Reactor Accidents and Afterthoughts

NUCLEAR ENERGY in general is controversial, but one of the most passionately felt controversies of all has revolved around the question of accidents in nuclear reactors. Are reactors safe? The central issue stems from the awkward technical fact that the heating process generating power in nuclear plants is difficult to turn off. In case of operating problems, the main nuclear reaction in the uranium fuel can be stopped quickly, but the heat from waste products remains at a high level for long periods of time. This means that the nuclear fuel must be kept cooled, or else the metal-encased fuel rod will melt (a meltdown), releasing radioactive gases and clouds of particulates.

Naturally, precautions are taken against such an event. These are backup water supply systems to supply cooling in case of failure in the regular cooling systems that transport the heat used to generate steam. Nuclear power stations also have a large variety of other safety systems, designed to prevent a meltdown or to keep the radioactive gases from escaping into the atmosphere should a meltdown occur. Nevertheless, should the gases and particulates escape, the amount of radioactivity that would be spread about could be quite large, depending on how well the safety systems work.

Fortunately, such an accident has never happened. Unfortunately, this makes it difficult to analyze whether nuclear power plants are sufficiently safe. We must estimate the likelihood that such an accident would happen, guess at its consequences, and decide whether the risk is too great for the advantages of using nuclear power in the first place. Naturally, no such theoretical answer to the problem will ever satisfy the most demanding critics of nuclear energy. And one cannot pretend that theory is preferable to experiment. But even so, theory may be better than nothing.

It may be true, strictly speaking, that no meltdown occurred at Three Mile Island. However, the fuel rods did break, and some radioactivity was released.

Accidents and Predictions of Accidents. Nuclear power is unusual among modern technologies in that such a theoretical analysis does exist. Trying to clear up the doubts about safety that have clouded the future of nuclear electricity, the U.S. Atomic Energy Commission (AEC) and its successor, the Nuclear Regulatory Commission (NRC), issued in 1975 a complex mathematical study of the probabilities of nuclear reactor accidents. Since no large accidents, and relatively few minor ones, have ever occurred, the study relied on estimates of the probability that each of the different pieces of equipment and various safety procedures in the plant might fail. All these individual probabilities were combined, as they might enter into the chances of nuclear meltdowns and other serious accidents, and numerical predictions of reactor safety were made.

It seems strange to realize that any one single accident such as the one at Three Mile Island still tells us almost nothing about probabilities of these events. We would need hundreds of accidents, which is probably not a result that anyone is anxious to achieve.

The Reactor Safety Study, often called the Rasmussen report for Norman C. Rasmussen, the director of the study, predicted that on the average the damages to health and property by radioactive contamination, as a result of nuclear reactor accidents on a year-in—year-out basis, would be small. Here, the word small means as compared with effects from naturally occurring radiation or with risks from coal-fired power plants. The predicted fatalities (ranging from 1 to 20) and illnesses (ranging from 20 to 200) are significantly larger than those predicted for normal operations (at least for short-lived isotopes). Still, the incidence of fatalities from nuclear reactor accidents is less than that for many other coal- and nuclear-related ones, such as the fatal occupational diseases and accidents to coal workers. The predicted incidence of nonfatal illnesses is also small as compared with coal-related illnesses.

Here I used the error bounds more or less suggested in the Rasmussen report itself. In view of the Lewis report,* which also came out after my book was published, and the Three Mile Island incident, I would feel happier with much wider error limits—both ways—or from, say, 0.00002 to 2 predicted fatalities. Such a change would mean that the statement about fatalities from nuclear reactor accidents being less than those for any other coal-related accidents or nuclear-related events would no longer necessarily hold.

What the average numbers fail to show is that most of the predicted effects do not stem from a series of small accidents happening all the time, but are a result of a number of medium-sized accidents that have a chance of happening with relative frequency, and a very few serious accidents that should happen very rarely indeed. The worst-case accident, for example, was predicted to kill some 3,000 people by acute radiation sickness and some 45,000 others by radiation-induced cancers over a period of years. The yearly probability of such an accident occurring in any one reactor was predicted, at the outside, as being only five chances in one billion. This turns out to be a very small risk when compared with each citizen's chances of being struck by lightning, for example. But some people might think that the risks of a catastrophic accident happening only rarely are a somewhat different proposition—maybe worse, maybe better—than, for example, the few random deaths yearly caused by constant small leaks of radioactive material into the air near a power reactor. After all, normal radiation from reactors is uncertain in its effects because there is no answer to the question, Who will be affected? Radiation from reactor accidents is doubly uncertain, because the answers are unknown to the questions, When might an accident happen? How much radiation might be emitted?

Other unknowns are troublesome as well. The effect of an accident would depend on the nature of the winds and rains that would carry the radioactive material from the plant and deposit it where it could do damage to the surrounding population. The Rasmussen report averaged out weather and population over a great number of sites. It is perfectly proper to average these factors out. But our future projections of accident hazards might be slightly more pessimistic if additional nuclear plants tended to be located closer to densely populated centers or, more likely, if rural areas around existing plants become more urbanized.

*(H. W. Lewis, chairman, Ad Hoc Risk Assessment Review Group, Risk Assessment Review Group Report to the U.S. Nuclear Regulatory Commission, NUREG/CR-0400 (Springfield, Va., National Technical Information Service, September 1978.)

Critics and Catastrophes. Does the Rasmussen report solve all our questions regarding safety? Hardly. In the first place, even though the report has received approval in some quarters, it has not gone without criticism elsewhere. Various critics, who have maintained that the risks are underestimated, have disagreed as to whether they are underestimated by a factor of 10 or 100, or more. Obviously, as we shall see in comparing these risks with the total risks for coal, the degree of understatement could be important. Perhaps these disagreements will be resolved by a comprehensive review of the whole report that is presently planned.

The Lewis report said the group was unable to draw up error limits on the accident probabilities but that it was evident that the error limits had been underestimated. If important accident sequences were omitted, such as appears to have happened at Three Mile Island, this criticism becomes indeed trenchant.

In the meantime, it should be noted that a recent review of the subject, while exceedingly critical of many parts of the Rasmussen report, has calculated on actuarial grounds that the dangers of a nuclear reactor accident were, at most, equal to 500 times Rasmussen's value. Five hundred times a minor impact turns out to be a larger but not overwhelming effect.

The effect may not be "overwhelming" as I stated, but it does mean that nuclear risks are up there with the coal risks.

Naturally, any study is only as good as its data. And data problems exist. Critics have pointed out the great difficulty of estimating the probability that an event initiating an accident will occur. Some possibilities are especially difficult to treat, such as a string of accident-causing events that occur together, and it could be that the study makes mistakes in this area. The behavior of the Emergency Core Cooling Systems (ECCS) cannot really be tested until there is an actual accident.

It is unclear whether the shutdown state of the cooling pumps on the secondary loop at Three Mile Island because of an oversight was an example of an area that is not treated correctly by the report, and the entire role of human error in the "common mode" failures may have been underestimated. As far as the ECCS systems are concerned, we have to find out whether they worked correctly; however, the irony is that the mysterious hydrogen bubble, rather than the pressure problems that have been of particular concern in theoretical analyses, seems to have gotten in the way of the correct operation of the system.

A Psychological Note. One odd risk is that of public reaction to nuclear catastrophes. It is widely thought that if there were a large accident, resulting in deaths and injuries, public reaction would force a shutdown of all reactors. If so, the resulting waste of resources would be an expensive unpaid cost of nuclear power.

This is a curious point. Even though there were no deaths—except for possible latent cancer fatalities of which we will be hearing a good deal more in the months and years ahead—I would have expected there to have been more outcry from the public about the Three Mile Island incident. While everyone took the accident seriously enough, and people were indeed frightened, I have yet to see a groundswell for closing down nuclear power plants.

The most interesting aspect of such a shutdown is that a nationwide reaction of this kind would be more surprising if the catastrophe were to involve hydroelectric power (a dam break) or an air pollution catastrophe related to coal-fired power. Naturally, there are generally differences in the nature of the health risks concerned for catastrophes involving the different technologies. Nevertheless, there may well be special aspects to the way the public perceives nuclear risks, even when those risks appear to be equivalent to other nonnuclear risks in type and magnitude. The reasons for such a special status are not clear, but anxieties related to nuclear war and to fears of radiation and radiation-related death have been suggested as possible motivations.

On the other hand, the unfamiliarity and newness of the nuclear power industry may play an important role, in that the public must learn to judge a whole array of potential consequences of a new technology.

Ironically, it may be that Three Mile Island is making nuclear less "unfamiliar."

Fragmentary evidence that unfamiliarity may be important has been observed in studies showing that people have more anxiety about a local nuclear plant if they live moderately close to the reactor site (several kilometers away) than in the immediate neighborhood (less than a kilometer away).

It would be interesting to know whether this phenomenon took place in the communities near Harrisburg.

But one cannot tell whether such phenomena are indeed psychological, that is, irrational, or whether they represent a strong distaste for catastrophic risks, or whether they merely reflect a public skepticism as to the opinions (or capabilities) of nuclear experts. Furthermore, is such an attitude consistent with recent polls showing strong public support for developing nuclear technology? If such special attitudes toward nuclear risks are important, they are fully as legitimate—though perhaps not as immutable—an unpaid cost of nuclear power as any other. And such costs would become tangible in monetary terms if the occurrence of a catastrophic accident led to a lengthy or even permanent shutdown of all reactors in the United States.

I would guess the purely psychological part of the anti-nuclear reaction mentioned here was not prominent and that a healthy skepticism about experts and a distaste for catastrophic risks showed up as one might have expected from a rational public.

Safer Reactors? What could be done to improve reactor safety? A significant amount of money already goes into research and regulation to improve safety by preventing and mitigating the results of accidents in nuclear power plants. Expenses for extra standby safety systems are already large. For example, the ECCS required within the last few years for all plants have inevitably added to the cost of nuclear electricity. Every new problem that arises, from cracking of the stainless steel pipes that carry the essential cooling water, to fires in so-called flame-resistant materials that can paralyze control circuits, leads to new and improved safety regulations and therefore to new costs. But there is a question of how much safety can be improved in any pressurized system in a cost-effective manner. Not to mention the problem of the "reliability" of the human beings that run the system; automation can help to alleviate this latter problem but cannot do away with it entirely.

It might be that pressurized systems, at least the light water reactor, are just too unsafe. Maybe nuclear policy should examine gas reactors or low-pressure systems such as liquid metal coolants to develop inherently safer systems.

It would help if we were sure that the safety level of new reactors was the same or better than that of the Rasmussen calculation. The Rasmussen results depend on the actual finished engineering of the two reactors studied in detail. Almost every individual reactor—at least up to now—has been a new and different engineering creation, even if the differences have sometimes been minor. Each reactor receives an exceedingly thorough safety review, but even so, the lack of homogeneity means that the Rasmussen results do not necessarily apply to all reactors at all sites. If each plant were to be analyzed separately in the Rasmussen style, the expense would be enormous, since it is unlikely that the detailed calculation can be carried out from blueprints alone.

If a point of sharply diminishing marginal returns has been reached in safety systems, it still should be possible to increase safety levels by siting nuclear power plants farther away from population centers. It has been suggested that plants be sited underground. Unfortunately, there are drawbacks to these solutions. If a small cloud of radioactivity were released just west of Fort Worth and blown eastward by the wind, the individual dose would fall off with distance, but the widening cloud would include more people; therefore, the impacts in both Fort Worth and Dallas would be important, even though the site would be relatively "remote" from Dallas. Underground siting promises to be costly and is good only to the extent that the radioactivity released cannot penetrate to the surface—a question of some uncertainty. But it would be difficult to achieve levels of safety high enough to satisfy some conceivable standards without moving nuclear power plants to the North Pole, or at least to isolated islands in the Pacific—as indeed has been proposed in recent futuristic studies.

Siting power plants farther away from population centers, while it might help the emergency evacuation problem, is probably a no-win policy and appears even less promising after the Three Mile Island incident. The AEC was right not to get into urban siting of power plants, but it is difficult to place a site far enough away that it will not cause trouble to somebody somewhere.

All this legitimately raises the question of how safe is safe enough. It is notoriously difficult to pursue any human activity without risks. Usually we have to balance risks against benefits, and on this basis the Rasmussen results seem reassuring. But the catastrophic possibilities involved suggest that improving the estimates of nuclear accident risks be given high research priority.

Subsidies and Risks. Is the electricity-using public paying for the possible costs of nuclear catastrophes? This question comes up because if an electric utility has a serious accident in one of its nuclear power plants, its liability is restricted by law—currently to $560 million —under the Price-Anderson Act. Comparing this liability restriction to the probable losses from an accident could give an indirect indication of the value of the unpaid costs of nuclear accidents.

There is also the related implied question that has excited public interest in this law: that is, if nuclear reactors are so safe, why does the nuclear industry need a special dispensation against liability? This is a good question, and the Rasmussen study was designed to answer it. The answer seems to have an Alice in Wonderland quality in that the report concluded that the limitation was justifiable because the accidents were so unlikely that no limitation was needed! Therefore, if we accept the Rasmussen results as reliable for an estimate of reactor accident dangers, clearly we have to accept the implication that Price-Anderson liability limitation is not a serious problem. Prorated costs would be less than $40,000 per reactor every year, or about 1 cent on the "monthly electric bill." We would infer higher estimates of risk from the premiums now actually being charged to utilities (for the $125 million for which they are liable). But even on the basis of such risk estimates, the effective subsidy for higher liabilities would still be small, representing an unpaid cost of, say, 6 cents on the "monthly electric bill." But again, the billed cost of an actual catastrophic accident in any given year would be more like $28, or $25 billion for 1 USW.

The insurance problem obviously has to be rethought. Even Three Mile Island liability claims may be substantial. Claims after a real meltdown could be astronomical.

The Breeder and Plutonium. Is there a breeder reactor economy in the U.S. for the future? It is difficult to tell. But it may be mildly reassuring to realize that safety problems are likely no worse for the breeder than for ordinary LWRs. Some new instability problems do arise that could conceivably lead to large releases of energy and radiation, because each gram of fuel contains a relatively large number of active (fissionable) atoms. The liquid metal breeder (LMFBR) that has been under development in the United States would also have problems involving the corrosive, extremely chemically active metal sodium that is used to transfer heat from the reactor to the steam generator. On the other hand, the sodium is not under pressure, so that it is easier to avoid the sudden losses of the cooling liquid that can occur in ordinary pressurized LWRs. Safety concerns for the LMFBR could turn out to be similar in degree, if not in kind, to those for LWRs.

Other new types of accidents might occur within the other parts of the plutonium economy, however. The reprocessing plants that separate the plutonium from the spent uranium fuel could be hazardous. Sudden, large accidental releases of the radioactive wastes in the fuel are not likely to occur during reprocessing; in fact, official predictions of possible accidents show that the probability is small that even one death in such an accident would occur—which would have little impact if the occurrence of the accident is improbable also. There has been some public concern about the handling and shipment of plutonium in general, because—judging from large doses given in animal experiments—even small quantities of plutonium are carcinogenic if inhaled into the lungs. But it has been estimated that only a small amount of plutonium and related elements would be released during normal operations from a large breeder reactor fuel cycle. And it has been argued that in practice much more plutonium than the microgram quantities sometimes quoted would have to be initially dispersed in order to be ultimately inhaled by human lungs and cause cancer. Even if such estimates were correct, some claim that plutonium is still a critical safety hazard when it is deposited nonuniformly in the lung. Such a hot particle theory is possible, but there is little convincing evidence to back it up. This is not to say that all these plutonium problems might not turn out to be more severe than we think. We would be dealing with a new kind of technology, one we do not fully understand. But based on what we know at present, and assuming prudent design, only the level of concern appropriate for any technology having certain obvious risks seems called for. And it is fair to also take into account that the use of a breeder would reduce the expected number of yearly fatalities from uranium mining and milling operations.

Conclusion. Even though no catastrophic accidents have taken place, there is an inherent danger of serious, accidental radioactive emissions from nuclear power plants. A large theoretical study has calculated that risks from this problem are small—on the average—as compared with other electricity-related risks. Although the Rasmussen study is not above criticism, pending further investigation the results constitute a reassuring best estimate of the risk of nuclear accidents. We must also recognize that there is always the chance that a fairly large catastrophe will occur; it is a legitimate value choice to view such a risk as worse than a noncatastrophic risk.

There is no compelling reason to expect dramatic improvements in the safety of reactors. Breeders and plutonium economies introduce some new safety problems—such as dangers of cancer from inhalation of plutonium and of a more dangerous type of reactor accident—but they may eliminate or reduce other problems such as other specific types of reactor accidents and part of the impacts connected with the uranium mining and milling industry.