Low Built-Up Density Emerges as a Key Driver of Heat Extremes over the Sahel

Urban heat islands threaten fast-growing Sahelian cities, yet causal drivers of surface heating remain unknown. Here we combine machine-learning classification (XGBoost, Random Forest, SVM) with spatial causal inference to disentangle correlation from causation among hotspot drivers in Ouagadougou, Burkina Faso. XGBoost generalised best (F1 = 0.70, κ = 0.67) while SHAP analysis identified built-up density as the dominant predictor. Geographical convergent cross-mapping confirmed it as a unidirectional cause of surface temperature, while spectral indices showed only bidirectional coupling despite strong correlations. Opposite to humid tropical cities, lower built-up density increases hotspot risk due to exposed bare soil. These findings point to compact urban form as a heat mitigation strategy.

This micropublication was created as part of the Neuromatch Impact Scholars Program 2025.

Project Website: https://impact-scholars.git…

Repository: https://github.com/impact-s…