Satellite-Based Detection of Urban Heat Islands Using Machine Learning: A Global Analysis

Urban heat islands (UHIs) affect over 3.5 billion urban residents globally, contributing to increased mortality, energy consumption, and air pollution. We analyze UHI effects across 500 global cities using Landsat-8 thermal infrared data combined with Sentinel-2 multispectral imagery, applying a gradient boosting model that explains 87% of UHI variance. Population density, impervious surface fraction, and green space ratio emerge as dominant predictors, with coefficients varying substantially across climate zones. Our model identifies 23 cities with anomalously high UHIs relative to population density—hotspots for urban greening interventions. Seasonal analysis reveals peak UHI intensities in summer afternoons, with monsoon climates showing distinctive attenuation patterns.