Integrated machine learning and AHP-based framework for spatial delineation of rainwater harvesting potential

Groundwater scarcity and its declining situation in the semi-arid area of Jaipur district, Rajasthan make it imperative to adopt scientific solutions to locate suitable rainwater harvesting sites. This study combines Geographic Information Systems (GIS), Analytical Hierarchy Process (AHP) and machine learning (ML) algorithms for delineating the RWH suitability zones. Eleven biophysical and anthropogenic factors, namely elevation (280–647 m), slope (0–35°), topographic wetness index (TWI), drainage density, soil profile, land use/land cover (LULC), rainfall (350–700 mm yr⁻ 1 ), hillshade, aspect, lineament density, and proximity to roads, were standardized and incorporated into a spatial database. A georeferenced inventory of 150 locations was divided into training (70%; n = 105) and testing (30%; n = 45) datasets and analyzed using XGBoost, Support Vector Machine (SVM), Gradient Boosting Machine (GBM), CatBoost, and AHP models. Among the evaluated algorithms, XGBoost exhibited the highest predictive performance with a ROC-AUC of 0.78, followed by SVM (0.76), AHP (0.73), CatBoost (0.69), and GBM (0.68). The XGBoost model identified 38.1% (3902.6 km 2 ) of the study area as highly suitable to very highly suitable for RWH implementation, whereas 34.6% (3533.7 km 2 ) and 27.3% (2783.7 km 2 ) were classified as moderate and low-to-very low suitability classes, respectively. Variable importance analysis revealed that rainfall (15.7%), drainage density (15.6%), elevation (11.5%), and TWI (8.8%) were the dominant controls on RWH suitability. The AHP framework showed acceptable consistency (CR < 0.10) and model validation showed an overall accuracy of 88% and a Kappa coefficient of 0.64. The integrated framework successfully identified priority recharge areas in the southeastern lowlands and offers a scalable decision support tool for sustainable groundwater management and water resource planning that is resilient to climate change in semi-arid areas.