Dissimilarity of parameterized maps generated in geostatistics and the assumption of an automatic software model in precision agriculture

Abstract

Geostatistics is the main technique used to efficiently determine spatial variability. The objective of this study was to evaluate the applicability of the principles of geostatistics in the use of semivariograms elaborated through parametric monitoring and the assumption automatically made by software in the map preparation of soil chemical attributes. Available phosphorus (P), potassium (K+), calcium (Ca2+), magnesium (Mg2+), base saturation (V%), sulfur (SO42-), and pH were compared from the soil chemical attributes of 60 samples of a Typical Oxisol collected at a 0-20 cm depth and a distance of 300 m between the points. The maps were compared using error matrices and evaluated by the Global Accuracy (GA), Kappa (K), and Tau (T) indexes. The parameterized semivariograms and the automatic software model assumption did not present a high coincidence for the available P and Mg2+, making it necessary to adjust the semivariogram variables in the spatial analysis as a function of the outliers, sum of squares of residuals, coefficient of determination, and cross-validation to better represent the variability of the data and thus avoid distortions of the sample point range that would affect the adequate representativeness of the attributes, which contrasts with the automatic model generated by the software.