Spatiotemporal changes of surface water and its drivers using geographically weighted regression: A case study of Gazipur City corporation, Bangladesh

Abstract Understanding the changes of surface water (SW) in rapidly urbanizing regions is crucial for sustainable urban planning and environmental management. This study investigates the spatiotemporal variation of surface water in Gazipur City Corporation (GCC), Bangladesh, and explores the influence of land use/land cover (LULC) change and topography on SW distribution. Landsat-based surface reflectance imagery from 2004 and 2023—pre- and post-establishment of GCC—was used to assess seasonal variations and long-term LULC changes. Classification of LULC types (impervious, vegetation, bare, and waterbody) was performed using the Random Forest (RF) algorithm in Google Earth Engine (GEE), supported by spectral indices and ground control points (GCPs). Seasonal changes in surface water were assessed using the optimum threshold value of the Modified Normalized Difference Water Index (MNDWI) obtained by Receiver Operating Characteristic (ROC) curve analysis, while the rate of change in LULC was calculated for 500 spatial sub-regions. To explore spatially varying relationships between surface water loss and potential drivers, a Geographically Weighted Regression (GWR) model was applied using Python. Results show a significant reduction in both permanent and seasonal surface water over the study period, primarily in areas undergoing rapid urbanization. GWR results revealed strong negative relationships between surface water change and impervious, bare, and vegetation change rates, while elevation was found to be statistically insignificant (p = 0.88). The study emphasizes the importance of localized modeling in understanding hydrological responses to urban growth and provides valuable insights for flood risk management and sustainable land planning in urban Bangladesh.