Stochastic modeling of macrodispersion in heterogeneous porous media

A method for the prediction of dispersion processes of inert solutes occurring in heterogeneous porous media, is presented. Its main features are as follows: (1) The velocity field is assumed to be a space random function. Its moments are expressed through the physical parameters of the log transmissivity and head random fields by linearizing the flow equation. The random velocity, being a linear function of the head and log transmissivity, which are assumed to be jointly multivariate normal (MVN), is thus also MVN. Its pdf is completely defined by its first two moments. (2) The statistics of the dispersion process are obtained by the particle‐tracking method through Monte Carlo simulations which are based on Gaussian conditioning. The method is applied to the Borden natural gradient tracer test (Freyberg, 1986), showing favorable results.