Relationship Between Sediment Toxicity and Contamination in San Francisco Bay

Prepared by
Bruce Thompson, San Francisco Estuary Institute
Brian Anderson, University of California, Institute of Marine Sciences
John Hunt, University of California, Institute of Marine Sciences
Karen Taberski, San Francisco Bay Regional Water Quality Control Board
Bryn Phillips, University of California, Institute of Marine Sciences

Prepared for the
San Francisco Estuary Regional Monitoring Program
San Francisco Estuary Institute
2nd Floor
7770 Pardee Lane

Oakland, CA 94621

December 1997

RMP Contribution #27

Table of Contents

Abstract

Introduction

Methods

Results and Discussion

Conclusions

Acknowledgments

References

Abstract

Sediment contamination and toxicity were monitored at 14 sites throughout San Francisco Bay between 1991-1995. Sediment bioassays using the amphipod Eohaustorius estuarius and elutriate bioassays using larval bivalves (Mytilus sp., Crassostrea gigas) indicated different patterns of sediment toxicity in space and time. Overall, sediments were most toxic to Eohaustorius at Redwood Creek (88% of the tests). Amphipod toxicity was significantly inversely related to the mean ERM quotient, suggesting that cumulative concentrations of several contaminants could be responsible for toxicity. Further analysis identified mixtures of specific contaminants at each site that were most closely related to amphipod toxicity. Of those contaminants, chlordanes were related to amphipod survival at five sites where concentrations above 0.28 ng/dry g were always associated with toxicity. Seasonal patterns in low, and high molecular weight PAHs were related to toxicity at two stations where concentrations above 474 and 1,983 ng/dry g respectively were always associated with toxicity. Sediment elutriates severely reduced normal bivalve larval development at the San Joaquin and Sacramento Rivers where several metals in sediments were weakly related to toxicity. However, those results were confounded because contaminant measurements were not made on the elutriates.
 

Introduction

When toxicity is indicated in sediment bioassays it is generally assumed that some component(s) of the sediment to which the test organisms were exposed caused the observed biological effect (e.g., mortality, impairment of growth or development, etc.). However, since sediments tested in monitoring programs are usually complex mixtures of numerous potential toxicants, the exact sediment component(s) that caused toxicity in the bioassay are usually difficult to identify. In fact, based on sediment bioassays and measurements of sediment contamination from monitoring data, it is usually not possible to ascribe cause because the relationships are correlational. However, rigorous numerical analysis of relationships between sediment toxicity and contamination may identify significant associations, priority chemicals, or locations that can be used to form testable hypotheses for further experiments that could determine the actual causes of toxicity.

Sediment toxicity may occur under several conditions: A single contaminant could be present in sediments (bulk or pore water) at concentrations sufficient to cause toxicity (e.g., Hong and Reish, 1987; Word et al., 1987; Swartz et al., 1990; Hoke and Ankley, 1991). Several contaminants could be present in sediments in low concentrations that together cause toxicity (e.g., Hermans et al., 1984; Swartz et al., 1988, 1995; Plesha et al., 1988). Natural sediment components such as ammonia (Knezovich et al., 1995), sulfide (Thompson et al., 1991), or other natural toxins (e.g., Gribble, 1994) may cause toxicity. Under any condition, an inverse relationship between contaminant concentration and a biological effect is expected. Sediment contaminant concentration gradients may occur with distance from a source, as concentrations change over time, or in a series of experimental laboratory exposures, to which sediment bioassay endpoints would be expected to respond inversely.

The purpose of this study was to determine the relationships between sediment toxicity and sediment contamination in San Francisco Bay, and identify the contaminant(s) that were statistically associated with the observed toxicity. These analyses are an important step in developing understanding about which sediment components may be contributing to toxicity in the Bay. While the results presented do not demonstrate the cause of sediment toxicity, they have facilitated the articulation of testable hypotheses about possible causes. Environmental management requires information about the causes of sediment toxicity in order to target source control or remedial action.

Previous studies of the relationships between sediment contamination and toxicity in San Francisco Bay have reported conflicting findings. Long and Markel (1992) showed that abnormal bivalve development was most highly correlated with PAHs and that amphipod mortality was most highly correlated with organic carbon and benzo(a)pyrene. However, no strong associations between sediment toxicity and individual contaminants were reported by Hoffman et al. (1994). Organic content of sediments was shown to be more closely correlated with sediment bioassay results than any single contaminant by Risebrough (1994). Swartz et al. (1995) unique study at a superfund site in the Bay unequivocally showed that sediment toxicity was caused by a gradient of DDTs in sediments. Studies in other areas of the U.S. have had varying levels of success in relating specific sediment contaminants to toxicity (e.g. Swartz et al., 1986; Wolfe et al., 1996; Carr et al., 1996).
 

Methods

Sediment monitoring data from two related programs were analyzed. The State's Bay Protection and Toxic Cleanup Program (BPTCP) sampled sediments in 1991 and 1992. The San Francisco Estuary Regional Monitoring Program (RMP) began in 1993, monitoring many of the same sites as the BPTCP, as well as some new sites. Each program used similar methods. Details of sampling and analyses for the BPTCP samples are in Flegal et al. (1994) and for the RMP in the Quality Assurance Project Plan (Lowe and Hoenicke, 1996), and are summarized below.

Data were analyzed from 14 sites sampled from 1991 to 1995 (Figure 1). Seven estuary sites were sampled during eight sampling periods and six sites were sampled four or five times. Sampling occurred during wet (February-April) and dry (August-September) months in the region. As part of the BPTCP, a spatial gradient was sampled at Castro Cove in May 1992. Four sites extending from an old oil refinery discharge were sampled (Flegal et al., 1994).

Sediment samples were collected using a modified 0.1 m² van Veen grab. In the RMP, the top 5 cm of sediment were scooped from each of two replicate grabs and homogenized to provide a composite sample for each site. The BPTCP sampled oxic surface sediments, typically 1-10 cm deep, above the black, sulfide-rich layer. Aliquot samples were taken for each analytical laboratory and for sediment bioassays. All collection apparatus was constructed of Teflon^® or coated with Dykon^® or Kynar^®.
 

Sediment Analyses

Sediment grain-size was determined by standard dry sieving and pipette analysis (1991 and 1992), or x-ray transmission using a Sedigraph 5100 (1993-1995). Total organic carbon was analyzed using a Leeman 440 Elemental Analyzer (1991-1992) or a UIC Coulometrics CM 150 Analyzer (1993-1995).

Metals in sediments were extracted with aqua regia to avoid decomposition of the silicate matrix of sediments (Flegal et al., 1981). Thus, metals tightly bound in sediments may not have been fully recovered as with hydrofluoric acid digestion. Trace metals (except As, Hg, Se) were measured using graphite furnace atomic absorption spectrometry preceded by sample pre-concentration using the APDC/DDC organic extraction method (Bruland et al., 1985; Flegal et al., 1991). Mercury samples were photo-oxidized with bromium chloride and quantified using cold vapor atomic fluorescence. Hydride generation coupled with atomic absorption spectroscopy was used to quantify As and Se.

Trace organics were analyzed by two separate laboratories. Samples were freeze-dried, mixed with kiln-fired sodium sulfate, and soxhlet-extracted with methylene chloride in both laboratories. In samples analyzed by Texas A & M‚s, Geochemistry and Energy Research Group (GERG) laboratory (1994-1995), extracts were concentrated and purified using EPA Method 3611, alumina column purification to remove matrix interferences. Chlorinated hydrocarbons were analyzed on a Hewlett Packard 5890A Series II capillary gas chromatograph utilizing electron capture detectors with 30 m x 0.25 mm fused silica capillary columns with DB5 and DB17 bonded phase. PAHs were quantified using a Hewlett-Packard 5890 Series II capillary gas chromatograph equipped with a mass spectral detector in the selected ion monitoring mode using a 30 m x 0.32 mm fused silica capillary column with DB5 bonded phase. In samples analyzed by R. Risebrough, University of California (1991-1993), extraction was conducted using hexane and florisil column chromatography. Chlorinated hydrocarbons were eluted in the first fraction, and PAHs and pesticides were eluted in the second fraction. A Varian 3400 electron capture gas chromatograph, and Saturn II GC/MS, and a DB5, 30 m columns were used. Both laboratories provided quality assurance data, including standard reference materials and internal standards, which indicated that the results from the two laboratories were comparable.

The sediment variables evaluated in this study are listed on Table 1. The detectable components of major classes of trace organic compounds were summed to reduce the total number of variables used in multivariate analyses, and because different numbers of component compounds were measured in each survey. Concentrations below detection levels (MDL) were set to values of one-half of the MDL. The GERG laboratory provided uncensored trace organic compound concentrations. Percent fine sediments (<63 mm) and total organic carbon (TOC) were included in multivariate analyses to account for contaminant covariation with those parameters.

Bulk sediment contaminant concentrations are often expressed as the presumed „bioavailable‰ fraction. For trace organics, data analysis was conducted using both dry weight concentrations and concentrations normalized to TOC. Neither monitoring program measured acid volatile sulfides and simultaneously extracted metals (AVS/SEM). All concentrations are expressed on a dry weight basis, except where noted.

Table 1. List of sediment parameters used in data analysis.

Sediment Bioassays

Two different sediment bioassays were conducted at all sites. The estuarine amphipod Eohaustorius estuarius was exposed to bulk sediment (ASTM, 1992) with percent survival after 10 days as the endpoint. Overlaying water salinities were adjusted to 15-28 parts per thousand (depending on ambient salinity at the station) by mixing clean seawater and distilled water. Larval mussel (Mytilus sp.) or oyster (Crassostrea gigas) sediment elutriate (water soluble fraction) bioassays were also conducted (ASTM, 1991) with percent normal larval development after 48 hours as the endpoint. The species used depended on seasonal larval availability. The elutriate test was included in the RMP by the local regulatory agency to extend a historical database using that test and to provide information about elutriate toxicity at „background‰ sites in the Bay, against which similar tests of dredge materials could be compared. Elutriate solutions were prepared by adding 100 g of sediment to 400 ml of clean sea water, shaking for 10 seconds, settling for 24 hours, and carefully decanting the eluate (U.S. EPA and COE, 1977). Elutriate test salinity was 28 ± 3 parts per thousand. The results of the tests using larval mussels and oysters were combined and are referred to as the larval bivalve test. Such combination assumes similar responses to the sediment elutriates by the embryos of each species.

Measurements of total ammonia and sulfides were made on interstitial or overlaying water samples during the bioassays. However, the measurements made on the 1991-1992 samples were not considered reliable because the detection limits were above those known to elicit effects in the species used. Measurements made on the 1993-1995 samples collected from the same sites as the 1991-1992 samples were always below concentrations known to cause toxicity (U.S. EPA, 1994; Knezovich et al., 1995; Anderson et al., 1995). Therefore, it was assumed that apparent toxicity due to sulfides or ammonia did not confound the results.

A sediment sample was considered toxic if there was a significant difference (p<0.05) between the laboratory control and sample replicates using a t-test, and if the difference between control and sample means was greater than the minimum significant difference (MSD). The MSD criterion prevents the designation of toxicity in samples with very low replicate variance in the controls. The MSD values were established separately for each test by examination of a cumulative frequency distribution of the minimum significant difference (MSD) from analyses of 119 Eohaustorius tests, and 34 larval bivalve tests conducted in San Francisco Bay. Based on those power curves, 90% of the Eohaustorius tests made a statistically significant distinction between the sample and the control if the difference between them was * 18.8%. The 90^th percentile MSD value for the bivalve test was 21% (Hunt et al., 1996). Similar analyses have been used by Thursby and Schlekat (1993), Schimmel et al. (1994), and Carr et al. (1996) for other species.
 

Data Analysis and Interpretation

Sediment quality guidelines (Long et al., 1995; Long and Morgan, 1990) were used to evaluate if sediment concentrations were within ranges that have been previously associated with biological effects. Those guidelines were derived from a large national database and are currently the most widely used and accepted sediment effects guidelines available. However, Effects-Range values only exist for 28 contaminants, Effects-Range values for Ni and DDTs are not considered to be very accurate predictors of effects, and values for some pesticides, such as chlordane and dieldrin, were not included in the more recent listing (Long et al., 1995) because of limited data. In interpreting the guidelines, concentrations below the Effects Range-Low (ERL) are „rarely‰ associated with adverse effects, concentrations between the ERL and Effects Range-Median (ERM) are „occasionally‰ associated with adverse effects, and concentrations above the ERM are „frequently‰ associated with adverse effects (Long et al., 1995).

The probability for amphipod toxicity is higher for samples with contaminant concentrations that exceed ERMs than for those that exceed ERLs (Long et al., in press). However, as will be shown, only Ni was consistently above the ERM in the Bay, therefore, in this study, concentrations are compared to the ERLs where exceedances are interpreted as a potential for „occasional‰ adverse effects.

ERM values were also used to calculate a mean ERM quotient (mERMq, Long et al., in press). The concentration of each contaminant was divided by its ERM to produce a quotient, or proportion of the ERM. For the PAHs, only the ERMs for individual compounds were used, not ERMs for total PAHs, LPAHs, and HPAHs. Similarly, only the ERM for total DDT was used. Quotients calculated for all contaminants in each sample were summed, then divided by the number of contaminants whose ERMs were used to calculate each sum. The last step is useful since the number of contaminants measured at each site changed over time. Mean ERM quotient values were used to evaluate the contribution of many sediment contaminants at each site to toxicity. Similar approaches have been used by Carr et al. (1996) and Canfield et al. (1996).

Multivariate analyses of relationships between sediment contaminant concentrations and toxicity test endpoints, as recommended by Green et al., 1993, were conducted in several steps using Statistical Analysis System (1995) software. The first step was to carefully consider which samples in space and time could be analyzed together. Different sites may have different patterns of contamination, and concentrations may change over time. Sediment contaminant patterns at various combinations of sites in space and time were examined using cluster analysis and principal components analysis (PCA). Multivariate analyses were only conducted at sites where toxicity occurred more than once. The next step was to eliminate contaminants that were not inversely correlated with the bioassay endpoints to further reduce the number of variables in the PCA, enhancing the ability of the analysis to select sediment variables related to toxicity. PCA was conducted using the remaining sediment variables. Only factors with eigenvalues greater than one were retained, and varimax rotation was applied. Next, multiple regression analyses (R² analysis, stepwise variable selection) was conducted to determine statistical relationships between the PCA factor scores (independent variables) and the bioassay endpoints (dependent variable) at each site. The final step was to evaluate the concentrations of each contaminant that composed the PCA factor(s) that were most closely related to toxicity. Correlations (product-moment) between concentrations and bioassay endpoints were examined and comparisons to the ERLs were made to determine whether the contaminants were above concentrations that have been associated with biological effects.
 

Results and Discussion

Patterns in Sediment Contamination

Sediment contaminant concentrations at most sites changed over the 5 years sampled. Detailed descriptions of those variations are beyond the scope of this paper. However, as examples, at Sacramento River and Grizzly Bay, chlordane concentrations were elevated in 1992 and decreased over time to concentrations below the old ERL of 0.5 ppb (Figure. 2). Concentrations also changed seasonally at several sites. Low molecular weight PAHs (LPAHs) at Alameda and San Bruno Shoal had higher concentrations during wet sampling periods than dry sampling periods (Figure. 2). Other contaminants, such as Cr at San Joaquin River, showed more random changes in concentration over time (Figure. 3).

Principal components analysis of sediment contaminant patterns at various combinations of sites and times showed that each site was somewhat different, and changed over time. Those changes in concentrations established temporal patterns to which patterns in sediment toxicity were expected to respond if the contaminant(s) actually influenced toxicity. Pooling data from several adjacent sites tended to average out important site-specific variations in sediment concentration patterns.
 

Patterns in Sediment Toxicity

The amphipod and larval bivalve bioassays indicated that sediment toxicity was widespread in the Bay in space and time (Figure 1). However, the endpoints of the amphipod and larval bivalve bioassays were not directly related (r=-0.23, p=0.031), and each test exhibited a different pattern of toxicity. Sediments were most frequently toxic to amphipods at Redwood Creek where 88% of the tests were toxic. The incidence of toxicity decreased, and mean percent survival increased with distance from that site, suggesting local sources of contamination. However, there was no obvious contaminant gradient from Redwood Creek (average mERMq=0.151) towards South Bay (average mERMq=0.184) and San Bruno Shoal (average mERMq=0.135). Sediments were toxic in 50-60% of the tests at several sites in the northern Estuary, but sediments from Horseshoe Bay, Red Rock, Davis Point, and San Joaquin River were never toxic to amphipods, probably due to the relatively uncontaminated coarse sediments at those sites (mean=45-87% sand). At Castro Cove, 80% of the tests were toxic, but samples from the site farthest from the old refinery outfall was not toxic.

There was temporal variability in the results of the amphipod tests at most sites between 1991 and 1995. Percent survival increased significantly over time at Grizzly Bay, Napa River, and South Bay (slope t-test, p<0.05; Figure 2). Although there appears to be an increasing trend at Sacramento River and Redwood Creek, it was not statistically significant; there was no obvious trend in percent survival at the remaining sites. There were also seasonal (wet, dry) differences in amphipod toxicity. Percent survival during the wet periods was significantly lower than during the dry periods (Wilcoxon 2- sample test, p=0.003). Seasonal differences were obvious at San Joaquin River, Grizzly Bay, Alameda, and San Bruno Shoal (Figure 2).

Sediment elutriates were always extremely toxic to larval bivalves at the Sacramento and San Joaquin River sites. Fewer than 2.5% of larvae developed normally in samples from the Sacramento River, and fewer than 4.6% developed normally at San Joaquin River. The incidence of larval bivalve toxicity decreased, and mean percent normal development increased with distance from those sites, but since elutriate chemistry was not measured, it is not known if a corresponding contaminant gradient existed. Toxicity also occurred in half the tests at Alameda. No larval bivalve toxicity was observed at Pinole Point, Davis Point, San Bruno Shoal, or South Bay (Figure 1).

Larval bivalve development decreased over time at Grizzly Bay (slope t-test, p=0.013), but was highly variable (Figure 3); larval bivalves did not develop at all in samples from this site during the two wettest years (February 1993 and 1995). There were no significant trends in larval

development over time at any other sites, ant there were no wet-dry seasonal differences in normal larval development (Wilcoxon 2-sample test, p=0.145).

Relationships Between Sediment Toxicity and Sediment Quality Guidelines

At the estuary sites, Ni was the only contaminant in sediments with concentrations above the ERM (51.6 µg/g), and it was elevated in nearly all samples. The primary source of Ni is probably serpentine rock, a common geological feature in the region (S. Luoma, USGS, pers. comm.). Arsenic, Cr, Cu, Hg, Ni, DDTs, dieldrin, and chlordanes often exceeded ERLs at most sites. At the Castro Cove gradient sites, Hg, Ni, HPAHs, total PAHs, and chlordanes were occasionally above their ERMs, especially at the location adjacent to the old outfall. All of the other contaminants, except Ag, were occasionally to frequently above their ERLs.

Percent survival of Eohaustorius was significantly inversely correlated with the mean ERM quotient (mERMq) when all data was combined (Figure 4). That relationship suggests that mixtures of many contaminants in sediments, in relatively low concentrations, may influence amphipod toxicity. That analysis also provided values that reflect the potential for sediment samples from the Bay to be toxic: samples with values below 0.105 were never toxic, values between 0.114 and 0.182 were toxic in about half the tests, values above 0.185 were toxic in 89% of the tests, and values above 0.219 were always toxic. Despite the significant relationship between amphipod survival and mERMq using all samples, similar analyses for each site showed that amphipod survival was significantly inversely correlated (p>0.05) with mERMq only at Sacramento River and Castro Cove.

Eohaustorius is apparently more sensitive to mixtures of contamination than other amphipods commonly used in sediment bioassays (Rhepoxinius abronius, Ampelisca abdita). The mERMq values associated with toxicity were much higher for those species than listed above for Eohaustorius (Long et al., in press): mERMqs below 0.10 were toxic in up to 12% of the tests, and values above 1.0 were toxic in 71% of the tests.

No meaningful relationship between mERMq and larval bivalve development was observed (Figure 4). Instead, normal development increased with increasing mERMq, whereas an inverse relationship was expected. The lack of a meaningful relationship with the bivalve test is probably because mERMq values were calculated from measurements of contaminants in bulk sediment, whereas sediment elutriates were used in the larval bivalve bioassays.
 

Site Specific Relationships Between Sediment Concentrations and Bioassay Endpoints

Since sediment contaminant patterns at each site were different from each other, all samples collected at each site over time were analyzed together providing a site-specific evaluation. Analyses using TOC normalized trace organics concentrations produced similar results as using dry weight concentrations except for the amphipod test at the Sacramento River and Redwood Creek, and the bivalve test at Alameda. Therefore, results from analysis using dry weight concentrations are reported below with the TOC normalized exceptions noted.

Eohaustorius Bioassays. Multiple regression analysis identified one or two PCA factors at each site that accounted for more than 61% of the variation in amphipod survival at all sites except South Bay, and resulted in significant regression models (p<0.05) at half of the sites (Table 2). The contaminants that composed the PCA factor(s) most closely associated with toxicity were probably those that contributed most to toxicity. Further evaluation of those individual contaminants showed that some contaminants were significantly inversely correlated with percent survival over time and/or were above their ERL concentration indicating that those concentrations have been „occasionally‰ associated with biological effects.

Chlordanes were included in the PCA factors related to toxicity at five sites: four in the northern Estuary (Sacramento River, Grizzly Bay, Napa River, Pinole Pt.), and at Redwood Creek. Chlordanes were significantly inversely correlated with amphipod survival at all sites except the Napa River, and were above the old ERL (0.5 ppb) in most of the toxic samples at those sites.

Combining the data from those five sites showed that chlordanes concentrations above 0.28 ng/g were always associated with toxicity (Figure 5). That concentration is very near the old ERL of 0.5 ng/g, and the concentration of 0.3 ng/g predicted by equilibrium partitioning for chronic effects (Pavlou et al., 1987). There is no sediment LC50 for chlordanes effects on Eohaustorius for comparison. However, chlordanes were not always associated with toxicity above those concentrations. Chlordanes were positively correlated with amphipod survival at three sites (Table 2), and 13 samples collected from sites where chlordanes were not identified by PCA had concentrations above 0.28 ng/g and were not toxic. Using TOC normalized chlordane concentrations did not rectify those differences. Other studies in the Bay have also reported chlordanes above 0.28 ng/g without toxicity (Risebrough, 1994). Further, chlordanes in sediments have generally decreased below the ERL of 0.5 ng/g in the Bay since February 1994. Therefore, chlordanes influence on sediment toxicity in the Bay appeared to have been restricted to specific sites and times.

Using TOC normalized trace organic concentrations produced different results at the Sacramento River. None of the three PCA factors accounted for large proportions of the variance in percent survival. Factor 2 (Ni, OC-HPAH) accounted for most (23%) of the variation, but neither contaminant was significantly correlated with percent survival. At Redwood Creek, analysis using TOC normalized organic concentrations removed PCBs from the factors related to toxicity, and strengthened their statistical relationships with percent survival (R²=0.79, p=0.028).

Another pattern was observed at Alameda and San Bruno Shoal in the Central Bay, where LPAHs and HPAHs were related to toxicity. Although the PCA factor at each station accounted for more than 75% of the variation in amphipod survival, LPAHs were significantly inversely correlated with survival only at Alameda, due to the small sample size (n=4). However, LPAHs and HPAHs were usually above the ERL at both sites, and the correspondence between toxicity and seasonal patterns of LPAHs was striking (Figure 2) suggesting that wet weather runoff may be the source of PAHs associated with toxicity. Combining the data from those two sites showed that samples with LPAH concentrations above 474 ng/g were always with toxic and samples with HPAH concentrations above 1,983 ng/g were always toxic (Figure 6). Both concentrations are near their respective ERL values (552 and 1,700 ng/g). No LC₅₀ for sediment LPAHs or HPAHs effects on Eohaustorius were found in the literature. Three or four samples from sites where LPAHs or HPAHs were not identified by PCA had concentrations above those values and were not toxic. Using TOC normalized concentrations did not rectify the differences.

Analysis of the spatial gradient at Castro Cove showed that seven trace metals were associated with the toxicity (Table 2). All seven metals were significantly inversely correlated with percent survival, and five of them were usually above the ERL. The mERMq was significantly inversely correlated with percent survival at those sites, suggesting that additive effects, possibly of the metals identified, were associated with toxicity.

At Yerba Buena Island, none of the PCA factor contaminants were significantly inversely correlated with percent survival (p>0.05), but Ni, DDTs, Cr, and Cu were often above their ERLs (Figure 2). However, DDTs and Ni were also above the ERLs in the six non-toxic samples. DDTs probably were not a very important determinant of toxicity since TOC normalized DDTs were 3 orders of magnitude below the toxicity threshold of 300 µg/g-OC reported for San Francisco Bay (Swartz et al., 1994). The mERMq values were 0.143 and 0.149 in the two toxic samples suggesting that the additive effects of several contaminants could have influenced toxicity.

Analysis of the South Bay site did not produce meaningful results. The PCA factor represented only Ag, and formed a weak relationship with percent survival (Table 2). All three PCA factors together only accounted for about half the variation in survival. Silver was not significantly correlated with percent survival (r=-0.54, p=0.17) and was always below the ERL. Although not identified by PCA, or significantly inversely correlated with amphipod survival, Cu, HPAHs, chlordanes, DDTs, and PCBs were all above their ERLs in at least half the samples, and the mERMq was above 0.169 in all samples suggesting the possibility of cumulative effects. Alternatively, that site is adjacent to freshwater inflows to the Bay where the pesticide diazinon in sediments was measured in elevated concentrations (WWC, 1996). Other unmeasured anthropogenic or natural contaminants could also have been involved.

Silver was associated with toxicity at six sites and was significantly inversely correlated with survival at three of them, but was only above the ERL (1.0 µg/g) at Redwood Creek (Table 2). In samples from those six sites, Ag concentrations above about 0.5 µg/g were always associated with toxicity. Similarly, Cd, Cu and Pb were each associated with toxicity at four sites where concentrations above 0.39, 48.0, and 32.0 µg/g, respectively, were always associated with toxicity.

Conversely, several contaminants may be ruled out as important factors in amphipod toxicity in the Bay. Mercury, As, and HCHs were not related to toxicity by multivariate analyses at any site, and Se was only related at one site. Additionally, those contaminants, as well as Ni, were positively correlated with amphipod survival at half of the sites or more. However, although not significantly correlated with toxicity, concentrations of those contaminants could contribute cumulatively to toxicity.

Although TOC is not toxic per se, patterns of TOC in sediments were related to amphipod survival at half of the sites (Table 2), but only where TOC covaried with chlorinated organics (chlordanes, DDTs). TOC probably mediates toxicity through its interactions with lipophylic trace organic contaminants (DiToro et al., 1991). TOC did not appear to negatively affect amphipod survival. The range of TOC in toxic samples (0.3-2.1%) was similar to the range of TOC in non-toxic samples (0.1-2.2%). Similarly, sediment grain-size (percent fines) appeared to have little influence on amphipod survival. Percent fines in the non-toxic samples ranged between 2-99%. Eohaustorius was not expected to respond negatively to grain-size (U.S. EPA, 1994).

Larval Bivalve Bioassay. Analysis of five sites where toxicity occurred more than once produced much weaker associations between sediment contamination and bivalve development than for amphipod survival. PCA factors accounted for greater than 68% of the variation in percent normal larval development at only two sites, San Joaquin River and Alameda, and regression analyses were only significant at the San Joaquin River site (Table 3). Contaminant factors accounted for less than 47% of the variation in bivalve development at the remaining three sites.

Arsenic was the only contaminant associated with toxicity at three adjacent sites in the northern Estuary: Sacramento and San Joaquin Rivers, and Grizzly Bay, and it was only significantly inversely correlated with bivalve development at Grizzly Bay. At those sites, arsenic concentrations were usually above the ERL of 8.1 µg/g (Table 2), and arsenic concentrations above 12.1 µg/g in sediments at those sites were always associated with toxicity (Figure 7).

Analyses using TOC normalized trace organic concentrations at Alameda changed the results. PCBs were included in factor 3, and the statistical relationship between factor 3 and normal development was much stronger (R²=0.99, p=0.005). Specifically, the inverse correlation between OC-DDTs and normal development was improved (r=-0.98, p=0.017), and suggests that chlorinated hydrocarbons may have contributed to larval bivalve toxicity.
 

Conclusions

Sediment bioassays conducted by monitoring programs in San Francisco Bay over the past five years indicated that toxicity was widespread in space and time. However, the two tests conducted showed markedly different patterns of toxicity, each providing a different type of information about the sediments, emphasizing the importance of using multiple tests in monitoring.

The sites analyzed in this paper were monitored to provide information on background, or ambient Bay conditions, and do not provide a comprehensive assessment of all Bay sediments. However, many other locations in San Francisco Bay not sampled by the RMP have also been shown to be toxic. In particular, areas near some of the major harbors, closed military bases, and superfund sites may have very toxic sediments (Chapman et al., 1987; Long and Markel, 1992; Swartz et al., 1994).

Cumulative concentrations of many contaminants in sediments, as mERMq, were statistically associated with decreased survival of Eohaustorius. That general relationship was investigated further using multivariate analysis which identified mixtures of specific contaminants that were related to toxicity at each site. Those mixtures probably contributed most to the toxicity observed because their concentration patterns were most closely associated with patterns in toxicity. Further evaluation of the individual contaminants associated with toxicity revealed strong relationships between chlordanes, LPAHs and HPAHs and toxicity at several sites suggesting that those contaminants could be key contributors to toxicity at those sites. Contaminants identified by multivariate analysis that were significantly inversely correlated with percent survival, and with concentrations above ERLs were probably the most important determinants of amphipod toxicity at most sites. However at Sacramento River and Castro Cove, mERMq was significantly inversely correlated with percent survival indicating that mixtures per se were possibly most important.

The use of mERMq to evaluate sediment toxicity in San Francisco Bay was probably affected by the inclusion of Ni values, which are not considered to be very reliable. Nickel contributed between 19-48% to mERMq values. Therefore, mERMq values associated with amphipod toxicity may be inflated. However, Ni was among the contaminants associated with amphipod toxicity at two sites and the contribution of Ni to toxicity in sediment mixtures cannot be ruled out.

All of the apparent threshold concentrations for amphipod toxicity determined in this study were closer to their respective ERLs than their ERMs demonstrating that in San Francisco Bay, ERLs provide reasonable indicators of the potential for sediment toxicity to Eohaustorius.

In contrast to the results for the amphipod bioassay, a statistically significant relationship between contaminants and reduced larval bivalve development was observed only at the San Joaquin River where several metals were implicated. Metals in sediments at the northern Estuary sites showed a consistent, yet weak, pattern related to larval bivalve toxicity. Those results correspond with results from independent toxicity identification evaluation (TIE) analyses on sediment elutriates from those sites where EDTA addition removed the toxicity to the larval bivalves (Phillips et al., 1997).

The weak relationships between larval bivalve toxicity and sediment contamination were probably the result of attempting to relate concentrations measured in bulk sediments with the results of bioassays using sediment elutriates. However, it is generally believed that the elutriate test provides useful information about sediment contamination. Measurements of contaminants in the elutriates, or TIEs are recommended for future RMP efforts.

The results presented in this study represent an important intermediate step in the determination of causes of sediment toxicity in San Francisco Bay. The associations between individual sediment contaminants and amphipod survival provide information that can be used to pose hypotheses about the causes of toxicity in the Bay. One hypothesis is that total chlordanes concentrations in sediments above 0.28 ng/g causes significant mortality to Eohaustorius. Another hypothesis is that HPAHs above 1,984 ng/g, and LPAHs above 474 ng/g cause significant mortality. Other hypotheses about the toxicity of mixtures of those contaminants identified by multivariate analysis could also be tested. Further research should be conducted on elutriate or pore water chemistry and trace metals bioavailability at the San Joaquin and Sacramento Rivers to determine whether metals could be causing acute toxicity to bivalve larvae. Determination of the causes of sediment toxicity observed in monitoring will ultimately require evidence from numerical analysis of monitoring data and manipulative experiments. Such experiments could include toxicity identification evaluations (TIEs), laboratory and/or in situ sediment spiking and dose-response tests at concentrations shown to be associated with toxicity in this paper.  

Acknowledgments

The authors wish to thank the investigators and staffs of the laboratories that conducted the analyses presented in this paper: For trace metals: Russ Flegal, U.C. Santa Cruz, and Eric Prestbo, Brooks-Rand, Seattle, WA. For trace organics: Jose Serricano, Texas A&M, GERG, Bob Risebrough, UC Berkeley. Max Puckett, Laboratory Director of California Department of Fish and Game‚s Marine Pollution Studies Laboratory at Granite Canyon, provided facilities and logistical support for the sediment bioassays. Witold Piekarski, Hilary McNulty, Matt Englund, Michelle Hester, Lisa Weetman, Steve Osborn, and Steve Clark, U.C. Santa Cruz, assisted with the sediment bioassays. Thanks to Dr. Bob Smith, EcoAnalysis, Inc. for discussion of analytical methods. Applied Marine Sciences, Livermore, CA coordinated field sampling. Jung Yoon, Ted Daum, Sarah Lowe, and Adrienne Yang, SFEI, assisted with creation of data bases, graphics, and editing. This is RMP Contribution No. 27.
 

References

Anderson, S.L., J.P. Knezovich, J. Jelinski, and D.J. Steichen. 1995. The utility of pore-water toxicity testing for development of site-specific marine sediment quality objectives for metals. Report No. 37615, Lawrence Berkeley Laboratory, Berkeley, CA.

ASTM. 1991. Designation E 724-89: Standard Guide for Conducting Static Acute Toxicity Tests Starting with Embryos of Four Species of Saltwater Bivalve Molluscs. Volume 11.04. American Society for Testing and Materials, Philadelphia, PA.

ASTM. 1992. Designation E 1367: Standard Guide for Conducting 10-day Static Sediment Toxicity Tests with Marine and Estuarine Amphipods. Volume 11.04. American Society for Testing and Materials, Philadelphia, PA.

Bruland, K.W., K.H. Coale, and L. Mart. 1985. Analysis of seawater for dissolved cadmium, copper, and lead: An intercomparison of voltametric and atomic absorption methods. Marine Chem. 17:285-300.

Canfield, T.J., F.J. Dwyer, J.F. Fairchild, P.S. Haverland, C.G. Ingersoll, N.E. Kemble, D.R. Mount, T.W. LaPoint, G.A. Burton, and M.C. Swift. 1996. Assessing contamination in Great Lakes sediments using benthic invertebrate communities and the sediment quality triad approach. J. Great Lakes Res. 22:565-583.

Carr, R.S., E.R. Long, H.L. Windom, D.C. Chapman, G. Thursby, G.M. Sloane, and D.A. Wolfe. 1996. Sediment quality assessment studies of Tampa Bay, Florida. Environ. Toxicol. Chem. 15:1218-1231.

Chapman, P.M., R.N. Dexter, and E.R. Long. 1987. Synoptic measures of sediment contamination, toxicity and infaunal community composition (the sediment quality triad) in San Francisco Bay. Mar. Ecol. Prog. Ser. 37:75-96.

DiToro, D.M., C.S. Zarba, D.J. Hansen, W.J. Berry, R.C. Swartz, C.E. Cowan, S.P. Pavlou, H.E. Allen, N.A. Thomas, and P.R. Paquin. 1991. Technical basis for establishing sediment quality criteria for nonionic organic chemicals using equilibrium partitioning. Environ. Toxicol. Chem. 10:1541-1583.

Flegal, A.R., L.S. Cutter, and J.H. Martin. 1981. A study of the chemistry of marine sediments and wastewater sludge. California State Water Resources Control Board, Sacramento, CA. 69 pp.

Flegal, A.R., R.W. Risebrough, B. Anderson, J. Hunt, S. Anderson, J. Oliver, M. Stephenson, and R. Packard. 1994. San Francisco estuary pilot regional monitoring program: Sediment studies. San Francisco Bay Regional Water Quality Control Board, Oakland, CA.

Green, R.H., J.M. Boyd, and J.S. MacDonald. 1993. Relating sets of variables in environmental studies: The sediment quality triad as a paradigm. Environmetrics 4:439-457.

Gribble, G.W. 1994. The natural production of chlorinated compounds. Environ. Sci. Technol. 28:310A.

Hermans, J., H. Canton, N. Steyger, and R. Wegman. 1984. Joint effects of a mixture of 14 chemicals on mortality and inhibition of reproduction of Daphnia magna. Aquat. Toxicol. 5:315-322.

Hoffman, E.R., S.L. Anderson, and J.P. Knezovich 1994. Determinants of sediment toxicity in San Francisco Bay. LBL-36592, UC-000. Lawrence Berkeley Lab.

Hoke, R. and G. Ankley. 1991. Results of dieldrin sediment spiking study conducted in support of U.S. EPA development of sediment quality criteria. In, Sediment Quality Criteria for the Protection of Benthic Organisms: Dieldrin. Draft Report. U.S. Environmental Protection Agency, Washington, DC.

Hong, J.S. and D.J. Reish. 1987. Acute toxicity of cadmium to eight species of marine amphipod and isopod crustaceans from southern California. Bull. Environ. Contam. Toxicol. 39:884-888.

Hunt, J.W., B.S. Anderson, S. Tudor, M.D. Stephenson, H.M. Puckett, F.H. Palmer, and M. Reeve. 1996. Marine Bioassay Project, Eighth Report: Refinement and implementation of foureffluent toxicity testing methods using indigenous marine species. Report No. 94-4 WQ. State Water Resources Control Board, Sacramento, CA. pp. 85-104.

Knezovich, J.P., D.J. Steichen, J.A. Jelinski, and S.L. Anderson. 1995. Sulfide tolerance in four marine species used to evaluate sediment and pore-water toxicity. Report UCRL-JC-121893. Lawrence Livermore National Laboratory, CA.

Long, E.R. and L.G. Morgan. 1990. The Potential for Biological Effects of Sediment-sorbed Contaminants Tested in the National Status and Trends Program. NOAA Tech. Memo. NOS OMA 52, 175 pp.

Long, E.R., and R. Markel. 1992. An Evaluation of the Extent and Magnitude of Biological Effects Associated with Chemical Contaminants in San Francisco Bay, California. NOAA Tech. Memo. NOS ORCA 64, Seattle, WA.

Long, E.R., D.D. MacDonald, S.L. Smith, and F.D. Calder. 1995. Incidence of adverse biological effects within ranges of chemical concentrations in marine and estuarine sediments. Env. Mgmt. 19:81-97.

Long, E.R., L.J. Field, and D.D. MacDonald, (In press). Predicting toxicity in marine sediments with numerical sediment quality guidelines. Environ. Toxicol. and Chem.

Lowe, S., and R. Hoenicke. 1996. Quality Assurance Project Plan: Regional Monitoring Program for Trace Substances. San Francisco Estuary Institute, Richmond, CA. 29 pp.

Pavlou, S., R. Kadeg, A. Turner, and M. Marchlik. 1987. Sediment quality criteria methodology validation: Uncertainty analysis of sediment normalization theory for nonpolar organic contaminants. Work Assignment 56, Task 3. Battelle, Washington, DC. 95 pp.

Phillips, B., J. Hunt, and B. Anderson. 1997. Investigating classes of compounds assoicated with sediment toxicity at Regional Monitoring Program river stations. In, 1996 Annual Report, San Francisco Estuary Regional Monitoring Program, San Francisco Estuary Institute, Richmond, CA.

Plesha, P.D., J.E. Stein, M.H. Schiewe, B.B. McCain, and U. Varanasi. 1988. Toxicity of marine sediments supplemented with mixtures of selected chlorinated and aromatic hydrocarbons to the infaunal amphipod Rhepoxynius abronius. Mar. Environ. Res., 25, 85-97.

Risebrough, R.W. 1994. Contaminants in San Francisco bay sediments: Relationships with toxicity studies undertaken by the National Oceanic and Atmospheric Administration, 1990. Report to the San Francisco Bay Regional Water Quality Control Board, the State Water Resources Control Board, and the U.S. Environmental Protection Agency. The Bodega Bay Institute, Berkeley, CA.

SAS Institute, Inc. 1995. SAS Procedures Guide, Version 6. SAS Institute, Inc., Cary, NC.

Schimmel, S.C., B.D. Melzian, D.E. Campbell, D.J. Strobel, S.J. Benyi, J.S. Rosen, H.W. Buffum, and N.I. Rubinstein. 1994. Statistical summary EMAP˜Estuaries Virginian Province˜1991. EPA/620/R-94/005.

Swartz, R.C., F.A. Cole, D.W. Schults, and W.A. DeBen. 1986. Ecological changes on the Palos Verdes Shelf near a large sewage outfall: 1980-1983. Mar. Ecol. Prog. Ser. 31:1-13.

Swartz, R.C., P.F. Kemp, D.W. Schultz, and J.O. Lamberson. 1988. Effects of mixtures of sediment contaminants on the marine infaunal amphipod, Rhepoxinius abronius. Environ. Toxicol. Chem. 7:1013-1020.

Swartz, R.C., D.W. Schultz, T.H. DeWitt, G.R. Ditsworth, and J.O. Lamberson. 1990. Toxicity of fluoranthene in sediment to marine amphipods: a test of the equilibrium partitioning approach to sediment quality criteria. Environ. Toxicol. Chem. 9:1071-1080.

Swartz, R.C., F.A. Cole, J.O. Lamberson, S.P. Ferraro, D.W. Schults, W.A. DeBen, H. Lee II, and R.J. Ozretich. 1994. Sediment toxicity, contamination, and amphipod abundance at a DDT- and dieldrin-contaminated site in San Francisco bay. Environ. Toxicol. Chem. 13:949-962.

Swartz, R.C., D.W. Schults, R.J. Ozretich, J.O. Lamberson, F.A Cole, T.H. DeWitt, M.S. Redmond, and S.P. Ferraro. 1995. SPAH: A model to predict the toxicity of poylnuclear aromatic hydrocarbon mixtures in field-collected sediments. Environ. Toxicol. Chem. 14:1997-1987.

Thompson, B., S. Bay, D. Greenstein, and J. Laughlin. 1991. Sublethal effects of hydrogen sulfide in sediments on the urchin Lytechinus pictus. Mar. Env. Res. 31:301-321.

Thursby, G.B. and C.E. Schlekat. 1993. Statistical analysis of 10-day solid phase toxicity data for amphipods. Abstract, 14^th Annual Meeting, Society of Environmental Toxicology and Chemistry.

U.S. EPA and COE. 1977. Technical committee on criteria for dredged and fill material, ecological evaluation of proposed discharge of dredge material into ocean waters; implementation manual of Section 103 of Public Law 92-532 (Marine Protection and Sanctuaries Act 1972), July 1977 (2^nd printing April 1978). U.S. Environmental Protection Agency; Environmental Effects Laboratory, U.S. Army Corps of Engineers, Waterways Experiment Station, Vicksburg, MI.

U.S. EPA. 1994. Methods for Assessing the Toxicity of Sediment-associated Contaminants with Estuarine and Marine Amphipods. EPA 600/R-94/025. Office of Research and Development, Narragansett, RI.

Wolfe, D.A., E.R. Long, and G.B. Thursby. 1996. Sediment toxicity in the Hudson-Raritan Estuary: Distribution and correlations with chemical contamination. Estuaries 19:901-912.

Word, J.Q., J.A. Ward, L.M. Franklin, V.I. Cullinan, and S.L. Kiesser. 1987. Evaluation of the Equilibrium Partitioning Theory for Estimating the Toxicity of the Non-polar Organic Compound DDT to the Sediment Dwelling Amphipod Rhepoxynius abronius. CSD 11, U.S. Environmental Protection Agency, Washington, DC.

WWC. 1996. Sediment Diazinon Special Study. Report to Alameda Countywide Cleanwater Program, Hayward, CA. From Woodward Clyde Consultants.