FORECASTING SCALE-DEPENDENT DISPERSION FROM SPILLS IN HETEROGENEOUS AQUIFERS

Existing theories of flow and contaminant transport in aquifers are either based on Monte Carlo simulations or small perturbation solutions of the governing stochastic partial differential equations, which limit the applications to cases of small variances in the physical parameters, In this article scale-dependent models to predict mean contaminant concentration from point sources (well injection), and non-point sources (ground surface spills) in heterogeneous aquifers are developed and tested. A general analytic technique, the Neumann expansion, is used in the solution of the equations. This method does not require the assumptions of small perturbation, logarithmic transformations, or a specific probability law in the random quantities. In addition, aquifer parameters such as the mean pore velocity are functionally defined in terms of the underlying groundwater flow problem, and field-measurable aquifer hydrogeologic properties such as mean transmissivity, mean hydraulic gradient, mean porosity and aquifer thickness. Comparison with theoretical models, and verification with field tracer tests at the Borden aquifer indicated that the proposed models reproduced the enhanced longitudinal dispersion as a function of distance reported in the literature.