Comparison of different toxic effect sub-models in ecosystem modelling used for ecological effect assessments and water quality standard setting.

Ecosystem models, combining a food web model with a toxic effect sub-model, have been proposed to incorporate ecological interactions in ecological effect assessments. Toxic effect sub-models in different studies tend to differ in (1) the used single-species toxicity data, (2) the effects they consider, (3) the concentration-effect function used. In this paper, we constructed four ecosystem models, each with a different toxic effect sub-model, and tested their capacity to predict biomass changes, and no observed effect concentrations (NOECs) established in an experimental microcosm. For most populations, these predictions depended heavily on the type of ecosystem model. The ecosystem model with a toxic effect sub-model incorporating mortality effects using a logistic concentration-effect function made accurate predictions for most populations. Additional incorporation of sub-lethal effects did not result in better predictions. Ecosystem models using linear concentration-effect functions predict biomass decreases at concentrations that are four times lower than the observed NOECs.