Abstract:
In this study we used the Kozeny-Carman (K-C) equation as a semi-physical model for estimating the soil permeability using data derived from microscope observations. Specific surface areas and porosities were obtained from two-point correlation functions derived from scanning electron microscope images of thin sections using a magnification of 50 and a resolution of 1.88μmpixel-1. Permeabilities were predicted using two published ('Ahuja' and 'Berryman') and one generalized variant of the K-C equation. The latter model was similar to the Berryman variant, but used a free parameter C rather than a porosity dependent formation factor. All K-C model variants were optimized on measured permeabilities. The Ahuja and Berryman K-C models performed relatively poorly with R2 values of 0.36 and 0.57, respectively, while the generalized model attained R2 values of 0.91. The parameter C was strongly related to texture and, to a lesser extent, particle density. The general model still required measured surface area and porosity. However, we showed that it was possible to estimate these parameters from texture resulting in an R2 of 0.87. A fully empirical model that did not assume K-C concepts performed slightly worse (R2=0.84). The results indicate that after developing the model using microscope information, only macroscopic data are necessary to predict permeability of soils in a semi-physical manner with the K-C equation.