Proposta metodológica para mapeamento da cobertura da terra em grande escala no semiárido brasileiro: um estudo de caso em Serra Negra do Norte-RN

Abstract

Planning and management of the territory goes through legislative paths supported by the Master Plan and application of land cover mapping methodologies to municipal planning, where climate seasonality directly interferes with physical and socioeconomic aspects in Brazilian Semiarid. In this context, realizing the lack of information on land cover and the difficulty of municipalities to have a cartographic basis to aid municipal planning, it is suggested a low-cost alternative to obtain information on land surface. Thus, the present study aims to propose a large-scale land classification methodology for municipalities in Brazilian Semiarid Region that can be generated with low-cost Remote Sensing products, using the municipality of Serra Negra do Norte-RN as an experiment. Thus, it were used hybrid images from the fusion of Sentinel 1 satellites with radar images and Sentinel 2 with multispectral images in 843RGB (red, green, blue) color composition in rainy and dry semiarid period. Four hybrid images were obtained, two with VH polarization and the others with VV polarization, respectively in rainy and dry periods. The dates of Sentinel 1 and 2 images were chosen according to the precipitation occurred in 2016. For Digital image processing used free software SNAP 6.0, QGIS 218.16 and Terraview 5.3.1 with the GeoDMA (Geographic Data Mining Analyst) plugin for object-oriented classification. Thus, the results obtained from Kappa for the land cover mapping with the four hybrid images classified in the VH channel polarization (Vertical-horizontal) that had better accuracy was for the rainy season with the value of 0.34 and 0.35 for the dry period.While Kappa values for VV (verticalvertical) channel were 0.25 for rainy month and 0.31 for dry season, indicating the mapping as reasonable. To the detriment of this, it is pointed out that the classification originated from hybrid images in VH channel had better accuracy, although the hybrid images with VV polarization presented better the land cover classes in the map. Thus, the land cover map in dry period is indicated due to the accuracy pointed out, but it must be taken into account that from rainy season in the Brazilian Semiarid there is the green (Caatinga) vegetation more evident, and the increase of temporary and perennial springs. Therefore, this thesis comes with an alternative in the proposed land cover mapping.