Groundwater Potential Mapping Using AHP And Machine Learning in the Upper Subarnarekha River Basin, India

Groundwater is a key water source, and identifying high-potential regions is essential for effective management. Analytical Hierarchy Process (AHP) and Random Forest (RF) model have been used to map Groundwater Potential Zones (GWPZs) in upper Subarnarekha River Basin (SRB), India. Ten factors influencing groundwater occurrence were selected for inter-thematic correlation analysis. The model’s performance has been assessed using Receiver Operating Characteristics (ROC). AHP shows approximately 15% of the area has good Ground Water Potential (GWP), primarily located in the south and southeast region. Additionally, around 57% of the area demonstrated moderate GWP in the central part of the basin, whereas 24% of the area shows poor GWP in the northwestern regions. ROC revealed that AHP and RF models achieved success rates of 71%, and 73%, respectively. Due to the limited availability of well data (126 data points across a $14,037 \mathrm{~km}^{2}$ study area), the Machine Learning (ML) model struggled to accurately delineate groundwater potential zones. In contrast, the AHP-based weighted overlay analysis proved more effective. The findings can support watershed management efforts and guide the selection of optimal well locations in future projects.