Calibração e validação de métodos baseados em espectroscopia de refletância no visível e infravermelho próximo para levantamento da variabilidade espacial de solos mato-grossenses

• Ano de defesa: 2015
• Autor(a) principal: Marcilio, Norka da Silva Albernaz
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Mato Grosso; Brasil; Faculdade de Agronomia, Medicina Veterinária e Zootecnia (FAMEVZ); UFMT CUC - Cuiabá; Programa de Pós-Graduação em Agricultura Tropical
• Palavras-chave em Português: Sensor; Refletância do solo; Sensoriamento proximal; Atributos do solo; CNPQ::CIENCIAS AGRARIAS; Reflectance of the soil; Proximal sensing; Soil properties
• Link de acesso: http://ri.ufmt.br/handle/1/2438

Currently, VisNIR spectroscopy has been seen as a fast and reliable method to express the spatial variability of soil, with high spatial coverage on maps and is useful to soil management in specific site and precision agriculture. These features surely contribute to fill the knowledge gap caused by the weaknesses of the conventional methods in solving the spatial patterns of soil fertility, mainly in fieldscale. In this study, the spectral data were correlated with the results of conventional analytical laboratory, considering chemical (P and SOM - soil organic matter) and physical (texture and color) attributes of soil in agricultural areas of the state of Mato Grosso in Brazil. We worked with VisNIR spectroradiometer in the field and in the laboratory, and sampled both simple and composite soil. Variations in clay content made it possible the separation in four textures: CT1 - clay content <150 g kg-1 soil; CT2 - clay content between 150 and 350 g kg-1 soil; CT3 - clay content between 350 and 600 g kg-1 soil; CT4 - clay content > 600 g kg-1 soil. The models were developed using partial least squares regression (PLSR). For the models of texture bands, the best results were obtained with the composite samples of laboratory spectra, with better performance for medium and loam texture bands. Contrarily, the predictions with the field spectra did not suffer interference from the type of sampling in place. The clay, MOS and chroma were estimated with a reasonable level of accuracy for all types of collection, considering the R2, RPD and RMSE values, while the predictions for phosphorus were assessed as poor with RPD <1.4. In the case study, excellent performances were recorded for MOS (RPD = 2.04) and chroma (RPD = 3.0) in laboratory spectra of composite samples. Researchers have shown that the spectral measurements taken with the Veris VisNIR spectroradiometer have great potential for use in precision agriculture in the Brazilian soils conditions and can be used to determine several soil properties that are crucial in defining management zones.