Machine learning-based groundwater flow prediction: a case study of Bina region, Madhya Pradesh, India

The present study examines the utilization of machine learning (ML) methodologies for predicting groundwater levels (GWLs) in the Bina River basin of Madhya Pradesh, India. The study incorporates several ML techniques, such as regression trees (RT), support vector machines (SVM), and ensemble tree models, to predict groundwater dynamics using an extensive array of hydrogeological, meteorological, and anthropogenic factors. The dataset was partitioned into training and testing subsets, and models were assessed employing conventional statistical measures including R2, root mean square error, mean absolute error, and Akaike information criteria. Of the evaluated models, the fine RT had the best accuracy (R2 = 0.976), illustrating its proficiency in capturing intricate non-linear interactions. The SVM models, namely the linear and medium Gaussian variations, demonstrated competitive performance. The ensemble boosted trees model exhibited poor performance, perhaps attributable to overfitting. The research highlights the significance of data-driven models in groundwater management, particularly in areas with restricted access to physical monitoring facilities. The results indicate that ML-based forecasting offers a dependable alternative to conventional approaches and may substantially improve decision-making for sustainable groundwater resource management. This study presents a unique regional application and establishes a benchmark for scalable, AI-driven hydrological modeling in comparable catchments.