Predicting Redox Conditions in Groundwater at a National Scale Using Random Forest Classification

Area of Science:

• Environmental Science
• Hydrogeology
• Geochemistry

Background:

• Groundwater redox conditions critically impact contaminant fate and human health.
• Spatial variability in redox rates complicates regional assessments.
• Understanding redox is vital for managing water resources and environmental risks.

Purpose of the Study:

• To predict groundwater redox conditions at a national scale for the contiguous U.S.
• To identify key natural and anthropogenic factors influencing redox states.
• To develop a model for predicting elevated manganese concentrations in groundwater.

Main Methods:

• Employed random forest classification using water quality data from over 30,000 wells.
• Related measured water quality to geological, hydrological, and soil properties.
• Developed a secondary model for predicting high manganese concentrations.

Main Results:

• Achieved 78-79% accuracy in predicting oxic/suboxic conditions.
• Geology, hydrology, soil properties, and hydrologic position were key predictors.
• Found that proximity to streams, depth, base flow index, and topographic wetness index influence redox conditions.

Conclusions:

• National-scale redox prediction is feasible and identifies key environmental drivers.
• The models can highlight areas vulnerable to redox-driven contaminant transport and manganese issues.
• Results aid in assessing groundwater and stream vulnerability to contaminants.