AN ALGORITHM FOR GENERATING SPATIALLY AUTOCORRELATED UNSATURATED FLOW PROPERTIES

A methodology is presented for using particle size distribution data to generate distributed fields of soil hydraulic properties which are described locally by van Genuchten's (VG) model. The VG model is parameterized by the saturated water content, saturated conductivity, "residual" water content, and two shape parameters for the water retention function. Local particle size distribution (PSD) is characterized by a log-normal cumulative distribution function defined in terms of a local log-mean and log-variance which are each considered to be independent stationary random functions with known mean, variance, and autocovariance functions. Saturated water content also is regarded as a stationary random function with stipulated mean, variance and autocovariance function. Spatially distributed fields of PSD parameters and saturated water content are generated by a stochastic moving average algorithm. Via a quasiphysical pore structure model, local PSD data are used to predict the local soil-water retention function which is fitted to the van Genuchten model. Saturated hydraulic conductivity is estimated from VG model parameters and saturated water content. Distributions of soil hydraulic properties generated using this indirect approach are similar to field observed distributions. © 1990.