Urban Change Detection and Prediction using SVM and CA Markov Model for Prayagraj City

Urban planning is an important aspect of land management due to the rapid growth in built-up structures, better known as urbanization. Accordingly, urbanization change detection with remote sensing technology has been a hot spot for research in recent times. In the present work, a method for detecting changes using Sentinel-2 optical data is proposed over a large area of Prayagraj City, Uttar Pradesh, India. A hybrid prediction model involving a support vector machine (SVM) and cellular automata-Markov (CA Markov) model is proposed to find the changes in the earth’s features, viz., built-up, vegetation, water bodies, and barren land, followed by prediction of urban growth. In the classification, the mixing of two earth surface features, i.e., built-up and bare soil is observed. To avoid this mixing, the spectral responses of different bands of Sentinel-2 are analyzed, and after a rigorous analysis, it is observed that bands 3, 4, 11, and 12 are suitable for the study of urban feature extraction. The baseline years for change detection are chosen as 2015 and 2020, and the transition probability matrix (TPM) is computed using classified images of 2015 and 2024. Based on TPM, the Land Use Land Cover (LULC) for the year 2030 is predicted using the Markov model, which is spatially mapped with CA. It is predicted that built-up cover will continue to increase while vegetation and barren land decrease during 2024 and 2030. This study will help to make a plan to control unplanned development and assist the vision of smart city development.