A Saturated Zone (SZ) Site scale Flow Model was developed for Yucca Mountain, Nevada, to incorporate new data and analyses including new stratigraphic and water–level data from Nye County wells, single and multiple well hydraulic testing data, and new hydrochemistry data. New analyses include the 2004 transient Death Valley Regional (ground water) Flow System (DVRFS) model, the new hydrogeologic framework model (HFM), and the 2003 unsaturated zone (UZ) flow model. This model includes the following: (1) reflect the current understanding of SZ flow, (2) enhance model validation and uncertainty analyses, (3) improve locations and definitions of fault zones, (4) enhance grid resolution (500 m grid spacing to 250 m grid spacing), and (5) incorporate new data not used in the previous revision. The flow model was completed using the three dimensional, finite element heat and mass transfer computer code, FEHM V2.24. The SZ site scale flow model was calibrated with the commercial parameter estimation code, PEST to achieve a minimum difference between observed water levels and predicted water levels, and also between volumetric/mass flow rates along specific boundary segments predicted by the SZ regional scale and site scale models. 161 water level and head measurements having different weights were used for calibration. A comparison between measured water level data and the potentiometric surface yielded an RMSE of 20.7 m (weighted RMSE of 8.8 m). The calibrated model was used to evaluate the impact of alternative models on flow paths and specific discharge predicted by the model.

Confidence in the results of the mathematical model was built by comparing: (1) calculated to observed hydraulic heads; and (2) calibrated to measured permeabilities and therefore specific discharge. In addition, it was confirmed that the flowpaths leaving the region of the repository are consistent with those inferred both from gradients of measured head and from independent water chemistry data. Uncertainties in the SZ site scale flow model were quantified because all uncertainty contributes to inaccuracy in system representation and response. Null space and solution space uncertainties were investigated and assessed.