Atmospheric Mercury in the LakeMichigan Basin: Influence of theChicago/Gary Urban AreaM A T T H E W S . L A N D I S , * , †

A L A N F . V E T T E , † A N D G E R A L D J . K E E L E R

The University of Michigan Air Quality Laboratory,Ann Arbor, Michigan 48109

The relative importance of the Chicago/Gay urban areawas investigated to determine its impact on atmosphericmercury (Hg) concentrations and wet deposition in the LakeMichigan basin. Event wet-only precipitation, totalparticulate, and vapor phase samples were collected forHg, and trace element determinations from five sites aroundLake Michigan from July 1994 through October 1995 aspart of the Lake Michigan Mass Balance Study (LMMBS).In addition, intensive over-water measurements wereconducted aboard the EPA research vessel Lake Guardianduring the summer of 1994 and the winter of 1995 aspart of the Atmospheric Exchange Over Lakes and OceansStudy. Atmospheric Hg concentrations were found to besignificantly higher in the Chicago/Gary urban areathan surrounding sites: Hg in precipitation was a factorof 2 and particulate Hg was a factor of 6 times higher. Over-water measurements found elevated Hg concentrations19 km off shore of Chicago/Gary suggesting an enhancednear field atmospheric deposition to Lake Michigan.Meteorological transport analyses also determined thatlocal sources in the Chicago/Gary urban area significantlyimpacted all of the LMMBS sites indicating a broadimpact to the entire Lake Michigan basin.

IntroductionMercury (Hg) is a toxic pollutant of concern in aquaticecosystems because of its ability to bioaccumulate up thefood chain and its demonstrated link to human health effectsfrom consuming contaminated fish. Atmospheric depositionis widely recognized as an important link in the cycling ofHg in the environment (1-5) and has been implicated as theprimary pathway for inputs of Hg to Lake Michigan (6-8).Consequently, Hg has been identified as a critical pollutantfor study and has been specifically targeted in the Great LakesWater Quality Agreement Amendments of 1987 and Section112(m) of the Clean Air Act Amendments of 1990 (“GreatWaters Provision”).

On a global basis, it is estimated that between 50 and 75%of total atmospheric Hg emissions are of anthropogenic origin(9, 10). Natural emissions are typically assumed to beelemental gaseous Hg0 (11). Anthropogenic emissions areprimarily Hg0, divalent reactive gaseous mercury (RGM), andparticulate Hg (Hg(p)). The dominant form of Hg in theatmosphere is Hg0 (12). Because it is relatively insoluble anddeposits very inefficiently, the mean residence time for Hg0

in the atmosphere is estimated to be approximately 1 year(1, 13) allowing for truly global circulation. Any RGM that isdirectly emitted to the atmosphere is expected to depositefficiently on a local or regional scale near major sourceslargely because of its solubility. Atmospheric deposition atany particular location can, therefore, be a complex com-bination of local, regional, and global emissions and trans-port/transformation processes (10).

The importance of local, regional, and global sources ofmercury to observed atmospheric deposition is a topic ofcontentious debate. Representatives from the Electric PowerResearch Institute purport that Hg pollution is a globalproblem in which the U.S. plays a minor role, emitting only158 tons year-1 of the estimated 2000-6000 tons year-1

emitted globally (14). The contamination of remote lakesand arctic ice cores has been attributed to long range andeven global transport of atmospheric Hg (15). However,researchers in the U.S. and Europe have observed significantspatial gradients in atmospheric Hg deposition around urbanand industrial areas indicating local anthropogenic influences(7, 16-18). Dated sediment cores from the Great Lakesindicate that annual net deposition rates increased from7-fold in Lake Erie to over 100-fold in Lake Ontario frompreindustrial times to their peak in the mid 1940s (19). Pirroneet al. (19) reported the Hg deposition rate for Lake Michiganincreased by a factor of 15 from the early 1800s to its peakin 1946. The relative importance of atmospheric depositionand direct discharges into the Great Lakes or their tributariescontributing to the observed sediment core results areunknown.

Most recent estimates of atmospheric deposition of traceelements to the Great Lakes have been made using data fromthe Integrated Atmospheric Deposition Network (IADN) (20,21). One sampling site was established on each of the GreatLakes to monitor regional trends and estimate atmosphericdeposition. The sites were established in remote locationsto avoid being unduly impacted by local sources. The IADNsite for Lake Michigan is located in the northeastern part ofthe lake in the Sleeping Bear Dunes National Lakeshore. Themaximum local source density near Lake Michigan occursalong its southwestern shoreline, which is dominated by thegreater Chicago, IL and Gary, IN urban area. With apopulation of over 8 000 000, this is the third largestmetropolitan area in the country. A 4-week intensive studydemonstrated that the Chicago/Gary urban area had asignificant impact on the dry deposition of Hg into southernLake Michigan (7), and a multisite monitoring study con-ducted in Michigan suggested the impact of the Chicago/Gary urban area was also observed when elevated concen-trations of Hg in precipitation were found to be associatedwith air transport from that region (16).

Major anthropogenic Hg sources in the Lake MichiganBasin and preliminary estimates of their annual emissionsinto the atmosphere have recently been reported (22). Sourcesinclude fossil fuel utility boilers, municipal and hospital wasteincinerators, iron and steel production, coke production, limeproduction, hazardous waste recycling facilities, and sec-ondary copper and petroleum refining. However, the sourcesof Hg are numerous and many are not well characterized.As a result, an accurate emissions inventory that includesspeciated anthropogenic as well as natural Hg sources is stillnot available. This reality, coupled with an incompleteunderstanding of atmospheric processes for Hg, limits thepresent reliability of deterministic models in predicting theatmospheric behavior and deposition of Hg over shorttemporal and large spatial scales (11, 16). To investigate

* Corresponding author phone: (919)541-4841; fax: (919)541-1153;e-mail: landis.matthew@epa.gov.

† Current address: U.S. EPA, National Exposure Research Labora-tory, Research Triangle Park, NC 27711.

Environ. Sci. Technol. 2002, 36, 4508-4517

4508 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 21, 2002 10.1021/es011216j CCC: $22.00 2002 American Chemical SocietyPublished on Web 10/08/2002

transport and deposition of Hg in the Lake Michigan Basin,accurate, long-term measurements at multiple receptorlocations were needed.

Two studies were conducted concurrently to investigatethe role of atmospheric Hg deposition to Lake Michigan:the Lake Michigan Mass Balance Study (LMMBS) from July1994 through October 1995 and the Atmospheric ExchangeOver Lakes and Oceans Study (AEOLOS) from May 1994through January 1995. This manuscript summarizes ourmeasurements of Hg in precipitation, Hg(p), and total vaporphase Hg (Hg(v)) in the Lake Michigan region from theLMMBS network of land-based sites and over-water mea-surements made aboard the EPA research vessel LakeGuardian during AEOLOS. In addition, a meteorologicalcluster analysis will be presented that demonstrates thesignificant impact of the Chicago/Gary urban area onatmospheric Hg concentrations across the entire LakeMichigan Basin.

Study DesignSite Descriptions. Five sampling sites were chosen for theLMMBS (Figure 1). Four sites were located around the lakein Kenosha (KEN), WI (42.50°N, 87.81°W); Chicago (IIT), IL(41.83°N, 87.62°W); Sleeping Bear Dunes (SBD), MI (44.76°N,86.06°W); and South Haven (SHN), MI (42.46°N, 86.16°W).The fifth site was located in Bondville (BON), IL (40.03°N,88.22°W). The BON site was considered an “upwind back-ground” location since there were no nearby sources and

the predominant wind direction was from the southwest (23).The IIT site was located on the roof of a four story buildingon the Illinois Institute of Technology campus in southeastChicago about 1.6 km west of the lake. This area of the cityis mixed commercial and residential. There is heavy urbanand industrial development in all directions from the site,with the heaviest concentration 10-20 km to the southeastin Chicago and northwest Indiana (Gary). The KEN site waslocated approximately 200 m from Lake Michigan in a largeopen field about 5 km south of Kenosha. The surroundingarea was semirural, with mostly residential and some lightcommercial activity. The SHN site was located 7 km east ofLake Michigan and 12 km northeast of South Haven in anopen field. The surrounding area was rural, dominated bygeneral agriculture. The SBD site was located about 1 kmeast of the lake in a large open field on a secondary duneapproximately 5 km south of Empire, MI. The surroundingarea is rural, dominated by wooded areas and generalagriculture. Site operators were hired and trained to collectsamples according to the University of Michigan Air QualityLaboratory (UMAQL) ultraclean protocols (24).

The AEOLOS utilized the five LMMBS land-based sites,an additional urban site at George Washington High School(GWS) in Chicago, IL (41.68°N, 87.53°W), and two over-watersites (42.00°N, 87.42°W; 41.77°N, 87.33°W) (Figure 1). TheGWS site was located on the roof of the school approximately18 km southeast of the IIT site in the heavily industrializedarea of Chicago. Over-water measurements were madeaboard the Lake Guardian from two stations approximately19 km off shore of Chicago. All ambient samples werecollected aboard the Lake Guardian approximately 5 m abovethe water surface from a retracting boom extended 2 m fromthe bow of the ship. The ship was stationary during samplingand anchored into the wind to prevent contamination fromonboard activities or engine exhaust.

Data Description. Precipitation samples were collectedfrom July 1, 1994 through October 31, 1995 during the LMMBSusing automatic wet-only precipitation collectors. Sampleswere collected on an event basis from April through Octoberand on a weekly basis from November through March toconstrain analysis costs. The event and weekly precipitationcollection methods were previously found to be equivalent(25). Precipitation samples were analyzed for Hg and a suiteof trace elements including copper (Cu), zinc (Zn), strontium(Sr), and lead (Pb).

Twenty-four h integrated total aerosol and Hg(v) sampleswere collected every sixth day from July 1, 1994 throughOctober 30, 1995 during the LMMBS. Sampling began at8:00 a.m. local time and ended the following morning at8:00 a.m. local time. During AEOLOS, 12-h total Hg(p) andHg(v) samples were collected at all sites except SBD. Inaddition, 12-h fine fraction aerosol (<2.5 µm) samples werecollected at IIT and onboard the Lake Guardian. Aerosolsamples were analyzed for Hg(p) and a suite of trace elements.An undetected severed sampling line prevented any validover-water Hg(v) samples during the AEOLOS Januaryintensive aboard the Lake Guardian.

Sample Collection and Analysis MethodsAcid Cleaning Procedure. All field and analytical suppliesused in the collection and analysis of Hg and trace elementsamples were cleaned in an 11-day procedure as describedby Landis and Keeler (25).

Precipitation. Wet-only event samples were collectedusing modified MIC-B (MIC, Thornhill, ON) automaticprecipitation collectors (25). Precipitation samples wereprocessed and analyzed in a Class 100 clean room to avoidcontamination. The volume of each precipitation samplewas determined gravimetrically. Mercury samples wereoxidized with concentrated BrCl to a 1% solution (v/v) and

FIGURE 1. Location of sampling sites used in the Lake MichiganMass Balance Study and the Atmospheric Exchange Over Lakesand Oceans Study.

VOL. 36, NO. 21, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 4509

stored in the dark at 4 °C overnight before being analyzed.Total Hg was quantified by a dual amalgamation techniquefollowed by cold-vapor atomic fluorescence spectrometry(CVAFS) (26). The method detection limit (MDL) for Hg inprecipitation was 0.3 ng L-1, determined using 3σ of theanalytical reagent blank (n ) 275). The precision of 34collocated samples was 8.1 ( 6.8 (relative percent difference(RPD) ( standard deviation (STDEV)) and the analyticalprecision was 4.0 ( 2.9 (n ) 208).

Trace element samples were acidified with concentratedHNO3 to a 0.2% solution (v/v) and were stored for a minimumof 2 weeks in the dark at 4 °C prior to analysis. Trace elementdeterminations were made using a Perkin-Elmer 5000AInductively Coupled Plasma Mass Spectrometer (ICP-MS)equipped with a thin film electron multiplier. The sampleswere introduced into the ICP-MS by pneumatic nebulization.All trace element samples were analyzed in triplicate, andreported concentrations are based upon the mean of thereplicate analyses. The method detection limits for Cu, Zn,Sr, and Pb were 0.024, 0.237, 0.026, and 0.037 µg L-1,respectively. The MDLs were generated for each elementusing a modification of EPA method 200.8 (27), in which aprecipitation sample was substituted for the fortified reagentwater to simulate the actual sample matrix.

Aerosols. Total aerosol samples were collected using 47mm open-faced Teflon filter packs inserted into custom-made fiberglass sampling enclosures mounted 3 m aboveground level. A nominal flow rate of 30 LPM was maintainedusing computerized vacuum pumps (URG Corporation,Chapel Hill, NC). Mercury aerosol samples were collectedonto baked glass fiber filters (Gelman type A/E), and traceelement aerosols were collected onto 2.0 µm PTFE membranefilters (Gelman R2PJ047) (24). Fine fraction aerosol filterswere inserted into Teflon filter packs, attached to a Teflon-coated aluminum cyclone (URG Corporation), and sampledat a nominal flow rate of 10 LPM.

All analytical procedures for determination of Hg(p) werecarried out in a Class 100 clean room. Hg(p) filters weremicrowave digested in HNO3 and subsequently analyzed byCVAFS in a manner described by Keeler and Landis (26). Thesystem detection limit (SDL) for Hg(p) was 1 pg m-3,calculated using 3σ of the field blanks (n ) 90). The precisionof 31 collocated samples was 9.1 ( 5.8 (RPD ( STDEV) andthe analytical precision was 3.0 ( 2.6 (n ) 391). Trace elementaerosol filters were analyzed by the EPA using energy-dispersive X-ray fluorescence (28). The 3σ MDLs for Cu, Zn,Sr, and lead were 1.2, 1.0, 1.0, and 2.0 ng m-3, respectively.All aerosol concentrations were corrected to standard tem-perature and pressure (STP; 0 °C, 1 atm).

Vapor Phase Hg. Hg(v) samples were prefiltered througha baked glass fiber filter and then quantitatively capturedonto two gold-coated glass bead traps in series at a massflow controlled flow rate of 300 cm3 min-1. Gold-coated beadtraps were heated to approximately 50 °C during samplingto prevent condensation from interfering in the Hg amal-gamation process. Vapor phase Hg samples were analyzedusing dual amalgamation CVAFS (26). The SDL for Hg(v) was45 pg m-3, calculated using 3σ of the field blanks (n ) 101).The precision of 29 collocated samples was 8.2 ( 6.2 (RPD( STDEV). All Hg(v) concentrations were corrected to STP.

During the January 1995 AEOLOS sampling period, aTekran model 2537A continuous vapor phase Hg analyzer(Tekran, Inc., Toronto, Canada) was deployed at the IIT site.The 2537A instrument utilizes two parallel gold matrix trapsto preconcentrate Hg(v) that is subsequently thermallydesorbed into a CVAFS analyzer (29). The 2537A instrumentwas configured to sample at 1.5 lpm for 20-min integratedsamples to provide high-resolution Hg(v) observations.

Meteorological Cluster Analysis. A meteorological clusteranalysis was conducted to evaluate how much of the

variability in the LMMBS atmospheric Hg data could beexplained by differences in air mass transport. The transportof air parcels to the LMMBS sites was estimated using theHybrid Single-Particle Lagrangian Integrated Trajectory (HY-SPLIT) Model Version 3.0 (30). HY-SPLIT 24-h sigma-layerback trajectories were calculated using input data from theNational Meteorological Center’s Nested Grid Model. Thesigma-layer trajectories represent the most probable path ofthe advected air parcels reaching the LMMBS sites whilesamples were collected. It is important to note that trajectorycalculations are not exact and have associated uncertaintiesthat generally increase with the time and/or distance upwindof the receptor location (31, 32).

The period of maximum precipitation intensity and themidpoint of sample period were used to match the trajectoryarrival time with each precipitation event and ambientsample, respectively. Several of the weekly precipitationsamples collected from November 1994 through March 1995were not included in the transport analysis. Weekly samplescomposed of more than a single event were excluded becauserepresentative meteorological transport could not be ascer-tained. Weekly samples containing snow and mixed pre-cipitation samples were also excluded from transport analysisbecause the potential bias imparted due to different in-cloudprocesses (see Results and Discussion). On average, theexcluded samples contributed less than 8% of the total Hgwet deposition at the five sites.

The 24-h back trajectories were grouped together usingWard’s minimum variance hierarchical clustering method,in a manner similar to Moody and Samson (33). Trajectoryendpoints representing both wind speed and direction every2 h upwind were introduced into the stepwise optimalclustering routine, which measured similarity using Euclideandistance between the endpoints. The resulting trajectoryclusters represent distinct meteorological transport regimesto the LMMBS sites.

Statistical Analysis. Data processing and all statisticalanalyses were performed using SAS v.6.12 (SAS Institute, Cary,NC). The assumptions of the parametric procedures wereexamined using residual plots, skewness and kurtosis coef-ficients, Shapiro-Wilk test, and the Brown-Forsythe test. TheLMMBS Hg in precipitation, Hg(p), Hg(v), and clustered Hgin precipitation data sets were log-normal. The data setswere transformed, and the following parametric procedureswere applied where appropriate: (i) t-test for independentsamples and (ii) one-way analysis of variance (ANOVA). Onlythe trace element aerosol and clustered ambient Hg datasets seriously violated the assumptions of the parametricprocedures and the following nonparametric procedures wereused, respectively: (i) Wilcoxon test and (ii) Kruskal-Wallistest. Two-sided tests were used unless otherwise stated. Alevel of significance of R ) 0.05 was used for all statisticalprocedures.

Results and DiscussionPrecipitation. An ANOVA found a significant (p < 0.0001)difference in Hg concentrations between the LMMBS moni-toring sites. The number of samples collected, volume-

TABLE 1. Volume Weighted Average Concentrations of Mercuryin Precipitation and Wet Deposition Measured during theLake Michigan Mass Balance Study (July 1, 1994-October31, 1995)

site n Hg (ng L-1) range (ng L-1) Hg deposition (µg m-2)

BON 82 16.1 5.3-136.8 18.7IIT 74 21.5 5.4-74.5 26.9KEN 76 16.3 4.5-132.0 18.3SBD 101 10.8 2.1-63.6 14.9SHN 86 13.9 3.2-110.3 17.3

4510 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 21, 2002

weighted average (VWA) Hg concentrations, and total Hgwet deposition at each site is summarized in Table 1. Asubsequent Bonferroni t-test found the IIT site (21 ng L-1)to be significantly higher and the SBD site (10.8 ng L-1) tobe significantly lower than the Hg concentrations at the othersites, indicating that local and regional sources are contrib-uting significant quantities of mercury to wet deposition inthe southern portion of the Lake Michigan Basin.

A high degree of variability was observed in the concen-tration of Hg in precipitation at each site (Table 1), withconcentrations ranging by more than an order of magnitude.A simple linear regression analysis was preformed to evaluatethe importance of precipitation depth to the observedLMMBS Hg concentrations. The analysis revealed that Hgconcentration was significantly (p ) 0.002) related toprecipitation depth at all sites, explaining from 8 (SBD) to

FIGURE 2. Regression plots showing the relationship between precipitation depth and mercury concentration in rain (a-e) and snow(f) at each of the Lake Michigan Mass Balance Study Sites.

VOL. 36, NO. 21, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 4511

20% (KEN) of the variability. This volume dependence withhigher Hg concentrations associated with lower precipitationdepths has also been observed in Michigan (16) andWisconsin (34).

An ANOVA was preformed to evaluate the importance ofprecipitation type. The analysis found that snow and mixedprecipitation (rain and snow) samples were significantly lowerin Hg concentration than rain samples (p < 0.0001).Regression analysis also found no significant relationshipbetween Hg concentration and snow amount, suggestingthat Hg may not be incorporated into precipitation formedin cold cloud processes (e.g. accretion, vapor deposition) asefficiently as warm cloud processes (e.g. water vapor diffusion,collision-coalescence). Subsequent regression of Hg con-centration to precipitation amount for rain samples onlyrevealed a more significant relationship (p < 0.0001),explaining from 16 (SBD) to 26% (KEN) of the variability(Figure 2). Since snow and mixed precipitation samples havelower Hg concentrations, it was not surprising that an ANOVAfound a significant difference (p < 0.0001) in Hg concentrationin precipitation by season. A subsequent Bonferroni t-testindicated that the spring and summer had significantly higherVWA Hg concentrations than the fall and winter (Figure 3),even though the spring and summer events typically hadgreater precipitation depths. The higher Hg concentrationsresulted in substantially higher Hg deposition in the springand summer than the fall and winter at all the LMMBS sites.A similar temporal trend was seen at a site in Dexter, MI (25)and in a multisite network in Michigan (16).

A meteorological cluster analysis was conducted todetermine if a source region(s) contributed to elevated Hgin event rain samples during the LMMBS could be identified.Six clusters were identified for the BON, IIT, SBD, and SHNsites, and four clusters were identified for the KEN siterepresenting statistically unique meteorological flow regimes.Figure 4 depicts the meteorological clusters for IIT andSupporting Information Figures 1-4 depict the meteorologicalclusters for the other sites. Each back trajectory plotted inFigure 4 represents the most probable path of the advectedair parcels for the 24 h preceding the measured rain event.

Three predominantly anthropogenic trace elements (Cu,Zn, and Pb) and one predominantly crustal trace element(Sr) were included into the meteorological analysis to clarifythe role of local and regional sources to the observed Hgconcentrations. The VWA concentrations of Cu, Zn, Pb, andSr from the LMMBS event rain samples are summarized in

Table 2. As with Hg, the anthropogenic trace elements areall significantly elevated at the IIT site (p < 0.0001). Tovisualize the relationship between precipitation chemistryand air transport, relative concentration factors (RCF) werecalculated for Hg, Cu, Zn, Pb, and Sr by cluster (35). The RCFis defined here as the quotient of the cluster VWA concen-tration and the site VWA concentration minus one. Any RCFgreater than zero indicates a trace element is enhancedrelative to the overall site mean. Figure 5 shows the RCF of

FIGURE 3. Seasonal volume-weighted average concentrations of mercury in precipitation measured during the Lake Michigan MassBalance Study (July 1, 1994 -October 31, 1995).

FIGURE 4. Results of meteorological cluster analysis for the IIT siteshowing the relationship between volume-weighted averagemercury in event rain samples and the direction of air mass transport.

4512 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 21, 2002

Hg, Cu, Zn, Sr, and Pb for each site by cluster.It is clear from the data in Figure 5 that concentrations

of Hg, Cu, Zn, Sr, and Pb in rain varied with storm transport.The difference in Hg concentration was typically greater thana factor of 2 between the highest and lowest transport cluster.An ANOVA found a significant difference between transportcluster and Hg concentration at BON, IIT, and KEN explaining19, 32, and 17% of the variance, respectively. In general,transport from large urban/industrial areas resulted inenhanced concentrations of Hg and other anthropogenicelements. Figure 5a shows enhanced RCFs at the BON sitewith transport from the southwest out of St. Louis (cluster1) and with transport from the east-northeast out of theIndianapolis and Chicago/Gary areas (cluster 4). Figure 5c-ealso show transport from the Chicago/Gary area resulting inelevated RCFs for anthropogenic elements at the KEN (cluster2), SBD (clusters 3 and 4), and SHN (clusters 1 and 4) sites.The influence from anthropogenic sources in other areaswas observed at SBD with transport from the northeast(cluster 6) and at SHN with transport from the southeast(cluster 5). While the influence of anthropogenic sourcesfrom multiple locations were observed in precipitation ateach of the LMMBS, the impact of the Chicago/Gary urbanarea was clearly dominant.

Surprisingly, even in downtown Chicago transport clusteranalysis found significantly different local source areas. Figure5b shows that transport from the southeast out of thesouthern Chicago/Gary area (clusters 1 and 3) resulted inthe highest RCFs for Hg, Zn, and Cu at the IIT site. Transportfrom the southwest (cluster 2) also resulted in elevatedconcentrations of Hg, Cu, Zn, and Pb, indicating thatprecipitation at IIT was influenced by multiple sources aroundthe urban area.

The rate of transport also affected the concentrations ofboth the anthropogenic and crustal elements in rain. At theurban IIT site, slow stagnant transport (cluster 1) fromsouthern Chicago/Gary area resulted in enhanced concen-trations of anthropogenic elements. Rapid transport fromthe same source region (cluster 3) resulted in lower Hg, Zn,and Pb concentrations and substantially higher Sr concen-trations. The higher concentrations observed with slowtransport were likely a result of concentrated local emissionswhile faster transport aided in rapid advection. Conversely,higher concentrations of anthropogenic elements wereobserved at SHN with rapid transport from the Chicago/Gary area (cluster 4) rather than from slower transport (cluster1). Rapid transport to SHN probably minimized the amountof aerosol mass that was wet deposited upwind resulting inelevated concentrations. At the semirural BON (clusters 6and 3) and SHN (clusters 5 and 2) sites, elevated levels of Srwere observed with rapid transport versus slower transportfrom similar source regions, indicating that additional crustalmaterial was incorporated into rain with higher wind speeds.

Although the variability of Hg concentration in rain wassignificantly related to precipitation depth, on average theprecipitation depth explained only 20% of the Hg variability.There was substantial variability in Hg concentrations for

events of similar magnitude. For example, the LMMBS meanHg concentration for rain events from 1 to 2 cm was 17.7 (9.3 ng L-1, with a range of 2.1-42.0 ng L-1. The range for theassociated wet deposition was 0.04-0.65 µg m-2. Theextremely large variation in Hg concentrations observed fromsuch a narrow range of precipitation depths demonstratethe importance of transported emissions from local andregional anthropogenic sources to Hg in the LMMBSprecipitation samples.

Aerosols and Vapor Phase Hg. An ANOVA found sig-nificant (p < 0.0001) differences in Hg(p) and Hg(v) con-centrations between the LMMBS monitoring sites. Subse-quent Bonferroni t-tests found the IIT site Hg(p) and Hg(v)to be significantly higher and the SBD site Hg(p) to besignificantly lower than the other sites. Tables 3 and 4summarize results from each of the sites during the LMMBSfor Hg(p) and Hg(v), respectively. The average Hg(p) con-centration at IIT (70 ( 67 pg m-3) was almost a factor of 6higher than the average Hg(p) concentration at SBD (12 (8 pg m-3). With the exception of IIT, all the sites showremarkably similar mean Hg(v) concentrations with relativelylittle variation. Unlike the Hg in precipitation or Hg(p), theIIT site Hg(v) concentration was less than a factor of 2 higher(3.6 ( 2.9 ng m-3).

During the July, 1994 AEOLOS sampling intensives theGWS site provided additional evidence of an enhancementof Hg(p) (Table 5) and Hg(v) (Table 6) in the Chicago/Garyurban area. In fact, the GWS Hg(p) mean concentration (133pg m-3) was nearly twice as high as the IIT Hg(p) meanconcentration (69 pg m-3), indicating a substantial near fieldHg(p) enhancement adjacent to the heavily industrializedGWS site. The Hg(v) concentrations measured at GWS werelower than the Hg(v) values at IIT, suggesting different sourceswere responsible for Hg(p) and Hg(v) at the two sites. TheAEOLOS size characterized Hg(p) data also implicates ananthropogenic source to the observed concentrations. Onaverage, over 70% of the measured Hg(p) was in the finefraction (Table 5). These small particles are generallyproduced in high-temperature anthropogenic combustionprocesses (36).

The elevated over-water Hg(p) concentrations measuredaboard the Lake Guardian during AEOLOS were a significantfinding. Elevated Hg(p) concentrations were observed notonly in the Chicago/Gary urban area but also over the

TABLE 2. Volume Weighted Average Concentrations ofAnthropogenic (Cu, Zn, Pb) and Crustal (Sr) Trace Elements inEvent Rain Samples Measured during the Lake Michigan MassBalance Study (July 1, 1994-October 31, 1995)

site n Cu (µg L-1) Zn (µg L-1) Pb (µg L-1) Sr (µg L-1)

BON 77 0.56 2.87 0.89 0.85IIT 67 1.96 12.15 3.64 1.51KEN 65 0.70 4.98 1.27 0.87SBD 90 0.38 2.80 0.80 0.76SHN 75 0.89 4.38 1.30 0.86

TABLE 3. Summary of Total Particulate Phase Mercury (Mean( Standard Deviation) and Trace Element Aerosols (Median)Measured during the Lake Michigan Mass Balance Study (July1, 1994-October 31, 1995)

site nHg(p)

(pg m-3)

Hg(p)range

(pg m-3)Cu

(ng m-3)Zn

(ng m-3)Pb

(ng m-3)Sr

(ng m-3)

BON 74 19 ( 11 4-63 1.8 14.1 6.3 1.4IIT 79 70 ( 67 8-494 11.4 91.0 28.7 3.8KEN 79 24 ( 18 3-108 3.5 24.2 7.4 1.4SBD 80 12 ( 8 1-41 1.5 7.0 3.2 0.8SHN 81 19 ( 12 2-69 1.6 14.8 6.7 1.2

TABLE 4. Concentrations of Total Vapor Phase Mercury (Mean( Standard Deviation) Measured during the Lake MichiganMass Balance Study (July 1, 1994 - October 31, 1995)

site n Hg(v) (ng m-3) range (ng m-3)

BON 74 2.0 ( 0.5 1.3-3.8IIT 78 3.6 ( 2.9 1.6-22.1KEN 73 2.2 ( 0.7 1.1-5.7SBD 78 2.1 ( 0.7 1.4-5.0SHN 77 2.2 ( 0.7 1.4-6.1

VOL. 36, NO. 21, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 4513

adjacent water body where the aerosols were available forwet and dry deposition. Although the overall IIT site mean(65 pg m-3) appears quite different from the over-water mean(38 pg m-3), the concentrations measured on the LakeGuardian were equal or higher on 10 of the 23 samplingperiods. A similar situation was observed with several othertrace elements such as Fe, Mn, and Zn (37). The IIT sitemean was heavily influenced by two elevated observationson the evenings of July 17 (206 pg m-3) and July 19 (275 pgm-3). Conversely, the over-water Hg(v) concentrationsmeasured aboard the Lake Guardian were lower than at IITon every sampling period except one, again suggesting thatthe sources of Hg(p) and Hg(v) are different.

During both the LMMBS and the AEOLOS there wereseveral sampling periods at IIT that the Hg(v) concentrationexceeded 10 ng m-3. In fact, on July 31, 1994 a 24-h compositesample at IIT reached 22.1 ng m-3. On January 20, the 12-hintegrated Hg(v) concentration was 25.4 ng m-3, which agreedwell with the concurrent Tekran 12-h integrated concentra-tion of 25.9 ng m-3. The high-resolution Tekran data indicatedthat the elevated Hg(v) 12-h average sample was a result ofa 3-h event that had a maximum 20-min average concentra-tion of 220.4 ng m-3. Other smaller magnitude events werealso observed on January 18 (12.9 ng m-3) and January 20(5.6 ng m-3). If most of the elevated integrated Hg(v)concentrations measured at IIT were the result of similar

FIGURE 5. Relative concentration factors for mercury, copper, zinc, strontium, and lead in event rain samples associated with meteorologicaltransport clusters.

4514 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 21, 2002

short-term events or “local plume impacts”, then it isunderstandable that the Hg(v) measured aboard the LakeGuardian were lower since the probability of being affectedby a small spatial event were substantially reduced.

The spatial gradient in Hg(p) observed between IIT andGWS and the high resolution Hg(v) data collected by theTekran analyzer indicate that the air concentrations of Hgin the Chicago/Gary urban area were not homogeneous. Infact, the urban airshed was a complex system with numerouspoint sources and distinctive meteorology. The ability of over-water measurements made from one research vessel over a3-week period were insufficient to definitively conclude if

IIT, or any other one site, is representative of the Chicago/Gary Urban area.

The results of meteorological cluster analysis for theLMMBS ambient samples are depicted in Supporting In-formation Figures 5-9. Four clusters were identified for theBON, SBD, and SHN sites, and five clusters were identifiedfor the IIT and KEN sites. An ANOVA found a significantdifference between transport cluster and Hg(p) concentrationat IIT, KEN, SBD, and SHN explaining 13, 14, 23, and 13%of the variance, respectively. A Kruskal-Wallis test alsoindicated a significant difference between transport clusterand Hg(v) concentration at IIT and KEN.

FIGURE 6. Relative concentration factors for total particulate phase mercury, total vapor phase mercury, copper, zinc, strontium, and leadin ambient samples associated with meteorological transport clusters.

VOL. 36, NO. 21, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 4515

RCFs demonstrating the relationship between ambientconcentrations and air transport were calculated for Hg(p),Hg(v), Cu, Zn, Sr, and Pb by cluster (Figure 6). A very clearinfluence of the Chicago/Gary urban area was observed atthe three semirural sampling sites on the lake. Bonferonnit-tests found significantly higher Hg(p) concentrations at theKEN (cluster 3), SBD (cluster 1), and SHN (cluster 2) siteswith transport from the Chicago/Gary area. In addition, thehighest concentrations of Hg(v), Cu, Zn, Sr, and Pb at allthree sites were also associated with the same urban transportregime. The lowest Hg(p) concentrations at KEN (cluster 5),SBD (cluster 4), and SHN (cluster 4) were associated with airmass transport from a northerly direction.

The relationship between aerosol concentrations andtransport was more complex at the IIT site. At IIT, the highestconcentrations of Hg(p) and Pb were observed with transportfrom the east-southeast (cluster 3) out of the southernChicago/Gary area, and the highest concentrations of Hg(v),Cu, and Sr were seen with transport from the southwest(cluster 1). IIT was the only site that average Hg(p) and Hg(v)maximums were observed from different meteorologicaltransport regimes supporting the data collected at the GWSsite during AEOLOS and suggesting different sources for Hg-(p) and Hg(v). Elevated concentrations of anthropogenicaerosols from different air mass transport regimes at the IITsite indicate that multiple local/regional sources are locatedaround the Chicago site.

Implications. While some industry trade groups arguethat Hg is only a global problem, the spatial pattern ofatmospheric mercury and meteorological cluster modelingresults from the LMMBS clearly indicate that sources in theChicago/Gary urban area were contributing to enhanced Hgin precipitation and Hg(p) concentrations across the entireLake Michigan basin. Hg in precipitation and Hg(p) con-centrations observed at the IIT, KEN, SBD, and SHN sitesdemonstrated a factor of 2 difference between transport fromthe Chicago/Gary urban area and transport from the lowestclusters. In addition, transport relationships were correlatedto trace elements known to be from anthropogenic origin.Overall, the spatial pattern of LMMBS Hg(v) indicated a

regional or global influence with additional local sourceenhancement observed routinely in the Chicago/Gary urbanarea and less frequently at the more remote locations (e.g.SBD).

The Hg monitoring strategies for the LMMBS and AEOLOSwere specifically designed to investigate the impact of theChicago/Gary urban area on Lake Michigan. Other urbanand industrial locations in the Lake Michigan basin may alsocontribute on a smaller scale to enhanced atmospheric Hg(e.g. Milwaukee, Green Bay). The results of this study clearlyindicate that atmospheric monitoring in urban/industriallocations is necessary to evaluate the relative importance oflocal and regional sources to atmospheric Hg deposition.

AcknowledgmentsThe U.S. Environmental Protection Agency through its Officeof Research and Development (Cooperative AgreementCR820909-01) and its Great Lakes National Program Office(Grant GL995569-01) funded the research described hereunder to the University of Michigan Air Quality Laboratory(UMAQL). It has been subjected to Agency review andapproved for publication. Mention of trade names orcommercial products does not constitute an endorsementor recommendation for use. We would like to thank theoperators of the LMMBS sampling sites Mary Barden (SHN),Brant Englund (KEN), Beth Laulley (SBD), Mike Snider (BON),and Tom Holsen who supervised the numerous graduatestudent operators at (IIT) and graciously provided uslaboratory space during the AEOLOS intensives. In addition,we would like to thank the captain and crew of the LakeGuardian and the following students and staff from theUMAQL for their efforts: Kabrina Askwith, Jim Barres, JanetBurke, Scott DeBoe, Tim Dvonch, Ganda Glinsorn, JoeGraney, Marion Hoyer, Bei Di Huang, Tamar Krantz, AnnKruger, and Jeremy Landis.

Supporting Information AvailableEvent rain sample meteorological cluster plots for BON, KEN,SBD, and SHN in Figures 1-4, respectively, and ambientsample meteorological cluster plots for all sites in Figures5-9. This material is available free of charge via the Internetat http://pubs.acs.org.

Literature Cited(1) Lindqvist, O.; Rodhe, H. Tellus 1985, 37B, 136-159.(2) Fitzgerald, W. F.; Vandal, G. M.; Mason, R. P. Wisconsin Mercury

Research Annual Report; Electric Power Research Institute:Concord, CA, 1990.

(3) Mierle, G. M. Environ. Toxicol. Chem. 1990, 9, 843-851.(4) Iverfeldt, A. Water, Air, Soil Pollut. 1991, 56, 251-265.(5) Lindberg, S. E.; Turner, R.; Meyers, T.; Taylor, G., Jr.; Schroeder,

W. Water, Air, Soil Pollut. 1991, 56, 577.(6) Fitzgerald, W. F.; Mason, R. P.; Vandal, G. M. Water, Air, Soil

Pollut. 1991, 56, 745-767.(7) Keeler, G. Technical Report EPA/600/SR-94/191; U.S. EPA At-

mospheric Research and Exposure Assessment Laboratory:Research Triangle Park, NC, 1994.

(8) Mason, R. P.; Sullivan, K. A. Environ. Sci. Technol. 1997, 31,942-947.

(9) Nriagu, J. Environment 1990, 32, 28-32.(10) Expert Panel on Mercury Atmospheric Processes. Technical

Report EPRI/TR-104214; 1994.(11) Schroeder, W. H.; Munthe, J. Atmos. Environ. 1998, 32, 809-

822.(12) Fitzgerald, W. Atmospheric and oceanic cycling of mercury. In

Chemical oceanography; Academic Press: San Diego, CA, 1989;Vol. 10.

(13) Slemr F.; Schuster, G.; Seiler, W. J. Atmos. Chem. 1985, 3, 407-434.

(14) Hanisch, C. Environ. Sci. Technol. 1998, 32, 176a-179a.(15) Landers, D. H.; Gubala, C.; Verta, M.; Lucotte, M.; Johanssin, K.;

Vlasova, T.; Lockhart, W. L. Atmos. Environ. 1998, 32, 919-928.(16) Hoyer, M.; Burke, J.; Keeler, G. J. Water, Air, Soil Pollut. 1995,

80, 199-208.

TABLE 5. Concentrations of Total Particulate Phase Mercury(Mean) Measured during the Atmospheric Exchange overLakes and Oceans Study (July 17-31, 1994 and January15-21, 1995)

site nJuly 1994(pg m-3) n

Jan 1995(pg m-3)

BON 21 18 7 12IIT (total) 30 69 14 84IIT (fine) 30 49 14 77KEN 30 19 7 11SBD 15 16 7 6SHN 23 23 7 8Lake Guardian (total) 18 36 5 44Lake Guardian (fine) 18 29 5 32GWS 9 133

TABLE 6. Concentrations of Total Vapor Phase Mercury (Mean)Measured during the Atmospheric Exchange over Lakes andOceans Study (July 17-31, 1994 and January 15-21, 1995)

site n July 1994 (ng m-3) n Jan 1995 (ng m-3)

BON 21 1.7 7 1.9IIT 30 4.0 14 4.1KEN 30 1.8 7 2.0SBD 15 1.8 6 2.1SHN 23 1.8 7 2.1Lake Guardian 15 1.9GWS 9 3.5

4516 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 21, 2002

(17) Iverfeldt, A. Water, Air, Soil Pollut. 1991, 56, 251-265.(18) Dvonch, J. T.; Graney, J. R.; Marsik, F. J.; Keeler, G. J.; Stevens,

R. K. Sci. Total Environ. 1998, 213, 95-108.(19) Pirrone, N.; Allegrini, I.; Keeler, G. J.; Nriagu, J. O.; Rossmann,

R.; Robbins, J. A. Atmos. Environ. 1998, 32, 929-940.(20) Hoff, R. M.; Strachan, W. M. J.; Sweet, C. W.; Chan, C. H.;

Shackleton, M.; Bidleman, T. F.; Brice, K. A.; Burniston, D. A.;Cussion, S.; Gatz, D. F.; Harlin, K.; Schroeder, W. H. Atmos.Environ. 1996, 30, 3505-3527.

(21) Gatz, D. F.; Sweet, C. W.; Basu, I.; Vermette, S.; Harlin, K.; Bauer,S. Technical Report SWS Miscellaneous Publication 158; IllinoisState Water Survey: Champaign, IL, 1994.

(22) Mercury Study Report to Congress. EPA-425/R-97-003; U.S.Environmental Protection Agency, Office of Air Quality Planningand Standards, Office of Research and Development; U.S.Government Printing Office: Washington, DC, 1998.

(23) Dolske, D. A.; Sievering, H. Water, Air, Soil Pollut. 1979, 12,485-502.

(24) Keeler, G. J.; Landis, M. S. Technical Report EPA 905-R-97-012a;U.S. EPA Great Lakes National Program Office: Chicago, IL,1997.

(25) Landis, M. S.; Keeler, G. J. Environ. Sci. Technol. 1997, 31, 2610-2615.

(26) Keeler, G. J.; Landis, M. S. Technical Report EPA 905-R-97-012b;U.S. EPA Great Lakes National Program Office: Chicago, IL,1997.

(27) Methods for the Determination of Metals in EnvironmentalSamples. U.S. Environmental Protection Agency Office of

Research and Development, Environmental Monitoring SystemsLaboratory; CRC Press: Boca Raton, FL, 1992; pp 95-137.

(28) Dzubay, T. G.; Stevens, R. K.; Gordon, G. E.; Olmez, I.; Sheffield,A. E.; Courtney, W. J. Environ. Sci. Technol. 1988, 22, 2, 46-52.

(29) Schroeder, W. H.; Keeler, G. J.; Kock, H.; Roussel, P.; Schnee-berger, D. R.; Schaedlich, F. H. Water, Air, Soil Pollut. 1995, 80,611-620.

(30) Draxler, R. R. Technical Memorandum ERL ARL-195; NOAA AirResources Laboratory: Silver Springs, MD, 1992.

(31) Kahl, J. D.; Samson, P. J. J. Appl. Meteorol. 1988, 27, 1024-1035.(32) Kahl, J. D. Atmos. Environ. 1996, 30, 2945-2957.(33) Moody, J. L.; Samson, P. J. Atmos. Environ. 1989, 23, 2117-

2132.(34) Lamborg, C. H.; Fitzgerald, W. F.; Vandal, G. M.; Rolfhus, K. R.

Water, Air, Soil Pollut. 1995, 80, 189-198.(35) Moody, J. L.; Galloway, J. N. Tellus 1988, 40B, 463-479.(36) Seinfeld, J. H. Atmospheric Chemistry and Physics of Air Pollution;

Wiley-Interscience: New York, 1986.(37) Caffrey, P. F.; Ondov, J. M.; Zufall, M. J.; Davidson, C. I. Environ.

Sci. Technol. 1998, 32, 1615-1622.

Received for review August 21, 2001. Revised manuscriptreceived July 24, 2002. Accepted July 30, 2002.

ES011216J

VOL. 36, NO. 21, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 4517