Modelling scale-dependent urban thermal dynamics through landscape structure: A multiscale spatial analysis of Lucknow, India

Rapid urbanization across the Global South is reshaping urban thermal environments, yet the scale-dependent mechanisms linking landscape composition and configuration to land surface temperature (LST) remain poorly quantified in rapidly growing Tier-2 cities. This study aims to evaluate the multiscale spatial relationships between urban green spaces (UGS), urban built-up surfaces (UBS), landscape configuration metrics, and LST across the Lucknow Municipal Corporation (LMC), India, from 2003 to 2025. Using multi-temporal Landsat Collection 2 Level-2 imagery, landscape metrics, spatial statistics, and directional transect profiling, we tracked multi-decadal landscape shifts and calibrated spatial regression models (OLS, GWR, and MGWR) on the contemporary 2025 baseline ( ). Between 2003 and 2025, built-up surfaces expanded from to , while vegetation cover contracted and fragmented, accompanied by a increase in mean surface temperature across the pre-monsoon observation dates ( to ). Among the evaluated models, the Multiscale Geographically Weighted Regression (MGWR) model provided the best empirical fit ( , Adjusted , ) compared to global OLS ( , Adjusted ) and standard GWR ( , Adjusted ). MGWR revealed distinct scale dependencies: green space configuration (UGS Edge Density) operated at a localized neighborhood scale (bandwidth = ), whereas built-up land-cover variables operated across broad municipal scales (bandwidths ). Directional transects confirmed the outward migration of thermal anomalies and intensified peri-urban heating. While based on instantaneous pre-monsoon satellite snapshots and two-dimensional landscape metrics, these findings demonstrate that urban thermal mitigation requires scale-tailored planning: preserving compact, connected green patches at the neighborhood level while managing impervious expansion through municipal-wide master planning.