Predictive modelling of metal mixture toxicity to Daphnia magna populations

Current practice of environmental risk assessment lacks ecological realism, because it depends mostly on toxicity of single substances to individual organisms. It is desirable to develop mechanistic, predictive models that take mixture toxicity on higher levels of organization into account. We conducted a population experiment with Daphnia magna exposed to Cu-Ni-Zn mixtures and the single metals, in order to calibrate a Dynamic Energy Budget Individual-Based Model (DEB-IBM) with single-metal population data and generate blind predictions on mixture toxicity. Metals with different physiological modes of action (PMoA) can be implemented independently in the DEB-IBM, without making further assumptions concerning mixture toxicity. For metals with the same PMoA, we assume no interactions between metals.We first explored approaches to calibrate a DEB-IBM with population-level data, which imposes constraints on parameter estimation as compared to conventional DEB-IBM calibration with individual-level data.We further evaluated the predictive capacity of the DEBbased approach in comparison with common reference models IA and Concentration Addition (CA). While the performance of CA and IA was concentration-dependent, the DEB-IBM has the capacity to capture such trends, because mixture toxicity is an emergent property and interactions between organisms can be taken into account. We conclude that an approach based on DEB-IBMs is a promising way forward to generate predictive models and enhance understanding of mixture toxicity at higher levels of biological organization.Current practice of environmental risk assessment lacks ecological realism, because it depends mostly on toxicity of single substances to individual organisms. It is desirable to develop mechanistic, predictive models that take mixture toxicity on higher levels of organization into account. We conducted a population experiment with Daphnia magna exposed to Cu-Ni-Zn mixtures and the single metals, in order to calibrate a Dynamic Energy Budget Individual-Based Model (DEB-IBM) with single-metal population data and generate blind predictions on mixture toxicity. Metals with different physiological modes of action (PMoA) can be implemented independently in the DEB-IBM, without making further assumptions concerning mixture toxicity. For metals with the same PMoA, we assume no interactions between metals.We first explored approaches to calibrate a DEB-IBM with population-level data, which imposes constraints on parameter estimation as compared to conventional DEB-IBM calibration with individual-level data.We further evaluated the predictive capacity of the DEBbased approach in comparison with common reference models IA and Concentration Addition (CA). While the performance of CA and IA was concentration-dependent, the DEB-IBM has the capacity to capture such trends, because mixture toxicity is an emergent property and interactions between organisms can be taken into account. We conclude that an approach based on DEB-IBMs is a promising way forward to generate predictive models and enhance understanding of mixture toxicity at higher levels of biological organization.C