Markov Chain Monte Carlo Simulation to Determine Soil Hydraulic Properties from Air Permeability Measurements.

Researchers are continuously searching for improved and quicker methods to determine soil hydraulic properties. The objective of this study was to determine soil hydraulic properties from air permeability values determined as a function of water content. An equivalent capillary model consisting of parallel sequences, with each sequence containing sections with different diameters, was assumed to represent the porous medium. The radii of the individual sections were derived through Bayesian statistics using Markov Chain Monte Carlo simulation with the recently developed DiffeRential Evolution Adaptive Metropolis (DREAM) sampling scheme (Vrugt et al., 2008). The DREAM algorithm estimates the most likely pore size distribution, and its underlying (nonlinear) uncertainty distribution within a single optimization run. Consequently, this results in an ensemble of pore size structures that are statistically consistent with the observed experimental data. The sampled equivalent capillary pore size distributions were subsequently used to predict water retention curves and effective water permeability relations. These relations were shown to be in good agreement with experimentally determined data.