Delineating hydro-geologically constrained groundwater zones in the Himalayan River basins of India through an innovative ensemble of hypsometric analysis and machine learning algorithms

Hydro-geologically constrained groundwater (HGCG) zones have restricted water-holding capacity and low natural recharge rates, leading to reduced groundwater storage potential. These restrictions stem from unfavorable landform shapes and geological conditions that hinder the natural replenishment, flow, and storage of groundwater. Hypsometric analysis by assessing the area-elevation relationship, indicates geological stage and landform shape, thereby offering insights into groundwater storage. Thus, in this study, an innovative approach by combining machine learning (ML) algorithms with hypsometric analysis, in-situ precipitation data, and Depth to Water Level data for delineating HGCG zones is proposed. The methodology was tested over the Indian Himalayan River basins of the Indus, Ganga, Brahmaputra, and Barack. Following hypsometric analysis and derivation of hypsometric parameters (HP), sub-watersheds (SW) were clustered into mountainous, shallow (< 10 m below ground level, mbgl), and deep (> 10 mbgl) Depth to Water Level (DWL) zones using unsupervised and supervised ML algorithms of k-means clustering and linear discriminant analysis (LDA), respectively. Out of the 140 SWs analyzed, 41 were identified as mountainous, 70 were characterized by shallow DWL, and 29 were found to exhibit deep DWL, with a classification accuracy of 95%. The classified DWL layer, in conjunction with observed seasonal in-situ precipitation and depth to water level data, was further utilized to identify 8 SWs covering a total area of 148,015.80 sq. km (11.17%) as hydro-geologically constrained groundwater zones. This innovative methodology highlights the significance of ML algorithms for identifying such zones based on hypsometric analysis and underscores the importance of delineating these zones for sustainable groundwater extraction and utilization.