Patterns of co-contamination in freshwater and marine fish of the northeastern United States

doi:10.21203/rs.3.rs-1762592/v1

Download PDF

Research Article

Patterns of co-contamination in freshwater and marine fish of the northeastern United States

https://doi.org/10.21203/rs.3.rs-1762592/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 26 Jun, 2023

Read the published version in Environmental Modeling & Assessment →

You are reading this latest preprint version

Persistent, bioaccumulative, and toxic contaminants (PBTs) are known to co-occur in fish tissue, yet this covariance has not been explicitly incorporated into model-based risk assessments that inform fish consumption advisories. We utilize available US EPA datasets to statistically model the covariance among PBT concentrations in fish tissue and the dependence of this covariance on waterbody and watershed conditions. We find that most PBTs positively covary, whether fish were collected in rivers, lakes, or coastal waters. Mercury in lakes, and mercury, PFCs, and heptachlor in rivers, covary negatively with the other PBTs. While much of the variance and covariance in PBTs can be statistically related to fish characteristics and watershed and waterbody conditions, a large amount remains in model residuals. This implies that single contaminant models, even if highly precise, can misestimate total health risk by neglecting the substantial covariance with other PBTs that is left unmodelled.

EMAP

NRSA

NCA

fish consumption advisories

risk profile

persistent

bioaccumulative and toxic chemicals

Human activities involving the use of chemicals and fossil fuels in industrial, residential, and agricultural settings have led to the presence of contaminants throughout the natural environment [1–3]. Persistent, bioaccumulative and toxic substances (PBTs) continue to pose human health risk despite efforts to ban their use or curb their emissions [4]. Of particular concern are mercury and organic compounds including pesticides, fungicides, industrial chemicals, and their metabolites. The concentrations of these environmental contaminants are particularly high in the northeastern U.S. due to dense urbanization, a history of agriculture, and proximity to coal fired power plants [5].

The primary route of human exposure to PBTs is through the consumption of fish [6]. These substances bioaccumulate in food chains [7, 8], leading to concentrations in fish tissue that can cause a variety of health problems, especially to vulnerable populations such as children, the elderly, pregnant women, and their unborn children [9]. The northeastern U.S. has one of the highest consumption rates of fish in the country, so its residents are particularly at risk of contaminant exposure via fish tissue [10].

In the U.S., the health risks posed by contaminated fish are managed using consumption advisories in the form of recommended meal restrictions. However, advisories vary from state to state in how they are calculated and the number of contaminants included. In a comparative analysis of state consumption advisories across the US, it was found that about a third of states based advisories on a single contaminant either statewide or for each waterbody. Almost half of states include advisories based on multiple contaminants but offer no explanation as to how this advice is developed [11]. There is evidence that many of the common contaminants included in these advisories have additive or synergistic health effects when consumed simultaneously, but many chemical mixtures have yet to be tested. The toxicity of chemical mixtures is therefore a critical area of health research [11–14], and the CDC has an ongoing effort to build a database of interaction profiles under the Agency for Toxic Substances and Disease Registry [15]. As the nature of most interactions is currently unknown, the EPA recommends that, as a precaution, additivity should be assumed for carcinogens and contaminants with similar mechanisms of toxicity [16].

Even if the health impacts of chemical mixtures were to be well-characterized, a separate but related question is, “What are the particular mixtures to which various populations are being exposed?” Answering this question requires analysis of the covariance of contaminants in fish tissue and the dependence of this covariance on local watershed and waterbody conditions. Most studies of fish tissue concentrations to date have only considered contaminants individually or in small groups [4, 17–20]. However, because we expect many contaminants to have common sources and similar environmental fate and transport mechanisms. A complete assessment of health risk should consider many contaminants simultaneously and with an increased focus on measuring many contaminants at once, the data required to do such an assessment is more available than ever [21].

In this paper, we apply a novel, multivariate Bayesian modeling method to analyze publicly available fish tissue data from the EPA to address the following three questions: 1) How do measured PBT concentrations in fish tissue covary across different waterbodies and how does this covariance differ between lakes, rivers, and coastal zones? 2) How much of the observed PBT covariance can be explained by models that account for local watershed and waterbody factors? 3) What are the practical implications of PBT covariance that cannot be addressed by models and therefore remains a component of model predictive uncertainty?

2.1 Data Description

Our analysis employed three datasets that are publicly available from the U.S. EPA: the Environmental Monitoring and Assessment Program (EMAP) Northeast Lakes data (1991–1994), the National Rivers and Streams Assessment (NRSA) data (2008–2009), and the National Coastal Assessment (NCA) northeast region data (2000–2006) [22–24]. The sample sites included in this study are mapped in Fig. 1. Measured PBTs are listed in Table 1, and fish, waterbody, and watershed variables are listed in Table 2. Most data come from the original datasets, however land use characteristics were derived from the National Land Cover Database (NLCD) for the years 2000 for the EMAP data and 2006 for the NRSA data [25]. A combination of the 2000 and 2006 NLCD data was used for NCA based on which year was closest in time to the fish tissue samples. Because watersheds of the NCA coastal locations are not well-defined, land use characteristics were derived from any land area within 200 km of the point at which the fish tissue samples were collected. Fish species trophic levels were determined through literature review (see Appendix A for details).

2.2 Data Analysis

Data were analyzed using the R package GJAM [26]. This package is designed to model multivariate response variables that can be both discrete and continuously valued. Importantly for our purpose, all response variables representing PBT concentrations can be modeled simultaneously, with explicit representation of their covariance structure. Also importantly, GJAM appropriately handles data which are variously censored, a common trait of fish tissue measurements that may fall below method detection limits (MDLs) that differ by contaminant. GJAM treats values reported to be below the MDL as being unknown on the interval between zero and the MDL, then simulates these values from a latent scale derived from the truncated distribution of measured values above the MDL. This approach avoids biases that are incurred when censored values are replaced by zeros, the MDL, or any other specific value [27–29].

Our GJAM model structure is depicted graphically in Fig. 2. In this model, the measured (possibly censored and transformed) concentrations Y_ik of contaminant k in fish tissue sample i are assumed to be the manifestation of the true (uncensored) concentrations W_ik, with mean µ_ik (each modeled as a linear function of j measured environmental and fish attribute predictors X_ij and parameters B_jk) and model error described by a common multivariate distribution with variance-covariance matrix ∑_kk. Thus, the covariance of predicted mean concentrations µ_ik represents the portion of the overall contaminant covariance that is explainable by differences in environmental and fish attributes, while the matrix Σ_kk represents the portion of the covariance that cannot be explained by these factors and is based on the joint distribution of model residuals. This remaining covariation should not be ignored, but rather explicitly considered in waterbody assessments, as it represents the residual risk of simultaneous exposure to multiple contaminants.

GJAM employs a Bayesian approach to model fitting. Parameter and variance-covariance estimates are represented by the posterior distribution $P\left[B,{\Sigma }|X, Y\right]$, which is derived from Bayes theorem as the product of a likelihood function of the fish tissue concentration measurements, P$\left[Y|X,B,{\Sigma }\right]$, and a joint prior distribution of the parameters and model error variance-covariance matrix, $P[B,{\Sigma }]$,:

$$\begin{array}{c}P\left[B, {\Sigma }|X,Y\right]= \frac{P\left[Y|X,B, {\Sigma }\right]\times P\left[B,{\Sigma }\right]}{\iint P\left[Y|X,B, {\Sigma }\right]\times P\left[B,{\Sigma }\right]dBd{\Sigma }} \#\left(1\right)\end{array}$$

We used a multivariate normal likelihood function, after a square-root transformation of concentration measurements to counter skewness and maintain strictly positive values. To minimize the influence of the prior, we used a widely-dispersed multivariate normal distribution on the parameters with mean of zero and a Jeffrey’s prior on the variances and covariances.

We simulated the posterior distribution using Gibbs sampling, a version of Markov Chain Monte Carlo (MCMC) simulation implemented in GJAM. This allows for estimation of missing and censored values, which were a common occurrence in our data. At each iteration of the Gibbs sampler, missing and censored data are replaced by inferred values derived from that iteration’s posterior parameter estimates.

Each of the three EPA datasets was analyzed separately using the variables available. Environmental predictor variables (Table 2) included those used in previous analyses of individual contaminants [18, 30–32]. Fish length, trophic level, lipid content, and species were used to control for fish characteristics, except for EMAP which does not include fish length and NRSA which does not include lipid concentration. Further, in the NRSA and NCA datasets, trophic level and species are redundant, so only species data were considered. Species data is included as a categorical variable in the model with GJAM selecting the species with the most data points as the reference species. Contaminants included all PBTs reported by each dataset (Table 1), with efforts to achieve consistency when possible. For example, not all datasets report specific PCB congeners, so we considered only total PCBs. Contaminant properties (e.g., octanol-water partition coefficient) were not considered explicitly, as their effects are presumed to be captured by the estimated model parameters. If only one observation existed for a species in a dataset, that observation was removed. Our analysis thus employed 222 (EMAP), 1001 (NRSA), and 118 (NCA) fish tissue samples across 158, 679, and 118 locations, respectively.

2.3 Model Selection

We present results for the model of each dataset that includes the combination of environmental predictor variables leading to the minimal value of the deviance information criterion (DIC). The DIC is a measure of goodness of fit, penalizing for model complexity [33, 34]. Unlike other criteria such as the Akaike Information Criterion (AIC), the DIC is easily calculated from MCMC samples. We explored the space of possible models using a two-way stepwise algorithm, initiated with a minimal model including only tissue type and species as predictors, and proceeding to alternatingly add and remove additional predictors to reduce the DIC value.

Covariation among contaminants is presented for each dataset in the form of three normalized variance-covariance matrices of (square-root transformed) fish tissue concentrations representing, respectively: (a) variation in the original measured data, (b) variation attributable to modeled factors, and (c) variation left unexplained by modeled factors. Normalization involved dividing each entry of each variance-covariance matrix by the product of standard deviations of the measured values of the two respective contaminants. Thus, in matrix (a) all diagonal elements equal one and off-diagonal elements represent the correlations between measured contaminant values. In matrix (b), the (off-)diagonal elements represent the normalized (co-) variance explained by the model (equivalent to model R² values in the case of the diagonal elements). The elements of matrix (c), therefore, represent the (co-)variance not explained by the model.

Parameter estimates for each data set are presented as matrices of the normalized sensitivity of (square-root transformed) fish tissue contaminant concentrations to each fish attribute and environmental predictor variable contained in the final model. Sensitivity values were calculated by multiplying each mean model parameter estimate by the average of its corresponding predictor variable and dividing by the average of the corresponding (square-root transformed) contaminant concentration. To better compare these values across all predictor-contaminant combinations, they are displayed as the signed quantiles across all sensitivity values, and thus range from − 1 to 1.

3.1 EMAP Lakes Data

The covariance structure of the EMAP data (Fig. 3, matrix a) reveals that in the tissues of lake fish, mercury is negatively correlated with all other PBTs, which are positively correlated with one another. A linear model accounting for environmental and fish attributes is able to explain a moderate proportion of the variance in individual contaminants, with R² values ranging from 0.29 to 0.72 (diagonal elements in matrix b, with specific values given in Appendix B), and a similar proportion of the covariance among contaminants (ratio of the off-diagonal elements in matrix a to those in matrix b) including the negative correlations between mercury and other contaminants. Overall, the chlordanes and related chemicals are predicted most accurately, both individually and in concert, while aldrin, mercury, hexachlorobenzene, and DDT and its breakdown products, as well as their covariances, are predicted least well. A moderate amount of (co-)variance remains unexplained (matrix c), as well as a mild negative covariance between hexachlorobenzene and both aldrin and oxychlordane that was induced by the model’s slight overestimate of the covariance between these contaminants.

The model best accounting for the variances and covariances among fish tissue contaminants in the EMAP data is one that includes, in addition to trophic level, lipid content, and fish species, the environmental variables percent agriculture, alkalinity, latitude, urban land, and zooplankton abundance (Fig. 4). Significant parameters in this model indicate that the measured contaminants generally increase with greater trophic level and lipid content and, relative to yellow perch, are generally greater in lake trout and white perch and lower in Allegheny pearl dace, blue gill, brown bulkhead, and chain pickerel, with a few exceptions. Both lake trout and white perch have higher concentrations of all contaminants. Contaminant concentrations are generally lower at sites from higher latitudes, except for aldrin and mercury which are higher. Greater levels of urban land and zooplankton abundance are each generally associated with higher contaminant concentrations. Fish tissue concentrations are generally most sensitive to trophic level, species, and latitude.

3.2 NRSA River Data

The NRSA data (Fig. 5, matrix a) reveal that mercury, heptachlor, and total PFC concentrations negatively covary with most other PBTs in the tissue of fish from rivers and streams, and that these other PBTs positively covary with each other. Only a fraction of this covariance structure was captured in the modeled mean concentrations (matrix b) with proportions of explained co-variance ranging from zero to 0.59. Endosulfan II was predicted especially poorly, as the model only captured 16% of the variation in this contaminant, and a comparably low proportion of its covariance with other contaminants. Mercury, heptachlorepoxide, the nonachlors, the chlordanes, and their covariances were predicted reasonably well, with R² values all above 0.3. A substantial amount of variance and covariance in the NRSA data remains unattributable to modeled factors (matrix c).

The model best accounting for the variances and covariances of fish tissue contaminants in the NRSA data is one that includes, in addition to lipid content, total length, and species, the environmental variables: riparian vegetation, instream dissolved organic carbon (DOC) and pH (Fig. 6). Riparian vegetation is a positive and significant predictor for total PFCs but a significant negative predictor for heptachlor. DOC is a significant positive predictor for mercury, total PFCs, endrin ketone, and heptachlorepoxide. For nearly all contaminants, pH is a significant and positive predictor of fish tissue contaminant concentration. Heptachlor and mercury generally show quite different relationships across species (relative to smallmouth bass) than the other contaminants.

3.3 NCA Data

In coastal waterways, all contaminants positively covary (Fig. 7, matrix a). This dataset is limited, however, in that many contaminants were not measured concurrently. DDT and related metabolites covary particularly strongly with each other and with transnonachlor, with normalized covariance (correlation) values close to 1. While the best fitting model yields R² values for 4,4-DDT,4,4-DDD, 4,4-DDE, total PCBs, and transnonachlor from 0.43 to 0.66, most others are quite low. Additionally, only small portion of the observed covariance among contaminants is reproduced by the model (matrix b), with the vast majority remaining unexplained (matrix c).

For the NCA data, the best fitting model (Fig. 8) includes the fish characteristics: tissue type (offal generally lower than filet and whole fish generally greater), lipid content (weak positive and negative relations with contaminant concentration), fish length (positive when significantly different from zero), and species (mixed sign and significance, but perch and catfish species notably positive). Significant environmental variables include: the percentages of agricultural, forested, and urban land, as well as the total nitrogen concentration, all of mixed sign and significance depending on the contaminant. Comparisons between species show channel catfish and white perch having relatively high concentrations of all contaminants in their tissues.

The data on contaminant concentrations in fish tissue show that PBTs generally covary positively across all types of waterbodies, with the exception of mercury against other contaminants in both lakes and rivers, and total PFCs and heptachlor against other contaminants in rivers. In coastal waters, these particular contaminants were not measured concurrently with others. The positive covariance among most contaminants implies that, lacking any more detailed information on fish characteristics or environmental conditions, one should assume that if one contaminant is elevated at a site, most others will be as well. However, in lakes and rivers, mercury is low when other contaminants are high in fish tissue. Gandhi et al. (2017) show empirically that the covariance patterns present in fish tissue data from the Great Lakes imply that the current practice of basing fish consumption advisories on the single most restrictive contaminant is not adequately protective. Gandhi et al. note, however, that both the magnitude and sign of the covariance among contaminants can differ according to location and fish species. They thus call for a more generalizable statistical approach for performing chemical mixture assessments than the simulation-based method they employ. In particular, they call not for multiple single-contaminant models, but rather for single multi-contaminant models such as that employed in the current study.

The performance of our multivariate modelling approach differed across the three data sets. In lakes, the joint distribution of multiple contaminants was captured reasonably well, with R² values as high as 0.72 and never below 0.29, and similar proportions of covariance among contaminants explained by the model. This is a level of explanatory power that is similar to previously published, single-contaminant models [5,30, 35]. The model performed less well in rivers, with captured proportions of variances and covariances ranging from 0.16 to 0.59. This may be because riverine fish may occupy a variety of locations over the course of their lifetimes and therefore land use variables are not predictive fish tissue contaminant concentrations. The model of coastal marine systems was able to account for only a small proportion of the variance and covariance among fish tissue contaminant concentrations (R² values ranging from 0.04 to 0.66, with most in the lower half of this range). This may be because of the wider variety of tissue types, species, and corresponding life cycles of fish included in this data set. Also, as for rivers, fish living in coastal areas are highly mobile and may not spend the majority of their time near the land used to calculate land use percentages. In general across all waterbodies, the contaminants best fit by the models were mercury, total PCBs, DDT and its metabolites, the nonachlors, and oxychlordanes and their covariances.

The majority of the (co-)variation accounted for by the models could be attributed to fish characteristics, such as tissue type, lipid content, length, trophic level, and species. However, watershed land uses and waterbody characteristics were also significant predictors – again in line with the results of single-contaminant modeling [5, 30, 36]. In the model of the EMAP lake and river data, for example, lakes with watersheds having more urban land showed consistently higher fish tissue contaminant concentrations (except for mercury). For coastal fish, agricultural and forested land were significant predictors of contaminant concentrations, although the sign of these relationships varied by contaminant.

Water chemistry has been shown to be important in the bioaccumulation of Hg but is less well explored for organic contaminants [30, 37–39]. We found that fish tissue mercury concentrations decreased with greater alkalinity in lakes and pH in rivers. This is consistent with the finding that acidity is associated with enhanced mercury methylation and uptake in lower alkalinity and pH systems [38]. Concentrations of contaminants other than mercury were positively associated with pH levels in rivers. While metal contaminants are more bioavailable at lower pH, there is little research on the relationship of pH with organic contaminant bioavailability. Mercury, PFCs, and some other contaminants were found to be positively associated with DOC in riverine fish from the NRSA dataset. Like pH, the influence of DOC on metal transport and bioavailability are known but not commonly studied for organic contaminants. Mercury concentrations in water are positively related to DOC, but DOC appears to reduce the bioaccumulation of mercury particularly at high DOC ranges [40, 41]. In the few studies that exist for organic contaminants, the influence of DOC does not show consistent patterns [42–45].

Our intent with this study was to demonstrate a method of modeling multiple contaminants simultaneously, using a Bayesian approach that can account for both missing and censored values (i.e. below detection limit). With such a model, one can assess how much of the covariance present among fish tissue contaminants can be accounted for by similar responses to modeled factors, and how much remains unaccounted for. Because we used a simple linear approach to model many contaminants, in many species, across many aquatic environments, each with particular fish characteristics, water chemistry attributes, and land use variables, our model results may not be as precise or mechanistically interpretable as more focused analyses. Yet, the R² values we were able to obtain are comparable to those of previously-published statistical models of single contaminants.

A more novel contribution of this study is our demonstration that, while a multivariate model can explain some of the covariance inherently present among contaminants, the proportion of covariance that remains unexplained is at least as large as the proportion of variance left unexplained for individual contaminants. This means that the application of multiple single-contaminant models – even those that fully account for predictive uncertainty – misestimate true human health risk. This is because unexplained covariance implies that, even after all modeled factors are taken into account, the likelihood of unusually high (or low) levels of one contaminant at a site are not independent of the likelihood of unusually high (or low) levels of another contaminant. The fact that most contaminants (except mercury and heptachlor in rivers) display unexplained variation with positive covariance means that unexpectedly high levels of multiple contaminants are likely to occur simultaneously, exacerbating human health risk.

It is important to emphasize that this underestimation of health risk that results from ignorance of positive error covariance among contaminants is a systematic bias and not just an overlooked contribution to predictive uncertainty. This is clear from the fact that the variance of the sum of two correlated random variables X₁ and X₂, for example, is computed as:

$$\begin{array}{c}{\sigma }_{{\text{X}}_{1}+{\text{X}}_{2}}^{2}={\sigma }_{{\text{X}}_{1}}^{2}+{\sigma }_{{\text{X}}_{2}}^{2}+2\bullet Cov\left({\text{X}}_{1}, {\text{X}}_{2}\right)={\sigma }_{{\text{X}}_{1}}^{2}+{\sigma }_{{\text{X}}_{2}}^{2}+2\bullet \rho \bullet {\sigma }_{{\text{X}}_{1}}\bullet {\sigma }_{{\text{X}}_{2}}\left(2\right)\end{array}$$

where σ² represents variance, Cov( ) represents covariance, and ρ represents correlation between X₁ and X₂. Thus, with positive covariance/correlation between two variables, the variance of the sum is always greater than the sum of the variances. Thus, the probability of high concentrations will be larger than would be calculated by assuming independence between the variables. This result can easily be extended to more than two random variables and to any linear combination of variables, making it applicable to the analysis of contaminants assumed to have additive health effects [16].

Models can be useful tools for predicting fish tissue contaminant concentrations at unsampled sites or under new conditions as a function of watershed attributes, water quality measures, and fish characteristics. However, model-based estimates of human dietary exposure for the purposes of setting fish consumption advisories need to account for the full distribution of possible contaminant concentrations, not just single point estimates. Several studies have demonstrated the value of a Bayesian approach in this context, as these methods allow for the characterization of both parametric and structural error [35, 46, 47]. This approach has long been advocated in the academic literature [48, 49] and is starting to make its way into fish consumption advisory decisions, at least for individual contaminants [50]. The present study is the logical extension of that approach to multiple contaminants, acknowledging that, if one contaminant happens to be unexpectedly elevated at a site, others are likely to be as well.

While our study demonstrates a moderate ability to predict contaminant concentrations with a multivariate model, much of the covariance among contaminants remains in the residuals. This indicates that the presence and composition of mixtures are relatively unpredictable, even knowing watershed attributes, water quality measures, and fish characteristics. Further analyses might reveal the importance of including additional predictor variables or formulating nonlinear models, and thus more precisely characterize the patterns of suites of contaminants. In the meantime, uncertainty in both single contaminants and their co-occurrence needs to be reflected in fish consumption advisories and in more holistic assessments of risk-risk and risk-benefit tradeoffs [51, 52].

6.1 Acknowledgements

The authors would like to thank Jim Clark, the creator of GJAM, for his support in model trouble shooting and aiding in the interpretation of the model results.

6.2 Author Contributions

K.B. performed the analysis and created the figures. A.C., J.C. and K.B. collected the data. M.E.B. and K.B. wrote the manuscript. All authors reviewed the manuscript.

6.3 Declaration of interests

The authors of this manuscript declare no competing financial or non-financial interests.

6.4 Funding

Funding to support this work was provided by the National Institute of Environmental Health Sciences grant number P42 ES007373-S1 awarded to Dr. Celia Y. Chen.

6.5 Ethics Declaration Statement

As this study utilizes previously collected data, there are no ethics standard or forms that apply.

6.6 Consent to Participate

Not Applicable

6.7 Consent for Publication

All authors of this manuscript have reviewed it and consent to its publication.

6.8 Data Availability

Data analyzed in this study can be found at the following sources. Additional data taken from multiple publications is described in the appendices.

Chase, M. E., Jones, S. H., Hennigar, P., Sowles, J., Harding, G. C. H., Freeman, K., Wells, P. G., Krahforst, C., Coombs, K., Crawford, R., Pederson, J., & Taylor, D. (2001). Gulfwatch: Monitoring Spatial and Temporal Patterns of Trace Metal and Organic Contaminants in the Gulf of Maine (1991–1997) with the Blue Mussel, Mytilus edulis L. Marine Pollution Bulletin, 42(6), 490-504. https://doi.org/10.1016/S0025-326X(00)00193-4
Driscoll, C. T., Han, Y.-J., Chen, C. Y., Evers, D. C., Lambert, K. F., Holsen, T. M., Kamman, N. C., & Munson, R. K. (2007). Mercury Contamination in Forest and Freshwater Ecosystems in the Northeastern United States. BioScience, 57(1), 17-28. http://dx.doi.org/10.1641/B570106
Phillips, P. J., Nowell, L. H., Gilliom, R. J., Nakagaki, N., Murray, K. R., & VanAlstyne, C. (2010). Composition, distribution, and potential toxicity of organochlorine mixtures in bed sediments of streams. Science of The Total Environment, 408(3), 594-606. https://doi.org/10.1016/j.scitotenv.2009.09.052
Stahl, L. L., Snyder, B. D., Olsen, A. R., & Pitt, J. L. (2009). Contaminants in fish tissue from US lakes and reservoirs: a national probabilistic study. Environmental Monitoring and Assessment, 150(1-4), 3-19. https://doi.org/10.1007/s10661-008-0669-8
Chen, C. Y., Stemberger, R. S., Kamman, N. C., Mayes, B. M., & Folt, C. L. (2005). Patterns of Hg Bioaccumulation and Transfer in Aquatic Food Webs Across Multi-lake Studies in the Northeast US [journal article]. Ecotoxicology, 14(1), 135-147. https://doi.org/10.1007/s10646-004-6265-y
Dorea, J. G. (2006). Fish meal in animal feed and human exposure to persistent bioaccumulative and toxic substances. Journal of food protection, 69(11), 2777-2785.
Opperhuizen, A., Wijbenga, A., & Hutzinger, O. (1984). Bioaccumulation of super‐lipophilic chemicals in fish AU - Bruggeman, W. A. Toxicological & Environmental Chemistry, 7(3), 173-189. https://doi.org/10.1080/02772248409357024
Petersen, G. I., & Kristensen, P. (1998). Bioaccumulation of lipophilic substances in fish early life stages. Environmental Toxicology and Chemistry, 17(7), 1385-1395. https://dx.doi.org/10.1002/etc.5620170724
Dórea, J. G. (2008). Persistent, bioaccumulative and toxic substances in fish: Human health considerations. Science of The Total Environment, 400(1), 93-114. https://doi.org/10.1016/j.scitotenv.2008.06.017
USEPA. (2014). Estimated Fish Consumption Rates for the U.S. Population and Selected Subpopulations (NHANES 2003-2010). Washington, DC
Carpenter, D. O., Arcaro, K. F., Bush, B., Niemi, W. D., Pang, S., & Vakharia, D. D. (1998). Human health and chemical mixtures: an overview. Environmental Health Perspectives, 106(Suppl 6), 1263-1270. https://doi.org/10.1289/ehp.98106s61263
Yang, R. S. H., Hong, H. L., & Boorman, G. A. (1989). Toxicology of chemical mixtures: experimental approaches, underlying concepts, and some results. Toxicology Letters, 49(2), 183-197. http://doi.org/10.1016/0378-4274(89)90032-5
Monosson, E. (2005). Chemical Mixtures: Considering the Evolution of Toxicology and Chemical Assessment. Environmental Health Perspectives, 113(4), 383-390. https://doi.org/10.1289/ehp.6987
Spurgeon, D. J., Jones, O. A. H., Dorne, J.-L. C. M., Svendsen, C., Swain, S., & Stürzenbaum, S. R. (2010). Systems toxicology approaches for understanding the joint effects of environmental chemical mixtures. Science of The Total Environment, 408(18), 3725-3734. http://doi.org/10.1016/j.scitotenv.2010.02.038
ASTDR. (2018). Interaction Profiles for Toxic Substances. Agency for Toxic Substances and Disease Registry.
USEPA. (2000). Guidance for assessing chemical contaminant data for use in fish advisories Washington, DC
Gewurtz, S. B., Shen, L., Helm, P. A., Waltho, J., Reiner, E. J., Painter, S., Brindle, I. D., & Marvin, C. H. (2008). Spatial Distributions of Legacy Contaminants in Sediments of Lakes Huron and Superior. Journal of Great Lakes Research, 34(1), 153-168. https://doi.org/10.3394/0380-1330(2008)34[153:SDOLCI]2.0.CO;2
Nicklisch, S. C., Bonito, L. T., Sandin, S., & Hamdoun, A. (2017). Geographic differences in persistent organic pollutant levels of yellowfin tuna. Environmental Health Perspectives, 67014, 1. https://doi.org/10.1289/ehp518
Nowell, L. H., Moran, P. W., Gilliom, R. J., Calhoun, D. L., Ingersoll, C. G., Kemble, N. E., Kuivila, K. M., & Phillips, P. J. (2013). Contaminants in Stream Sediments From Seven United States Metropolitan Areas: Part I: Distribution in Relation to Urbanization. Archives of Environmental Contamination and Toxicology, 64(1), 32-51. https://doi.org/10.1007/s00244-012-9813-0
De Vault, D. S., Hesselberg, R., Rodgers, P. W., & Feist, T. J. (1996). Contaminant Trends in Lake Trout and Walleye From the Laurentian Great Lakes. Journal of Great Lakes Research, 22(4), 884-895. https://doi.org/10.1016/S0380-1330(96)71009-2
Gandhi, N., Drouillard, K. G., Arhonditsis, G. B., Gewurtz, S. B., & Bhavsar, S. P. (2017). Are fish consumption advisories for the Great Lakes adequately protective against chemical mixtures? Environmental Health Perspectives, 125(4), 586-593. https://doi.org/10.1289/EHP104
USEPA. (2010). National Coastal Assessment. https://archive.epa.gov/emap/archive-emap/web/html/index-149.html.
USEPA. (2016). National Rivers and Streams Assessment 2008-2009: A Collaborative Survey. https://www.epa.gov/national-aquatic-resource-surveys/national-rivers-and-streams-assessment-2008-2009-report.
USEPA. (1999). EMAP Surface Waters Lake Database. https://archive.epa.gov/emap/archive-emap/web/html/index.html.
Dewitz, J., 2019, National Land Cover Database (NLCD) 2016 Products (ver. 2.0, July 2020): U.S. Geological Survey data release, https://doi.org/10.5066/P96HHBIE.
Clark, J. S., Nemergut, D., Seyednasrollah, B., Turner, P. J., & Zhang, S. (2017). Generalized joint attribute modeling for biodiversity analysis: median‐zero, multivariate, multifarious data. Ecological Monographs, 87(1), 34-56.
Helsel, D. R. (1990). Less than obvious - statistical treatment of data below the detection limit. Environmental Science & Technology, 24(12), 1766-1774. Retrieved 1990/12/01, from https://doi.org/10.1021/es00082a001
Shumway, R. H., Azari, R. S., & Kayhanian, M. (2002). Statistical Approaches to Estimating Mean Water Quality Concentrations with Detection Limits. Environmental Science & Technology, 36(15), 3345-3353. https://dx.doi.org/10.1021/es0111129
Antweiler, R. C., & Taylor, H. E. (2008). Evaluation of Statistical Treatments of Left-Censored Environmental Data using Coincident Uncensored Data Sets: I. Summary Statistics. Environmental Science & Technology, 42(10), 3732-3738. https://dx.doi.org/10.1021/es071301c
Shanley, J. B., Moore, R., Smith, R. A., Miller, E. K., Simcox, A., Kamman, N., Nacci, D., Robinson, K., Johnston, J. M., Hughes, M. M., Johnston, C., Evers, D., Williams, K., Graham, J., & King, S. (2012). MERGANSER: An Empirical Model To Predict Fish and Loon Mercury in New England Lakes. Environmental Science & Technology, 46(8), 4641-4648. Retrieved 2012/04/17, from http://dx.doi.org/10.1021/es300581p
Sunderland, E. M., Gobas, F. A., Branfireun, B. A., & Heyes, A. (2006). Environmental controls on the speciation and distribution of mercury in coastal sediments. Marine Chemistry, 102(1-2), 111-123.
Stahl, L. L., Snyder, B. D., Olsen, A. R., Kincaid, T. M., Wathen, J. B., & McCarty, H. B. (2014). Perfluorinated compounds in fish from U.S. urban rivers and the Great Lakes. Science of The Total Environment, 499, 185-195. Retrieved 2014/11/15/, from http://www.sciencedirect.com/science/article/pii/S0048969714011632
Spiegelhalter, D. J., Best, N. G., Carlin, B. P., & Van Der Linde, A. (2014). The deviance information criterion: 12 years on. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 76(3), 485-493. https://dx.doi.org/10.1111/rssb.12062
Gelman, A., Hwang, J., & Vehtari, A. (2014). Understanding predictive information criteria for Bayesian models. Statistics and Computing, 24(6), 997-1016. https://dx.doi.org/10.1007/s11222-013-9416-2
Visha, A., Gandhi, N., Bhavsar, S. P., & Arhonditsis, G. B. (2015). A Bayesian assessment of the mercury and PCB temporal trends in lake trout (Salvelinus namaycush) and walleye (Sander vitreus) from lake Ontario, Ontario, Canada. Ecotoxicology and Environmental Safety, 117, 174-186. Retrieved 2015/07/01/, from http://www.sciencedirect.com/science/article/pii/S0147651315001219
De Vault, D. S., Hesselberg, R., Rodgers, P. W., & Feist, T. J. (1996). Contaminant Trends in Lake Trout and Walleye From the Laurentian Great Lakes. Journal of Great Lakes Research, 22(4), 884-895. https://doi.org/10.1016/S0380-1330(96)71009-2
Lange, T. R., Royals, H. E., & Connor, L. L. (1993). Influence of Water Chemistry on Mercury Concentration in Largemouth Bass from Florida Lakes. Transactions of the American Fisheries Society, 122(1), 74-84. https://doi.org/10.1577/1548-8659(1993)122%3C0074:IOWCOM%3E2.3.CO;2
Miskimmin, B. M., Rudd, J. W. M., & Kelly, C. A. (1992). Influence of Dissolved Organic Carbon, pH, and Microbial Respiration Rates on Mercury Methylation and Demethylation in Lake Water. Canadian Journal of Fisheries and Aquatic Sciences, 49(1), 17-22. https://doi.org/10.1139/f92-002
Ravichandran, M. (2004). Interactions between mercury and dissolved organic matter––a review. Chemosphere, 55(3), 319-331. https://doi.org/10.1016/j.chemosphere.2003.11.011
Tsui, M. T. K., & Finlay, J. C. (2011). Influence of dissolved organic carbon on methylmercury bioavailability across Minnesota stream ecosystems. Environmental Science & Technology, 45(14), 5981-5987. https://doi.org/10.1021/es200332f
Broadley, H. J., Cottingham, K. L., Baer, N. A., Weathers, K. C., Ewing, H. A., Chaves-Ulloa, R., Chickering, J., Wilson, A. M., Shrestha, J., & Chen, C. Y. (2019). Factors affecting MeHg bioaccumulation in stream biota: the role of dissolved organic carbon and diet. Ecotoxicology, 28(8), 949-963. https://doi.org/10.1007/s10646-019-02086-2
Gourlay, C., Tusseau‐Vuillemin, M. H., Garric, J., & Mouchel, J. M. (2003). Effect of dissolved organic matter of various origins and biodegradabilities on the bioaccumulation of polycyclic aromatic hydrocarbons in Daphnia magna. Environmental Toxicology and Chemistry: An International Journal, 22(6), 1288-1294. https://doi.org/http://dx.doi.org/10.1002/etc.5620220615
Bejarano, A. C., Decho, A. W., & Thomas Chandler, G. (2005). The role of various dissolved organic matter forms on chlorpyrifos bioavailability to the estuarine bivalve Mercenaria mercenaria. Marine Environmental Research, 60(1), 111-130. https://doi.org/10.1016/j.marenvres.2004.10.001
Akkanen, J., Penttinen, S., Haitzer, M., & Kukkonen, J. V. K. (2001). Bioavailability of atrazine, pyrene and benzo[a]pyrene in European river waters. Chemosphere, 45(4), 453-462. https://doi.org/10.1016/S0045-6535(01)00038-8
Haitzer, M., Akkanen, J., Steinberg, C., & Kukkonen, J. V. K. (2001). No enhancement in bioconcentration of organic contaminants by low levels of DOM. Chemosphere, 44(2), 165-171. https://doi.org/10.1016/S0045-6535(00)00269-1
Mahmood, M., Bhavsar, S. P., & Arhonditsis, G. B. (2013). Fish contamination in Lake Erie: An examination of temporal trends of organochlorine contaminants and a Bayesian approach to consumption advisories. Ecological Informatics, 18, 131-148. https://doi.org/10.1016/j.ecoinf.2013.08.001
Visha, A., Gandhi, N., Bhavsar, S. P., & Arhonditsis, G. B. (2016). Guiding fish consumption advisories for Lake Ontario: A Bayesian hierarchical approach. Journal of Great Lakes Research, 42(1), 70-82. https://doi.org/10.1016/j.jglr.2015.11.005
Craig, P., & Moreton, P. (1986). Total mercury, methyl mercury and sulphide levels in British estuarine sediments—III. Water Research, 20(9), 1111-1118. https://doi.org/10.1016/0043-1354(81)90180-9
Visha, A., Gandhi, N., Bhavsar, S. P., & Arhonditsis, G. B. (2016). Guiding fish consumption advisories for Lake Ontario: A Bayesian hierarchical approach. Journal of Great Lakes Research, 42(1), 70-82. https://doi.org/10.1016/j.jglr.2015.11.005
Hope, B. K., Lut, A., Aldrich, G., & Rubin, J. (2007). Environmental management with knowledge of uncertainty: a methylmercury case study. Integrated Environmental Assessment and Management: An International Journal, 3(1), 144-149. https://doi.org/10.1897/1551-3793(2007)3[144:EMWKOU]2.0.CO;2
Cohen, J., Bellinger, D., Connor, W., Krisetherton, P., Lawrence, R., Savitz, D., Shaywitz, B., Teutsch, S., & Gray, G. (2005). A Quantitative Risk–Benefit Analysis of Changes in Population Fish Consumption. American Journal of Preventive Medicine, 29(4), 325-325. https://doi.org/10.1016/j.amepre.2005.07.003
Rheinberger, C. M., & Hammitt, J. K. (2012). Risk Trade-Offs in Fish Consumption: A Public Health Perspective. Environmental Science & Technology, 46(22), 12337-12346. https://doi.org/10.1021/es302652m
Cabana, G., & Rasmussen, J. B. (1996). Comparison of aquatic food chains using nitrogen isotopes. Proceedings of the National Academy of Sciences, 93(20), 10844-10847. https://doi.org/10.1073/pnas.93.20.10844
Paterson, G., Drouillard, K. G., & Haffner, G. D. (2006). Quantifying resource partitioning in centrarchids with stable isotope analysis. Limnology and Oceanography, 51(2), 1038-1044. https://doi.org/10.4319/lo.2006.51.2.1038
Vander Zanden, M. J., Cabana, G., & Rasmussen, J. B. (1997). Comparing trophic position of freshwater fish calculated using stable nitrogen isotope ratios (δ15N) and literature dietary data. Canadian Journal of Fisheries and Aquatic Sciences, 54(5), 1142-1158. https://doi.org/http://doi.org/10.1139/cjfas-54-5-1142
Sullivan, M. J., & Moncreiff, C. A. (1990). Edaphic algae are an important component of salt marsh food-webs: evidence from multiple stable isotope analyses. Marine ecology progress series, 62(1), 149-159. https://doi.org/http://dx.doi.org/10.3354/meps062149
Schwinghamer, P., Tan, F. C., & Gordon Jr, D. C. (1983). Stable carbon isotope studies on the Pecks Cove mudflat ecosystem in the Cumberland Basin, Bay of Fundy. Canadian Journal of Fisheries and Aquatic Sciences, 40(S1), s262-s272. https://doi.org/10.1139/f83-288
Pastershank, G. M. (2002). Unifying ecosystem concepts and mercury biomagnification in an estuarine environment using stable isotopes (delta-¹³C and delta ¹⁵N). National Library of Canada, Ottawa. https://doi.org/http://doi.org/10.20381/ruor-16286
MacAvoy, S., Macko, S., McIninch, S., & Garman, G. (2000). Marine nutrient contributions to freshwater apex predators. Oecologia, 122(4), 568-573. https://doi.org/10.1007/s004420050980
MacAvoy, S. E., Macko, S. A., & Garman, G. C. (2001). Isotopic turnover in aquatic predators: quantifying the exploitation of migratory prey. Canadian Journal of Fisheries and Aquatic Sciences, 58(5), 923-932. https://doi.org/10.1139/f01-045
Bucci, J., Showers, W., Rebach, S., DeMaster, D., & Genna, B. (2007). Stable isotope analyses (δ15N and δ13C) of the trophic relationships ofCallinectes sapidus in two North Carolina estuaries. Estuaries and Coasts, 30(6), 1049-1059. https://doi.org/10.1007/BF02841395
Carrozzo, L., Potenza, L., Carlino, P., Costantini, M. L., Rossi, L., & Mancinelli, G. (2014). Seasonal abundance and trophic position of the Atlantic blue crab Callinectes sapidus Rathbun 1896 in a Mediterranean coastal habitat. Rendiconti Lincei, 25(2), 201-208. https://doi.org/10.1007/s12210-014-0297-x
Szczebak, J. T., & Taylor, D. L. (2011). Ontogenetic patterns in bluefish (Pomatomus saltatrix) feeding ecology and the effect on mercury biomagnification. Environmental Toxicology and Chemistry, 30(6), 1447-1458. https://doi.org/10.1002/etc.516
Melville, A. J., & Connolly, R. M. (2005). Food webs supporting fish over subtropical mudflats are based on transported organic matter not in situ microalgae. Marine Biology, 148(2), 363-371. https://doi.org/10.1007/s00227-005-0083-5
Gu, B., Schelske, C., & Hoyer, M. (1996). Stable isotopes of carbon and nitrogen as indicators of diet and trophic structure of the fish community in a shallow hypereutrophic lake. Journal of fish biology, 49(6), 1233-1243. https://doi.org/10.1111/j.1095-8649.1996.tb01792.x
Power, M., Power, G., Caron, F., Doucett, R. R., & Guiguer, K. R. A. (2002). Growth and dietary niche in Salvelinus alpinus and Salvelinus fontinalis as revealed by stable isotope analysis. In P. Magnan, C. Audet, H. Glémet, M. Legault, M. A. Rodríguez, & E. B. Taylor (Eds.), Ecology, behaviour and conservation of the charrs, genus Salvelinus (pp. 75-85). Springer Netherlands. https://doi.org/10.1007/978-94-017-1352-8_6
Jardine, T. D., Gray, M. A., McWilliam, S. M., & Cunjak, R. A. (2005). Stable isotope variability in tissues of temperate stream fishes. Transactions of the American Fisheries Society, 134(5), 1103-1110. https://doi.org/10.1577/T04-124.1
Keough, J. R., Sierszen, M. E., & Hagley, C. A. (1996). Analysis of a Lake Superior coastal food web with stable isotope techniques. Limnology and Oceanography, 41(1), 136-146. https://doi.org/10.4319/lo.1996.41.1.0136
McHugh, P., Budy, P., Thiede, G., & VanDyke, E. (2008). Trophic relationships of nonnative brown trout, Salmo trutta, and native Bonneville cutthroat trout, Oncorhynchus clarkii utah, in a northern Utah, USA river. Environmental Biology of Fishes, 81(1), 63. https://doi.org/10.1007/s10641-006-9171-8
French, W. E., Vondracek, B., Ferrington Jr, L. C., Finlay, J. C., & Dieterman, D. J. (2014). Winter feeding, growth and condition of brown trout Salmo trutta in a groundwater-dominated stream. Journal of freshwater ecology, 29(2), 187-200. https://doi.org/10.1080/02705060.2013.847868
Garcia, E., & Carignan, R. (2005). Mercury concentrations in fish from forest harvesting and fire‐impacted Canadian Boreal lakes compared using stable isotopes of nitrogen. Environmental Toxicology and Chemistry: An International Journal, 24(3), 685-693. https://doi.org/10.1897/04-065r.1
Sierszen, M. E., Keough, J. R., & Hagley, C. A. (1996). Trophic analysis of ruffe (Gymnocephalus cernuus) and white perch (Morone americana) in a Lake Superior coastal food web, using stable isotope techniques. Journal of Great Lakes Research, 22(2), 436-443. https://doi.org/10.1016/S0380-1330(96)70968-1
Fry, B., Mumford, P. L., Tam, F., Fox, D. D., Warren, G. L., Havens, K. E., & Steinman, A. D. (1999). Trophic position and individual feeding histories of fish from Lake Okeechobee, Florida. Canadian Journal of Fisheries and Aquatic Sciences, 56(4), 590-600. https://doi.org/10.1139/f98-204
Harvey, C. J., & Kitchell, J. F. (2000). A stable isotope evaluation of the structure and spatial heterogeneity of a Lake Superior food web. Canadian Journal of Fisheries and Aquatic Sciences, 57(7), 1395-1403. http://doi.org/10.1139/cjfas-57-7-1395
Vander Zanden, M. J., & Vadeboncoeur, Y. (2002). Fishes as integrators of benthic and pelagic food webs in lakes. Ecology, 83(8), 2152-2161. http://doi.org/10.1890/0012-9658(2002)083[2152:FAIOBA]2.0.CO;2
Stephenson, R., Tan, F., & Mann, K. (1986). Use of stable carbon isotope ratios to compare plant material and potential consumers in a seagrass bed and a kelp bed in Nova Scotia, Canada. Marine Ecology Progress Series, 30, 1-7. http://www.jstor.org/stable/24817006
Tucker, J., Sheats, N., Giblin, A., Hopkinson, C., & Montoya, J. (1999). Using stable isotopes to trace sewage-derived material through Boston Harbor and Massachusetts Bay. Marine Environmental Research, 48(4-5), 353-375. https://doi.org/10.1016/S0141-1136(99)00069-0
Steneck, R. S., Vavrinec, J., & Leland, A. V. (2004). Accelerating trophic-level dysfunction in kelp forest ecosystems of the western North Atlantic. Ecosystems, 7(4), 323-332. https://doi.org/10.1007/s10021-004-0240-6
Freedman, J., Curry, R., & Munkittrick, K. (2012). Stable isotope analysis reveals anthropogenic effects on fish assemblages in a temperate reservoir. River research and applications, 28(10), 1804-1819. https://doi.org/10.1002/rra.1576
Tom, K. R., Newman, M. C., & Schmerfeld, J. (2010). Modeling mercury biomagnification (South River, Virginia, USA) to inform river management decision making. Environmental Toxicology and Chemistry: An International Journal, 29(4), 1013-1020. https://doi.org/10.1002/etc.117
Froese, R., D. Pauly. (2014). FishBase. www.fishbase.org
Zeug, S. C., & Winemiller, K. O. (2008). Evidence supporting the importance of terrestrial carbon in a large‐river food web. Ecology, 89(6), 1733-1743. https://doi.org/10.1890/07-1064.1
Dodder, N. G., Strandberg, B., & Hites, R. A. (2002). Concentrations and spatial variations of polybrominated diphenyl ethers and several organochlorine compounds in fishes from the northeastern United States. Environmental Science & Technology, 36(2), 146-151. https://doi.org/10.1021/es010947g
Heady, W. N., & Moore, J. W. (2013). Tissue turnover and stable isotope clocks to quantify resource shifts in anadromous rainbow trout. Oecologia, 172(1), 21-34. https://doi.org/10.1007/s00442-012-2483-9
Angradi, T. R. (1994). Trophic linkages in the lower Colorado River: multiple stable isotope evidence. Journal of the North American Benthological Society, 13(4), 479-495. https://doi.org/10.2307/1467845
Dempson, J. B., & Power, M. (2004). Use of stable isotopes to distinguish farmed from wild Atlantic salmon, Salmo salar. Ecology of Freshwater Fish, 13(3), 176-184. https://doi.org/10.1111/j.1600-0633.2004.00057.x
Piraino, M. N., & Taylor, D. L. (2009). Bioaccumulation and trophic transfer of mercury in striped bass () and tautog () from the Narragansett Bay (Rhode Island, USA). Marine Environmental Research, 67(3), 117. https://doi.org/10.1016/j.marenvres.2008.12.006
Hughes, E. H., & Sherr, E. B. (1983). Subtidal food webs in a georgia estuary: δ13C analysis. Journal of Experimental Marine Biology and Ecology, 67(3), 227-242. https://doi.org/10.1016/0022-0981(83)90041-2
Winemiller, K. O., Akin, S., & Zeug, S. C. (2007). Production sources and food web structure of a temperate tidal estuary: integration of dietary and stable isotope data. Marine Ecology Progress Series, 343, 63-76. https://doi.org/http://doi.org/10.3354/meps06884
Overman, N. C., & Parrish, D. L. (2001). Stable isotope composition of walleye: 15N accumulation with age and area-specific differences in δ13C. Canadian Journal of Fisheries and Aquatic Sciences, 58(6), 1253-1260. https://doi.org/10.1139/f01-072
Buchheister, A., & Latour, R. J. (2010). Turnover and fractionation of carbon and nitrogen stable isotopes in tissues of a migratory coastal predator, summer flounder (Paralichthys dentatus). Canadian Journal of Fisheries and Aquatic Sciences, 67(3), 445-461. https://doi.org/10.1139/F09-196
Weinstein, M. P., Litvin, S. Y., Guida, V. G., & Chambers, R. C. (2009). Is global climate change influencing the overwintering distribution of weakfish Cynoscion regalis? Journal of fish biology, 75(3), 693-698. https://doi.org/10.1111/j.1095-8649.2009.02304.x
Fry, B. (1988). Food web structure on Georges Bank from stable C, N, and S isotopic compositions. Limnology and Oceanography, 33(5), 1182-1190. https://doi.org/10.4319/lo.1988.33.5.1182

Table 1

PBTs modelled for each data source.
EMAP	NRSA	NCA
2,4'-DDE 2,4'-DDD 2,4'-DDT 4,4'-DDD Aldrin Alpha Chlordane Cis-Nonachlor Dieldrin Gamma Chlordane Hexachlorobenzene Mercury Oxychlordane Total PCBs Trans-Nonachlor	Alpha Chlordane Cis-Nonachlor Dieldrin Endosulfan II Endrin Endrin Ketone Gamma Chlordane Heptachlor Heptachlorepoxide Mercury Oxychlordane Total PCBs Total PFCs Trans-Nonachlor	2,4'-DDE 2,4'-DDT 4,4'-DDD 4,4'-DDE 4,4'-DDT 2,6'-Dimethylnaphthalene Acenaphthylene Anthracene Benzo(b)fluoranthene Biphenyl Chrysene Fluoranthene Mercury Methylnaphthalene Naphthalene Total PCBs Pyrene Trans-Nonachlor

Table 2

Fish, waterbody, and watershed variables considered by data source. (See Appendix A for definitions and units.)
EMAP	NRSA	NCA
Estimated Trophic Level Species Lipids Total Zooplankton Abundance Alkalinity DOC Chlorophyll a Percent Human Disturbance Percent Mining Percent Urban Watershed Area Percent Agriculture Latitude Longitude	Fish Length Species Percent Ephemeroptera Total Nitrogen pH DOC Watershed Area Percent Agriculture Percent Urban Percent Shrubgrass Percent Wetland Percent Impervious Barrier Disturbances Riparian Vegetation Longitude Latitude	Chlorophyll a Total Nitrogen Sediment Toxicity Abundance of Tolerant Benthic Organisms Abundance of Sensitive Benthic Organisms Tissue Type Percent Agriculture Percent Forest Percent Impervious Percent Urban Percent Wetlands Latitude Longitude Lipids Length Species

No competing interests reported.

Appendix.docx

Download PDF

Journal Publication

published 26 Jun, 2023

Read the published version in Environmental Modeling & Assessment →

Editorial decision: Major revision
30 Oct, 2022
Reviews received at journal
16 Aug, 2022
Reviewers agreed at journal
25 Jul, 2022
Reviewers invited by journal
05 Jul, 2022
Editor assigned by journal
01 Jul, 2022
Submission checks completed at journal
17 Jun, 2022
First submitted to journal
15 Jun, 2022

You are reading this latest preprint version

Patterns of co-contamination in freshwater and marine fish of the northeastern United States

Status:

Journal Publication

Version 1

Abstract

Figures

1.0 Introduction

2.0 Methods

2.1 Data Description

2.2 Data Analysis

2.3 Model Selection

3.0 Results

3.1 EMAP Lakes Data

3.2 NRSA River Data

3.3 NCA Data

4.0 Discussion

Declarations

References

Tables

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1