Phthalate ester pollution poses a potential threat to soil organisms: Insights from a global meta and machine-learning analysis.

The widespread accumulation of phthalate esters (PAEs) in agricultural soils has raised global concern, yet their comprehensive ecological impacts on soil organisms remain poorly quantified. To address this critical knowledge gap, we conducted a systematic meta-analysis of 2063 observations from 43 publications to evaluate the effects of PAE on soil organisms. Machine-learning models were further developed to predict the PAE-induced impacts on biological indicators. Our meta-analysis revealed that PAEs induce significant adverse effects across multiple biological endpoints, including a 22.01 % reduction in soil animal survival, a 42.24 % decrease in reproduction rate, an 8.53 % diminishment in plant biomass, a 29.71 % reduction in root length, a 5.48 % inhibition of seedling growth, and a 2.04 % and 4.32 % lowering of the microbial Shannon index and dehydrogenase activity, respectively. To predict these quantified impacts, the eXtreme Gradient Boosting (XGBoost) model was developed and achieved acceptable prediction accuracy. Subsequent Shapley Additive Explanation (SHAP) analysis further identified soil pH (5.25 %), PAE concentration (4.80 %), exposure time (3.16 %), and PAE type (3.08 %) as the primary factors driving these ecotoxicological responses. Overall, this study provides a quantitative basis for risk assessment and novel predictive tools for evaluating the ecological risks of PAE pollution to soil health and agricultural sustainability.