Classificação e mapeamento preditivos do solo na região de Volta Grande do Rio Uruguai – SC/RS

• Ano de defesa: 2020
• Autor(a) principal: Sandra Cristina Deodoro
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; Brasil; IGC - INSTITUTO DE GEOCIENCIAS; Programa de Pós-Graduação em Análise e Modelagem de Sistemas Ambientais; UFMG
• Palavras-chave em Português: Solo; Granulometria; Modelagem estatística; Assinatura espectral; Modelagem de dados – Aspectos ambientais; Solos – Uruguai, Rio, Bacia; Materiais granulados; Modelagem gráfica (Estatística)
• Link de acesso: http://hdl.handle.net/1843/34379; https://orcid.org/0000-0002-6141-7577

Soil is a natural resource that can be analysed from its anthropic and (geo)ecosystem functions. The knowledge of soil texture (proportion of particle size) is important under these contexts, mainly of the upper layers (surface-based), which are the first to be eroded. In addition, the texture is an important characteristic due to its relationship with other soil properties such as structure, porosity, permeability, fertility, chemistry and moisture content. There is a growing need for spatially continuous and quantitative soil information for environmental modelling and management, at different cartographic scales. The lack of data sampling is overcome, generally, by prediction and modelling results whose procedures, known as predictive soil mapping, are specially developed to estimate the spatial distribution of soil variables. Digital soil mapping is a useful approach for spatial prediction of soil attributes. Such an approach involves a relationship between the soil and the environmental variables, based on statistical and geostatistical models, to create a predictive map or derive soil property values in unsampled locations. The present study aims to map the granulometry of the soil (topsoil) in the Uruguay River basin (between the states of Santa Catarina and Rio Grande do Sul, Brazil), in the stretch known as Volta Grande. The methods used in this research were the spectral data collected from the MSI sensor of the Sentinel-2 satellite, fieldworks sampling (soil particle-size analysis), predictive statistical modelling (Discriminant Analysis) and IDW interpolation. The results showed an accuracy of 71% in the soil texture classification, according to the Kappa index, with predominance of clay. Based on the morphometric data and in the MRVBF index – Multiresolution Index of Valley Bottom Flatness – derived from the SRTM GL1 DEM (12,5m), most of the area was represented by clayey colluviums, which is coherent with field observations and with extensive sloping segments of slopes widely distributed in that area. It can be concluded that the occurrence of colluvium-alluvium on the banks of the Uruguay River, in the lowland areas, indicates the contribution of the slopes in the pedogeomorphological dynamics of the study area and not only of river dynamics. Based on the results, therefore, the methodology applied in this research demonstrated that remote sensing products and techniques, together with multivariate statistics and statistical modelling, have potential utility as both knowledge and auxiliary techniques for obtaining, analysing and mapping soil texture. This thesis is original and involves interdisciplinary concepts of remote sensing and pedogeomorphology, integrated into statistical modelling and geographic information system. As a practical implication, it presents the soil as an important natural resource for environmental analysis. It also shows the capabilities and limitations of the Sentinel-2 satellite for predicting and modelling soil attributes, since the MSI sensor is relatively recent (2015) when compare with others such as the Landsat.