From Cooling to Warming: Local Reversals in Urban Factor–Land Surface Temperature Relationships Using a Spatially Explicit Structural Modeling Framework

Urban heatwaves are intensifying under climate change, yet the relationships between urban morphological and environmental factors and land surface temperature (LST) remain insufficiently understood at fine spatial scales. Coarse thermal resolution and administrative or grid-based units can obscure building-level heterogeneity. An integrated analytical framework was proposed to identify microscale spatial heterogeneity and local reversals in urban factor–LST relationships. Landsat-derived LST was downscaled to 10 m using Sentinel-2 imagery, while building-centered tessellations representing building and road configurations served as spatial units. Two- and three-dimensional indicators of albedo, sky view factor, green and road environments, building density, and building scale were integrated into latent variables using partial least squares structural equation modeling (PLS-SEM). Latent variable scores were subsequently analyzed using multiscale geographically weighted regression (MGWR). Application to Jung-gu and Seo-gu in Daegu, South Korea, revealed positive and negative local relationships with LST for most factors. Albedo-related results differed between variable representations, with the underlying cause remaining unresolved, and the estimated spatial scale of the Albedo Environment (AE)–LST relationship was sensitive to MGWR estimation settings. Based on openly available data and general morphological indicators, the framework provides a transferable basis for locally tailored heat-mitigation strategies.