Identifying covariates to assess the spatial variability of saturated soil hydraulic conductivity using robust cokriging at the watershed scale

Abstract

The mapping of saturated soil hydraulic conductivity (KSat) is essential to understanding soil water dynamics and is a sensitive input in hydrological modeling. The objectives of this study were to provide a reference for the selection of soil hydrology and other environmental attributes that can be used as covariates for estimating KSat and to compare the efficiency of univariate ordinary kriging versus ordinary robust cokriging, using selected soil hydrology and environmental attributes. Data sets were obtained from a sample grid of 179 points established in the Ellert creek watershed (ECW), located in Rio Grande do Sul state, Southern Brazil. KSat, macroporosity, microporosity, total porosity, and bulk density were determined from soil sampled at each point. Data of land use and elevation were also applied. All data sets were firstly submitted to classical statistics. Boxplot graphics were constructed to evaluate the relationship between KSat and land uses. Spearman coefficient of correlation between KSat and the other attributes was also assessed. For the assortment of covariates, cluster analysis was applied. Classical and robust estimators were applied to calculate the auto and cross-semivariograms and hereafter the ordinary kriging and cokriging. The Spearman coefficient showed some inconsistencies among the applied variables, suggesting that the multivariate method was more appropriate. All cross-semivariograms, except for land use, showed results with better accuracy than the auto-semivariograms. From the methods applied, the best estimates of KSat were obtained using the robust cokriging method, using macroporosity and soil bulk density as covariates.