ABSTRACT Groundwater, the largest global source of freshwater, is under increasing stress due to over‐extraction, leading to a significant decline in groundwater levels (GWLs) in many regions around the world. This global groundwater crisis, driven by consistent overdraft, seriously threatens water security and requires immediate action for sustainable management strategies. This study aims to predict and forecast monthly GWLs at three critical observation wells, such as Ramachandrapuram, Palakollu, and Jangareddigudem, located in the Lower Godavari River Basin, India, to support sustainable groundwater management. Univariate artificial intelligence (AI) models, namely, random forest (RF), least‐squares support vector machine (LS‐SVM), and radial basis function SVM (RBF SVM), were utilized for GWL simulation and prediction. The time‐series features were extracted from historical groundwater data (January 1998–December 2012) to develop prediction models for training (January 1998–June 2008) and testing (July 2008–December 2012) periods. The models were then applied to project the monthly GWLs from January 2013 to December 2018. RF outperformed LS‐SVM and RBF SVM models, achieving R 2 values of 0.89, 0.86, and 0.82 for Jangareddigudem, Ramachandrapuram, and Palakollu during testing phase. The superior performance of the RF model demonstrates its robustness in modeling GWLs with high predictive accuracy. This data‐driven approach, leveraging AI techniques for time‐series prediction, presents a novel methodology for GWL estimation in data‐sparse regions. The developed models provide valuable insights for sustainable groundwater management and inform policy decisions to mitigate impacts of groundwater overdrafts and ensure long‐term water security in vulnerable regions.
Patel et al. (Fri,) studied this question.