Machine-learning models to map pH and redox conditions in groundwater in a layered aquifer system, Northern Atlantic Coastal Plain, eastern USA
Leslie A. DeSimone, Jason P. Pope, Katherine M. Ransom
Abstract
The study was conducted in the Northern Atlantic Coastal Plain aquifer system, in the eastern USA. Groundwater pH and redox conditions are fundamental chemical characteristics controlling the distribution of many contaminants of concern for drinking water or the ecological health of receiving waters. In this study, pH and redox conditions were modeled and mapped in a complex, layered aquifer system. Machine-learning methods (boosted regression trees) were applied to data from 3000 to 5000 wells. Predicted pH and the probability of anoxic conditions, defined by three thresholds of dissolved oxygen (0.5, 1, and 2 mg/L), were mapped at the 1-km2 scale for each of 10 regional aquifer layers. Maps depict the extent of acidic groundwater and oxic conditions in the shallow, unconfined surficial aquifer and in unconfined, recharge-proximal areas of underlying aquifers, in contrast to alkaline and anoxic groundwater elsewhere. Geographic patterns and influential predictors–including elevation, overlying confining-units thickness, and simulated groundwater age and flux–are consistent with prior understanding of the processes controlling pH and redox in the aquifer system. The model-based maps support robust estimates of aquifer proportions, either areal or volumetric, likely to contain groundwater of a specified quality or be vulnerable to specific pH- or redox-sensitive contaminants. The machine-learning methods were an effective tool to map groundwater quality at the regional scale.