Scaling urban structure-temperature analysis: Global application of the STURLA classification to 11,214 urban areas

Urban heat is a growing threat to human health and livability, yet mitigation strategies based on individual-city studies may not generalize across climates and surface contexts. The Structure of Urban Landscapes (STURLA) classification uses grid-based characterization in which composite classes emerge empirically from observed combinations of land cover (trees, grass, bare soil, water, paved surfaces) and building height (lowrise, midrise, highrise). Prior applications in three temperate cities (Berlin, New York City, Philadelphia) showed consistent temperature signatures, but global generalization was untested. This study applies STURLA globally for the first time, classifying 5.16 million grid cells across 11,214 urban areas in 178 countries at 500 m resolution using ESA WorldCover, GHS-BUILT-C, and Landsat surface temperature data processed through Google Earth Engine. Of 220 observed classes, just 27 cover 90% of global urban area, confirming the structural concentration previously observed at city scale. Classes exhibit a 24 °C temperature range, with 89% of pairwise comparisons showing practically significant differences (≥1 °C). Temperature rankings are remarkably consistent across urban areas (mean Spearman correlation 0.71, median 0.86), confirming that structure-temperature relationships from three cities generalize worldwide. Component-level analysis reveals strong context-dependence: tree cooling is greatest over bare and paved surfaces (4.9 °C) and negligible near water (∼0 °C), water body cooling is climate-dependent (9.6 °C in arid vs. 3.4 °C in humid climates), and building height effects are modest but variable. These findings establish that STURLA can be applied consistently at global scale, that its structure-temperature relationships generalize across diverse urban contexts, and that structural and climatic context is central to effective urban heat mitigation.