A novel and simple method for accurate prediction of soil particle-size distribution from limited soil texture data

Abstract

The full range of soil particle-size distribution (PSD) is not often readily accessible in most soil databases and conventional studies. In the present work, an Exponential Model (EM) was developed to predict PSD in the particle diameter range of 0.002–0.05 mm. Furthermore, its combination with two cubic polynomial-based models (Very Coarse Sand-Dependent, VCS-D, and Very Coarse Sand-Independent, VCS-I) that accurately predict PSD in the particle diameter range of 0.05–2 mm was tested. The VCS-D and VCS-I models (for 0.05–2 mm diameter range) were also combined with the Revised Skaggs (R-Skaggs) model, as a logistic function (for 0.002–0.05 mm diameter range) to predict PSD. These combinations thus allow us to predict the full range of PSD from 0.002 to 2 mm, which was tested using 665 measured soil PSD data from Iran (245), UNSODA (143), Brazil (116), and Belgium (161). The EM, VCS-I, and R-Skaggs methods use clay, silt, and sand fractions to predict PSD, while the VCS-D utilizes the mentioned fractions along with the fraction of very coarse sand particles. It was shown that the PSD predicted between 0.05 and 2 mm by the VCS-D and VCS-I methods and between 0.002 and 0.05 mm by the EM method, well-matched with the measured ones in all studied textural classes. The PSD curves predicted by the R-Skaggs method were closely matched with the measured values only within specific ranges of particle diameter and soil textures. Combining the VCS-D or VCS-I methods with EM showed superior performances in predicting the full range of PSD in a wide range of textural components compared to the other investigated models. The mentioned combination of models predicted the full range of PSD with excellent (root mean square error, RMSE < 5 %) performance in most soils containing > ∼33 % clay content and excellent to good (RMSE < 10 %) accuracies in most soils with ∼20–33 % clay, < ∼65 % silt, and > ∼25 % sand contents. The obtained mean values of RMSE from the combination of VCS-D or VCS-I methods with EM to predict PSD were significantly (p < 0.05) lesser than those obtained from VCS-D or VCS-I methods alone and also their combination with R-Skaggs in the mentioned ranges of textural components. In soils with > ∼65 % silt content, irregular patterns of the predicted PSD curves were observed. Generally, we recommend using the combination of VCS-I and EM methods to predict the full range of PSD due to their high simplicity, applicability, and accuracy.