The transferability of Australian pedotransfer functions for predicting water retention characteristics of French soils.

A French data set was used in evaluating how well two widely used analytical functions describe measured soil water characteristic (SWC) data. Both the van Genuchten (sigmoidal) and Campbell (power-law) equations gave good descriptions of the data (mean R2 of 98.1% and 97.1% respectively). Methods of predicting SWC data were also evaluated. When a power-law equation was parameterised using just two measured SWC points and bulk density (the 'two-point' method), a very good SWC prediction was obtained for the French data (mean R2 of 94.8%). An empirical equation for prediction of the SWC was also assessed using the French data set. This method was developed using multiple regression analysis from Australian soil data and requires soil texture and bulk density as input. The predictions (mean R2 of 85.2%) lacked accuracy and precision in comparison to the two-point method but uses more readily available input data. The accuracy of prediction from both methods was similar to that observed previously for Australian data sets. The empirical approach developed from Australian soil data has reasonable applicability to French soils. The approach of assuming a power-law model and empirically predicting slope and air entry potential is shown to have merit. A strategy for achieving adequate coverage of soil hydraulic property data for France is suggested incorporating hydraulic prediction methods such as those evaluated here