Evaluation of different representations of the particle-size distribution to predict soil water retention

Detailed particle-size distribution (PSD) data or parameters derived from it are used in almost all pedotransfer functions (PTFs) to estimate soil hydraulic properties. Most commonly, the PSD is represented by the sand, silt and clay content of a soil. However, a number of standards exist and are in use internationally to describe PSD and particle-size classes. Additionally, some authors use derived particle-size information, such as the geometric mean diameter and its standard deviation as input to their predictive equations. We compared the usefulness of some of the most common representations of the PSD in estimating water retention at − 33 and − 1500 kPa, using Group Method of Data Handling and data of three databases. The cutoff limit between silt and sand content has been simulated to be at 20, 50 and at 63 μm, as used in various countries. We also used the geometric mean diameter and its standard deviation as predictors. We examined the effect of using the wrong cutoff limit in the PSD curve while applying a PTF. Overall, we could not point out one particular representation of the PSD that would clearly provide better results in estimating the selected soil hydraulic properties. Using geometric mean diameter and its standard deviation to represent the entire PSD curve with two parameters did not result in significantly different estimates of the examined soil hydraulic properties, than models that use distinct particle-size classes as input. We found no evidence suggesting that using interpolated data would reduce the accuracy of the estimation of these soil hydraulic properties. It is not straightforward to pre-determine if using the wrong definition of the silt/sand boundary in the PSD data will result in significant loss of estimation accuracy or not; therefore, using such data should be avoided. Our study suggests that using interpolated data–which of course carries a certain magnitude of interpolation error–poses fewer risks while using a PTF than using measured data with the wrong silt/sand boundary.