Pacific Salmon (Oncorhynchus spp.) introduced to the Great Lakes bioaccumulate pollutants over their lifespan. Salmon return to tributaries to spawn and die, and in the process deposit contaminated tissue and eggs. This ecosystem linkage is a key factor influencing the concentration and pattern of polychlorinated biphenyls (PCBs) in stream fish. However, uncertainty exists regarding salmon-mediated transport of heavy metals, such as mercury (Hg), and how watershed and instream factors interact to influence the retention, transformation, and subsequent uptake by stream-resident fish of biotransported contaminants. To address these uncertainties, we conducted a large-scale survey across 15 watersheds in the Upper Great Lakes region. Within each watershed, we sampled one reach with salmon spawners present and one reach with salmon absent. In each stream reach, we collected stream-resident fish and analyzed their tissues for a suite of contaminants including PCBs and Hg. In addition, within each reach we estimated salmon spawner biomass and measured multiple attributes of instream geomorphology and watershed condition. Resident fish contaminant levels were related to various environmental attributes using a suite of generalized linear mixed models with model selection. Model performance and uncertainty were assessed using AICc, which ranks models based upon the principle of parsimony.
Our results suggest that contaminant concentrations in stream-resident fish vary depending on the contaminant of interest, biological variables, and elements of instream geomorphology and watershed condition. For PCBs, all of the top models included presence or absence of salmon spawners, salmon biomass, and species identity as important predictors of resident fish contaminant burden. In addition, instream and landscape factors were included in 50% of the top models suggesting that environmental context also influences this relationship. The best model explained 88% of total variation in PCBs. In contrast, the influence of salmon on Hg concentrations was minimal and salmon variables were only included in three of the top models. However, similar to PCB models, species identity was an important variable in each of the top models. We also found evidence that watershed variables, including % wetland cover and % forest cover, positively influenced this relationship. However, the Hg models explained only 26% of the overall variance, suggesting that unmeasured factors contributed to the patterns. Our preliminary modeling effort suggests the role of salmon as a contaminant vector varies depending upon the contaminant considered, but elements of environmental heterogeneity can modulate biotransport by salmon.