Contribution of Sentinel-2 multispectral satellite images to study salinization effect of the Touggourt agricultural region (Algeria)

Abstract

In this paper, we propose a new spatio-morphological mapping of a region affected by the problem of salinization of its agricultural land. The rise of salt, which occurs naturally by infiltration of water passing through a saline layer through the faults, is accentuated by poor irrigation management, particularly by the use of Mio-Pliocene’s water with an excessive salt content and the inadequate drainage system. We use optical remote sensing data that allows us to study a spacious area with high accuracy (up to 10-m resolution) using Sentinel-2 images in combination with in situ data. In addition to the use of heterogeneous data, we also combine several methods (salinity and vegetation index, K-means clustering, and artificial neural networks classification) to achieve more accurate and reliable results by using software specially designed for this type of processing such as SNAP with its two versions (python dependent and stand-alone). We obtain an occupation map with six classes (dry salty soils, fine sand, medium sand, vegetation, agricultural land, and buildings) with a surface area of salt soils reaching 23% of the total surface area against only 14% of vegetation that is still likely to decrease. These methods allowed us not only to delimit the region affected by the excess salt with more or less intensity but also to highlight its impact on the quality of the vegetation.