Species Sensitivity Distribution revisited: a Bayesian nonparametric approach

We present a novel approach to ecological risk assessment by recasting the Species Sensitivity Distribution (SSD) method within a Bayesian nonparametric (BNP) framework. Widely mandated by environmental regulatory bodies globally, SSD has faced criticism due to its historical reliance on parametric assumptions when modeling species variability. By adopting nonparametric mixture models, we address this limitation, establishing a statistically robust foundation for SSD. Our BNP approach offers several advantages, including its efficacy in handling small datasets or censored data, which are common in ecological risk assessment, and its ability to provide principled uncertainty quantification alongside simultaneous density estimation and clustering. We utilize a specific nonparametric prior as the mixing measure, chosen for its robust clustering properties, a crucial consideration given the lack of strong prior beliefs about the number of components. Through simulation studies and analysis of real datasets, we demonstrate the superiority of our BNP-SSD over classical SSD methods. We also provide a BNP-SSD Shiny application, making our methodology available to the Ecotoxicology community. Moreover, we exploit the inherent clustering structure of the mixture model to explore patterns in species sensitivity. Our findings underscore the effectiveness of the proposed approach in improving ecological risk assessment methodologies.

💡 Research Summary

The paper introduces a Bayesian non‑parametric (BNP) mixture‑model framework for Species Sensitivity Distribution (SSD), a cornerstone method in ecological risk assessment used to estimate hazardous concentrations such as HC5 from limited Critical Effect Concentration (CEC) data. Traditional SSD implementations rely on a single parametric family (log‑normal, log‑logistic, triangular, Burr III, etc.), an approach that is increasingly criticized because (i) small sample sizes (often < 15 species) provide little power to justify any particular distribution, (ii) real‑world data frequently exhibit multimodal patterns reflecting taxonomic, habitat or mode‑of‑action heterogeneity, and (iii) censored observations (left, right, interval) are either discarded or crudely imputed.

To overcome these limitations, the authors recast SSD as a non‑parametric mixture of normal kernels on log‑transformed CECs, with the mixing distribution modeled by a Normalized Random Measure with Independent Increments (NRMI). Specifically, they adopt the Normalized Stable Process (a special case of NRMI) as the prior for the mixing measure, characterized by a stability parameter γ = 0.4 and a base measure P₀ that factorizes into independent priors for location (μ) and scale (σ). The location prior is a normal distribution with hyper‑priors on its mean and precision, while σ receives a uniform prior on