Before recent drilling and characterization efforts in the vicinity of Los Alamos National Laboratory (LANL), conceptual models had been developed for recharge and discharge in the regional aquifer on the basis of sparse data. By integrating site-wide data into a numerical model of the aquifer beneath the plateau we provide new insight into large-scale aquifer properties and fluxes. This model is useful for understanding hydrologic mechanisms, assessing the magnitudes of different terms in the overall water budget, and, through sampling, for interpreting contaminant migration velocities in the overlying vadose zone. Modeling results suggest that the majority of water produced in well fields on the plateau, extracted at rates approaching 70% of total annual recharge, is derived from storage. This result is insensitive to assumptions about the percentage of total recharge that occurs in the near vicinity of water supply wells, because of strong anisotropy in the aquifer that prevents fast transport of local recharge to deeper units from which production occurs. Robust estimates of fluxes in the shallow portion of the aquifer immediately down gradient of LANL are important for contaminant transport simulations. Our model calculations show that these fluxes have decreased in the past 50 years by approximately 10% because of production in water supply wells. To explore the role of parameter uncertainty in flux prediction, a predictive analysis method was applied. Results showed that predicted flux through older basalts in the aquifer can vary by a factor of three because of uncertainty in aquifer properties and total recharge. We explored the impact of model parameter uncertainty on these results; however, the true uncertainty of our predictions, including the impact of possible conceptual model errors, is likely to be larger and is difficult to quantify.