Publications Details

Approaches for the Validation of Models Used for Performance Assessment of High-Level Nuclear Waste Repositories

Olague, N.E.

The purpose of this report is to provide general approaches and concepts that can be applied in validation of models used in performance assessment of high-level waste (HLW) repositories. The approaches are based on a validation strategy that Sandia National Laboratories has implemented as participants in the International Transport Validation Study (INTRAVAL). This strategy focuses on the demonstration that performance assessment models are adequate representations of the real systems they are intended to represent, given the pertinent regulatory requirements rather than proving absolute correctness from the purely scientific point of view. Positions that are taken consist of the following: due to the relevant time and space scales, models that are used to assess the performance of a HLW repository can never be validated; therefore, validation is a process that consists of building confidence in these models and not providing "validated" models; in this context, model validation includes comparisons to "reality," however, adequacy for the given purpose is the overall goal; comparisons to "reality" consist of comparing model predictions against laboratory and field experiments, natural analogues, and site-specific information; when comparing experimental data to model predictions, a model can be either "invalid" or "not invalid," based on the null hypothesis concept, however, confidence in the model arises in finding a model to be "not invalid" over a wide range of conditions; an attempt should be made to consider in the validation process all plausible conceptual models; and when comparing experimental data to model predictions, a logical systematic approach should be followed. This report discusses the definition of validation in the context of performance assessment for HLW repositories, the need for validation, an approach to validation, and an approach to comparing model predictions with experimental data proposed by the authors.