Interpretable machine learning predictions reveal climate-driven escalation of arsenic in Tibetan alpine headwater rivers.

Climate warming is changing elemental dynamics in alpine headwater rivers. The Tibetan Plateau, known as the Asian Water Tower, is characterized by naturally high arsenic (As) levels. However, the spatiotemporal dynamics and future concentration-exceedance probability-flux changes of As under warming remain poorly understood. Here, we conducted high-frequency monitoring of As in river water and suspended particulate matter (SPM) at eight hydrological stations spanning the Sanjiangyuan region. Results show river water As decreases during floods periods, whereas SPM-As peaks during drought periods, with the highest levels in the Yangtze headwaters. Machine learning models, including Random Forest (RF), XGBoost, Support Vector Machine (SVM), and K-Nearest Neighbors (KNN), identified climate variables as dominant drivers of As concentrations and exceedance probabilities across river water, groundwater, and SPM. Under future climate scenarios, As concentrations in river water and SPM are projected to increase more rapidly after 2040, reaching 5.1 - 5.7 μg/L and 88 - 93 mg/kg, respectively, by 2050. The total As flux from river water and SPM, currently ∼1800 t, is predicted to increase to approximately 2314 - 2744 t by 2050. This study enhances our understanding of As dynamics in the Sanjiangyuan region, with implications for alpine rivers in Asia and globally.