Expert suggests way of assessing safety of nuclear waste disposal site
RENO, Nev.—The Nuclear Regulatory Commission‘s most important criterion for assessing the safety of the Yucca Mountain nuclear waste repository is based on a complicated mathematical analysis that can be misleading, due to large uncertainties in its results, says a University of Michigan professor who has studied radioactive waste disposal for more than 20 years.
A better approach—that should be used along with the mathematical analysis—would be to develop and evaluate the safety of a series of independent barriers, such as the canisters in which nuclear waste is sealed, the rock in which it is buried and the rate of ground water flow through the system, says Rodney Ewing, who served on a scientific committee that spent two years evaluating the present assessment method. Ewing will present his views—published last year in the journal Science—during a panel discussion Nov. 14 at the annual meeting of the Geological Society of America.
“I am certainly not saying that the site isn’t safe,” Ewing explains. “But I don’t believe that the results of the performance assessment should be the sole criterion for determining the safety of the site. I’m asking—not so much as a scientist, but as a member of the public—why should I accept the results of an extremely complicated analysis for which there can be no verification?”
The problem of how to safely dispose of radioactive waste generated by commercial nuclear power plants and the production and dismantling of nuclear weapons has been under study for the past 40 years. Experts agree that nuclear waste should be permanently buried deep underground, and Congress has directed that the Yucca Mountain site be investigated and licensed as soon as it is shown to meet the regulations of the Environmental Protection Agency and the NRC. Plans call for the licensing process to begin in 2002 and for the repository to open to receive waste in 2010. In 1998, the U.S. Department of Energy completed a systematic analysis of the site’s expected performance and declared that there were no “show stoppers”—no problems with the site that should halt its development as a nuclear waste repository.
But Ewing maintains that the method DOE used for its analysis—called probabilistic performance assessment, or PPA—may not be able to recognize a “show stopper” because of the very large uncertainties in the analysis. A site such as Yucca Mountain may in fact be safe, but because the uncertainties are so large, it is difficult to know whether the results of the analysis truly reflect the site’s safety.
“This is because the analysis involves thousands of variables, many of which are not based on experiments or actual observations but rather are established by panels of experts,” explains Ewing, who has joint appointments in the departments of geological sciences, nuclear engineering and radiological sciences, and materials science and engineering at U-M.
The original concept for nuclear waste disposal called for a “belt and suspenders” approach—a series of engineered and geologic barriers that would independently prevent the release of radioactivity to the environment. The idea was that if one barrier failed, the others would back it up. To assure that this redundant system would work, the performance of each of the independent barriers was to be evaluated separately. The present approach no longer requires a specific evaluation of each barrier’s performance. Instead, the probabilistic performance assessment combines all of the data on the site’s geology and design into mathematical descriptions of the relevant processes. Hundreds of calculated results are combined to project the range of future behaviors of the repository site over the next tens of thousands of years. The final result is a number—or a range of numbers—that will be used as the most important criterion for determining the site’s safety.
But Ewing cautions against putting too much faith in the results. “There is very limited, relevant experience in modeling geologic systems of this size and over extended periods of time. One should not expect greater success with such a prediction than we have in other fields, such as predicting which presidential candidate gets the electoral votes from Florida.”
Ewing notes that the analysis provides no indication of how reliably the final number or range of numbers describes the actual long-term behavior of the repository site.
“With the large uncertainty it may not be possible to distinguish between a repository that meets the regulation’s requirements for safe exposure and one that does not,” says Ewing. “How can I tell the difference between these two possibilities?”
Complicated models have their place, Ewing concedes. PPA has long been used to analyze the safety of nuclear reactors; for example, the method can accurately predict how often valves will need to be replaced over a reactor’s 50-year lifetime. But that’s a relatively simple system, analyzed over a short time period, says Ewing. Moreover, analysts can use direct observations and experience to predict how various parts of the system are likely to perform—and to confirm how accurate their predictions were.
A nuclear waste disposal site is a far more complex system with many more unknowns. To use PPA on such a system, analysts must evaluate a tangle of processes and probabilities, such as the probability that water will reach a disposal canister; the probability that the canister will corrode; the probability that water will enter the canister; and the probability that, after several other possible steps, contaminated water will leak out into the rock and finally reach the water table. Information on hydrology and geochemistry must be incorporated in order to predict how the radioactive material might travel through ground water. Possible climate change and volcanic activity must be factored in. Even human behavior is taken into account. The analysis must consider the likelihood that someone will drill a well into the contaminated ground water, the probability that person will drink the water, and the amount of water the person is likely to drink.
“Many of these events and processes cannot be known,” says Ewing. So instead of being based on exact measures of performance, the analysis depends on many assumptions. A set of overly optimistic assumptions may exaggerate the safety of the site, and by the same token, a series of overly conservative assumptions may cause a regulatory agency to abandon a perfectly good site.
“The key point is that one must carefully review the analysis, but even then, one must view the analysis as a very qualitative description of the long-term behavior of the site,” says Ewing. “The results of the analysis should not depend on the assumptions in the analysis, but rather should depend on the actual, confirmable properties of the site. If an airplane were built using only probabilistic analysis—that is, smaller versions of the plane had never been test flown, but you were assured by good and competent engineers and scientists that they had successfully modeled the plane’s ability to fly—would you fly on the first airplane, based on those analyses? Or would you want to see the results of some test flights?”