Reducing the uncertainty of hydrogeological parameters by co-conditional simulations: lessons from practical applications in aquifers and in low permeability layers

Stochastic simulation of aquifer heterogeneity is now often performed to provide a confidence interval of the modelled results for flow and solute transport problems. In practice, due to the few available measurements of the hydraulic conductivity (hard data), it is useful to integrate several other properties of the medium as indirect data (soft data). The additional conditioning obtained from the use of these secondary data allows reduction of the variance of the distribution and consequently decrease of the uncertainty of the results. This practice can also be extended to low permeability clay layers. For example, stochastic sequential simulation can be performed involving hydraulic conductivity values as hard data, and grain size measurements, electrical resistivity log, gamma ray log and a description of the lithology variation as soft data. However, other important properties can also be considered. The possible fracturing of clay strongly influences the flow and solute transport. On the other hand, in very low permeability media, diffusion can be considered as the dominant transport mechanism, so that heterogeneity in terms of the effective diffusion coefficient becomes important. Examples of application are summarized considering aquifers and low permeability clay layers. It clearly shows the great advantage of collecting multiple data sets of inter-correlated data on the same geological medium to be modelled. In high conductivity aquifers as well as in low permeability layers, this kind of additional conditioning obtained from various data is always useful when considering applications such as, among many others, well capture zones delineation, impact studies and geological confinement of wastes.