SREL Reprint #3707

Evaluating the predictive power of multiple regression models for groundwater contamination using PyCaret-23489

Trang Huynh1, Hirak Mazumdar1, Hardik Gohel1, Hilary Emerson2, and Daniel Kaplan3

1University Of Houston-Victoria
2Pacific Northwest National Laboratory
3University of Georgia

Abstract: The U. S. Department of Energy manages several sites with complex subsurface contamination due to release of legacy waste from the historic development of nuclear weapons. In compliance with federal regulations, extensive and costly sampling and monitoring plans are being implemented at these sites which require long-term operations. As researchers are working to optimize these processes, there are significant opportunities for improvement and future cost savings that could be realized through artificial intelligence. We present a novel Machine Learning (ML) algorithm calibrated with aggregated groundwater data from Hanford Site’s 100-area using PyCaret, an open-source, low-code ML library in Python. We demonstrate the library’s capability for automatic calibration of ML workflows for different regression modeling tasks. For increased spatiotemporal resolution and accuracy of groundwater monitoring and prediction, we compare different regression models where the Area Under the Receiver Operating Characteristic Curve threshold value of 80% was achieved for Cr-VI after training on 16,457 observations and testing on 4,116 observations. The model accuracy achieved a range of 0.92–0.97 for different ML regression models. Thus, this research is bringing us closer to utilizing ML models for monitoring groundwater contamination.

Keywords: Groundwater, Contamination, Prediction, ARIMA, Machine Learning, Optimization

SREL Reprint #3707

Huynh, T., H. Mazumdar, H. Gohel, H. Emerson, and D. Kaplan. 2023. Evaluating the predictive power of multiple regression models for groundwater contamination using PyCaret-23489. WM2023 Conference. Phoenix, Arizona.

This information was provided by the University of Georgia's Savannah River Ecology Laboratory (srel.uga.edu).