Earthquakes and the H.B. Robinson Plant: A Case Study of the NRC’s Failure to Learn from the Fukushima Accident

October 22, 2020 | 10:30 am
Nuclear Regulatory Commission/Flickr
Ed Lyman
Director, Nuclear Power Safety

In response to the 2011 Fukushima Daiichi triple-meltdown accident in Japan, the Nuclear Regulatory Commission (NRC) ordered nuclear reactor owners to conduct studies to revise and update the seismic hazard profiles at their sites. After nearly a decade, the results are in. Which nuclear plant in the United States has the reactor (or reactors) most susceptible to melting down after an earthquake? If you guessed the answer is the two-unit Diablo Canyon, the only nuclear power plant currently operating in seismically challenged California, you would be wrong.

The “winner” (or more accurately, “loser”) is apparently—wait for it—the H.B. Robinson Unit 2 reactor near Hartsville, South Carolina, owned by Duke Energy Progress (“Duke”). There is only one nuclear reactor at the site: Unit 1 is a shut-down coal-fired plant. Duke’s risk assessment found that, on average, there is a one in 7,700 chance per year that an earthquake could cause a meltdown at Robinson—five times higher than the average estimated risk for each of Diablo Canyon’s two reactors.

This shouldn’t be a big surprise. South Carolina, where the city of Charleston suffered a devastating earthquake in 1886, is one of sixteen states with the highest seismic risk, according to the United States Geological Survey. But this danger is compounded at Robinson—partly because of poor plant design and partly because of a phenomenon known as liquefaction: the potential for ground shaking to cause the underlying soil to lose strength and its ability to support buildings and other structures. Of particular concern is the possible failure of the adjacent Robinson Dam, which was built to create Lake Robinson, the major source of cooling water for the plant’s radioactive fuel and critical safety equipment such as emergency diesel generators.

Robinson was one of seventeen nuclear plants that the NRC identified as having the largest differences between the severity of the seismic hazards revealed by the new studies and the original hazards that were assumed when the plants were designed and licensed. The NRC required this group of plants to performed detailed analyses called “seismic probabilistic risk assessments.”

The ten US nuclear reactor sites with the highest likelihoods of meltdown triggered by an earthquake are listed in Table 1, in order of decreasing risk. UCS compiled this list from plant seismic probabilistic risk assessment reports and NRC reviews of those reports. The NRC itself does not provide such information to the public in a form that would facilitate ranking the safety of different plants.

(1) The Vogtle plant in Georgia was required to conduct a seismic probabilistic risk assessment, but unlike other plants, the value of the seismic meltdown frequency does not appear in the public documentation on the NRC website. Thus it is not publicly known how Vogtle’s seismic risk compares with the plants on this table. (2) This is the annual likelihood that a single reactor will experience a core melt following an earthquake. The total site risk is actually the cumulative risk considering the possibility of multi-reactor accidents. However, the NRC’s (irrational) policy is to consider only a single reactor accident when evaluating nuclear plant safety in the context of its Safety Goals. (3) The risk values of different reactors at multi-unit sites are not always the same. In those cases, the highest site values are reported. (4) Unlike other plants, the publicly available Peach Bottom seismic documentation only specifies the point value of the meltdown frequency, not the mean value over the uncertainty distribution, which would be higher.

As Table 1 shows, the Robinson reactor has the highest mean (or average) likelihood of experiencing an earthquake-induced meltdown of all the reactors that were required to conduct seismic probabilistic risk assessments. One of the reasons for this is that much of the risk does not come from the most severe earthquakes believed to be possible at the Robinson site, but from relatively mild (and more frequent) earthquakes. This is “atypical,” according to the NRC, since nuclear plants are supposed to be able to withstand small earthquakes by design.

Not every core meltdown accident will result in significant radioactive contamination off-site. However, Duke calculated that there was a 1 in 40,000 risk that a seismic event at Robinson could cause not only a core meltdown but also a large near-term release of radioactivity to the surrounding area, which includes the city of Florence, South Carolina (within thirty miles) and the cities of Columbia, South Carolina, and Charlotte, North Carolina (within eighty miles).

While these may seem like small numbers, they are large in the context of nuclear plant safety. The NRC has concluded that “the risk from seismic events at Robinson is high.” The agency considers a core damage frequency of one in 10,000 per year from all potential accidents—including pipe breaks, fires, and floods—to be the upper limit for acceptable reactor safety risk. Similarly, it considers the highest acceptable frequency of a large near-term (“early”) release of radioactivity to be one in 100,000 per year. (These are called the Safety Goals.) But at Robinson, the risks due to earthquakes alone exceed the Safety Goals.

However, the NRC’s Safety Goals are non-binding targets and not regulatory requirements. Accordingly, the NRC has not ordered Robinson to shut down or to make structural modifications to mitigate the high seismic risk, such as reinforcing the plant’s seismically vulnerable turbine building, which could collapse and disable critical backup cooling systems needed to keep the reactor’s fuel from overheating.

Instead, Duke has committed on a voluntary basis, to make upgrades only to the emergency backup safety systems that the NRC required nuclear plants to acquire after Fukushima. This less-than-half measure simply is not sufficient to mitigate the risk. Given that Robinson may operate for at least another thirty years, the NRC should revisit this decision.

The lessons of Fukushima

The Fukushima accident, which resulted in three nuclear reactor meltdowns, was triggered by a massive earthquake that generated tsunami waves far higher than the plant was designed to withstand. As it happened, the reactors were not seriously damaged by the earthquake itself, but the catastrophic flooding from the tsunami damaged the plant’s emergency power supplies and electrical systems, causing a near-total loss of electrical power and disabling the pumps that provided critically important cooling water to the hot nuclear fuel.

Two of Fukushima’s main lessons were that (1) nuclear plants need to have additional means of coping with a prolonged loss of electrical power; and (2) plant owners need to ensure that their assessments of external hazards such as earthquakes and floods are accurate and up-to-date, and then implement additional measures where necessary to protect the plants from these hazards.

Following the Fukushima accident, the NRC required, among other things, that all US plant owners (1) develop plans and acquire emergency (“FLEX”) equipment to provide backup power and cooling if power is lost for long periods; and (2) reevaluate the earthquake and flooding hazard profiles at their sites based on updated information and methods.

The hazard reevaluation studies that plant owners conducted found that nearly all nuclear plants were vulnerable to more severe earthquakes and floods than they were prepared to withstand. However, the NRC, inexplicably, has not required a single plant to upgrade its defenses in response to these findings despite having the regulatory authority and responsibility to do so.

The risk of a meltdown at Robinson

In the case of Robinson, the NRC’s decision not to take regulatory action is clearly misguided. The NRC noted in its review of the Robinson seismic study that “the risk, using present-day information and more sophisticated tools, is an order of magnitude [factor of 10] higher and more consequential than previously known.” Particularly alarming was the revelation that Robinson is not only at risk from earthquakes that exceed its current design requirement—the so-called “Safe Shutdown Earthquake”—but also from smaller earthquakes that the reactor should have been able to withstand.

For example, the study estimates that there is a twelve percent chance a Safe Shutdown Earthquake would cause the dam to collapse and the lake to drain due to liquefaction, causing a loss of the primary water source relied on to remove heat from the reactor. Therefore, Duke’s plant is not in compliance with the NRC’s fundamental requirements for earthquake protection.

Another high-risk scenario is the collapse of the building containing the plant’s turbines, which was not built to be highly seismically resistant. This could result in multiple safety system failures by disabling critical electrical equipment and emergency pumps, including the recently acquired FLEX equipment that was intended to be used as backup in loss-of-power scenarios caused by natural disasters. Without these pumps and other equipment, plant workers would have no means to provide water to the steam generators and remove heat from the nuclear fuel in the reactor core, eventually resulting in a meltdown.

Duke’s mean seismic meltdown risk estimate of one in 7,700 per year is the sum of the risks of these and other potential accident initiators. One should note that these risk values are only rough estimates— predicting the chance of an earthquake is an uncertain business and error bars are large. Taking those uncertainties into account, Duke calculated that there was a five percent chance the actual meltdown risk could be four times greater than the mean value, or about one in 2,000 per year.

Robinson’s license to operate, which has already been renewed for an additional twenty-year term for a total of sixty years of operation, is slated to expire in 2030. However, Duke has announced that it is planning to apply for subsequent twenty-year license renewals for its entire fleet, including Robinson. If approved, the likelihood of a seismically induced meltdown at Robinson over its remaining lifetime would be about 0.4 percent on average, and could be as high as 1.5 percent.

An inadequate response

The most straightforward way to reduce the high seismic risk at Robinson is for Duke to reinforce the turbine building and other vulnerable structures so that they are less likely to collapse if an earthquake occurs. But that option is apparently too costly for Duke.

Though Duke plans to make modifications to reduce the vulnerability of the emergency backup (FLEX) equipment and emergency water sources so that they would have greater chances of surviving an earthquake, this is a poor substitute for structural reinforcements. While this action could increase the survivability of the emergency equipment that personnel could use to mitigate earthquake destruction, it would not help prevent that destruction from occurring. While functional mitigation capabilities are important, they require actions by site personnel, which introduces additional risks—for example, if workers are injured or are prevented from setting up and using the emergency equipment in time to restore core cooling because of earthquake damage at the site. It is far preferable to reduce the likelihood of structural damage in the first place.

The limited benefit of these changes is apparent from Duke’s own analysis, which found they would reduce the risk of core damage from earthquakes by only forty percent. When other accident initiators, such as fires, are taken into account, the total core meltdown risk would be about one in 7,750 per year—which still fails to meet the NRC’s one in 10,000 per year Safety Goal.

So why did Duke decide to pursue this alternative approach? Because of the price tag. Duke told the NRC that the “Class III [Turbine Building] structural modification and other dominant risk contributor modifications were not being pursued based on the comparison of the modification cost estimates to the proposed modification strategy cost estimate.” In other words, Duke decided beforehand on the amount of money it was willing to spend to fix the problem, and the NRC acquiesced.

But that’s not the way the process should work. The NRC should require that the plant meet its safety goals, regardless of cost. And Duke should consider the cost of the necessary seismic upgrades relative to investments in safer, less risky low carbon alternatives like energy efficiency and renewable energy.

What the NRC should do

The NRC has determined that Robinson is out of compliance with original assumptions about the plant’s seismic risk. In fact, the reactor has been vulnerable to earthquakes that it was supposed to have been designed to withstand since it began operation in 1970. Unfortunately, this issue was not considered when the NRC renewed Robinson’s license in 2010, because the agency’s license renewal regulations do not require a comprehensive review of plant safety and a confirmation of the validity of the original licensing assumptions.

The NRC should not repeat that mistake—and should not approve a second license renewal, allowing Robinson to operate until 2050, unless Duke rectifies this long-standing and dangerous situation.

Other nuclear plants seeking subsequent license renewals, such as North Anna in Virginia and Oconee in South Carolina (see Table 1), also have seismic risks high enough so that their overall meltdown risks likely exceed the NRC Safety Goals. The NRC should thoroughly assess the need for seismic reinforcements as part of its review of license renewals for these plants.