Assessment of spatiotemporal risks for nationwide groundwater nitrate contamination

National assessments of groundwater contamination risks are crucial for sustaining high-quality groundwater supplies. However, traditional methods often treat groundwater contamination risk as a steady-state indicator without considering spatiotemporal variation in risk, both geographically and over time, caused by anthropogenic and climatic factors. In this work, XGBoost, a tree-based algorithm, was applied to comprehensively analyze the drivers of groundwater contamination from nitrate, using data on 13 physical features (as used by the index-based ranking method DRASTIC) and 30 anthropogenic features from 1985 to 2010 in the contiguous United States (CONUS). The results indicate that physical features controlling the transport processes, particularly those affecting contaminant travel time from land surface to groundwater (depth to water table and transmissivity), were the dominant factors for nitrate contamination in groundwater. This was followed by features representing the potential nitrogen loading. Positive correlations between most features and the nitrogen loading time (year) were found, suggesting their growing influence on contamination risk. Based on the drivers identified for nitrate concentrations exceeding 10 mg/L in groundwater and their varying temporal contributions, this study proposes a reformulated index-based method for contamination risk assessment. With this method, an overall accuracy of around 70 % was achieved based on the validation data set. The predicted high-risk areas are mainly intensive irrigation regions, such as the High Plains, northern Midwest, and Central Valley. This new approach contributes to a more accurate and effective assessment of the contamination risks of groundwater on a regional and national scale under temporally varying environmental conditions.