Modelling Pesticide Degradation in Soil: Influence of Chemical and Soil Properties on DT50 Prediction

Abstract Assessing pesticide degradation in soil is crucial for environmental risk assessment and regulatory decision-making. Traditional quantitative structure-activity relationships models rely on chemical structure to predict pesticide behavior, often overlooking the influence of soil properties. In this study, soil properties and chemical structures were integrated to investigate the key variables affecting pesticide degradation. Using two datasets—an industry-internal dataset and the publicly-available Pesticide Properties DataBase—eXtreme Gradient Boosting regression modelling and SHapley Additive exPlanations were applied to identify the most important predictors of pesticide half-life of degradation (DT50). The important molecular descriptors derived from the analysis are structural variables (number of aromatic atoms, hybridization ratios), molecular connectivity, reactivity (topological polar surface area efficiency), electronic properties (HOMO-LUMO gap, the energy difference between the lowest unoccupied and highest occupied molecular orbitals), and hydrophobicity (octanol–water partition coefficient). Equally, the following soil properties: pH, soil organic carbon content, clay content, and microbial biomass, improved predictive performance, reducing mean squared error and increasing the coefficient of determination values. Despite differences in compound composition and experimental conditions across datasets, a substantial overlap in the key predictors found to be important, was observed. The findings underscore the importance of chemical structure and soil characteristics in degradation dynamics, with the former providing the greatest predictive power and the latter still adding some important improvement to the final model. By integrating multiple variables, our approach allows for more realistic and reliable predictions of pesticide persistence and can be extended to field-related conditions, mixtures, and multiple stressors in the future, and provide support for improved environmental risk assessment and regulatory frameworks.