Macropore Flow in Soil Columns: Investigations with Computer Tomography and Lattice Boltzmann Simulations

Soil structure greatly affects the ability of soil to transmit and to retain water, chemicals, and colloidal particles that can carry contaminants or be contaminants themselves, e.g. pathogenic microorganisms. No theory or empirical relationships have been developed to date to quantitatively relate parameters of soil structure and parameters of the contaminant transport in soils. The absence of theoretical advances in this area seriously hampers the ability to address issues of public concern, e.g. spread of contaminants introduced in the environment by agricultural activities. Recently, computer tomography of soils has become available to generate detailed images of soil pore space with high resolution and density. Successful applications of computer tomography in medical and material sciences show the great potential of this technique to create an exhaustive characterization of soil structure heterogeneity. In this presentation we investigate saturated flow through twelve undisturbed macroporous soil columns (7.62- cm sample diameter and 18-cm length) with lattice Boltzmann simulations. Saturated flow was measured for the complete columns, as well as on 2 cm sections for selected columns. Computed X-Ray tomography was performed on each of the columns, using the 420 kV X-ray source of a HYTEC FlashCT high-speed industrial CT scanner. The resolution was 116 microns per voxel, yielding a final tomography image of 656x656x1482 (~ 6.3 10E8) voxels. X-Ray CT observations typically provide "gray-scale" representations of the imaged object that must be segmented to yield discrete pore and particle geometry. Many segmentation algorithms are available, each yielding different final pore geometries thus potentially creating uncertainties in subsequent flow analyses. Lattice Boltzmann (LB) simulations will be presented only for some of the columns as the simulations are extremely computationally intensive (each simulation requires ~ 60 GB of computer RAM at the observed resolution). The main objectives of this presentation are: 1) to define an optimal resolution for the LB simulations in terms of simulated saturated hydraulic conductivity and computational requirements, 2) study the effect of different types of segmentation algorithms on simulated flow, and 3) compare simulated to observed hydraulic conductivities. Successful LB simulations may lead to improved insights into flow patterns within macroporous structures and lead to better understanding of solute and colloid transport.