A Probabilistic Approach for Predicting Methane Occurrence in Groundwater.

Area of Science:

• Environmental Science
• Hydrogeology
• Geochemistry

Background:

• Groundwater geochemistry data are valuable for environmental baseline assessment (EBA) in areas with unconventional hydrocarbon resource development.
• Existing datasets often lack critical methane concentration data necessary for EBA in shale gas regions.

Purpose of the Study:

• To develop and validate a logistic regression (LR) model for predicting methane occurrence in Alberta's aquifers.
• To apply the model to a large-scale groundwater monitoring program to enhance EBA resolution.

Main Methods:

• Calibrated and tested an LR model using existing groundwater geochemistry data with methane concentrations.
• Applied the validated LR model to a province-wide public health groundwater monitoring program dataset (n = 52,849 samples).
• Utilized basic hydrochemical parameters to predict methane occurrence where direct measurements were unavailable.

Main Results:

• The LR model achieved high prediction accuracy for methane occurrence: 89.8% (n = 234) and 88.1% (n = 532) in calibration/testing datasets.
• Methane presence was significantly correlated with electron donors (sulfate), well depth, and total dissolved solids.
• Successfully predicted methane occurrence across a large dataset lacking direct gas measurements.

Conclusions:

• The developed LR model effectively predicts methane occurrence in groundwater using readily available hydrochemical parameters.
• This approach significantly improves the spatial resolution and effectiveness of EBA in shale gas development areas.
• Leveraging existing high-density monitoring data enhances environmental protection strategies.