From above and below: integrating remote sensing and citizen science weather stations to model Local Climate Zones and Urban Heat Islands in Padua

As climate change intensifies urban heat exposure, understanding intra-urban temperature variability is essential for adaptation. This thesis integrates satellite-based Local Climate Zone (LCZ) mapping with observations from institutional and citizen-science weather stations in Padova (northeastern Italy). The general aim is to investigate intra-urban microclimatic variability and urban heat island patterns during summer 2026, establishing a methodological and observational baseline for urban climate monitoring, planning and adaptation. Specifically, the study aims to: i) map and classify LCZs; ii) assess the potential of a citizen-science-based meteorological network to capture spatial and temporal temperature variability; and iii) examine relationships between observed thermal patterns and LCZs. This study produced the first high-resolution (100 m) LCZ map of the Municipality of Padova, representing its main urban morphological and land-cover configurations. An adapted WUDAPT workflow combined Sentinel-1 and Sentinel-2 imagery with a Random Forest classifier in Google Earth Engine. Validation against held-out test polygons yielded an overall accuracy of 77.4% and a Cohen’s kappa of 0.74. Meteorological observations from 18 stations were analysed for June-August 2026. A subset of 13 stations supported the interpolation of seasonal mean, mean daily minimum and mean daily maximum temperatures using Inverse Distance Weighting and Ordinary Kriging. Leave-one-out cross-validation assessed interpolation performance, and the resulting temperature surfaces were compared across LCZs. The LCZ map distinguished Padova’s compact centre, open residential districts, industrial areas and vegetated outskirts. The strongest thermal differentiation concerned minimum temperatures: seasonal station averages ranged from 19.1 to 22.3 °C, with warmer conditions in compact built-up areas and lower minima in open or vegetated settings. Differences between mean interpolated temperatures of aggregated LCZ groups were approximately 1.0-1.1 °C for minimum temperature, compared with about 0.3-0.4 °C for mean and maximum temperatures. Both interpolation methods yielded broadly comparable cross-validation errors, with neither consistently outperforming the other. These findings suggest that urban compactness was more clearly associated with reduced nocturnal cooling than with higher daytime maxima, consistent with the predominantly nocturnal expression of the urban heat island. Despite limitations arising from the small, unevenly distributed station network and single-season analysis, the integrated framework provides a basis for urban heat assessment. It also supports targeted monitoring expansion towards underrepresented LCZs, improving observational coverage and informing climate adaptation planning.