AbstractA geological disposal facility (GDF) is the widely accepted long-term solution for the management of higher-activity radioactive waste. It consists of an engineered facility constructed in a suitable host rock. The facility is designed to inhibit the release of radioactivity by using a system consisting of engineered and natural barriers. The engineered barriers include the wasteform, used to immobilize the waste, the waste disposal container and any buffer material used to protect the container. The natural barrier includes the rocks in which the facility is constructed. The careful design of this multi-barrier system enables the harmful effects of the radioactivity on humans and biota in the surface environment to be reduced to safe levels.Bentonite is an important buffer material used as a component of a multi-barrier disposal system. For example, compacted bentonite rings and blocks are used to protect the copper container, used for the disposal of spent fuel, in the KBS-3 disposal system. As the bentonite saturates, through contact with groundwater from the host rock, it swells and provides a low hydraulic conductivity barrier, enabling the container to be protected from deleterious processes, such as corrosion. The characteristic swelling behaviour of bentonite is due to the presence of significant quantities of sodium montmorillonite.Recently, there have been detailed in situ experiments designed to understand how bentonite performs under natural conditions. One such experiment is the Buffer–Rock Interaction Experiment (BRIE), performed at the Äspö Hard Rock Laboratory near Oskarshamn in the SE of Sweden. This experiment is designed to further understand the wetting of bentonite from the groundwater flow in a fractured granite host rock.In this paper, the observations from the BRIE are explained using an integrated model that is able to describe the saturation of bentonite emplaced in a heterogeneous fractured rock. It provides a framework to understand the key processes in both the rock and bentonite. The predictive capability of these models was investigated within the context of uncertainties in the data and the consequence for predictions of the wetting of emplaced bentonite. For example, to predict the wetting of emplaced bentonite requires an understanding of the distribution of fracture transmissivity intersecting the bentonite. A consequence of these findings is that the characterization of the fractured rock local to the bentonite is critical to understanding the subsequent wetting profiles. In particular, prediction of the time taken to achieve full saturation of bentonite using a simplified equivalent homogeneous description of the fractured host rock will tend to be too short.
Natural Gradient Flow in the Mixture Geometry of a Discrete Exponential Family
