Efeitos da amostragem na análise da variabilidade espacial de variáveis georreferenciadas

Abstract

In precision agriculture, efforts have been done to characterize the spatial variability of soil attributes, aiming at establishing sampling procedures that guarantee representativeness of georeferenced samples. Thus, this trial aimed at evaluating stationary and isotropic data, with or without directional tendency (non-stationary or anisotropic processes, the influence of sample configuration in geostatistical model estimation and in non-sampled locations estimation. In order to obtain some data, the following sampling systems were simulated: the randomized one with 100 points, the systematic one in 10x10, 5x20 and 20x5 versions, and the lattice plus close pairs with nearby added points toward tendency and anisotropy, and toward orthogonal to the anisotropy. These results will serve as a scientific basis for a more efficient analysis of the spatial variability regarding chemical attributes of an agricultural area with isotropic, non-stationary and anisotropic variables. The results of simulated data have shown that lattice plus close pairs sampling (in all simulated versions) has presented the best results on quality of parameters estimation of model and spatial prediction. Considering the simulated results and the analysis of spatial variability regarding soil chemical attributes, it is proposed that in subsequent experiments in this agricultural area, the increase in number of nearby points and radius reduction of the nearby points should be considered. Also, for anisotropic variables and with directional tendency, the nearby points addition toward these phenomena and in orthogonal direction to these ones.