Bayesian species sensitivity distribution modeling for microplastic particles: Integrating particle characteristics...

Because of the global concern about the environmental impacts of microplastic particles (MPs), scientifically defensible ecological risk assessments are increasingly required. However, such assessments remain challenging because of factors such as the diversity of MP characteristics (e.g., particle length and shape). In this study, we developed species sensitivity distributions (SSDs) for MPs using hierarchical Bayesian modeling, which accounted for intraspecies variation as well as the influence of MP characteristics on chronic no observed effect concentrations (NOECs). We also incorporated highest observed no-effect concentrations (HONECs) into SSD estimation by appropriately treating them as right-censored data. Using data from a recently updated ecotoxicity database, we analyzed two datasets excluding HONECs (21 species) and including HONECs (33 species). For the HONEC-excluded dataset, the best SSD model, selected based on the widely applicable information criterion, included size category (<83 vs. ≥83 μm) and fiber shape, suggesting that smaller particles and fibers were associated with lower chronic NOECs. For the HONEC-included dataset, the best model included particle length and shape (fragment and fiber), indicating that shorter particle lengths and nonspherical shapes were linked to lower chronic NOECs. Median estimates of the hazardous concentration for 5% of species ranged from 0.06 μg/L (fiber, particle length <83 μm) to 111 μg/L (nonfiber, particle length ≥83 μm) in the HONEC-excluded dataset. For the HONEC-included dataset, median estimates of the hazardous concentration for 5% of species ranged from 0.003 to 167 μg/L depending on particle length and shape, while their 95% Bayesian credible intervals spanned approximately 5 to 7 orders of magnitude. Despite uncertainties in modeling, our SSD modeling framework provided a generalizable and data-informed approach to the incorporation of diverse MP characteristics and censored effect concentrations to improve ecological risk assessments for MPs as well as other substances.