Avaliação de índices de dependência espacial de modelos geoestatísticos sobre a krigagem

Abstract

The semivariogram is used in geostatistics to predict the degree of spatial dependence, inferring about the relation of a spatialized variable. Nowadays there are methodologies that use with more efficiency the semivariogram parameters and the choice of method is important for selecting models for kriging. The aim of this study was to evaluate spatial dependence index to the selection of theoretical models to Kriging. The data were obtained from clonal stands of Eucalyptus sp. in three regions of Minas Gerais. Spherical and exponential models were fitted to volume, looking for obtain sets of parameters for the functions. To all adjustment were classified the structure of spatial continuity by using the methods GDE and IDE. Approximately 60% of the adjustments were classified as strong spatial dependence by the GDE method, while approaching 50% showed classification as strong by the method IDE. The GDE index ranked 117 adjustment as strong spatial dependece, being that by the new index (SDI) 40% would change the classification from strong to weak. When comparing the methods of least square adjustment and maximum likelihood, there were alterations in 28% of the analyzes. Although the kriging maps show high correlation, it was possible to observe the difference of area to the volumetric classes and consequently the mean volume. Using a robust database and more information from the semivariogram, the IDE showed to be more efficient to selection of models. Thus it is recommended to describe a spatial dependence because it includes all semivariogram parameters and correction factors for each model.