This paper focuses on the relationship between remotely-sensed urban site characteristics (USCs) and land surface temperature (LST). Particular emphasis is put on an extensive comparison of two-dimensional (2D) and three-dimensional (3D) USCs as potential indicators of the surface urban heat island (UHI) effect and as potential predictors for thermal sharpening applications. Both widely-used as well as more recently proposed metrics of the urban remote sensing literature are investigated within a single experiment. While some of these USCs have already been used earlier, others have never been analyzed before in the context of urban temperature studies. In addition to the comparison of 2D and 3D USCs, the spatio-temporal dependencies of their relation to LST are examined. To this end, the experimental setup of this work includes two study areas, 26 USCs, and 16 LST scenes covering four seasons. Use is made of a comprehensive database compiled for the cities of Berlin and Cologne, Germany. After data preparation, very high resolution (VHR) multi-spectral and height data are employed to map fine-scale urban land cover (LC). The resulting LC maps are then used in conjunction with the height information to compute 2D and 3D USCs. Subsequently, multi-temporal LST images are retrieved from Landsat Enhanced Thematic Mapper Plus (ETM+) scenes. The spatio-temporal investigation of the USC–LST connection constitutes the final stage of the workflow and is achieved in the framework of a dedicated correlation analysis. The results of this study highlight that the linkage between USCs and LST sensed at small scan angles is not stronger when 3D parameters are considered. Even though they may offer more holistic representations of the urban landscape, 3D USCs are consistently outperformed by some of the most widely-used 2D metrics. The analysis of spatial dependencies reveals that the USC–LST interplay does not only differ between, but also within the two test sites. This is due to their distinct geographies, with urban form and compactness, green spaces and street trees, and the structural composition of LC elements being some of the determining factors. The examination of temporal dependencies yielded that the association between USCs and LST is fairly stable over time but can be subject to larger inter- and intra-season variations for different reasons, including the season of acquisition, vegetation phenology, and meteorological conditions. Since previous research was based on the analysis of a single study area, a limited number of (mainly 2D) USCs, and/or only a few LST scenes acquired in specific seasons, it is concluded that the findings of this study provide researchers and practitioners with a more complete picture of the USC–LST relationship. Display omitted •Spatio-temporal analysis of the statistical relationship between 2D/3D USCs and LST•Detailed inspection of 2 study areas, 26 USCs, and 16 LST scenes covering 4 seasons•3D USCs are consistently outperformed by some of the most widely-used 2D indicators.•Correlations are spatially dependent due to the distinct geographies of the cities.•Larger inter-/intra-season variations are mainly driven by environmental conditions.