Threshold models for species sensitivity distributions applied to aquatic risk assessment for zinc

Species sensitivity distributions (SSDs) are used in ecological risk assessment to derive maximum acceptable concentrations of toxicants in the environment from a limited set of ecotoxicity data obtained in the laboratory. Such distributions usually employ continuous bell-shaped functions such as the normal and the logistic distribution, which have the disadvantage that an arbitrary cut-off value must be chosen (usually the 5-percentile) to designate the concentration below which the fraction of species exposed above their no-effect level is considered acceptably small. In this paper the possibility is explored of introducing a true no-effect principle in the SSD framework by considering models with a finite lower threshold. Four of these distributions are elaborated, the uniform, triangular, exponential and Weibull distributions. The mathematical representations of these functions were re-parameterized allowing direct estimation of the threshold parameter by nonlinear regression. By way of example, a data set comprising chronic ecotoxicity of zinc to 21 different aquatic organisms was used. The exponential distribution did not describe the data well. The other distributions provided estimates for HC0 (hazardous concentration for none of the species) between 1.66 and 7.83 mug/l. The triangular distribution fitted best to the data and was consistent with previous models. Since threshold-SSDs incorporate a true no-effect level they may be better communicable as a principle for environmental protection in comparison to the approach based on '95% protection'. (C) 2002 Elsevier Science B.V. All rights reserved.