Statistical Analysis of Salinity and Texture Effects on Spatial Variability of Soil Hydraulic Conductivity

Abstract

A statistical analysis that identifies the dominant factors responsible for the spatial variability of the saturated conductivity (K_s) in a selected field is presented. The variables are considered random functions of the space coordinates. Two types of stationarity are investigated: constant mean and stationary variogram (stationarity of order 2) and trend of the means (nonstationarity of the mathematical expectation) with stationary covariances. Linear models of statistical dependence were used. Elimination of factors of minor importance was based on correlation analysis. Electrical conductivity (EC) and percent sand were the most important independent variables, whereas K_s was the dependent variable. The analysis showed that soil salinity (EC) accounted for only 10 to 15% of the variability of K_s texture (percent sand) explained 25 to 45% of the K_s variability, and 10 to 15% of the variability was explained by the interaction between salinity and texture. Various soil factors other than salinity and texture accounted for 30 to 50% of the variability in K_s. Semivariograms and covariograms of K_s, EC, and percent sand did not have a sill because of a significant horizontal linear trend. Calculations showed that the trend accounted for about a half of the variability of the original data. Semivariograms and covariograms of a new population of residuals after extracting the trend gave values of the sills equal to the variances. The integral scale of the residuals of K_s and percent sand was estimated to be about 15 to 20 m, whereas salinity (EC) did not display a spatial structure. The results can be generalized for other studies with different values of variances and covariances of soil salinity and soil texture.