Integrating Metapopulation Dynamics into a Bayesian Network Relative Risk Model: Assessing Risk of Pesticides to Chinook Salmon (<i>Oncorhynchus tshawytscha</i>) in an Ecological Context
Chelsea J. Mitchell, Eric J. Lawrence, Valerie R. Chu, Meagan J. Harris, Wayne G. Landis, Katherine von Stackelberg, John D. Stark
Abstract
Abstract The population level is often the biological endpoint addressed in ecological risk assessments (ERAs). However, ERAs tend to ignore the metapopulation structure, which precludes an understanding of how population viability is affected by multiple stressors (e.g., toxicants and environmental conditions) at large spatial scales. Here we integrate metapopulation model simulations into a regional-scale, multiple stressors risk assessment (Bayesian network relative risk model [BN-RRM]) of organophosphate (OP) exposure, water temperature, and DO impacts on Chinook salmon (Oncorhynchus tshawytscha). A matrix metapopulation model was developed for spring Chinook salmon in the Yakima River Basin (YRB), Washington, USA, including 3 locally adapted subpopulations and hatchery fish that interact with those subpopulations. Three metapopulation models (an exponential model, a ceiling density-dependent model, and an exponential model without dispersal) were integrated into the BN-RRM to evaluate the effects of population model assumptions on risk calculations. Risk was defined as the percent probability that the abundance of a subpopulation would decline from their initial abundance (500 000). This definition of risk reflects the Puget Sound Partnership's management goal of achieving “no net loss” of Chinook abundance. The BN-RRM model results for projection year 20 showed that risk (in % probability) from OPs and environmental stressors was higher for the wild subpopulations—the American River (50.9%–97.7%) and Naches (39.8%–84.4%) spring Chinook—than for the hatchery population (CESRF 18.5%–46.5%) and the Upper Yakima subpopulation (21.5%–68.7%). Metapopulation risk was higher in summer (58.1%–68.7%) than in winter (33.6%–53.2%), and this seasonal risk pattern was conserved at the subpopulation level. To reach the management goal in the American River spring Chinook subpopulation, the water temperature conditions in the Lower Yakima River would need to decrease. We demonstrate that 1) relative risk can vary across a metapopulation's spatial range, 2) dispersal among patches impacts subpopulation abundance and risk, and 3) local adaptation within a salmon metapopulation can profoundly impact subpopulation responses to equivalent stressors. Integr Environ Assess Manag 2021;17:95–109. © 2020 SETAC KEY POINTS Fine-scale risk estimates can be produced by integrating metapopulation model simulations into a Bayesian network relative risk model (BN-RMM). Risk differed across spring Chinook salmon subpopulations in the Yakima River Basin, Washington, USA, highlighting the importance of considering life history differences when making risk predictions for environmental stressors. Water temperature is the risk factor that requires management to prevent loss of Chinook population abundance, given our modeling efforts. Small sample sizes for organophosphate in winter introduced sufficient uncertainty to increase the influence of toxicological effects in winter.