Risk Assessment using the Species Sensitivity Distribution Method: Data Quality versus Data Quantity

Species sensitivity distributions (SSDs) are cumulative distributions of measures of species sensitivity to a stressor or toxicant, and are used to estimate concentrations that will protect p% of a community (PCp). There is conflict between the desire to use high-quality sensitivity data in SSDs, and to construct them with a large number of species forming a representative sample. Trade-offs between data quality and quantity were investigated using the effects of increasing salinity on the macroinvertebrate community from the Hunter River catchment, in eastern Australia. Five SSDs were constructed, representing five points along a continuum of data quality versus data quantity and representativeness. This continuum was achieved by the various inclusion/exclusion of censored data, nonmodeled data, and extrapolation from related species. Protective concentrations were estimated using the Burr type III distribution, Kaplan-Meier survival function, and two Bayesian statistical models. The dominant taxonomic group was the prime determinant of protective concentrations, with an increase in PC95 values resulting from a decrease in the proportion of Ephemeropteran species included in the SSD. In addition, decreases in data quantity in a SSD decreased community representativeness. The authors suggest, at least for salinity, that the inclusion of right censored data provides a more representative sample of species that reflects the natural biotic assemblage of an area to be protected, and will therefore improve risk assessment.