Técnicas de diagnósticos em modelos espaciais lineares gaussianos

Abstract

Tracking and management concepts of the process of agricultural production are being used as a great option of strategy management in agriculture. Such concepts consider the spatial variability of the variables at study. The modeling of the spatial dependence structure of the geoestatistic approach is fundamental importance for the definition parameters that define this structure and are used in the interpolation of values in places not sampled, by kriging techniques. However, the estimation of parameters can be greatly affected by the presence atypical observations in the data sampled. The development of this work was aimed at using diagnostics techniques in spatial linear gaussians models, used in geoestatistics, to evaluate the sensitivity of the maximum likelihood estimators and restrict maximum likelihood to small perturbations in the data. Studies were performed with simulated data, with literature data and with experimental data, collected in a commercial agricultural area in the region West of Paraná. The study with simulated data showed that the techniques used in diagnostics were efficient in identifying the perturbation data. The restrict maximum likelihood estimator produced more robust estimates for the parameters spatial dependence. Those results obtained from the study of real data, it was concluded that the presence atypical values between the sampled data can exert strong influence on thematic maps, changing, therefore, the spatial dependence. The application the diagnostic techniques should be part of any geoestatistic analysis, ensuring that the information contained in thematic maps have better quality and can be used with greater security by the farmer.