Semi-Unsupervised Threshold-Based and Jenks Optimization Classification for Analyzing Urban Expansion in BSD City Using Multi-Temporal Landsat Data

Abstract

This study analyzes land cover changes in BSD City from 2002 to 2023 using the Semi-Unsupervised Threshold-Based and Jenks Optimization Classification (SUT-JOC) method. Five land cover types were mapped for 2002, 2013, and 2023: built-up area, bare soil, vegetated area, sparsely vegetated area, and water bodies. The results show significant urban expansion from 2002 to 2023, with built-up areas increasing from 2,202 hectares to 4,498 hectares. Land transitions followed a clear pattern, starting from vegetated areas to bare soil, then to built-up land, with spatial development shifting from east to west. Classification accuracy ranged from 88% in 2002 to 96% in 2023, with Kappa values approaching 0.95. The integration of NDVI, MNDWI, and NDBI with rule-based classification reduced manual intervention and improved reliability. These findings highlight the dynamics of peri-urban growth, environmental impacts such as urban heat island effects, and offer insights for sustainable urban planning. SUT-JOC provides a scalable and interpretable framework for multi-temporal urban land monitoring.