Geospatial modelling of groundwater potential zones in Baramati tehsil, India using AHP and F-AHP approaches

Abstract

Geospatial techniques provide an effective framework for groundwater potential zonation by integrating multiple hydrogeological and terrain-related factors in a spatial environment. In the present study, geospatial technology, Analytical Hierarchy Process (AHP), and Fuzzy Analytical Hierarchy Process (F-AHP) techniques were applied to delineate groundwater potential zones in Baramati tehsil for sustainable groundwater resource management in an agriculturally important region. Eight thematic layers, namely geomorphology, slope, lineament density, vadose zone, soil, drainage density, rainfall, and land use/land cover, were prepared and integrated for groundwater potential mapping. The relative weights of these thematic layers and their subclasses were assigned using AHP and F-AHP based on their inferred influence on groundwater potential. The resulting groundwater potential maps were classified into five categories: very high, high, moderate, low, and very low. The analysis revealed that the study area is predominantly characterized by moderate groundwater potential, covering 67.72% of the area in the AHP model and 78.40% in the F-AHP model. High groundwater potential zones account for 19.35% and 14.42% of the area, while low groundwater potential zones cover 12.76% and 7.12% under the AHP and F-AHP approaches, respectively. Very high and very low potential zones occupy comparatively smaller areas. The predictive performance of the generated maps was evaluated using groundwater-level observations from 38 wells and Receiver Operating Characteristic/Area Under Curve (ROC/AUC) analysis. The AHP model yielded an AUC value of 0.772, while the F-AHP model showed a comparatively higher AUC value of 0.835. Because the ROC analysis was based on a presence/background comparison and the same 38 well observations were also used to derive the vadose-zone input layer, these AUC values are interpreted as internal support for the models rather than fully independent predictive validation, with F-AHP performing relatively better in the present study area.