A steady state, three-dimensional finite element model is calibrated for the groundwater system underlying the Frenchman Flat basin at the Nevada Test Site. The model is designed to predict radionuclide migration from ten underground nuclear tests over a period of 1,000 years. The model implements five detailed hydrostratigraphic framework models; several recharge models, and site-specific and regional data to constrain the model parameterization. A novel approach was used to construct the finite element mesh that allows alternative interpretations of the hydrostratigraphy and fault structures in the flow model. The calibration followed a multiple-step approach that combined optimized parameter estimation and manual adjustment of parameters guided by local sensitivity analyses results. Further sensitivity and perturbation analyses of calibrated models are used to establish the (local) optimality of calibrated parameters. As with other models covering large areas, considerable uncertainty exists in the calibrated model parameters due to model parameter correlations and incomplete information about regional groundwater fluxes and boundary heads. Two approaches were used to analyze the uncertainties associated with conceptual model development and parameter calibration. Additional work is currently being done to compare modeled flow paths with major ion and stable-isotope data in an effort to validate model results. Currently, the results indicate that radionuclide migration from the underground test areas can be bounded by the model in spite of these uncertainties and that, within a 1,000-year time frame, the likelihood that radionuclides will reach the underlying regional aquifer is extremely small.