Application of geostatistical models to define the variability of a soil profile

Abstract

Geostatistical analysis is used to visualize and graph the spatial distribution of various soil properties. In this context, the objectives were: to describe several physicochemical properties of the soil of two profiles in a banana farm; then to estimate the best interpolation method, and through graphs to interpret the spatial variability. Two 2.50 m wide by 0.70 m deep pits were constructed diagonally to a banana plant at a distance of 40 cm. A 2 m long by 0.6 m wide mesh with 20 x 20 cm squares was placed in each profile, from which samples were taken for soil analysis. The geostatistical analysis was performed in ArcGis version 10.3. To show the distribution of the profile data, the kriging interpolation method was used. The empirical semivariogram models used were: linear, spherical, circular and Gaussian. Areas with different variability in their direction and data concentration as a function of distance were identified. The spherical interpolation model was more accurate. The maps maintained continuity of the values in depth, an aspect that allowed the characterization of the two soil profiles, when bought with the descriptive statistics.