Arsénico en aguas subterráneas de Taiwán: de la crisis de mediados del siglo XX a modelos predictivos para...

Taiwan's mid-20th-century Blackfoot disease epidemic highlighted the public health implications of geogenic arsenic in groundwater and prompted extensive water-supply interventions. Despite these measures, arsenic persists in aquifers, continuing to affect drinking water, irrigation, and aquaculture. This study developed a scalable machine learning model to identify areas across Taiwan at risk of groundwater arsenic concentrations exceeding the WHO guideline of 10 µg/L. The model demonstrated robust performance (mean AUC 0.93, balanced accuracy 0.87). A detailed evaluation and interpretation of the principal predictor variables governing arsenic dissolution provides insights into geochemical and hydrogeological controls on mobilization. Furthermore, the resulting hazard map of Taiwan was used to estimate the human population at risk as well as the locations and types of exposed agriculture and aquaculture. Hotspots are clustered in the Chianan, Pingtung and Lanyang plains, where many rural communities still rely on untreated groundwater. Population exposure is substantial, with an estimated 148,000 people at risk in 2023, which is about half the 303,000 reported in 1998. Agricultural impacts are also considerable: >80 % of aquaculture areas, and 33 % of paddy rice fields occur within high-risk zones. Several high-risk zones lie outside Taiwan's Groundwater Control Areas, revealing regulatory blind spots. By connecting Taiwan's historical arsenic crisis to novel predictive tools, this work supports evidence-based strategies to reduce current exposures and prevent future public arsenic-related health impacts.