Environ. Sci. Technol. 2006, 40, 7629-7635
Factors Influencing the National Distribution of Polycyclic Aromatic Hydrocarbons and Polychlorinated Biphenyls in British Soils ELIZABETH HEYWOOD,† JULIAN WRIGHT,† CLAIRE L. WIENBURG,† H E L A I N A I . J . B L A C K , ‡,§ S A R A M . L O N G , † DAN OSBORN,‡ AND D A V I D J . S P U R G E O N * ,† Centre for Ecology and Hydrology, Monks Wood, Abbots Ripton, Huntingdon, Cambridgeshire, PE28 2LS, UK, and Centre for Ecology and Hydrology, Lancaster, Bailrigg, Lancaster, LA1 4AP, UK
The polycyclic aromatic hydrocarbons (PAHs) naphthalene, acenaphthylene, acenaphthene, fluorene, fluoranthene, pyrene, benzo[a]anthracene, chrysene, benzo[b]fluoranthene, benzo[k]fluoranthene, benzo[a]pyrene, ideno[1,2,3,-cd]pyrene, dibenz[a,h]anthracene, benzo[g,h,i]perylene and the polychlorinated biphenyls (PCBs) 8, 18, 28, 29, 31, 52, 77, 101, 105, 114, 118, 123, 126, 128, 138, 141, 149, 153, 156, 157, 163, 169, 170, 171, 180, 183, 187, 189, 194, 199, 201, 206, and 209 were measured in ∼200 rural soils across Great Britain (GB). Dominance of soil PAH profiles by heavier compounds (4-6 rings) provided initial evidence for the importance of source in governing soil PAH concentrations. No relationship was found between soil organic matter (SOM) and sum concentration of total and “heavy” PAHs, although there was a weak positive relationship with lighter compounds. A spatial statistical technique showed that highest soil PAH concentrations were usually found close to industrial/urban centers where presumably source intensity is highest. PCBs clustered into seven groups, five of which contained a single “dioxin like” PCB, one contained lighter congeners (2-4 chlorines), and one contained heavy congeners (5-10 chlorines). Linear regressions with SOM explained up to 24.3% of variation for the sum concentration of penta- to deca- congeners, but LOD
50.51 5.91 15.75 15.5 113 95.3 87.5 128 122 54.8 88.3 84.3 16.9 89 976
98 14.2 30 27.7 194 168 180 197 204 165 163 125 28.8 152 1530
15.6 1.49 6.21 5.36 40.2 40.7 29.4 58.2 47.1 16 33.7 34.9 6.55 41.5 487
1.44 0.433 0.256 0.242 4.39 3.53 1.11 0.766 2.77 1.07 1.11 2.72 0.298 3.03 40.4
543 100 232 155 1430 1530 1670 1560 1500 1690 1440 710 205 1450 14100
5.18 0.606 1.61 1.59 11.6 9.76 8.97 13.1 12.5 5.61 9.05 8.64 1.73 9.12
81 63 95 98 87 94 93 96 100 99 97 99 80 99
138, 141, 149, 153, 156, 157, 163, 169, 170, 171, 180, 183, 187, 189, 194, 199, 201, 206, and 209 by gas chromatographymass spectrometry (GC-MS). Briefly, soil was extracted with dichloromethane (DCM) and the extracts were cleaned up using activated alumina followed by automated highresolution size exclusion chromatography (SEC) using a highpressure liquid chromatography (HPLC) system, and the solvent was changed to n-hexane. The cleaned up extract was then separated by HPLC into three fractions containing aliphatic/monocyclic aromatic compounds, PAHs, and PCBs, and run on a gas chromatograph fitted with a mass selective detector. The detector was run in electron impact SIM mode at the highest mass resolution available. PAHs were separated on a 30 m × 0.25 mm i.d. HP5-MS column fitted with a 5 m × 0.25 mm i.d. Siltek deactivated guard column. Injection was carried out using the solvent vent technique. PCBs were separated on a 50 m × 0.22 mm i.d. HT8 column (SGE). PAH and PCB standards were used as internal standards and to calculate losses during analysis (isotope dilution). Each batch of samples included a sample blank, a standard reference material (SRM) SETOC 738, a forest soil control, and a spiked forest soil. Limits of detection (LOD) were set as the mean of 7 sample blanks plus 3 standard deviations. (See Supporting Information for details on sample preparation and extraction and statistical analyses of the data).
Results and Discussion Data Summary. Limits of detection (LoDs) for PAHs ranged from 0.242 (acenaphthene) to 4.39 (fluoranthene) ng/g and from 0.003 (PCBs 157, 189, 209) to 0.055 (PCB 180) ng/g for the PCBs. These LoDs were sufficient to allow detection of residues of the majority of PAHs in most soils (Table 1) and between 0.5% of samples for PCB 169 to 98% for PCB 118 (Table 2). Average agreement between measured and certified concentrations for PAHs in SETOC 738 estuarine SRM was 101%. Additionally for naphthalene, the extraction and analysis method used can result in limited loss and for this reason absolute concentrations for this compound should be treated with caution, although comparison among samples remains valid. Average agreement between measured and certified concentrations of present PCBs (28, 31, 52, 101, 105, 118, 128, 138, 149, 153, 156, 180) was 90%, ranging from 58% for PCB 156 to 171% for PCB 31. Given the low levels present in the reference material, these recoveries indicate the GCMS method used is reliable for low level PCB quantification in soil. Concentrations of PCBs assessed in multiple locations from the single grassland site were close to or below the detection limit for PCBs in many of the collected samples. 7630
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 24, 2006
TABLE 2. Summary Data for PCB Concentrations (ng/g Dry Weight Soil) in Soil Samples Collected as Spatially Referenced Sites Across Great Britain as Part of the Countryside Survey Project PCB congener
mean
SD
median
min
max
8 18 28 29 31 52 77 101 105 114 118 123 126 128 138 141 149 153 156 157 163 169 170 171 180 183 187 189 194 199 201 206 209 Total PCB
0.3127 0.431 0.4437 0.0213 0.3663 0.1628 0.02875 0.1824 0.1156 0.01375 0.2739 0.027 0.00873 0.0929 0.3493 0.0818 0.2808 0.4546 0.03819 0.0164 0.14 0.00899 0.1311 0.0409 0.3037 0.0843 0.2272 0.00898 0.0915 0.01612 0.1319 0.0586 0.083 5.028
0.845 1.072 1.306 0.008 1.036 0.33 0.061 0.445 0.299 0.039 0.794 0.084 0.026 0.181 0.751 0.184 0.621 1.025 0.09 0.038 0.278 0.014 0.308 0.088 0.86 0.201 0.51 0.022 0.249 0.044 0.353 0.117 0.153 8.411
0.045 0.035 0.02 0.023 0.021 0.039 0.009 0.078 0.05 0.004 0.109 0.014 0.003 0.044 0.171 0.037 0.111 0.223 0.018 0.005 0.069 0.008 0.06 0.019 0.118 0.034 0.1 0.002 0.034 0.006 0.047 0.026 0.036 2.524
0.008 0.002 0.004 0.001 0.001 0.001 0.002 0.001 0.0006 0.001 0.008 0.001 0.002 0.004 0.015 0.002 0.012 0.02 0.001 0.001 0.007 0.004 0.002 0.002 0.013 0.003 0.006 0.001 0.003 0.001 0.002 0.002 0.002 0.274
8.512 9.074 12.188 0.087 9.933 2.11 0.521 5.527 3.772 0.293 10.288 0.863 0.253 2.128 8.807 1.828 6.739 11.926 1.082 0.297 3.181 0.188 2.931 0.817 8.938 1.915 4.395 0.158 2.589 0.416 2.924 1.053 1.102 80.579
% percent total >LOD 6.22 8.57 8.82 0.42 7.29 3.24 0.57 3.63 2.30 0.27 5.45 0.54 0.17 1.85 6.95 1.63 5.58 9.04 0.76 0.33 2.78 0.18 2.61 0.81 6.04 1.68 4.52 0.18 1.82 0.32 2.62 1.17 1.65 -
60 54 46 18 51 67 60 88 90 26 98 12 15 94 92 92 84 95 86 60 95 0.5 87 82 75 81 89 42 68 64 79 92 93 -
PAH determinations were within the detection range. Capability to compare and contrast the within-site heterogeneity of PCB and PAHs (e.g., use of box and whisker plots or calculation of coefficient of variation) was hampered by the frequency of nondetected in the PCB analysis. The cursory comparison of the range of measured concentrations does, however, suggest that local scale heterogeneity of PCBs is more likely to hinder the identification of national and regional trends than for PAHs, since maximum difference between PCB samples was approximately 1 order of magnitude compared to only 3-fold for the PAHs.
Soils from each of the 109 GB wide sites were analyzed (Tables 1 and 2). Three sites had no associated longitude and latitude and were discounted. Where analytes were present below the LOD, missing values were replaced by a value of half of the LOD. As the data were positively skewed with outliers on the high end and unequal variation, concentrations were logarithmically transformed prior to analysis. Compared with the soils from remote areas of the American Great Plains (18), the analyzed GB soils contain mean total PAH loads that are higher by approximately a factor of 8. Dominant PAHs detected in most GB soils were of higher molecular weight (5-6 ring) (Table 1). This was in contrast to Wilcke and Amelung (18) where low molecular weight PAHs (2-4 ring), that were considered characteristic of long-range transport, were present at highest concentrations. The overall elevated levels of PAHs and in particular of the heavier compounds suggest that, even though they were collected from ostensibly rural locations, the collected soil are contaminated by long term aerial deposition of both light and heavy PAHs derived probably from a range of industrial and domestic combustion sources. The low ppb levels for total PCB and sub ppb levels for single isomers detected (Table 2) are in agreement with the concentrations found previously by Creaser and Fernandes (19) and Creaser et al. (20). Based on medians, the dominant congeners were 153, 138, 180, 149, 118, and 187. Creaser et al. (20) and Meijer et al. (21) also found higher concentrations of PCB 153 and 138 than other measured congeners, while congeners 138, 153, and 180 were the most dominant congeners in UK birds of prey (9). Compound Correlation and Clustering. Construction of a Pearson correlation matrix for the PAHs indicated a significant correlation (P < 0.05) between all compounds except acenaphythlene and fluoranthene (P ) 0.079) (see Supporting Information, Table SI-1). Notably these two PAHs had the highest and fourth highest number of concentrations below LOD. Clustering based on log concentrations identified three PAH clusters (Figure 1a). The most correlated cluster comprises all five- and six-ring PAHs and the two heavy fourring PAHs (the heavy group). The second group comprises only the two lighter four-ring PAHs fluoranthene and pyrene (the medium group). The remaining group comprises the less strongly correlated light two- and three-ring PAHs (benzo[a]anthracene and chrysene) (the light group). Since distinct clusters were found in the data it was decided to reduce the number of separate PAHs for further analysis by combining the compounds into four groups, the sum concentration of (1) the 14 measured PAHs, (2) the light cluster, (3) the medium weight group, and (4) the heavy PAH group; the individual PAHs in each group are shown in Figure 1a. Analyses concentrated particularly on groups 1, 2, and 4 as these represent the extremes for molecular weight and hydrophobicity. Many PCB concentrations were correlated (see Supporting Information, Table SI-2), with the strongest relationship being between PCB 183 and 187 (r ) 0.957, p < 0.01). Clustering gave seven groupings (Figure 1b). Five of these contained a single PCB congener, four of which (PCB 114, 123, 126, 169) exhibit a dioxin-like toxicity, and PCB29. All these congeners were not major components of the commercial PCB mixtures (sold under the trade name Aroclor in the UK), so the potential for release into the environment may have been limited. This is confirmed by a high number of measurements below the LOD for these compounds (74-99.5%), which are probably the cause of the separate clusters for each of these compounds. Of the remaining two clusters, one included all the lighter congeners with 2, 3, and 4 chlorine substitutions (except PCB77) and the second contained the heavier PCBs (5-10 chlorines) and PCB77. PCB77 may cluster into this group because it has transport properties similar to the
FIGURE 1. Dendrogram with average linkage and correlation coefficient distance of the natural log concentrations of individual PAH (a) and PCB (b) compounds in soils (ng/g dry weight soil) collected from 1 km squares distributed across GB. heavier PCBs and so co-occurs with these congeners. All the PCBs in the light cluster were major components of Aroclor 1016 and 1242 (13), while in the cluster of heavier PCBs the most correlated congeners (PCB118, 128, 187, 183, 171, 170, and 206) and the sub-group of PCBs 101, 153, 138, 180, and 209 were found in Aroclor 1254 and/or 1260. These clustering patterns suggest that the relative concentrations of PCBs in GB soils can be linked to the composition of commercial PCB mixes, which act as the major environmental source, with transport properties as a further influence. Longitude and Latitude. Scatter plots of total PAHs fitted with a locally weighted Scatterplot smoother (LOWESS) line against longitude indicated lower concentrations in western GB than in the east (Figure 2a). Against latitude, concentrations appeared to increase from position 0 (south coast of GB) to a latitude of 200,000 (approximately a line from London to Cardiff) and then follow a decreasing trend further north (Figure 2b). These complex spatial trends broadly reflect population densities in GB, with the lowest populations occurring in upland areas to the north and west and the highest in the midlands and particularly the southeast. Separate LOWESS plots for the sum of the light and heavy PAHs showed that the trend for lower concentrations in the west than east was conserved for the heavy PAHs, but not for the lighter PAHs (data not shown). Although both heavy and light PAHs showed an increase from the south coast to the London-Cardiff line, light PAHs showed no trend northwards whereas heavy PAHs showed a linear decline. The consistency of the trend for heavy PAHs with broad population trends suggests a link between concentrations of these compounds and anthropogenic sources, while the VOL. 40, NO. 24, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
7631
FIGURE 2. Scatterplot with fitted LOWESS line of the natural log concentrations of total measured concentration of 14 PAHs in soils (ng/g dry weight soil) and longitude (a), latitude (b), and sum concentration of 33 PCBs and longitude (c), latitude (d), and SOM content (e), and natural log PAHs and mean January temperature (f). weaker trends for lighter PAHs may indicate a greater influence of regional/national scale processes on soil concentrations. This issue, however, clearly requires further investigation to confidently attribute cause to the observed differences between compound groups. LOWESS plots showed no visual trend between the sum concentration of PCBs and longitude (Figure 2c) or latitude (Figure 2d). In a study of PCBs in 191 ‘background’ soil samples collected worldwide, Meijer et al. (21) found weak evidence for fractionation (i.e., increasing concentration at high latitudes) for a number of PCB congeners north of 60° North, but no evidence of fractionation at lower latitudes (30-60° N). This absence of a northerly trend at middle latitudes was attributed to the influence of local sources on soil concentrations and the retarding effect of soils on redistribution (22). The absence of evidence for fractionation 7632
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 24, 2006
found in the current study supports these observations and assumptions. Soil Organic Matter. No correlation was found between SOM content and the sum of total or heavy PAH concentrations. A weak, but nonetheless significant (P < 0.05) positive correlation (R 2 ) 5.4%) was found with the sum of light PAHs. Wilcke and Amelung (18) found a strong positive correlation between total PAH concentration and SOM content in soil from remote sites in North America. Notably the PAH profiles at these sites were dominated by lighter compounds supporting our observation for the “light” PAH cluster. LOWESS plots indicated a trend for higher total PCB concentrations at high SOM (Figure 2e). Initial scatterplots of SOM against the log sum of total PCBs showed that the data were more linear and less skewed using the logarithm of SOM content. Regressions calculated against log-SOM for
each chlorination group were significant for the penta- (p < 0.001, R 2 ) 24.3%), hexa (p < 0.001, R 2 ) 18.2%), hepta- (p < 0.001, R 2 ) 11.5%), octa- (p ) 0.001, R 2 ) 6.2%), and nona- (p < 0.001, R 2 ) 11.3%) groups, but not for the di- to tetra- congener groups or for the single deca- PCB (see Supporting Information, Table SI-3). More chlorinated PCBs evidently have a greater affinity for organic carbon than less heavily chlorinated PCBs, potentially affecting their fate in soil (23). A positive relationship between SOM and soil PCB concentrations would be consistent with the concept of PCB “hopping”, whereby repeated processes of volatilization and deposition lead to the accumulation of PCBs in soil with a high SOM (24). The fact that a similar relationship was not found for PAHs indicates that the strong binding of PAHs to SOM that has been shown previously may prevent the redistribution of this group via air (25). For PCBs, “hopping” offers a route for the local and global scale redistribution of PCBs into higher organic soils. However, conversely to results here, Meijer et al. (2003) (21) found that the relationships between PCB concentration and % SOM were stronger for the lighter homologues and became less clear with increasing chlorination. The authors interpreted these trends as evidence that, over time, the more volatile PCBs are moving toward equilibrium with the OM burden of the soil compartment, on a European regional and possibly global scale. The fact that this latter study was conducted for soils collected worldwide (including many remote areas away from potential sources), while the present study is at the national scale, may account for the differing results, since this more localized study is unlikely to pick up the signature of global fractionation seen by Meijer et al. (21). Climate. Winter and summer temperatures have the potential to influence the deposition and volatilization of organic compounds. Figure 2f shows a scatter plot of total PAH against mean winter temperature (calculated from UK meteorological office data) with a fitted LOWESS line. This indicates that lower soil total PAH concentrations are associated with warmer winter temperatures. A linear model provides a weak, but nonetheless significant fit, accounting for 7% variation for sum concentration of total PAH (p < 0.001), 8.8% for sum of light PAHs (p < 0.001), and 4.4% for sum of heavy PAHs (p < 0.01). Previously, Lee et al. (26, 27) implicated domestic solid fuel burning as the principal cause of increased airborne POPs loads and Halsall et al. (28, 29) measured higher concentrations, particularly of heavier PAHs, in urban air during winter. These previous studies, thus, implicate increased domestic fuel burning in colder regions, leading to greater PAH deposition, as the most important cause of the relationship observed here. Influences of climate on sum concentrations of the PCB congener groupings showed very weak correlations with mean precipitation and mean summer and winter temperature. Maximum variance explained was only 3.9% (for decaPCB and winter temperature). Further analysis by construction of multiple regression relationships including all variables (spatial, SOM, climate) indicated the dominance of SOM in explaining variance for the heavier congener groups (pentato deca-). For all congener groups, a number of other parameters were significantly correlated with sum concentration. The nature of the explanatory factors varied among the different congener groups and in all cases correlations were weak explaining only a relatively small portion of total variance (
