Integrated Micro-Watershed Prioritization Using a Novel Multi-Criteria Decision-Making Approach

Watershed management is necessary to supply water resources and manage soil erosion, especially in semi-arid areas where land degradation and groundwater loss are serious issues. In the current research, morphometric analysis is combined with multi-criteria decision-making (MCDM) methods to prioritize micro-watersheds in the Bandu sub-watershed of Purulia district, West Bengal, India. The region is characterized by rolling terrain, hard rock, and semi-arid climate with extreme water scarcity and soil erosion. 13 morphometric parameters like Stream Length Ratio, Bifurcation Ratio, Drainage Density, Stream Frequency, and Hypsometric Integral were derived from Shuttle Radar Topography Mission (SRTM) Digital Elevation Models (DEMs). Micro-watersheds were divided using Principal Component Analysis (PCA) and K-means cluster classification according to the geomorphological differences. TOPSIS, COPRAS, and VIKOR, three various MCDM models, were utilized in ranking micro-watersheds using Fuzzy Shannon Entropy in weighing of the parameters. Each MCDM model was applied to assign priority ranks to micro-watersheds based on integrated performance across groundwater potentiality and soil erosion risk. The resulting rankings were then compared with an unsupervised classification generated using K-means clustering, which objectively groups watersheds with similar hydro-environmental characteristics. The comparison revealed a strong alignment between the TOPSIS model and the K-means derived clusters, particularly in micro-watersheds such as 2A2B5k and 2A2B5m, both of which received similar ranks in both approaches. To further evaluate model performance classification metrics were computed. The TOPSIS model emerged as the most consistent and accurate, achieving an accuracy of 0.73, outperforming both VIKOR and COPRAS. This consistency between supervised (MCDM) and unsupervised (K-means) methods strengths confidence in prioritization outputs.