Spatial Mapping of Saltwater Intrusion in Coastal Odisha, India using Hydrochemical Signatures and Machine Learning Tools

Abstract. Freshwater resources in coastal regions are increasingly at risk of salinization as a result of seawater intrusion, posing serious challenges to source sustainability. This study introduces a systematic approach for delineating vulnerable zones through hydrogeochemical assessment integrated with spatial and machine learning techniques. To demarcate the extent of contamination, groundwater samples were collected from 33 tubewells of Ersama block, Odisha during both pre and post-monsoon seasons of 2023, key ionic constituents such as Cl⁻, HCO₃⁻, and Na⁺ were analysed. Two critical hydrochemical ratios, Cl⁻/HCO₃⁻ and Na⁺/Cl⁻ were computed to quantify the level of salinization and to distinguish the sources of salinity respectively. Based on established threshold values, contamination levels were categorized into six classes, which were further refined by combining them with salinity origin to form 11 hybrid classes. These hybrid classes were used as labelled categories for classification using the Random Forest (RF) algorithm. Spatial distribution of classified zones was visualized using ArcGIS, enabling effective identification of contaminated areas. During the pre-monsoon, zones classified as high to extreme contamination due to seawater intrusion accounted for 19.1% of the area, while other sources contributed 31% in the same severity range. In the post-monsoon, seawater-related contamination showed a notable decrease to 6.4%, whereas contamination from other sources significantly elevated to 43%. This suggests a seasonal dilution effect on saline water intrusion immediately after monsoon due to the reversal of hydraulic gradient, though contamination from non- marine sources intensified. This approach offers a practical and scalable solution for monitoring saltwater intrusion dynamics in coastal settings.