Environ. Sci. Technol. 2007, 41, 2331-2337
Watershed Processing of Atmospheric Polychlorinated Biphenyl Inputs A M Y A . R O W E , † L I S A A . T O T T E N , * ,† GREGORY J. CAVALLO,‡ AND JOHN R. YAGECIC‡ Department of Environmental Sciences, Rutgers University, 14 College Farm, Road, New Brunswick, New Jersey 08901, and Delaware River Basin Commission, 25 State Police Drive, West Trenton, New Jersey 08628
Indirect atmospheric deposition of PCBs was examined in subwatersheds of the Delaware River Estuary. Tributary PCB loads and atmospheric PCB concentrations were used to understand the pass-through efficiencies for nine rivers/ creeks for which PCB inputs appeared to be dominated by atmospheric deposition. The pass-through efficiency, E, was calculated from tributary loads and atmospheric deposition fluxes. Unfortunately, uncertainties in the gaseous and dry particle deposition velocities, vg and vd, respectively, render the calculated atmospheric deposition fluxes highly uncertain. In order to circumvent this problem, export of PCBs from the watershed was related directly to atmospheric PCB concentrations via a new mass transfer coefficient, the watershed delivery rate or vws, which describes the process by which the watershed transfers PCBs from the air to the River’s main stem. vws increases with increasing chlorination and is significantly correlated with vapor pressure. This trend suggests that the transfer of PCBs from the atmosphere to the River via the watershed is more efficient for high molecular weight PCBs than for low molecular weight PCBs. This may indicate that the selected watersheds are at or close to equilibrium with respect to gaseous exchange of PCBs, such that lower molecular weight congeners undergo substantial revolatilization after deposition. The magnitude of the pass-through efficiency, E, depends on the deposition velocities used to calculate the atmospheric deposition flux, but when congener-specific deposition velocities are used, E is independent of vapor pressure and is relatively constant at about 3%.
Introduction Indirect atmospheric deposition is atmospheric deposition to land surfaces that can be remobilized to enter surface waters. Because most water bodies have a high ratio of watershed surface area to water surface area, indirect atmospheric deposition is potentially a more important source of pollutants than direct atmospheric deposition. Polychlorinated biphenyls (PCBs) are one class of chemicals for which atmospheric deposition has shown to be an important delivery mechanism to watersheds (1-8). There is little published research on indirect atmospheric deposition * Corresponding author e-mail:
[email protected]. † Rutgers University. ‡ Delaware River Basin Commission. 10.1021/es062136o CCC: $37.00 Published on Web 03/01/2007
2007 American Chemical Society
of PCBs. In a previous publication, indirect atmospheric deposition parameters for pentachlorinated PCBs were calculated for the Delaware River watershed (9). This study expands on our previous work by including congeners from homologues 3-8 and more carefully examining watershed characteristics. The Delaware River acts as a natural boundary between the mid-Atlantic states of Delaware, New Jersey, and Pennsylvania, eventually flowing into the Atlantic Ocean. The Delaware River Basin Commission (DRBC) recently established a Total Maximum Daily Load (TMDL) for PCBs in zones 2-5 of the river (10). The TMDL was developed after construction of a detailed water quality model, which required estimates of all PCB loadings to the river, including atmospheric deposition (10). Due to a court-imposed deadline for the TMDL, it was calculated based on the results of a water quality model for the pentachlorinated PCB homologue (penta-PCB) and extrapolated for total PCBs. Thus previous estimates of indirect atmospheric deposition and watershed pass-through were based on these penta-PCB loading estimates (9). Loadings for additional congeners are now available, allowing for an expanded investigation of the watershed processing of atmospheric PCB inputs. The purposes of this paper are to investigate the magnitude of indirect atmospheric deposition of PCBs and the fate of those PCBs in the subwatersheds of the Delaware River, including the fraction that passes through the watershed to eventually reach the River.
Methodology Site Characterization. The Delaware River flows out of the Catskill Mountains in New York State, eventually emptying into the Atlantic Ocean (Figure 1). The Delaware River Estuary begins south of the head of tide at Trenton, NJ, (river mile 133.4), but the river is not saline until south of Philadelphia, around river mile 60. The estuary is divided into 5 water quality zones (Figure 1). The 2001 USGS National Land Cover Data Set (http://www.mrlc.gov/mrlc2k_nlcd.asp) indicates that the land use in this area is a complex mixture of lowintensity residential, mixed and deciduous forests, row crops, and wetlands. PCB Measurements. Tributary PCB loads to the main stem Delaware River were estimated by DRBC using gaged or estimated daily flows for each tributary and multiplying those by tributary-specific wet weather or dry weather PCB concentrations (11). Tributaries were sampled between September 2001 and March 2003. Details of the process used to calculate the tributary loads are presented in the TMDL report (11). All tributary water samples were analyzed for whole-water PCB concentrations via EPA method 1668a. PCB loads from 20 tributaries were calculated for individual congeners, coeluting congener groups, and homologues for the period from September 1, 2001 to March 31, 2003 (the model calibration period). Atmospheric deposition was characterized via sampling at seven air monitoring stations (Figure 1) along the river during 2001-2002 (9). Here, only the concentrations at Washington’s Crossing, Lum’s Pond, and Alloways Creek will be used in the atmospheric deposition calculations. These stations have background concentrations of PCBs and are not likely to be influenced by the urban areas of Philadelphia and Camden (7). Although portions of some of the watersheds are closer in proximity to sites such as Swarthmore and Northeast Philadelphia Airport, which have been shown to exhibit elevated atmospheric PCB concentrations (9), the data from these sites were not used because passive air VOL. 41, NO. 7, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2331
FIGURE 1. Map of the water quality zones, the monitoring sites (circles), and the tributaries of the Delaware River. Tributaries in capital letters are thought to receive PCB inputs primarily from the atmosphere.
samplers installed at several locations within these watersheds displayed low PCB masses, similar to masses observed at the regional background site of Lum’s Pond (data not shown). For the atmospheric deposition measurements, atmospheric samples were analyzed for 60 PCB peaks representing 93 congeners (9). Because the atmospheric data and tributary data were collected using different analysis methods, the congener list from the tributary data had to be composited to represent the coeluting PCB congeners detected via the ECD method. For example, the tributary data reported concentrations of PCBs 52 and 43 separately. These concentrations were summed for comparison to the data on atmospheric concentrations of PCBs 52 + 43, which coelute via the ECD method. 2332
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 7, 2007
Calculation of Deposition Fluxes. The total atmospheric flux, Fatm, is the sum of the wet (Fwet), dry particle (Fdry), and gaseous (Fgas) deposition fluxes (ng m-2 d-1):
Fatm ) Fdry + Fwet + Fdry ) Cp‚vd + CVWM‚vp + Cg‚vg
(1)
Cp and Cg are the PCB concentrations in the particle and gas phases (ng m-3), respectively. CVWM is the volume-weighted mean concentration of PCBs in precipitation (ng m-3). vd and vg are the dry particle and gaseous deposition velocities (m d-1), and vp is the average precipitation intensity (meters of rain per day). The geometric mean was calculated for all Cg + Cp concentrations from all samples at the Washington’s Crossing
(WC), Lum’s Pond (LP), and Alloways Creek (AC) sites. These sites were chosen because they are suburban sites that show regional background PCB concentrations and are in close proximity to the Delaware River (7). The geometric mean was used because the data were log-normally distributed. CVWM was obtained by assuming a constant particle scavenging coefficient (Wp) of of 1.0 × 105 (9) and the geometric mean Cp concentration:
CVWM ) Wp‚Cp
Calculation of Pass-Through Efficiency (E). In a previous investigation, the pass-through efficiency (E) of penta-PCB was calculated for each watershed in which atmospheric deposition was thought to be the dominant source of PCBs (9). E was defined as the ratio of the tributary yield (Fws) and the atmospheric deposition to the watershed
E)
(2)
Precipitation data were obtained from the National Weather Service’s Philadelphia International Airport monitoring station, and a daily average precipitation intensity for the model calibration period was used in the deposition calculations. A major difficulty encountered in estimating atmospheric deposition fluxes from concentration data is the choice of deposition velocities. Here we examine two approaches to the problem of selecting appropriate deposition velocities: applying one deposition velocity to all congeners and varying deposition velocity with molecular weight. Dry deposition velocities (vd) for PCBs and other semivolatile organic compounds (SOCs) have been estimated and measured via many techniques (Table S-1 of Supporting Information and refs 6 and 12-25). Most studies report one dry particle deposition velocity for ΣPCBs, although Lee et al. (20) found that vd increases with molecular weight. To select a single vd value for application to all congeners, the entire data set of reported vd values was examined. The reported vd values display a log-normal distribution with 12 values falling between the minimum value (0.09 cm s-1) and 0.5 cm s-1. Another 7 values fall between 0.5 and 1.8 cm s-1, and 3 values exceed 1.8 cm s-1. The best way to characterize a log-normal distribution is to use the geometric mean (0.46 cm s-1), which in this case is in good agreement with the median (0.5 cm s-1). Here, vd for PCBs was chosen to be 0.5 cm s-1, which is in good agreement with the vd value of 0.49 ( 0.23 cm s-1 calculated in a PCB study of the Chesapeake Bay (26). However, it should be noted that the dry particle deposition velocity, vd, is expected to be higher in terrestrial systems, which have very high surface roughness, than for water surfaces or the kinds of surrogate surfaces used in the studies listed in Table S-1. Very few values of the gaseous deposition velocity, vg, exist in the literature (Table S-2 of Supporting Information and refs 27-30). Measurement of vg is even more problematic than measurement of vd because gaseous deposition is reversible, and the measured deposition velocity depends on whether this reversible process is kinetically limited or has reached equilibrium (30). Also, the type of land surface appears to make a substantial difference in vg, with coniferous canopies and deciduous canopies giving rise to PCB deposition velocities that, in one study, differed by an order of magnitude (29). Ould-Dada (31) demonstrated that dry particle deposition velocities also vary with height in a forest canopy. Horstmann and McLachlan calculate faster gaseous deposition velocities for heavier PCBs in both coniferous and deciduous forests (29). Although there is a general consensus in the literature that the gas deposition velocity should be 1 order of magnitude lower than the dry particle deposition velocity (27), the published data do not reflect this (Table S-2). Thus, the choice of vg for the mixed land-use areas of this study is particularly difficult. Here again, we choose one deposition velocity to apply to all congeners, but we will consider the effect of varying vg on the pass-through efficiencies. The values in Table S-2, once again, appear to be log-normally distributed, and the geometric mean (0.47 cm s-1) and the median (0.46 cm s-1) are again similar. Thus, a value of 0.5 cm s-1 was chosen for the gaseous deposition velocity for this study.
Fws Fatm
(3)
where Fws is the watershed yield defined as the PCB load exiting the watershed (i.e., tributary load to the main stem Delaware River) divided by the surface area of the watershed (ng m-2 d-1). Fatm is also in these units, so the pass-through efficiency, E, is dimensionless. Given the levels of uncertainty associated with the atmospheric deposition velocities (vd, vg) involved in calculating Fatm, it may be more useful to relate the watershed yield directly to the atmospheric PCB concentrations, because these parameters are easier to measure. This was accomplished by defining a new mass transfer coefficient (vws) for the process by which the watershed transfers PCBs from the atmosphere downstream (9):
Fws ) vws‚Catm = vws‚(Cg + Cp)
(4)
Catm should reflect the concentrations of PCBs in all atmospheric phases: gas, aerosol, and precipitation. Our previous work suggested a very constant particle-phase scavenging ratio for PCBs in this region of 1 × 105 (9). This implies that Cp and CVWM are directly related, and therefore Cp can be used as a surrogate for CVWM in this equation. In our previous work we investigated only the pentachlorinated homologue and therefore assumed that both wet and dry particle deposition were negligible (9). This work examines heavier PCB homologues, however, so Cp is not negligible and remains in eq 4. vws is a useful parameter because it can be used to generate estimates of tributary loads to the river if the atmospheric PCB concentrations and watershed surface area (Aws) are known:
load ) vws‚(Cg + Cp)‚Aws
(5)
Results and Discussion Watershed PCB Yields. The PCB loads exiting each tributary, and the watershed surface areas are presented in the Supporting Information (Table S-3 of the Supporting Information). The loads are normalized to surface area and presented in Figure 2 as Fws. Note the emergence of the high molecular weight homologues 9 and 10 in the Raccoon, Salem, and Alloways creeks, which are all tributaries located on the southeastern Delaware near a documented source of PCBs 206, 208, and 209. In order to examine the importance of indirect atmospheric deposition, tributaries in which the PCB inputs are dominated by atmospheric deposition must be identified. One way to do this is to examine Fws. Watersheds with high Fws values can be assumed to receive substantial PCB inputs from sources other than the atmosphere. Thus the tributaries were ranked by their Fws values. Tributaries in Figure 2 are listed from north to south. The urban area of Philadelphia and Camden is located roughly between the Pennsauken and Chester Creeks. The Assunpink flows through Trenton, NJ. These tributaries appear to be influenced by an urban signal, while those outside of these urban zones may be affected mainly by atmospheric deposition. The nine tributaries with the lowest Fws values are as follows: Crosswicks, Neshaminy, Rancocas, Pennypack, Raccoon, Brandywine, Christina, Salem, and Alloways Creeks. These nine will VOL. 41, NO. 7, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2333
FIGURE 2. Watershed flux (Fws; load divided by surface area) for each homologue for 20 tributaries to the main stem Delaware River. Inset is a blowup of the nine tributaries thought to receive PCB inputs primarily from the atmosphere. Tributaries are arranged from north (left) to south (right).
TABLE 1. Watershed PCB Loads to the Delaware and Average Watershed Delivery Mass Transfer Coefficients (vws) by Homologue Group for the Nine Atmospherically Driven Tributaries, Listed from North to South homologue 3 4 5 6 7 8
Crosswicks 25 33 49 38 22 5.7
3 4 5 6 7 8
0.010 0.013 0.026 0.037 0.044 0.022
Neshaminy 14 18 30 26 11 3.5
Rancocas 29 80 130 140 74 26
0.0029 0.0044 0.0041 0.012 0.0096 0.025 0.013 0.049 0.014 0.056 0.0070 0.035 North of Philadelphia/Camden
Pennypack
Load (g y-1) 10 9.3 35 33 43 36 44 28 32 14 8.1 7.5
vws (cm s-1) 0.0090 0.034 0.057 0.097 0.15 0.067
hereafter be referred to as the “atmospherically driven tributaries”. A second method of determining which watersheds are atmospherically driven is to regress PCB loads against watershed surface area. This approach is based on the assumption that watersheds in which atmospheric deposition dominates the PCB inputs would show a significant correlation between watershed area and tributary PCB load. The R2 value increases as tributaries with higher Fws are eliminated. Best R2 values (ranging from 0.56 to 0.66 depending on homologue) were obtained with the nine atmospherically driven tributaries. Another way to eliminate tributaries with internal PCB sources was to examine National Pollutant Discharge System (NPDES) permits. Not surprisingly, tributaries with high Fws were often located near NPDES permittees (32). The final method of determining which watersheds are atmospherically driven was to compare congener patterns 2334
9
Raccoon
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 7, 2007
0.0095 0.034 0.053 0.074 0.073 0.059
Brandywine 44 86 150 170 92 37
Christina 4.3 19 83 79 25 7.5
Salem 7.2 23 21 16 7.6 4.5
0.0065 0.0013 0.0029 0.013 0.0061 0.011 0.028 0.039 0.012 0.056 0.058 0.017 0.066 0.041 0.017 0.045 0.027 0.015 South of Philadelphia/Camden
Alloways 5.9 17 25 22 11 9.3 0.0045 0.012 0.025 0.040 0.048 0.047
among the tributaries. Because the atmospheric data suggest that there exists a regional background PCB level with a congener pattern that is largely constant across the region (9), it follows that watersheds that are atmospherically driven should show PCB congener patterns that are similar to each other. Congener patterns were compared by constructing x-y scatter plots of the congener-specific PCB concentrations in each pair of samples. Dry weather samples frequently yielded much lower correlations than wet weather samples, although the reason is not immediately clear. The nine atmospherically driven tributaries displayed highly correlated congener patterns. In these nine tributaries, a total of 28 samples were taken for construction of their tributary loads, giving rise to 378 possible correlations between samples. Of these, 83% were strongly correlated (R2 > 0.6). When the 14 dry weather samples are removed from the data set, the remaining 14 wet weather samples give rise to 159 possible correlations, of which 92% yield R2 > 0.6. The prevailing
maximum amount of PCBs that could be present in the tributaries due to direct interaction with the atmosphere was calculated on a congener specific basis. This calculation assumed equilibrium between the gas phase, the dissolved phase, and the particulate organic carbon (POC). The Henry’s Law constant of Bamford et al. (34, 35) were used, and the organic carbon normalized partition coefficient (KOC) was calculated from Karickhoff’s equation (36):
log KOC ) 1.00 log KOW - 0.21
FIGURE 3. Log pass-through efficiency (E) vs log vapor pressure (log pL in Pa) for a representative tributary, Neshaminy Creek. When a constant deposition velocity is applied to all congeners (solid triangles), the slope of the line is significant and negative (-0.31). When coniferous deposition velocities are used (open diamonds), the slope is significant and positive (0.59). When deciduous deposition velocities are used (open squares), the slope is shallow and not significant, suggesting that pass-through efficiency (E) may be constant and independent of the physical-chemical properties of the PCB congener. congener pattern in these nine tributaries did not correlate highly with the congener pattern in the atmosphere (R2 typically < 0.1). This does not necessarily indicate that the PCBs in the tributary are not atmospherically derived, since the differences in congener patterns could be due to the different analytical techniques used and the extensive processing of PCBs in the watershed (as described below). The above criteria revealed nine rivers/creeks listed in Table 1 that follow a linear load vs surface area pattern, are located at a reasonable distance from the Philadelphia/ Camden urban area, contain no notable NPDES permittees, and have highly correlated congener patterns. Although PCB inputs to the Salem, Alloways, and Pennypack appear to be atmospherically driven, the loading data for those tributaries were based on only one or two samples each, so caution must be used in interpreting results from these three watersheds. Ironically, smaller numbers of samples were collected in these tributaries because it was thought they would exhibit low PCB loads due to their location and their lack of NPDES dischargers (i.e., the atmosphere would be the dominant source of PCBs). This sampling plan served the needs of the TMDL well but is obviously inconvenient for the present study. Atmospheric Deposition Fluxes. Atmospheric deposition fluxes were calculated for homologues 3-8. Homologues 1, 2, and 10 were not measured in the atmosphere, and homologue 9 was not included in this study due to the documented non-Aroclor source of PCBs 206, 208, and 209 in the Delaware River system (33). Here the atmospheric fluxes, Fdry, Fwet, and Fdry, are the same for all of the tributaries due to the use of an average value from the three land sites for Cp, Cg, and CVWM, as discussed above. Fatm was calculated to be 31, 22, 15, 6.2, 1.7, and 0.93 ng m-2 d-1 for homologues 3-8, respectively. As described above, the deposition velocities used to calculate these fluxes are highly uncertain, and therefore the uncertainty in these fluxes is large, at least 50%. Gaseous deposition is the dominant atmospheric deposition mode, especially for the lesser-chlorinated homologue groups, representing 96%, 96%, 94%, 84%, 79%, and 62% of Fatm for homologues 3-8, respectively. To ensure that the tributary loads primarily arise from indirect, rather than direct, atmospheric deposition, the
(6)
These calculations suggest that direct atmospheric deposition accounts for a maximum of 22%, an average of 4%, and a geometric mean of 2% of the ΣPCBs in the tributary samples. Watershed Delivery Rates. The pass-through efficiency (E) and watershed delivery rate (vws) follow the same trends; both increase with molecular weight, peaking at homologue 7, followed by a decrease for homologue 8 (Table 1). This pattern is followed closely for all the tributaries, although the drop off in homologue 8 is more pronounced for some tributaries than for others. In general, E is e5% for homologues 3-5 and as high as 43% for PCB 180 in the Pennypack watershed. The vws values range from 0.0004 to 0.24 cm s-1. The congener specific log vws (in cm s-1) values display a significant correlation (p < 0.01) with both liquid vapor pressure (log pL in Pa) and octanol-water partitioning coefficient (log KOW) with R2 values ranging from 0.17 to 0.67 for individual tributaries. For the pooled data set of all nine tributaries, the correlations were
log vws ) -0.30 ((0.05) log pL - 2.6 ((0.2) R2 ) 0.32 (7) log vws ) 0.42 ((0.08) log KOW - 4.5 ((0.5) R2 ) 0.30 (8) Pass-Through Efficiency. The pass-through efficiency (E) and watershed delivery rate (vws) are related, since both parameters dependent on the watershed yield (Fws) and PCB concentrations in the particle and gas phases (Cp and Cg). When constant values for vd and vg are used, E and vws are related by a nearly constant factor where E ≈ 1.9 * vws, and E, like vws, is correlated with the physical-chemical properties of the PCB congeners. However, vws and E are only coupled when deposition velocities (vd and vg) that are constant for all congeners are used. If congener-specific deposition velocities are used, vws and E are no longer directly proportional, and E may no longer depend on the physical-chemical properties of the congener. To investigate this possibility, the vg values of Horstmann and McLachlan (29) and the vd values of Lee et al. (20) were used to recalculate Fatm. The following congeners were investigated because congener specific vg values for them have been published by Horstmann and McLachlan (29): PCBs 31, 49, 92+84, 153+132, 141, 202+171+156, and 170+190. vg for PCB 138+158 was applied to the congener measured in the atmospheric samples that had the most similar vapor pressure, PCB 202+171+156. The resulting E values were calculated for a hypothetical deciduous and coniferous forest (since the Delaware River basin includes both types of forests) and compared with those calculated using constant vg and vd values of 0.5 cm s-1. The average E value resulting from using the coniferous, deciduous, and constant vg values were 31%, 3%, and 10%, respectively. Log E values were regressed against log pL. The results for a typical atmospherically driven tributary are shown in Figure 3. The choice of vd had virtually no effect on the correlation between log E and log pL because dry particle deposition is a small part of the overall Fatm. In contrast, the choice of vg made a significant difference in the strength of the correlation between log E and log pL. The correlation was strongest when vg was constant for all VOL. 41, NO. 7, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2335
congeners, and weakest when the deciduous congenerspecific vg values were used. When vg was assumed to be constant, eight of the nine atmospherically driven tributaries displayed a significant (p < 0.05) correlation between log E and log pL, and the average ((SD) slope was negative (-0.44 ( 0.10). When coniferous vg values were used, four of the nine atmospherically driven tributaries displayed a significant (p < 0.05) correlation between log E and log pL, and the average slope was positive (0.59 ( 0.09). When the deciduous vg values were used, none of the tributaries displayed a significant correlation between log E and log pL, and the average slope was therefore not significantly different from zero. In other words, the choice of congener-specific deposition velocities can completely remove the dependence of watershed pass-through efficiency on the physical-chemical properties of the PCB congeners. This makes sense if atmospherically deposited PCBs are removed from the watershed via a process that does not depend on their physical-chemical properties, such as particle erosion. The average ((SD) pass-through efficiency (E) obtained using deciduous deposition velocities was 3.3 ( 2.9%. Uncertainty in vws. Because of the high uncertainty associated with the deposition velocities, it is difficult to assign an uncertainty to the pass-through efficiencies. vws is associated with less uncertainty. Recall
vws =
Fws (Cg + Cp)
(9)
Since (Cg + Cp) is log-normally distributed, the uncertainty is also log-normal, and we write
log vws = log Fws - log (Cg + Cp)
(10)
Thus the absolute uncertainties in log Fws and log(Cg + Cp) can be summed to yield the absolute uncertainty in log vws. The uncertainties in the atmospheric concentrations log(Cp + Cg) averaged 0.37 log units for all congeners (one standard deviation about the log mean). The uncertainty in Fws is dominated by the uncertainty in the tributary loading. DRBC conducted Monte Carlo simulations to estimate the uncertainties in the penta-PCB watershed loads. This analysis determined that the main source of uncertainty in the tributary loads was the small number of water samples analyzed for PCBs, especially those collected under wet weather conditions. The analysis assumed that the tributary PCB concentrations were log-normally distributed and therefore resulted in log-normally distributed uncertainties. When loads are expressed as pg PCB per m2 of watershed surface area per day, the uncertainties are about 0.23 log units for the 9 tributaries thought to be atmospherically driven. We assume that the uncertainties in the tributary loads for the all individual congeners will be similar on a log-normal basis. Adding these uncertainties results in an uncertainty in the log vws of ( 0.60 log units. This suggests that the true value of vws has a 68% probability ((1 σ) of lying between 25% and 400% of the value given in Table 1. While this is a large range of uncertainty, it applies to the absolute value of vws. Relative vws values across congeners are more certain because they are ultimately a function of the congener patterns observed in the air and water. Since both of these congener patterns are highly conserved, the relative vws values are also conserved. This assessment of mathematical uncertainty assumes, of course, that all underlying assumptions are correct. In reality, the biggest source of the uncertainty in these calculations lies in the assumption that atmospheric deposition dominates PCB sources in these watersheds, which is difficult or impossible to prove. One indication that sources other than atmospheric deposition occur in these watersheds 2336
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 7, 2007
is the presence of PCBs 206, 208, and 209 in the tributary loads for Alloways, Salem, and Raccoon Creeks. It is unlikely that these heavy congeners could have been deposited in these watersheds via atmospheric deposition alone. PCB 206 represents less than 0.5% of ΣPCBs in both the gas and particle phases at Lum’s Pond, the site nearest the documented source of these congeners, yet it represents about 15% of ΣPCBs in Alloways Creek. Mechanism of Watershed Pass-Through. The increase in vws with increasing chlorination suggests that transfer of PCBs from the atmosphere to the river via the watershed is more efficient for high molecular weight than low molecular weight PCBs. We can rationalize this observation if we consider that the transfer process occurs in three steps. First, PCBs must be deposited to the watershed surfaces from the atmosphere. This process is described by the deposition velocities, vd and vg. Second, the PCBs may revolatilize from the watershed surfaces to the atmosphere. Some consideration of this revolatilization can be incorporated by using congener-specific vg values, which account for the effect of revolatilization via a lowering of the deposition velocity. Third, PCBs move from the watershed surfaces into the water column and are flushed out of the system. Experimentally, it is difficult to distinguish between steps 1 and 2. The vg values of ref 37 demonstrate that lower molecular weight congeners come to equilibrium with respect to gaseous absorption more rapidly than high molecular weight congeners, and this results in a lowering of their effective vg values. Shoeib and Harner (38) describe the same effect for uptake of PCBs by passive air samplers. Thus, although high molecular weight PCBs do have a higher affinity for organic carbon, which could make them less mobile within the watershed, this effect appears to be outweighed by their decreased ability to revolatilize, which can be described by a lowering of their gaseous deposition velocities. Sorption to organic carbon is an important process for retarding the movement of PCBs in groundwater. However, most of the PCB burden exiting the watershed resides on particles, which have presumably been mechanically scoured from the system. The fate of PCBs in the watershed, therefore, appears to be dominated by their volatility. Because revolatilization of the low molecular weight PCBs appears to be extensive, these watersheds, which are relatively remote from the urban zone, appear to be at or close to equilibrium with respect to gaseous exchange of PCBs. The fact that use of congenerspecific vg values removed the dependence of pass through efficiency (E) on the chemical’s vapor pressure suggests that once at equilibrium with respect to air/surface exchange, all PCB congeners are processed similarly within the watershed, with about 3% of the total atmospherically deposited mass exiting the watershed via the tributary.
Acknowledgments DRBC provided the funding for this work. Support from the New Jersey Agricultural Experiment Station is also gratefully acknowledged. Thanks go to Dawn Kaczorowski for Figure 1. We thank the members of AAR’s Ph.D. committee, John Reinfelder, Chris Uchrin, and Kevin Farley, for their useful comments. We also thank two anonymous reviewers whose comments greatly improved the manuscript.
Supporting Information Available Literature values for dry particle deposition velocity (Table S-1) and dry gas deposition velocity (Table S-2) and Fws and atmospheric geometric mean concentrations for PCB congeners in the subwatersheds of the Delaware River (Table S-3). This material is available free of charge via the Internet at http://pubs.acs.org.
Literature Cited (1) Eisenreich, S. J.; Looney, B. B.; Thornton, J. D. Airborne organic contaminants in the Great Lakes ecosystem. Environ. Sci. Technol. 1981, 15, 30-38. (2) Hillery, B. R.; Basu, I.; Sweet, C. W.; Hites, R. A. Temporal and spatial trends in a long-term study of gas-phase PCB concentrations near the Great Lakes. Environ. Sci. Technol. 1997, 31, 1811-1816. (3) Offenberg, J. H.; Baker, J. E. Influence of Baltimore’s urban atmosphere on organic contaminants over the Northern Chesapeake Bay. J. Air Waste Manage. Assoc. 1999, 49, 959965. (4) Offenberg, J. H.; Baker, J. E. PCBs and PAHs in Southern Lake Michigan in 1994 and 1995: urban atmospheric influences and long-term declines. J. Great Lakes Res. 2000, 26, 196-208. (5) Brunciak, P. C.; Dachs, J.; Gigliotti, C. L.; Nelson, E. D.; Eisenreich, S. J. Atmospheric polychlorinated biphenyl concentrations and apparent degradation in coastal New Jersey. Atmos. Environ. 2001, 35, 3325-3339. (6) Tsai, P.; Hoenicke, R.; Yee, D.; Bamford, H. A.; Baker, J. E. Atmospheric concentrations and fluxes of organic compounds in the northern San Francisco Estuary. Environ. Sci. Technol. 2002, 36, 4741-4747. (7) Totten, L. A.; Gigliotti, C. L.; VanRy, D. A.; Offenberg, J. H.; Nelson, E. D.; Dachs, J.; Reinfelder, J. R.; Eisenreich, S. J. Atmospheric concentrations and deposition of polychlorinated biphenyls to the Hudson River Estuary. Environ. Sci. Technol. 2004, 38, 25682573. (8) Wethington, D. M.; Hornbuckle, K. C. Milwaukee, W. I. as a source of atmospheric PCBs to Lake Michigan. Environ. Sci. Technol. 2005, 39, 57-63. (9) Totten, L. A.; Panangadan, M. V.; Eisenreich, S. J.; Cavallo, G. J.; Fikslin, T. J. Direct and indirect atmospheric deposition of PCBs to the Delaware River Watershed. Environ. Sci. Technol. 2006, 40, 2171-2176. (10) Fikslin, T. F.; Suk, N. Total Maximum Daily Loads for Polychlorinated Biphenyls (PCBs) for Zones 2-5. of the Tidal Delaware River; Delaware River Basin Commission: 2003. (11) Yagecic, J.; Cavallo, G.; MacGillivray, R.; Liao, D.; Suk, N.; Fikslin, T. J. Calibration of the PCB Water Quality Model for the Delaware Estuary for Penta-PCBs and Carbon; Delaware River Basin Commission: 2003. (12) McClure, V. E. Transport of heavy chlorinated hydrocarbons in the atmosphere. Environ. Sci. Technol. 1976, 10, 1223-1229. (13) Slinn, W. G. N.; Hasse, L.; Hicks, B. B.; Hogan, A. W.; Lal, D.; Liss, P. S.; Munnich, K. O.; Sehmel, G. A.; Vittori, O. Some aspects of the transfer of atmospheric trace constituents past the airsea interface. Atmos. Environ. 1978, 12, 2055-2087. (14) Doskey, P. V.; Andren, A. W. Modeling the flux of atmospheric polychlorinated biphenyls across the air/water interface. Environ. Sci. Technol. 1981, 15, 705-711. (15) Atlas, E.; Foster, R.; Glam, C. S. Air-sea exchange of high molecular weight organic pollutants: laboratory studies. Environ. Sci. Technol. 1982, 16, 283-286. (16) Eisenreich, S. J. The chemical limnology of non-polar organic contaminants: PCBs in Lake Superior. In Sources and Fates of Aquatic Pollutants; Hites, R. A., Eisenreich, S. J., Eds.; American Chemical Society: Washington, DC, 1987; Vol. 216, pp 393469. (17) Swackhamer, D. L.; McVeety, B. D.; Hites, R. A. Deposition and evaporation of polychlorobiphenyl congeners to and from Siskiwit Lake, Isle Royale, Lake Superior. Environ. Sci. Technol. 1988, 22, 664-672. (18) Holsen, T. M.; Noll, K. E.; Liu, S. P.; Lee, W. J. Dry deposition of polychlorinated biphenyls in urban areas. Environ. Sci. Technol. 1991, 25, 1075-1081. (19) Lee, W. J.; Lin, Lewis, S. J.; Chen, Y. Y.; Wang, Y. F.; Sheu, H. L.; Su, C. C.; Fan, Y. C. Polychlorinated biphenyls in the ambient air of petroleum refinery, urban, and rural areas. Atmos. Environ. 1996, 30, 2371-2378.
(20) Lee, W. J.; Su, C. C.; Sheu, H. W.; Fan, Y. C.; Chao, H. R.; Fang, G. C. Monitoring and modeling of PCB dry deposition in urban area. J. Hazard. Mater. 1996, 49, 57-88. (21) Granier, L. K.; Chevreuil, M. Behaviour and spatial and temporal variations of polychlorinated biphenyls and lindane in the urban atmosphere of the Paris Area, France. Atmos. Environ. 1997, 31, 3787-3802. (22) Franz, T. P.; Eisenreich, S. J.; Holsen, T. M. Dry deposition of particulate polychlorinated biphenyls and polycyclic aromatic hydrocarbons to Lake Michigan. Environ. Sci. Technol. 1998, 32, 3681-3688. (23) Wania, F.; McLachlan, M. S. Estimating the influence of forests on the overall fate of semivolatile organic compounds using a multimedia fate model. Environ. Sci. Technol. 2001, 35, 582590. (24) Tasdemir, Y.; Odabasi, M.; Vardar, N.; Sofuoglu, A.; Murphy, T. J.; Holsen, T. M. Dry deposition fluxes and velocities of polychlorinated biphenyls (PCBs) associated with particles. Atmos. Environ. 2004, 38, 2447-2456. (25) Tasdemir, Y.; Holsen, T. M. Measurement of particle phase dry deposition fluxes of polychlorinated biphenyls (PCBs) with a water surface sampler. Atmos. Environ. 2005, 39, 1845-1854. (26) Leister, D. L.; Baker, J. E. Atmospheric deposition of organic contaminants to the Chesapeake Bay. Atmos. Environ. 1994, 28, 1499-1520. (27) Sehmel, G. A. Particle and gas dry deposition: a review. Atmos. Environ. 1980, 14, 983-1118. (28) Schroder, J.; Welsch-Pausch, K.; McLachlan, M. S. Measurement of atmospheric deposition of polychlorinated dibenzo-p-dioxins (PCDDs) and dibenzofurans (PCDFs) to a soil. Atmos. Environ. 1997, 31, 2983-2989. (29) Horstmann, M.; McLachlan, M. S. Atmospheric deposition of semivolatile organic compounds to two forest canopies. Atmos. Environ. 1998, 32, 1799-1809. (30) Bohme, F.; Welsch-Pausch, K.; McLachlan, M. S. Uptake of airborne semivolatile organic compounds in agricultural plants: field measurements of interspecies variability. Environ. Sci. Technol. 1999, 33, 1805-1813. (31) Ould-Dada, Z. Dry deposition profile of small particles within a model spruce canopy. Sci. Tot. Environ. 2002, 286, 83-96. (32) EPA. EPA National Pollutant Discharge Elimination System (NPDES) - Office of Wastewater Management, 2006. http:// cfpub.epa.gov/npdes/index.cfm (accessed March 17, 2006). (33) Liebert, D. P. PCBs in Chesapeake Bay fish, M.S. Thesis, University of Maryland, 2006. (34) Bamford, H. A.; Poster, D. L.; Baker, J. E. Henry’s Law constants of polychlorinated biphenyl congeners and their variation with temperature. J. Chem. Eng. Data 2000, 45, 1069-1074. (35) Bamford, H. A.; Poster, D. L.; Huie, R. E.; Baker, J. E. Using extrathermodynamic relationships to model the temperature dependence of Henry’s Law constants of 209 PCB congeners. Environ. Sci. Technol. 2002, 36, 4395-4402. (36) Karickhoff, S. W.; Brown, D. S.; Scott, T. A. Sorption of hydrophobic pollutants on natural sediments. Water Res. 1979, 13, 241-248. (37) McLachlan, M. S.; Welsch-Pausch, K.; Tolls, J. Field validation of a model of gaseous SOC in Lolium multiflorum (rye grass). Environ. Sci. Technol. 1995, 29, 1998-2004. (38) Shoeib, M.; Harner, T. Characterization and comparison of three passive air samplers for persistent organic pollutants. Environ. Sci. Technol. 2002, 36, 4142-4151.
Received for review September 7, 2006. Revised manuscript received January 11, 2007. Accepted January 23, 2007. ES062136O
VOL. 41, NO. 7, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2337
