Modeling PFAS sorption characteristic patterns across the conterminous United States and utilizing these data for...

Per- and polyfluoroalkyl substances (PFAS) are persistent environmental contaminants that pose significant risks to groundwater due to their widespread presence and resistance to degradation. The soil-water partitioning coefficient (K_d) is one of the key parameters used for predicting PFAS transport processes, yet the availability of K_d data is currently limited. This study developed a data-driven framework to map PFAS sorption characteristics (K_d values) at a large scale and employed the dataset to assess groundwater vulnerability across the conterminous United States (CONUS). We trained a stacked ensemble machine learning model, which combined Random Forest, XGBoost, GBM, and LightGBM with a ridge regression meta learner, on a literature derived K_d dataset. Soil properties, such as organic carbon content, CEC, silt, sand and clay percentages, molecular weight, chain length, and pH served as predictors and measured K_d values as target. Organic carbon content emerged as the dominant control for predicting PFAS sorption. Overall, the stacked ensemble model achieved strong predictive performance with R² and RMSE values of 0.87 and 0.32, respectively. We applied this model to the Soil Survey Geographic Database (SSURGO) dataset to generate spatially explicit K_d maps for multiple PFAS species across CONUS. The predicted K_d values were subsequently integrated with groundwater recharge, depth to water table, and proximity to known PFAS sources data within a Random Forest classification model to map groundwater vulnerability to PFAS contamination. The vulnerability map identified approximately 33,400 km² as high-risk (probability >0.67) areas. These areas are primarily concentrated in the Northeast, upper Midwest, and portions of the Southeast regions. This study provides useful CONUS-scale K_d data and utilizes these data to develop a modeling framework that can help prioritize monitoring and remediation strategies to minimize PFAS risks.