Groundwater quality and risk in the Ganga River Basin: an integrated machine learning appraisal
摘要
Groundwater supports the livelyhoods of hundreds of millions across the Ganga River Basin (GRB), yet its quality is increasingly stressed by geogenic and anthropogenic factors. Using a high-density 2022-dataset from 3417 wells, this study integrates Self-Organizing Maps (SOM), saturation indices, Nonnegative Matrix Factorization (NMF), the Entropy Water Quality Index (EWQI), and Monte Carlo-based health risk assessments. SOM analysis identified four distinct clusters (Davies-Bouldin = 1.45), resolving three dominant regimes: salinity-driven (Na+–Cl−) zones in the west and deltaic regions, widespread freshwater (Ca2+–Mg2+–HCO3−) recharge zones, and localized nitrate (NO3−) and sulfate (SO42−) hotspots. NMF decomposition further attributed these to four processes: salinization/mixing, carbonate/sulfate dissolution with nitrate inputs, geogenic cation exchange, and anthropogenic overprints on saline backgrounds. The study shows that only 30% of samples fall into excellent or moderate quality categories, while poor quality categories cluster in the western basin. Health risk modeling indicates significant vulnerability for children, with Hazard Quotient (HQ) exceedance in 28% of sampling locations for nitrate and in 25% for fluoride. These results provide a 2022-baseline to target salinity controls and protect vulnerable communities through coupled geospatial and machine learning diagnostics.