Selection of a semivariogram model in the study of spatial distribution of soil moisture

Abstract

AbstractThe paper presents a selection of a semivariogram model in the study of spatial variability of soil moisture in a loess agricultural catchment. Soil moisture tests were carried out in the Moszenki village, 15 km northwest of Lublin. Soil moisture measurements were performed at two dates at 104 points, located on a rectangular surface measuring 700 × 1200 m. These points were laid out in the corners of a grid of squares with sides 100 m. In addition, 6 measurements were made at a distance of less than 100 m from the nearest points. Soil moisture was measured in the soil surface (0–5 cm). ArcGis software with Geostatistical Analyst extension was used for modelling semivariograms. In both terms, five models of semivariograms were used: stable, circular, spherical, exponential and Gaussian. Kriging was used for the estimation of soil moisture values. Among the semivariogram models analyzed in this study, the largest errors in the determined values of soil moisture relative to the empirical data were observed for the exponential model, and the smallest for the Gaussian model. However, it should be emphasized that the values of the analysed errors for the individual semivariogram models were similar. Application of the ordinary kriging method for interpolation of spatial distribution of soil moisture yields good results, but it has to be kept in mind that the final shape of the spatial distribution is influenced by the choice of the semivariance function model.