Evaluate the level effectiveness of selected bare soil indices in Semi-Arid land using Sentinel-2 data

The Semi-Arid area is a specific climate region, where the accurate mapping of dry soil is a crucial issue and a real challenge. For that reason, this paper investigated this question over a semi-arid land that located in North-East of Algeria, by using Sentinel-2 data with a comparative analysis of selected Index-Based Methods, in addition to apply the Sup-port Vector Machine-SVM method. The finding showed that the Modified Normalized Difference Bare Index (MNDBaI) and the Modified Bare Soil Index (MBSI) which are combining visible bands at 10 m of resolution, worked better than the Bare Soil Index (BSI) and the Dry Bare-Soil Index (DBSI) that are combining VNIR-SWIR bands; the overall accuracy (OA) of the MNDBaI and MBSI are about 92.29 % and 90.86 %, respectively. Meanwhile, the DBSI provided the weakest (OA) that achieved 86%. Based on this, the indices MNDBaI and MBSI are successfully tested in semi-arid land, since they were more discriminative toward dry soil, especially the MNDBaI index that manifested an advanced behavior. Under climate changes and urban expansion impacts in the region, the MNDBaI index can contribute for accurate ecological studies and better spatial management, as it can produce an accurate bare soil information and land cover maps updates; the MNDBaI index is promising to be more adoptable in survey land degradation, indicating soil condition and boosting agriculture management. However, more tests over other areas, are suggested. Additionally, developing new soil indices that are more sensitive to dry soil, is highly needed.