Detection and Mapping of Seawater Intrusion Using Machine Learning Integration of Hydro-Chemical, Soil Salinity, and Remote Sensing Indices

The process of groundwater salinization and seawater intrusion affects coastal aquifers of arid and semiarid areas owing to excessive abstraction of groundwater, land use change, and climatic variability. In this study, hydrochemistry, Remote sensing, Geographic Information System, and machine learning techniques are combined to analyze the level of groundwater salinity and seawater intrusion in Ramanathapuram District, Tamil Nadu, India. Thirty samples were collected and analyzed for electrical conductivity (EC), total dissolved solid (TDS), Ca 2+ , Mg 2+ , Na + , K + , HCO 3 − , SO 4 2 − , Cl − , and NO 3 − parameters. A wide spatial variability was observed where EC varied between 2,450 and 12,450 µS/cm, and TDS between 1,650 and 7,980 mg/L. Also, high concentrations of Na + (average value: 43.85 meq/L) and Cl − (average value: 65.47 meq/L) indicate the predominance of seawater in the coastal areas. Soil salinity indices obtained using the Landsat-8/9 imagery were integrated with the parameters of hydrochemical analysis and then classified using the random forest (RF) technique. The RF classification model reached an overall accuracy of 91.8% with a Kappa coefficient of 0.89. SSI-2 (29%) and NDSI (24%) were selected as the major predictors, followed by salinity index (19%), SSI-1 (17%), and shortwave infrared (SWIR)-1 reflectance (11%). From spatial analysis, it was found that Sayalgudi and Keelakarai are severe impacted zones (Seawater Intrusion Indicator > 10), Ramanathapuram, Devipattinam, and Thondi are high-impact zones, RS Mangalam and Perunali are moderate impact zones, while Kamudi, Paramakudi, and Bogalur are low to very low salinity zones. The combined methodology ensures high predictive accuracy (91.8%), showing that the salinity indices calculated based on the SWIR are the most sensitive predictor of seawater intrusion in coastal aquifers with sparse data. The results provide useful guidance for management and mitigation of coastal aquifers in Ramanathapuram District and other semiarid coastal areas across the world.