time.26 The committee is in general agreement with the appraisals of the reactor safety study conducted by the American Physical Society study group27 and more recently by the Reactor Safety Review Group.28 WASH-1400 contains some estimates that are excessively conservative and others that are almost certainly too optimistic. Which way this would shift the median probabilities for accidents of various severities is uncertain. The consequences of given accidents are apparently underestimated, but probably by not more than a factor of 3. However, the uncertainties in the probability estimates are almost surely several times larger than estimated in WASH-1400. If larger uncertainties are used, the mean, or expected number of fatalities from nuclear accidents, could be higher by a factor of 10 or more than the median values given by WASH-1400 (namely, 0.025 delayed deaths per reactor-year).*†
Catastrophic accidents can also occur in the case of other energy sources, especially large hydroelectric facilities. Between 1918 and 1958, an average of 40 deaths per year resulted from dam failures in the United States, though fewer in the more recent period. Some individual failures killed hundreds. Worst-case scenarios for both dams and LNG facilities lead to numbers of casualties comparable to those associated with the more severe nuclear accident possibilities. The calculated probabilities are higher, although the analyses on which they are based have been much less thorough and systematic than those for nuclear plants.
In the case of the most likely nuclear accidents, most fatalities would be delayed deaths that could not be specifically attributed to nuclear power, due to the exposure of a large population to low-level radiation (chapter 9). Casualties from dam failures and LNG accidents are immediate, with fewer delayed effects. Because such a high proportion of the reactor-related deaths are delayed, and because large populations may be at risk (even though the enhanced risk to any individual may be small), reactor accidents may create much greater apprehension than other types of catastrophic accidents that can cause the same number of fatalities.
Nuclear plants, dams, and LNG facilities are probably similarly vulnerable to sabotage, but nuclear plants are presently better guarded and may be inherently easier to guard. The consequences of sabotage of nuclear plants appear to be in about the same range as those of the severest postulated accidents discussed in WASH-1400.‡ The possible severe consequences could be much higher, though, because saboteurs could choose times and places for maximum effect. The safety analysis techniques developed for assessing nuclear reactor accidents ought to be