Spatiotemporal dynamics and drivers of surface urban heat island intensity in Dhaka, Bangladesh using geospatial remote sensing

Abstract

Rapid urbanization has a profound impact on urban thermal environments, particularly in densely populated megacities. This study aims to assess the spatiotemporal dynamics of land surface temperature (LST), surface urban heat island (SUHI) intensity, land use/land cover (LULC) changes, and their primary driving factors in Dhaka City Corporation (DCC), Bangladesh, during the pre-monsoon season from 1990 to 2024. Using multi-temporal Landsat imagery and ancillary spatial datasets on Google Earth Engine (GEE), we observed a 194.54% increase in urban land, alongside substantial declines in vegetation (45.99%) and water bodies (75.57%). These classifications were validated with strong accuracy metrics (Overall Accuracy > 92%, Kappa > 89%, K-fold > 95%). Concurrently, LST rose dramatically, with areas exceeding 40 °C expanding to 70.06% by 2024, and traditional cool zones nearly disappearing. SUHI analysis showed a growing heat island effect, with recent nocturnal SUHI values surpassing daytime levels, indicating increased nighttime heat retention. Correlation and multiple linear regression analyses (R² = 0.801) identified NDBI, population density, and nighttime light intensity as the dominant positive drivers of LST, while vegetation and water bodies offered significant cooling potential. These findings highlight the urgent need to integrate ecological benchmarks (such as the 3-30-300 rule) and Sponge City designderscorn paradigms into Dhaka’s climate-resilient urban planning framework. Ultimately, this cloud-based multi-factor modeling framework provides a highly replicable, open-access blueprint to help rapidly growing megacities worldwide diagnose thermal vulnerability and mitigate escalating heat stress.