Per- and polyfluoroalkyl substances (PFAS) are emerging contaminants that pose significant challenges in mechanistic fate and transport modeling due to their diverse and complex chemical characteristics. Machine learning provides a novel approach for predicting the spatial distribution of PFAS in the environment. We used spatial location information to link PFAS measurements from 1207 private drinking water wells around a fluorochemical manufacturing facility to a mechanistic model of PFAS air deposition and to publicly available data on soil, land use, topography, weather, and proximity to multiple PFAS sources. We used the resulting linked data set to train a Bayesian network model to predict the risk that GenX, a member of the PFAS class, would exceed a state provisional health goal (140 ng/L) in private well water. The model had high accuracy (ROC curve index for five-fold cross-validation of 0.85, 90% CI 0.84-0.87). Among factors significantly associated with GenX risk in private wells, the most important was the historic rate of atmospheric deposition of GenX from the fluorochemical manufacturing facility. The model output was used to generate spatial risk predictions for the study area to aid in risk assessment, environmental investigations, and targeted public health interventions.

Private well users are potentially exposed to a range of chemical contaminants through their drinking water. Point-of-use (POU) water treatment represents one potential solution to reduce harmful exposures through well water, but well users frequently do not adopt household treatment even if they learn their water is contaminated. This study elucidates the experiences, perceptions, and beliefs of 17 households on private wells in North Carolina that participated in a pilot-scale POU water treatment intervention to better understand the drivers and barriers of POU treatment adoption among well users. The intervention consisted of an under-sink activated carbon block POU filter designed to remove lead and two long-chain perfluoroalkyl acids. Filter effluents and influents were tested monthly for eight months. Questionnaires administered before and after the intervention showed a significant decrease in participants' perceived vulnerability to well water contamination, with 77% feeling vulnerable to poor well water quality before, compared to 23% after the filter was installed. However, the POU filters did not fully eliminate feelings of water insecurity (for example, concerns about exposure to contaminants when bathing remained). Lack of knowledge and skills associated with installing and maintaining POU treatment were important barriers to adoption for some well users. Perceptions of POU treatment were also significantly correlated with the intent to implement other well stewardship behaviors such as well water testing. The results highlight the need for strengthened outreach and support programs that provide technical assistance, education, and financial support for households relying on private wells.

Lead in drinking water continues to put children at risk of irreversible neurological impairment. Understanding drinking water system characteristics that influence blood lead levels is needed to prevent ongoing exposures. This study sought to assess the relationship between children's blood lead levels and drinking water system characteristics using machine-learned Bayesian networks. Blood lead records from 2003 to 2017 for 40,742 children in Wake County, North Carolina were matched with the characteristics of 178 community water systems and sociodemographic characteristics of each child's neighborhood. Bayesian networks were machine-learned to evaluate the drinking water variables associated with blood lead levels ≥2 μg/dL and ≥5 μg/dL. The model was used to predict geographic areas and water utilities with increased lead exposure risk. Drinking water characteristics were not significantly associated with children's blood lead levels ≥5 μg/dL but were important predictors of blood lead levels ≥2 μg/dL. Whether 10% of water samples exceeded 2 ppb of lead in the most recent year prior to the blood test was the most important water system predictor and increased the risk of blood lead levels ≥2 μg/dL by 42%. The model achieved an area under the receiver operating characteristic curve of 0.792 (±0.8%) during ten-fold cross validation, indicating good predictive performance. Water system characteristics may thus be used to predict areas that are at risk of higher blood lead levels. Current drinking water regulatory thresholds for lead may be insufficient to detect the levels in drinking water associated with children's blood lead levels.