Abstract For over two decades, the Target Lipid Model (TLM) has been used to characterize the aquatic toxicity of hydrocarbons, petroleum products, ionic and other non-ionic organic chemicals across a wide range of organisms. The TLM has also been applied to derive hazard concentrations intended to protect 95% of aquatic species (i.e., HC5). The TLM database has expanded to include multiple acute and chronic data (over 100 species from 8 major taxonomic groups) across over 400 individual chemicals for both standard apical (e.g., mortality, reproduction, growth) and relevant non-apical endpoints (e.g., cardiotoxicity). Given these improvements, there is an opportunity to re-evaluate the lipid-water partitioning models so they are common across the TLM and PETROTOX models. The ability of each partitioning model to predict acute and chronic effects was evaluated using statistical goodness-of-fit descriptors in addition to considerations for model accessibility and ease of application. While all lipid-water partitioning models performed similarly, the Abraham-type formulation of the Polyparameter Target Site Model (pp-TSM) has been selected as the preferred lipid-water partitioning model. Previous estimations of chronic effects HC5s used a statistical extrapolation procedure with distributional constraints or applied boot-strapped error estimation to species sensitivity only. To eliminate distributional constraints and account for additional model prediction errors, an alternative approach is proposed to calculate HC5s using boot-strapped error estimation of the TLM calibration to derive prediction errors which include the uncertainty in model parameters and the empirical toxicity distribution. Predicted HC5s using this method are protective of 95% of species based on comparison to measured acute, chronic, and sublethal data. Advantages of this calculation method are discussed along with implications for future TLM applications.
Fanelli et al. (Mon,) studied this question.