Explainable geo-informatics for spatial solar suitability analysis in Gauteng Province, South Africa

Abstract The global transition toward low-carbon energy systems requires spatially explicit, transparent, and sustainable approaches for identifying suitable renewable energy sites, particularly in rapidly urbanising regions. This study develops an explainable geospatial artificial intelligence framework to support sustainable solar energy planning in Gauteng Province, South Africa, where competing land uses and limited ground-based solar data constrain evidence-based decision-making. The objective was to identify environmentally and operationally suitable areas for solar development while ensuring model transparency and spatial interpretability. Remote sensing-derived vegetation, surface moisture, built-up land, and topographic variables were integrated into a gradient boosting regression model, and feature contributions were interpreted using explainable artificial intelligence techniques. Spatial clustering was applied to classify suitability zones. The model demonstrated strong internal predictive performance (area under the curve = 0.98). Results show that surface moisture conditions and elevation exert the strongest influence on suitability, with large portions of the province exhibiting high relative suitability and distinct spatial hotspots emerging in central regions. The findings demonstrate that urban and peri-urban landscapes offer substantial sustainable solar potential. This study provides a transparent, scalable, and policy-relevant decision-support framework that advances sustainable energy planning in data-scarce environments, particularly across rapidly developing regions of the Global South.