Stochastic analysis of unsaturated flow: One‐dimensional Monte Carlo simulations and comparisons with spectral perturbation analysis and field observations

Abstract

A numerical experiment was designed to study the stochastic behavior of one‐dimensional transient unsaturated flow in a Monte Carlo setting. Soil hydraulic properties, log‐saturated hydraulic conductivity ln Ks, pore size distribution parameter α, and the specific water capacity C are assumed to be statistically homogeneous random fields described by exponential correlation functions with identical correlation lengths. Fifty realizations of each soil hydraulic property, with statistical properties obtained from a field experiment, are generated by using a nearest‐neighbor model. Numerical solutions of the one‐dimensional flow equation are used repeatedly to obtain realizations of soil water pressure head ψ and flux q corresponding to the realizations of In Ks, α, and C. The dependence of ψ and q on the statistical properties of ln Ks, α, and C and the magnitude of the hydraulic head gradient G prevailing at the lower boundary are investigated. In addition, results of Monte Carlo analysis and spectral perturbation analyses are compared with field observations. The greatest variability of ψ and q are observed when soil hydraulic properties are uncorrelated and have large variances and integral scales and when G at the lower boundary is unity. The Monte Carlo and spectral perturbation analyses tend to agree reasonably well for the flow domains in which ergodicity of local soil hydraulic properties is assured. Results of the Monte Carlo and spectral perturbation analyses are also supported by field observations.