The Nuclear Regulatory Commission’s Break With Reality

Engineers who must make predictions about accidents, health effects, and the behavior of the complex systems they are building will sometimes use “conservative” assumptions; that is, when asked “what’s the worst that could happen?” they will invent something so awful that they are certain that reality can’t be that bad, and thus that their bottom line can’t possibly understate the risk.

The Nuclear Regulatory Commission (NRC) has been writing worst-case scenarios for years, but now the staff has gone to a new extreme. In its framework established to calculate the risk from advanced reactors, it has proposed the impossible. The conditions it postulates are impossible, and the rules are impossible to meet.

Among the assumptions:

• The reactor suffers the worst possible accident, is promptly repaired, and has the same accident again the next year, and the year after that, and the year after that, every year for as long as it is licensed, 40 years.

• Residents of the area never move, so they are available to soak up the cumulative dose of radiation, every year for 50 years.

• Residents never die of anything else over the course of the 50 years, so a person who is 65 years old at the time of the first accident is assumed to live to be 115, during which time he or she soaks up doses from each of the next 39 catastrophic accidents.

NRC regulations and guidance around nuclear reactors are filled with “conservative” assumptions. For example, exposure from an accident at today’s reactors is usually calculated for individuals who stand outside, in the location where the dose would be highest, without doing anything to protect themselves or leave the area. And when a reactor is decommissioned, the exposure to people in the future from residual radioactive material left at the site is calculated by looking at the most extreme case: a subsistence farmer who grows all their food on that site. These rules are applied to places that are unlikely to become farms, because they are industrial sites, in areas where there is no farming, let alone irrigated farming or subsistence farming.

As unlikely as subsistence farming on a nuclear site is, accidents every single year are even further from plausible. The rules that assume such annual accidents appear in the NRC’s proposed Part 53, which is the new licensing framework meant to govern advanced reactors. As we have written elsewhere, there is a lot wrong with Part 53. It was supposed to set standards and let the reactors figure out how to meet them, but instead it falls back on the same prescriptive rules as in current regulations—rules that will impose useless burdens on new reactors. Part 53 was also intended to recognize the inherent safety features of advanced reactor designs, but it does not.

The assumptions about accident frequency and the surrounding population are particularly egregious, though. They appear in a section of Part 53 called “Alternative Evaluation of Risk Insights,” or AERI, which is meant to be an easier, simpler way to calculate risk. The standard method is a Probabilistic Risk Assessment (PRA), a technique that involves building a fault tree, assigning a probability to each component that could go wrong, and then a probability that a combination that could cause damage would occur. PRA is slow and expensive, and not particularly suited for some designs like microreactors. But AERI as written in regulatory guidance, may be totally useless. The Breakthrough Institute recently released an analysis that reveals the flaws in the AERI approach.

AERI’s mathematical derivation used to justify a 25 rem dose assumes one catastrophic nuclear accident total— assuming every plant will have a catastrophic failure is already a very conservative assumption. But on paper, the guidance multiplies that risk to assume that a catastrophic accident occurs every year for the duration of the reactor licensing period.

To put this into context, look to Three Mile Island, the most serious commercial nuclear power plant accident in U.S. history. The damaged reactor unit (TMI-2) never ran again, while the undamaged reactor unit (TMI-1) did not restart until 6 years after the accident at its next-door twin. So, how is it that the NRC can expect a reactor to be operational within 12 months of a previous catastrophic accident?

Another problem in AERI is the dose consequence. AERI stipulates that for one catastrophic accident, the population will be exposed to 2.5 rem in the first year and 0.46 rem in the second and subsequent years. This means that by the end of the 50 year dose period, the cumulative exposure will be 25 rem. If we apply the annual frequency of catastrophic accidents as stated in the regulatory guide, the cumulative dose consequence becomes outrageous, clocking a 1002 rem exposure 50 years after the last accident. Compare that to the dose consequence of TMI-2, where the maximum exposure to the population for the duration of the accident was only 0.001 rem (1 millirem) above background radiation. In other words, the allowable dose from AERI is 1,000,000 times as much as that from the TMI-2 accident.

This is an issue that the Breakthrough Institute has brought to the NRC’s attention on several occasions; however, the NRC staff have not recognized nor addressed the problem. It is an easy fix, as our analysis shows, that the NRC staff seems unwilling to make. AERI was derived based on one catastrophic accident per license period, the regulatory guide should state the same.

This is yet another case where the agency could use some reality testing.

Click here to read our whitepaper on the flaws with AERI.