Uma abordagem bayesiana para estudo estatístico e geoestatístico de estimativas de salindade do solo utilizando sensor de indução eletromagnética

• Autor(a) principal: PESSOA, Antônio Lopes
• Data de Publicação: 2006
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UFRPE
• Texto Completo: http://www.tede2.ufrpe.br:8080/tede2/handle/tede2/4487
• topic: Medidas de condutividade elétrica; Variabilidade espacial; Geoestatística; Salinidade do solo; Krigagem bayesiana; Electrical conductivity measurement; Spatial variability; Geostatistics; Soil salinity; Bayesian kriging; CIENCIAS EXATAS E DA TERRA::PROBABILIDADE E ESTATISTICA

This study analyzed the existing relationship among measurements of soil apparent electrical conductivity in an alluvial valley in the Agreste region of Pernambuco State and its spatial variability in the subsurface. The soil apparent electric conductivity was investigated through an electromagnetic induction EM 38 equipment. The readings have been carried out both in the vertical and horizontal modes. The measurements have been analyzed through the classic descriptive statistics as well as geostatistics and bayesian approach. The statistical analyses had shown that the data of apparent electric conductivity had adjusted to a normal distribution, presenting a high space variability for the horizontal mode and an average space variability for the way of vertical operation. In order to allow the use of the geostatistical methodology, the experimental semivariogram was constructed, and fitted to a theoretical model. Considering the spatial dependence mapping of the salinized areas have been performed. The best theoretical models for the vertical mode and for the horizontal operation were the gaussian model and the exponential model, through the crossed validation and using the Akaike’s Information Criterion .The bayesian approach focused the spatial predictionrelating the method of the maximum likelihood with the functions of prioris distributions for each parameter, considering the uncertainty associated to each one of these distributions. It was verified that the adjusted semivariograms had not presented significant differences in the validation of the geostatistics methodology and in the bayesian approach.