Publications Details

A comparison of parameter estimation and sensitivity analysis techniques and their impact on the uncertainty in ground water flow model predictions

Zimmerman, D.A.; Hanson, R.T.; Davis, P.A.

This work documents a comparison of sensitivity and uncertainty analysis techniques that are likely to be used in support of repository performance assessments to determine compliance with the Nuclear Regulatory Commission (NRC) and the Environmental Protection Agency (EPA) regulations for high-level radioactive waste (HLW) repositories. A variety of parameter estimation and sensitivity analysis techniques were applied to a model of the Avra Valley aquifer, Arizona. Two approaches to sensitivity analyses were used, statistical and deterministic; these were applied to evaluate the sensitivity of the ground water travel time to changes in transmissivity. The effect of different boundary conditions on the calculated sensitivity derivatives was also evaluated. Parameter estimates and estimation errors were obtained via geostatistical and inverse techniques. The throughput'' of the kriging techniques suggests that the mean estimates derived from these techniques are frequently off the mark'' or inconsistent with the conceptual model. With no screening of the input parameter estimates for realism, non- conservative travel time estimates were obtained. The differential analysis sensitivity technique is shown to be dependent on the choice of design point, providing only a local measure of the sensitivity. The statistical approach to sensitivity identifies parameters which are both sensitive and uncertain, i.e., it shows when the uncertainty in a model parameter is important. Sensitivity estimates are also shown to be dependent on the choice of boundary conditions used. 92 refs., 55 figs., 13 tabs.