Document Type : Research Paper
In the present study, five log-normal models, normal model (N), and two four- parameter models namely Gompertz (G) and Fredlund (Fr) were tested on 71 PSD data from 10 soil textural classes. The Jaky (J) and Fr models had the smallest and largest R2 values for all of the soils, respectively. The results of comparing the N and simple log-normal (SL) models showed that the accuracy of the N model increased with decrease soil clay content. In other words, the mass-size distribution of soil samples tends to a normal statistical pattern from the finer to the coarser textural classes. The result of F, Cp and AIC statistics showed that the Fr was the best model to describe PSD for majority of soils and the modified log-normal models namely ORL and ONL was the next in rank, respectively. The range of R2 values for the sandy and sandy loam soil texture classes indicated that the G, ONL, ORL, N and Fr models had a better fit on the data than the SL and Shiozawa-Campbell (SC) models. However, the performance of the SL and SC models improved compare to G and N models in clay and silty clay soil texture classes.