PubMed · 2026-03-31
Scientists have developed a roadmap for using AI and computer modeling to predict how long chemical pollutants — like pesticides and microplastics — persist in the environment before breaking down, moving the field from outdated rules-of-thumb to modern machine learning.
Over 60 years of development separates early simple chemical models from today's machine learning approaches, which can now handle vastly more diverse and complex chemical datasets.
Modern models can predict not just how long a chemical persists, but also its breakdown products and pathways — critical for understanding whether degradation makes a contaminant safer or more toxic.
The review identifies a concrete roadmap including standardized reporting, shared benchmark datasets, and hybrid models that combine physical chemistry rules with AI to make predictions decision-ready for regulators and land managers.