Efficient uncertainty propagation of lognormal hydraulic conductivity in a three dimensional hydrogeological model of flow and transport on very large regular grids

Abstract

This work illustrates an efficient method for Monte Carlo uncertainty analysis in a three dimensional hydrogeological context involving very large grids. Although Monte Carlo simulation using Simple Random (SR) sampling is the de facto method for such an uncertainty analysis, it quickly becomes expensive in terms of time and computer resources, particularly for the case of complex models. To alleviate this problem, a new Latin Hypercube (LH) sampling method is proposed in this work for efficiently generating realizations of secondorder stationary lognormal random fields on large (order of million pixels) regular grids. he performance of the proposed LH sampling method is compared to that of SR sampling in the context of a 3D hydrogeological flow and transport model. The results show that the proposed LH sampling is more efficient than SR sampling, in that it can overall reproduce to a similar extent statistics of the conductivity (and associated concentration) field, yet with smaller sampling variability than the latter.