Monitoring desertification in Biskra, Algeria using Landsat 8 and Sentinel-1A images

Abstract

[EN] Desertification is the persistent degradation of ecosystems caused by environmental changes and human activities. This is a global problem closely related to climate change with severe consequences in urban locations. For that reason, monitoring those locations with low cost and freely available satellite images could be useful for local agencies. This work studies a strategy for the observation of desertification in Biskra (Algeria) with Landsat 8 images and Synthetic Aperture Radar (SAR) images from Sentinel-1A satellite. Radar images are now available from a growing number of missions. These microwave images add valuable additional information to the existing optical products involving soil roughness and moisture content. They are also a very valuable tool to detect man made objects. However, radar images are still difficult to exploit due to their inherent Speckle noise and their characteristics. This study searches the best methodology for the insertion of radar data to a previously designed approach that uses optical data. Several algorithms are implemented and evaluated and the best technique in terms of overall accuracy and Kappa coefficient is selected for the final change map production. The approach achieves Land Use Land Cover (LULC) change detection using Support Vector Machine (SVM) and segmentation. The most useful change indices are obtained for the best methodology product. The simple improved methodology including radar images provides excellent results and it clearly outperforms the baseline optical technique.