Use of class pedotransfer functions to predict water retention properties of soils: Are they still of some interest?

The use of pedotransfer functions (PTFs) for the estimation of soil hydraulic properties has become increasingly popular in recent years. Moreover, the estimation of hydraulic properties by PTFs can be an alternative to troublesome and expensive measurements. New approaches to develop PTFs are continuously being introduced within the last years. However, PTFs application or use in locations other than those of data collection has been rarely reported. Most PTFs are continuous-pedotransfer functions (continuous-PTFs) which are continuous functions between the basic soil properties which are used as estimates and the predicted hydraulic properties. Other PTFs called class-pedotransfer functions (class-PTFs) are functions that predict hydraulic properties from the soil class (very often classes of texture). Many studies during the last two decades have concerned that second type of PTF because most provide directly a mathematical model for the entire water retention curve. Despite their possible inaccuracies, class-PTFs are easy to use because most require little soil information and are well adapted to prediction of water retention properties of soils over large areas. The objective of this study was to compare the quality of the prediction of the water retention properties when using continuous-PTFs or class-PTFs. We derived continuous-PTFs and class-PTFs for 320 horizons of the data base SOLHYDRO 1.0 that gathers together the water retention properties of soil samples collected in France. The continuous-PTFs were regression equations with the clay and silt content, organic carbon content and bulk density. They were developed for all types of horizons together and after separating topsoil and subsoil horizons. The class-PTFs were developed for all horizons together using the texture classes associated to the European soil data base (textural class-PTFs), or both the texture and bulk density (texturo-structural class-PTFs). These continuous- and class-PTFs and those established in the literature by using data of the European data base HYPRESS were applied to 108 horizons belonging to French soils. Comparison of the results was based on the mean error of prediction (MEP) and standard deviation (SDP) of prediction that enable the bias and accuracy of the prediction to be discussed, respectively. Results showed similar bias between the continuous- and class-PTFs developed with SOLHYDRO 1.0 but the bias was already very small whatever the PTFs tested (< 0.01 cm3 cm-3). The bias was much greater with the continuous- and class-PTFs established with HYPRESS. Results showed also similar accuracy etween the continuous- and class-PTFs developed with SOLHYDRO 1.0. The values of SDP ranged from 0.040 to 0.051 cm3 cm-3. The values of SDP were slightly greater with the continuous- and class-PTFs developed with HYPRESS. They ranged from 0.043 to 0.065 cm3 cm-3. Thus, our results show that there is still some future for class-PTFs because they produce similar bias and accuracy than continuous-PTFs when applied to soils close (i.e. originated from similar pedological context) to those used to established the PTFs. They appeared particularly well adapted to the prediction of water retention properties at the scale of large areas such as a country territory, scale at which the so called easily accessible basic soil properties are often class of texture, organic carbon content and bulk density rather than well defined numerical values