Groundwater Contamination Forecasting Using Automated Machine Learning

Abstract

Because of the subsurface's inherent geologic unpredictability, it is difficult to forecast the fate and transit of groundwater contaminants. To solve the equation for advection, dispersion, and reactivity, forecasting the flow of pollutants has been done using simplified geology and accepted assumptions. It may soon be possible to use extensive groundwater quality data from long-term polluted sites to feed machine learning algorithms that predict the spread of pollution plumes and enhance site management. The objective of this study was to first utilise extensive historical data from groundwater monitoring well samples to better understand the complex relationships between groundwater quality parameters, and then to construct a useful model for predicting the time until site closure.