Predictive Modeling of Selenium Accumulation in Brine Shrimp in Saline Environments

Transcription

1 Integrated Environmental Assessment and Management Volume 7, Number 3 pp ß 2011 SETAC Predictive Modeling of Selenium Accumulation in Brine Shrimp in Saline Environments Earl R. Byron,*y Harry M. Ohlendorf,y Aaron Redman,z William J. Adams, Brad Marden,k Martin Grosell,# and Marjorie L. Brooksyy ych2m Hill, 2485 Natomas Park Drive, Suite 600, Sacramento, California 95833, USA zhydroqual, Providence, Utah, USA Rio Tinto, Murray, Utah, USA kparliament Fisheries, Ogden, Utah, USA #University of Miami, Miami, Florida, USA yydepartment of Zoology, Southern Illinois University, Carbondale, Illinois, USA (Submitted 6 September 2010; Returned for Revision 11 October 2010; Accepted 5 January 2011) Case Study ABSTRACT Great Salt Lake, Utah, is a large, terminal, hypersaline lake consisting of a northern more saline arm and a southern arm that is less saline. The southern arm supports a seasonally abundant fauna of low diversity consisting of brine shrimp (Artemia franciscana), 7 species of brine flies, and multiple species of algae. Although fish cannot survive in the main body of the lake, the lake is highly productive, and brine shrimp and brine fly populations support large numbers of migratory waterfowl and shorebirds, as well as resident waterfowl, shorebirds, and gulls. Selenium and other trace elements, metals, and nutrients are contaminants of concern for the lake because of their concentrations in municipal and industrial outfalls and runoff from local agriculture and the large urban area of Salt Lake City. As a consequence, the State of Utah recently recommended water quality standards for Se for the southern arm of Great Salt Lake based on exposure and risk to birds. The tissue-based recommendations (as measured in bird eggs) were based on the understanding that Se toxicity is predominately expressed through dietary exposure, and that the breeding shorebirds, waterfowl, and gulls of the lake are the receptors of most concern. The bird egg based recommended standards for Se require a model to link bird egg Se concentrations to their dietary concentrations and water column values. This study analyzes available brine shrimp tissue Se data from a variety of sources, along with waterborne and water particulate (potential brine shrimp diet) Se concentrations, in an attempt to develop a model to predict brine shrimp Se concentrations from the Se concentrations in surrounding water. The model can serve as a tool for linking the tissue-based water quality standards of a key dietary item to waterborne concentrations. The results were compared to other laboratory and field-based models to predict brine shrimp tissue Se concentrations from ambient water and their diet. No significant relationships were found between brine shrimp and their dietary Se, as measured by seston concentrations. The final linear and piecewise regression models showed significant positive relationships between waterborne and brine shrimp tissue Se concentrations but with a very weak predictive ability for waterborne concentrations <10 mg/l. Integr Environ Assess Manag 2011;7: ß 2011 SETAC Keywords: Selenium Brine shrimp Bioaccumulation Great Salt Lake INTRODUCTION Great Salt Lake, Utah, is a large, terminal, hypersaline lake consisting of a northern more saline arm and a southern arm that is less saline (although currently between 85 and 165 g/l total dissolved solids) because it receives most of the freshwater drainage entering the lake (Gwynn 2002; Brix et al. 2004). The 2 arms are incompletely separated by a Southern Pacific Railroad causeway and embedded culverts. The southern arm supports a seasonally abundant community of low diversity consisting of brine shrimp (Artemia franciscana), 7 species of brine flies (Ephydra sp.), and multiple species of algae dominated by 2 species of the green alga, Dunaliella (Marden 2008). Although fish can not survive in the main body of the lake, the lake is highly productive and brine shrimp and brine fly populations support large numbers of migratory waterfowl and shorebirds, as well as resident * To whom correspondence may be addressed: ebyron@ch2m.com Published online 31 January 2011 in Wiley Online Library (wileyonlinelibrary.com). DOI: /ieam.179 waterfowl, shorebirds, and gulls. In addition, the harvest of brine shrimp cysts (encysted brine shrimp embryos) for aquaculture is a multimillion dollar industry (Kuehn 2002; Marden 2008). Selenium and other trace elements, metals, and nutrients are contaminants of concern for Great Salt Lake because of inputs from municipal and industrial point sources and runoff from local agriculture and the large urban area of Salt Lake City, which borders the southeastern end of the lake. Generic, national recommended water quality criteria (USEPA 2009) have not been applied to the hypersaline waters of the lake, although recommendations for the possible development of site-specific standards for Se were proposed in the past (Brix et al. 2004). Driven by ongoing concerns related to increasing discharges of Se to the lake from mining, agriculture, and urban sources, the State of Utah recently recommended water quality standards for Se for the southern arm of Great Salt Lake (Utah 2010) based on exposure and risk to birds. The tissue-based recommendations (as measured in bird eggs) are based on the understanding that Se toxicity occurs predominately through dietary exposure, and that the breeding shorebirds, waterfowl,

2 Selenium in Brine Shrimp Integr Environ Assess Manag 7, and gulls of the lake are the receptors of most concern and can be assessed for direct exposure and risk through measurement of Se in their eggs and their diet. Concerns also focused on the potential effects on migrant avian species (e.g., eared grebes [Podiceps nigricollis]) and overwintering waterfowl (e.g., common goldeneye [Bucephala clangula]). However, no clear threshold has been defined for those species. The final, proposed Se standard for the lake that would constitute a state of water quality impairment is a geometric mean bird egg concentration of 12.5 mg Se/kg dry weight (dw), averaged over the nesting season from at least 5 eggs. Lower egg concentrations (between 5 and 12.5 mg Se/kg) were proposed as triggers to initiate further studies, an antidegradation review, and a review of water quality conditions (Utah 2010). The bird egg based recommended standards for Se require a model to link egg Se concentrations to their dietary concentrations and water column values. Such a link facilitates the estimation of allowable loading from discharges and the creation of scientifically defensible water discharge based on concentration. Brine shrimp and brine flies are the major dietary items of birds nesting at Great Salt Lake. A model to predict bird egg Se concentrations from prey items and water was developed as part of the creation of tissuebased standards, but researchers recognized that some uncertainties warranted further evaluation. This article presents a refined model to predict Se concentrations in brine shrimp from those in ambient waters, thereby providing a statistical tool to address the critical link between key dietary items and waterborne Se concentrations. The results are compared to existing models (Table 1) that predicted Se concentrations in brine shrimp tissue from ambient water and from their diet (seston). The creation of a new Se model was justified based on the availability of a new, comprehensive database of brine shrimp bioaccumulation and the need for a logistically simple means of assessing potential threats to waterfowl at all locations and life stages. MATERIALS AND METHODS Selenium concentrations in colocated samples collected between 1988 and 2009 and analyzed for Se content of brine Table 1. Previous models used to estimate brine shrimp tissue concentrations of Se based on water or dietary concentrations Source Brooks (2007) Grosell (2008) Marden (2008) Skorupa and Ohlendorf (1991) Characteristics Laboratory estimates of uptake of Se from food and water for Great Salt Lake Laboratory estimates of uptake of Se from food and water for Great Salt Lake Partitioning coefficients and trophic transfer factors computed from field-collected seston, water, and brine shrimp data for State of Utah, Great Salt Lake model Central Valley dataset of brine shrimp and cocollected water Se concentrations as a simple regression relationship on log-transformed data shrimp tissue, seston, and water (as total or dissolved fractions) were combined in a large database from sources across the western United States. Brine shrimp tissue concentrations spanned more than 3 orders of magnitude. All brine shrimp samples were whole-body composites of many individuals as adults or near-adult life stages. Basic data summaries by source are given in Table 2. The data set was examined for statistically significant relationships between water and seston, between seston (as a surrogate for dietary Se) and brine shrimp, and between water and brine shrimp. Values for seston were corrected for salt content on the filters used to separate seston from water. The ultimate goal was to predict brine shrimp tissue Se on the basis of waterborne concentrations; surrogates for brine shrimp diet (seston, chlorophyll) were examined as possible modeling covariates. To facilitate the construction of the model, an all-water variable was produced that consisted of total Se when both total and dissolved fractions were available or by using the dissolved fraction if total values were missing. This is a reasonable assumption for Se, because most of the waterborne concentration was in the dissolved fraction for Table 2. Sources and basic statistics for data set; brine shrimp, water, and seston Se concentrations (brine shrimp and seston as mg Se/kg dw, water as mg Se/L) Location Parameter Geometric mean, n Range Source Great Salt Lake, Utah Brine shrimp 1.77, ,2,3 Water (total) 0.748, ,2,3 Water (dissolved) 0.977, ,2,3 Seston 0.941, Central Valley ponds, California Brine shrimp 9.26, Water (total) 26.1, Water (dissolved) 17.4, Hailstone NWR, Montana Brine shrimp 21.9, Water (dissolved) 17.9, Sources: 1. Marden (2008); 2. Brix et al. (2004); 3. Brix (2006); 4. Skorupa and Ohlendorf (1991); Fan et al. (2002); Shelton et al. (1990); 5. Nelson and Reiten (2009)

3 480 Integr Environ Assess Manag 7, 2011 ER Byron et al. samples where both fractions were available (average of 89%, ranging from 82 to 100% in our dataset). Most of the data (70%) were available only as total Se and the creation of an all-water variable allowed the maximized comparisons of waterborne concentrations to tissue and seston (Table 2). No separate dataset for brine shrimp Se was readily available to test a bioaccumulation model; therefore, the full dataset was split randomly so that 2/3 of the data were used in model construction (referred to as a test set ) and 1/3 were held back to validate the applicability of the model ( validation set ). This strategy allowed us to compare the model s performance to that of other models. The results of empirical modeling from the test set were compared to water versus brine shrimp relationships as described by other models from the literature, both mechanistic and empirically derived. Grosell s controlled laboratory study used a commercially prepared salt matrix to evaluate the influence of salinity and variable waterborne and dietary concentrations of Se on Se bioaccumulation by Artemia (Grosell 2008). The resulting model predicted brine shrimp tissue Se as affected by variable environmental concentrations. Similarly, Brooks and others measured brine shrimp uptake from varying concentrations of aqueous and dietary Se under controlled laboratory conditions using Great Salt Lake water as the background matrix (Brooks et al. 2007). Both laboratory studies used algal cultures (Dunaliella viridis) with experimentally determined Se concentrations. Statistical analyses of the data were performed using the Statview (SAS 1998) program for transformations, linear regression, and analysis of variance and SPSS (IBM 2010) for the piecewise regression. RESULTS The brine shrimp tissue spanned a large range in Se concentrations, from the relatively lower concentrations of Great Salt Lake to the higher, average concentrations and maximum values of Hailstone National Wildlife Refuge (NWR), Montana, and the Central Valley, California, evaporation ponds (Table 2). The range of Se concentrations in water (dissolved and total), seston, and brine shrimp tissue provided some useful information on associations among variables but did not provide an empirically derived model to match literature expectations. Specifically, the Se concentrations in cocollected seston and brine shrimp revealed no clear associations that could be used as a model for the uptake of Se from diet. This was true for chlorophyll-normalized seston Se concentrations, as well (Table 3). The weak positive associations between total Se in water and seston Se and between seston and brine shrimp Se are undoubtedly related to the inclusion of nonfood seston particles as part of any total Se measurements. The strongest association between variables occurred between waterborne Se (dissolved or total fractions) and brine shrimp tissue Se, inferring both dietary and waterborne exposure (Table 3). A predictive model for brine shrimp tissue Se from water is shown in Figure 1, using the test set. Although the linear model shows a strong positive relationship between water column concentrations and those of cocollected brine shrimp, there is high variability for any given water column concentration (1 or 2 orders of magnitude in tissue concentrations). In particular, note the high variability in brine shrimp tissue Se concentrations for water concentrations <10 mg Se/L. The test and validation data sets possessed nearly identical ranges and central tendencies and were statistically similar ( p < 0.05, unpaired t-test). Normality was improved by log-transformation of the variables. The resulting linear model shown in Figure 1 is: Selenium in brine shrimp ðmg=kg dwþ ¼ 10^ð0:48ðLog mgse=l waterþ þ 0:354Þ A closer examination of the tissue versus water relationship limited to the lower range of water concentrations (<10 mg Se/L) reveals a possible hockey stick, piecewise regression relationship, with a lower range of concentrations averaging 1.72 mg Se/kg dw (0 slope), changing to a significant positivesloped regression above the inflection point where water column concentrations exceed 0.87 mg Se/L (Figure 2). The piecewise regression model shown in Figure 2 is: Selenium in brine shrimp (mg/kg dw) ¼ 10^(0.56(Log mgse/l water ( 0.064)) þ 0.24), for water > mg Se/L and brine shrimp ¼ 1.72 mg/kg dw for water < mg Se/L. Table 3. Tests of significance from dataset for cocollected data pairs. All field data. Linear regressions on log-transformed data. Regression tests Outcome p Level, n, r 2 Seston Se as predicted from total or dissolved water Se Weak positive, or Total ¼ 0.02, 99, 0.06 Not significant Dissolved ¼ 0.247, 34, Brine shrimp Se as predicted from seston Se Not significant (or as weak negative) 0.06, 98, 0.04 Brine shrimp Se as predicted from chlorophyll-normalized seston Se Brine shrimp Se as predicted from total or dissolved water Se or mixed water fractions in model Not significant 0.88, 76, Significant, positive Total ¼ <0.0001, 158, 0.59 Dissolved ¼ <0.001, 63, 0.69 Model ¼ <0.0001, 110, 0.64 Model (subset <10 mg Se/L) ¼ <0.0001, 90, 0.19

4 Selenium in Brine Shrimp Integr Environ Assess Manag 7, Figure 1. Brine shrimp tissue Se as predicted from total Se in cocollected water samples of the test dataset. The 3 data sources are Central Valley, California, ponds; Great Salt Lake, Utah (GSL); and Hailstone Reservoir, Montana. Figure 2. Piecewise regression model for brine shrimp as predicted from Se in water with <10 mg Se/L in the test dataset. DISCUSSION AND CONCLUSIONS Previous models examined relationships between brine shrimp tissue concentrations of Se and ambient Se in food or water (Table 1). One approach, as applied by the State of Utah to set a standard, was to use a biodynamic model based on field data to quantify the uptake of Se from water (partitioning coefficient or enrichment factor ) and the subsequent transfer of Se from particles to biota and on through the food chain to consumers and predators (trophic transfer factors) (Utah 2010; Marden 2008). This approach was modeled after the protocol for setting a Se standard as proposed for San Francisco Bay (Presser and Luoma 2006), although the Bay has a very different aquatic environment and community. Regardless, this simple, empirically derived modeling approach is based on field measurements of Se in water, seston (water column particles), brine shrimp tissue, and bird egg tissue. Efforts were made to link the Se concentrations in the various media as closely as possible in time and space, using all samples from the southern arm of Great Salt Lake. Key components of the model were used to derive a partitioning coefficient (particulate Se versus dissolved water Se) and trophic transfer factor (brine shrimp tissue Se versus particulate Se) from summaries of monitoring data. The simple multipliers can be used to estimate brine shrimp tissue Se concentrations from water. Other models linking waterborne concentrations of Se to brine shrimp Se concentrations relied on mechanistic or a combination of field-derived and mechanistic parameters. It is recognized that the water measured at one point in time is an inherently poor measure of all waterborne exposure relevant to brine shrimp bioaccumulation; the laboratory studies control for that effect. For example, the Grosell (2008) model provided predictions of brine shrimp tissue concentrations based on similar salinity to Great Salt Lake water and diet, as derived from controlled, laboratory experiments using Dunaliella as food. The Grosell model was incorporated as a possible predictive tool in the State of Utah standard setting spreadsheet model (Utah 2010). The scenario 1 model, used here, is a linear relationship between brine shrimp, their food, and water, for water concentrations <2.5 mg/l, as is appropriate for Great Salt Lake (Grosell 2008). The Brooks (2007) model also included laboratory and field-derived parameters to establish relationships between brine shrimp versus waterborne and dietary concentrations of Se. The Skorupa and Ohlendorf (1991) model is a field-derived regression analysis relating Se in brine shrimp tissue to waterborne concentrations, similar to the approach shown in Figure 1, but limited to populations from the Central Valley, California. The estimates of statistical error derived from the model in Figure 1 and the others in Table 1 was evaluated by comparing the model residuals derived from model predictions versus actual brine shrimp Se concentrations in the validation data set. The average residuals (measured predicted tissue Se concentrations, mg Se/kg dw, not as absolute values) for the brine shrimp tissue models, limited to water concentrations <10 mg Se/L, were: Figure 1, with data limited to water <10 mg Se/L: 0.75 Marden (2008): 2.11 Brooks (2007): 0.87 Skorupa and Ohlendorf (1991): 1.14 (where a perfect model would include average residuals of 0). Note that the Grosell (2008) model, Scenario 1, was constructed only for water concentrations <2.5 mg Se/L, so it is not included in the list above. For waterborne values <10 mg Se/L, the regression model of test samples in Figure 1 yielded the lowest residuals (bullet list, above). However, when all models were compared for waterborne values <2.5 mg Se/L, the Grosell model appears to provide the best predictions of the test dataset (Grosell residuals averaged 0.28). Most of the models predicted high (þ residuals) whereas the models of Grosell, and Skorupa and Ohlendorf predicted low ( residuals). Each model was constructed with unique ranges for the datasets, but for predicting brine shrimp tissue Se concentrations, in general, the simple regression model, as in Figure 1, has similar performance as compared to the other models for the lower range of waterborne concentrations (i.e., similar to Great Salt Lake). All models demonstrate the similar patterns of brine shrimp bioaccumulation of Se. An examination of the test dataset as shown in Figures 1 and 2 reveals the high variability for brine shrimp tissue Se concentrations at water concentrations <10 mg Se/L. As one approach, the Brooks model is based on a constant, average value of tissue concentrations of 3.9 mg Se/kg corresponding to waterborne

5 482 Integr Environ Assess Manag 7, 2011 ER Byron et al. concentrations <48 mg Se/L. Similarly, a piecewise regression of the test data (Figure 2) reveals a 0 slope prediction of 1.72 mg Se/kg for water Se concentrations <0.87 mg Se/L. Marden (2008) concluded that no significant relationship could be used to accurately predict brine shrimp tissue concentrations in Great Salt Lake. Apparently, there is little justification for quantifying a positive relationship for brine shrimp tissue Se concentrations derived from ambient Se for low concentrations of waterborne and/or dietary Se. The combination of varied sources of model error and physiological regulation of tissue Se concentrations at lower levels of ambient exposure preclude accurate predictions. Only higher ranges of ambient Se correspond to predictably higher concentrations of brine shrimp tissue Se (e.g., Figure 1, or Skorupa and Ohlendorf 1991). The piecewise regression results of Figure 2 are included for their potential to help explain the high variability in tissue concentrations of brine shrimp when waterborne Se environments are below 1 mg Se/L. Regardless, the overall model represented by Figure 1 (and for a similar regression for values <10 mg waterborne Se/L) is structurally simpler and has a slightly higher coefficient of determination (corrected r 2 of 0.19, p < for values <10 mg Se/L) and remains, therefore, a better choice to model bioaccumulation of brine shrimp Se than the piecewise approach of Figure 2 (corrected r 2 of 0.17, p < 0.01). However, Figure 2 highlights the lack of predictability for brine shrimp Se concentrations in environments of low water column concentrations. The finding is similar to that presented by Brooks (2007). Presumably, environmental exposure to Se at those concentrations results in physiologically regulated levels of Se in the brine shrimp but as ambient concentrations rise above 1 mg Se/L, bioaccumulated tissue concentrations begin to increase in association with exposure. The breakpoint for rising tissue Se concentrations in Great Salt Lake could occur as low as 0.9 mg Se/L in water (Figure 2) to as high as 48 mg Se/L (Brooks 2007). Other, nonlinear models did not produce an improved relationship. In conclusion, a regression model based on a comprehensive database of brine shrimp tissue Se concentrations revealed a positive relationship between waterborne and brine shrimp tissue concentrations of Se, but no discernable relationships with measures of Se in food. However, as verified for this dataset as well as previous studies, the high variability in tissue Se concentrations precludes the ability to create meaningful models based on causal linkages between brine shrimp tissue Se and environmental exposures for low ambient concentrations. In the case of Great Salt Lake, with waterborne total Se concentrations <1 mg/l, variations in waterborne concentrations do not yield predictable changes in brine shrimp tissue Se. Acknowledgment We wish to thank Joseph Skorupa for compiling brine shrimp and water data from various sources. Lucinda Tear performed the hockey stick regression analyses. Chris Kaiser and Kelly Payne at Kennecott Copper assisted with funding for this project. REFERENCES Brooks ML Final Report. Bioaccumulation of selenium by brine shrimp (Artemia franciscana) and indigenous algae (Dunaliella viridis) in Great Salt Lake waters. Prepared for Kennecott Utah Copper. January Brix KV Summary of selenium concentrations in water, sediments, invertebrates, and birds of the Great Salt Lake, Utah. Prepared for William Adams, Rio Tinto. March Brix KV, DeForest DK, Cardwell RD, Adams WJ Derivation of a chronic sitespecific water quality standard for selenium in the Great Salt Lake, Utah, USA. Environ Toxicol Chem 23: Fan TWM, Teh SJ, Hinton DE, Higashi RM Selenium biotransformations into proteinaceous forms by foodweb organisms of selenium-laden drainage waters in California. Aquat Toxicol 57: Grosell M Final report for the Brine Shrimp Kinetics Study, Project 5. Prepared for Utah Department of Environmental Quality, Division of Water Quality. Gwynn JW Great Salt Lake, an overview of change. DNR Special Publication, Utah Department of Natural Resources. Utah Geological Survey. IBM SPSS statistical package. Somers (NY): IBM. Kuehn D The brine shrimp industry in Utah. In: Gwynn JW, editor. Great Salt Lake, an Overview of Change. DNR Special Publication, Utah Department of Natural Resources. Utah Geological Survey. p Marden BT Preliminary results. Great Salt Lake water quality studies. Development of a selenium standard for the open waters of Great Salt Lake. Project 2B, synoptic survey of the pelagic zone: selenium in water, seston, and Artemia. Prepared for Utah Department of Environmental Quality, Division of Water Quality. Nelson KJ, Reiten JC Saline seep impacts on Hailstone and Halfbreed National Wildlife Refuges in South-Central Montana. US Fish and Wildlife, Region 6. Environmental Contaminants Program Report. DEC ID: , FFS: N47. [cited 2011 March 15]. Available from: montanafieldoffice/environmental_contaminants.html Presser TS, Luoma SN Forecasting selenium discharges to San Francisco Bay- Delta estuary: Ecological effects of a proposed San Luis drain extension. Professional Paper Reston (VA): US Department of Interior, US Geological Survey. SAS Institute Statview 5.0. Cary (NC): SAS Institute. Shelton JM, Hoffman-Floerke MD, Jacobsen D Bioaccumulation of trace elements in agricultural evaporation pond organisms in the San Joaquin Valley, California. Presentation at the Western and Northwest Sections Meeting of the Wildlife Society, Sparks (NV), Skorupa JP, Ohlendorf HM Contaminants in drainage water and avian risk thresholds. In: Dinar A, Zilberman D, editors. The Economics and Management of Water and Drainage in Agriculture. Dordrecht (NL): Kluwer Academic. p [USEPA] US Environmental Protection Agency National recommended water quality criteria. Office of Water. [cited 2011 March 15]. Available from: Utah State of Utah, Division of Administrative Rules. Rule R Standards of Quality for Waters of the State. In effect 1 March [cited 2011 March 15]. Available from: htm#t7