Beyond local climate zones: Continuous morphology archetypes for heat mitigation across 66 cities
Sustainable Cities and Society, cilt.150, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 150
- Basım Tarihi: 2026
- Doi Numarası: 10.1016/j.scs.2026.107814
- Dergi Adı: Sustainable Cities and Society
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Compendex, Geobase, INSPEC
- Anahtar Kelimeler: Building footprint data, Climate-smart design, Machine learning, Multi-city study, Satellite remote sensing, SHAP, Urban heat island, Urban morphology
- Ankara Yıldırım Beyazıt Üniversitesi Adresli: Evet
Özet
Urban block morphology is fixed at masterplan stage and locks in pedestrian-level thermal performance for 50–100 years, yet design-stage guidance for climate-smart morphology remains poorly generalised. Most published analyses are single-city, and the cross-city literature relies predominantly on Local Climate Zone categorical classes that collapse within-class morphology variance. We address this gap with a measurement-first multi-city framework (prioritising harmonised empirical measurement over categorical assumptions) built on a rigorous, fully open data stack: 33 715 cloud-masked one-square-kilometre patches harmonised across 66 cities and 15 major Köppen climate zones (33 498 retained after a physical-plausibility quality screen on the surface-UHI target), pairing 3-summer rolling-mean Landsat 8/9 surface-temperature anomalies with continuous morphology features extracted from the latest open global building-footprint-with-heights products (Microsoft Global ML, Google Open Buildings v3, GlobalBuildingAtlas, GLAMOUR), an ESA WorldCover land-cover mask, ERA5-Land summer air temperature, WUDAPT LCZ, and SRTM elevation. A linear mixed-effects model with city random intercept attains a marginal R2=0.25 on the within-city-centred UHI target and a conditional R2=0.39 (ICC =0.18); all 8 morphology features are statistically significant at the 95% level, though morphology still explains only a limited share of total surface-UHI variability, with climate and synoptic context dominant. Tree-canopy fraction is the dominant cooling lever (β=−0.64, p<10−122); water fraction (β=−0.58), 80th-percentile NDVI (β=−0.31), building-height coefficient-of-variation (β=−0.18) and mean building height (β=−0.09) also cool, while bare-soil (β=+0.42) and built-up (β=+0.31) fractions are the dominant warming factors. Per-zone mixed-effects models lift the patch-weighted pooled marginal R2 to 0.38, reaching 0.55 in Dwa and 0.51 in Dfb and Cwb. Tree-canopy fraction is the only universal lever, with a negative coefficient in all 14 zones and significance in 13. We extract 144 empirically Pareto-optimal patches in (UHI, density) space and cluster them into three exploratory climate-smart archetypes (silhouette =0.38). The harmonised patch-level table, ingest-to-archetypes pipeline, and archetype card library are released under an open licence.