| Local climate, land cover, topography, and socio-economic factors shape urban temperature patterns, with Urban Heat Islands (UHIs) posing significant environmental and societal challenges. Surface UHI (SUHI) analysis often relies on remote sensing of Land Surface Temperature (LST), but temporal limitations complicate assessment. This study applies a holistic framework that combines remote sensing, geospatial analysis, and machine learning (ML) to parameterise, correlate, and quantify SUHI drivers in Sofia using LST and spectral indices from five Landsat 8/9 images (2017, 2022) processed in Google Earth Engine. Predictor variables included population and building data (GHSL), land cover (Sentinel-2, Urban Atlas), elevation, and road networks. Among the five algorithms tested, Random Forest achieved the best performance. Sensitivity analysis showed terrain elevation, water distance, land cover, building height, and road distance as the most influential predictors, with building height having a much stronger effect under SUHI conditions. The study’s main contribution lies in systematically linking land cover, morphology, and topography to LST, providing insights into urban heat drivers and offering a transferable framework for heat-risk mapping and climate-resilient planning. |
*** Title, author list and abstract as submitted during Camera-Ready version delivery. Small changes that may have occurred during processing by Springer may not appear in this window.