Land cover mapping using remote sensing data in the Apure River Flood Plain (Venezuela)

Abstract

The soil cover is a fundamental indicator to identify the factors that act in the development of the geomorphology of an alluvial plain. This coverage is characterized by the control exercised by the vegetation in the hydromorphological processes, as well as the maintenance and stability of the fluvial channels. A record on the distribution of land cover in the middle course of the anastomosed system of the Apure River is presented. The distribution of geomorphological environments in an area of 65 km2 is analyzed from a combination of data from Landsat-8 and Sentinel-2 images, integrated into a Geographic Information System (GIS). A supervised classification was established using the Support Vector Machine and Maximum Likelihood algorithms. The Landsat image was processed through an atmospheric correction, to later calculate the spectral signatures. Six covers were found: a) wooded savannah, b) forest, c) open savannah, d) crops, e) bodies of water, and f) scrub. There are no substantial differences in the reliability achieved with the Support Vector Machines and Maximum Likelihood classification algorithms. It was shown that the woodland cover is the most representative in the study area with a total extension of 5,717.26 ha (39%), out of 14,658.77 ha. The classification presented a global thematic accuracy of 98.08% and a Kappa index of 0.98. As a result, a soil cover cartography was generated from the best classifier, based on the Kappa index. These findings serve as a reference to increase the records of soil cover characterization and can be useful in studies on management and use of the territory, to identify places more susceptible to degradation and propose measures for the management and conservation of water resources, which can be potentially applicable in similar fluvial environments in other latitudes.