This study was focused on the metropolitan area of Florence in Tuscany (Italy) with the aim of mapping and evaluating thermal summer diurnal hot- and cool-spots in relation to the features of greening, urban surfaces, and city morphology. The work was driven by Landsat 8 land surface temperature (LST) data related to 2015–2019 summer daytime periods. Hot-spot analysis was performed adopting Getis-Ord Gi* spatial statistics applied on mean summer LST datasets to obtain location and boundaries of hot- and cool-spot areas. Each hot- and cool-spot was classified by using three significance threshold levels: 90% (LEVEL-1), 95% (LEVEL-2), and 99% (LEVEL-3). A set of open data urban elements directly or indirectly related to LST at local scale were calculated for each hot- and cool-spot area: (1) Normalized Difference Vegetation Index (NDVI), (2) tree cover (TC), (3) water bodies (WB), (4) impervious areas (IA), (5) mean spatial albedo (ALB), (6) surface areas (SA), (7) Shape index (SI), (8) Sky View Factor (SVF), (9) theoretical solar radiation (RJ), and (10) mean population density (PD). A General Dominance Analysis (GDA) framework was adopted to investigate the relative importance of urban factors affecting thermal hot- and cool-spot areas. The results showed that 11.5% of the studied area is affected by cool-spots and 6.5% by hot-spots. The average LST variation between hot- and cold-spot areas was about 10 °C and it was 15 °C among the extreme hot- and cool-spot levels (LEVEL-3). Hot-spot detection was magnified by the role of vegetation (NDVI and TC) combined with the significant contribution of other urban elements. In particular, TC, NDVI and ALB were identified as the most significant predictors (p-values < 0.001) of the most extreme cool-spot level (LEVEL-3). NDVI, PD, ALB, and SVF were selected as the most significant predictors (p-values < 0.05 for PD and SVF; p-values < 0.001 for NDVI and ALB) of the hot-spot LEVEL-3. In this study, a reproducible methodology was developed applicable to any urban context by using available open data sources.