Comparação de Imagens OLI/Landsat-8 e MSI/Sentinel-2 no mapeamento de cobertura e uso da terra no Maciço de Uruburetama, Ceará

Abstract

Orbital images of the Landsat series have been systematically employed in the mapping of land cover and use. However, some areas, due to the relief characteristics or the strong antropic influence, impose difficulties in this characterization. The Uruburetama massif, in the state of Ceará, represents an area with such peculiarities. This work compares images of the OLI / Landsat-8 and MSI / Sentinel-2 orbital sensors in order to determine which product can best be used studies. The methodology was based on obtaining orbital images of the area, including pre-processing stages, NDVI generation, segmentation by region growth, supervised classification, classification validation and production of thematic maps. The NDVI products had very strong positive correlation, evidencing spectral compatibility among the sensors. In the segmentation stage, we noticed the influence of the best spatial resolution of the MSI sensor with the creation of almost eight times more polygons and a mean area corresponding to 12.5% the measurement of the OLI sensor. The supervised classification using the Bhattacharya algorithm allowed to map the two products in seven thematic classes of coverage and land use of the Uruburetama massif: Mata Humid; Mata Seca; Dense Shrub Caatinga; Open Shrub Caatinga; Urban / Ground Exposure; Water Bodies and Crops. The validation of the classifications attested to the best accuracy of the MSI / Sentinel-2 product through the Kappa indices and global accuracy. The results demonstrate that the MSI / Sentinel-2 image, due to its better spatial resolution, allows a greater detail of the targets, and a better accuracy in the classification, which allows even its application in studies with larger scales of analysis. The OLI / Landsat-8 image, on the other hand, has been shown to be more suitable for studies with lower levels of detail, or with more homogeneous targets. Keywords: coverage and land use; NDVI; segmentation by regions; supervised classification.