U.S. date and military time notation throughout this manuscript.In this contribution, we comparethe performance of three individual flux chamberdesigns within close proximity of each other, and evaluate their qualitative and quantitative agreement,as well as their response to environmental changes. We also discuss the correlations betweenHg fluxes and environmentalparameters, and attempt to deduce some information about the underlying biogeochemicalprocesses drivingHg air-surfaceexchange. 2.
Methods
2.1. Experimental Site
During the Nevada STORMSexperiment,groupsoccupied sevendifferentsites in an areaof approximately200 by 200 m on both sides of a dry arroyo. Within a 20 by 20 m subplot south of that creek,G6teborg University, G6teborg, Sweden (GU), Frontier Geosciences, Seattle, Washington, United States (FGS), and GKSS Research Center, Geesthacht, Germany (GKSS) conducted parallel flux chamber measurements.The soils were arid sandy soils with scarcescrubby vegetation. They were very dry at the beginning of the experiment becauseit had not rained at the site for over 3 months,but there were several severe rainstormsduring the courseof the experiment. The study site was preselectedwith regardto apparenthomogeneity,which was probably valid with regardto meteorologicalparametersand vegetation,but subsequentchemicalanalysisrevealedsignificant differences
amongthe soils underthe individual measurement plots. For example,the soilsunderour3 flux chambers weresampledafter
theexperiment andhad Hg concentrations of 1.4-2.9•tgg'•
(Table 1), which provedto be significantlylower than those regardto blanks(which derivefromadsorptionor desorption determined at the siteswhereothergroupsmeasuredHg fluxes on the interior chambersurface),and it appearsthat Teflon during this intercomparison.For further informationon site [Kim and Lindberg, 1995] interactsless with gaseousH g and meteorology,consult an accompanyingoverview pub-
been conducted
on the choice
of the chamber material
with
than stainlesssteel [Xiao et al., 1991]. There are also reports
lication [Gustin et al., this issue].
thatemphasize theimportance of obtainingthebestpossible seal between the chamber and the investigated surface [Wallschl•ger et al., 1999b]. Most other important para-
2.2. Flux Chamber Techniques
meters, however, are left at the discretion of the individual flux
There were significant differences between the three individualflux chamberdesigns,and in the ways in which the
chamberdesigner,without any systematicinvestigationson how these parametersaffectthe measuredfluxes for any trace gas in general, or for Hg in particular. Examples of uninvestigatedkey factorsinclude the velocity, turbulence and flow pathof the chamberflushing stream,and the alteration of the meteorologicalconditions inside the chamber(versusthe unimpacted outside conditions) through the choice of chambermaterial;for example,the effectsof radiation exclusion and temperaturechanges. In September1996, nine international researchgroupsmet in Reno, Nevada, to intercomparetheir micrometeorological and flux chambertechniques for measuringHg exchange between soils and the atmosphere.This intercomparison exercise,referredto here as "Nevada Study and Tests of the Release of Mercury from Soils (STORMS)", also aimed at establishing the source strength of natural geogenic H g emissions,and at identifying someimportant environmental
flux measurements and calculations were conducted. GU and
GKSS used basically the same flux chamber [Carpi and Lindberg, 1998], which consistsof a thin transparentTeflon bag, sealedto the groundby a bottomskirt, and supportedby an external frame.Both groups introduced sufficient experimental modifications to the original setup, used during Nevada STORMS by the group from Oak Ridge National Laboratory(ORNL, Oak Ridge, Tennessee),to qualify their techniquesas similar,but differentfrom eachotherand from the original. In contrast,FGS used a very differentnovel type of flux chamber(introduced here for Hg flux measurements), consisting of a 3 mm thick Plexiglas pastry cover (SC25, Carlisle, OklahomaCity, Oklahoma).The cover has a bottom lip, which was pressed into the ground during the flux measurements to seal off the flux chamber with
the soil. Some
crudeexperimentsregardingthe airflow in this chamberwere
WALLSCHI,,•GER ET AL.: MEASUREMENT OF MERCURY EMISSIONSFROM SOILS
21,861
Table 1. Details of the Three ComparedFlux ChamberExperiments group
SoilHgcontent, ggg4 Chamber
Covered ground area,m2 Volume, L Inlet/outlet
holes
Flow direction Sealingto ground
Precleaning
FGS
GKSS
GU
1.4
2.9
2.5
Plexiglaspastrycover
Teflon bag
0.27
0.12
Teflon bag 0.12
28
24
24
9/2
1/1
4/1
front -• rear
front -• top
front --• rear
lip pressed in ground
stainlesssteelframe
lead clampsand
alkalinedetergent;DI water
flushingwith ambientair
aluminum
frame
acid;
Hg free air Detector
Analysismode Inlet/outletmeasurement cycle
1 Tekran-Analyzer
2 Tekran-Analyzers
on-line
on-line
in/in/out/outswitching
continuous
1 GARDIS
off-line in/out
alternating Inlet TGM
concentration
20
5
samplingheight,cm
Sampling flowrate,L min'l
1.5
1.5
Samplingvolume,L
7.5
7.5
Chamber flushing rate,L min-I
15.6
1.5
1.5 (flushing)/ 2.4 (sampling)
20
5
10
Temporalresolution,min
conductedwith water vapor generatedover liquid nitrogen, and seemedto indicatethat the flow was relatively parallel to the bottom,but not laminar.This poses a theoretical dilemma, because on one hand, the application of flux chambers demandsthat the flow in the chamberdoes not contain any verticalcomponent(laminar flow) [Gao et aL, 1997], while on the otherhand,turbulenceis assumedto be the major driving force behind environmental air-surfaceexchangeof gases. Thereforethe presenceof a certain (unwanted) degree of turbulencein the flux chambermay actually recreatesomeof the natural conditionsthat are excludedfromthe investigated surfaceby the chamber.For the purposeof an intercomparison like this, it is problematic, though, that this "artificial" turbulenceintroduced into the chamberdependsstrongly on the chosenflow rate and path, so that the air flow in each of the chamberswill have a differentdegreeof turbulence,which could have significant impacts on the measuredflows. The radiationabsorbanceof the Plexiglas and Teflon chambermaterials was also compared.The Plexiglas pastry cover (3 nm thick) transmitsvisible radiation with wavelengthsdown to 380 nm, but blocks radiation with shorter wavelengths (i.e., all UV radiation)effectively.The Teflon foil (0.2 rrrnthick) additionally transmits near-UV radiation, down to about
0.9 0.9 -4.5
250nm, and blocks only far-UV significantly. Table 1 summarizes potentially relevant experimental differences and similarities.
For
the TGM
concentration
measurements
at
the
flux
chamberinlet and outlet, GU used an automatedsinglechannel double-amalgamation dual-beam CV-AAS system (GARDIS l-A, Ekoservis,Vilnius, Lithuania) [Urba et al., 1995]. FGS and GKSS measuredTGM with automated dualchannelsingle-amalgamationCV-AFS systems(model 2537, Tekran Inc., Toronto, Canada; referred to
as "Tekran-
Analyzer" in the following) [Schroederet al., 1995]. All analyzers were repeatedlycalibratedautomaticallyor manually during the field campaign and showed no significant analytical drift. In a previous international field intercomparison, the two automated TGM-measurement methods (GARDIS and Tekran-Analyzer)have been comparedto each other,as well as with severalcorrespondingmanualmethods involving separatecollection and analysis; on average,all techniquesagreedwithin 20% or better [Ebinghaus et al., 1999]. Untreated ambient air was pulled through all flux chambersat a massflow-controlled flushing rate. TGM was collected through Teflon tubing (5-10 m long, 6 nan or 1/8 inch ID), and the tubinginlet was shelteredfromdust and
21,862
WALLSCHL•GERET AL.: MEASUREMENTOF MERCURYEMISSIONSFROM SOILS
rain with a glass funnel. FGS calculated Hg fluxes by subtracting the average of four inlet concentrations (two before and two after the corresponding outlet concentration was determined)from the averageof two outlet concentrations. Thus a flux value was calculatedevery 20 min and assigned to the time between the two outlet measurements.
GKSS
and G U
used simultaneous and alternating inlet/outlet concentrations,respectively,to calculatetheir Hg fluxes. 3.
Results
and
Discussion
3.1. Experimental Factors Affecting Flux Chamber Measurements of Mercury Air-Surface Exchange 3.1.1.
Flux
chamber
blanks.
In
flux
chamber
measurements, it is quite complicatedto assessthe impact of the chamber material
on the measured fluxes in
the
field.
Commonly,measuringthis "blank" involves sealing off the bottomof the chamberand then passing"ambient"air through it to determine whether the chambermaterial systematically adsorbs(negativeflux bias) or desorbs(positiveflux bias) the analyte of interest. This approachdepends on three critical assumptions, namely: (1) that the chamberinlet concentration remainsconstant,(2) that the chamberblank is independentof the inlet concentration (if it should vary), and (3) that no other external factorsinfluence the magnitudeof the chamber blank.
FGS measuredtwo chamberblank fluxes in the field by attachinga Plexiglas servingtray (PAL 235, Palmer Distributors Inc., Clinton Township, Michigan) to the pastry cover as
determined Hg fluxes because they were comparably insignificant. The same Teflon flux chambersemployed by GKSS and GU in this study have been used for investigating the Hg air-surfaceexchangeat uncontaminatedsites, for which
typical Hgfluxratesaround 5 ngm'2h'l arecurrently assumed. However, this required extensive precleaning by equilibrating the chamberwith uncontaminated ambient air for prolongedperiodsof time [Carpi and Lindberg,1998]. This indicatesthat the averageblank fluxes determinedin this study are mostly a result of the elevated TGM concentrations at the SteamboatSprings site, and can therefore not be directly transposedto studies at background sites. Thus the results of our blank flux determinationsneither prove nor preclude the suitability of the described flux chamber techniques for measurements of smallerHg fluxes. 3.1.2. Small-scale variability. To estimate small-scale inhomogeneitywithin the experimentalsite, GKSS and FGS moved
their
measurement
sites
on 9/3/97
at
1500.
GKSS
movedinto the exact sameplot whereFGS had been measuring before, and FGS moved into a new location close to GKSS'
old plot. For the 2 daysbeforeand after the site exchange,only the periods between 1500 and 0800 were compared,because they containedneither experimentalmanipulationsnor drastic meteorologicalchanges.The correlation coefficientsbetween the Hg fluxes measuredby one group at the two sites were 0.703 for FGS and 0.719 for GKSS, and the correlations between GKSS and FGS were 0.754 before and 0.880
after the
site exchange,which is much better than over the whole Nevada STORMS experiment.These data confirm that both a bottom plate.These fluxblanksaveraged 2.5 _+0.3 ngm-2h'• experimentaldesignsagreedwell qualitatively, and document (duringa periodof low Hg fluxes)and2.5 _+19.6ngm-2h'• that the Hg air-surfaceexchangepatterns at both sites were (during a high Hg emission period), which constituted less qualitatively very similar. However, after initially high and then decreasingHg emissions during the afternoons,both than 6 % of the averageHg fluxesmeasuredin the respective groups observedmostly Hg emission during the first night intervals two hours prior to and two hours after the blank (afterthe initial rain event), whereas mostly Hg deposition measurements, and GU obtained one field blank flux of was measuredduring the second night. The average ratios 2.4+ 1.3ngm'2h'• in a similarmanner. However,dueto the unusual Hg air-surfaceexchangecharacteristicsat the Steam- FGS(beforemove)to FGS(aftermove)andGKSS(aftermove) to boat Springs site, the ambient TGM concentration varied GKSS(beforemove) were 0.46 and 0.67, respectively,indicat-
ing that the originalGKSS site emittedmoreHg than the old FGS site, and demonstratingthat the small-scaleregional inwhich the second FGS blank measurementwas performed, which directly violates assumption (1), even though the homogeneityof the measuredHg fluxeswas of the order of chamberwas elevated to approximately50 cm in order to 100%. The meanof thesetwo ratios(0.57) matchesthe ratio of reducethe intrusionof near-surface air masseswith high TGM the soil Hg contents(0.48) relatively well, suggestingthat concentrations. The much highervarianceof the secondblank the differentHg fluxes are primarily causedby varying H g levels in the soils. measurement (comparedto the first) also seemsto suggestthat 3.1.3. Qualitative and quantitative comparison. Another assumption(2) was not valid. Finally, GKSS performed several blank determinationsduring the course of the way of separatingnatural variability fromexperimentaldifferexperimentand noticed that the chamberblank seemedto de- ences is to look at the TGM concentrations that the three pendstronglyon radiation/temperature; they measuredblanks groupsmeasuredthroughoutthe campaign(Figure 1) because around 0.5ngm'2h'• atnight,butup to 2 ngm'2 h4 during those shouldbe independentof the used flux chambersetups. daytime. A similar effect has been reported before for a Generally, GKSS measured the widest range of TGM
strongly (between10 and30 ngm'2h-•) duringtheperiodin
Plexiglaschamber[Wallschl•iger et al., 1999b], and probably just illustratesthat the air-surfaceexchangeof Hg on the chambermaterialis governedby similar factorsas at the airsoil boundary,which appearsto invalidateassumption (3). Consequently,the measured"blanks" maynot be entirely suitableto accountfor the interaction of the volatile H g compounds with the chambermaterials,althoughthey maybe usedas indicative values.The materialsalone appearto have
concentrations (3.5-180:average 32.6 + 27.6 ngm'3),while
both othergroupsmeasuredlower peak concentrations(FGS:
2.1-108;average 28.6+ 22.9ngm-3),andGU alsomeasured higher
background concentrations (5.8-101;
average
15.7+12.1ngm'3).ElevatedTGM concentrations wereob-
served around midday, which coincides with high H g emissionsfrom the soils,and during nighttime periods,which is probably causedby the formationof a nocturnal inversion introduced experimental artefacts of between 1 and5 ngm'2h-1 layer close to the ground [Gustin et al., this issue]. TGM measuredby GKSS and FGS agreedvery well intothe daytimeHg fluxesmeasured at SteamboatSprings,but concentrations noneof the three groups subtractedthese "blanks" fromtheir throughout the whole experiment (solid regression line;
WALLSCHL•GER ETAL.'MEASUREMENT OFMERCURY EMISSIONS FROMSOILS
21,863
140
120
ß
FGS
o o
GU day (7:00-20:00) GU night(20:00-7:00) linearregression FGS data
- - - linearregression GUnightdata o
100
ß
80
o o
$ooß 40
20
0
0
20
40
60
80
100
120
140
160
180
TGM (GKSS) [ng/m3] Figure 1. Correlation betweenthe TGM concentrationsmeasuredby the three groups during the Nevada STORMS intercomparison.
r=0.87), and the ratio TGM(FGS)/TGM(GKSS) averaged 0.97 + 0.42, with individualvaluesrangingfrom 0.32 to 3.14. AlthoughTGM concentrations are not a direct measureof H g fluxes,they characterizethe overall Hg air-surfaceexchange behavior to some extent, if the investigated surface is relatively homogeneous. Consequently,we could estimatethe averagenaturalvariationbetweenthe GKSS and FGS plots to be of the order of 50%, but individual variations being as high as 300%. By comparison,the GU TGM data show a completely different distribution, in that they do match the elevated concentrationsduring daytimeHg emissions(Figure 1; open circles), but fail to rise during the assumed nighttime inversion situations (Figure 1; gray circles). We have to assumethat their way of measuringambient TGM (at 20 an height on top of the chamber)yields different results than measuringdirectly at the chamberinlet (like GKSS and FGS) because TGM concentration can be extremely heightdependent[Wallschlgigeret al., 1999a], especiallyif upward mixing is suppressed by the formationof a nocturnalinversion layer. Alternatively, their samplingsite could have been less impactedby Hg-contaminated air masses;for example,the tent in which all three groupsran their analyzers(located north of the GU site andwestof the GKSS and FGS sites) could have blockedthe advectivetransportof Hg fi'omlocal sources(hot springs)to the GU site. Quantitatively,the GU TGM data fall into two groups: midday Hg emissionperiods, where their resultscomparereasonablywell with the othergroups,and all
outlets weretypically of theorderof 10ngm-3forlowHg flux situations andwereashighas1000ngm4 duringsome peak flux periods.Due to the higher chamberflushing rate, TGM concentrationdifferencesfor the FGS setupwere about 1 order of magnitude lower, but still high enough to exclude analytical uncertainty as a major source of error in all three groups'flux measurements. The moststraightforwardway to intercomparethe Hg fluxes measuredby the three groups is linear correlation analysis (Figure2 and Table 2). For this purpose,the simultaneousflux measurements had to be assigned to commonpoints in time (x values) by running time-weighted averaging,becauseall groupshad differenttemporalresolution (Table 1), and GU additionally started their individual flux measurementsat slightly differenttimesthan GKSS and FGS. As the lowest commondenominator,FGS' temporalresolutionof 20 min was chosenfor this time normalization.Higher temporalresolution cannot be used, because GKSS and GU have theoretical chambervolume turnover times of up to 16 min, and lower
suggeststhat the natural variability of the derived Hg fluxes couldbe even higher than estimatedfromthe GKSS and FGS TGM data. However, since this analysis demonstratesa completelydifferentdiurnal TGM profile for GU, we have to considerthat experimentalfactorscould be superimposedon the natural variation. Finally, typical TGM concentration
temporalresolution should not be used,becauseit smoothes out the impactof fast environmentalchanges(suchas the "flux dip" beforethe first rain event, describedin section 3.2.2). Correlations between all three completedata sets (Table 2, first three lines) were statistically significant at the 99.9% I•'.-,vel, so all three groupsqualitatively measuredthe sameH g flux trends during the experimentalperiod. As expected,the two similar flux chamberdesigns (GKSS and GU) agreed betterwith each other than with the apparentlydifferentFGS technique. The slopes of all three correlation curves are noticeably different from one, suggesting quantitative differencesbetweenthe three experiments. Two of the reasons for these statistical discrepancies becomeapparentin Figure2. First, GU - contraryto FGS and GKSS - did not detectHg depositionevents(negativefluxes), which is caused by their lack of elevated ambient TGM concentrations duringthe night (as explainedabove),so that TGM at their chamberoutlet was always higher than ambient TGM. Second,FGS measuredlong periods of Hg emissions
differences between
> 200ngm'2h'l (especially afterthefirstrainevent,shownin
otherperiods,whereGU measured TGM around15 ngm'3 totally uncorrelatedto the other groups.Consequently,the overall ratio TGM(GU)/TGM(GKSS) averaged0.92 + 1.08, with individual
values between 0.08 and 6.55, which
the GKSS
and GU
chamber
inlets
and
21,864
WALLSCHL3.GER ET AL.' MEASUREMENTOF MERCURYEMISSIONSFROM SOILS
Hgflux[ng/m•h]
-m--FGS
500--
a
..... * ....GU
400-
300-'
IJ
'"
IJ
IJ
•'
.......
.......... ' •
GKSS Hg fl•
[n•m2'h]
-•0
Figure 2. CorrelationbetweenHg fluxesmeasured by GU and FGS duringthe Nevada STORMSexperiment and those determinedby GKSS (all data sets are normalizedto 20 min temporalresolution, and only experimentallyunmanipulatedflux situationsare used).
Figure4), whereasGKSS and GU did not detectany fluxes secondsetof correlation analyses was performed (Table2, last
above200 ngm'•- h'l, despitemeasuring on soils that threelines),whichwasrestricted to fluxes radiation > wind direction (as A from 180ø) > T(soil) > A T (soil - air) > T(air) > relative humidity, both with regard to correlation coefficientsand to number of groups that found significant correlations. This result is somewhatsurprising, becausealthough wind and radiation (as shownin section3.2.3) are significantfactorsaffectingthe Hg fluxes,flux chambersexcludemostof the wind, and at least
certain regions of the radiation spectrum.These findings suggest that the Hg fluxes measuredwith flux chambersto somedegreereflectprocesseshappeningin soils adjacentto
theflux chamber. Elevatedsoil temperature increasesboth the vapor pressures of volatile Hg compounds and their desorption from the solid phase. Thereforethe correlation betweenthe Hg fluxes(mostlyemissions)and higher soil temperatures,or higher temperaturegradients between soil and air, appearreasonable.In contrastto all other correlations, the relationbetweenHg flux and relative humidity was negative,
the questionwhethersomefeatureof the micrometeorological approaches preventsaccurateresponseto environmentalvariations, or whether the flux chambersartificially create correlationsthat do not exist in the unperturbedenvironment, which would renderthe entire previous discussionobsolete. Thus it is of fundamentalimportanceto answerthis question in orderto understandthe factorsdriving the air-surfaceexchangeof trace gasesin the environment. 3.2.2. Response of flux measurements to rain events. Right at the beginning of the Nevada STORMS intercomparison, a heavy rainfall event with a precipitation
ratearound10L m-2(equalto 1 cm)[Lindberg et al., this issue] occurred on 9/2/97 from about 1200 to 1300. This event altered
the
environmental
situation
at the site com-
pletely becausethe originally arid desertsoils were now very wet. All threegroupsnoticeda strongand immediateresponse (i.e., with the nextflux measurement; between5 and 20 min) of the observedHg fluxes to the rain event (Figure4). Fluxes
before therainwerearound 50ngm'2h'•, whichis comparable
which suggeststhat the transportof volatile Hg compounds tomidday Hg emissions fromthedrysoilsof20-60ngm-2h'l, could be coupledto a water vapor gradient between soil and
measuredby FGS and GKSS on the previous day. Both FGS and GU measureda pronounceddip of the Hg fluxes just As expected,wind directiondid not have a straightforward beforethe rain event, which may have been caused by the correlationwith Hg fluxesbecausethe spatialangleis not a observedrainclouds moving in, which blocked the sunlight linear variable. However, it was noticed that most of the and thereby reduced radiation and temperature.This "flux higher emissionfluxes were observed when the wind came dip" was accompanied by a sharp increasein ambientTGM from anglesaround180ø (south).So, when the wind direction concentrations (measuredat the inlet of the FGS chamber)from is expressed as the differenceto 180ø, the Hg fluxesmeasured 5.8ngm-3at 1215to 24.4ngm4 at 1220,coinciding approxiby all three groups correlatesignificantly with it, meaning matelywith the startof the rain, while the TGM time profile at that the highestHg fluxeswere observedduring periodsof the chamberoutlet tracked the overall Hg flux behavior. This southerly winds. Analysis of the TGM concentration data suggeststhat Hg emissionsfromthe soil coveredby the flux air.
21,868
WALLSCHL•GERET AL.' MEASUREMENTOF MERCURYEMISSIONSFROM SOILS
chamber decreasedwhen the clouds moved in, and then started
checkedby introducinga moisturesensorin the flux chamber. increasedstronglyduring the rain event, with a time lag in the Artificial irrigation of the soils under the flux chambers responsecomparedto the outside soils. Peak fluxesmeasured (which had not yet been in direct contact with water) with at the endor afterthe rain event (1300-1400) ranged fi•om150 deionizedwater(1.2 L per plot, or 10 L m-2, which is to 500 ngm-2h-1,meaning a 3 to 10-foldincrease. Thisin- comparableto the discussed rain event) caused dramatic crease in Hg fluxes was matched by TGM concentrations increasesin Hg emissions[Lindberget el., this issue].The H g risingfrom5-20ngm'3upto 100ngm'3(FGSandGU) or 180 fluxes measured by GKSS rose from around 50 up to
ngm'3(GKSS).A second andlighterrainevent(99.5%) with each other, sensor in the flux chamber. The fact that the FGS chamber but not with the micrometeorologicaltechniques. Quantishows an even stronger correlation with radiation than the tatively, there were significant differences between the two Teflon chambers(Table 3), even though it blocks UV individual flux chambermethods,due to site inhomogeneity radiation completely,suggeststhat visible or infraredradi- and experimentaldifferences.Finally, the quantitative agreement with the micrometeorologicaltechniques was even ation might be responsible for the process.However, this worse; on average,the micrometeorologicalapproachesestiargumentonly holds up if the reactionto sunlight is taking place exclusively inside the chamber, but the previous matedHg fluxes about 4 times higher than the flux chamber sections demonstrate that the flux chambers also seem to reflect techniques.It has been suggestedthat this discrepancyis causedby approach-inherent differences,which will causeflux someprocessesgoing on in the surrounding soils. This is chamber techniques to underestimate and/or micrometemade even clearer by the flux chamberdata from AES, who useda totally opaquechamber,and still observedthe shading orological methodsto overestimatethe fluxes [WallschNiger et al., 1999b]. However, sincemuch better agreementhas been effectanda strongcorrelationwith radiation [Poissant et al., this issue]. observed in other intercomparison studies [Carpi and 1300, the shadingwas extendedadditionally to a 30 cm wide strip of soil surroundingthe chambers(so, in the caseof the
3.3. Mercury Volatilization at SteamboatSprings
During the Nevada STORMS intercomparison, an experimentallyunperturbed24-hour-diurnal cycle of the H g
air-soil exchangewas measured by all participatinggroups
Lindberg, 1998; Poissant and Casimir, 1998], it is also possiblethat the observedbad correlationis causedby problems thatare specificto the Steamboat Springssite. Studiesover "background"sites, definedas locations that are (apparently)not influenceddirectly by any natural or anthropogenic Hg sources,indicatethat the magnitudeof the
21,870
WALLSCHL•GER ET AL.: MEASUREMENT OF MERCURY EMISSIONS FROM SOILS
current environmentalHg air-soil exchangeis of the order of
comparison to a referencechambercould be used to study the
5 ngm-2h'l [Kimet al., 1995;Carpiand Lindberg,1998; influencesof experimentalparameters;this should be much Poissantand Casimir, 1998]. By comparison,the averageH g fluxes measured during this study are about 1 order of magnitudehigher, which roughly matchesthe ratio of total H g
easier on waters than on soils.
Additionally,our experiencesduringthe Nevada STORMS experiment lead to the following recommendationsfor concentrations in the investigated soils(1-5 lagg-l)comparedimprovingfuture studiesof Hg air-surfaceexchangewith flux to thosein background soils(-0.1 lagg-l) [Andersson, 1979]. chambers.When intercomparingdifferent techniquesin the Individualfluxevents,rangingfrom-50 to +500ngm-2h'l, field, it is critical that all groups follow the samesampling were as much as 2 orders of magnitude above that assumed protocol,so that averagingand temporalnormalizationdo not background.Those peak periodsrepresentsomeof the largest introduce artificial variability or distort the responseto Hg emissions ever measuredon not directly anthropoge- environmental changes. The (common) temporal resolution nically impactedsites.Under the assumptions that the average should be as high as experimentallypossible, which can be systems Hgfluxduring theexperiment was200ngm'2h'l (whichis in achievedby using two separateTGM measurement between flux chambersand micrometeorologicaltechniques), monitoring the inlet and outlet concentrations.Higher and that a seasonalcycle cuts that value in half on an annual chamberflushingrateswill also indirectly improvetemporal average,just the part of the SteamboatSpringsareaon which resolution because they decrease the response time to Nevada STORMS was conducted would emit around 35 kg (sudden) environmental changes, and thereby reduce the (obtained before Hg y-1.Thispointsoutthatnatural Hg sources cancontribute numberof "intermediate"flux measurements significantly to regional atmosphericHg budgets and em- at leastone flux chambervolumehas been exchanged), which phasizesthe need to study their contribution to global H g should be discarded. cycling in comparisonto direct and indirect anthropogenic The measurement of Hg air-surfaceexchangeover an arid sources[Ebinghauset al., 1998]. geothermalarea established a strong net emission, which emphasizes the importanceof accountingfor these processes in regional and globalHg budgets.Experimentalshadingand 4. Conclusions and Outlook rainfall events strongly influencedthe measuredHg fluxes, This field intercomparison of threeflux chamberexperiments even though the soil under the chamberwas not directly for measuring air-surface exchange of mercury showed wetted. This suggeststhat the footprint areaof a flux chamber relatively good qualitative agreement,but also revealedsome actuallyextendsbeyondthe coveredarea, probablybecauseof quantitative discrepancies, especially during deposition soil air diffusion within the soil body. Several highly periods and high emissionevents. The study points out that significant correlationswere identified between the measured flux chambermeasurementsappear to have the same large Hg fluxesandenvironm6ntal parameters, although it hasto be variability that is generally associatedwith micrometeoro- emphasizedagain that both the experimentalsite and the logical approaches or with diffusion-based models. Our prevailing meteorologicalconditionsduring the study were results suggestthat the perception of flux chambermeasure- quite differentfromother studies of environmentalHg airmentsbeing moreaccuratethan the other techniquesis, if at surfaceexchange,so that the quantitative conclusionsand all, only true for one particular flux chamberon one defined qualitative correlationscannotbe transposedeasily to other experimentalplot, but that diurnal variancein the Hg air-sur- sites. The most pronounced correlations exist with wind face exchangeprocess, geographic variability, and experi- speedand radiation, and transmissionspectraof the different mental differencesintroduce significant uncertainty. The des- chambermaterialssuggestthat visible or infrared radiation are cribed experimentsdemonstratethat the used flux chamber the primary factorscausing this effect.In the case of wind techniques are capable of measuringenvironmental Hg flux speed,it is not obvious why flux chamberexperimentswould processesat a near-real time temporal resolution. They be affectedby this parameter,and since there are strong respond sufficiently fast to environmental changes and intercorrelations between most of the measured environmental experimentalmanipulations to allow observation of natural parameters,caution has to be used here to distinguish processesand parametrical"in situ" investigations. Their betweencoincidental(artificial) correlationsand real causeapplicability to measuring 'Hg air-surface exchange in effect relationshipsresulting fromunderlying biogeochemical uncontaminatedecosystemswas not demonstratedin this reactions. These uncertainties should be addressed in future study, but this appearsfeasible, if appropriatemeasuresare field studies, and because it is not possible to control taken to assure low chamber blanks. environmentalparametersin those experiments,they should These studies also indicate that purely experimental be conductedby performingcontrolled manipulationson an parametersinfluence the determination of Hg air-surface experimental chamber, while monitoring environmental exchangewith flux chambers.For example,somepreliminary changeswith a referencechamberon a parallel plot. Finally, resultspresentedhere suggestthat the measuredHg fluxes and flux studies should be extendedto individual Hg species,in the correlationswith environmentalparameterswere higher order to identify all mechanismsdriving the Hg air-surface when the chamberwas flushed at elevated flow rates; this exchangeprocessin different ecosystems,and to assessthe finding could be critical for the design of future flux chamber importanceof theseindividualspeciesfor regional and global experiments.To resolve the question as to what extent atmospheric Hg budgets and for ecotoxicological experimental parameters,like, for example, the chamber considerations. flushing rate or the chambermaterial,affectthe measuredflux, we suggestconductingfuture parallel studies with multiple chambers. For that purpose,experimentalplots with identical Acknowledgments.We thank Dan Schneeberger and Frank (or at least parallel) air-surfaceexchangebehaviors must be Schaedlich(Tekran,Toronto,Canada)for valuableassistance with their identified,and then the behavior of a manipulatedchamberin analyzersat•erthe powerfailure problems.Steve Lindberg(Oak Ridge
WALLSCHL•GERET AL.: MEASUREMENT OF MERCURYEMISSIONSFROMSOILS National Laboratory, Oak Ridge, Tennesee) and Mae Gustin (UNR, Reno, Nevada) did an excellentjob organizing this field intercomparison and the accompanyingworkshop, and Mae's studentsare thankedfor their assistance duringadverseweather-periodsof the field measurements. We are gratefulto Laurier Poissantand his group (AES, Montr6al, Canada) for permittingus to use their micrometeorological data in the preparationof this manuscript,and to Steve Lindberg for supplyingthe reference data for the flushingrate experiment.Finally, we appreciatethe financial supportprovided the Electric Power ResearchInstitute,PaloAlto, California(EPRI), and stimulatingpromotion by Mary Ann Allen and Don Porcella(EPRI), withoutwhich thisstudy and the associatedworkshopwould have been impossible. Finally, we acknowledgethe criticalcommentsof two anonymousreviewers,which led to significantimprovements overan olderversionof thismanuscript.
21,871
Kim, K. H., and S. E. Lindberg,Design and initial testsof a dynamic enclosure chamber for measurementsof vapor-phase mercury fluxesover soils, WaterAir Soil Pollut., 80, 1059-1068, 1995. Kim, K. H., S. E. Lindberg, and T. P. Meyers, Micrometeorological measurements of mercuryvaporfluxesover backgroundforest soils in EasternTennessee,Atmos.Environ., 29, 267-282, 1995. Lindberg, S. E., K. Kim, T. P. Meyers, and J. G. Owens, Micrometeorological gradientapproachfor quantifyingair/surface exchangeof mercuryvapor:Testsover contaminatedsoils,Environ. Sci. Technol.,28, 126-135, 1995. Lindberg,S. E., et al., Increasesin mercuryemissions from desertsoils in response to rainfallandirrigation,d. Geophys.Res.,thisissue. Poissant,L., and A. Casimir,Water-air and soil-air exchange rate of total gaseous mercury measured at background sites, Atmos. Environ., 32, 883-893, 1998.
References Advokaat, E. M., and S. E. Lindberg, Effect of rainfall exclusionon air/surface exchange rates of mercury over forest soils, paper presentedat the 4th InternationalConference on Mercury as a GlobalPollutant,Hamburg,Germany,Aug. 4-8, 1996. Amyot, M., G. Mierle, D. R. S. Lean, and D. J. McQueen, Sunlightinducedformationof dissolvedgaseousmercury in lake waters, Environ. Sci. Technol.,28, 2366-2371, 1994. Andersson,A., Mercury in soils, in The Biogeochemistryof Mercury in theEnvironment,editedby J. O. Nriagu, pp. 79-112, Elsevier,New York, 1979. Carpi,A., and S. E. Lindberg,Applicationof a TeflonTMdynamicflux chamberfor quantifyingsoil mercury flux: Tests and resultsover backgroundsoil,Atmos.Environ.,32, 873-882, 1998. Carpi, A., S. E. Lindberg, E. M. Prestbo,and N. S Bloom, Methyl mercury contaminationand emissionto the atmospherefrom soil amendedwith municipalsewagesludge,d. Environ.Qual., 26, 16501655, 1997. Chan,A. S. K., J. H. Prueger,and T. B. Parkin, Comparisonof closedchamber and Bowen-ratio methodsfor determiningmethane flux from peatlandsurfaces, d. Environ.Qual., 27, 232-239, 1998. Ebinghaus,R., R. M. Tripathi, D. Wallschlager,and S. E. Lindberg, Natural and anthropogenic mercurysourcesand their impacton the air-surface exchangeof mercury on regional and global scales, in Mercury ContaminatedSites -- Characterization,Risk Assessment and Remediation,edited by R. Ebinghauset al., pp. 3-50, SpringerVerlag, New York, 1998. Ebinghaus, R., et al., Internationalfield intercomparison measurements of atmosphericmercury speciesat Mace Head, Ireland, Atrnos. Environ., 33, 3063-3073, 1999. Gao, F., S. R. Yates, M. V. Yates, J. Gan, and F. F. Ernst, Design, fabrication, and application of a dynamic flux chamber for measuringgas emissions from soil,Environ.$ci. Technol.,31, 148153, 1997. Gustin,M. S., et al., The Nevada STORMS project:Measurementof mercuryemissionsfrom naturallyenrichedsurfaces,d. Geophys. Res., this issue.
Poissant, L., M. Pilote,andA. Casimir,Mercury flux measurements in a naturallyenrichedarea: Correlationwith environmentalconditions duringtheNevadaStudyandTestsof the Releaseof Mercury from Soils(STORMS),d. Geophys.Res.,thisissue. Schroeder, W. H., and R. A. Jackson, Environmental measurements
with an atmospheric mercurymonitorhavingspeciationcapabilities, Chemosphere, 16, 183-199, 1987. Schroeder,W. H., C. Lamborg,D. Schneeberger,W. F. Fitzgerald,and B. Srivastava,Comparisonof a manual method and an automated analyzerfor determiningtotal gaseous mercuryin ambientair, paper presented at the 10thInternationalConferenceon Heavy Metals in the Environment,Hamburg,Germany,Sept. 18-22, 1995. Urba, A., K. Kvietkus,J. Sakalys,Z. Xiao, and O. Lindqvist,A new sensitiveandportablemercuryvapor analyzerGARDIS-1A, Water Air Soil Pollut., 80, 1305-1309, 1995.
Wallschlager,D., H. Hintelmann,R. D. Evans, and R. D. Wilken, Volatitizationof dimethylmercury andelementalmercuryfrom river Elbefloodplainsoils,WaterAir SoilPollut.,80, 1325-1329,1995. Xiao, Z., J. Munthe,W. H. Schroeder,and O. Lindqvist,Vertical fluxes of volatilemercuryover forestsoil and lake surfacesin Sweden, Tellus,Ser.B, 43, 2.67-279,1991. J. London,R. R. Turner and D. Wallschlager,FrontierGeosciences Inc., 414 Pontius Ave. N, Seattle, WA 98109, USA (email
DirkW•Frontier.WA.com) R. Ebinghausand H. H. Kock, Instituteof Physicaland Chemical Analytics,GKSS ResearchCenter,Max-Plank-Str.,21502 Geesthacht, Germany(email ralf.ebinghaus•gkss.de) J. Sommar and Z. Xiao, Department of Inorganic Chemistry, University of G6teborg, 412 96 G6teborg, Sweden (email sommar@inoc. chalmers.se)
(ReceivedJanuary14, 1999;revisedApril 19, 1999; acceptedMay 7, 1999.)
