Development of species sensitivity distributions and estimation of HC5 of organochlorine pesticides with five statistical approaches

Abstract

Eighteen organochlorine pesticides (OCPs) were studied to develop species sensitivity distributions (SSDs) and calculate hazardous concentration thresholds for 5% of species (HC5), using both parametric (log-normal and log-logistic) and nonparametric bootstrap methods. In order to avoid picking repetitive values in each resample when performing bootstrap, and to determine the influence of fluctuation of toxicity data of single species on the SSDs and HC5, a modified bootstrap method was introduced, which can generate unrepetitive sampling data other than original elements in datasets. This method can enlarge a dataset without any assumption of a special distribution. Combined with parametric methods, modified bootstrap was also used to develop SSDs and determine HC5. The HC5 estimated by five approaches coincide well with each other with good positive correlation. Even if there is intra-species variation in a certain range of toxicity data; SSDs and HC5 are not very sensitive to the local fluctuation of toxicity of single species. The studied OCPs were classified according to their estimated HC5. A lower HC5 indicates higher ecological toxicity potentials. Endrin, DDTs and Endosulfan are OCPs with very high ecological toxicity potential. alpha-HCH has the lowest ecological toxicity potential in the studied OCPs. For OCPs with high ecological potential, more attention should be paid to their ecological risk.

Aquatic predicted no-effect-concentration derivation for perfluorooctane sulfonic acid

“…The AF approach is simple and is applied when few toxicity data are available. In contrast, the statistical extrapolation techniques are based on species sensitivity distributions (SSDs), which are increasingly used to determine the ecological adverse effects of chemicals [6,7]. Generally, various species in an ecosystem possess different sensitivities to a given toxic substance.…”

Ecological Risk Assessment of Agricultural Pesticides throughout the Shadegan Wetland, Iran

“…The mean and the standard deviation of the in-transformed dataset of a number of toxicity end points (LC 50 or EC 50 ) of a specific pesticide were used to estimate the parameters which describe the distribution. From this distribution, in order to protect 95 % of species, the hazardous concentration for 5% of the species (HC 5 ) in an ecosystem was calculated using the following model based on species sensitivity distributions (Aldenberg and Slob, 1993;Steen et al, 1999;Wang et al, 2008).…”

Setting Water Quality Criteria in China: Approaches for Developing Species Sensitivity Distributions for Metals and Metalloids

“…For this purpose, SSDs are visualized as a plot of a cumulative distribution function against the logarithm of the concentrations of toxicity data (Solomon et al 2000). Also, SSDs offer greater statistical confidence in calculating a predicted no effect concentration (PNEC) for use in risk assessments than does the commonly used quotient approaches Wheeler et al 2002;Wang et al 2008). The latter approaches are usually calculated by applying a safety factor to the statistical summary of a single toxicity test such as no observed effect concentration (NOEC) or a 50%-effect concentration (EC 50 ) (van Dam et al 2012).…”

“…The bootstrap regression was further developed by combining a nonparametric bootstrap with a parametric log-logistic model to solve the difficulty of the limited toxicity data available . Based on the standard bootstrap, we applied artificial interpolations to avoid repetitive values in each resample and to expand the data beyond the limited original datasets (Wang et al 2008).…”