Deconvoluting the influence of environmental forcings on chemical weathering rates using machine learning model

Quantifying the contributions of silicate and carbonate weathering to global carbon dioxide sequestration is crucial for understanding Earth’s climate regulation mechanisms. In this study, we introduce a framework that integrates explainable machine learning with partial dependence analysis to understand the relationships between environmental forcings and chemical weathering rates. We trained machine learning models on a geochemical dataset from the Tibetan Plateau together with climatic and environmental factors, achieving a weathering model with high predictive ability (R 2 = 0.84–0.96). Furthermore, we separated direct environmental driving effects from indirect mediating effects through systemic partial dependence plots. The results indicate that runoff is the dominant controlling factor of both riverine dissolved SiO 2 and HCO 3 ⁻ flux. Warming promotes riverine HCO 3 ⁻ flux in the river, but its effect on riverine dissolved SiO 2 flux is bidirectional-first inhibiting, then promoting. Further, our analysis reveals that the effect of precipitation on riverine dissolved SiO 2 flux is mainly mediated indirectly by runoff. Our study establishes a framework for investigating chemical weathering, highlighting the potential of machine learning to advance predictive modeling and reveal complex interactions between environmental forcings and chemical weathering.