Species non-exchangeability in ecotoxicological probabilistic risk assessment


Current ecotoxicological risk assessment for chemical substances is based on the assumption that tolerances of all species in a specified ecological community are a priori exchangeable for each new substance. We demonstrate non-exchangeability by using a large database of tolerances to pesticides for fish species and extend the standard statistical model for species tolerances to allow for the presence of a single species which is considered non-exchangeable with others. We show how to estimate parameters and adjust decision rules that are used in ecotoxicological risk management. Effects of parameter uncertainty are explored and our model is compared with a previously published less tractable alternative. We conclude that the model and decision rules that we propose are pragmatic compromises between conflicting needs for more realistic modelling and for straightforwardly applicable decision rules.

Journal of the Royal Statistical Society, Series A: Statistics in Society