Groundwater Potential Mapping and Borehole Prioritization in Togo: An Integrated Machine Learning and Multi-Criteria Decision Framework

Access to safe groundwater remains a critical challenge across sub-Saharan Africa, particularly in Togo where much of the rural population still lacks functional water infrastructure. Deciding \emph{where} to drill a borehole---balancing geological viability, population needs, poverty, and logistical constraints---currently relies on costly expert surveys, leaving high-potential aquifers untapped in the most vulnerable communities. This study presents an integrated, data-driven framework that maps groundwater potential and produces a ranked national list of priority borehole locations, directly usable by water engineers and programme managers.Methodologically, a binary classification model---a Random Forest algorithm with features curated by an XGBoost meta-evaluator---was trained on historical borehole records from the Togolese Ministry of Water Resources and Sanitation using 374 geospatial covariates spanning topography, hydrology, climatology, lithology, pedology, and land cover. The model achieves an AUC of 0.844, correctly ranking productive over unproductive sites in 84\,\% of sampled pairs. Calibrated probability scores were deployed across a 20\,m national inference grid (${\approx}12.9 \times 10^{6}$ points), yielding a continuous groundwater potential map for the entire territory. A Bayesian-optimized Analytic Hierarchy Process (AHP) then integrates this hydrogeological signal with demographic demand, infrastructure coverage gaps, socio-economic vulnerability, and road accessibility into a composite priority score. The quantile-based priority zones achieve 90.2\% precision and 84.8\% recall, identifying 6.25~million people in urgent need of water access intervention. The resulting maps and locality rankings constitute an operational, evidence-based roadmap for water infrastructure deployment in Togo, adaptable to any sub-Saharan territory with open geospatial data.