Thermosolutal fluid flow has a strong influence on the evolution of solidification microstructures. While porous media theory and volume-averaged permeability relations give a basis to quantify these phenomena, traditional methods of permeability estimation used for random porous media fail to adequately characterize the full relation of microstructural morphology to volume-average permeability. Most significantly, the link between microstructural parameters and permeability is treated as a deterministic function at all scales, ignoring the variability inherent in porous media.The variation in permeability inherent to random porous media is investigated by the numerical solution of Stokes equations on an ensemble of porous media, which represent of many scales of sampling and morphological character. Based on volume-averaging and statistical treatment, the stochastic character of tensoral permeability in porous media is numerically investigated. Quantification of permeability variation and autocorrelation structure are presented as conditions, which future realistic stochastic permeability fields must respect.