Review of permeability in buried-valley aquifers: Centimeter to kilometer scales

Aquifer systems are considered in which measurements of logpermeability (Y) at a 1E-2 m support scale are unimodally distributed as taken within the 1E0 to 1E1 m scale of individual sand and gravel (sg) lithofacies, are weakly multimodal at the 1E1 to 1E2 m scale of assemblages of these facies, and are strongly multimodal at the 1E3 m scale of complexes of sg facies assemblages juxtaposed with mud and diamicton (md) facies assemblages. Contaminant plumes resulting from decades-old disposal have grown to the 1E3 m scale with a high probability of sampling both facies assemblages in the complex. The relevant aspects of the complicated residence time distribution (varying orders of magnitude between the time of travel in preferential pathways through sg facies assemblages and the longer time of travel if in part through md facies assemblages) of mass at this scale is explained by the heterogeneity in the pattern of facies assemblages in the complex; in-facies permeability structure is not important. However, recent spills and tracer tests create smaller plumes at the 1E1 to 1E2 m scale. Here the spatial structure of in-facies permeability and the spatial structure of facies within the sg assemblage is relevant. A hierarchical random space function model gives the global spatial structure of Y as a linear function of the mean, variance, the two point auto and cross-covariance of Yi for each i facies, as weighted by the auto and cross-transition probabilities representing the proportions, the modality, mean and variance in facies lengths, and the juxtapositioning pattern of the facies. This is illustrated with data from an outcrop analog study.