Spatio-Temporal Patterns of Land Use/Land Cover Change in the Heterogeneous Coastal Region of Bangladesh between 1990 and 2017

Although a detailed analysis of land use and land cover (LULC) change is essential in providing a greater understanding of increased human-environment interactions across the coastal region of Bangladesh, substantial challenges still exist for accurately classifying coastal LULC. This is due to the existence of high-level landscape heterogeneity and unavailability of good quality remotely sensed data. This study, the first of a kind, implemented a unique methodological approach to this challenge. Using freely available Landsat imagery, eXtreme Gradient Boosting (XGBoost)-based informative feature selection and Random Forest classification is used to elucidate spatio-temporal patterns of LULC across coastal areas over a 28-year period (1990–2017). We show that the XGBoost feature selection approach effectively addresses the issue of high landscape heterogeneity and…