A GIS based hybrid AHP and ensemble machine learning framework for identifying groundwater recharge and flood mitigation hotspots

Abstract Globally groundwater depletion threatens water security with aquifer storage declining at 300 km³/year. Climate change intensifies flood frequency and magnitude causing agricultural and other losses. Punjab, India also faces a dual crisis of declining groundwater levels and recurring flood events threatening agricultural sustainability, which contributes 30% of India’s wheat and rice production. The state experiences accelerated aquifer decline from extensive tube well use, while climate change induced extreme precipitation events cause devastating floods witnessed in August-September 2025. Traditional approaches treat groundwater recharge and flood control as isolated challenges. This paper presents an integrated framework combining Analytical Hierarchy Process (AHP) with ensemble machine learning to identify optimal groundwater recharge hotspots that serve dual purpose of groundwater recharging and flood control. Random Forest, Gradient Boosting and CART are used for analysis with opensource platform Google Earth Engine. Sentinel-1 SAR flood mapping, GRACE groundwater depletion, soil permeability, rainfall, land use and topography are used for analysis. JRC Global Surface Water occurrence dataset (1984–2023) is used as a proxy for flood proneness, replacing unreliable global flood hazard datasets. The ensemble model refines AHP suitability scores. Two strategic borewell categories are proposed, first, flood overflow borewells (50 sites) near rivers for flood diversion and recharge. Second, normal recharge borewells (50 sites) in inland waterlogged agricultural areas. Compared to existing methods like standalone AHP or ML model, our framework achieves performance with up to 90.4% accuracy. An interactive web application provides public accessibility for stakeholder engagement and policy formulation.