Conditional stochastic analysis of solute transport in heterogeneous, variably saturated soils

A method is developed for the conditional (Monte Carlo) simulation of steady stateflow and transient transport from point sources in variably saturated porous media. It combines the geostatistical method, a linearized approximation of the soil water tension perturbation solution, and afinite element numerical model. The method is used to investigate the usefulness of conditional simulation for predicting solute transport under a variety of sampling network designs applied to a number of hypothetical soils. Saturated hydraulic conductivity data yield the largest reduction of conditional uncertainty in relatively wet soils with mild heterogeneities. In highly heterogeneous soils or under dry conditions, soil water tension data by themselves, taken at a sampling density of one to two correlation scales along the expected mean travel path, can greatly reduce prediction uncertainty about solute concentration. Parameter uncertainty about statistical properties of the independent random variables becomes less important as the number of conditioning data increases. However, even with a very high number of sampling data, uncertainty of predicted concentration levels remains significant.