PREDICTING PH AND REDOX IN THE GLACIAL AQUIFER SYSTEM

The glacial aquifer system in the northern contiguous United States ranks first in the Nation as a source of groundwater for both public and domestic supply, with a combined pumpage of about 2.6 billion gallons per day for these purposes. Occurrences and concentrations of constituents of concern in drinking-water supplies from the glacial aquifer can vary considerably due to differences in soil/aquifer chemistry and mineralogy, geochemical processes, aquifer position, and residence time. pH and redox conditions are important because they affect the solubility and mobility of contaminants through processes such as ion exchange, sorption, complexation, and mineral weathering. Machine learning methods are being used to develop aquifer wide, 3-D maps of predicted pH and redox conditions using data available from the USGS National Water Information System database, and State ambient and compliance monitoring data.

Model predictor variables – including soil chemistry, mineral-saturation indices, land use, aquifer texture and hydrology, well-construction information, and groundwater age –are compiled into training and hold-out data sets, respectively, and applied to Boosted Regression Tree (BRT) prediction models. The relative contributions of the various explanatory variables are evaluated and partial dependence plots generated to infer the direction and degree of influence that individual explanatory factors have on pH or redox conditions at depths used for domestic and public supplies. Finally, grids of predicted continuous pH or redox ranges at depths used for domestic and public supplies are generated by applying the final BRT model to gridded predictor variables. Output from the grids of predicted pH and redox ranges can then be used to better understand the distribution and concentrations of constituents of concern such as arsenic or manganese.