Potentialities and development of groundwater resources applying machine learning models in the extended section of Manbhum-Singhbhum Plateau, India

Abstract

Groundwater is essential for living earth including ecosystem functioning and development of society worldwide. In recent times, demand and pressure on groundwater resources are progressively increasing over time. Thus, the assessment and management of groundwater resources particularly in semi-arid region are very much crucial. Therefore, the principal objective of the present study is to categorize the groundwater potential areas using advanced machine learning (ML) approaches. In this study, random forest (RF), support vector machine (SVM), and extreme gradient boosting (XGBoost) algorithms have been applied. The accuracy of each model has been estimated using the receiver operating characteristics (ROC) curve. About 60.63%, 65.39%, and 53.75% of areas come under moderate to very low groundwater potential. XGBoost indicates the highest predictive capacity (AUC 0.97). The innovation of this study lies in the combination of hydrological, topographical and geological datasets into machine learning platform. This research will support water resource management worldwide.