Conceptual uncertainties in modelling the interaction between engineered and natural barriers of nuclear waste repositories in crystalline rocks

Abstract

AbstractNuclear waste disposal in geological formations relies on a multi-barrier concept that includes engineered components – which, in many cases, include a bentonite buffer surrounding waste packages – and the host rock. Contrasts in materials, together with gradients across the interface between the engineered and natural barriers, lead to complex interactions between these two subsystems. Numerical modelling, combined with monitoring and testing data, can be used to improve our overall understanding of rock–bentonite interactions and to predict the performance of this coupled system. Although established methods exist to examine the prediction uncertainties due to uncertainties in the input parameters, the impact of conceptual model decisions on the quantitative and qualitative modelling results is more difficult to assess. A Swedish Nuclear Fuel and Waste Management Company Task Force project facilitated such an assessment. In this project, 11 teams used different conceptualizations and modelling tools to analyse the Bentonite Rock Interaction Experiment (BRIE) conducted at the Äspö Hard Rock Laboratory in Sweden. The exercise showed that prior system understanding along with the features implemented in the available simulators affect the processes included in the conceptual model. For some of these features, sufficient characterization data are available to obtain defensible results and interpretations, whereas others are less supported. The exercise also helped to identify the conceptual uncertainties that led to different assessments of the relative importance of the engineered and natural barrier subsystems. The range of predicted bentonite wetting times encompassed by the ensemble results were considerably larger than the ranges derived from individual models. This is a consequence of conceptual uncertainties, demonstrating the relevance of using a multi-model approach involving alternative conceptualizations.