Hybrid AHP-machine learning modelling of groundwater recharge potential under land use and climate change in tropical Basins: Implications for sustainable water management

Groundwater recharge in sub-Saharan Africa is increasingly threatened by climate change and land use/land cover (LULC) changes, yet integrated assessments remain limited for tropical basins. This study evaluates groundwater recharge potential in Ghana's Pra and Ankobra River Basins usings a novel hybrid approach combining Analytical Hierarchy Process (AHP) and machine learning to assess coupled climate-LULC impacts. The framework integrates statistically downscaled CMIP6 projections (SSP1-2.6, SSP2-4.5, SSP5-8.5), Random Forest-based LULC modelling, and AHP-weighted multi-criteria analysis. The Analog method achieved accurate rainfall downscaling (RMSE = 5.56 mm/day, R² = 0.79), while Land Change Modeller predicted LULC transitions (precision = 0.81, Kappa = 0.55). Results indicate climate change dominates recharge variability, with SSP1-2.6 expanding very high recharge zones (+91.90 % mid-future) and SSP5-8.5 reducing very low zones (−67.94 % far-future). Nonlinear responses emerged, including an initial high-recharge decline (−8.72 % near-future) followed by recovery (+34.31 % mid-future). Spatially, the Ankobra Basin and mid-western Pra Basin exhibited the highest recharge potential (14.84 % very high), while southeastern areas remained vulnerable (5.79 % very low). AHP identified rainfall (weight = 0.22), geology (0.20), and lineament density (0.14) as key controls. The hybrid AHP-machine learning approach outperformed conventional methods, providing robust quantification of climate-LULC interactions. Findings emphasise the need for adaptive management, prioritising high-recharge conservation (e.g., Tarkwa) and alternative solutions in vulnerable zones (e.g., Shama). This study offers a transferable framework for tropical basins, supporting sustainable groundwater planning under global change.