Natural systems prediction of radionuclide migration

This paper reviews the application (and limitations) of data from natural systems to the verification of performance assessments, particularly as they apply to the evaluation of the long-term performance of waste forms, backfill, canister materials, and finally, the integrity of the repository itself. Two specific examples, the corrosion of borosilicate glass and the formation of alteration products of spent fuel, will be discussed. In both cases, inferences are of three types: 1) directly applicable data (i.e. radiation effects, stable phase assemblages): 2) inferences based on the analogous behaviour of the natural and repository systems (e.g. long-term corrosion rate); 3) specific identification of new phenomena that could not have been anticipated from the short term laboratory data (i.e. new mechanisms for the retention or release of radionuclides). The latter can only be derived from the observation of natural systems. Finally, specific attention will be paid to the limitations in the use of natural systems, particularly as the spatial and temporal scales expand, and to the inherent limitations of prediction and verification. (J.P.N.)