Confronting pesticide exposure predictions from different models to observations from a monitoring study in small freshwater streams in Germany

Plant protection products are integral to European agriculture but can cause unwanted environmental impacts. Before authorisation, predicted concentrations in environmental compartments are compared with effect thresholds in a regulatory risk assessment. This study evaluates the agreement between predicted and measured concentrations for the established FOCUS surface water models (Steps 1-3) and the recently published PEC-CKB model. Model results were compared with monitoring data from lowland streams in Germany, and particular attention was paid to the models' conservatism. The conservative character of FOCUS Step 1 can be confirmed, but underestimations were observed for FOCUS Step 2 and 3 models. PEC-CKB results are similar to those of the higher-tier FOCUS models, while having lower model complexity and requiring less input data. Using real application rates and landscape information generally improved model predictions by nearly halving the bias, but led to increased underestimations of measured concentrations. Linking prospective and retrospective environmental risk assessment (ERA) by incorporating real data can make prospective ERA more realistic and identify opportunities for simplification. Finally, we discuss the challenges in evaluating prediction models for pesticide concentrations in surface waters, particularly with regard to the environmental variability of measured concentrations.