1997

Rasilainen, Kari

Matrix diffusion is an important retarding and dispersing mechanism for substances carried by groundwater in fractured bedrock. Natural analogues provide, unlike laboratory or field experiments, a possibility to test the model of matrix diffusion in situ over long periods of time. This thesis documents quantitative model tests against in situ observations, done to support modelling of matrix diffusion in performance assessments of nuclear waste repositories, which must cover long, even geological time periods. A model testing methodology was developed by combining measurements of long-lived natural decay chains yielding the experimental reference, rock matrix characterisations, and matrix diffusion simulations. Valuable support to matrix diffusion simulations was provided by independent uranium-series disequilibria simulations, done to date radiometrically the accumulated uranium. Repeated model tests, applying the methodology, were carried out at two Finnish natural analogue study sites, the U-Th deposit at Palmottu, and a boulder sample near Hämeenlinna; both sites show elevated natural radioactivity. The deposit at Palmottu is extremely old, while the uranium accumulation in the boulder appears postglacial. The most important lesson learned at Palmottu is that the experimental reference must be conceptually in line with the model being tested. In this view, reconciling the respective concepts of attachment of radionuclides on rock is absolutely essential, otherwise quantitative model testing is pointless. In situ simulations require a complete and consistent site-specific data base for the model, because in model testing even partly generic input data may lead to false conclusions. The boulder sample indicates that recent uranium accumulations are easier to interpret than old ones, because matrix diffusion response is easier to trace, and boundary conditions for the simulations are easier to quantify. The iterative model validation process is a powerful method to refine both theoretical and experimental tools. The tests with the boulder sample support the classical matrix diffusion concept used in performance assessments. The tests at Palmottu indicate, however, a clear need to improve the technique used to obtain the experimental reference.