Spatio-temporal change detection of built-up areas with Sentinel-1 SAR data using random forest classification for Arnavutköy Istanbul

Abstract

As one of the most populated cities in Turkiye and the world, the Istanbul metropolis has always attracted the crowd of people masses. Arnavutköy Town has become one of the critical points of Istanbul City with increasing built-up areas (BAs). The spatial-temporal change detection of the expansion of the BAs of this district is essential data on behalf of Istanbul City. This research aims to determine urban areas expansion zones, also defined as the BAs footprint, from Sentinel-1 radar data. The determination of Sentinel-1A data of the urban area change detection encountered in Arnavutköy Town between 2018-2021 with Random Forest (RF) classification machine learning algorithm is investigated in this study. Based on the changes in spatial-temporal data, the direction of urban development has been determined. In addition, to visually compare the Normalized Difference Built-up Index (NDBI) and optical Sentinel-2A's false color urban RGB composite, which is a distinct data format, the processes have been proved. As a result of the study, SAR satellite data was found to be more appropriate than optical satellite data since not being affected by atmospheric conditions for extracting BAs with remotely sensed data.