Flow of groundwater with variable density and viscosity was simulated at the Atikokan Research Area (ARA) in northwestern Ontario, Canada. An empirical viscosity-concentration equation was modified to include total-dissolved-solids (TDS) data from the ARA. The resulting equation was used successfully to estimate reasonably accurate viscosity values over the expected range of temperature and concentration, in comparison with experimental values derived for sodium chloride solutions. A three-dimensional finite-element code, MOTIF, developed by Atomic Energy of Canada Limited, was used in the simulations. The inclusion of the effects of depth-increasing temperature and TDS-dependent fluid-density distribution, while maintaining only a temperature-dependent viscosity relationship in a simulation, resulted in a more penetrative flow against expected buoyancy effects (i.e., the physics of the system was not honored). Accounting for concentration in the viscosity equation caused water to be less penetrative and more in accordance with the expected physics of the system. A conclusion is that fluid concentration should be considered simultaneously in calculating the density and viscosity of a fluid during modeling of variable-density flow in areas underlain by fluids with high TDS. Results of simulations suggest that both flow directions and magnitudes should be employed simultaneously during the calibration of a model. Large-scale groundwater movement in the ARA may be analyzed with carefully selected vertical no-flow boundaries. By incorporating the geothermal temperature gradient, groundwater recharge increases by 12%; thus, this gradient plays a significant role in groundwater flow at the ARA. Variability in the fluid concentration at the ARA neither decreases nor increases recharge into the groundwater system. The hypothesis that an isolated continuous regional flow system may exist at depth in the ARA is not supported by these simulations.