Stochastic characterization of geological heterogeneity and its impact on groundwater contaminant transport

The spatial variability of aquifer parameters is in itself deterministic. But, because of the erratic nature of the hydrogeological parameters observed in field data and the insufficient information about the spatial distribution of these parameters, it would be more realistic to describe these parameters in a probabilistic context (e.g. correlated random fields). Traditional random field models, reviewed in chapter 2, are based on the notions of stationarily and uni-modal Gaussian distribution of the hydrological parameters. These models are attractive from statistical point of view but most of them are less applicable because they are either poorly represent geological features or they need intensive hard data to get reliable estimates of their parameters. Therefore these models must be guided by classical geological knowledge. Some field studies manifested various types of heterogeneity: continuous variability [see e.g. Gelhar, 1986], discrete features and compound heterogeneity [see e.g. Killy and Moltyaner, 1988]. These observation have motivated the idea of characterizing spatial structures using geologically-based stochastic model rather than the use of traditional correlated random fields.