Environ. Sci. Technol. 2003, 37, 4038-4042
Variable Air Temperature Response of Gas-Phase Atmospheric Polychlorinated Biphenyls near a Former Manufacturing Facility M A R K H . H E R M A N S O N , * ,† CHERYL A. SCHOLTEN,† AND KEVIN COMPHER‡ Department of Earth & Environmental Science and Department of Chemistry, University of Pennsylvania, Philadelphia, Pennsylvania 19104
Many investigations of gas-phase atmospheric PCB show a strong relationship between concentration and air temperature, especially near PCB sources. Comparative gasphase atmospheric PCB trends during an annual temperature regime at two sites near a former PCB manufacturing plant and nearby PCB landfills in Anniston, AL, indicate a departure from this trend. The Mars Hill sampling site, located closest to the plant and landfills, shows an annual average ∑PCB concentration of 27 ng m-3 (ranging from 8.7 to 82 ng m-3) three times the average at Carter, 1.5 km away (9 ng m-3, ranging from 1.1 to 39). However, total PCB and congener concentrations vary more with air temperature at Carter where PCB are evaporating from surfaces during warmer weather. The slopes of the ClausiusClapeyron plots of 18 of the most concentrated congeners representing dichloro- through heptachlorobiphenyl homologues are significantly higher at the Carter site. While some of the atmospheric PCB at Mars Hill is derived from ground surface evaporation, the source of much of it apparently is the material buried in the landfills, which has different thermal properties than surface materials and is not in equilibrium with air temperature.
Introduction Gas-phase atmospheric PCB concentrations are often higher in summer than winter, resulting from influences of increasing air temperatures on surfaces and the fugacities or partial pressures of PCB congeners condensed there while atmospheric pressure remains relatively constant (1-6). The cyclical nature of this process has led to several efforts to identify the energy required (∆H) for a phase change of PCB and other organochlorine compounds from the surfacecondensed to the gas phase using the Clausius-Clapeyron (CC) equation (4, 7-10):
ln P (atm) ) -∆H/R (T-1) + c
(1)
where ∆H ) kJ mol-1, R ) gas constant, T ) temperature (K), and c ) constant. Air temperature data are more commonly used with the CC equation as a surrogate for surface temperatures, which are less frequently observed. The ∆H/R * Corresponding author e-mail:
[email protected]; telephone: (215)573-8727; fax: (215)898-0964. † Department of Earth & Environmental Science. ‡ Department of Chemistry. 4038
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 18, 2003
term is the slope of the equation, which can be determined empirically. Many investigations using this common form of the CC equation have determined a slope (∆H/R) for various compounds. However, in regions where there is no local source (no contaminant is condensed on the ground or other surfaces) or where contaminant concentrations in air are controlled by long-range transport, the air temperature concentration dependence may be weak or not exist, in which case a slope cannot be measured (9-11). Our objective is to use the CC process to understand the effects of air temperature on atmospheric PCB in an area where PCBs were manufactured for over 40 yr and where large amounts of PCB-contaminated waste remained more than 25 yr after the end of production. The use of the CC equation will help identify the thermal processes that play a role in emissions from a large, unregulated point source. Previous research in PCB-contaminated areas suggests a hypothesis that atmospheric PCB concentrations at the study siteswith multiple PCB sourcesswill show a strong seasonal variation in gas-phase atmospheric PCB (4, 6). However, since PCB production ended in North America during the 1970s and since remediation steps have been taken at many PCBcontaminated sites, there are few chances to observe atmospheric PCB trends where a source may exist in its original state. This study investigates thermal and spatial atmospheric PCB trends near a landfill that has received little remedial attention. It is the first investigation of atmospheric PCB at a production site in North America. Study Site. We collected air samples from two sites in the western part of Anniston, AL, where an estimated 400 000 t of PCB was manufactured from the 1920s until 1971. PCB manufacturing wastes were discarded in what is known as the west landfill, a dump located about 300 m to the WSW from the plant, from the 1930s until 1960 when the land was sold for other industrial uses (Figure 1). It contains an unknown amount of PCB, but soil samples collected in 1994 have shown ∑PCB concentrations as high as 800 µg g-1 dry mass (dm); a composite tar sample from the landfill surface at the time contained 33 000 µg g-1 (12). The landfill has no artificial bottom liner (13) and had only soil and vegetation on the surface until 1995 when a different capping system was installed (12). Also beginning in the 1920s and until 1975 (4 yr after PCB manufacturing ended), PCB wastes were discarded in a portion of what is known as the south landfill, an old quarry on the northern flank of Coldwater Mountain, about 500 m to the SE from the plant. In 1979, it contained an estimated minimum of 4550 t of PCB, approximately 2200 t dumped after PCB production ended in Anniston in 1971. Part of the material dumped after 1971 apparently included demolition residue of the PCB manufacturing facility. No PCBs are known to have been moved from the site. Like the west landfill, it has no artificial liner at the bottom, and the landfill segment containing PCB continues to be covered with unspecified vegetation growing on a soil cover of unknown depth (12). No other efforts are known to have been made to restrict PCB gas emissions from either landfill. PCB contamination in Anniston is not restricted to the landfills and plant site: 18 residential properties within an 800 m radius of the plant are known to have an average ∑PCB soil concentrations of 1.4 µg g-1 dm, which exceeds the U.S. EPA action level (1.0 µg g-1 dm) (13). Residents of Anniston have ∑PCB blood concentrations up to 2115 µg (g of serum)-1 (14): a value of 10 µg (g of serum)-1 is considered to be high (13, 14). No PCB-contaminated properties in Anniston are on the National Priority List (NPL). 10.1021/es030332e CCC: $25.00
2003 American Chemical Society Published on Web 08/13/2003
FIGURE 1. Map of southern Anniston, AL (outline), and surrounding areas. Samples were collected from Mars Hill and Carter (b). Landfills (1) and the PCB plant site (9) are also shown.
Methods Sampling Methods. In May 1997, we set up high-volume air samplers at two sites in west Anniston (Figure 1). The Mars Hill site was located about 500 m ENE from the plant site and about 400 m north from the south landfill. In April 1998, access to Mars Hill was lost, so the sampler was moved 300 m NE to a residential property where the final four samples (out of 23) were collected. We consider both sites to represent the same conditions considering the proximity of both to the south landfill and the dominance of SSW winds in Anniston. The Carter site, located at a residential property on Carter Street, was about 1.5 km NW from Mars Hill. Carter was chosen because it is away from the immediate influence of landfill and plant site emissions and did not have high PCB soil concentrations characteristic of many residential properties located to the NE from the plant and landfill sites. Samples were collected simultaneously from these sitess with the exception of two datesson an average 17-d cycle until June 1998. All sites were located at similar elevations (∼245 m asl). Both samplers were calibrated when repaired or moved. Approximately 200 m3 of air was collected during each ∼24-h sampling event. A glass fiber filter (GFF) was used to capture particles >0.1 µm and associated PCB. Gas-phase PCBs were collected behind the GFF on a cartridge consisting of a 1.5-cm layer of XAD-2 between layers of polyurethane foam (PUF). This gas concentrating system and the procedures used to clean the sampling materials before collection are identical to those used in other studies (3, 15). The site operator installed and removed sampling materials from the samplers handling them only with the original foil packaging material, which was removed from the site during sampling. All samples were stored and shipped frozen. Sampler clock reading, flow rate, start and stop times, and basic weather conditions were noted at the beginning and end of each sampling event. Blanks were collected at 10 sample intervals by installing materials on the samplers, removing them immediately, and then taking them to frozen storage. Air temperatures for each sampling event were averages of hourly observations at the Anniston airport. In addition, temperatures at the sites were noted at the beginning and
end of each sampling period, which were consistent between sites and with airport data. Analytical Methods. Gas-phase samples were spiked with surrogate standard (100 ng of PCB 14 and 50 ng each of PCB 65 and PCB 166) and extracted overnight in 1/1 (v/v) acetone/ hexane. After volume reduction, samples were cleaned by elution with hexane on a column of silica gel deactivated 3.5% by mass with H2O. The samples were again reduced in volume and spiked with internal standards (80 ng of PCB 30 and 60 ng of PCB 204) and were analyzed by GC-ECD. Up to 120 PCB congeners were qualified and quantified using internal standards against a mixed-Aroclor standard, the same procedure used in other investigations (3, 4). Total PCB is the sum of analyzed congeners. Quality Control. Average surrogate recoveries were 119.8% and 121.3% for PCB 14 at Mars Hill and Carter, 89.8% and 115% for PCB 65, and 112% and 121% for PCB 166. All are within our limit of 50-125%. Our LOD was 0.573 ng m-3, calculated as the sum of mean blank and 3 SD and assuming a 200 m3 average sample size. To test for saturation and possible PCB breakthrough of sampling materials, a test sample was collected at Mars Hill. The XAD-2 and two PUF layers were analyzed separately. All of the 37 ng m-3 ∑PCB in this sample was in the upper PUF and XAD-2 layers, showing no breakthrough.
Results and Discussion Total PCB concentrations at Mars Hill ranged from 8.7 to 82 ng m-3, averaging 27 ng m-3, while Carter ranged from 1.1 to 39 ng m-3, averaging 9 ng m-3 (Figure 2). The higher average and the much higher minimum concentrations at Mars Hill show that it was nearer to a larger, active atmospheric PCB source than Carter, only 1.5 km away. Both Mars Hill and Carter maximum and average PCB concentrations are higher than any site in Bloomington, IN, in the late 1980s (also shown in Figure 2), an area with three PCB sites on the NPL. The maximum observed at the Bloomington Courthouse site in 1988 is about 0.25 that of Mars Hill and 0.5 that of Carter (6). However, the comparison between Anniston and Bloomington needs to consider that some remedial work had been done at the Bloomington landfills before air sampling and VOL. 37, NO. 18, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4039
FIGURE 2. Comparative maximum, average, and minimum gasphase ∑PCB from sites in Anniston (Mars Hill and Carter), AL; Bloomington, IN (Courthouse, Batchelor, and Sanders) (6); and Michigan (Traverse, Saginaw, and Soo) (3). The U.S. EPA Region 3 risk-based concentration guideline value for ambient air is also shown, which is about the same as the average gas-phase concentration observed in Chicago in 1994 (16). that the samplers were located at least 2250 m from the nearest landfill, a greater distance than Carter from the nearest Anniston landfill. Perhaps the most intensive studies of atmospheric PCB have taken place in the Great Lakes region. Samples from Michigan shoreline sites without known PCB sources, sampled in the early 1990s, show a gas-phase maximum of 1.8 ng m-3 at Traverse, about 45 times less than Mars Hill and 21 times less than Carter maxima. More recent results from Chicago, considered to have the highest average urban values along the Great Lakes, were about 1.3-1.9 ng m-3 in 1997-1998 when we sampled in Anniston, while the maximum annual average observed there since 1993 was about 3 ng m-3 in 1994 (16). The Chicago maximum is similar to the U.S. EPA Region 3 risk-based concentration guideline for ambient air for 2002, 3.1 ng m-3 (see www.epa.gov/ reg3hwmd/risk/index.htm), which is about 9 times less than the Mars Hill average and 3 times less than Carter. Obviously a large area of Anniston area is influenced by very strong PCB sources that are unknown in the Great Lakes region and perhaps elsewhere. In addition to the spatial PCB concentrations being different at Mars Hill and Carter, their thermal behaviors are also significantly different. A regression of ∑PCB and air temperature, which is the same at both sites, shows that Carter concentrations have greater response to air temperature changesas measured by regression slopesthan those at Mars Hill (Figure 3). The correlations in Figure 3 are both high and significant (r ) 0.644 and F ) 0.0047 at Mars Hill; r ) 0.647 and F ) 0.0066 at Carter). This trend is not consistent with what would be expected where fugacity of PCB in a major source is influenced by air temperature: Mars Hill, being near major atmospheric PCB sources, would likely have an air temperature-related trend similar to or stronger than Carter. Total PCB can be difficult to compare, however, because of the variable contribution of individual congeners over time and between sites. To resolve this issue, we consider the association between partial pressure (P) of major PCB congeners in the air samples and inverse air temperature, one form of the Clausius-Clapeyron relationship (see eq 1) that has been widely used to understand air temperature effects on gas-phase contaminant concentrations (2, 4, 5, 4040
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 18, 2003
FIGURE 3. Regressions of ∑PCB related to air temperature at Mars Hill and Carter (with 95% standard error intervals). Slopes (m), correlations (r), and significance levels (G) are shown. Both trends are correlated and are significant.
FIGURE 4. Trends of ln P of PCB 8+5 vs inverse air temperature at Mars Hill and Carter. Regression slope, correlation, significance, and standard error are shown in Table 1. 7-9, 11). PCB congener concentrations can be converted to P using a modified version of the ideal gas law (8), where
P ) CRT (10-15)/MM
(2)
where P ) partial pressure (atm), C ) congener concentration (pg m-3), and MM ) molecular mass. The factor 10-15 accounts for unit conversions. We regressed 1000 T-1 on ln P using eq 1. We selected 18 of the congeners most concentrated in Anniston air samples: CC plots for six of these congeners, representing the most concentrated from a respective homologue (dichloro- through heptachlorobiphenyl), are shown in Figures 4-9. The slopes for all 18 congeners (Table 1) clearly show that ln P at Mars Hill varies less with air temperature changes than at Carter for all congeners except PCB 45 and perhaps PCB 180 (figure 9), which does not have a significant trend at Carter. Wania et al. (9) found a significant relationship between CC slope and number of Cl atoms on the PCB molecule in several data sets. This is analogous to a relationship between ∆H/R and vapor pressure: vp and number of Cl atoms for our 18 congeners is highly correlated (r ) -0.9678). We plotted ∆H/R values for all 18 congeners against log vp, calculated from Falconer and Bidleman (17), for both sites (Figure 10). If the fugacity of a compound is a function of vp, then the less volatile congeners should have lower ∆H/R. The results
FIGURE 5. Trends of ln P of PCB 31+28 vs inverse air temperature at Mars Hill and Carter. Regression slope, correlation, significance, and standard error are shown in Table 1.
FIGURE 6. Trends of ln P of PCB 52 vs inverse air temperature at Mars Hill and Carter. Regression slope, correlation, significance, and standard error are shown in Table 1.
FIGURE 7. Trends of ln P of PCB 101 vs inverse air temperature at Mars Hill and Carter. Regression slope, correlation, significance, and standard error are shown in Table 1. in Figure 10 show clearly that PCB congeners measured at Mars Hill do not follow this trend. Carter results are more scattered, but SE are large enough that no trend is established. However, there is a significant difference between Carter and Mars Hill: PCB congeners measured at Carter have higher ∆H/R values than Mars Hill, with the exception of PCB 45, showing that PCBs at Carter are more prone to phase change related to air temperature than Mars Hill.
FIGURE 8. Trends of ln P of PCB 153+132 vs inverse air temperature at Mars Hill and Carter. Regression slope, correlation, significance, and standard error are shown in Table 1.
FIGURE 9. Trends of ln P of PCB 180 vs inverse air temperature at Mars Hill and Carter. Regression slope, correlation, significance, and standard error are shown in Table 1. What is the difference between these sites that causes variable thermodynamic behavior of PCB? We assume that equilibrium conditions between the atmosphere and the ground surface are the same at both sites because air temperature is the same. The PCB concentrations and their predictable variability with air temperature at Carter are characteristic of a site where PCB contamination is likely found on the ground or other surfaces, where air temperature can influence surface temperatures leading to a phase change (9, 11). Since there is also an air temperature-related trend at Mars Hill, it is likely that surfaces in the area contain PCB and react to air temperature changes similar to Carter. But there is also a large subsurface PCB component that must be considered. Subsurface soil temperatures are known to be much less influenced by air temperature variability than those at the surface (18). While we do not know burial depth of PCB in these landfills, the phase change of PCB is more likely influenced by the thermal properties of materials in the landfills, including PCB, than by air temperature. The thermal properties of PCB may contribute an additional explanation: The specific heats of the 18 major atmospheric PCB congeners observed in Anniston range from 292 (PCB 5) to 345 (PCB 180) J mol-1 K (calculated from ref 19), about 2.1-2.5 times the specific heat of H2O (138 J mol-1 K). The thermal response of landfill materials saturated with PCB would be slower than with H2O saturation given the same temperature changes. A further consideration is that average air temperature during our sampling was never below VOL. 37, NO. 18, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4041
TABLE 1. Slopes (∆H /R), Correlations (r), Significance Levels (G), and Standard Errors (SE) of CC equations for 18 PCB Congeners and Pairsa PCB
∆H/R
r
G
SE
8+5 18 26 25 31+28 45 52 47+48 70+76 91 92+84 101 82 151 149 153+132 163+138 180
4.13 3.88 5.03 3.51 4.31 7.7 4.08 4.51 4.7 4.64 4.25 4.49 4.59 4.14 4.52 4.1 4.85 8.99
Mars Hill 0.5262 0.4967 0.5875 0.4819 0.5463 0.5674 0.6656 0.6835 0.7475 0.7292 0.7304 0.7961 0.7846 0.7199 0.8389 0.5132 0.8386 0.7752
0.0099 0.0159 0.0032 0.0428 0.0070 0.0059 0.0007 0.0003
