Classificação e mapeamento da textura superficial do solo a partir de dados de sensoriamento remoto e análise discriminante, na região de Volta Grande do Rio Uruguai – Brasil

Abstract

Measuring and evaluating soil components and properties are often costlyand time-consuming procedures.Prediction and modelling, whose proceduresare known as predictive soil mapping,are useful for solving the lack offield data sampling. The objective of this work is to map the surface-based soil texturein the region of Volta Grande do Rio Uruguay (SC / RS-Brazil) by using soil particle-size sampling (44 points)or soil granulometry, soil spectral reflectance (141 points) in the MSI sensor (Sentinel-2), predictive statistical modelling (Discriminant Analysis) and IDW interpolation. The methodology aims to find discriminant functions to obtain a statistical classifier that, based on soil spectral data, is able to discriminate the soil texture (surface-based) in terms of sand, silt and clay. The results showed that (i) the statistical model classified the soil texture at an accuracy of 71% according to the Kappa Index; (ii) silt and sand were similarboth in terms of spectral signature and of content(g/kg); (iii) there was a predominance of clay corroborating the lithological characteristic of the basaltic area. Discriminant Analysis provides a basis to identify which independent variables (reflectance) contribute most to the response of the dependent variable (soil texture). Thus, thestudied methodcan be useful to indicatethe spectral ranges inregression models forpredicting soil texture at locations not sampled.