Modeling groundwater in multimodal porous media with localized decompositions (Q1042528)

The authors introduce a new stochastic framework to approximate the stochastic flow equations of groundwater in multimodal porous media in the case when the variance \(\sigma^2_y\) of log hydraulic conductivity is not smaller than 1, and the homogeneous approach fails mainly due to the fact that the residual terms in the Reynolds decomposition are not statistically small. The authors also consider a porous medium with a multimodal conductivity structure that can be categorized into \(N\) different categories of geologic units. The authors believe that the Reynolds decomposition is inadequate and that new decompositions for log hydraulic conductivity \(Y(x)\) are needed to separate various factors causing largeness of overall variance. Therefore, the authors propose a localized Reynolds decomposition as \(Y(x)= I_M(x) + \overline Y(x)\), where the leading approximation is a space random function that contains information about media heterogeneity. Based on this decomposition method, the authors develop a stochastic model for flows in highly heterogeneous porous media. In particular, they present integral representations for hydraulic heads and fluxes. These closed integral expressions provide a better insight into how the overall behavior is influenced by individual units. These expressions could also potentially produce better predictions for flow and solute transport under the assumption that within each unit, either the variance of log conductivity is small or its volume fraction is small. The paper is of strong mathematical character with many new interesting results.