Beware of scaling artefacts and implicit model characteristics when fitting soil water release and moisture capacity data

Abstract

The primary objectives of this study were to: ( i) elucidate the impacts of nonlinear scale transformations on the shapes and parameter values of soil water release and moisture capacity curves; and ( ii) demonstrate how implicit characteristics of some established soil water release and moisture capacity models can impact model-data fits and estimates of model parameters. Nonlinear scale transformations of the tension head ( h) axis (e.g., log10 h,  h1/2) were found to distort release and capacity curve shapes, create fictitious curve inflections and modes, and occasionally erase visual evidence of actual inflections and modes. The popular van Genuchten–Mualem and Assouline–Grant models were shown to always generate a release curve inflection and a capacity curve mode, even when inflections and modes did not exist in the data, and this in turn caused poor model-data fits in the critical near-saturated region. The van Genuchten model with four independently fitted parameters and the Dexter–Weibull model could accurately fit data sets with no inflection or mode, but this resulted in a physically unrealistic zero-angle intersection between the release curve and the water content axis. It was concluded that nonlinear h axis transforms should not be used when determining inflections, modes, pore size distributions, soil structure parameters, or soil quality indexes from soil water release and moisture capacity data-sets. It was also recommended that more flexible release curve models should be developed that do not assume the existence of inflections and modes, and also produce physically realistic angles of intersection between the water content axis and the fitted model.