Predictors of mercury spatial patterns in San Francisco Bay forage fish

Ben K. Greenfield * , Darell G. Slotton , Katherine H. Harrold

Accepted for publication on July 22, 2013 Environmental Toxicology and Chemistry http://onlinelibrary.wiley.com/journal/10.1002/%28ISSN%291552-8618

San Francisco Estuary Institute, Richmond, CA, USA

Environmental Health Sciences Division, School of Public Health, University of California - Berkeley, CA, USA

Department of Environmental Science and Policy, University of California - Davis, CA, USA Current address: Environmental Sciences and Engineering, Gillings School of Global Public Health, University of North Carolina - Chapel Hill, NC, USA

* Address correspondence to bengreenfield@berkeley.edu

ABSTRACT Pollution reduction efforts should be targeted towards those sources that result in the highest bioaccumulation. For mercury (Hg) in estuaries and other complex water bodies, carefully designed biosentinel monitoring programs can help identify predictors of bioaccumulation and inform management priorities for source reduction. We employed a probabilistic forage fish Hg survey with hypothesis testing in San Francisco Bay. The study goal was to determine what pollution sources, regions, and landscape features were associated with elevated Hg bioaccumulation. Across 99 sites, whole body Hg concentrations in Mississippi silversides (Menidia audens) and topsmelt (Atherinops affinis) followed a broad spatial gradient, declining with distance from the Guadalupe River (Pearsons r = -0.69 and -0.42), which drains historic mining areas. Site landscape attributes and local Hg sources had subtle effects which differed between fish species. Topsmelt Hg increased in embayment sites (i.e., enclosed sites including channels, creek mouths, marinas, and coves) and sites with historic Hg contaminated sediment, suggesting an influence of legacy industrial and mining contamination. In 2008, Mississippi silverside Hg was reduced at sites draining wastewater treatment plants. Fish Hg was not related to abundance of surrounding wetland cover but was elevated in some watersheds draining from historic Hg mining operations. Results indicated both regional and site-specific influences for Hg bioaccumulation in San Francisco Bay, including legacy contamination and proximity to treated wastewater discharge.

Keywords: Mercury, Prey fish, Estuary, Biosentinel, Bioaccumulation

INTRODUCTION Mercury (Hg) is a global pollutant that adversely impacts ecosystems and human health. Global Hg concentrations are elevated due to widespread human use and inadvertent release, creating a need for coordinated efforts to curtail Hg release, transport, and exposure [1]. Anthropogenic Hg sources to estuarine and coastal ecosystems include runoff from Hg and gold mining operations, atmospheric emissions (e.g., coal combustion), and point sources associated with historic or current industrial activity [1-5]. Methylmercury (MeHg) is highly toxic and bioaccumulative [6], triggering reproductive effects in wildlife [7] and potential developmental and neurological effects to humans [8, 9]. At the regional scale, carefully designed research and monitoring is needed to prioritize MeHg management actions in the presence of multiple spatially-distributed sources. Comparative studies of MeHg in forage fish (small, short-lived prey fish, consumed by piscivorous wildlife) aid in describing spatiotemporal patterns and explanatory variables for MeHg food web accumulation [10-18]. Forage fish integrate exposure across a several month time period, and have limited ranges in age, diet, and movement area [17]. Thus, forage fish are often used to describe spatial patterns in food web MeHg, and the factors that contribute to elevated MeHg [10-13, 18]. However, probabilistic spatial surveys and hypothesis testing approaches are rarely employed to evaluate forage fish contamination within a single water body. San Francisco Bay (the Bay) is influenced by Hg watershed loads and sediment deposits from historic mining operations and industrial sources, making it an important system for characterizing ecosystem MeHg exposure [3, 19-21]. Local sources targeted for management reduction include Hg mines, stormwater runoff from urban and industrial watersheds, municipal publically owned wastewater treatment works (POTWs), drainage from the Central Valley

watersheds, and industrial facilities [3, 22]. Historic Bay sediment contamination also contributes Hg to the water column and food web [23-25]. As in other estuaries, the spatiotemporal dynamics of MeHg concentrations, bioavailability, and bioaccumulation in the Bay are influenced by complex biogeochemical factors, including variable primary productivity and sulfate reduction, in addition to spatial differences in Hg loading [26-29]. Due to this biogeochemical and spatial complexity, the relative importance of different source categories for Hg bioaccumulation is poorly understood. In addition to sources, there are several spatial factors that may influence MeHg bioaccumulation within the Bay. Bay sediment and biota Hg are elevated in proximity to a historic Hg mining district (New Almaden Mining District), in salt ponds and other semienclosed embayments, and in interior wetlands [3, 13, 30-32]. Wetlands are frequently sites of MeHg production, and consequently sources to adjacent ecosystems and biota [33-36]. Enclosed environments, channels, and freshwater tributaries are also frequently associated with increased MeHg in water, sediment, and biota [26, 37-40], due to the combined effects of watershed Hg loading, legacy industrial sources, elevated organic carbon deposition, and spatial variation in biota diets [4, 37, 41-43]. Forage fish sampling could indicate whether proximity to wetlands or embayment areas (such as enclosed marinas, backwater sloughs, stream drainages, and natural coves) predicts differences in biotic MeHg exposure within an estuary. We report Hg spatial patterns in Bay forage fish collected from 99 sites between 2008 and 2010. Unlike many ambient monitoring programs, the study design was hypothesis-based. Monitoring strata were defined, selected, and randomly subsampled to identify what kinds of locations within the Bay exhibit elevated Hg concentrations in forage fish. Since MeHg is the predominant Hg form in these fish [13], analyses of total Hg is assumed to indicate MeHg. Four

questions are addressed: (1) What are the spatial trends in Bay forage fish Hg? (2) Are Hg concentrations elevated in embayments, relative to open water sites within the Bay? (3) Does the extent of fringing wetland habitat correlate with Hg concentrations? (4) Are concentrations elevated at potential Hg source sites, relative to randomly selected sites? In addition to randomly selected sites, 4 types of source sites were evaluated: sites draining watersheds impacted by historic Hg mining (mine sites), sites draining urbanized and industrial watersheds (industrial watershed sites), sites receiving treated effluent from wastewater treatment facilities (POTW sites), and sites known to have elevated sediment Hg (contaminated sediment sites). METHODS Study design and site selection The study employed a stratified sampling design, intended to evaluate the 4 study questions based on a priori hypotheses (Supplemental Data Text). The sample design included 99 sites distributed along the entire shoreline of San Francisco Bay from Lower South Bay to Suisun Bay (Supplemental Data, Figure S1). Wetland channels and estuarine tributaries were included but salt ponds and tidal lakes were excluded. Sites were probabilistically selected from this sample frame using a Generalized Random Tessellation Stratified (GRTS) spatially balanced sampling design [44]. Two sample draws were performed: the first was for random locations across the entire Bay shoreline, and the second was from all identified points within the 4 source categories, treating each category as a stratum. The random sample draw included 2 categories (i.e., strata): open water sites (N = 25 sites) and embayment sites (N = 23 sites; Supplemental Data, Figure S2). The source sample draw included 4 categories: Hg mine creeks (N = 4 sites), watersheds draining urban and industrial areas (N = 13), POTW drainages (N = 7), and areas with relatively elevated sediment

THg or MeHg (N = 15). For each subcategory, appropriate sampling locations were identified using GIS, literature, and unpublished data (further detailed in Supplemental Data Text). Due to limited sample sizes for POTWs and Hg mine sites, all sites within these categories were sampled. To ensure sufficient coverage of wetland habitats, 12 additional sites adjacent to nearshore wetlands were sampled in 2008, including 6 sites fringing the South Bay and 6 sites fringing San Pablo and Suisun Bays (Supplemental Data Text). These wetland sites were only included in the analysis of fringing wetland habitat versus fish Hg. Fish sampling All fish sampling was performed by beach seine in 2008, 2009, and 2010. To minimize confounding seasonal variation with spatial variation, study analysis was restricted to the fall season (August 27 to November 30 of each year). The target species were topsmelt (Atherinops affinis; target total lengths of 60 - 100 mm) and Mississippi silverside (Menidia audens; target total lengths of 40 - 80 mm), both of which have been successfully employed in the Bay as Hg biosentinels [13, 16, 25]. Four composites of 5 individuals each per species were targeted for total Hg at each sampling event. Target composites each included similar-sized individuals, with the composites distributed in ascending 10 mm size windows spanning the overall size range targeted for each species (i.e., for Mississippi silverside: composite 1, N = 5 at 40 - 50 mm through composite 4, N = 5 at 70 - 80 mm). For the Bella Oaks and Borges Hg mine sites, target species were not available. At these 2 sites, prickly sculpin (Cottus asper, 52 - 100 mm), California roach (Hesperoleucus symmetricus, 54 - 82 mm), and three-spined stickleback (Gasterosteus aculeatus, 34 - 50 mm) were collected. Like the target species, these are all invertivores previously employed as Hg biosentinels in California [14, 15, 45-47].

Sample preparation and analysis All fish collection and preparation followed protocols developed at UC Davis, with slow cooling euthanasia method certified by the UC Davis Veterinary School's Institutional Animal Care and Use Committee. Fish were measured for total length, rinsed with site water, and sorted into labeled, freezer-grade plastic bags as composites for analysis, field frozen with air excluded and water surrounding, on dry ice, and subsequently transferred to a -20 C laboratory freezer. Composite whole body fish samples were subsequently thawed, weighed, dried to constant weight at 55 C, dry weight and percent solids recorded, and ground to a fine homogenous powder. Samples were analyzed for total Hg at the University of California-Davis. Analysis employed standard cold vapor atomic absorption (CVAA) spectrophotometry, using a dedicated Perkin Elmer Flow Injection Hg System (FIMS) with an AS-90 autosampler, following a 2 stage digestion at 90 C in a mixture of concentrated nitric and sulfuric acids with potassium permanganate. Routine analytical QA/QC included 20 QA/QC samples for every 30 analytical samples, and included blanks, aqueous standards, continuing control standards, standard reference materials with certified levels of Hg, laboratory split samples, matrix spike samples, and matrix spike duplicates. All results met QA protocols of the Regional Monitoring Program for Water Quality in San Francisco Bay and were well within laboratory control limits. All study Hg results are reported on a wet weight basis. Geospatial data Geospatial data were developed in ArcGIS v10. The Bay shoreline was partitioned into open water versus embayment site categories based on visual inspection of a Bay shoreline vector file with depth data overlay, and satellite imagery. Inclusion criteria were depth, degree of separation from the rest of the Bay, and presence of channels or sloughs. The embayment layer

included habitats throughout the Bay, with the largest areal coverage north of San Pablo and Suisun Bays (Supplemental Data, Figure S2). Two numeric geospatial attributes were examined for association with fish Hg: percent surrounding wetland area and distance from the Guadalupe River. Percent surrounding wetland area was based on a 500 m buffer, using data from Bay Area Aquatic Resource Inventory and Association of Bay Area Governments 2005 land-use polygons. Percent surrounding wetland was defined as the sum of the depressional, marsh, and tidal ditch land cover categories. Distance from the Guadalupe River, defined as the nautical distance from the westernmost tidal point of Coyote Creek, was negatively correlated with forage fish Hg at 22 sites sampled previously [13]. It was calculated following along the deep Bay channel, extending from the starting point to the upstream study extent of Suisun Bay (Mallard Island, near the confluence of the Sacramento and San Joaquin Rivers). Distance from the Guadalupe River indicates how close the sites are to the Hg contaminated New Almaden Mining District, which drains into the Lower South Bay near the community of Alviso. However, distance from the Guadalupe River also indicates general position along the Bay axis, with the most distant north Bay segments (Suisun Bay, San Pablo Bay) having potentially different net MeHg production and distribution from the progressively closer Central Bay, South Bay, and Lower South Bay [13, 26]. Data analysis Data analyses were performed using the linear mixed effects model function in R, version 2.15 [48]. Hg data were log 10 transformed to improve residual normality and variance homoskedasticity. Topsmelt and Mississippi silverside were analyzed separately. In line with the study questions, 4 separate linear models were built to examine the potential effect of spatial trend (distance from Guadalupe River); embayment category (embayment versus open Bay);

surrounding wetland abundance; or site type (i.e., POTW, contaminated sediment, industrial watershed, and random sites) on fish Hg. Model evaluation was performed manually, using backwards elimination of non-significant model terms. Parameter inclusion was based on the likelihood ratio test (with an alpha = 0.05 to retain a parameter) in combination with information theoretic criteria (i.e., AIC and BIC) [49]. Random effects were included to account for variability among sampling sites [50]. When significant, the slope for fish length was also allowed to vary by site. Four samples (3 Mississippi silverside and 1 topsmelt) were removed from the statistical analyses because their inclusion would have violated assumptions of residual normality and variance homoskedasticity. However, when analyses were performed with these samples included, the results were essentially unchanged. More details on the modeling approach and outlier removal are provided in the Supplemental Data Text. Of the 4 mine sites, Mississippi silversides were only present at the Guadalupe River upstream of Alviso Slough and topsmelt were only present at American Canyon Creek, draining Borges Mine. Since this was insufficient to statistically evaluate a mine site effect for these species, each mine site was compared to other data reported for additional species on an ad hoc basis. To provide context, data on additional species were compared to previously published Hg concentration data from mine sites [14, 46, 47] and unpublished data from reference (i.e., no known mine influence) sites. Unpublished data were obtained via queries performed on March 23, 2013 of the California Environmental Data Exchange Network (www.ceden.us), a collaboratively developed statewide environmental water quality database [51]. RESULTS Graphical analysis indicated a spatial trend in average forage fish Hg concentrations, with the highest concentrations in and adjacent to Lower South Bay, and concentrations progressively

10

decreasing towards South, Central, San Pablo, and Suisun Bays (Figure 1, Figure 2). This spatial gradient was more pronounced for Mississippi silverside (Figure 1), whereas topsmelt exhibited more local scale spatial heterogeneity, particularly within Central Bay (Figure 2). Hg concentrations were higher in Mississippi silversides (0.090 0.058 g/g, mean SD, N = 237) than topsmelt (0.041 0.015, N = 239), as reported in a prior study [13]. Total length was positively related to Hg and therefore included as a covariate in all models. Sampling year differences (treated as a categorical variable) were significant and included in some models (Table 1). Mixed models were needed to account for correlations among samples within a site, with site treated as a random effect. Additionally, a significant effect of site on the length covariate was observed for some models, and was thus incorporated as a random site effect on the length vs. Hg slope (Table 1). Distance from Guadalupe River Distance from the Guadalupe River (Question 1) was negatively related to Hg in Mississippi silverside (r = -0.69, N = 237, Figure 1) and topsmelt (r = - 0.42, N = 239, Figure 2). Distance was also a significant predictor in mixed models accounting for site effect (p < 0.0005, Table 1), and was therefore included as a covariate in models testing for other effects. The distance effect varied among sampling years. For silverside, in 2009, the decrease in Hg with distance from the Guadalupe River was weaker than other years (DistanceGuadalupe*Year2009 interaction, Figure 3). Based on model predicted concentrations, in 2009, the closest site to the Guadalupe River (Coyote Creek upstream of Alviso Slough) exhibited two-fold higher Hg concentrations than the furthest site (Kirker Creek near Pittsburg; 0.11 vs. 0.053 g/g), whereas in 2010, the predicted difference was four-fold (0.16 vs. 0.039 g/g). For topsmelt, in 2008, the

11

overall Hg decrease with distance from the Guadalupe River was weaker, compared to other years (Year2008*DistanceGuadalupe). Embayment and fringing wetland effects Hg concentrations in Mississippi silversides collected in embayment sites were not significantly different from Hg in Mississippi silversides collected in open water sites (Question 2; likelihood ratio test p = 0.096; Supplemental Data, Table S2). However, for topsmelt, Hg concentrations were significantly elevated in embayment sites compared to open water sites (p = 0.012), and embayment site Hg significantly increased with distance from the Guadalupe River and with fish length (Table 1, Supplemental Data, Table S3). Embayment sites in Central and San Pablo Bays were more often elevated in topsmelt Hg versus adjacent open water sites (Figure 2, Figure 4). For example, at the embayment site furthest from the Guadalupe River (the Petaluma River site), model predicted topsmelt Hg would be 0.043 g/g, whereas an open site at the same distance would have a predicted Hg of 0.029 g/g. Percent surrounding wetlands (Question 3) was not a significant predictor of Hg for Mississippi silverside or topsmelt. For both species, the final model included a significant increase with body length, a significant decrease with distance from the Guadalupe River, and no effect of wetlands (Table 1, Supplemental Data, Tables S4 and S5). For Mississippi silverside, the highest mercury wetland sites were in channels surrounding San Pablo and Suisun Bays, and had lower Hg than Lower South Bay and South Bay sites (Figure 1). Source site type effects Source site effects (Question 4) varied between Mississippi silverside and topsmelt. In 2008, Mississippi silverside Hg was lower at POTW sites than other site types (SourcePOTW*Year2008 interaction, Figure 3, Supplemental Data, Table S6). Based on model

12

predictions for average length fish, in 2008, POTW sites had approximately one-half of the Hg of non-POTW sites (0.035 vs. 0.068 g/g); in 2009 and 2010, POTW sites were predicted to be 0.015 g/g lower than non-POTW sites. Graphical analysis indicated POTW sites to be lower than nearby sites in both 2008 and 2010 (Figure 3). In 2009, there was no apparent pattern of POTW versus other sites. Topsmelt were only obtained at 1 POTW site, the Hayward wastewater treatment plant discharge pond, monitored in 2010. Topsmelt Hg concentrations at that site (0.021 0.0005 g/g, N = 4) were less than half the concentrations at the nearest site measured in 2010, the Eden Landing Shoreline (0.045 0.007 g/g, N = 4). Hg in topsmelt was moderately elevated at contaminated sediment sites (p = 0.032, Table 1, Supplemental Data, Table S7), which were only present in Lower South, South, and Central Bays (Figure 5). The model predicted topsmelt Hg at a contaminated sediment site to be 1.2 times that predicted for another site type in the same location. Fish species captured varied across the mining sites (Table 2), likely due to variable salinity conditions. The Guadalupe River upstream of Alviso Slough, which drains from the New Almaden Mine watershed, was elevated in Mississippi silverside Hg, consistent with the general spatial gradient observed in the present study and elsewhere [3, 13, 21, 25]. Concentrations at this site were within the range of spatial variation observed in Lower South Bay (Figure 1), but higher than the Baywide average and Mississippi silversides from Hg contaminated Clear Lake [14]. The Guadalupe River site also had extremely high Hg in three-spined stickleback. In the Napa River below the Bella Oaks Mine watershed, prickly sculpin Hg was comparable to Clear Lake [14] and higher than the average of 8 sites from the Sacramento-San Joaquin Rivers Delta. Napa River California roach Hg was higher than the average of 8 CA statewide sites lacking mine influence, but well below the concentration previously measured by Slotton et al. [46] at
13

Marsh Creek, a mine dominated creek that drains into Suisun Bay. American Canyon Creek, which drains from the Borges mine, was not elevated in topsmelt Hg, relative to general Baywide concentrations. Dry Creek also had relatively low Hg concentrations in prickly sculpin and California roach, and unremarkable concentrations in three-spined stickleback, suggesting a lack of influence of the nearby La Joya mine. Three-spined stickleback were relatively low in Hg within the urbanized industrial watershed of Zone 4 Line A, compared to mine sites and 4 other Bay sites. The Zone 4 Line A sampling location is within a flood drainage channel, several km above the Bay shoreline [52]. DISCUSSION We found that forage fish Hg concentrations were elevated in the southern Bay. This pattern has been noted previously [13], as well as elevated Hg or MeHg in southern Bay water, sediment, and shorebirds [3, 26, 53], suggesting that greater attention be dedicated to MeHg management in this region. The association between Hg and distance to the Guadalupe River indicates that Hg bioaccumulation in San Francisco Bay exhibits a spatial gradient across the relatively large distance of approximately 146 km from Coyote Creek upstream of Lower South Bay to the east side of Suisun Bay downstream of the Sacramento-San Joaquin Rivers Delta. A spatial gradient is also apparent in 202Hg stable isotope along the Bay for sediment and forage fish, suggesting increased bioaccumulation from Hg sources from the New Almaden Mining District, draining into Lower South Bay [21, 25]. The presence of tidal mixing and fluvial transport over time may create a relatively smooth gradient of both Hg source material and MeHg biogeochemistry. Results from the present and previous studies suggest that in the western United States, fish Hg bioaccumulation exhibits regional spatial gradients, often following

14

gradients of Hg mass in the sediments due to proximity to mines and other legacy Hg sources [5, 11, 14, 54, 55]. The higher Hg concentrations in Lower South Bay and South Bay likely result from multiple factors including historical Hg loading from the New Almaden Hg Mining District, South Bay hydrodynamics, and methylation in the Bay and adjacent habitats. Relatively long water residence times in Lower South Bay and South Bay may result in reducing conditions that favor sediment and water column MeHg production [3], with additional elevated MeHg production in the extensive salt pond complexes adjacent to Lower South Bay [31, 32], and periodic anoxia along Alviso Slough itself [16]. Elevated Hg in topsmelt but not Mississippi silverside at embayment sites (e.g., marinas, creeks, and backwater sloughs), and no relationship between surrounding wetlands and fish Hg, suggest a limited ability to predict biotic MeHg exposure based on natural landscape attributes. We hypothesized that surrounding wetland abundance would correlate with forage fish Hg based on the established role of freshwater wetlands as MeHg sources to adjacent waters [34], the consequent association between proximity to wetlands and freshwater fish Hg [12, 18, 35, 36], and evidence of elevated MeHg production in estuarine wetland sediment [30, 33, 56]. However, for these forage fish that reside within the subtidal open waters of the Bay, MeHg concentrations were decoupled from fringing wetland abundance. This is in contrast to northern temperate lakes, which are dominated by atmospheric deposition, and frequently exhibit at least a moderate effect of catchment wetland abundance and other landscape attributes on MeHg bioaccumulation [12, 18, 34, 57]. This finding suggests that the extensive wetland restoration activities planned for San Francisco Bay are not likely to adversely affect MeHg exposure for these subtidal Bay forage fish or their predators. Nevertheless, the Bay and associated wetlands comprise a range of

15

habitats, and changes in Hg exposure to organisms residing within the wetlands or ponds fringing the Bay could still occur as a result of restoration or other management activities [32]. Our hypothesis that embayment status could predict increased MeHg exposure in forage fish was based on elevated Hg accumulation in Bay forage fish species that heavily utilize intertidal and shoreline areas (e.g., Mississippi silverside) [13], elevated sediment and biota MeHg in proximity to freshwaters in the Bay and other estuaries [26, 37-40], increased exposure to anthropogenic Hg pollution at embayment sites [4, 31], and the possible importance of fringing wetlands, intertidal mudflat habitat, and shallow sediments for MeHg production at embayment sites [30, 38, 56]. The increase in topsmelt Hg from embayment sites was related to spatial location; differences were primarily observed in Central Bay, where Mississippi silversides were not readily available. We speculate that the embayment pattern for topsmelt largely stems from exposure to historic industrial contamination, because topsmelt Hg was also increased near legacy contaminated sediment. Historic industrial activity was abundant along the Central Bay shoreline, and is associated with elevated concentrations of PCBs, another legacy and industrial pollutant, in sediment and forage fish [58, 59]. This pattern suggests that regional priorities for minimizing MeHg production might focus on identifying and restoring those embayment sites with elevated sediment and biota MeHg. Source site type effects included higher topsmelt Hg near contaminated sediments, higher Hg near some historic mine drainages, and lower Hg adjacent to POTWs in 2008 and possibly 2010. Previous research suggests that Bay forage fish Hg and PCBs are sediment derived [25, 59], and we observed that topsmelt but not Mississippi silverside Hg corresponded to contaminated sediment sites. Other studies have also exhibited variable relationships between fish and sediment Hg (or MeHg), with associations observed in Texas rivers [11], the Hudson

16

River (New York/New Jersey) [60], and the Willamette River (Oregon/Washington) [5], but not in a survey of northeastern US freshwaters [57] or a Columbia River (Washington) reservoir [61]. In the Gulf of Maine, biota Hg is generally elevated in regions with elevated sediment Hg, but the bioaccumulation factor is lower in more contaminated areas, due to elevated total organic carbon reducing bioavailability [43]. The complexity of Hg methylation and bioavailability, biota movement, and food web structure all contribute to the weak and variable relationships between fish and sediment Hg [43, 61, 62]. The negative effect of POTWs on forage fish Hg was unexpected given that average total MeHg detected in discharge water from the 16 largest Bay POTWs was 0.37 ng/L, versus 0.096 ng/L in Bay ambient water [20]. We speculate that lower than expected forage fish Hg concentrations at some POTW sites may result from biodilution, in which increased primary and secondary production decreases Hg bioaccumulation and biomagnification [63, 64]. The reduced Mississippi silverside Hg concentration at 4 South Bay POTW sites is associated with elevated discharge water ammonium concentrations, compared to ambient Bay conditions [65]. This may result in increased rates of primary production, higher densities of Mississippi silversides and their invertebrate prey, or more rapid growth rates, all resulting in decreased tissue Hg concentrations. Forage fish densities during collections at these sites were also observed to be substantially greater than typical for the Bay (D.G. Slotton, personal observation). We observed inconsistent impacts of local mining, especially compared to the broad spatial gradient across the Bay. Concentrations in proximity to mining-impacted sites varied widely: the Guadalupe River downstream of New Almaden Mining District and the Napa River below the Bella Oaks Mine were elevated in fish Hg, and at or above prior measurements of the same fish species in mine influenced sites [14, 46]. In contrast, American Canyon Creek and Dry

17

Creek were not elevated, suggesting lower levels of residual contamination. In California roach monitored closer to the New Almaden Mines (Guadalupe Creek at Meridian Ave and Alamitos Creek at Harry Road), Hg concentrations were even greater, versus other sites in the local Guadalupe River watershed [66], and Hg isotopes indicate a New Almaden mining source signal in sediment and forage fish [21, 25]. In freshwater lakes and rivers, fish Hg concentrations are frequently elevated in sites impacted by mining waste versus reference sites, and tend to decrease with increasing distance from mining sources [5, 14, 55, 67, 68]. Hg is elevated near mines, processing facilities, or waste tailings even in areas with naturally occurring Hg deposits, and even with Hg mining completed several decades before fish collection. This indicates a remaining concern for mine Hg in the food web and a potential benefit of ongoing remediation focused on Hg mining sources such as the New Almaden Mining District [3]. Finally, sites adjacent to industrial watersheds hypothesized to be Hg-contaminated did not exhibit elevated forage fish Hg concentrations. This is consistent with the relatively small Hg mass discharged from these industrial watersheds, compared to other sources and Bay sediment. The San Francisco Bay TMDL Staff Report [22] estimates urban stormwater runoff to contribute 92 kg/yr Hg to the Bay, which is only 7.5% of all sources (1222 kg/yr) [22], and Hg isotope studies found a significant relationship between sediments and forage fish, without any notable deviations adjacent to more industrial sites [25]. Even stickleback collected within a small industrial watershed (Zone 4 Line A) were lower than at other sites, suggesting that industrial watersheds are not locations of elevated MeHg bioaccumulation. We demonstrated the use of biosentinel forage fish, combined with a stratified probabilistic survey design, to identify Hg bioaccumulation spatial patterns and sources in a large urbanized estuary. Both regional and local patterns were observed, reflecting the complex legacy

18

Hg sources and system hydrology. Regionally, there was a clear spatial gradient with distance from a historic Hg mining district. After accounting for that gradient, local differences among sites were subtle and varied between fish species. These findings suggest that forage fish Hg bioaccumulation predominantly exhibits broad regional variation, and that sources varying at local scales, including POTWs and legacy sediment Hg contamination, exhibit a secondary influence. SUPPLEMENTAL DATA Forage fish mercury spatial patterns and predictors in San Francisco Bay (Supplemental Text, Tables and Figures), and field and laboratory Hg data on 1260 forage fish samples (SFForageFishHg.csv). ACKNOWLEDGMENT N. Feger, C. Sommers, R. Looker, R. Schlipf, and the RMP Technical Review Committee assisted with site selection. S.M. Ayers, K. Ridolfi, A.R. Melwani, R.M. Allen, and M.A. Lent assisted with field sampling. A.R. Melwani and J. Griswold helped with statistical analysis, and R.M. Allen and M. Klatt helped with data interpretation. J.A. Davis, D.B. Senn, M.D. Sedlak, M. Lahiff, and two anonymous reviewers constructively reviewed drafts of the manuscript. Site access was provided by CDPR, CDFG, East Bay RPD, Mt. View Sanitary District, USACE, and SF Bay National Wildlife Refuge. The Regional Monitoring Program for Water Quality in San Francisco Bay funded the study (SFEI Publication 694), with additional support by a USEPA STAR Fellowship to B.G. REFERENCES 1. Lambert KF, Evers DC, Warner KA, King SL, Selin NE. 2012. Integrating mercury science and policy in the marine context: Challenges and opportunities. Environ Res 119:132-142.

19

2. Pacyna EG, Pacyna JM, Sundseth K, Munthe J, Kindbom K, Wilson S, Steenhuisen F, Maxson P. 2010. Global emission of mercury to the atmosphere from anthropogenic sources in 2005 and projections to 2020. Atmos Environ 44:2487-2499. 3. Davis JA, Looker RE, Yee D, Marvin-Dipasquale M, Grenier JL, Austin CM, McKee LJ, Greenfield BK, Brodberg R, Blum JD. 2012. Reducing methylmercury accumulation in the food webs of San Francisco Bay and its local watershed. Environ Res 119:3-26. 4. Sunderland EM, Amirbahman A, Burgess NM, Dalziel J, Harding G, Jones SH, Kamai E, Karagas MR, Shi X, Chen CY. 2012. Mercury sources and fate in the Gulf of Maine. Environ Res 119:27-41. 5. Hope BK, Rubin JR. 2005. Mercury levels and relationships in water, sediment, and fish tissue in the Willamette Basin, Oregon. Arch Environ Contam Toxicol 48:367-380. 6. Wiener JG, Spry DJ. 1996. Toxicological significance of mercury in freshwater fish. In Beyer WN, Heinz GH, Redmon-Norwood AW, eds, Environmental Contaminants in Wildlife: Interpreting Tissue Concentrations. Lewis, Boca Raton, FL, pp 297-339. 7. Scheuhammer AM, Meyer MW, Sandheinrich MB, Murray MW. 2007. Effects of environmental methylmercury on the health of wild birds, mammals, and fish. Ambio 36:1218. 8. National Research Council. 2000. Toxicological Effects of Methylmercury. Committee on the Toxicological Effects of Methylmercury. National Academies Press, Washington, DC. 9. Karagas MR, Choi AL, Oken E, Horvat M, Schoeny R, Kamai E, Cowell W, Grandjean P, Korrick S. 2012. Evidence on the human health effects of low-level methylmercury exposure. Environ Health Persp 120:799-806.

20

10. Snodgrass JW, Jagoe CH, Bryan AL, Brant HA, Burger J. 2000. Effects of trophic status and wetland morphology, hydroperiod, and water chemistry on mercury concentrations in fish. Can J Fish Aquat Sci 57:171-180. 11. Smith A, Abuzeineh AA, Chumchal MM, Bonner TH, Nowlin WH. 2010. Mercury contamination of the fish community of a semi-arid and arid river system: Spatial variation and the influence of environmental gradients. Environ Toxicol Chem 29:1762-1772. 12. Wiener JG, Knights BC, Sandheinrich MB, Jeremiason JD, Brigham ME, Engstrom DR, Woodruff LG, Cannon WF, Balogh SJ. 2006. Mercury in soils, lakes, and fish in Voyageurs National Park (Minnesota): importance of atmospheric deposition and ecosystem factors. Environ Sci Technol 40:6261-6268. 13. Greenfield BK, Jahn A. 2010. Mercury in San Francisco Bay forage fish. Environ Pollut 158:2716-2724. 14. Suchanek TH, Eagles-Smith CA, Slotton DG, Harner EJ, Colwell AE, Anderson NL, Mullen LH, Flanders JR, Adam DP, McElroy KJ. 2008. Spatiotemporal trends in fish mercury from a mine-dominated ecosystem: Clear Lake, California. Ecol Appl 18:A177-A195. 15. Eagles-Smith CA, Ackerman JT. 2009. Rapid changes in small fish mercury concentrations in estuarine wetlands: implications for wildlife risk and monitoring programs. Environ Sci Technol 43:8658-8664. 16. Greenfield BK, Melwani AR, Allen RM, Slotton DG, Ayers SM, Harrold KH, Ridolfi K, Jahn A, Grenier JL, Sandheinrich MB. 2013. Seasonal and annual trends in forage fish mercury concentrations, San Francisco Bay. Sci Total Environ 444:591-601. 17. Wiener JG, Bodaly RA, Brown SS, Lucotte M, Newman MC, Porcella DB, Reash RJ, Swain EB. 2007. Monitoring and evaluating trends in methylmercury accumulation in aquatic biota.

21

In Harris R, Krabbenhoft DP, Mason R, Murray MW, Reash R, Saltman T, eds, Ecosystem Responses to Mercury Contamination: Indicators of Change. CRC Press, Pensacola, FL, pp 87-122. 18. Greenfield BK, Hrabik TR, Harvey CJ, Carpenter SR. 2001. Predicting mercury levels in yellow perch: use of water chemistry, trophic ecology, and spatial traits. Can J Fish Aquat Sci 58:1419-1429. 19. Conaway CH, Black FJ, Grieb TM, Roy S, Flegal AR. 2008. Mercury in the San Francisco Estuary: A review. Rev Environ Contam Toxicol 194:29-54. 20. Yee D, McKee LJ, Oram JJ. 2011. A regional mass balance of methylmercury in San Francisco Bay, California, USA. Environ Toxicol Chem 30:88-96. 21. Gehrke GE, Blum JD, Marvin-DiPasquale M. 2011. Sources of mercury to San Francisco Bay surface sediment as revealed by mercury stable isotopes Geochim Cosmochim Acta 75:691-705. 22. SFBRWQCB. 2006. Mercury in San Francisco Bay Total Maximum Daily Load (TMDL) Proposed Basin Plan Amendment and Staff Report for Revised Total Maximum Daily Load (TMDL) and Proposed Mercury Water Quality Objectives. Final Report. California Regional Water Quality Control Board San Francisco Bay Region, Oakland, CA 23. Hornberger MI, Luoma SN, van Geen A, Fuller C, Anima R. 1999. Historical trends of metals in the sediments of San Francisco Bay, California. Mar Chem 64:39-55. 24. Conaway CH, Watson EB, Flanders JR, Flegal AR. 2004. Mercury deposition in a tidal marsh of south San Francisco Bay downstream of the historic New Almaden mining district, California. Mar Chem 90:175-184.

22

25. Gehrke GE, Blum JD, Slotton DG, Greenfield BK. 2011. Mercury isotopes link mercury in San Francisco Bay forage fish to surface sediments. Environ Sci Technol 45:1264-1270. 26. Conaway CH, Squire S, Mason RP, Flegal AR. 2003. Mercury speciation in the San Francisco Bay estuary. Mar Chem 80:199-225. 27. Marvin-DiPasquale M, Agee JL. 2003. Microbial mercury cycling in sediments of the San Francisco Bay-Delta. Estuaries 26:1517-1528. 28. Hammerschmidt CR, Fitzgerald WF. 2004. Geochemical controls on the production and distribution of methylmercury in near-shore marine sediments. Environ Sci Technol 38:14871495. 29. Luengen A, Flegal AR. 2009. Role of phytoplankton in mercury cycling in the San Francisco Bay estuary. Limnol Oceanogr 54:23-40. 30. Heim WA, Coale KH, Stephenson M, Choe K-Y, Gill GA, Foe C. 2007. Spatial and habitatbased variations in total and methyl mercury concentrations in surficial sediments in the San Francisco Bay-Delta. Environ Sci Technol 41:3501-3507. 31. Miles AK, Ricca MA. 2010. Temporal and spatial distributions of sediment mercury at salt pond wetland restoration sites, San Francisco Bay, CA, USA. Sci Total Environ 408:11541165. 32. Grenier JL, Davis JA. 2010. Water quality in South San Francisco Bay, California: current condition and potential issues for the South Bay Salt Pond Restoration Project. Rev Environ Contam Toxicol 206:115-147. 33. Bergamaschi BA, Fleck JA, Downing BD, Boss E, Pellerin B, Ganju NK, Schoellhamer DH, Byington AA, Heim WA, Stephenson M, Fujii R. 2011. Methyl mercury dynamics in a tidal wetland quantified using in situ optical measurements. Limnol Oceanogr 56:1355-1371.

23

34. Rudd JWM. 1995. Sources of methyl mercury to freshwater ecosystems: a review. Water Air Soil Pollut 80:697-713. 35. Chumchal MM, Drenner RW, Fry B, Hambright KD, Newland LW. 2008. Habitat-specific differences in mercury concentration in a top predator from a shallow lake. T Am Fish Soc 137:195-208. 36. Chumchal MM, Hambright KD. 2009. Ecological factors regulating mercury contamination of fish from Caddo Lake, Texas, USA. Environ Toxicol Chem 28:962-972. 37. Farmer TM, Wright RA, DeVries DR. 2010. Mercury concentration in two estuarine fish populations across a seasonal salinity gradient. T Am Fish Soc 139:1896-1912. 38. Fry B, Chumchal MM. 2012. Mercury bioaccumulation in estuarine food webs. Ecol Appl 22:606-623. 39. Harris R, Pollman C, Landing W, Evans D, Axelrad D, Hutchinson D, Morey SL, Rumbold D, Dukhovskoy D, Adams DH, Vijayaraghavan K, Holmes C, Atkinson RD, Myers T, Sunderland E. 2012. Mercury in the Gulf of Mexico: Sources to receptors. Environ Res 119:42-52. 40. Benoit JM, Gilmour CC, Mason RP, Riedel GS, Riedel GF. 1998. Behavior of mercury in the Patuxent River estuary. Biogeochemistry 40:249-265. 41. Turekian KK. 1977. The fate of metals in the oceans. Geochim Cosmochim Acta 41:11391144. 42. Lambertsson L, Nilsson M. 2006. Organic material: The primary control on mercury methylation and ambient methyl mercury concentrations in estuarine sediments. Environ Sci Technol 40:1822-1829.

24

43. Chen CY, Dionne M, Mayes BM, Ward DM, Sturup S, Jackson BP. 2009. Mercury bioavailability and bioaccumulation in estuarine food webs in the Gulf of Maine. Environ Sci Technol 43:1804-1810. 44. Stevens DL, Jr., Olsen AR. 2004. Spatially balanced sampling of natural resources. J Am Stat Assoc 99:262-278. 45. Moyle PB. 2002. Inland Fishes of California. University of California Press, Berkeley, CA. 46. Slotton DG, Ayers SM, Reuter JE. 1996. Marsh Creek Watershed 1995 Mercury Assessment Project Final Report to Contra Costa County, Davis, CA 47. SFBRWQCB. 2008. Total Maximum Daily Load for Mercury in Walker Creek Watershed. California Regional Water Quality Control Board San Francisco Bay Region, Oakland, CA. http://www.waterboards.ca.gov/sanfranciscobay/water_issues/programs/TMDLs/walker mercury/Staff_Report_Final.pdf 48. R Core Team. 2013. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org/ 49. Burnham KP, Anderson DR. 2002. Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach, 2nd ed. Springer, Berlin, Germany. 50. Zuur AF, Ieno EN, Walker NJ, Saveliev AA, Smith GM. 2009. Mixed Effects Models and Extensions in Ecology with R. Springer, Berlin, Germany. 51. CEDEN. 2012. California Environmental Data Exchange Network General Information. Fact Sheet. http://www.ceden.org/docs/ceden_intro_010912.pdf 52. McKee LJ, Gilbreath AN, Eads R. 2009. Concentrations and Loads of Trace Contaminants in the Zone 4 Line A Small Tributary, Hayward, California: Water Year 2007. 563. San

25

Francisco Estuary Institute. http://www.sfei.org/sites/default/files/563_Z4LA_Year1final4_web.pdf 53. Ackerman JT, Eagles-Smith CA, Takekawa JY, Demers SA, Adelsbach TL, Bluso JD, Miles AK, Warnock N, Suchanek TH, Schwarzbach SE. 2007. Mercury concentrations and space use of pre-breeding American avocets and black-necked stilts in San Francisco Bay. Sci Total Environ 384:452-466. 54. Melwani AR, Bezalel SN, Hunt JA, Grenier JL, Ichikawa G, Heim W, Bonnema A, Foe C, Slotton DG, Davis JA. 2009. Spatial trends and impairment assessment of mercury in sport fish in the SacramentoSan Joaquin Delta watershed. Environ Pollut 157:3137-3149. 55. Gustin MS, Saito L, Peacock M. 2005. Anthropogenic impacts on mercury concentrations and nitrogen and carbon isotope ratios in fish muscle tissue of the Truckee River watershed, Nevada, USA. Sci Total Environ 347:282-294. 56. Marvin-DiPasquale MC, Agee JL, Bouse RM, Jaffe BE. 2003. Microbial cycling of mercury in contaminated pelagic and wetland sediments of San Pablo Bay, California. Environ Geol 43:260-267. 57. Kamman NC, Chalmers A, Clair TA, Major A, Moore RB, Norton SA, Shanley JB. 2008. Factors influencing mercury in freshwater surface sediments of northeastern North America. Ecotoxicology 14:101-111. 58. Davis JA, Hetzel F, Oram JJ, McKee LJ. 2007. Polychlorinated biphenyls (PCBs) in San Francisco Bay. Environ Res 105:67-86. 59. Greenfield BK, Allen RM. 2013. Polychlorinated biphenyl spatial patterns in San Francisco Bay forage fish. Chemosphere 90:1693-1703.

26

60. Levinton JS, Pochron ST. 2008. Temporal and geographic trends in mercury concentrations in muscle tissue in five species of Hudson River, USA, fishes. Environ Toxicol Chem 27:1691-1697. 61. Munn MD, Short TM. 1997. Spatial heterogeneity of mercury bioaccumulation by walleye in Franklin D. Roosevelt Lake and the Upper Columbia River, Washington. T Am Fish Soc 126:477-487. 62. Fowlie AR, Hodson PV, Hickey MBC. 2008. Spatial and seasonal patterns of mercury concentrations in fish from the St. Lawrence River at Cornwall, Ontario: Implications for monitoring. J Great Lakes Res 34:72-85. 63. Chen CY, Folt CL. 2005. High plankton densities reduce mercury biomagnification. Environ Sci Technol 39:115-121. 64. Essington TE, Houser JN. 2003. The effect of whole-lake nutrient enrichment on mercury concentration in age-1 yellow perch. T Am Fish Soc 132:57-68. 65. Driscoll CT, Chen CY, Hammerschmidt CR, Mason RP, Gilmour CC, Sunderland EM, Greenfield BK, Buckman KL, Lamborg CH. 2012. Nutrient supply and mercury dynamics in marine ecosystems: a conceptual model. Environ Res 119:118-131. 66. Tetra Tech Inc. 2005. Guadalupe River Watershed Mercury TMDL Project Technical Memorandum 5.3.2. Data Collection Report. Prepared for Santa Clara Valley Water District, Lafayette, CA 67. Weech SA, Scheuhammer AM, Elliott JE, Cheng KM. 2004. Mercury in fish from the Pinchi Lake Region, British Columbia, Canada. Environ Pollut 131:275-286. 68. Leady BS, Gottgens JF. 2001. Mercury accumulation in sediment cores and along food chains in two regions of the Brazilian Pantanal. Wetl Ecol Manag 9:349-361.

27

28

Table 1. Results of study model evaluations

Final model Test (Questions) a Distance from Guadalupe b (1) Distance from Guadalupe (1) Embayment (2) Embayment (2) Embayment * Length Interaction (2) Embayment * Distance from Guadalupe Interaction (1, 2) Wetland (3) Wetland (3) Source: POTW * 2008 Interaction (4) Source: Contaminated Sediment (4)
a b

Species Silverside Topsmelt Silverside Topsmelt Topsmelt Topsmelt Silverside Topsmelt Silverside Topsmelt

N 237 239 116 133 133 133 278 269 237 231

Likelihood ratio 44.8 12.4 2.77 6.30 40.8 17.1 0.27 3.08 10.4 4.58

p value < 0.0001 0.0004 0.096 (NS) 0.012 < 0.0001 0.0007 0.61 (NS) 0.079 (NS) 0.0012 0.032

Fixed terms (Effect sign) Distance (-) Distance (-) Distance (-), Length (+) Distance (-), Length (+), 2009 (+), Embayment (+), Embayment * Distance (+), Embayment * Length (+) As above As above Distance (-), Length (+), 2008 (-), 2009 (-), Distance*2009 (+) Distance (-), Length (+) Distance (-), Length (+), 2008 (-), 2009 (-), Distance*2009 (+), POTW (-), POTW*2008 (-) Distance (-), Length (+), 2008 (+), ContaminatedSediment (+), 2008*Distance (+)

Random terms Intercept, Length c Intercept, Length Intercept, Length Intercept Intercept Intercept Intercept, Length Intercept, Length Intercept, Length Intercept

Likelihood ratio tests were employed to answer the 4 study questions, on mixed models which account for additional significant predictor variables. Distance is always centered. c Length = fish total length (centered). NS = not significant. POTW = Publically owned wastewater treatment works.

29

Table 2. Forage fish Hg at mine sites and selected comparison sites Site a Mine influence

Species

Guadalupe River upstream of Alviso Slough New Almaden Mines Mississippi silverside Study average (57 remaining sites) Reference Mississippi silverside Clear Lake [14] Sulphur Bank Mine Mississippi silverside American Canyon Creek Borges Mine Topsmelt Study average (55 remaining sites) Reference Topsmelt Napa River Bella Oaks Mine Prickly sculpin Dry Creek La Joya Mine Prickly sculpin Clear Lake [14] Sulphur Bank Mine Prickly sculpin c Sacramento-San Joaquin Rivers Delta Reference Prickly sculpin Napa River Bella Oaks Mine California roach Dry Creek La Joya Mine California roach Marsh Creek [46] Mt. Diablo Mine California roach c California statewide Reference California roach Guadalupe River upstream of Alviso Slough New Almaden Mines Three-spined stickleback Dry Creek La Joya Mine Three-spined stickleback Zone 4 Line A Industrial watershed Three-spined stickleback Four additional Bay sites Reference Three-spined stickleback Marsh Creek [46] Mt. Diablo Mine Three-spined stickleback Putah Creek, CA Central Valley c Reference Three-spined stickleback Walker Creek [47] Gambonini Mine Three-spined stickleback a Data were from the present study or other referenced studies, where noted. b Whole body samples, collected 1999 to 2004 c www.ceden.us data query, March 23, 2013 d Average of site averages for 8 sites, sampled by D.G. Slotton in 1998 e Average of site averages for 8 sites, sampled 1995 to 1997 f Average of site averages for 26 sites, sampled 1991 to 2001 g Single composite of 36 individuals, collected June, 1992

Hg concentration Mean SD (N) 0.16 0.020 (4) 0.09 0.059 (236) 0.10 0.055 (97) b 0.030 0.006 (4) 0.042 0.020 (236) 0.13 0.009 (4) 0.068 0.012 (4) 0.13 0.044 (5) 0.098 0.062 (15) d 0.14 0.003 (4) 0.061 0.003 (4) 0.27 0.21 (6) 0.084 0.037 (45) e 0.30 0.027 (4) 0.099 0.005 (2) 0.052 0.004 (4) 0.096 0.034 (11) 0.082 0.021 (6) 0.065 0.007 (2) f 0.19 (1) g

30

Figure 1. Site average Hg concentrations in Mississippi silverside.

Figure 2. Site average Hg concentrations in topsmelt.

Figure 3. Hg concentrations in Mississippi silverside, as a function of distance from Guadalupe River, sampling year, and site category. Each point represents a composite sample, and lines represent linear model fits to the associated data type for the given year. POTW sites (i.e., draining wastewater treatment plants) (); all other sites (). Note log scale y axis.

Figure 4. Hg concentrations in topsmelt, as a function of distance from Guadalupe River and embayment category. Each point represents a composite sample, and lines represent linear model fits to the associated data type. Embayment sites (); open sites (). Note log scale y axis.

Figure 5. Hg concentrations in topsmelt, as a function of distance from Guadalupe River and site category. Each point represents a composite sample, and lines represent linear model fits to the associated data type. Contaminated sediment sites (); all other sites (). Note log scale y axis.

31

! (
! ( ! ( ! ( ! ( ! (
( !

! ( ! (
( ! ( ! ( !

! (
( ! ( ! ( ! ( Suisun ! Bay ( ! ( ! ! ( ( !

( !

! ( ! (

San Pablo Bay

! (

( ! ( !

( ! ( ! ( !

( !

! (! ! ( ( Central Bay ! (

Silverside
Hg (g/g)
( !

< 0.060 0.060 0.106 0.106 0.152 0.152 0.197 > 0.197

! (

Pacic Ocean
! ( ! (

! (
( ! ! (

! (

! (

! (
! (
( ! ( !

! (
C

South Bay
! ( ! (

! (

! (
o

Pacic Ocean

a
l

! (

! (! ( ! ( ! ( (! ! ( ( (! ! (!
Lower South ! ( Bay

10

20 Km

! (! ! ( (

i
f

! (
! ( ! ( ! (
! (

San Pablo Bay

! ( ! (

Suisun Bay
! ( ! ( ! ( ! (

! (

! (

! (

! ( ! Central( ( Bay !
! (
! ( ! ( ! ( !( ( ! ! (

! (

! ( ! (

Topsmelt
Hg (g/g)
( !

! ( ! (

< 0.031 0.031 0.045 0.045 0.058 0.058 0.072 > 0.072

Pacic Ocean

! ( ( !

! ( ! (
! (

! (

! (

! (

! ( ! ( ! ( ! ( ! ( ! ( ! ( ! (

South Bay

! (
! (
! (

Pacic Ocean

! ( ! (
! (

a
l

! (

! ! South ( ( Bay ! (
a

!( ( ! ! ( Lower

! ! ((

10

20 km

i
f

o
r

2008
qq q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q qq qq q

Mercury (g/g wet wt) 0.02 0.05 0.20

POTW q Other 0
q q q q q q q qq q q q q q q q q q q q q

50 2009
q q q

100

Mercury (g/g wet wt) 0.02 0.05 0.20

q q q qq q q q

q q q q q q q q q qq

q q q qq qq q q q q q q q q q q q

POTW q Other 0

50 2010

100

Mercury (g/g wet wt) 0.02 0.05 0.20

q q q q q q q

q qq q q qq q q q q q q q q

q q q q

q q q q q q q q q q q q q q q q q q q q q q qq q q q q q q q q q q q q q q

POTW q Other 0 50 100 Distance from Guadalupe River (km)

0.06

Mercury (g/g wet wt) 0.02 0.03 0.04

q q q q q q q q q q q q q q q q

q q q q q q q q q q q q q q q q q

q q q q q q q q q q q q

Open q Embayment

20 40 60 80 100 120 Distance from Guadalupe River (km)

0.10
q q q q q q q qq qq q q q q q qq q q q q q q qq q q q qq q q q qq q qq q q q qq q q qq q q q q q q q q q q q q q

Mercury (g/g wet wt) 0.02 0.04 0.06

q q q q q q q q q q q q q q q q q qq q q q q q q q q q q q q q q q q q q q q q q q q q q qq q qq q q q q q q qq q q q q q q qq q q q q q Contaminated sediment q q Other

q q q q q q

q q q q qq q q q q q q q q q q

20 40 60 80 100 120 Distance from Guadalupe River (km)

Supporting document for Predictors of mercury spatial patterns in San Francisco Bay forage fish Corresponding author: Ben K. Greenfield Environmental Health Sciences Division School of Public Health University of California-Berkeley, 50 University Hall #7360 Berkeley, CA 94720-7360 Tel: 1 510 507 2365 E-mail: bengreenfield@berkeley.edu

14 pages total Section I: Supplemental Text. p. 2 Study hypotheses Study parameters Site selection description Data analysis methods Removal of outliers Section II: Supplemental Tables. p. 5 Section III: Supplemental Figures. p. 8 Section IV: Description of Supplemental Data. p. 10 Section V: References. p. 13

Section I: Supplemental Text Study hypotheses The study design and site selection were intended to answer the 4 questions listed in the Introduction. Based on prior Bay studies [1-5], input from local natural resource managers, and current conceptual models regarding MeHg cycling and bioaccumulation in estuaries, we developed the following 4 hypotheses regarding factors that influence small fish Hg concentrations: 1. Mercury concentrations will increase with proximity to the Guadalupe River, in the Lower South Bay. 2. Concentrations will be higher in fully or partially enclosed areas, defined to include natural or man-made coves or channels, or estuarine creeks draining into the Bay (defined in this paper as embayments). These areas would tend to have low hydraulic mixing of subtidal water (i.e., locations with low water turnover rate), resulting in higher MeHg production. 3. Concentrations will be positively correlated with nearby wetland abundance. 4. Concentrations will be higher near mercury source areas, including urban or industrial watershed drainages, wastewater treatment plants (i.e., publicly operated treatment works, or POTWs), areas with historically Hg or MeHg contaminated sediments, and mine drainages. Study parameters Average total body length (mm) was based on field measurements of all individuals in a composite sample. Length is a widely reported correlate of fish Hg, and length correction is needed [6, 7], including for small fish evaluated in the Bay [2, 8]. Sampling was performed in 3 years: 2008, 2009, and 2010, which are treated as categorical variables (with 2010 as the base condition). There are also 4 categorical variables indicating different kinds of Hg source sites (Table S1), as well as randomly chosen sites that represent background conditions (statistically treated as the base condition). Site selection description Creeks draining Hg mines (N = 4) were chosen based on the priority scheme of Abu-Saba [9]. Sites were included based on connectivity to the Bay, evidence of mine waste discharging into State waters, and risk to fishery resources or other sensitive habitat areas. Three creeks were chosen based on drainage from mines meeting the inclusion criteria: Napa River (drainage from La Joya and Bella Oaks mines), American Canyon Creek (Borges Mine), and Guadalupe River (New Almaden Mining District). Due to the small number of candidate sites, all Hg mine drainage sites were sampled. To further evaluate the potential mine signal from La Joya mine, a freshwater site adjacent to and draining the mine (Dry Creek) was added and sampled in 2009. POTW sites (N = 7 sites) were selected from among those Bay POTWs that are identified by the San Francisco Bay Regional Water Quality Control Board (Regional Board) as shallow-water discharge POTWs [10]. The selection of shallow water discharges was based on the expectation of greater potential impact of POTW discharge to the nearshore area biogeochemistry in shallow-water environments than deep-water or offshore environments. That is, shallow-water POTW sites had a relatively high water volume ratio of discharge water to natural Bay water, potentially resulting in discharge water impacting Hg biogeochemistry. The original pool of
2

sites was further reduced to only include sites having summer discharge because sampling occurred in the fall. The final POTW sites were City of American Canyon Wastewater Treatment Plant, Fairfield-Suisun, Hayward Pond 3B, Mountain View Sanitary District Peyton Slough, City of Palo Alto, City of Sunnyvale, and Sonoma Valley County Sanitary District Schell Creek. All 7 POTW sites were sampled. Candidate sites with relatively elevated sediment total Hg or MeHg (N = 35) were selected based on 2 prior sediment surveys from the Bay Protection and Toxic Cleanup Program [11] and a study funded by the Regional Board and the California Bay-Delta Authority program [3]. Both studies targeted areas known to be currently or historically industrial or otherwise suspected of having high contamination. Sites were included when sediment THg concentrations were greater than 700 ng g-1 or MeHg concentrations were greater than 2 ng g-1. In the GRTS sample draw, 15 sites were sampled from this category. Candidate sites draining industrial and urban watersheds (N = 21) were selected based on a combination of 4 attributes in the watersheds: the documented presence of historic mercury spills, density of historic industrial sites, density of railway lines, and density of car recyclers. Toxic Hg spills were identified using database queries from the CA Department of Toxic Substances Control, USEPA Superfund, and Toxic Release Inventory. Railway line density was considered as an indicator of historic industry, while auto recycler density indicated current industry and potentially an additional Hg source. All attributes were characterized using GIS layers developed from Lower South Bay to southern San Pablo Bay [12]. To increase geographical coverage, 3 sites were added from central San Pablo Bay and Suisun Bay (Kirker Creek, Corte Madera Creek, and Alhambra Creek). In the GRTS sample draw, 13 sites were sampled from this category. In 2008, 12 sites adjacent to fringing wetlands were sampled to aid in evaluating the potential impact of wetland proximity on biosentinel Hg (Hypothesis 4). The wetland sites ranged in anticipated frequency of wetting and drying; wetlands with a high expected wetting and drying frequency were hypothesized to exhibit higher MeHg production and consequently higher Hg in biosentinels [13]. Results from 2008 indicated no apparent correspondence with fringing wetland abundance or type, and GIS indicated that a range of proximity to wetlands was achieved using the random samples in the GRTS design. Therefore, in 2009 and 2010 wetland site targeting was discontinued to increase sample sizes in the other strata. The final study analysis compared fish Hg based on percent wetlands at the site. The wetland sites were included in analysis of proximity to wetlands vs. small fish Hg (Question 3). To avoid confounding site categorization, the 2008 wetland sites were excluded from the comparison of embayment versus open sites (Question 2) and the comparison of random versus source sites (Question 4). Data analysis methods We employed the 10 step approach to nested data recommended by Zurr et al. [14] (Sections 4.2.3 and 5.10), to assess the need for mixed models. In all cases, model fit and residual behavior significantly improved when including a random term for sampling site. Therefore, all models included a random intercept effect for site, and a random slope (i.e., length) effect for site if warranted. The initial model fixed structure always included year terms (2008 and 2009), fish length, distance from the Guadalupe River, and the effect under consideration (embayment category, site type, or surrounding wetland), and one way interaction terms between site effects and the other model terms. For example, the initial model to evaluate embayment effect was: Log(Hg) = Year2008 + Year2009 + FishLength + DistanceGuadalupe + Embayment +
3

Year2008*Embayment + Year2009*Embayment + FishLength*Embayment + DistanceGuadalupe*Embayment. The embayment category statistical evaluation (Question 2) was performed on random sites only, to avoid confounding site type versus embayment category. The silverside site type evaluation (Question 4) compared contaminated sediment, industrial watershed sites, POTW, and random (open and embayment) sites. The topsmelt site type evaluation did not include POTW sites because topsmelt were only collected at 1 POTW site. Removal of outliers Four outlier samples were removed from statistical analysis because their inclusion would have violated linear model assumptions. For silversides, the Cooley Landing site (N = 3 samples, located west of Lower South Bay) had extreme variance heterogeneity due to only having 1 individual per sample, which caused violation of normality in residuals. For topsmelt, 1 of the composite samples collected in Alviso Slough (south of Lower South Bay) in 2009 was an outlier (standardized residual = 4.68), with an Hg concentration of 0.235 g/g, versus a range of 0.015 to 0.114 g/g for the remaining 239 topsmelt samples. Other than Figure 1, all results are reported excluding these outlier samples. However, when the analysis was performed including these samples, results (statistical significance and effect size and direction) were unchanged.

Section II: Supplemental Tables Table S1. Parameters examined in study. ID in Models Description Hg Hg concentration in composite fish sample, wet weight. Obtained from single species composite samples, log 10 transformed for model inclusion TLengthCen Fish total body length, averaged across composite sample, centered before model inclusion Station Site Location ID. N = 99 stations; but only a subset for each species and analysis. DistGuadCen Distance from Guadalupe River of collection site, centered before model inclusion Y2008 Indicates samples collected in 2008 Y2009 Indicates samples collected in 2009. Default (i.e., baseline) year is 2010 WetlandAbund Surrounding wetland area, based on a 500 m buffer Embayment Indicates sample collected from enclosed embayment, rather than open Bay shoreline (Figure S1) SourcePOTW Publicly operated treatment works (POTW) at station. This is influenced by wastewater treatment plant discharge. This and remaining source station categories are compared to randomly selected stations. SourceContamSed Historic Hg contaminated sediment at station SourceWatershed Station adjacent to industrial or urban watershed hypothesized to be high in Hg SourceMine Station adjacent to watershed containing historic Hg mine a. All categorical variables are nominal, rather than ordinal Type Numeric Units g g-1

Numeric Categoricala Numeric Categorical Categorical Numeric Categorical

mm

km 1=Yes 1=Yes % 1=Yes

Categorical

1=Yes

Categorical Categorical Categorical

1=Yes 1=Yes 1=Yes

Model structures Table S2. Silverside final model for embayment effects (N = 116). For this and remaining tables, likelihood ratio test is performed, and the difference in AIC (AIC) is determined, between a model containing all parameters listed and a model with the current parameter removed. NA: since embayment was not significant, it is not included in the final model. Fixed effects Value SE Likelihood ratio p value AIC (Intercept) -1.20219 0.01802 TLengthCen 0.00516 0.00190 39.5 <0.0001 37.5 DistGuadCen -0.00345 0.00041 7.00 0.0082 5.0 Embayment NA NA 2.77 0.096 0.8 (NS) Random effects: Formula: ~1 + TLengthCen | Station

Table S3. Topsmelt final model for embayment effects (N = 133) Fixed effects Value SE Likelihood ratio (Intercept) -1.47553 0.02565 TLengthCen 0.00238 0.00082 8.50 DistGuadCen -0.00229 0.00068 11.1 Y2009 0.06071 0.03098 4.18 Embayment 0.09145 0.03752 6.30 DistGuadCen:Embayment 0.00298 0.00119 17.1 TLengthCen:Embayment 0.00308 0.00128 40.8 Random effects: Formula: ~1 | Station

p value 0.0035 0.0008 0.041 0.012 0.0007 <0.0001

AIC 6.5 9.1 2.2 4.3 11.1 34.8

Table S4. Silverside final model for wetland effects (N = 278). NA = wetland was not significant, and was therefore not included in the final model. Fixed effects Value SE Likelihood ratio p value (Intercept) -1.07663 0.02652 TLengthCen 0.00481 0.00103 18.7 <0.0001 Y2008 -0.10544 0.02671 14.6 <0.0001 Y2009 -0.03914 0.03234 1.59 0.21 DistGuadCen -0.00405 0.00046 46.9 <0.0001 Y2009:DistGuadCen 0.00169 0.00065 5.96 0.015 WetlandAbund NA NA 0.27 0.61 (NS) Random effects: Formula: ~1 + TLengthCen | Station

AIC 16.7 12.6 -0.4 44.9 4.0 -1.7

Table S5. Topsmelt final model for wetland effects (N = 269). NA = wetland was not significant, and was therefore not included in the final model. Fixed effects Value SE Likelihood p value AIC ratio (Intercept) -1.39659 0.01656 TLengthCen 0.00379 0.00051 38.7 <0.0001 36.7 DistGuadCen -0.00229 0.00049 20.9 <0.0001 18.9 WetlandAbund NA NA 3.08 0.079 1.1 (NS) Random effects: Formula: ~1 + TLengthCen | Station Table S6. Silverside final model for site effects (N = 237) Fixed effects Value SE Likelihood ratio (Intercept) -1.10565 0.02760 TLengthCen 0.00463 0.00122 12.7 Y2008 -0.06108 0.03153 3.44 Y2009 -0.00657 0.03191 0.04 DistGuadCen -0.00414 0.00048 52.7 SourcePOTW -0.09216 0.06358 2.15 Y2008:SourcePOTW -0.19625 0.06058 10.4 Y2009:DistGuadCen 0.00189 0.00063 60.8 Random effects: Formula: ~1 + TLengthCen | Station Table S7. Topsmelt final model for site effects (N = 231) Fixed effects Value SE Likelihood ratio (Intercept) -1.43702 0.01879 TLengthCen 0.00332 0.00047 45.4 Y2008 0.03358 0.02967 1.31 DistGuadCen -0.00228 0.00052 17.8 SourceContamSed 0.07831 0.03727 4.58 Y2008:DistGuadCen 0.00198 0.00087 5.24 Random effects: Formula: ~1 | Station

p value

AIC

0.0004 0.064 0.84 <0.0001 0.14 0.0012 <0.0001

10.7 1.4 -2.0 50.7 0.2 8.4 54.8

p value

AIC

<0.0001 0.25 <0.0001 0.032 0.022

43.4 -0.7 15.8 2.6 3.2

Section III: Supplemental Figures

Figure S1. Study sample locations.

Figure S2. Final stratification of the Bay shoreline for the random sample draw. The red line represents the embayment stratum. Blue areas not parallel to a red line represent the open stratum.

Section IV: Description of Supplemental Data File All forage fish Hg data collected in San Francisco Bay from 2008 to 2011 are provided as a comma separated (CSV) file (filename: SFForageFishHg.csv). Comprising 1260 composite samples, these data are more extensive than the samples analyzed for the present study. They include Bay forage fish Hg data analyzed for temporal patterns and published elsewhere [8], as well as samples collected and analyzed for Hg but falling outside the scope of both studies. Data on fish Hg and other contaminants are also presently available via the California Environmental Data Exchange Network (www.ceden.us) [15] This file contains 30 data fields, including Hg concentrations, site and sample descriptive information, and ancillary information. Each row corresponds to a separate composite fish sample analyzed for Hg: Sample ID Laboratory identifier indicating the composite fish sample analyzed Site ID Unique identifier for each collection site Collection Code Unique identifier for each collection event (i.e., each site and sample date combination) Site Description Descriptive name for each collection site Site Type Indicates the site category each sample was included in. Table S8 (below) lists all possible categories Date Collection date Latitude In degrees and decimal degrees Longitude In degrees and decimal degrees Species Common name of the sampled species. Table S9 lists corresponding scientific names, as well as number of samples collected per species HgDw Mercury concentration (g g-1 dry weight) Moisture Tissue proportion moisture HgWw Mercury concentration (g g-1 wet weight) N Number of individuals in the composite sample Total Length Average total length of individuals in the composite sample (mm) Distance Guadalupe Distance from the Guadalupe River, following along the deep Bay channel (further described in Methods) WetlandAbund Percent surrounding wetland area within a 500 m buffer of collection location The remaining variables are binary categorical variables (1 = yes; 0 = no), indicating which study model evaluation (Table 1) or sampling strata a sample was included in.

10

Question1SpatialTrend Sample included in study Question 1: What are the spatial trends in forage fish Hg? Question2Embayment Sample included in study Question 2: Are Hg concentrations elevated in embayments relative to open water sites? Question3Wetlands Sample included in study Question 3: Does extent of fringing wetland habitat correlate with Hg concentrations? Question4SiteType Sample included in study Question 4: Are Hg concentrations elevated at potential source sites, relative to randomly selected sites? LongTerm Sample included in prior analysis of long term Hg trends [8] Source Sample is in 1 of the 4 source site categories Random Sample is in the random strata (either embayment or open) sampled from entire Bay shoreline SourcePOTW Wastewater treatment plant source site sample SourceMine Historic Hg mine source site sample SourceContamSed Hg or MeHg contaminated sediment source site sample SourceWatershed Urban or industrialized watershed source site sample Embayment Site falls within the embayment stratum (1 = embayment; 0 = open; Figure S2) OffSeason Sample collected outside the Fall season sampling window required for study inclusion. The study sampling window was August 27 to November 30 Wetlands Sample targeted to increase wetland coverage Table S8. Site type descriptions. Site type a Embayment Open Water Source Contaminated Sediment Source Wastewater Treatment Plant (POTW) Source Industrial Watershed Source Hg Mine Wetlands

Description Random Bay shoreline site; embayment stratum (Figure S2) Random Bay shoreline site; open water stratum (Figure S2) Source site with elevated sediment THg or MeHg

Source site draining wastewater treatment plant Source site draining urban and industrial watershed Source site draining historic Hg mine watershed Site adjacent to nearshore wetland, targeted to increase wetland coverage, sampled in 2008 Long Term Monitoring Site monitored for annual variation from 2005 to 2010, analyzed elsewhere [8] Seasonal/Long Term Site monitored for seasonal and annual variation, analyzed Monitoring elsewhere [8] Not in Hg Sampling Design Site targeted for PCB study [16], analyzed for Hg but not included in Hg study analyses a. Some sites are in more than 1 category
11

Table S9. Common and scientific names, and total number composite samples collected for each species. Number samples Common Name Scientific Name 7 American shad Alosa sapidissima 105 Arrow goby Clevelandia ios 8 California roach Hesperoleucus symmetricus 6 Diamond turbot Hypsopsetta guttulata 3 Golden shiner Notemigonus crysoleucas 1 Longjaw mudsucker Gillichthys mirabilis 357 Mississippi silverside Menidia audens 24 Northern anchovy Engraulis mordax 8 Prickly sculpin Cottus asper 27 Rainwater killifish Lucania parva 3 Shimofuri goby Tridentiger bifasciatus 1 Shiner perch Cymatogaster aggregata 33 Staghorn sculpin Leptocottus armatus 24 Striped Bass Morone saxatilis 2 Threadfin shad Dorosoma petenense 24 Three-spined stickleback Gasterosteus aculeatus 610 Topsmelt Atherinops affinis 7 Western mosquitofish Gambusia affinis 10 Yellowfin goby Acanthogobius flavimanus

12

Section V: References 1. Conaway CH, Squire S, Mason RP, Flegal AR. 2003. Mercury speciation in the San Francisco Bay estuary. Mar Chem 80:199-225. 2. Greenfield BK, Jahn A. 2010. Mercury in San Francisco Bay forage fish. Environ Pollut 158:2716-2724. 3. Heim WA, Coale KH, Stephenson M, Choe K-Y, Gill GA, Foe C. 2007. Spatial and habitatbased variations in total and methyl mercury concentrations in surficial sediments in the San Francisco Bay-Delta. Environ Sci Technol 41:3501-3507. 4. Marvin-DiPasquale MC, Agee JL, Bouse RM, Jaffe BE. 2003. Microbial cycling of mercury in contaminated pelagic and wetland sediments of San Pablo Bay, California. Environ Geol 43:260-267. 5. Conaway CH, Black FJ, Grieb TM, Roy S, Flegal AR. 2008. Mercury in the San Francisco Estuary: A review. Rev Environ Contam Toxicol 194:29-54. 6. Tremblay G, Legendre P, Doyon J-F, Verdon R, Schetagne R. 1998. The use of polynomial regression analysis with indicator variables for interpretation of mercury in fish data. Biogeochemistry 40:189-201. 7. Wiener JG, Bodaly RA, Brown SS, Lucotte M, Newman MC, Porcella DB, Reash RJ, Swain EB. 2007. Monitoring and evaluating trends in methylmercury accumulation in aquatic biota. In Harris R, Krabbenhoft DP, Mason R, Murray MW, Reash R, Saltman T, eds, Ecosystem Responses to Mercury Contamination: Indicators of Change. CRC Press, Pensacola, FL, pp 87-122. 8. Greenfield BK, Melwani AR, Allen RM, Slotton DG, Ayers SM, Harrold KH, Ridolfi K, Jahn A, Grenier JL, Sandheinrich MB. 2013. Seasonal and annual trends in forage fish mercury concentrations, San Francisco Bay. Sci Total Environ 444:591-601. 9. Abu-Saba K. 2003. Managing inactive mercury mine sites in the San Francisco Bay Region. Applied Marine Sciences, Inc. http://www.bacwa.org/LinkClick.aspx?fileticket=RUiczIW5flU%3D&tabid=126&mid=5 72 10. California Regional Water Quality Control Board San Francisco Bay Region. 2011. San Francisco Bay Basin (Region 2) Water Quality Control Plan (Basin Plan). Web Document. Accessed October 11, 2012. http://www.waterboards.ca.gov/sanfranciscobay/basin_planning.shtml. http://www.wate rboards.ca.gov/sanfranciscobay/basin_planning.shtml 11. Hunt JW, Anderson BS, Phillips BM, Newman J, Tjeerdema RS, Taberski K, Wilson CJ, Stephenson M, Puckett HM, Fairey R, Oakden J. 1998. Sediment quality and biological effects in San Francisco Bay: Bay Protection and Toxic Cleanup Program Final Technical Report. San Francisco Bay Regional Water Quality Control Board, Oakland, CA. http://www.swrcb.ca.gov/water_issues/programs/bptcp/docs/reg2report.pdf 12. SFEI. 2010. A BMP tool box for reducing polychlorinated biphenyls (PCBs) and mercury (Hg) in municipal stormwater. San Francisco Estuary Institute, Oakland, CA. http://www.sfei.org/documents/3813 13. Snodgrass JW, Jagoe CH, Bryan AL, Brant HA, Burger J. 2000. Effects of trophic status and wetland morphology, hydroperiod, and water chemistry on mercury concentrations in fish. Can J Fish Aquat Sci 57:171-180.

13

14. Zuur AF, Ieno EN, Walker NJ, Saveliev AA, Smith GM. 2009. Mixed Effects Models and Extensions in Ecology with R. Springer, Berlin, Germany. 15. CEDEN. 2012. California Environmental Data Exchange Network General Information. Fact Sheet. http://www.ceden.org/docs/ceden_intro_010912.pdf 16. Greenfield BK, Allen RM. 2013. Polychlorinated biphenyl spatial patterns in San Francisco Bay forage fish. Chemosphere 90:1693-1703.

14