Groundwater Potential and Borehole Priority Zone Datasets for Togo

Overview This dataset contains three geospatial output files produced by the analytical pipeline described in the associated article. Together they constitute an operational, national-scale decision support tool for borehole site selection in Togo. The pipeline consists of: 1. A Random Forest classifier (374 geospatial covariates, XGBoost feature pre-selection, sigmoid calibration) trained on historical borehole records from the Togolese Ministry of Water Resources and Sanitation to predict the probability of a productive aquifer at any location. 2. A Bayesian-optimized Analytic Hierarchy Process (AHP) that integrates the ML probability with five additional socio-demographic and logistical criteria to produce a composite intervention priority score. 3. A Delaunay triangulation + dissolve step to convert the 12.9-million-point inference grid into vector geometries. Territory covered: Togo (excluding the Greater Lomé Metropolitan Area for the priority analysis). Reference coordinate system: EPSG:32631, WGS 84 / UTM Zone 31N. Files 1. togo_groundwater_potential_20m.tif Type: GeoTIFF raster — single band Resolution: 20 m (approx. 12.9 million pixels covering Togo) CRS: EPSG:32631 (WGS 84 / UTM Zone 31N) No-data value: pixels outside Togo boundary are masked Band Field Type Unit Description 1 proba_calibree Float Sigmoid-calibrated Random Forest probability of encountering a productive aquifer (flow rate ≥ 0.8 m³/h). Range: 0.097 – 0.899. Values close to 1 indicate high groundwater potential; values close to 0 indicate low potential. Productivity threshold: A borehole is classified as *productive* (positive class, Y=1) when its measured flow rate Q ≥ 0.8 m³/h (≥ 800 L/h), consistent with the threshold applied by the Togolese Ministry of Water Resources and Sanitation and with the Sub-Saharan literature. Model performance (test set): AUC = 0.844 2. togo_potentiel_eau_souterraine_delaunay.gpkg Type: GeoPackage (vector) — layer triangles Geometry type: Polygon (Delaunay triangles) CRS: EPSG:32631 Number of features: 599,971 triangles (generated from a probability-weighted sample of 300,000 points drawn from the 12.9-million-point inference grid) Total area: ~56,703 km² Field Type Unit Description i Integer Index of the first vertex of the triangle in the sampled point array j Integer Index of the second vertex k Integer Index of the third vertex centroid_x Float X-coordinate of the triangle centroid (EPSG:32631) centroid_y Float Y-coordinate of the triangle centroid (EPSG:32631) area_m2 Float m² Triangle area (computed from vertex coordinates via the shoelace formula) area_m2_real Float m² Actual geometric area of the polygon after clipping to Togo boundary area_km2 Float km² area_m2_real / 1000000 proba_value Float Mean calibrated RF probability across the 3 vertices. This is the continuous groundwater potential signal. geometry Polygon Triangle geometry clipped to Togo national boundary How to use: This file is the vector equivalent of togo_groundwater_potential_20m.tif. It supports spatial queries (e.g., extracting potential values at a given location) in environments where raster processing is not available. For continuous analysis, the raster TIF is preferable. 3. togo_borehole_priority_zones.gpkg Type: GeoPackage (vector) — dissolved priority zones Geometry type: MultiPolygon CRS: EPSG:32631 Number of features: 4 (one per priority class) Field Type Unit Description optimized_quant_label String Priority class label. One of: Faible (Low), Modérée (Moderate), Élevée (High), Très élevée (Very High). Derived from quantile-based classification of the composite AHP score. area_km2 Float km² Total area of the dissolved zone geometry MultiPolygon Zone boundary, clipped to Togo national boundary Priority class boundaries (composite AHP score, quantile-based Borehole Training Data The borehole dataset used to train the Random Forest model was provided by the Togolese Ministry of Water Resources and Sanitation under a data-sharing agreement. It is not included in this deposit and cannot be made publicly available without Ministry approval. Requests may be directed to the corresponding author. Total boreholes in inventory with available discharge: 8690 (national coverage) Training/test split: 80 / 20 (stratified, random_state = 42) Class balance (training set): 44.8 % positive (productive), 55.2 % negative Positive label: Q ≥ 0.8 m³/h (800 L/h), converted from field bucket-fill measurements