Studies on the spatial variability of soil attributes influencing crop productivity are important for the development of new technologies beneficial to agriculture. Geostatistical techniques can be used to evaluate the variability of soil attributes and contribute to soil mapping and management. The purpose of this paper was to evaluate the quality of the theoretical spatial model adjustments according to the Akaike Information and Filiben Criteria, Cross Validation and the maximum value of the log-likelihood function, of the soil humidity, of the soil density data and soil resistance to penetration, in the layers 0-0.1; 0.1-0.2; and 0.2- 0.3 m and the soybean yield in the 2004-2005 growing season. The parameters of the spatial variability models were estimated by the methods of least ordinary squares, least weighted squares and maximum likelihood. The experiment was developed in an area of 57 ha with a regionally typical distrofic Red Latosol (Oxisol). A spatially georeferenced 75 x 75 m regular mesh was used. Based on the results of the evaluation of adjustments it was concluded that the Cross Validation criterion was the most adequate to choose the best adjustment of the spatial variability model, resulting in more precise thematic maps.
geostatistics; estimation methods; adjustment validation
