Improved regression of the water retention curve for highly aggregated soils

The van Genuchten (VG) model is widely used to represent soil water retention curves. Durner's dual-VG model, which combines VG functions linearly, is suitable for expressing retention curves of highly aggregated soils like Andisols, which are commonly found in Japan's arable land. However, we encountered difficulties in determining appropriate model parameters for the dual-VG model using the regression method. To overcome this difficulty, we improved the algorithm for nonlinear regression using the dual-VG model. This was accomplished by separating the water retention curve into subcurves that represent regions of high and low water content. Each subcurve was individually fitted with VG models, after which the entire curve was fitted with the dual-VG curve. The new algorithm was implemented in SWRC Fit, which is the web-based application to determine model parameters automatically and its accuracy was validated using measured data from agricultural soils throughout Japan. The results demonstrated that the proposed method generated retention curves with a goodness of fit comparable to the global solution, while maintaining significantly low computational costs. The developed method is expected to contribute to improve the results of numerical simulation with well-aggregated soil in arable land by accurately estimating water holding capacity and hydraulic conductivity.