Evaluation of Two Different Approaches for Predicting Soil Water Contents at Field Capacity and Wilting Point.

Soil hydraulic properties, such as the water content, are used in many hydrological models for predicting soil and environmental processes such as water erosion, soil erodibility, solute transport and evapotranspiration. Measuring soil water content in the field is typically hardworking and needs a large amount of soil for measuring properties in laboratory. Because of that, soil scientists have developed equations, also known as Pedotransfer Functions (PTFs), to predict the entire water retention curve (e.g. Van Genuchten) or to compute the water content at specific potentials. These potentials are typically at field capacity (FC) and wilting point (WP).

This work evaluated 12 PTFs with an independent soil dataset. The PTFs included Gupta and Larson (1979), Lal (1979), Rawls et al. (1982, 1983, 2004) and Aina and Periaswamy (1985). The results showed that PTFs tend to overestimate the water contents at FC, while underestimate the contents at WP. A second approach was the use of Rosetta model (v2.0-alpha) to compute the parameters of the van Genuchten model. Even though Rosetta has several methods for predicting the water contents, for a fair comparison in this study we used those that do not require a known value of soil water retention. These results showed that Rosetta underestimate both water content, FC and WP. This paper discusses the differences in these approaches in order to predict a precise and accurate value or method to obtain the soil water estimates, and the effect of that when predicting other soil hydraulic processes. The authors are developing a set neural networks for predicting in a directly manner the soil water content at field capacity and wilting point, and preliminary results show a precise prediction.