Diagnosing the Mechanisms of Catastrophic Deforestation in an African Savanna City: An Integrated Geospatial and Statistical Modelling Approach

This study provides a diagnostic analysis of the mechanisms driving deforestation in Minna, Nigeria, between 2004 and 2024. We integrated multi-temporal Landsat imagery with a rigorously validated Random Forest classifier (Overall Accuracy: 92%) and a binomial logistic regression model (ROC-AUC = 0.87) that controlled for spatial autocorrelation. This integrated geospatial-statistical framework advances beyond descriptive mapping and conventional driver assessments by explicitly linking high-accuracy change detection with spatially explicit, inferential modeling of proximate causes, providing a scalable diagnostic protocol for data-scarce environments. Results reveal a catastrophic decline in dense vegetation cover, which decreased by 88% at an annual rate of 4.4%, while non-vegetated area expanded by 228%. The regression model identified proximity to roads (Odds Ratio: 1.8; 95% CI: 1.5–2.1) and existing built-up areas as the most significant predictors of deforestation. While accessibility-driven loss converges with global deforestation frontiers, the underlying land use, characterized by smallholder agriculture and biomass dependence, differs from large-scale commercial drivers. The study demonstrates a replicable, policy-relevant methodology for diagnosing deforestation drivers, offering a transferable template for actionable land-use zoning and sustainable governance in rapidly urbanizing savanna environments across West Africa and beyond.