ABSTRACT
The purpose of this paper was to compare thematic maps constructed from data sets related to soybean productivity with different regular sampling grids of 25x25 m, 50 x50 m, 75 x 75 m and 100x100 m, using kriging techniques. In the theoretical models fitted to the experimental semivariances, it was used the maximum likelihood method for parameter estimation. The comparison of thematic maps were made by accuracy indices, obtained from the error matrix. The results showed that factors such as sample size and sample density between points interfere in the choice of the theoretical spatial model, the parameter estimates and the construction of thematic maps.
geostatistics; maximum likelihood; spatial dependence
