Unravelling the Spatio-Temporal Relationship between Land Use and Land Surface Temperature Through Longitudinal Geospatial Modelling in South-Eastern Bangladesh

This study investigated the relationship between land use changes and land surface temperature (LST) in Southeastern Bangladesh, particularly focusing on Chittagong. By integrating remote sensing and geospatial analysis, along with predictive modelling techniques, we examined the environmental impacts of rapid urban expansion over the past three decades and forecasted future scenarios up to 2043. Utilizing Landsat satellite imageries from 1993 to 2023, we quantified how urbanization influenced in the ecological landscapes and temperature patterns. Our analysis revealed that built-up areas expanded from 62.11 km(2) (1.45%) in 1993 to 722.36 km(2) (16.52%) in 2023, while vegetation and barren land decreased by 271.59 km(2) (-6.22%) and 209.02 km(2) (-4.78%), respectively. These land use and land cover (LULC) transitions were closely linked to rising LST, especially in urban regions, where the temperature peaked at 40.77 degrees C in 2023. Moreover, the positive association with NDBI values further highlighted the heat retention of urban structures. To achieve future projection, an integrated Artificial Neural Networks (ANN) and Geographically Weighted Regression (GWR) based predictive model was developed to simulate the spatio-temporal patterns of LULC and LST changes. LST is expected to continue rising, 41.89 degrees C in 2033 and 42.76 degrees C in 2043, exacerbating the urban heat island (UHI) effect. These findings emphasize the need for sustainable urban planning and green infrastructure to mitigate the adverse effects of urbanization and climate change, particularly in rapidly urbanizing coastal cities like Chittagong.Graphical AbstractThe graphical abstract presents an integrated analysis of land use and land surface temperature (LST) changes in Chittagong, Bangladesh, using remote sensing and AI-based predictions. The Landsat imagery (1993-2023) was used and to project future scenarios, a predictive model integrating Artificial Neural Networks (ANN) and Geographically Weighted Regression (GWR) was developed to simulate the spatiotemporal dynamics of LULC and LST changes. The LULC change maps reveal significant urban expansion, with built-up areas increasing while vegetation and barren land decline. Correspondingly, the LST change maps indicate a steady rise in temperature, especially in urbanized regions. Statistical analyses, including correlation studies, reveal a strong positive relationship between NDBI and LST (R-2 = 0.337), while NDVI shows a negative correlation (R-2 = 0.126), emphasizing the heat-retention effects of urban structures. Future projections indicate that LST will reach 42.76 degrees C by 2043 with an r(2) of 0.952, the model explains 95% of the variation, indicating an excellent fit. The coefficient 0.173 suggests that the maximum temperature increases by approximately 0.17 degrees C each year, confirming a clear warming trend in Chittagong district. With 95% confidence level, the true annual increase lies between 0.134 degrees C and 0.212 degrees C. Additionally, environmental impact assessments using landscape metrics, hotspot analysis, and GWR offer spatial insights into the distribution of LST on LULC, as well as the accuracy of predictive modelling. Overall, the study results provide actionable insights for policymakers to mitigate UHI effects through targeted interventions such as increased vegetation, sustainable urban design, and green infrastructure development.