Significant variation in mercury (Hg) bioaccumulation is observed across the diversity of freshwater ecosystems in North America. While there is support for the major drivers of Hg bioaccumulation, the relative influence of different external factors can vary widely among waterbodies, which makes predicting Hg risk across large spatial scales particularly challenging. We modeled Hg bioaccumulation by coupling Hg concentrations in more than 21,000 dragonflies collected across the United States from 2008 to 2021 with a suite of chemical (e.g., dissolved organic carbon (DOC), pH, sulfate) and landscape (e.g., soil characteristics, land cover) variables representing external drivers of Hg methylation, transport, and uptake. Model predictions explained 85% of the variation in dragonfly Hg concentrations across the United States. Certain predictor variables were more important than others (e.g., DOC, pH, and percent wetland), and they varied among waterbodies. Variation in Hg bioaccumulation was explained by including habitat and ecosystem type in a hierarchical modeling framework, which confirms the context-dependency of external factors in explaining Hg bioaccumulation across disparate freshwater ecosystems. This continent-scale model provides valuable insights into the processes underlying landscape-scale patterns in Hg exposure risk and demonstrates that drivers of Hg methylation and bioaccumulation are habitat- and ecosystem-dependent.
Keywords: Bayesian predictive modeling; context-dependence; dragonfly larvae; mercury biogeochemistry; mercury risk; water quality.