PREDICTING GROUNDWATER CONTAMINATION HOTSPOTS: PROBABILISTIC MAPPING OF MANGANESE AND IRON HAZARDS IN THE SHENANDOAH VALLEY, VIRGINIA

The Shenandoah Valley Aquifer, part of the Valley and Ridge province of the Appalachians, spans approximately 300 km and serves as the primary water source for over 36% of its residents who rely on private wells. Water quality assessments conducted by the Virginia Household Water Quality Program (VAHWQP) and the U.S. Geological Survey (USGS) National Water Information System (NWIS) indicate elevated concentrations of manganese (Mn) and iron (Fe) in the region's groundwater. Although Mn and Fe are vital nutrients, excessive Mn exposure has been linked to neurological issues, congenital heart defects, increased cancer mortality, and other health concerns, while combined Mn and Fe presence correlates with low birth weights and additional risks. This study employs a machine learning approach, utilizing Random Forest (RF) classification on VAHWQP and USGS NWIS datasets to predict the spatial distribution of high-risk zones where Mn concentrations exceed 0.08 mg/L (World Health Organization’s health-based guideline) and Fe concentrations surpass 0.1 mg/L (a threshold associated with geogenic contaminant mobilization). The results indicate a notable prevalence of Fe-elevated zones in the northern Valley, while Mn contamination risk is highest along the NE-SW axis in the northern and central Valley. Scattered high-risk zones are also observed throughout the region. The models achieve Area Under the Receiver Operating Characteristic Curve (AUC-ROC) scores of 90.21% for Mn and 93.81% for Fe on validation data, demonstrating robust predictive performance. These findings highlight the critical need for systematic water-quality testing and the implementation of mitigation strategies to address contamination risks in potentially hazardous aquifer zones within the Valley, ensuring the provision of safe drinking water for local communities.