Stochastic Simulation of Non-Gaussian 3D Conductivity Fields in a Fractured Medium with Multiple Statistical Populations: Case Study

This paper applies a stochastic inverse method, named as gradual conditioning (GC) method, to the fractured site of Äspö, Sweden, which is an underground hard rock laboratory initially designed as a potential future deep geological repository for spent nuclear fuel. The aim of this paper is (1) the verification of GC method in a real three-dimensional (3D) fractured rock medium, showing that the GC method is a competitive stochastic tool in highly heterogeneous aquifers, furthermore, it gathers a set of capabilities so far not included in any existing method; (2) to characterize the site as adequately as possible, experimental data are reproduced closely; (3) to provide measures on the uncertainty of the estimates by means of using multiple equally likely realizations, thus showing the importance of conditioning to as much information as possible in order to reduce the uncertainty; (4) to prove that this fractured media can be adequately modeled by assuming a (pseudo-) continuum media, or equivalent porous medium, in which fractures are represented by the introduction of discretization blocks of very high effective hydraulic conductivity $\left(K\right)$ ; (5) to describe how to take advantage of the available geological information and how to properly integrate this secondary data in the GC method; (6) to efficiently model the statistical distribution of the hydraulic conductivity data by using the concept of multiple statistical populations (SPs); each fracture plane and the rock matrix represent an independent SP; and (7) to highlight that the GC method is able to deal with nonmulti-Gaussian distributions, thus allowing to reproduce existing preferential flow channels, and therefore, to obtain lower (safer) radionuclide travel times from the deep geological repository to the biosphere.