Determinazione delle Local Climate Zones da immagini satellitari per l’analisi delle isole di calore urbane: il caso di Bologna

This thesis addresses the issue of Urban Heat Islands (UHIs) and their relationship with urban morphology through the classification of Local Climate Zones (LCZ), applied to Bologna. The main objective was to produce high-resolution LCZ maps calibrated on the city’s specific characteristics. Multispectral data (Sentinel-2) and hyperspectral data (PRISMA) were used, integrated with ancillary variables such as building height, built-up fraction, soil impermeability, Sky View Factor, albedo, and land surface temperature. Training and validation polygons were digitized in QGIS, while the classification was carried out in Google Earth Engine (GEE) using the Random Forest algorithm. The results showed that overall accuracy depends on the integration of additional variables and on pre-processing. Sentinel-2 with 18 bands achieved 81% accuracy, while PRISMA reduced to 15 bands through Principal Component Analysis (PCA) provided the best result with 82.3%. These outcomes confirm that although hyperspectral data offer strong potential, Sentinel-2 represents a more sustainable solution for repeated applications over time, thanks to its finer spatial resolution and higher acquisition frequency. The comparison with global datasets at 100 m resolution highlighted the importance of local-scale studies, which allow a more detailed description of the urban fabric. The inclusion of derived layers proved essential for improving class discrimination, demonstrating that spectral information alone is not sufficient in complex contexts. Finally, the work explored future perspectives, particularly the role of citizen science and low-cost tools such as MeteoTracker, which enable distributed collection of microclimatic data in motion. These participatory approaches can complement satellite observations, enriching datasets and providing valuable insights for land management and climate change adaptation strategies.