You are here:
Predicting Violations of Nitrate in Drinking Water Across the Conterminous United States
Citation:
Pennino, M., S. Leibowitz, J. Compton, R. Hill, AND R. Sabo. Predicting Violations of Nitrate in Drinking Water Across the Conterminous United States. AGU Fall Meeting, Washington, D.C, December 10 - 14, 2018.
Impact/Purpose:
Under the Safe and Sustainable Water Resources Research Program, EPA scientists are modeling the spatial patterns of drinking water nitrate violations across the United States. Excess nitrate in drinking water has been a human health concern since the middle of the 20th century. Since 1979, public drinking water suppliers have been required to regularly monitor nitrate levels and EPA is notified when a public water system violates the 10 mg nitrate-N L-1 maximum contaminant level (MCL). This study found that for groundwater systems the most important variables in predicting nitrate drinking water violations are % cropland, irrigation extent, nitrogen surplus, nitrate treatment, and agricultural drainage, while for surface water systems the most important variables are irrigation extent, % shrubland, runoff, surplus precipitation, and mean precipitation. For groundwater systems, the random forest classification model correctly predicted 79% of catchments to have or have not have nitrate violations, while the random forest regression modeling approach explains about 43% of the variation in nitrate concentration. Regions predicted to have the highest probability of violations were central California, areas above the Oglala Aquifer in Texas, Oklahoma, Kansas, the Upper Midwest (Minnesota, Wisconsin, and Michigan), and Southeast Pennsylvania and Delaware. This analysis may inform decisions on how treatment, source water protection, and other management options could best protect drinking water from nitrate contamination. The results of this study could be important for a number of programs within EPA’s Office of Groundwater and Drinking Water, Regions, and States. Also, this work contributes to SSWR Task 4.03C "Monitoring and multi-media modeling approaches for verifying reductions" and SSWR Task 3.01B “Estimating and Predicting Water Resource Condition and Watershed Integrity.”
Description:
Excess nitrate in drinking water is a human health concern, especially for young children. When a public drinking water system exceeds the 10 mg nitrate-N L-1 maximum contaminant level (MCL), that system is reported as having a violation in the U.S. Environmental Protection Agency’s (EPA) Safe Drinking Water Information System (SDWIS). We used SDWIS data with random forest modeling to examine the drivers of nitrate violations across the conterminous U.S. and to predict where systems are most likely to exceed the nitrate MCL. As explanatory variables, we used EPA’s StreamCat data and additional metrics, including land cover, nitrogen input, soil/hydrogeology, and climate variables. For groundwater systems, the RF classification model correctly classified 79% of catchments and the RF regression model explained 43% of the nitrate concentrations above the MCL. The most important variables in the RF classification groundwater model were % cropland, agricultural drainage, irrigation-to-precipitation ratio, nitrogen surplus, and surplus precipitation. For surface water systems, the RF classification model correctly classified 90% of catchments and the RF regression model explained 52% of the variation in SW nitrate concentration. The variables most important for the RF classification surface water model were largely hydrologic variables including surplus precipitation, mean precipitation, irrigation-to-precipitation ratio, and % shrubland and canal density. Regions predicted to have greatest risk for nitrate violations from groundwater were central California, western Texas, Oklahoma, and Kansas, Nebraska, the Dakotas, parts of the Upper Midwest, and southeast Pennsylvania and Delaware. Areas at greatest risk for surface water nitrate violations were generally in the non-mountainous west and southwest. Understanding where violations are most prevalent, and the causes of violations, could help inform future management decisions on how source water protection and other management options could best protect drinking water from nitrate violations.