Using a Regression-Enhanced Random Forest Model to Improve Groundwater Monitoring and Management

Abstract

Groundwater is a vital natural resource essential for sustaining human life and ecosystems as a freshwater source. However, intensive groundwater pumping (GWP) can dangerously deplete groundwater levels, and exacerbate issues such as sea-level rise, groundwater inundation, and saltwater intrusion in coastal areas [20], further affecting the availability and accessibility of groundwater. To address these challenges, accurate monitoring and recording of water table depth (WTD), a key indicator of groundwater levels, are imperative for sustainable groundwater management. Therefore, this study investigates the implementation of the regression-enhanced random forest (RERF) model to predict WTD anomalies with pumping as a major input. Focusing on New Jersey, the study provides insights into the relationship between WTD and its critical input variables. Moreover, significant gaps in WTD observations persist in New Jersey, which further highlights the need for comprehensive monitoring efforts. Thus, by employing advanced ML techniques and leveraging available data, this research contributes to enhancing groundwater management practices and informing future decision-making processes.