Assessing Annual Water- Air Fluxes of ... - ACS Publications

Feb 1, 1995 - Engineering, University of Minnesota, P.O. Box 100,. Navarre, Minnesota 55392 .... with the flux of H 2 0 vapor using laboratory and fie...
0 downloads 0 Views 2MB Size
Environ. Sci. Technol. 1995, 29, 869-877

Assessing Annual Water- Air Fluxes of Pdychlorinated Biphenyls in Lake Michigan KERI C . H O R N B U C K L E , + CLYDE W. SWEET,* ROGER F. PEARSON,§ DEBORAH L. SWACKHAMER,S AND STEVEN J. EISENREICH*'+ Gray Freshwater Biological Institute and Department of Civil Engineering, University of Minnesota, P.O. Box 100, Navarre, Minnesota 55392, Illinois State Water Survey, Champaign, Illinois 61820, and Environmental and Occupational Health, School of Public Health, Box 807 UMHC, University of Minnesota, Minneapolis, Minnesota 55455

Air-water exchange of PCBs was determined in Lake Michigan on an event and seasonal basis in 19911993. Instantaneous fluxes of CPCB (sum of 77 congener peaks) based on air-water concentration gradients drawn from air and water samples collected simultaneously aboard ship demonstrated net volatilization in September 1991. Air samples collected on the northeastern shore of Lake Michigan (Sleeping Bear Dunes State Park) between December 1991 and July 1993 showed no seasonal trend in vapor-phase CPCB concentrations and ranged from 30 to 400 pg/m3. These air concentrations were used t o calculate seasonal water-air fluxes of CPCB that ranged from -18 ng m-2 day-' (net deposition) to 60 ng m-2 day-' (net volatilization). The seasonal variation of vapor-phase and dissolved-phase PCBs in the impacted southern quarter of the lake are unknown, thereby hindering estimation of fluxes in this region. The estimated annual net CPCB flux is 12.3 p g m-2 yr-', which corresponds to 520 kg for the northern three-quarters of Lake Michigan.

Introduction Air-water exchange is an important process for the delivery and removal of semivolatile organic compounds (SOCs) from natural waters (1-5). Jeremiason et al. (a have demonstrated through a detailed mass balance that volatilization is the dominant pathway for the removal of PCBs from Lake Superior, a conclusion verified by Hornbuckle et al. (7). Mackay (8)has drawn a similar conclusion for Lake Ontario based on a dynamic mass balance approach. Air-water exchange is likely to be a dominant process in all large aquatic ecosystems. The magnitude of atmosphericvapor exchange as a sink or source of PCBs to Lake Michigan has previously been estimated by poorly constrained mass budgets (2,9).This method is limited by the propagated error of all other components of the mass balance, assumes that a l l components are known, and yields no information on seasonal variations or factors controlling air-water exchange. The goal of this study is to estimate the direction and magnitude of air-water exchange of PCBs in Lake Michigan on daily, seasonal, and annual time scales. PCBs in Lake Michigan are of concern because of their presence in commercial and sport fish (10, 11) requiring the issuance of fish consumption advisories. Through gas exchange, the atmosphere may be an important source and sink of PCBs to the water column and ultimately to fish. PCB vapor exchange is calculated as a function of local meteorology and "skin" surface water temperatures, air and water PCB concentrations, and physical-chemical properties of the organic compounds. The result is independent of estimates of sedimentation, direct inputs, and PCB speciation in the air and water. Ah-Water Gas Exchange Model. The modified airwater exchange model applied here has been presented elsewhere (5, 7) and is summarized below. The net flux of PCBs across the air-water interface is calculated from the mass transfer coefficients across the air-side and waterside layers and the PCB concentration difference between the bulk air and water phases (12). The mass transfer coefficients are derived from laboratory and field experiments using gaseous tracers such as SFs, 0 2 , and H20 vapor. The operative flux expression is

flu = k,, (C, - C )

(1)

where llkol is the overall resistance to mass transfer and

and

C = Cd,IH

(3)

Equation 1 is written to describe net volatilization as a positive flux;net deposition will have a negative value. The * To whom correspondence should be addressed e-mail address: [email protected]; FAX: 612-471-9070. + Gray Freshwater Biological Institute and Department of Civil Engineering. Illinois State Water Survey. 5 Environmental and Occupational Health, Schoolof Public Health.

*

0013-936W95/0929-0869$09.00/0

0 1995 American Chemical Society

VOL. 29. NO. 4, 1995 ENVIRONMENTAL SCIENCE

TECHNOLOGY m 8es

overall mass transfer coefficient is kol (m/day); ka (m/day) is the rate coefficient describingtransfer across the stagnant air layer; and k, (mlday) is the rate coefficient describing transfer across the stagnant water layer. Cair is the PCB congener or homolog concentration in the vapor phase (pg/m3);C,is the PCB concentration in the dissolved phase in water (pg/m3). The dimensionless Henry's law constant, H,defines the equilibrium distribution of the chemical in air and water [e.g., (pg/m3)/(pg/m3)l.Since Henry's law constants decrease two to three times for every 10 "C decrease in temperature, the constants for the individual congeners reported for 25 'C were corrected for ambient temperature using the method of ten Hulscher et al. (13). Details on the temperature correction for PCBs are described elsewhere ( 7 ) . The flux of PCBs through the air-side layer is correlated with the flux of H20vapor using laboratory and field results as summarized by Schwarzenbach et al. (14) as

+

ka,H,O= 0 . 2 ~0.3 ~ ~

(4)

where ka,~20 (cmls) is correlated to PCBs through HzO and PCB diffusivities in air (15), and (5) where ka,PCB (cm/s)is the mass transfer coefficient through the stagnant air layer for a PCB congener, ul0 (m/s) is the wind speed at a reference height of 10 m, and &CB,air and DHzO,air are the diffusivities of PCB and water in air, respectively. PCB molecular diffusivities in air were calculated using the appropriate surface water temperatures by the method of Fuller (16) and range from 0.041 to 0.063 cm2/s. The value of 0.61 for the exponent is based on laboratory observations (15). The flux of PCBs across the water-side layer is correlated with the flux of COZ across the air-water interface. The correlation is based on whole-lake experiments involving the transfer of SFs, a volatile compound with dominant resistance to transfer across the water layer (17-20). This correlation is normalized to the appropriate Schmidt number, Sc (the ratio of the kinematic viscosity and the molecular diffusivityin water). The molecular diffusivities of PCB congenerswere calculated using the method of Wilke and Chang (21). From Wanninkhof et al. (17) 1.64

~ , c o=, 0.45U~~

and (7)

where kW,pc~ (cm/h)and kW,co2 (cm/h) are the mass transfer coefficientsthrough the water-side layer for PCB and carbon dioxide,respectively. SCCO, and SCPCB were calculated using the surface water temperatures appropriate to each sample and time period. A Schmidt number exponent of -0.5 is consistent with experimental evidence (19, 22) and the surface renewal model (23). Resistance to mass transfer for PCBs through the water-side layer is 70-80% of the total resistance to transfer at ambient temperatures. All 870

ENVIRONMENTAL SCIENCE &TECHNOLOGY / VOL. 29, NO. 4,1995

mass transfer parameters were calculated for each homolog group using the appropriate water temperatures and wind speeds.

Experimental Methods Sampling Strategy. Air and water samples were collected in Lake Michigan aboard the RlVLake Guardian (US. EPA) in September 1991,May 1992,and August 1992. Two sets of air samples were collected and used separately for estimates of instantaneous PCB air-water exchange fluxes and for estimates of seasonal and annual fluxes. For the instantaneous fluxes, air samples were collected from the bowwith a high-volume air sampler (GPS1Graseby/GMW, Cleves, OH 45002)equipped with a quartz fiber filter (QFF) and polyurethane foam plug (PUF). The air sampler was operated onlywhen the wind was (60' off the bow, usually only during travel. Each sample consisted of approximately 400 m3 of air or -10 h of sampling at 0.5-0.7 m3/min. Our goal was to simultaneously collect air and water samples from the research vessel as the ship traveled between water sampling stations. Water samples (-100 L) from Lake Michigan were extracted on board by continuous flow liquid-liquid extraction using the Goulden large sample extractor (GLSE) (24, 25). The dissolved phase was operationally defined by filtration through a 0.7-mm (nominal pore size) pre-ashed glass fiber filter (293mm). Details of this sampling procedure are reported elsewhere (6). Air samples were also collected at Sleeping Bear Dunes National Lakeshore, MI (SBD site), on the northwestern Lake Michigan shore (44'45'40'' latitude, 86'03'31" longitude) and used to estimate seasonal and annual PCB airwater exchange. This remote site is operated under the direction of the International Atmospheric Deposition Network (IADN) of the U.S. EPA. Samples were collected approximately twice monthly from December 12,1991,to June 21,1993,using two high-volume air samplers (GMW, Cleves, OH). They were equipped with glass fiber filter and XAD-2resin or PUF and operated at 0.566 m3/min for 24 h. The two matrices are comparable for all homologs at temperatures below 20 'C. At warmer air temperatures, XAD-2 retains a higher percentage of the vapor-phase dichlorobiphenyls and trichlorobiphenyls (26). PUF plugs were used for collecting samples from December 12,1991, through April 21,1992;XAD-2resin was used from May 15, 1992,through June 21, 1993. Air temperature and wind speed and direction were recorded on a 10-m meteorological tower (-60 m over the lake surface) during air sampling. Analytical Procedures. Analytical procedures and quality control data for the three data sets are summarized in Table 1. Treatment of polyurethane foam plugs (PUF) used for the over-water vapor PCB samples is the same as described elsewhere (7). XAD-2 used for air sampling at the SBD site was rinsed with water and extracted in a Soxhlet apparatus for at least 24 h with each of the following pesticide-grade solvents in sequence: methanol, acetone, hexane, dichloromethane, hexane, acetone, and methanol. The XAD-2 was ovendried at 65 'C and stored at -20 "C. PUF plugs used for air sampling at the SBD site were rinsed with water and cleaned by extractions with dichloromethane and 1:l hexane acetone, dried in a vacuum desiccator, and stored in a sealed metal can at -20 'C. Prior to sample extraction, PCB surrogate standards [IUPAC No. 14 (3,5-dichlorobiphenyl), IUPAC No. 65 (2,3,5,6-tetrachlorobiphenyl), and IUPAC No. 166 (2,3,4,4',5,6-hexachlorobiphenyl)l were

TABLE 1

Summary of Analytical Methods and Quality Control Results water samples

sample collection method extraction method sample cleanup column material chromatography

submersible pump/ glass fiber filter (GFF) G LSE 100% activated alumina

HP5890 with 63Ni ECD, 60 m capillary column, 0.32 i.d., dimethyl

diphenyl polysiloxane stationary phase, 0.25 film thickness PCB surrogate recoverya(av f SD)

over-water vapor

high-volumeair sampler, PUF, GFF

Soxhlet/dichloromethane 10% deactivated alumina, 6% deactivated silica gel HP5890 with 63Ni ECD, 60 m capillary column, 0.25 i.d., dimethyl diphenyl polysiloxane stationary phase, 0.25 film thickness

replicate samples (rel.% difference)

No. 14: 51 & 15% No. 65: 60 i 15% No. 166: 60 f 10% 11.3%(split samples)

data not available

CPCB field blanks (av f SD)

2.3 f 0.1 ng

6.8 i 2.2 ng

limit of detection

50 pg/L

17 pg/m3

No. 14: 95 f 16% No. 65: 90 i 10% No. 166: 99 f 5%

SBD site vapor

high-volumeair sampler, PUF or XAD-2, GFF Soxhlet/l:l hexane:acetone 3% deactivated silica gel HP5890 with 63Ni ECD, 60 m capillary column, 0.32 i.d., dimethyl

diphenyl polysiloxane stationary phase, 0.25 film thickness No. 14: 105 f 16% No. 65: 87 f 16% No. 166: 96 f 16% 22% (side by side

sample collection)

added to allfield and quality control samples. PUF or XAD-2 samples were extracted with 300 mL of a 1:l mixture of acetone and hexane for 24 h. After concentration (rotary evaporator) to 3 mL and solvent exchange to hexane, the samples were eluted through a column of 3% deactivated silica with a sodium sulfate cap to remove residual water and most of the nontarget, interfering compounds. PCBs were eluted in hexane, reduced to 0.3-1 mL by rotary evaporation and nitrogen blow-down, and stored at -20 "C. Air samples were not corrected for surrogate recovery. Water sample extracts in dichloromethane were eluted through anhydrous sodium sulfate to remove residual water, reduced to -3 mL using a Kuderna-Danish apparatus, and solvent-exchanged to hexane. The extracts were cleaned and fractionated using a liquid-solid chromatography column packed with 5 g of 100% deactivatedneutral alumina (450"C for 4 h, Brockman activity I) and 1 g of anhydrous sodium sulfate. Extracts containingthe PCB fractions were eluted with 2% dichloromethane in hexane (vlv). Prior to chromatographic analysis, the extract is reduced to about 100 mL, and PCB internal standards are added (Nos.30 and 204 as above). Due to the slight solubility of dichloromethane and the mechanics of the GLSE, surrogate recovery is predictably low. All PCB congener concentrations were therefore adjusted for these samples. Sample congeners with two and three chlorines were corrected using No. 14,congeners with four and five chlorines were corrected using No. 65, and congeners with six through nine chlorines were corrected with No. 166. The PCBs were quantified by the internal standardresponse factor method and a standard mixture ofAroclors 1232,1248,and 1262 as outlined by Swackhamer (27)and Hornbuckle et al. (26). Each sample was spiked with internal standards (2,4,6-trichlorobiphenyl, IUPAC No. 30, and 2,2',3,4,4',5,6,6'-octachlorobiphenyl,IUPAC No. 204) prior to the final concentration to 50-500 mL. Samples were analyzed using a Hewlett-Packard 5890 gas chromatograph equipped with a Ni-63electron capture detector, a Hewlett Packard 7673 autosampler, and either a Maxima 820D Data System (Waters) or a Hewlett Packard 3365 Chem-Station. Appropriate quality control measures included field blanks, solvent blanks, PCB standard surrogate recoveries, and chromatographic replicates (Table 1).

6.4 f 4.5 ng (PUF) 23 f 12 ng (XAD) 24 pg/m3 (PUF) 72 pg/m3 (XAD)

The number of PCB congeners analyzed varied with sampling and analytical method. A set of 77 congeners or coeluting congener groups were common to all three methods. This set includes congeners found in the highest concentrations in natural systems, with the exception of congener groups 4+10 and 8+5. These dichlorobiphenyls are expected to be present in the environment but were eliminated due to chromatographic interferences.

Results and Discussion Instantaneous Fluxes. Instantaneous CPCB fluxes were calculated for pairs of air and water samples collected throughout Lake Michigan aboard ship in September 1991. Air samples were collected during travel between water sampling sites as outlined in Table 2 and in Figure 1. Concentrations of CPCBs in the atmospheric vapor and dissolved water phases are listed in Table 2,along with the average surface water temperature, wind speed, and resulting calculated fluxes. Concentrations of CPCBs were highest in southern Lake Michigan in both water and air (Figure 2). Dissolved CPCB concentrations ranged from 1.2 nglL in southwestern Lake Michigan to 0.29 ng/L in central Lake Michigan. The sites nearest Chicago, IL, Milwaukee,WI, and the southwest shore of Lake Michigan (sites 17 and 11)exhibited dissolved CPCB levels four times higher than measured at the other sites (18-47q). Vapor CPCB concentrations ranged from a high of 1.6 ng/m3over southern Lake Michigan to a low of 0.16 ng/m3over northern Lake Michigan. Winds during the period of sampling were from the south and implicate the urban-industrial centers of the southern shore as major sources to the lake. Calculated fluxes for air and water sample pairs collected in September 1991 vary in response to wind speed and water to air concentration gradients. All fluxes are from water to air (net volatilization) and range from 24 to 220 ng m-2 day-'. There is no north-south gradient for calculated fluxes for two reasons. First, the north-south concentration gradient is greater for water than for vapor CPCBs. The disproportionate change in concentrations affects the (C,- C) term in the flux equation (eq 1) and results in a low net volatilization flux at the south central site and higher fluxes in other regions of the lake (Figure 3). Second, the average wind speed measured in the VOL. 29, NO. 4, 1995 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

871

TABLE 2

Dissolved and Vapor Total PCB Concentfations, Water Tenyerature, Wind Speed, and Calculated lnstananeous Fluxes sampling site 11,17 18, 19, 23, 27 34 40,41 b, 47, 47q Green Bay (GB)b GB and northern lakeb northern lakeb northern lakeb

dissolved PCB (pg/L)’ 1230 286 344 366

vapor sample name

(pS/m3’

1LM91 2LM91 3LM91 4LM91 1SUP92 2SUP92 1LM92 2LM92

1540 715 283 117 414 332 330 38 1

vapor PCB date Sept 6-7, 1991 Sept 7-8,1991 Sept 8-9, 1991 Sept 10-11, 1991 May 3, 1992 May 5, 1992 Aug 17,1992 Aug 18,1992

water temp (“C)

wind speed (m/s)

23.5 22.0 22.0 19.0 17.0 15.5 17.5 17.8

4.47 3.66 6.09 6.50 3.25 3.00 8.36 5.30

a Dissolved concentrations from listed sites were obtained by averaging individual congeners and summing the average. collected.

flux (ng m-2 day-’) 199 23.8 186 223

No water samples

......... ......... Sept. 10-11

Aug. 18

Chicago

1991 Sampling

1992 Sampling

FIGURE 1. Lake Michigan air and water sampling sites.

northern region of the lake was almost twice as high as that measured in the southern region. Higher wind speeds yield higher mass transfer coefficients which increase the calculated flux. The small variation in temperature (Table 2) over the lake did not significantly affect the calculated variation in fluxes. Annual Fluxes. The overall goal of this study was to determine the direction and magnitude of air-water exchange of CPCB and PCB homologs on a seasonal and annual time scale. Recent work in Lake Superior suggests that seasonal fluxes are a function of the variation in water 872

ENVIRONMENTAL SCIENCE &TECHNOLOGY / VOL. 29. NO. 4.1995

temperature, wind speed, and PCB vapor concentrations (7). Lake Superior CPCB volatilization fluxes were greatest in the fall when “skin”water temperatures were the highest and vapor concentrations were the lowest. Vapor deposition was predicted for the spring when vapor CPCB concentrations are high and water temperature is low. McConnell et al. (4) report similar results based on their calculations for hexachlorocyclohexanes (HCHs) in the lower Great Lakes. Seasonal and annual fluxes in Lake Michigan were calculated in this study using (a) vaporphase PCB concentrations from samples collected over 19

I

\

\P

YAid

0.0 VaporZF‘CB (ng/m3)

J

d.0 0.8 I .6 Dissolved ZPCB (ng/L)

FIGURE 2. Vapor and dissolved XPCBs in lake Michigan. B a n of the same pattern represent an air-water pair used in calculating instantaneousfluxes. Location of the bars on the map correspond to the sample collection region.

months on the northeastern shore of Lake Michigan, (b) wind speeds measured at the air sampling site during the sampling period, and (c) surface water temperatures. The ai-water exchange model includes the effect of temperature on diffusivity, waterviscosity and density,and

Henry’s lawconstants. The modified two-layer modelused here requires the temperature at the air-water interface. When calculating the 1991 instantaneous fluxes, we used the water temperature measured within the isothermal uppermost meter of the surface. For calculating the seasonal fluxes, we used surface temperatures that correspond to the upper millimeters of the surface water. The National Oceanic and Atmospheric Administration/Great Lakes Environmental Research Laboratory (NOAAIGLERL) provided “skin”water temperatures based on satellite and airborne surface temperature imaging measured by the Canadian Atmospheric Environment Service (AES) and NOAAIGLERL from 1988 to 1993 and modeled at G L E E (26, 29). Due to energy losses from long-wave radiation and evaporation, the surface skin is usually some tenths of a degree Celsius cooler than the temperature of the directly underlying, well-mixed water layer (26.30). For this study, all surface water temperatures north of 43” latitude were averaged in semimonthly (-2 week) periods. Wind speeds were recorded over the entire air sampling period from a 10-m tower at the land-based site. Because wind speed did not vary by season and we wanted the model to reflect atypical year, windspeedswereaveraged bymonth foruse in the air-water flux equations. Wind speeds, water temperatures, and the resulting mass transfer coefficients are plotted in Figure 4. Wind speed is an important factor in the estimation of the mass transfer coefficientfor transfer across the water-side layer, k,,and for transfer across the air-side layer, k,. Small variations in wind speed (Le., 1 m/s) affect the magnitude of k,as much as 10“C variations in water temperature. The air-sidemass transfer coefficient, k,, is strongly dependent on wind speed at the interface, but the resistance to air-side exchange, Ilk& is usually less than 20% of the overall resistance to mass transfer. Table 3 lists the mass transfer parameters calculated for the homologs at 0 and 15“C. Achman etal. (3detailed the

250

8

800

7

700

6

6W

200 h

s

mh

51

“E i;b

5 6 ’ a m

150

E.

v

4

X

a

E m

z

a

500

2 400

h

100

3

V e,

05

2

300

h

63u v

n

v

2

200

I

100

0

0

50

0

South

South Central

North Central

North

FIGURE3. Instantaneous XPCB flux across the air-water interface. PCB fluxes (ban) are calculated using the air-water concentration gradients. C, - ? I (long dashed line), wind speeds (solid line). and water temperatures (shon dashed line). All data collected in September 1991.

VOL. 29,

NO. 1.1995 I ENVIRONMENTAL SCIENCE &TECHNOLOGY.

873

.

n e

0.4

1I

. -_ . .- . I

.. .

*en

...

. . .. .

FIGURE 4. (A) Mass transfer coefficients calculated for each of the 37 days that air samples were collected at Sleeping Bear Dunes (SBD site). Plotted coefficients represent the average of seven homologs.(B) Normal surface temperaturesforthe whole lake (dotted line); normal surface temperatures for the northern three-quarters of the lake (solid line); and wind speeds recorded at SBD site.

effects of wind speed and water temperature on calculated volatilization fluxes from Green Bay, Lake Michigan, and found that an increase in wind speed from 4 to 8 m/s increased CPCB flux by a factor of 8 and that an increase in temperature from 10 to 20 "C resulted in a doubled CPCB flux. The air-water exchange of vapor CPCB in the southern quarter of the lake was not modeled here because of inadequate seasonal air and water concentration data to describe this region. The impact of the Gary-ChicagoMilwaukee corridor and its industrial activity on seasonal PCB water and vapor concentrations is unknown, but concentrations are expected to be higher than the region to the north (31, 32). Figure 4B describes the seasonal variation in surface water temperatures in Lake Michigan for the whole lake and for the area north of 43" latitude. Although the average surface temperatures of the whole lake and the northern part differ by no more than 2 "C,the northern temperatures are used to emphasize that the fluxes are applicable only to the northern three-quarters of the lake. Atmospheric concentrations of SOCs measured on land are reported to vary seasonally with maximum concentrations measured in the summertime (33-38). Samples collected seasonally on Siskiwit Lake on Lake Superior's Isle Royale; at Bloomington, IN; and at Egbert in southern Ontario have exhibited this trend (7, 33-35, 38). In this study of seasonal air-water exchange, 37 air samples collected between December 1991and June 1993were used. 874 1 ENVIRONMENTAL SCIENCE &TECHNOLOGY / VOL. 29, NO. 4, 1995

The samples were collected one to three times per month at Sleeping Bear Dunes National Lakeshore. The CPCB vapor concentrations varied from 30 to 400 pg/m3 and did not exhibit the expected seasonal trend (Figure 5B). The summer average CPCB value from the SBD samples was not significantly higher than the winter value (t-test, 95% confidence interval). It is not clear why no seasonal variation was observed at this site. No simple correlations of daily and seasonal CPCB vapor concentrations with air temperature, wind speed, or 4-day back-trajectories of the sampled air mass were found. Unlike vapor CPCBs measured near Lake Superior (7) and in southern Ontario (381,no seasonally dependent model could be fitted to the data used in this study. Therefore, actual daily vapor PCB concentrations were applied in this study. Vapor-phase and dissolved-phase PCB concentrations were used in calculating the homolog fluxes. In the vapor phase, the homolog distributions exhibited enrichment in the lower molecular weight, more volatile species. Trichlorobiphenyls made up 35%of the total, tetrachlorobiphenyls were 33%, pentachlorobiphenyls were 20%, and hexachlorobiphenyls were 8% of the total average vapor PCB concentrations. The homologs for the 10 dissolved-phase samples were similarly distributed: 27% of the average CPCB were trichlorobiphenyls, 29% were tetrachlorobiphenyls, 21% were pentachlorobiphenyls, and 15% were hexachlorobiphenyls. Dichlorobiphenyls were a small percentage of the total vapor and dissolved concentrations because several congeners that were expected to dominate this group were eliminated due to interference in the dissolved-phase samples. Seasonalvariation in dissolved-phase CPCB has not been published for the open waters of the Great Lakes. The dissolved PCB concentration used to calculate flux is an arithmetic average concentration, by homolog, of the nine water samples collected in 1991 at sites 18-47q (360 f 100 pglL). Samples collected near Milwaukee (site 17 and 11) were not included. The congeners were averaged and separated into homolog groups. One average value for each dissolved-phase homolog group was then used in calculating the seasonal fluxes. The seasonal variation in XPCB fluxwas calculated using the homolog vapor concentrations from 37 air samples collected over 19 months; the homolog dissolved concentrations from an average of 10 water samples; monthly average wind speed collected at the SBD site; and semimonthly normal surface water temperatures. Figure 5A describes the CPCB flux calculated for the 37 days. These fluxes are within the range of the instantaneous fluxes calculated from samples collected during the September 1991 sampling cruise. The annual flux of PCBs was calculated by averaging the fluxes in Figure 5 by month (ng rnW2day-') and multiplying by the number of days in the month and the surface area north of 43" latitude (41900 km2)or about 72% of lake area. The annual volatilization flux is 12.3pg m-2 yr-l or 520 kglyr from this portion of the lake. This is comparable to the range of 3.1-8.1 pg m-2 yr-' reported by Swackhamer et al. (9)but is lower than the value of 8 8 . 9 ~m-2 8 y r l of Strachan and Eisenreich (2)and lower than the flux of 34.2 pg m-* yr-l calculated from Achman etal. (5). Strachan and Eisenreich applied a mass budget calculation that covered all processes; Swackhamer et al. used a self-consistent set of measured data applied to a mass balance model and accounted for air-water

TABLE 3

Mass Transfer Parameters for February and July Water Temperatures and Wind Speeds: 0 "C (3 ink) and 15 "C (2 mls) log HLC

(mm3/mol)

PCB homolog

A. (CdISl

0. (cdls)

kw Idday)

h (ddavl

4 Idday)

0.223 0.218 0.214 0.209 0.204 0.199 0.197 0.192

273 265 258 252 246 242 231 232

0.161 0.153 0.122 0.083 0.055 0.029 0.020 0.011

0.206 0.202 0.197 0.192 0.187 0.185 0.182 0.178

334 325 316 308 302 295 290 285

0.181 0.174 0.154 0.122 0.093 0.056 0.042 0.024

AtO'C

hexachlorooipnenyl heptacnlorobiphenyl octach orooiphenyl nonachlorooophenyl

-4.41 -4.45 -4.69 -5.00 -5.26 -5.60 -5.77 -6.04

3.01 x 2.85 x 2.72 x 2.60 x 2.49 x 2.40 x 2.31 x 2.23 x

0.053 0.051 0.049 0.047 0.045 0.044 0.042 0.041

oichlorobiphenyl trichlorobiphenyl tetrachlorobipnenyl penrachlorooiphenyl hexacnlorobiphenyl heptach orob phenyl ocracnlorobipnenyl nonachlorobophenyl

-3.91 -3.95 -4.19 -4.50 -4.76 -5.10 -5.27 -5.54

4.96 x 4.70 x 4.48 x 10-6 4.28 x 4.11 x 3.95 x 10-6 3.81 x 10-6 3.68 x

0.059 0.056 0.054 0.052 0.050 0.048 0.047 0.045

dichlorobiphenyl trichlorobiphenyl tetrachlorobiohenvl

At 15 "C

~

~~

-H20 temp 60

24

50

20

40

16

30

12

20

8

10

4

0

0

-10

-4

-20

-8

400

350

300 250 200

150 100 50 0

exchange by difference. The annual flux of ZPCBs from the northern three-quarters of Lake Michigan is dominated by volatilization of tri- and tetrachlorinated biphenyls, w h i c h account for about 80% of the total annual flux. For

the area north of 43 "N. the lowest monthly flux of 15 kgl m o n t h was calculated for April while the largest flux, 70 kglmonth, was calculated for September. B y season, the greatest volatilization flux is in the fall, at 180 kg, and the

VOL. 29. NO. 4.1995 I ENVIRONMENTALSCIENCE &TECHNOLOGY. 875

smallest volatilization flux is in the spring, at 70 kg. No net depositional fluxes were calculated for any month although four individual days exhibited net vapor deposition (Figure 5). Gross volatilization and gross deposition is useful for determining total sources and sinks and chemical water column residence times. Gross fluxes may be calculated by factoring out eq 1. Gross annual CPCB volatilization fluxes for the northern three-quarters of Lake Michigan is 770 kg while gross annual CPCB vapor deposition is 250 kg. Gross fluxes by season and homolog are available as supplementary material to this article. The effect of local high wind speeds on the magnitude of the air-water exchange in the northern three-quarters of Lake Michigan is obscured in the exchange model outlined here. The wind speeds applied were collected on shore at one point and averaged over 1 month. A better estimate of over-lake winds would require more collection sites and some evaluation of the over-land and over-water differences in wind speed. In addition, this model applies the arithmetic average wind speed measured for the month in order to provide a seasonal estimate of the mass transfer coefficients. However, the mass transfer coefficient is expressed as an exponential function. Thus, short duration high wind speeds induce exponentiallyhighermass transfer coefficients (eq 6). For this reason, we believe that 520 kg/yr for the northern portion is a conservative estimate of volatilization from Lake Michigan. Analysis of the uncertainty in this method for calculating PCB fluxes was performed for similar work on Lake Superior (7). The error due to the use of averaged wind speeds in the calculation of mass transfer coefficients is not considered for the reasons discussed above. The remaining uncertainty was evaluated using propagation of error analysis and confirmation of the annual flux using an independent mass balance method. We conclude that the annual flux is reasonable within a factor of 2 for Lake Superior in 1991-1993 and for Lake Michigan for 1992.

Conclusion Gas exchange of PCBs from the northern two-thirds of Lake Michigan was calculated on an instantaneous and seasonal basis. Instantaneous volatilization fluxes determined for September 1991 ranged from 24 to 220 ng m-2 day-'. Seasonalestimates of PCB fluxes indicated net volatilization of 71 kglseason in the spring to 190 kg/season in the fall and is reflective of the seasonal variation in surface water temperature. Depositional flux was calculated for only 4 days in the spring. An annual flux of 520 kg/yrwas estimated for the lake north of 43" latitude. Due to the fact that periodic high wind speeds over the lake can lead to exponential increases in the mass transfer coefficients,this estimate is probably conservative. Volatilization of PCBs from Lake Michigan is a major sink of the compounds from the lake and a major source to the regional atmosphere. PCB removal from Lake Michigan via net volatilization is about one-third the loss viasedimentation (1600-1800 kg/yr) (39).Lake Michigan's contribution to atmospheric PCBs equals or exceeds volatilization from landfills (10-100 kg/yr) (40),Green Bay (130 kg/yr) (3,and Lake Superior (250-500 kglyr) (6, 7).

Acknowledgments We gratefully acknowledge the assistance of the captain and crew of the R/V Lake Guardian (U.S. EPA). We thank 876

ENVIRONMENTAL SCIENCE &TECHNOLOGY / VOL. 29, NO.

4, 1995

Sherman Bauer and Ilora Basu (IllinoisState Water Survey) for meteorological information at the SBD site and for analytical work on the samples collected at the SBD site, respectively, and Tom Van Zoeren (National Park Service) for sampling operation at the SBD site. Further acknowledgments are due to Tim Hunter (NOAAIGLERL)and Karl Schneider (formerlyNOAA/GLERL;currently University of Munich) for providing normal surface water temperatures. This research has been supported in part by the Great Lakes National Program Office of U.S. EPA (EPA/R995233),the Great Lakes Protection Fund (FG 6901029), and the MN Sea Grant College Program OR1231 supported by the N O M Office of Sea Grant, Department of Commerce.

Supplementaty Material Available Four tables (Tables 4-7) containing the average concentrations of each of the analyzed vapor-phase and dissolvedphase PCB congeners and homologs collected at SBD site (Table 4); the ZPCB vapor concentrations, wind speed, and wind direction data collected from SBD site and applied to the flux equations (Table 5);normal surface water temperatures in Lake Michigan, whole lake values and values for the area north of 43" latitude (Table 6);values (kg) for gross volatilization and gross deposition for four seasons and eight homologs (Table 7 ) (6 pages) will appear following these pages in the microfilm edition of this volume of the journal. Photocopies of the supplementary material from this paper or microfiche (105x 148 mm, 24x reduction, negatives) may be obtained from Microforms Office, American Chemical Society, 1155 16th St. N.W., Washington, DC 20036. Full bibliographic citation (journal, title of article, names of authors, inclusive pagination, volume number, and issue number) and prepayment, check or money order for $16.50 forphotocopy($18.50foreign)or $12.00 for microfiche ($13.00 foreign), are required. Canadian residents should add 7% GST.

Literature Cited (1) Atlas, E.; Giam, C. S. In The Role of Air-Sea Exchange in Geochemical Cycling;Buat-Menard, P., Ed.; D. Reidel Publishing Co.: Dordrecht, Holland, 1985; Vol. 185, pp 295-329. (2) Strachan, W. M. J.; Eisenreich, S. 1. Mass Balancing of Toxic Chemiicalsin the GreatLakes: theRole ofAtmosphencDeposition. Reuort to the International Joint Commission, Windsor, Ontario, Miy 1988, 161 pp. (3) Iwata. H.: Tanabe, S.: Sakai, N.; Tatsukawa, R. Environ. Sci. Technol. 1993, 27, 1080-1098. (4) McConnell, L. L.; Cotham, W. E.; Bidleman, T. F. Environ. Sci. Technol. 1993, 27, 1304-1311. (5) Achman, D. A.; Hornbuckle, K. C.; Eisenreich, S. J. Environ. Sci. Technol. 1993, 27, 75-87. (6) Jeremiason, J. D.; Hornbuckle, K. C.; Eisenreich, S. J. Environ. Sci. Technol. 1994, 28, 903-914. (7) Hornbuckle, K. C.; Jeremiason, J. D.; Sweet, C. W.; Eisenreich, S. J. Environ. Sci. Technol. 1994, 28, 1491-1501. (8) Mackay, D. J. Great Lakes Res. 1989, 15, 283-297. (9) Swackhamer, D. L.;Armstrong, D. E. Environ. Sci. Technol. 1986, 20, 879-883. (10) DeVault, D. S.; Willford, W. A.; Hesselberg, R. J.; Nortrupt, D. A. Arch. Environ. Contam. Toxicol. 1986, 15, 349-356. (11) Dar,E.;Kanarek, M. S.;Anderson, H.A.; Sonzogni, W.C. Environ. Res. 1992, 59, 189-201. (12) Liss, P. W.; Slater, P. G. Nature 1974, 247, 181-184. (13) ten Hulscher, T. E. M.; van der Velde, L. E.; Bruggeman, W. A. Environ. Toxicol. Chem. 1992, 11, 1595-1603. (14) Schwarzenbach, R. P.; Gschwend, P. M.; Imboden, D. M. Environmental OrganicChemistry;Wdey Interscience: NewYork, 1993. (15) Smith, 1.; Bomberger, D.; Haynes, D. Environ. Sci. Technol. 1980, 14, 1332-1337. (16) Reid, R. C.; Prausnitz, 1. M.; Poling, B. E. The Properties of Gmes and Liquids; McGraw-Hill Inc.: New York, 1987; pp 586-605. (17) Wanninkhof, R.; Ledwell, 1. R.; Crusius, J. In Air-Water Mass Transfer; Wilhelm, S. C., Gulliver, J. S., Eds.; American Society of Civil Engineers: New York, 1991; pp 441-458. (18) Wanninkhof, R. H.; Bliven, L. F. J. Geophys.Res. 1991, 96,27852796.

(19) Watson, A. J.; Upstill-Goddard, R. C.; Liss, P. S. Nature 1991,349, 145-147. (20) Upstill-Goddard, R.; Watson, A.; Liss, P.; Liddicoat, M. Tellus 1990, 42B, 364-377. (21) Wilke, C. R.; Chang, P. A.I.Ch.E.J. 1955, 1 , 264-270. (22) Holmen, K.; Liss, P. Tellus 1984, 36B, 92-100. (23) Danckwerts, P. V. Ind. Eng. Chem. 1951, 43, 1460-1467. (24) Pearson, R. F. M.S. Thesis, Environmental Health, University of Minnesota, 1994. (25) Neilson, M.; Stevens, R.; Biberhofer, J.; Goulden, P. D.; Anthony, D. H. J. Inland Waters Directorate Technical Bullentin 157; National Water Research Institute: Burlington, Ontario; 1987. (26) Hornbuckle, K. C.; Achman, D. R.; Eisenreich, S. J. Environ. Sci. Technol. 1993, 27, 87-98. (27) Swackhamer, D. L. Quality assurance plan for the Green Bay Mass Balance Study 1 . PCBs and Dieldrin; Great Lakes National Program Office of the United States Environmental Protection Agency: Chicago, IL, 1988. (28) Schneider, K.; Assel, R. A.; Croley, T. E., 11. NOM Technical Memorandum ERL GLERL-6’1; National Oceanic and Atmospheric Administration (NOAA), Great Lakes Environmental Research Laboratory: Ann Arbor, MI (GLERL),Aug 1993; 47 pp. (29) Croley, T. E., 11. Water Resour. Res. 1992, 28, 69-81. (30) Paulson, C. A.; Parker, T. W. J. Geophys. Res. 1972, 77,491-495. (31) Holsen, T.; Noh K.; Liu, S.; Lee, W. Environ. Sci. Technol. 1991, 25, 1075-1081. (32) Murphy, T . J. In Toxic Contaminants in the Great Lakes; Nriagu,

(33) (34) (35) (36)

(37) (38) .~ (39) (40)

J. O., Simmons, M. S., Eds.; John Wiley and Sons: New York, 1984, pp 53-79. Hermanson, M. H.; Hites, R. A. Environ. Sci. Technol. 1989,23, 1253-1258. Manchester-Neesvig, J. B.; Andren, A. W. Environ. Sci. Technol. 1989, 23, 1138-1148. Swackhamer, D. L.; McVeety, B. M.; Hites, R. A. Environ. Sci. Technol. 1988,22, 664-672. Sweet, C. W.; Basu, I.; Harlin, K. Presented at the 86th annual Meeting and Exhibition of the Air and Water Waste Management Association, Denver, CO, June 13-18, 1993; AWMA: Pittsburgh, PA; 93-RP-137.03. Monosmith, C. Final report to theMichigan GreatLakesProtection Fund Michigan Department of Natural Resources Air Quality Division: Lansing, MI, July 12, 1993; 80 pp. Hoff. R. M.: Muir, D. C. G.: Grift, N. P. Environ. Sci. Technol. 1992,26, 266-275. Golden, K. A. M.S.C.E. Thesis. Universitv of Minnesota, 1994. Murphy, T.; Formanski, L.; Brownawel1,’B.; Meyer, J. Environ. Sci. Technol. 1985, 19, 942-946.

Received for review April 11, 1994. Revised manuscript received November 18, 1994. Accepted December 14, 1994.@

ES940220A @

Abstract published in Advance ACS Abstracts, February 1, 1995.

VOL. 29, NO. 4, 1995 / ENVIRONMENTAL SCIENCE &TECHNOLOGY 1877