Premium
Fitting Performance of Particle‐size Distribution Models on Data Derived by Conventional and Laser Diffraction Techniques
Author(s) -
Bah Abdul R.,
Kravchuk Olena,
Kirchhof Gunnar
Publication year - 2009
Publication title -
soil science society of america journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.836
H-Index - 168
eISSN - 1435-0661
pISSN - 0361-5995
DOI - 10.2136/sssaj2007.0433
Subject(s) - sieve (category theory) , akaike information criterion , log normal distribution , range (aeronautics) , mathematics , exponential function , particle size , monte carlo method , particle (ecology) , particle size distribution , fractal , statistics , biological system , statistical physics , materials science , physics , mathematical analysis , geology , paleontology , combinatorics , composite material , biology , oceanography
Mathematical description of most classical particle size distribution (PSD) data is often used for estimating soil hydraulic properties. Fast laser diffraction (LD) techniques now provide more detailed PSDs, but deriving a function to characterize the entire range of sizes is a major challenge. The aim of this study was to compare the fitting performance of seven PSD functions with one to four parameters on sieve‐pipette and LD data sets of fine‐textured soils. The fits were evaluated by the adjusted R 2 , MSE, and Akaike's information criterion. The fractal and exponential functions performed poorly while the performance of the Gompertz model increased with clay content for the LD data sets. The Fredlund function provided very good fits with sieve‐pipette PSDs but not the corresponding LD data sets, probably due to underestimation of the clay fraction in the latter. The two‐parameter lognormal function showed better overall performance and provided very good fits with both sieve‐pipette and LD data sets.