Proof of future performance of a complex system such as a high-level nuclear waste package over a period of hundreds to thousands of years cannot be had in the ordinary sense of the word. The general method of probabilistic reliability analysis could provide an acceptable framework to identify, organize, and convey the information necessary to satisfy the criterion of reasonable assurance of waste package performance according to the regulatory requirements set forth in 10 CFR 60. General principles which may be used to evaluate the qualitative and quantitative reliability of a waste package design are indicated and illustrated with a sample calculation of a repository concept in basalt.

Analytic models have been developed and used to study the performance of engineered barriers in a basalt repository with waste packages vertically emplaced below the storage room floors. Models have been developed for the groundwater flow field and for waste transport through the waste package backfill and up to the top of the storage room. The lowest release rates at the top of the storage room occur for a room backfill with high hydraulic conductivity (greater than the surrounding rock), high porosity and good sorption. Such a room backfill is much more effective than waste package backfill at delaying releases from the engineered system, especially for short-lived nuclides. A repository design with horizontal emplacement of waste packages has also been studied and found to have higher release rates, owing to a larger cross-section to groundwater flow and the absence of an overlying storage room.

Ground-water flow and transport were modeled along a one-dimensional, vertical cross-section of the Carrizo aquifer from its outcrop down gradient for 100 km. Using a constant flux of 14C at the recharge boundary, the 14C mass fraction distribution downgradient was simulated until steady-state was attained. The match between the observed data and simulated results is excellent for a porosity of 35% and a KD of zero. The transport of both 234U and 238U was simulated with varying retardation coefficients. The best fit occurred with a RD of about 30, which corresponds to a KD of 6. The results indicate that modeling may be reliably conducted over time periods and distances of interest in radionuclide transport performance assessments and that regionally averaged hydrologic properties can provide appropriate values for predicting ground-water flow and nuclide transport.