Determining soil hydraulic properties by parameter estimation: On the selection of a model for the hydraulic properties

Publisher Website
Google Scholar
Abstract
Parameter estimation procedures involving solution of the inverse flow problem pertinent to a transient experiment constitute a powerful method of determining the soil hydraulic properties. One of the problems associated with these procedures is the selection of an appropriate model to describe the soil hydraulic properties. Three different models were considered: those of van Genuchten (VG) and Brooks and Corey (BC) model, and the exponential model of Gardner for the hydraulic‐conductivity‐soil‐water‐pressure relationships, coupled with a new expression for the soil‐water‐content‐soil‐water‐pressure relationship (GR model). For a given model the soil hydraulic properties were determined from simulated and measured outflow experiments supplemented with water content at a soil water pressure head of −15,000 cm H2O, using the parameter estimation procedure of Kool et al. (1985a) and data from two soils: a hypothetical sandy loam soil (assumed to be described by the VG model) and a silt loam soil. Model validation tests were performed, and the most appropriate model was selected from the candidate models by discrimination tests using the Akaike Information Criterion (AIC). In the case of the hypothetical sandy loam, the VG model was found to be the most accurate and most consistent with the data. The performance of both the BC and the GR models (in terms of the AIC) was less good but very similar to each other. In the case of the silt loam soil, the VG model with additional parameter m, which accounts for the correlation between pores and for the flow path tortuosity (m=2.02) rather than a constant m=0.5, was found to be most accurate and most consistent with the data. The performance of the BC and the GR models was similar when m was considered as an unknown parameter instead of a constant.