NUMERICAL SIMULATION OF GROUNDWATER CONTAMINANT TRANSPORT ON A SUPERCOMPUTER WITH INJECTION-PUMPING NETWORKS USING THE MODIFIED MOC AND MFE METHOD

To prevent the deterioration of groundwater quality, mathematical simulation models have been formulated to predict the transport of contaminants in complex aquifer systems and to design remedial schemes for the problems. Existing analytical and numerical approaches have serious disadvantages for large-scale nonhomogeneous field problems where well-pumping or injection is involved. The major difficulties relate to numerical dispersion and oscillations in highly advective-dominated simulations, computational accuracy, excessive computer expense, grid orientation problems, and an inability for simulating with random conductivity fields. Recent work by Ewing, Russell, and Wheeler (1983) has produced a very efficient and accurate method for miscible displacement in oil reservoirs. Their concept was adapted and then applied to groundwater contaminant transport problems in this thesis. The highly efficient code combines a mixed finite element procedure for groundwater flow and a modified method of characteristics and finite element procedure (MMOC) for the parabolic transport equation. The preconditioned conjugate gradient method was used to solve the resulting matrices for both equations. The method has been compared with two analytical solutions on a homogeneous domain. Excellent agreements were demonstrated through relative concentration contours and breakthrough curves. The method has also been compared with the currently popular USGS Solute Transport model. More accurate resolutions were achieved for the MMOC method than for the USGS Solute Transport model. In addition, much larger time steps were allowed in the MMOC method than the USGS Solute Transport model obtaining similar resolutions. The method has been applied to highly advective-dominated problems on a CRAY-XMP supercomputer and the results showed there are no dispersion or oscillation problems common in many existing numerical codes. The method has also been used to simulate cases with random hydraulic conductivity fields that were simulated from Turning Bands Method. Fingering phenomena developed because the concentration front is transported more rapidly in the zones of higher hydraulic conductivity. The method has been shown to be superior in many respects to currently used models in groundwater transport, especially in the presence of strong pumping or injection centers or heterogeneities in the flow field