Alternatives to the use of safety factors have been proposed, although no method is completely free from uncertainty. The use of data to estimate ACRs for a particular chemical has been discussed above, and is the oldest example of attempting to develop a robust data set for estimating beyond the tested species. More recently, the use of species sensitivity distributions (SSDs) has been gaining favor as a more accurate estimate of differences in sensitivity among species. However, no consensus exists on the number of species required to develop a sufficiently robust distribution, and what the minimum degree of taxonomic relationship should be among the species tested (i.e., is it sufficient to be of different genera or should they represent different orders?). The US EPA requires a minimum of eight genera representing at least three trophic levels when using SSDs for derivation of water-quality criteria or mitigation goals. In the European Union, eight to ten genera are required if an SSD is to be used instead of an interspecies extrapolation factor, and other jurisdictions are considering as many as 30 species. There is general agreement, however, that aquatic and terrestrial species should be analyzed separately, and within terrestrial systems animals and plants should have separate estimates (note that aquatic toxicity threshold estimates generally use algae as a surrogate for all aquatic plants, although it is likely that rooted macro-phytes (larger plants) will respond differently to pollutants in the water column, particularly those that are emergent above the water line).
Another source of uncertainty when using distributions of species responses (Figure 2) is the point on the curve that is chosen as the toxicity threshold. There is general agreement that the value that defines the concentration where only 5% of the species would show a response is appropriate for management purposes (i.e., 95% of the species would be protected). However, there is disagreement on how the uncertainty in this estimate should be incorporated. It is recognized that the species toxicity endpoints that are used to derive the SSD contain the variability discussed in the above sections on intra-species extrapolations and, therefore, that a curve based on single points for each species will contain significant uncertainty. Ideally, more than one value would be generated for each species and the species (or genus) mean value used to derive the distribution curve. The US EPA
uses the concentration where 5% of the species are protected (the 5th percentile of the distribution) for setting water quality criteria, while other decisions are based on the lower confidence limit of this value. The size of the confidence interval around the 5% value depends on the number of data points (i.e., species) used to generate the SSD and the inherent variability in their responses. The difference between using eight, ten, or 20 species appears to be a factor of 3. Some jurisdictions may consider using 3 as a safety factor on the 5% estimate if less than ten species are tested.
Other methods of uncertainty analysis that are applied to exposure estimates cannot be used in effects analysis. For example, uncertainty bounds could be applied, providing estimates of risk using the most sensitive and least sensitive organism. This bounds the possibilities of where the true risk lies, and allows the risk manager to determine a level of conservatism based on degree of risk aversion rather than assuming scientific certainty. The difficulty with this approach is that there still is no empirical way of determining what is the most or least sensitive organism and what exposure concentration would represent the threshold where these organisms would begin to show effects.
Was this article helpful?