In making safety decisions, the Nuclear Regulatory Commission uses accident probability calculations that are much more optimistic than anything that nuclear manufacturers like General Electric and Westinghouse actually believe. The result is weak public protection. A good example is the NRC commissioners’ rejection in 2014 of a proposal to limit the possible severe consequences of spent fuel pool fires in nuclear power plants because the proposal’s cost, however modest, exceeded the value of the expected reduction in “risk.”
Spent fuel pools are where highly radioactive (and thus thermally hot) used reactor fuel is stored after it is removed from the reactor core. If a pool loses its water supply, the spent fuel can overheat and eventually burn, releasing large quantities of radioactivity. The spent fuel pool issue gained prominence after the 2011 Fukushima accident. For a time during the accident the dominant concern was that spent fuel in Fukushima’s damaged Unit 4 pool might catch fire. It didn’t happen, but it could have multiplied the effects of the catastrophic Fukushima accident manyfold. The NRC staff told the commissioners in 2014 that a worst-case spent fuel pool fire in a US plant like those at Fukushima—of which there are nearly three dozen—could release 25 times more long-lasting radioactivity than escaped from the Fukushima reactor vessels, and perhaps even more. Such a release could render 10,000 square miles uninhabitable and (around the Pennsylvania nuclear plant the staff chose as an example) could require the evacuation of 4 million persons.
The specific proposal before the commissioners was to limit the amount of radioactive spent fuel in a pool and thus to reduce the consequences of a fire by a factor of ten. This would be accomplished by speeding up the transfer of radioactive spent (used) fuel from the pool into “dry cask” storage. The plant owners have to do this eventually, but earlier transfers increase the cost. The commissioners saw their role as deciding whether the safety benefit—the reduction in risk—warranted this cost increase.
In fact, they weren’t deciding anything. The commissioners lent an air of official seriousness to the proceeding, but the decision making was on autopilot. It involved calculating the average risk (R) of an accident by multiplying two numbers, the accident’s probability (P) and its consequence (C). If P is sufficiently small, the average risk (or P times C) will be negligible no matter how large the consequence. And, therefore, the possible reduction in risk will hardly be worth any expenditure. That is how it worked in the 2014 case of a possible spent fuel fire, and that is how it has worked in most cases involving protection against severe accidents.
Actually, most cases don’t get this far. The commission has a threshold for the staff to investigate a safety issue posed by a hypothetical accident. If the estimated probability of “prompt” deaths offsite is below 2 in 1 million per year, the NRC staff need not investigate further. This involves a kind of Catch-22. The NRC assumes effective evacuation of the surrounding area in the event of an accident, so there aren’t people to be irradiated, and even substantial accidents don’t exceed the commission’s threshold.
In the 2014 Commission meeting, the NRC staff told the commissioners that the probability of a severe spent fuel pool event was about 1 in 10 million per year, so the case did “not pass the safety goal screening” and required no further review. Nevertheless, “to provide information to the commission,” the staff generously proceeded to the next step: cost-benefit analysis.
In a way, the staff’s analysis in this case was novel in that it included contamination of land and evacuation of people. Traditionally the NRC only considered consequences to people’s health—“prompt” radiation deaths and subsequent cancers. Since Fukushima it has been obvious (and should have been so from the 1986 Chernobyl accident) that the main impacts of a severe accident are the radioactive contamination of land and the evacuation of large numbers of people, many of them permanently.
Consider the implications of NRC’s risk definition for the risk of long-term land contamination: The NRC staff’s projection of about 10,000 square miles, when multiplied by the staff-estimated accident probability, becomes an annual risk of about one-thousandth of a square mile, or less than an acre per year. Since valuable farmland runs at several thousand dollars per acre, the NRC conclusion is that any safety improvement that costs more than that isn’t worthwhile in terms of saving land. Similarly, the risk of displacing persons, becomes about half a person displaced per year, perhaps at a cost of tens of thousands of dollars, and so, again, per NRC logic, it is not worth spending more than that to avoid long-term evacuations to protect against severe spent fuel pool fires. This isn’t the conclusion most people would arrive at for themselves or their home towns.
There are several things wrong with the NRC’s cost-benefit approach to nuclear safety. To begin with, neither factor in the risk formula—probability times consequence—can be calculated with any accuracy. For example, the consequences of an accident requiring the long-term, possibly permanent, evacuation of 4 million will surely not be limited to the expense of such an evacuation. It would, for example, almost certainly spell the end of nuclear power use in the United States and likely in many countries, with huge economic consequences. We know the Fukushima accident resulted in the closing of all Japanese nuclear plants, hardly any of which have gone back into operation. The Fukushima accident cost estimate already runs into the hundreds of billions of dollars. A very much larger such accident in the United States could run into trillions. None of this is part of the NRC’s blinkered analysis.
Nor is the situation much better when it comes to estimating the accident probability. As there is little data on large accidents, the accident probability is a calculated number. The NRC staff relies increasingly on elaborate calculations that model the various failure modes of a nuclear plant. For outsiders, or for that matter the NRC commissioners themselves, the result essentially comes out of a black box. It is questionable whether a frequency as low as 1 in 10 million per year is meaningful. It amounts to saying that if the nuclear plant ran for 10 million years one should expect only one such accident. But one came very close to happening a few years ago at Fukushima Unit 4.
Which brings us to a deep flaw in NRC’s safety methodology—its reliance on the average risk as the figure of merit. It is by no means the only possible measure of risk. We know that in many statistical situations the average is not the best choice to characterize the data. It works where there are well-established data on both probabilities and consequences as, for example, in considering measures to reduce auto accidents. It doesn’t make sense for high consequence/low probability events, for one thing, because the numbers are so poorly known. Also, using average risk doesn’t reflect what most people—the people the NRC is supposed to be protecting—want to achieve. They don’t want to risk losing a city, no matter what the calculated probabilities. That is how the nuclear manufacturers—Westinghouse and General Electric—see it, too. They refuse to participate in any project unless they are guaranteed to be free of any liability for any offsite accident consequences. If they believed the NRC risk calculations, they would have no difficulty in accepting the litigation risk—but they obviously don’t. In short, the organizations most highly knowledgeable about nuclear safety don’t trust the NRC’s probabilistic calculations.
The handling of the spent fuel pool fire issue followed the pattern for current NRC safety decisions—safety goal screening and cost-benefit analyses based on staff-supplied numbers. The commissioners do have the authority to impose any safety requirement they judge necessary regardless of its cost, but they rarely use it. They prefer the protection of the average risk algorithm with inputs provided by their staff engineers. It gives the commissioners’ decision making a scientific gloss, and thus avoids their having to defend a safety judgment about “How safe is safe enough?”
The commissioners did use their statutory authority in 2012 to improve the capability of General Electric plants like the Fukushima units to vent high pressure (radioactive) steam from containments during an accident, to prevent over-pressurization and failure. The commissioners were under considerable pressure to act post-Fukushima, and, as containment venting was a prominent problem during that accident, they overrode their staff’s conclusion that the proposal did not pass “safety goal screening.” But the commissioners managed to put the obvious corollary—requiring filters to keep venting during accidents from spewing radioactivity into the surrounding area—on a slow rulemaking track. Once out of the political spotlight, the commissioners, in August 2015, directed the staff not to proceed further with the rulemaking.
Any change in the NRC’s approach to nuclear risk must come from the outside; the agency has too much invested in the current approach for internal reform to have a chance. When a witness at the 2014 Commission meeting on spent fuel pool fires, Clark University professor Gordon Thompson, questioned using the average risk as the figure of merit, only one commissioner took notice and that was to ridicule the notion. The commissioners should have paid more attention.
A definition of risk that placed greater emphasis on avoiding large-consequence events would be more in line with the common sense of the public whom the NRC is supposed to be protecting. If nuclear power is to have any long-term future, it will have to go beyond even that level of protection. A 2012 report of the American Society of Mechanical Engineers, a group heavily involved with the nuclear industry, called for a major step-up in nuclear safety and warned that severe accident impacts on people’s lives were “wholly inconsistent with an economically viable and socially acceptable use of nuclear energy.” Just as the nuclear manufacturers don’t want to bet their companies on calculations of nuclear safety, neither do people at large want to bet their cities and countrysides.