Statistical evaluation and choice of soil water retention models

Abstract

[1] This paper presents the results of statistical investigations for the evaluation of soil water retention models (SWRMs). We employed three different methods developed for model selection in the field of nonlinear regression, namely, simulation studies, analysis of nonlinearity measures, and resampling strategies such as cross validation and bootstrap methods. Using these methods together with small data sets, we evaluated the performance of three exemplarily chosen types of SWRMs with respect to their parameter properties and the reliability of model predictions. The resulting rankings of models show that the favorable models are characterized by few parameters with an almost linear estimation behavior and close to symmetric distributions. To further demonstrate the potential of the statistical methods in the field of model selection, a modification of the four-parameter van Genuchten model is proposed which shows significantly improved and robust statistical properties.

Using Pedotransfer Functions to Estimate the van Genuchten–Mualem Soil Hydraulic Properties: A Review

“…[1] with five independent parameters (including m ) provides the most flexibility in terms of describing measured data (van Genuchten and Nielsen, 1985), it is not the most efficient model for deriving PTFs (e.g., Cornelis et al, 2005). Reducing the number of parameters from five to four, for example by assuming m = 1 − 1/ n or m = 1, generally leads to improved PTF predictions at the expense of less flexibility in describing the observed data (Vereecken et al, 1990; Cornelis et al, 2005; Lennartz et al, 2008). The difference between the two most often used four‐parameter parameterizations is illustrated in Figure 1a shows that setting m = 1 creates a symmetrical MRC shape, while S is always equal to 0.5 when h = α −1 (i.e., α h = 1).…”