Discriminação de classes de solo, quantificação de atributos e determinação do comportamento espectral de minerais de ferro e argila esmectítica utilizando espectrorradiometria difusa

• Ano de defesa: 2012
• Autor(a) principal: Cézar, Everson
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual de Maringá; Brasil; Departamento de Agronomia; Programa de Pós-Graduação em Agronomia; UEM; Maringá, PR; Centro de Ciências Agrárias
• Palavras-chave em Português: Solos; Sensoriamento remoto; Análise multivariada; Raios-X; Paraná; Brasil.; Soil; remote sensing; multivariate analysis; X-ray; Brazil; Ciências Agrárias; Agronomia
• Link de acesso: http://repositorio.uem.br:8080/jspui/handle/1/1127

The rational use of soil has been constantly debated. The search for less expensive, more agile techniques, which can effectively help in soil studies and its use, have highlighted the use of remote sensing. This study aimed at evaluating the possibility of employing diffuse spectroradiometry as an alternative tool to traditional survey methods and soil analysis. In order to achieve this goal, the research was divided into two stages. In the first one, the hypothesis to be tested was whether the soils formed in the transition region sandstone/basalt could be discriminated by their reflected energy, and if their physico-chemical attributes could be quantified. A number of 200 soil samples collected in Lobato (a town in Parana State, Brazil) - PR, were used. They were collected at depths from 0 to 0.20 m and from 0.80 to 1.0 m. These samples were subdivided and sent to the soil laboratory so that their physical, chemical and mineralogical attributes could be determined. Part of the same samples was subjected to spectral reading in FieldSpec 3 JR equipment in the range of 350-2500 nm. After reading and tabulating the data, they were evaluated by multivariate analysis. Discriminant analysis, simulated discriminant analysis and multiple linear regression analyses were carried out. Both discriminant analysis and simulated discriminant analysis showed satisfactory results, with overall average hit rates for classification close to 75% and 62% respectively. The multiple linear regression analysis also obtained promising results for horizons, being clay, sand, sum of bases and iron oxides the attributes with larger capacity of estimation. In the second stage, two hypotheses were tested. The first one was if it would be possible to establish estimation models for silt, sand and clay fractions more robust than those presented in current studies. The second one was that iron oxides and bentonite would present reflectance factor with linear behavior. The soil samples used in the fractionation and in the generation of models for estimating attributes were collected at Cooperval (a local mill), at its Irrigation Training Center and at Maringa State University Experimental Farm, located in Iguatemi (Parana State, Brazil) at depths from 0 - 0.20 m and from 0.80 to 1.0 m. The iron oxides were synthesized at the laboratory, while the bentonite was obtained from geological repository. The physical and chemical analysis and spectral readings were performed under the same conditions described in stage 1. Regarding the sand and clay fractionated in the soil samples, the results showed that these elements had the reflectance factor with linear behavior, whereas the silt fraction showed a curvilinear behavior. The regression models for Cooperval and the Irrigation Training Center, at both depths, were able to estimate the fraction of sand in natural soil samples satisfactorily, without significant differences between the values observed and the values expected. However, the models adjusted for silt and clay fractions have not performed the expected results. Similarly, models adjusted for the Experimental Farm in Iguatemi have not been able to adequately estimate the attributes mentioned above, when used in natural soil samples. The results for the oxides of iron showed that the hematite, goethite and magnetite, subjected to spectral reading did not present linear behavior for the reflectance factor; there was a sharp fall in their concentrations levels, corresponding to 20%, 40% and 10% respectively, with relative stabilization from these proportions. On the other hand, the sample of bentonite showed linear behavior, not only in the range studied, but across the entire reading equipment spectrum. Given our results, we conclude that the diffuse spectroradiometry has promising future in soil studies; however, for it to be used effectively, more studies, which are incipient in Brazil, are still necessary.