Spatiotemporal dynamics of land surface temperature and its drivers within the local climate zone framework

Abstract

The relationship between land surface temperature (LST) and urban morphology, particularly through the lens of Local Climate Zones (LCZs), has attracted increasing scholarly interest. However, the spatiotemporal dynamics of LST and its underlying drivers, both within and across LCZ types, remain insufficiently explored. This study integrates remote sensing and geospatial big data to investigate the differentiated mechanisms shaping LST dynamics within the LCZ framework. Taking Nanjing as a case study, we first used the World Urban Database and Access Portal Tools to classify LCZs for the years 2012 and 2022, and derived corresponding LST datasets from Landsat imagery. We then employed a Geographically Weighted Random Forest (GWRF) model to systematically examine the spatial and temporal dynamics of LST and its key drivers. Results show that mean LST increased from 39.22 °C in 2012 to 40.61 °C in 2022, with built-up LCZs consistently 5–7 °C hotter than vegetated or water-dominated zones. Vegetation demonstrated the strongest cooling capacity (NDVI reduced LST by up to 15 °C), whereas population density contributed to warming (up to +6 °C). Importantly, the magnitude and direction of driver effects varied significantly across both time and LCZ types, with landscape pattern metrics (e.g., CONTIG, FRAC) gaining influence over the decade. These findings highlight the importance of the LCZ framework for capturing heterogeneous spatiotemporal patterns of urban thermal environments and provide context-sensitive guidance for mitigating urban heat in rapidly urbanizing regions.