Detection and mapping of land use land cover with Support vector machines SVM-based change monitoring using Landsat and Sentinel-2 data. The case of Quseir, Red Sea

Abstract

Land use land cover (LULC) mapping and spatial change monitoring are essential to manage the coastal cities and its resources. This study aims to investigate LULC changes of Quseir, a typical small Red Sea coastal city for a long period from 1984 to 2021. Support vector machines (SVMs) classifier and Post classification comparison (PCC) change detection technique were used to analyse two Thematic Mappers (TM), an Enhanced Thematic Mapper plus (ETM+), an Operational Land Imager (OLI) Landsat imagery in addition to a Multi-Spectral Instrument (MSI) Sentinel-2 image cover the study period. Twelve LULC classes have been identified for this study. Accuracy of the classified images and the LULC change were analysed. Along the thirty seven years, Quseir's urban has increased by nine folds and the green area was increased from nil to 6.10 m2 per person. SVM achieved high accuracy classification results for all the studied images of all sensors, while the MSI, 2021, was the highest accuracy. Through data-resampling, combining Landsat and Sentinel-2 satellite datasets for long-term monitoring studies using PCC resulted in more reliable and accurate outputs. Results obtained from this study will fill the gap of rare LULC maps and spatial change information of Quseir during the past four decades.