Assessment of urban expansion susceptibility in major urban units of Bangladesh leveraging machine learning and geostatistical approach

Abstract

This study examines the governing factors and susceptibility zones for urban expansion in Bangladesh's major urban areas, including Dhaka, Barisal, Chittagong, Comilla, Narayanganj, Gazipur, Khulna, Sylhet, Rajshahi, Rangpur, and Mymensingh. The main goals of this research are to determine the impact of governing factors and to identify susceptibility zones for urban expansion in major urban units using a data-driven approach. By using governing factors (DEM, Slope, LST, NDVI, Population, distance to (industry, growth center, settlement, facilities, waterbody, road), and machine learning (Random Forest) and geostatistical approach (Binary Logistic Regression), the research identifies the most important factors influencing urban expansion, including NDVI, LST, waterbodies, roads, and settlements. The RF model's ROC-AOC values showed the highest accuracy (1.00) in Comilla and Mymensingh, moderate accuracy (0.99) in Barisal, Chittagong, Narayanganj, Gazipur, Khulna, and Rajshahi, and lower accuracy in Dhaka (0.98), Sylhet (0.89), and Rangpur (0.85). For the Binary Logistic Regression model, Comilla, Narayanganj, Gazipur, and Mymensingh had the best fit (Nagelkerke R 2 ​= ​1.00), while Sylhet had the lowest significance (0.482). Furthermore, Khulna, a major urban unit, is the highest urban expansion susceptibility zone which is 35.72%. Rajshahi and Barisal are the moderate and low urban expansion susceptibility where 83.17% and 0.88% respectively. This unplanned and rapid urban expansion zone has also confronted policymakers and planners with an insurmountable challenge and stressed local governments' ability to manage and use their scarce land-based resources with geospatial data. Thus, this study's machine learning and geostatistical findings will help explain land cover change and urban expansion in Bangladesh's eleven metropolitan areas. This study will improve urban development understanding in Bangladesh. Findings will help planners, stakeholders, and policymakers understand urban expansion patterns, enabling better environmental planning. • Addresses the understudied phenomenon of urban expansion in Bangladesh, focusing on its major urban units amidst rapid growth in South Asia. • Leveraging machine learning (Random Forest) and geostatistical methods (Binary Logistic Regression Model), the research assesses the impact of various governing factors on urban growth susceptibility zones. • NDVI, LST, waterbody, road, and settlement emerge as critical factors influencing urban expansion in specific urban units, providing valuable insights for land use planning and management. • Regional variations in urban expansion susceptibility highlight the necessity for tailored strategies, with Khulna at 37.72% being the highest, and Rajshahi and Barisal at 83.17% and 0.88% respectively. • The study offers insights for policymakers and planners to address rapid urban expansion, informing future research and policy decisions in urban studies and land management.

Topics & Concepts