Validation of transport models for use in repository performance assessments: A view illustrated for INTRAVAL test case 1b

Abstract We present our views on validation. We consider that validation is slightly different for general models and specific models. For a general model validation consists of establishing the case for the model such that the model is widely accepted within the scientific community. However, for a specific model, which is unlikely to be reviewed by many people, validation consists of establishing a case such that one might reasonably expect that someone with relevant technical knowledge would consider that the model was acceptable. Thus we stress the importance of presenting for review the case for (or against) a model. We outline a formal framework for validation, which helps to ensure that all the issues are addressed. Our framework includes calibration, testing predictions, comparison with alternative models (which we consider particularly important), analysis of discrepancies, presentation, consideration of implications and suggesting improved experiments. We illustrate the approach by application to an INTRAVAL test case based on laboratory experiments. Three models were considered: a simple model that included the effects of advection, dispersion and equilibrium sorption, a model that also included the effects of rock-matrix diffusion, and a model with kinetic sorption. We show that the model with rock-matrix diffusion is the only one to provide a good description of the data. We stress the implications of extrapolating to larger length and time scales for repository performance assessments.