Mapping PFAS Exceedance Risk in China's Surface Water: A Machine Learning Approach Informed by Source Distribution.

Per- and polyfluoroalkyl substances (PFAS) cause pervasive contamination of surface water, which presents a substantial public health challenge. China is a leading global producer and consumer of fluorinated chemicals. Therefore, the country faces an urgent need to clarify the nationwide distribution of these persistent pollutants. In this study, we addressed the challenge of sparse monitoring data by developing a Geographically Weighted Random Forest (GWR-RF) model, integrating a comprehensive, spatially explicit inventory of over 280,000 potential PFAS sources. The model demonstrated robust predictive performance (accuracy = 0.83, ROC-AUC = 0.91) and generated a 1 km resolution spatial map of PFAS exceedance risk across China's surface water. According to the results, a portion of the nation's surface water areas is at high risk, and hotspots are concentrated in the eastern coastal plain and key inland industrial provinces. An estimated 80-90 million people live in the high-risk areas. Further analysis revealed that proximity to known PFAS users and annual precipitation are the dominant predictors associated with increased risk. This research provides key scientific evidence to support the development of targeted contamination mitigation strategies, optimization of PFAS management, and protection of vulnerable populations.