Identification of groundwater potential zones in data-scarce mountainous region using explainable machine learning

Abstract

Groundwater is a critical resource, yet its detailed assessment in mountainous regions is challenged by varying topography, complex hydrogeological characteristics and limited data. In this study, machine learning approaches were used to analyze the groundwater potential in five different watersheds in Nepal. Explainable machine learning models (EBM and GAMI-net) were used to identify zones with different groundwater potentials and controlling factors. The models were validated with k-fold cross-validation using the area under the receiver operating characteristics curve for the two groundwater potential models with unseen validation dataset of 0.87 and 0.88 respectively. We found that precipitation, elevation, soil bulk density, slope and lineaments primarily controls the groundwater potential in the study regions. The expected impact of each of the factors on groundwater potential was complex and multimodal. The results of this study can be used to improve water resource management and ensure sustainable groundwater use in the region.