A FORECAST ANALYSIS OF URBAN EXPANSSION USING TIME-COMPOSITE CHANGE DETECTION THROUGH MACHINE LEARNING TECHNIQUES

Abstract

The concept of Land Use and Cover Change (LUCC) encompassed the modifications occurring on the surface of Earth. Land use pointed to deliberate human actions on land, while land cover described the physical attributes of the land surface. LUCC was a crucial area of research in the context of global environmental change, particularly in developing and underdeveloped countries facing climate changes and urban expansion. Rapid urbanization and the scarcity of agricultural land were major drivers of LUCC on a global scale. This study proposed a novel approach to improve the accuracy of land cover classification and enhance the detection of urban land cover changes. Using remotely sensed Landsat images from 2000 to 2020 and a multi-date composite change detection technique, this research analysed the spatial dynamics and Urban expansion in Faisalabad, Division, Pakistan, over the past two decades. Google Earth Engine (GEE) was utilized to collect data from Landsat satellite. A Normalized Difference Vegetation Index (NDVI) was employed to extract features from the satellite images. Classification and Regression Tree (CART) , Random Fores (RF) and Support Vector Machines (SVM) were used as classification algorithms to classify the satellite images into different land cover classes for their ability to monitor expansion across four time periods (2000, 2010, 2015, 2020) in study area. The classification results indicate that SVM outperforms RF and CART, achieving a high accuracy of 98.03%. RF follows with 96.7% accuracy, while CART attains 73.63% accuracy. Based on Landsat results, there had been a decrease in vegetation and agricultural fields, accompanied by an increase in urban areas. The built-up areas had grown to over 60% in 2020, and agricultural land, water bodies, vegetation, and barren ground had experienced continuous reductions. The study highlighted the importance of forecasting agricultural land use changes in Pakistan through historical land use and land cover change detection.