Machine Learning-Driven Mapping of Heatwave Health Risks Across Local Climate Zones in a Mediterranean Context

Mediterranean cities are increasingly vulnerable to extreme heat events, driven by rapid urbanization and climate change. This study proposes a high-resolution framework for assessing heat-health risk (HHR) in Algiers, Algeria, by integrating the Heat Health Risk Index (HHRI) and Surface Urban Heat Island (SUHI) metrics within the Local Climate Zone (LCZ) classification 2020 system. Drawing on multi-temporal satellite data (2015–2023), demographic information, and meteorological records, we generated hazard, exposure, and vulnerability layers, with variable weighting derived from Principal Component Analysis (PCA). SUHI was estimated using Landsat-based Land Surface Temperature (LST) data, referencing rural LCZs as thermal baselines. Unsupervised K-means clustering was employed to classify combined HHRI–SUHI data, revealing four distinct urban heat risk profiles. The results indicate that LCZs 4, 5, and 8 are most affected by compounded heat-health risks, while LCZs 4, 6, and 8 display the highest surface heat accumulation. The resulting typologies identify zones where thermal stress intersects with social vulnerability, offering valuable insights for targeted adaptation. This is the first study in North Africa and the Mediterranean to apply this integrated clustering approach, demonstrating its applicability to other heat-prone, data-scarce urban environments.