Abstract:
Spatial variation in water flow and solute transport are important considerations when assessing field-scale re mediation options, and for managing water and chemicals during agricultural production. Here, we evaluated the relationship between laboratory-measured solute transport variables pore-water velocity (v), dispersivity (lambda), and retardation coefficient (R) and land use, soil type, and other independently measured soil properties in adjacent Typic Haplusteps, Mollic Ustifluvents, and Lithic Ustipsamments. We also characterized the spatial variation in v, R, coefficient of hydrodynamic dispersion (D), and lambda. We obtained 100 geo-referenced, undisturbed soil columns (8.4 cm diameter and of varying length) from the topsoil at sites with varying land uses, topographies, and soil types, and synchronized disturbed soil samples to analyze the basic soil properties. We conducted miscible displacement tests on all 100 columns, and predicted the retardation coefficient (R) and D using the computer program CXTFIT, and measured v. All four variables, except R, had strongly right-skewed and kurtotic distributions, and were highly variable. Log lambda (logarithm to base 10) was significantly correlated with land use, with the mean of lambda being significantly greater in grasslands, followed by orchards and croplands. By contrast, v was significantly correlated with soil type, with the mean being significantly greater in Lithic Ustipsamments. Log v (logarithm to base 10) was also significantly correlated with the soil CaCO3 content, log organic matter content, pH, cation exchange capacity, and log electrical conductivity, whereas neither v nor lambda was correlated with soil texture measures. An analysis of the spatial structure of solute transport variables using semivariograms showed that R and lambda had considerably high nugget effect. Log-transformed semivariograms performed well using ordinary point kriging for v and D. However, lambda and R exhibited a very weak spatial dependency that could not be interpolated using ordinary kriging at the current sampling resolution. Therefore, this should be considered when designing future sampling programs to analyze spatial structure-of R and lambda. (C) 2016 Published by Elsevier B.V.