Is it safe to eat fish from the Great Lakes? An adaptive modelling-monitoring framework to assess compliance with consumption advisories

Abstract In this study, we conducted a retrospective analysis of the spatio-temporal trends of fish contamination in the Canadian Great Lakes. We subsequently formulated consumption advisories that explicitly account for all sources of uncertainty, such as model error, sampling bias, and natural variability of fish assemblages. Our analysis generated exceedance frequency maps of the tolerable daily intakes (TDIs), which showed that mercury (Hg) concentrations are generally at a safer level for consumption, whilst polychlorinated biphenyls (PCBs) continue to result in restrictive advisories. Specifically, the initial decline in PCBs has transitioned into a steady state, possibly due to the combined effects of external contamination sources and internal abiotic and biotic factors. Our assessment of the PCB decline rates required to comply with TDI thresholds over the next twenty years suggest that a small degree of reduction is needed for Walleye (Sander vitreus) across all sampled sites, whereas much faster decay rates are needed for Lake Trout (Salvelinus namaycush), especially in lower Lake Huron, Lake Erie, and Lake Ontario. We also detected a distinct Hg gradient whereby the southernmost waterbody, Lake Erie, had the lowest average concentrations, whereas the highest levels amongst the fish species sampled were registered in St. Lawrence River. Finally, we show the ability of our Bayesian approach to fish consumption advisories to produce time-specific, highly customizable, risk-assessment statements that account for the inherent variability in natural fish communities and the variant degree of empirical evidence about the contamination history in different locations around the Great Lakes.