A re-evaluation of fifteen years of European risk assessment using effect models.

Ecological risk assessments of chemicals can be informed by a suite of effect models, including population and food web models. In the risk assessments conducted under EU regulation 793/93/EC, however, applications of such effect models are extremely scarce and toxicity-extrapolation approaches are often used instead. The objective of the present study was to re-evaluate these risk assessments using two types of effect models: species sensitivity distributions (SSDs, non-mechanistic), and food web models (mechanistic). Species sensitivity distributions significantly fitted the available toxicity data for up to 35% of the chemicals, depending on the trophic levels included and the amount of data available. Median hazardous concentrations for 5% of the species (HC5-50) estimated by the SSDs were less accurate predictors of measured community-level no observed effect concentration than food web model-derived HC5-50s, albeit data were available for seven chemicals only. For datasets with more than 10 data points, the 90% confidence interval of the estimated HC5s was narrower for the food web modeling approach than for the SSD approach. The HC5-50s predicted by the two approaches were two to five times (metals) and 10 to 100 times (organic chemicals) higher than the predicted no effect concentrations (PNECs) for the aquatic environment listed in the risk assessment reports. This suggests that the derived PNECs are protective for aquatic ecosystems.