The Use of Mechanistic Population Models in Metal Risk Assessment: Combined Effects of Copper and Food Source on Lymnaea stagnalis Populations

Environmental risk assessment (ERA) of chemicals aims to protect populations, communities, and ecosystems. Population models are considered more frequent in ERA because they can bridge the gap between the individual and the population level. Lymnaea stagnalis (the great pond snail) is an organism that is particularly sensitive to various metals, including copper (Cu). In addition, the sensitivity of this species to Cu differs between food sources. The first goal of the present study was to investigate whether we could explain the variability in sensitivity between food sources (lettuce and fish flakes) at the individual level with a dynamic energy budget (DEB) model. By adapting an existing DEB model and calibrating it with Cu toxicity data, thereby combining information from 3 studies and 2 endpoints (growth and reproduction), we put forward inhibition of energy assimilation as the most plausible physiological mode of action (PMoA) of Cu. Furthermore, the variation in Cu sensitivity between both food sources was considerably lower at the PMoA level than at the individual level. Higher Cu sensitivity at individual level under conditions of lower food quality or availability appears to emerge from first DEB principles when inhibition of assimilation is the PMoA. This supports the idea that DEB explained Cu sensitivity variation between food sources. Our second goal was to investigate whether this food source effect propagated to the population level. By incorporating DEB in an individual‐based model (IBM), population‐level effects were predicted. Based on our simulations, the food source effect was still present at the population level, albeit less prominently. Finally, we compared predicted population‐level effect concentration, x% (ECx) values with individual‐level ECx values for different studies. Using the DEB‐IBM, the range of effect concentrations decreased significantly: at the individual level, the difference in chronic EC10 values between studies was a factor of 70 (1.13–78 µg dissolved Cu/L), whereas at the population level the difference was a factor of 15 (2.9–44.6 µg dissolved Cu/L). To improve interstudy comparability, a bioavailability correction for differences in water chemistry was performed with a biotic ligand model. This further decreased the variation, down to a factor of 7.4. Applying the population model in combination with a bioavailability correction thus significantly decreased the variability of chronic effect concentrations of Cu for L. stagnalis. Overall, the results of the present study illustrate the potential usefulness of transitioning to a more modeling‐based environmental risk assessment. Environ Toxicol Chem 2019;00:1–16. © 2019 SETAC