Model Validation and Decision Making: An Example Using the Twin Lakes Tracer Test

Abstract

ABSTRACTAn approach to the validation of ground-water flow and transport models that are used in support of licensing a radioactive waste site is proposed and demonstrated. The approach attempts to interpret modeling results in light of their intended purpose. Instead of trying to find the ‘best’ fit model to the experimental data, we attempt to determine the conservativeness or bias invoked by applying different models to simulate the same experiment. In this way we are able to provide guidance to the regulatory community on where they should expect model results to fall relative to actual data and provide some evidence that certain model approaches lead to conservative results while other modeling approaches lead to non-conservative results. Applying this approach to the Twin Lakes Tracer Test demonstrated that a simple one-dimensional flow and one-dimensional dispersion model consistently over-predicts the maximum concentration. A one-dimensional flow with three-dimensional dispersion and two dimensional flow with two dimensional advection-dispersion are conservative only if they employ laboratory scale dispersivities. Therefore, if a decision-maker were to accept a similar site based on results from these types of models, there would be a relatively low chance that he has accepted an unsafe site. However, if decision-makers were to reject a similar site based on these types of results, there would be a relatively large chance that they rejected a safe site.