Stochastic modeling of spatial heterogeneities conditioned to hydraulic and tracer tests

Uncertainty concerning the physical and chemical nature of subsurface heterogeneities constitutes a severe technical barrier to assessing long term performance of nuclear waste repositories. This paper discusses an approach to generation of stochastic permeability fields through simultaneous inversion of flow and transport data. For tracer transport calculations, we have used a semianalytic transit time algorithm which is fast, accurate and free from numerical dispersion. The inversion of data has been accomplished through the use of simulated annealing. We have addressed the non-uniqueness associated with our results by shifting the focus from the search for a single model that fits the data best to inferences about the properties that are shared by an ensemble of acceptable models. We then determine a most likely model for heterogeneity. The approach has been illustrated through application to tracer migration in a synthetic fracture plane.