Environ. Sci. Technol. 2001, 35, 1709-1715
National-Scale, Field-Based Evaluation of the Biota-Sediment Accumulation Factor Model C H A R L E S S . W O N G , †,‡ P A U L D . C A P E L , * ,§ A N D LISA H. NOWELL| Department of Civil Engineering, University of Minnesota, Minneapolis, Minnesota 55455, U.S. Geological Survey, Minneapolis, Minnesota 55455, and U.S. Geological Survey, Sacramento, California 95819
The biota-sediment accumulation factor (BSAF) model has been suggested as a simple tool to predict bioaccumulation of hydrophobic organic compounds (HOCs) in fish and other aquatic biota from measured concentrations in sediment based on equilibrium partitioning between the sediment organic carbon and biotic lipid pools. Currently, evaluation of this model as a predictive tool has been limited to laboratory studies and small-scale field studies, using a limited number of biotic species. This study evaluates the model, from field data, for a suite of organochlorine HOCs from paired fluvial sediment and biota (fish and bivalves) samples throughout the United States and over a large range of biotic species. These data represent a real-world, worst-case scenario of the model because environmental variables are not controlled. Median BSAF values for fish (3.3) and bivalves (2.8) were not statistically different but are higher than theoretically predicted values (1-2). BSAF values varied significantly in a few species. Differences in chemical-specific BSAF values were not observed in bivalves but were statistically significant in fish. The HOCs with differing BSAF values were those known to be biotransformed. Sediment organic carbon content and biota lipid content had no effect on BSAF values in fish and only a weak effect in bivalves. This study suggests that the BSAF model could be useful under in situ riverine conditions as a first-level screening tool for predicting bioaccumulation; however, variability in BSAF values may impose limits on its utility.
Introduction Determining the bioaccumulation of sediment-associated hydrophobic organic compounds (HOCs) is crucial in the assessment of their risk to ecosystem and human health. One simple approach to estimate the bioaccumulation potential is based on equilibrium partitioning (1, 2), which assumes that HOCs partition between the carbon pools of biotic tissue lipids and sediment organic carbon. This approach also assumes that there is no chemical transfor* Corresponding author e-mail:
[email protected]; phone: (612)625-3082; fax: (612)626-7750. † University of Minnesota. ‡ Present address: Department of Chemistry, University of Toronto, Toronto, ON, M5S 3H6 Canada. § U.S. Geological Survey, Minneapolis. | U.S. Geological Survey, Sacramento. 10.1021/es0016452 Not subject to U.S. Copyright. Publ. 2001 Am. Chem. Soc. Published on Web 03/22/2001
mation, mass transfer resistance, or differential biotic uptake or depuration. Under these conditions, bioaccumulation can be assessed using the biota-sediment accumulation factor (BSAF), or its inverse, the preference factor (3, 4). The BSAF, which is also referred to by other names such as “accumulation factor” (5-8) or “biota-sediment factor” (2, 9) is defined as
BSAF ) (Cb/fl)/(Cs/foc)
(1)
where Cb is the biota HOC concentration (µg/kg wet weight), fl is the biota lipid concentration (fraction by weight), Cs is the sediment HOC concentration (µg/kg dry weight), and foc is the fraction organic carbon of the sediment (fraction by weight). Because the BSAF is based on equilibrium partitioning between the two carbon pools, its value is theoretically independent of sediment type, species, or HOC hydrophobicity. Therefore, in theory, bioaccumulation can be calculated easily using eq 1 and applied over a wide variety of environmental conditions for potential use by regulatory agencies (10) for establishing sediment quality guidelines. A predicted BSAF value of 1.6 also can be derived from nonequilibrium steady-state bioaccumulation models for nonmetabolized HOCs with log Kow < 6 (12). If the BSAF deviates from the expected value, then one or more of the assumptions behind BSAF are not satisfied. The BSAF model has attracted attention because it is simple and easy to use. As a result, there have been numerous studies undertaken, both in the laboratory and in the environment, to determine the values of BSAF and to evaluate the utility of the BSAF model in predicting bioaccumulation. BSAF values have been measured in several types of aquatic biota, such as fish (8, 13-16), bivalves (3-7, 17-22), crustaceans (14, 18, 19), and polychaetes (3, 4, 6, 15, 19). BSAF values have been measured for a wide assortment of HOCs, such as polychlorinated biphenyls (PCBs) (4-8, 11, 14, 16, 19-21), polycyclic aromatic hydrocarbons (PAHs) (5, 17, 22), pesticides such as dieldrin (23), DDT and its major metabolites (3, 5, 17), and polychlorinated dioxins and furans (13, 18, 19). These studies also have investigated factors that may influence BSAF values, such as the type and hydrophobicity of HOCs, the concentrations of HOCs (6), the metabolism of HOCs (24), the sediment and organic carbon composition (4, 5), the type of species and lipid composition (4), the effects of disequilibrium (8, 17, 19, 25), and the effects of biomagnification and food webs (12, 14, 16). BSAF values from many of these studies have been reviewed by Tracey and Hansen (26), who found that BSAF values for many hydrophobic compounds were similar for various species in a number of habitat groups (e.g., benthic, infaunal, scavenger). In general, prior research supports the use of the BSAF model but warns of caveats due to conditions that violate one or more of its assumptions (e.g., disequilibrium, biotransformation, mass transfer limitations). Although there have been studies done in the laboratory and in the field to support the use of BSAF values as a predictive tool in assessing bioaccumulation potential, the BSAF model has not been validated under in situ environmental conditions over a large spatial scale and for multiple species. Individual studies on BSAF values in the literature have focused mainly on laboratory or small-scale field studies and generally target only a very limited set of HOCs and(or) species. A previous review study of BSAF values (26) did evaluate the applicability of BSAF over many of these studies, but comparisons among multiple studies with different VOL. 35, NO. 9, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1709
TABLE 1. Biota Sampled and Used in BSAF Paired Measurements in This Study biota type bivalves benthic fish
pelagic fish
common name
scientific name
taxon levela
no. of PMb
mussels asiatic clam asiatic clam common carp white sucker largescale sucker bridgelip sucker sculpins paiute sculpin channel catfish rainbow trout brown trout Eastern mosquitofish smallmouth bass largemouth bass
Unionidae Corbicula Corbicula manilensis Cyprinus carpio Catostomus commersoni Catostomus macrocheilus Catostomus columbianus Cottus Cottus beldingi Ictalurus punctatus Oncorhynchus mykiss Salmo trutta Gambusia holbrooki Micropterus dolomieui Micropterus salmoides
F G S S S S S G S S S S S S S
1 7 64 18 136 11 9 19 1 4 14 1 3 6 2
a Taxon level: F, family; G, genus; S, species. b No. of PM, number of paired sediment-biota measurements in which both sediment and biota concentrations are above respective reporting limits.
analytical, experimental, and sampling methods can be problematic. In order for the BSAF model to be widely applicable, it should hold for many HOCs and over a broad range of in situ environmental conditions in a consistent manner. The objective of this paper is to evaluate the BSAF model for some common organochlorine compounds, ranging in log Kow from about 5.6 to 7.0 in river bed sediment and biota throughout the continental United States on the basis of samples from the National Water-Quality Assessment (NAWQA) Program of the U.S. Geological Survey (USGS). This analysis uses measurements of chlorinated HOCs in sediment and biota at 485 fluvial sites across the United States to test the BSAF model under in situ environmental conditions, to analyze the effects of different environmental parameters (e.g., species type, chemical type, sediment organic carbon), and to determine the applicability of this model for first-level screening of bioaccumulation potential under natural environmental conditions.
Materials and Methods Sampling. Samples of sediment and biota tissues were collected between 1992 and 1995, generally during summer or autumn low-flow conditions. Multiple samples were collected from each location and composited to help homogenize spatial variability within the stream. A total of about 600 paired sediment and biota samples from 485 sites, ranging from undeveloped to agricultural to urban land use, were analyzed. Specific details and a summary of chemical concentrations in sediment and biota can be found in Wong et al. (27). Sediment sample collection and processing methods are described in detail by Shelton and Capel (28). Briefly, finegrained sediment was collected from several depositional zones within a stream reach using a hand-held core sampler and then composited into a single sample, resulting in a reach average of fine-grained surficial sediment. In general, 5-10 depositional zones were used at a site. A sample from the surficial 2-3 cm of sediment in each depositional zone was collected and composited. The composite sample then was processed through a 2.0-mm stainless steel sieve, packed in ice, and shipped to the USGS National Water-Quality Laboratory (NWQL) for analysis. Biota sample collection and processing procedures varied for different taxa (29). Bivalves were collected by hand or with a rake, rinsed, and depurated in streamwater for 24 h, measured and weighed, then frozen and shipped on ice to the USGS NWQL for analysis. Fish were collected by 1710
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 9, 2001
electroshocking or seining, then sacrificed, rinsed, measured, and weighed. Whole fish were wrapped individually in aluminum foil, frozen and packed in ice, and shipped to the USGS NWQL for analysis. Multiple organisms of the same taxon from a given site were composited in order to average individual variability and to meet a minimum mass requirement for analysis. Bivalve samples typically were composites of about 50 individual organisms, and fish samples were composites of 5-8 organisms. Because no single aquatic biota taxon is available nationwide, multiple taxa were sampled. For HOC analysis, the taxa included bivalves (Corbicula species, the asiatic clam), bottom-feeding fish (e.g., carp, white sucker, channel catfish), and predator fish (e.g., brook or brown trout, largemouth bass). Biota sampled in this study are listed in Table 1 (and in the Table 1 of the Supporting Information). Details on taxon selection procedures and the underlying rationale are provided elsewhere (29). Chemical Analysis. Details on chemical analysis methods for chlorinated HOCs in sediment and biotic tissue are described by Foreman et al. (30) and Leiker et al. (31), respectively. Briefly, residual water was removed from sediment by centrifugation, and about 25 g dry weight equivalent was extracted with dichloromethane in a Soxhlet apparatus. The extracts were then reduced in volume and filtered to remove solids, and a portion was passed through a gel permeation chromatography (GPC). The GPC extract was solvent exchanged to hexane, cleaned up and fractionated using alumina/silica adsorption chromatography, and analyzed using gas chromatography/electron capture detection (GC/ECD). Target analytes and their reporting limits are listed in Table 2. Organic and inorganic carbon in sediment was determined by induction furnace oxidation and subsequent thermal conductivity measurement of evolved CO2 (32). Bivalves were analyzed as soft tissue and fish as whole fish. Biotic samples were analyzed by homogenizing frozen whole body biological tissue. Extracts were processed in a similar manner (31) as organochlorine compounds in sediment. Briefly, a sample of the tissue composite was extracted with dichloromethane in a Soxhlet apparatus, and an aliquot was removed for percent lipid determination. A portion was passed through a GPC system to remove lipids and other interferences. The extract was then solvent-exchanged into hexane, fractionated by alumina/silica adsorption chromatography, and analyzed by GC/ECD. Target analytes and their reporting limits are listed in Table 2. Lipid content in tissue samples was determined gravimetrically (31).
TABLE 2. Chemical Analytes Used in Individual BSAF Values (Both Media Above Reporting Limits) and Statistical Comparison of Compound-Specific BSAF Distributionsa fish chemical name
biota RLb
sediment RLb
all data all data except p,p′-DDE
cis-chlordane trans-chlordane p,p′-DDD p,p′-DDE p,p′-DDT dieldrin cis-nonachlor trans-nonachlor total PCBs
dacthal (DCPA) o,p′-DDD o,p′-DDE o,p′-DDT heptachlor epoxide lindane toxaphene
bivalves
median BSAF
Nc
3.3 2.3
24 153
5 5 5 5 5 5 5 5 50
Included in Statistical Comparison of Compound-Specific BSAF Distributions 1 2.9 18 1 1.1 13 1 2.6 28 1 8.6 71 2 0.7 25 1 3.4 18 1 2.1 8 1 4.5 18 50 2.4 11
5 5 5 5 5 5 200
Not Included in Statistical Comparison of Compound-Specific BSAF Distributionsf 5 0.1 3 1 1.8 4 1 2.8 1 2 1.1 4 1 0.8 1 1 2.5 1 200 0
median BSAF
Nc
d e
2.8 2.0
72 48
d e
AB D C A D C BCD AB BCD
1.0 6.1 1.7 5.8 3.5 2.8
6 3 5 24 6 5 0 4 5
E E E E E E
1.5 2.7
4.5 1.7 2.0 0.5 0.6
E E
2 3 4 3 0 0 2
a
Compound-specific BSAF distributions sharing the same letter are not statistically different. Analytes without letters were not included in compound-specific statistical analysis (Sediment-fish: two-way ANOVA on data ranks. R ) 0.05, P < 0.0001, and Tukey-Kramer multiple comparison test on data ranks, overall R ) 0.05. Sediment-bivalves: Kruskal-Wallis test, R ) 0.05, P ) 0.14). b RL, reporting limit (µg/kg dry weight in sediment, µg/kg wet weight in biota). c N, number of paired measurements. d Sediment-fish and sediment-bivalve BSAF distributions not statistically different (Mann-Whitney-Wilcoxon test, R ) 0.05, P ) 0.29). e Sediment-fish and sediment-bivalve BSAF distributions (excluding p,p′-DDE) not statistically different (Mann-Whitney-Wilcoxon test, R ) 0.05, P ) 0.86). f These were not included due to the limited number of observations.
Laboratory quality assurance samples (spiked samples, standard reference material samples, various blanks, and other standard solutions) were analyzed along with field samples. DDT breakdown in GC injector ports, which would bias measured BSAF values, was monitored by periodic injection of standard solutions, and kept within limits (e20%) such as those stated in EPA methods (33). Analytical method performance and quality assurance/quality control are discussed elsewhere (30, 31). Database Selection and Statistical Methods. A BSAF value was calculated for each individual data pair, which consisted of the measured concentrations of one chemical analyte in one sediment and one biota sample taken at the same site on the same date. Duplicate sediment samples taken at a site were matched to corresponding biota samples to produce multiple data pairs (i.e., one data pair for each duplicate) and vice versa. Duplicates made up 31% of the data pair between both sediment-fish and sediment-bivalves. The BSAF database potentially consisted of 4181 data pairs for individual target analytes in sediment-fish and 2835 data pairs for individual target analytes in sediment-bivalves. Only 5% of the total potential number of data pairs in sediment-fish (224) and 3% in sediment-bivalves (72) had concentrations above the reporting limits in both matrixes (Table 1 in Supporting Information). Statistical comparisons within and between the BSAF distributions (sediment-fish and sediment-bivalves) were calculated. Nonparametric significance tests were used because BSAF distributions were skewed and because some subgroupings (i.e., BSAF for specific chemical analytes) had small nonnormal distributions (34). Differences at the 95% confidence level (R ) 0.05) were considered significant. Comparisons between the two distributions were compared with Mann-Whitney-Wilcoxon rank sum tests (34). Comparisons of BSAF values among different analytes and biotic species were conducted with Kruskal-Wallis significance
FIGURE 1. Organic carbon-normalized sediment concentrations versus lipid-normalized biota concentrations of individual organochlorine chemicals for paired sediment and biota measurements in which both values were above the reporting limit. Lines have a slope of unity; points along a 1:1 line have the same BSAF value. tests (34). If the null hypothesis (no difference among BSAF distributions or medians) was rejected, the differences in BSAF values were determined by two-factor ANOVA on data ranks (34) and with the Tukey-Kramer honestly significantdifference multiple comparison test (MCT) (overall R ) 0.05).
Results and Discussion BSAF Values in Fish and Bivalves. Figures1 and 2 show the distribution of BSAF values of each data pair for both sediment-fish and sediment-bivalves. Figure 1 shows that there is a great deal of variability in the BSAF values, although most of the data pairs fall slightly above the lines corresponding to the theoretical BSAF values of 1-2. There is also no systematic bias between data pairs with fish and those VOL. 35, NO. 9, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1711
FIGURE 2. Distribution of individual BSAF values for (A) sedimentfish paired measurements and (B) sediment-bivalve paired measurements. Distribution of BSAF values for all chemical compounds in black bars; distribution of BSAF values for p,p′-DDE in lighter bars. with bivalves, as the data pairs for both media are intermingled, showing that both media have similar BSAF distributions. The total BSAF distributions for both fish and bivalves are approximately log-normal (Figure 2). The median BSAF is 3.3 for fish with a 95% confidence interval (34) from 2.8 to 3.9; the median BSAF for bivalves is 2.8 with a 95% confidence interval from 2.1 to 3.9. As suggested by the BSAF scatterplot (Figure 1), there is no significant difference (Table 2) between the median BSAF values of the fish and bivalve distributions (Mann-Whitney-Wilcoxon test, R ) 0.05). The overall BSAF distribution is the aggregate total of the BSAF values for individual data pairs of all analytes. Because the DDT metabolite p,p′-DDE constitutes the largest fraction for both fish and bivalves (Table 2), the BSAF values for p,p′-DDE could substantially affect the overall BSAF distribution. If the data pairs of p,p′-DDE are excluded, the median BSAF values for both fish (median of 2.3, 95% confidence interval from 1.8 to 2.6) and bivalves (median of 2.0, 95% confidence interval from 1.4 to 3.3) are lower than the median BSAF values calculated using all chemical analytes (Table 2), indicating that BSAF values for p,p′-DDE are generally higher than the BSAF values for the other compounds. However, the median BSAF values (excluding p,p′-DDE) for fish and bivalves still are not statistically different from one another (Mann-Whitney-Wilcoxon test, R ) 0.05). There is no significant difference between BSAF values for fish and bivalves despite the fact that these are field data and are not controlled for differences in environmental conditions and exposure to the HOCs. Most prior studies on BSAF have concentrated on benthic organisms such as bivalves and polychaetes rather than fish. Benthic organisms are in direct contact with sediment contaminated with HOCs and do not move appreciably. Fish, on the other hand, have relatively greater movement throughout the water column. 1712
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 9, 2001
Some fish migrate to other parts of the stream and feed on organisms besides those living in sediment. This could preclude the use of the BSAF model, if the fish are not exposed to the HOCs in the sediment at the locations they are sampled. The fish sampled for this study are mostly benthic fish (Table 1), which are in greater contact with the sediment than other types of fish. These results suggest that the BSAF model may be applicable to benthic fish and may explain why the BSAF values for fish are similar to those for bivalves, because both benthic fish and bivalves are in contact with sediment and are exposed to HOCs in sediment. Bierman (1) observed that BSAF values in sculpins and smelt in the Great Lakes were similar to those in oligochaetes, in agreement with observations in this study, and suggesting that the BSAF model may be applicable for benthic fish. The values of BSAF for this study are generally higher than the theoretical estimates of BSAF (from 1 to 2). Several other studies have also suggested that measured BSAF values tend to exceed the theoretical value (6-8, 11, 26). One possible explanation is that biomagnification of HOCs is occurring in the organisms that were sampled. Equilibrium partitioning models, such as the BSAF, assume that HOCs are in equilibrium between sediment, pore water, and the organism (2). Equilibrium should be maintained, whether an organism takes up HOCs by partitioning or by ingestion of contaminated material. Biomagnification results when food digestion and absorption from the gastrointestinal tract, accompanied by inflow of more contaminated food increases the concentration and fugacity of the HOC in the gastrointestinal tract relative to that in the original food. This creates a fugacity gradient driving the diffusion of HOCs from the gastrointestinal tract into the organism and raising the fugacity of the predator over that in the prey. The result is higher BSAF values than theoretical predictions (12). However, biomagnification is not the only possible explanation. The higher values of the BSAF could also be due to other factors, such as metabolism for some compounds (i.e., metabolites) or the amount and type of organic carbon pool present. These factors are discussed further below. BSAF values in this study are generally higher than theoretical predictions and higher than many values reported in the literature, especially from laboratory studies, for the same types of HOCs (4-8, 11, 13, 15, 19, 20, 26). Comparisons among different BSAF studies are complicated by the fact they are based on different analytical techniques, targeted HOCs, species, exposure times, and exposure routes. In many laboratory studies, HOCs were spiked into the sediment, and the uptake by biota was measured. The extent of bioavailability of HOCs in these laboratory studies may differ from that in environmental samples because exposure times generally are longer in the field than in laboratory experiments. HOCs sorbed to aged sediment in the environment may not be as bioavailable as spiked sediment, and HOCs on aged sediment may not be as readily removed from the sediment by the analytical method than freshly spiked sediment (8). There is a great deal of variability in BSAF values in most studies, including this one (Figures 1 and 2; 4-8, 11, 13, 15, 19, 20, 22, 26). Comparing results between studies is complicated by the way BSAF values are often reported in the literature. Many studies have reported measured BSAF values as means and distributions as standard deviations and(or) ranges. Means and standard deviations are not meaningful for highly skewed BSAF distributions (34) such as observed in this study, as shown by the log-normal BSAF distributions in Figure 2. With these caveats in mind, the mean BSAF values in this study (6.5 for fish and 4.6 for bivalves) are higher than the mean BSAF values for common organochlorine HOCs reported in the literature, such as those measured in laboratory studies, ranging from 0.2 to 5.4 (4,
5, 7, 8, 15, 19, 20), as well as mean BSAF values those from other field studies, ranging from 0.9 to 5.9 (6, 8, 13, 15). On the other hand, some of the reported BSAF values, such as for PCB 28 in molluscs (14) were quite high (BSAF ) 21.3). If the data pairs for p,p′-DDE (a metabolite) are excluded, the mean BSAF for fish in this study (3.2) is consistent with the mean values from other field studies (6,11). Effect of Chemical Type on BSAF. Many of the analytes in this study are known to degrade slowly in the environment, and a few are themselves metabolites that are recalcitrant to biodegradation (e.g., DDE and DDD). The BSAF model assumes that the chemical analyte is not transformed in the sediment or the biota. To assess the validity of the BSAF model as a screening tool, the effect of chemical transformation on BSAF needs to be examined. To investigate the effect of different chemical analytes on BSAF, a two-factor ANOVA on BSAF data ranks with subsequent Tukey-Kramer multiple comparison tests was performed on the BSAF data for fish (34). The first factor in the ANOVA was chemical analyte, which consisted of p,p′DDT and its major metabolites p,p′-DDD and p,p′-DDE, total PCBs, dieldrin, and cis- and trans-nonachlor and cis- and trans-chlordane. Analytes with fewer than 10 data pairs in fish were excluded from this analysis. The second factor was species type, which consisted of sculpins (Paiute sculpin and other Cottus genus fish), carp (common carp), catfish (channel catfish), mosquitofish (Eastern mosquitofish), bass (smallmouth and largemouth bass), suckers (white, largescale, and bridgelip suckers) and trout (rainbow and brown trout). A two-factor ANOVA was not done on bivalve BSAF data ranks because all but one of the data pairs for bivalves (Table 1) were for the asiatic clam (Corbicula genus). BSAF values in fish are influenced by both species type and chemical analyte (both factors significant, P < 0.0001). There is no significant interaction between the two factors (P ) 0.5530). The effect of species type on BSAF will be discussed in the next section. No significant compound-specific differences in BSAF values were found in bivalves (Kruskal-Wallis test, P ) 0.14, Table 2). It must be noted, however, that the extremely limited number of bivalve data pairs for some chemical analytes (Table 2) results in a loss of accuracy for the large-sample approximation to the Kruskal-Wallis test used by statistical software packages (34). Subject to this caveat, the lack of significant differences in compound-specific BSAF values for bivalves suggests that the BSAF model is valid for bivalves for the chemicals tested in this study. The agreement in compound-specific bivalve BSAF values contrasts with some prior bivalve studies that have reported significant differences (5, 6, 19, 20). For example, Pruell et al. (19) observed preferential accumulation of lower molecular weight PCB congeners, which was attributed to the very low lipid content of the species studied. Bivalve BSAF values also appear to decrease with increasing PCB chlorination (6, 20) suggesting that all congeners do not partition similarly. However, these studies (6, 19, 20) examined BSAF values for individual PCB congeners, whereas the PCB results in this study are for total PCBs. In addition, the effect of chemical hydrophobicity on the BSAF, which was observed in some previous studies, cannot be assessed in this study since the analytes measured have fairly similar Kow values (log Kow ∼ 5.6-7.0). In contrast to the bivalves, the two-factor ANOVA on data ranks showed significant differences among median BSAF values for different chemical analytes in fish (Table 2, P < 0.0001). The Tukey-Kramer analysis found that the compound-specific BSAF values for p,p′-DDE (median ) 8.6) and trans-nonachlor (median ) 4.5) are significantly higher than BSAF values for most other compounds; the BSAF values of p,p′-DDT (median ) 0.7) and trans-chlordane (median )
1.1) are significantly lower than most other BSAF values (Table 2). The BSAF values for p,p′-DDE make the overall BSAF distribution appear bimodal, with the p,p′-DDE values making up the majority of the higher BSAF values (Figure 2). The high BSAF values for p,p′-DDE and the low BSAF values for p,p′-DDT indicate preferential enrichment and depletion of these two compounds, respectively, in fish as compared to sediment. Clearly, the BSAF model assumption of no metabolism or degradation in either medium does not hold for these two chemicals. In soils and sediment, DDT is biotransformed into DDE under aerobic conditions and into DDD under anaerobic conditions. Laboratory studies have observed that many fish species can transform DDT into DDE and DDD (35). Similarly, the high BSAF values for transnonachlor and the low BSAF values for trans-chlordane are also consistent with differential metabolism and(or) degradation of the latter compound compared to the former. Some fish species (e.g., goldfish, bluegills, cichlids) have been observed to degrade chlordane isomers (35, 36). The shifting proportion of chlordane constituents over time, coupled with the presence of oxychlordane in fish but not sediment (27), suggest that the reason the BSAF values of trans-nonachlor and trans-chlordane differ from that of other compounds is the biotransformation of chlordane constituents in fish. In summary, BSAF values for the different compounds (within the log Kow range of about 5.6-7.0) in this study are similar, except for those of compounds that are known to be biotransformed, such as DDT and chlordane, implying that BSAF may distinguish between chemicals that are metabolized and those that are not. Effect of Species Type on BSAF. The two-factor ANOVA on BSAF data ranks with subsequent Tukey-Kramer multiple comparison tests was used to investigate the effect of different species types on the values of BSAF in fish. The fish, even those of the same species, probably have life histories and exposures to HOCs that are very different because they were sampled in rivers across the United States. The data from these fish are aggregated based on species for the analysis presented here. As stated previously, BSAF values in the fish in this study are significantly influenced both by the type of species and by the chemical analyte in question, with no significant interaction between the two (P < 0.0001). The Tukey-Kramer analysis showed that BSAF values for bass and mosquitofish were higher than those for other species types, as shown in Figure 3. It is important to bear in mind that there are only 8 data pairs in bass and 3 data pairs in mosquitofish in the present study in which both sediment and biota concentrations were above their respective reporting limits. The effect of species composition on BSAF depends on a number of different processes. One process is the route of exposure of the organism. Foster et al. (17) found in laboratory studies that BSAF values were much higher in the Baltic clam (Macoma balthica) than in the soft-shell clam (Mya arenaria), which bioaccumulated only trace amounts of the targeted HOCs (PAHs, pthalates, and diphenyl ether). This discrepancy was explained by Mya being a filter feeder, which would accumulate HOCs from the water column not the bed sediment. Lake et al. (6) found no significant difference in BSAF values of PCBs between the polychaete deposit feeder Nephtys incisa and the predator Glycera but observed lower BSAF values in the filter feeder Mercenaria mercenaria for organisms measured in the field. Ankley et al. (15) observed that fathead minnows had lower BSAF values of PCBs than did oligochaetes and attributed this effect to the lower exposure of the fish to HOC-contaminated sediment. But, if lower exposure to sediment is the reason behind the different BSAF values in bass and mosquitofish, then the BSAF would be expected to be lower for these two types of fish and not higher as observed. VOL. 35, NO. 9, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1713
FIGURE 3. Box plot of BSAF values (both values above respective reporting limits) by species type and by habitat. Each box shows median, 25th percentile, and 75th percentile; end whiskers show 10th and 90th percentiles. BSAF values for mosquitofish and bass are statistically higher than those for other fish species types (twoway ANOVA on data ranks, r ) 0.05, P < 0.0001, and TukeyKramer multiple comparison test on data ranks, overall r ) 0.05). Number on top of each box plot is number of paired measurements for that species type. Another possible explanation is that BSAF values increase at higher trophic levels due to biomagnification of HOCs. MacDonald et al. (16) found that BSAF values observed in several lakes increased with increasing trophic level. Bass are predatory fish. Thus, they may be subject to biomagnification. However, trout also are pelagic predatory fish, and BSAF values in these fish are not significantly different than those of the benthic fish. Effect of Sediment Organic Carbon and Biota Lipid Content on BSAF. To investigate the effect of the sediment and lipid carbon, the BSAF data pairs were examined for systematic biases in BSAF values due to these carbon pools. There does not appear to be any relationship between the fish BSAF and the amount of either carbon pool (sediment foc vs BSAF P ) 0.88, fish flipid vs BSAF P ) 0.63). Deletion of one unusually high fish BSAF value (BSAF ) 184, a data pair from an agricultural site heavily contaminated with DDTs) does not change the lack of relationship. However, for bivalves there are significant (R ) 0.05) but weak relationships between BSAF and the amount of both carbon pools present: a positive one for sediment organic carbon (foc vs BSAF P ) 0.0005, r2 ) 0.16) and a negative one for bivalve lipids (flipid vs BSAF P ) 0.0093, r2 ) 0.093). Literature reports on the effect of the two carbon pools on the value of BSAF are conflicting. Boese et al. (20) observed variability in the BSAF from various sediment types to be about 2-3-fold but noted that this variation was the same as the range of uncertainty for equilibrium partitioning. Ferraro et al. (5) found bivalve BSAF values to be lower (
