Large-scale pore network and continuum simulations of solute longitudinal dispersivity of a saturated sand column

Predicting the solute dispersivity in porous media is complicated by the known scale dependency of dispersion processes. In this study, we combined three complementary methods to investigate solute dispersivity at different length scales. The applied methods included 36-cm long column experiments, extremely large three-dimensional (3D) X-ray image-based pore network simulations, and Darcy scale modelling. We used X-ray imaging to extract the information on pore structures needed to construct a very long, 36-cm, pore network model. Doing so, for the first time, a direct comparison of the pore-scale model to the large experimental observations was performed. The longitudinal dispersivity was found to increase with length, with the rate of increase being lower at higher flow velocities. Results have shown that a pore network, which matches pore connectivity distribution and pore size distribution of a sub-sample of the column, is able to reproduce the experimentally observed solute breakthrough curves. Pore-scale modelling provided the extent of pore velocity variations corresponding to various locations along the experimental column as well as the one dimensional (1D) Darcy scale model. We have found that even in homogenous porous media immobile pores exist, with low velocities compared to the average velocity, which creates local concentration maximals.