Derivation and validation of pedotransfer functions for estimating soil water retention curve using a variety of soil data.

Soil textural components are the most common inputs in pedotransfer functions (PTFs), although other soil properties, such as bulk density, organic matter content and structural indexes, are also used. The possibility of using the geometric mean d g and the geometric standard deviation σ g of soil particle diameters instead of soil particle size distribution to derive PTFs was investigated. Soil samples (234) having different particle size distributions were collected randomly. The predictor variables were separated into two groups: (i) particle-size distribution and bulk density, (ii) bulk density, geometric mean and geometric standard deviation of particle diameters. Two point PTF types and two parametric PTF forms were developed to predict six points on the retention curve and the parameters of the van Genuchten model, using the stepwise regression method. The derived PTFs were compared with the Rosetta package. The results indicated that the second set of variables accounted for 65 and 90% of the variation in α and 0 s parameters of the van Genuchten, model respectively. The results also indicated that these variables better explained the variation of the n parameter (the shape parameter of the van Genuchten model) and the water contents at -30, -100, -300, -500 and -1500 kPa matric potentials. It was concluded that, at the drier end of the retention curve, d g and σ g predict the water contents better than the soil bulk density. Although the coefficients of determination of the derived PTFs with the second set of variables were higher than that of the first group, the first set of variables gave more reliable estimates. Results showed that the descriptors of the particle size distribution (d g and σ g ) did not improve the quality of soil water retention prediction. The reliability test indicated that the derived PTFs predict soil hydraulic properties better than the Rosetta package.