Answers are grounded in this paper and cite their sources.
Current regulation identifies environmentally persistent chemicals often using thresholds defined for single-media half-lives. This approach ignores the large and even dominating impact of multimedia partitioning on the (bio)availability of compounds that undergo degradation, and accordingly may underestimate the chemical lifetimes by orders of magnitude. This is demonstrated through comparing chemical half-lives in water with overall persistence (Pov) in terms of environmental lifetimes calculated through a steady-state multimedia mass-balance model. While Pov depends on the entrance compartment, Stroebe's τE, the temporal-remote-state expected environmental lifetime, is less sensitive to the initial entry compartment. Thus, τE appears most suited for the prediction of chemical persistence. In this study, a stepwise Ordinary Least Squares (OLS) regression analysis showed that only four chemical properties, log values of Kow (hydrophobicity), kdeg,w and kdeg,a (degradation rate constants in water and air, respectively), and Pv (vapour pressure), are sufficient to estimate τE. But this analysis also showed non-linear error patterns despite overall good statistics, unraveling peculiarities of the non-linear log-log dependence of τE on compound properties. A machine-learning k nearest neighbors (kNN) regression model operating in the property space yields even more accurate and linear results. Application to existing data sets demonstrates the usefulness of this approach. In particular, it shows that the reduction of bioavailability due to multimedia partitioning may cause compounds classified as "readily biodegradable" to be identified as persistent when they have sufficiently high hydrophobicity (e.g. phenanthrene with log Kow = 4.5: lifetime > 30 years). The developed tools (SimpleRisk4REACH enabling τE calculation through Stroebe's mathematical procedure; kNN regression model for τE prediction) are made available to the public.
No takes yet. Share an insight, caveat, or question.
Meent et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: