Considering Data Uncertainty in Species Sensitivity Distribution for Ecological Risk Assessment of Chemicals

摘要

Ecological risk assessment for chemical management favors multi-species as the protective object. Since different species respond differently, species sensitivity distribution (SSD) is applied to delineate the impact and to infer criterion concentration of a chemical. A SSD is usually established with ecotoxicity test data of different species, usually No-Observed-Effect-Concentrations (NOECs) derived from chronic bioassays. Based on the SSD, the concentration that begets impairment in 5% of the total species in the ecosystem, denoted as HC5, together with its 95% lower limit value, is used for regulation. The dependence of SSD on NOECs entails that the uncertainties in NOEC values will skew the shape of SSD and consequently the HC5, as is of much importance since NOEC value is generally prone to errors. This study demonstrated the bias in HC5 with Monte Carlo simulations by assuming different uncertainty levels in NOECs. With increasing NOEC uncertainty, the HC5 and its lower limit were found to shift downwards substantially, implying underestimation of the ecological risk if NOEC uncertainty is disregarded. By incorporating the NOEC variances during the HC5 inference procedure, the bias could be adequately corrected. However, simulating SSD through Monte Carlo approach is probably indispensable in the case of irregular NOECs uncertainties.