An evaluation of methodologies for the generation of stochastic hydraulic conductivity fields in highly heterogeneous aquifers

 Stochastic techniques, such as Monte Carlo experiments, are more and more frequently used for the study of flow and transport in heterogeneous aquifers. When the aquifer is composed of distinct hydrofacies, a common way to model heterogeneity is to first generate equally-possible hydrofacies fields, and then convert these hydrofacies fields into hydraulic conductivity (K) fields by assigning a single K value to each facies. This technique assumes relative homogeneity of K within each facies but may not be appropriate for the most conductive facies that often exhibits substantial variability. In this paper, we assessed the impacts of assigning multiple random K, rather than a uniform K value, to the highly conductive facies on the results of a flow and transport model. A set of fifty stochastic hydrofacies maps depicting an environment similar to the Snake River Plain aquifer (SRPA) in south-east Idaho were generated. Simulations demonstrated that a uniform K value, if carefully chosen, can reasonably reproduce the specific discharges and early particle arrival times produced by multiple K values. Yet, the results obtained with a uniform K value are dramatically less variable than those obtained with multiple K values. It is therefore concluded that stochastic simulations with uniform K assigned to the most conductive and variable facies do not necessarily portray the entire uncertainty in the analyses.