U.S. flag

An official website of the United States government

NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Institute of Medicine (US) Steering Committee for the Symposium on the Medical Implications of Nuclear War; Solomon F, Marston RQ, editors. The Medical Implications of Nuclear War. Washington (DC): National Academies Press (US); 1986.

Cover of The Medical Implications of Nuclear War

The Medical Implications of Nuclear War.

Show details

Casualties Due to the Blast, Heat, and Radioactive Fallout from Various Hypothetical Nuclear Attacks on the United States

William Daugherty, Barbara Levi, PH.D., and Frank Von Hippel, PH.D.

Princeton University, Princeton, New Jersey


We have developed the tools for calculating the deaths and injuries due to blast, thermal effects, and local fallout from hypothetical nuclear attacks on the United States. This is the first time that the capability to do such consequence calculations has existed outside the (mostly classified) government domain.

We have used this capability to explore the sensitivities of the consequences of a nuclear attack to various assumptions. The first was the sensitivity to the types of targets involved. We examined three different hypothetical ''limited" nuclear attacks on the United States, each involving a 1-megaton (Mt) airburst over approximately 100 targets of three different types:

  • The city centers of the 100 largest U.S. urban areas
  • 101 industries rated as the highest-priority targets for an attack on U.S. military-industrial capability
  • 99 key strategic nuclear targets.

The calculated ranges of fatalities and casualties (deaths plus severe injuries and illnesses) from blast, bums, and radioactive fallout for these "100-Megaton" attacks are shown in Table 1. This table indicates that more than 10 million deaths could result from these "limited" attacks, even if the targets were industrial or military and not population per se. The results also indicate that even a strategic defense system that was 99 percent effective might not protect the United States against potential catastrophe in a nuclear war with the USSR.

Table 1. Estimated Deaths and Total Casualties from the "100-Megaton" Attacks.

Table 1

Estimated Deaths and Total Casualties from the "100-Megaton" Attacks.

We also explored the sensitivity of these calculations to different models for predicting casualties. Lower numbers result if we use the predictions of the traditional "overpressure" model, which assumes that the same casualty rates will occur as those that occurred at Hiroshima at given levels of peak blast overpressure. Higher numbers result when we use a new "conflagration" model (Postol, 1986), which postulates that much higher fatality rates might be expected in the large "burnout" areas that would be caused by modem weapons than occurred in the burnout area of the much lower yield Hiroshima bomb.

We find, for 1-Mt airbursts, that the numbers of fatalities predicted by the conflagration model are 1.5 to 4 times higher than those predicted by the overpressure model, with the exact factor depending on the population distribution and the assumed scaling of the burnout area with yield. The predicted numbers of injured are significantly smaller for the conflagration model because many of the people injured in the overpressure model die from fire effects in the conflagration model. In view of the plausibility of the conflagration model, we believe that previous estimates of the deaths due to the blast and bum effects of nuclear attacks are very uncertain and probably low by a large factor.

Next, we calculated the consequences from a major "counterforce" attack on U.S. strategic-nuclear forces. We assumed an attack on more than 1,200 targets with almost 3,000 attacking warheads. Because such an attack would result in a great amount of local fallout from many ground bursts, our casualty models in this case included the effects of radioactive fallout as well as blast and thermal radiation. The estimated number of deaths ranged from 13 to 34 million people. The range reflects the varying predictions associated with different possible winds, the different models for blast and burns, and different assumptions about the susceptibility of the population to death from radiation. The corresponding final estimates made by the Department of Defense (DOD) in 1975 for a similar attack ranged from 3 million to 16 million deaths (U.S. Congress, Senate Foreign Relations Committee, 1975; pp. 12-24).

Our casualty estimates should still be considered as only a partial accounting of the potential human toll due to the attacks discussed here. Nuclear weapons are powerful enough to destroy both our social and environmental support systems, and the numbers of casualties from second-order effects such as exposure, starvation, or disease could be as great as or greater than the numbers presented in this paper for direct casualties.


An all-out nuclear war between the United States and the Soviet Union would destroy the urban areas of both countries and thereby the infrastructure that makes them modem industrial states. This fact makes the deliberate launching of such a war the ultimate act of folly. Nevertheless, military planners have felt that the United States should have "credible strategic nuclear options," and have worried about those credible nuclear options that the Soviets might devise. This concern led to debates in the 1970s over the possibility of "limited" nuclear wars that might produce significant military results but minimal civilian casualties. During this same period, according to Ball (1983; p. 19), U.S. policy was changed to exclude targeting "population per se"—presumably because "collateral" civilian casualties from the targeting of economic or military facilities were expected to be much lower than those from direct attacks on population centers. And recently, the Strategic Defense Initiative has provoked debates over whether strategic defenses could reduce U.S. casualties from an all-out nuclear attack to less than catastrophic levels.

How much would these options and policies actually buy in reduced casualties? Unfortunately, quantitative estimates of these reductions are hardly ever offered. Yet such estimates of casualties—and, just as important, the public disclosure of the assumptions behind them—are essential to the evaluation of these concepts.

In this paper, we describe the results of an exploration of the sensitivities of the estimates of direct casualties from limited nuclear attacks on the United States to various assumptions concerning the targets and the casualty models used. We have estimated the casualties from four different types of attacks: three involving approximately 100 targets each and the fourth a major counterforce attack on U.S. strategic-nuclear facilities.

Casualties from "100-Megaton" Attacks

Blast and Burn Casualty Models

The primary basis for models of the blast and burn effects of nuclear explosions is the casualty data from Hiroshima. The nuclear weapon used on Hiroshima, however, had a yield of only 15 kilotons (kt) (Loewe and Mendelsohn, 1982)—much less than most warheads in the current strategic arsenals of the superpowers. Therefore, casualty models must contain rules for extrapolating the number of casualties at Hiroshima to those caused by explosions of higher yields.

Overpressure Model

The standard method for extrapolation that is, to our knowledge, used in virtually all government calculations is to assume that casualty probabilities are a function of peak blast overpressure. Given the weapon yield and height of burst, the peak overpressure is calculated as a function of the distance from ground zero, and the Hiroshima blast and burn casualty rates for that overpressure are applied to the population at that distance (e.g., U.S. Congress, Office of Technology Assessment, 1979; p. 19). Figure 1 shows the casualties at Hiroshima as a function of distance from ground zero. Figure 2 shows the same data replotted as a function of the peak blast overpressure.

Figure 1. Mortality and casualty rates at Hiroshima.

Figure 1

Mortality and casualty rates at Hiroshima. The percent probability of total casualties (•) and fatalities (•) is shown as a function of distance from the hypocenter. (From Oughterson and Warren, 1956, Figure 3,10.)

Figure 2. Hiroshima casualty rates as a function of peak blast overpressure.

Figure 2

Hiroshima casualty rates as a function of peak blast overpressure.

For obvious reasons, we call the standard casualty model the "over-pressure" model. The use of blast overpressure as the explanatory variable does not mean that burns are ignored. At Hiroshima, the probability of blast injuries and burn injuries fell off with distance from ground zero in about the same manner (Oughterson and Warren, 1956; p. 43), and the cause of death was not generally known. Under these circumstances, it was natural to choose overpressure as a basis for scaling—especially since the distance corresponding to a given level of overpressure can easily be calculated, given the weapon's yield and height-of-burst.

Postol has recently challenged this "overpressure casualty model." He points out that the fires simultaneously ignited by a megaton-sized explosion over an urban area would merge into a "superfire" of such large extent and intensity, with asphyxiating gases and gale-force winds, that many more casualties would result than would be predicted by the over-pressure model (Postol, 1986). Postol notes that even those not fatally injured or trapped under collapsed buildings would have insufficient time to escape from the inner regions of superfires before the fire effects became lethal. Therefore, he argues that for higher-yield explosions, casualty rates would depend primarily on the area of the subsequent conflagration.

At Hiroshima, a mass fire, or conflagration, developed approximately 20 minutes after the explosion and ultimately consumed essentially all combustible materials in an area with a radius of about 2 km (see Figure 3). The growth of the area of the conflagration with weapon yield is difficult to predict because a nuclear explosion would cause fires through two mechanisms: (1) direct ignition by heat radiated from the fireball and (2) indirect ignition by blast-caused electrical short circuits, gas line breaks, ruptured fuel tanks, and other sources.

Figure 3. Map of Hiroshima damage areas.

Figure 3

Map of Hiroshima damage areas. (From Committee for the Compilation of Materials on Damage Caused by the Atomic Bomb in Hiroshima and Nagasaki, 1981; pp. 58-59. Reprinted with permission from Basic Books, Inc.)

If the radius of the conflagration area were scaled with peak overpressure, then the Hiroshima conflagration area would scale up to have a radius of 8 km for a 1-Mt airburst. There are at least two reasons why the conflagration radius could grow more rapidly than this: (1) the blast effects at a given peak overpressure could be much more damaging at higher weapon yields because the associated winds would last much longer (Wilton et al., 1981); and (2) given a relatively clear atmosphere, the radius of incendiary effects by direct ignition might reach out well beyond 8 km.

Brode and Small (1983; Figure 27 therein) have estimated that the conflagration caused by a 1-Mt airburst over an urban area could have a radius anywhere from 4 to 14 km, depending on the atmospheric conditions and the types of buildings involved. The lower end of the range is not relevant to considerations of conflagrations in ordinary urban areas, since it is associated with extremely blast-resistant, reinforced concrete buildings, while the upper end involves blast-caused fires in building types that are quite common in U.S. cities. Therefore, we have considered conflagrations with radii ranging from a minimum of 8 km to a maximum of 15 km, i.e., from the radius that would be predicted if the conflagration radius occurred at a fixed peak overpressure to a radius almost twice as large. Our medium-radius conflagration model has a fire radius of 12 km.

Given this range of conflagration radii, we have constructed a conflagration casualty model by dividing the distance from ground zero into three zones (see Figure 4):

Figure 4. Conflagration model.

Figure 4

Conflagration model.

  • An inner conflagration zone—the area further in than 2 km from the edge of the conflagration zone. Here we assume that there would be 100 percent fatalities because the population would not have time to escape before individual fires would merge into a single inferno.
  • An outer conflagration zone—the outer 2 km of the conflagration zone. Here we assume that there would be 50 percent fatalities and 33 percent severe injuries. These are approximately the average values observed within the 2-km radius Hiroshima burnout zone.
  • An overpressure injury zone—the area outside of the conflagration zone, where there would be blast effects and scattered fires. Here we assume that the fatality and injury rates are the same functions of peak blast overpressure as were observed outside the conflagration zone at Hiroshima.

In Figures 5a and 5b, we compare the probabilities of death and injuries as a function of distance from ground zero predicted by the overpressure and conflagration models for a 1-Mt airburst at a 2-km altitude. A midrange conflagration radius of 12 km has been assumed.

Figure 5. A: Fatality rates from a 1-Mt explosion at 2-km altitude: predictions of two casualty models.

Figure 5

A: Fatality rates from a 1-Mt explosion at 2-km altitude: predictions of two casualty models. B: Injury rates from 1-Mt explosion at 2-km altitude: predictions of two casualty models.

Ranges of Casualties Calculated for 100-Mt Attacks on U.S. City Centers, Military-Supporting Industry, or Strategic Nuclear Targets

The calculated results of an all-out attack on the U.S. population or economic targets involving thousands of megatons would be relatively insensitive to the casualty model used. The degree of overkill would be so high that, regardless of the casualty model assumed, the calculations would find that virtually the entire U.S. urban population would be killed by blast and bums. Much of the rural population would die of fallout-caused radiation illness, and most of the remainder would die of starvation and disease (e.g., Haaland et al., 1976; Harwell, 1984).

Therefore, in order to explore the sensitivity of blast and bum casualty estimates to the choice of casualty model and types of targets involved, we have considered much more limited hypothetical attacks on three different classes of targets in the United States—each containing approximately 100 ground zeros:

  • The city centers of the 100 largest U.S. urban areas;
  • 101 final-assembly factories selected by a Department of Defense contractor as the highest-priority targets for an attack on U.S. military-industrial capability;
  • 99 key strategic nuclear targets (34 home bases for nuclear bombers and their associated refueling aircraft; 16 nuclear naval bases; 9 nuclear weapon storage facilities; and 40 command, communication, and early-warning sites).

We have assumed 1-Mt airbursts for all these attacks. This provides a common basis for comparison and is certainly well within Soviet capabilities. More than 1,000 Soviet intercontinental and submarine-based ballistic missiles carry single warheads with estimated yields of approximately 1 Mt (Arkin, Burrows, et al., 1985).

Our data base for the U.S. population distribution was obtained from a tape prepared for the government from 1980 U.S. census data (Federal Emergency Management Agency, 1980). Since the data are for the residential population, our casualty estimates are for nighttime attacks. However, the results should be indicative for daytime attacks as well.

Figure 6a shows the cumulative populations around the ground zeros of the three hypothetical target sets. It will be seen that the populations around the military-industrial sites are almost as high as those around the city centers. This is because most of the military-industrial targets are located in major urban areas, including those around Boston, Detroit, Los Angeles, Minneapolis-St. Paul, Philadelphia, Phoenix, Rochester, Sacramento, St. Louis, San Diego, San Jose, Seattle, Tucson, and Wichita (Science Applications Inc., 1984).

Figure 6. A: Cumulative population around ground zeros for 100-Mt attacks.

Figure 6

A: Cumulative population around ground zeros for 100-Mt attacks. B: Cumulative population around 99 U.S. strategic nuclear targets.

The cumulative population around the 99 strategic nuclear targets is considerably lower than that around the city centers or military-industrial sites but still includes tens of millions of people (see the breakdown by classes of target in Figure 6b). Many of these targets are in urban areas (Table 2). For example,

Table 2. Some Major U.S. Urban Areas with Nearby Strategic Nuclear Facilities.

Table 2

Some Major U.S. Urban Areas with Nearby Strategic Nuclear Facilities.

  • Strategic nuclear bombers or their associated aerial refueling groups are based outside Chicago, Milwaukee, Phoenix, Salt Lake City, Sacramento, Wichita, Fort Worth, and Shreveport (Air Force Magazine, 1985).
  • Aircraft carriers carrying nuclear-armed fighter bombers are based in San Francisco Bay; and battleships equipped with long-range, nuclear-armed, land-attack cruise missiles are proposed for bases off both Long Beach, Calif., and Staten Island in New York harbor.
  • The Pentagon and the White House in the Washington, D.C., urban area and the Strategic Air Command headquarters outside Omaha are among the most important strategic nuclear weapons command posts. And a number of other major urban areas are the sites of key radio transmitters for communicating with ballistic missile submarines, bombers, and military early-warning and communications satellites (Arkin and Fieldhouse, 1985).

The numbers of blast and burn deaths and injuries predicted for attacks involving 1-Mt airbursts over each of these targets are listed in Table 1. Figure 7 shows the results from the overpressure and medium-radius conflagration models. The conflagration model predicts 1.5 to 4.1 times as many deaths as the overpressure model—up to 56 million—depending on the conflagration radius that is assumed.

Figure 7. Casualties for three target sets, 100-Mt attacks.

Figure 7

Casualties for three target sets, 100-Mt attacks.

Figure 7 also shows that the targeting of factories instead of city centers does not greatly reduce the number of deaths, and that there could still be more than 10 million fatalities even if the 100 warheads were targeted solely against strategic military targets. So, if Soviet targeteers were to declare that they did not target cities per se—as their American counterparts have done (Ball, 1983; p. 19)—that statement should not give us a great deal of comfort. Our finding that tens of millions of people might die from just 100 warheads also shows that any strategic defense system would have to reduce an attack to well below this level before the results would be less than catastrophic.

Casualties From A Major Attack On U.S. Strategic Nuclear Targets

U.S. strategic-nuclear capabilities would be the highest-priority targets for any Soviet attack because they represent by far the greatest threat posed by the United States to the Soviet Union. ''Counterforce" attacks, restricted to military bases containing strategic nuclear weapon delivery vehicles and to their command, control, and warning systems, have therefore become a staple of debates over nuclear weapons policy.

During 1974-1975, a series of hearings held by a subcommittee of the U.S. Senate Foreign Relations Committee focused on the potential U.S. casualties that might result from such counterforce attacks on the United States. Even in 1985, a decade later, the record of these hearings represents the fullest open discussion thus far of the U.S. government's views on this subject. We therefore briefly summarize these hearings here.

James R. Schlesinger, then Secretary of Defense, provoked the hearings with a section in the DOD report to Congress on the proposed budget for fiscal year 1975. Schlesinger argued in that report that the United States required additional nuclear options to be able to respond in case of a "limited attack on military targets that caused relatively few civilian casualties." When asked in the first hearing on March 4, 1974, what he meant by "relatively few casualties," Schlesinger responded: "I am talking here about casualties of 15,000, 20,000, 25,000 . . ." (U. S. Congress, Senate Foreign Relations Committee, March, 1974; p. 26). However, he never explained where these low numbers came from.

Schlesinger was recalled by the subcommittee on September 11, 1974 for a "Briefing on Counterforce Attacks." In this hearing, he presented the results of DOD calculations of hypothetical Soviet attacks on U.S. missile silos and bomber bases. Although other numbers were mentioned, he summarized the results as follows: "In an attack on the ICBMs alone, the mortalities would run on the order of a million and for SAC [bomber] bases the mortalities would be less than that—on the order of 500,000" (U.S. Congress, Senate Foreign Relations Committee, September, 1974; p. 12). He pointed out that these numbers were much smaller than the 95-100 million U.S. deaths that he believed would result from an all-out Soviet nuclear attack on the United States.

In response to a request from the Senate Foreign Relations Committee, the Congressional Office of Technology Assessment organized a panel of experts, chaired by Jerome Wiesner, then President of the Massachusetts Institute of Technology, to review the DOD casualty estimates. The panel questioned the credibility of the assumed attacks since they "were evidently not designed to maximize destruction of U.S. ICBMs and bombers. . . ." The panel also questioned the DOD assumption that "the urban population is familiar with the shelter areas available [and] that these shelters are all stocked with adequate supplies of essential material for sustaining inhabitants for 30 days." The panel suggested that the DOD analysts do additional calculations to determine the sensitivity of their results to changes in a number of the challenged assumptions. The most significant changes suggested were to assume groundbursts for the weapons attacking the missile silos, to use "pattern attacks" on the bomber bases, and to adopt a more realistic fallout shelter posture (U.S. Congress, Office of Technology Assessment, 1975; pp. 4-9).

The DOD submitted a written response on July 11, 1975. The number of deaths estimated in this report for "comprehensive attacks" on U.S. ICBM, bomber, and ballistic-missile submarine bases believed to be within Soviet capabilities ranged from 3.2 million to 16.3 million (U.S. Congress, Senate Foreign Relations Committee, 1975; p. 14). In the 10 years since this set of hearings, no other Congressional Committee has seen fit to pursue the subject further.

Description Of The Attack

Table 3 summarizes our hypothetical attack on 1,215 U.S. strategic-nuclear targets. The attack involves almost 3,000 warheads with a total yield of about 1,340 Mt. This represents perhaps one-third of the warheads and one-quarter of the megatonnage carried by Soviet strategic delivery vehicles (not including reloads) (Arkin, Burrows, et al., 1985).

Table 3. Targets for Full Attack on U.S. Strategic Nuclear Forces.

Table 3

Targets for Full Attack on U.S. Strategic Nuclear Forces.

We have made the usual assumption that each of the 1,116 U.S. missile silos and missile launch-control centers would be struck by two 0.5-Mt warheads—the first, a low-altitude airburst, and the second, a groundburst (e.g., U.S. Congress, Senate Foreign Relations Committee, 1975; p. 7). About 3,000 such warheads are carried by Soviet SS-18 intercontinental ballistic missiles (e.g., Arkin, Burrows, et al., 1985). We have assumed that the low-altitude airburst would produce one-half as much local fallout as the groundburst.

In the case of the attacks on the 34 U.S. strategic bomber and aerial refueling tanker bases, we have assumed "pattern attacks" with a number of nuclear warheads exploding in the air around the base in an effort to destroy any aircraft that have just taken off (Quanbeck and Wood, 1976; U.S. Congress, Senate Foreign Relations Committee, 1975; p. 23). We have assumed that each air base would be attacked by a 1-Mt groundburst, to destroy the runway and stored nuclear weapons, and by a pattern attack of 14 0.2-Mt airbursts. This attack would require only 102 of the Soviets' 966 submarine-based ballistic missiles (Arkin, Burrows, et al., 1985). We estimate that everyone within 13-17 km of the center of an air base under such an attack would be killed and those out to a radius of 18-19 km would be injured.

We have not included in our target set the many bases to which U.S. bombers and aerial refueling tankers would be dispersed during a crisis. During the Cuban missile crisis, U.S. strategic bombers were dispersed to 40 civilian airports—many of them near major cities (Allison, 1971; p. 139). Inclusion of such dispersal bases on our target list would have greatly increased the number of deaths calculated for this attack.

In the case of the attacks on the 16 nuclear navy bases, we have assumed that, because of their importance, they would be targeted with two warheads each. Since submarines are quite hard targets (U.S. Defense Intelligence Agency, 1969) and some of the port facilities might be also, we have assumed a 1-Mt airburst and a 1-Mt groundburst per port. Since the nuclear weapons at the nine major nuclear weapon storage depots are stored in hardened bunkers, we have assumed that they would be subjected to the same type of attack.

The highest-priority targets for a Soviet attack on U.S. strategic nuclear forces would be their early-warning radars, their command posts, the communications links between the command posts, and the ballistic missile submarines and bombers. According to Blair (1985; p. 182), the United States has about 400 primary and secondary command communication and control targets. Of these, 100 are the missile silo launch control centers and 10 are Minuteman missiles in silos at the Whiteman missile field that are equipped with emergency radio transmitters that would be fired aloft if all other forms of communication between the command posts and the missile fields and bombers failed (Arkin and Fieldhouse, 1985). These 110 targets have already been included in the countermissile silo attack component described above. Still other targets are located at bomber bases that have also been included in the bomber targets described above. We have therefore included in our attack on U.S. strategic nuclear forces only 40 additional key command, communication, and early-warning targets:

  • Seven major headquarter and alternate headquarter installations;
  • Five early-warning radar installations;
  • Ten key naval radio transmitters for communicating to submarines;
  • Nine key Strategic Air Command (SAC) transmitters for communicating to bombers;
  • Nine key terminals for communication with and control of satellites.

We assume that these 40 targets would be attacked with 1-Mt airbursts with seven exceptions:

  • The White House, the Pentagon, SAC headquarters at Offut Air Base, Nebraska; Cheyenne Mountain, Colorado; and the Alternate National Military Command Center near Fort Ritchie, Md. All of these are high-priority targets, and at least three have underground command posts.
  • SAC's two survivable low-frequency transmitters. It is assumed that each of these seven targets would be attacked by a 1-Mt groundburst as well as by a 1-Mt airburst.

Fallout Casualty Model

Nuclear explosions create a great deal of short-lived radioactivity—mostly associated with fission products. We have made the standard assumption in our calculations that one-half of the yield from the attacking weapons would be from fission. In the case of airbursts, the fireball would carry this radioactivity into the upper atmosphere, from which it would slowly filter down as a rather diffuse distribution called "global fallout" over a period of months to years. In the case of an attack on so-called "hard" targets such as missile silos, which can withstand high overpressures, the nuclear weapons would have to be exploded so close to the ground that surface material would be sucked into the fireball, mixed with the vaporized bomb products, and carried by the buoyancy of the fireball into the upper atmosphere. There, much of the bomb material and surface material would condense into particles, a large fraction of which would descend to the surface again within 24 hours in an intense swath of "local fallout" downwind from the target.

Various models have been developed to describe the effect of winds in distributing this early fallout. Our calculations were done using the WSEG-10 model developed by the Weapons System Evaluation Group of the Department of Defense (Schmidt, 1975).

Our wind data base, which also comes from the Department of Defense, contains the wind speed and direction at five different altitudes on a 2-degree latitude-longitude grid for the entire Northern Hemisphere for a "typical day" of each month of the year (U.S. Defense Communications Agency, 1981).

Radiation Protection Factors

The WSEG-10 model predicts the doses that would be received by a population standing fully exposed on a perfectly flat surface. These doses must be reduced by dividing by the protection factors that account for the shielding effects of the shelters in which the population would take refuge. Wooden and brick residences have average protection factors of about 2.5 and 5, respectively (U.N. Scientific Committee on the Effects of Atomic Radiation, 1982; p. 62), and below-ground residential basements typically offer protection factors of 10-20 (Glasstone and Dolan, 1977; p. 441). We have assumed in our calculations that one-half of the population would be in shelters with effective protection factors of 3, and one-half would be in shelters with protection factors of 10.

Protection factors of more than 100 are often discussed by those who argue that the USSR (or the United States) could effectively shelter its populations from radioactive fallout in improvised shelters (e.g., Nitze, 1979; p. S10080). Such protection factors are unrealistic, however, because they assume implicitly that it would be possible for the sheltered population to stay in shelters without interruption for several weeks. Even staying inside a perfect radiation shield for the first 2 days and coming out thereafter for only 2 hours a day would result in a decrease in the shelter's effective protection factor from infinity to about 20. Additional radiation doses would be absorbed because, within a relatively short time, most of the population would be drinking water and eating food contaminated with radioactivity.

Population Radiation Sensitivity

The WSEG-10 model takes into account the well-known fact that mammals can survive larger cumulative doses of radiation if these doses are delivered over a longer period of time. This is done by assuming that 90 percent of the radiation damage is repairable and that this repair proceeds exponentially at a rate of 3.3 percent of the remaining unrepaired damage per day (30-day average repair time). Making this assumption, we can calculate the total unrepaired radiation dose as a function of time. At some time, as the intensity of the fallout radiation declines, the rate of biological repair will exceed that of accumulation of new damage and the total radiation damage will peak. The level of this "peak equivalent residual radiation dose" determines the probability of death. If no biological repair were assumed, the total absorbed dose out to 6 months would be 30-60 percent higher than the peak residual dose calculated by the WSEG-10 model, depending on the fallout arrival time.

Since the end of World War II, the standard assumption used in official U.S. estimates of deaths from radiation has been that, in the absence of intensive medical treatment, a 450-rad peak residual dose would cause a 50 percent fatality rate from radiation sickness within 60 days (LD50 = 450 rads) (Lushbaugh, 1982; pp. 46-57). However, this number is near the top of the current range of uncertainty (Lushbaugh, 1982). In 1960, Cronkite and Bond estimated that, in the absence of treatment with antibiotics and blood transfusions, the LD50 would be 350 rads. Their estimates were based principally on a comparison of the hematological effects of radiation on a group of Marshall Islanders, after the fallout from a U.S. nuclear test in 1954 accidentally exposed them to 175 rads, with similar hematological effects in dogs subject to radiation exposures in the laboratory. Cronkite and Bond assumed that the lethality curve for humans would be parallel to that of the dogs and estimated that human fatalities would begin at about 200 rads (Cronkite and Bond, 1958; p. 249).

Recently Rotblat analyzed the recalculated doses from Hiroshima and estimated an LD50 for Hiroshima of 220 rads (Rotblat, 1986). This is much lower than that estimated for the Marshallese by Cronkite and Bond. Rotblat's lower value for the people of Hiroshima may reflect synergistic effects of radiation dosage with the traumatic effects of the explosion and its aftermath. The population surviving the immediate effects of a large-scale nuclear attack on the United States would certainly also find itself under multiple stresses, including emotional shock, hunger, unsanitary conditions, and possibly exposure to cold weather. We have therefore estimated fallout deaths for an LD50 of 250 rads as well as for 350 and 450 rads. We approximated the associated lethality curves as straight lines parallel to the Cronkite-Bond curve (see Figure 8). Numbers of cases of severe radiation illness were estimated by assuming that everyone would get seriously ill from radiation sickness at radiation doses where anyone died.

Figure 8. Three possible population sensitivities to ionizing radiation.

Figure 8

Three possible population sensitivities to ionizing radiation.


In addition to the short-term radiation illnesses and deaths from fallout, there would also be a large number of radiation-caused cancers in the longer term. In calculating these cancers, we use the "linear hypothesis" that the probability of developing a radiation-caused cancer is proportional to the radiation dose. According to the linear dose-effect model in the 1980 report of the National Academy of Sciences' Committee on the Biological Effects of Ionizing Radiation, a radiation dose of 109 person-rads would result in 0.4-1.25 million extra cancers in the general population, of which 0.17-0.5 million would be fatal. (National Academy of Sciences, Committee on the Biological Effects of Ionizing Radiation, 1980.)

The cancer dose is calculated assuming no biological repair. This means that the population cancer dose would continue to increase even after radiation levels had fallen to the point where the population sheltered below ground felt that they could emerge without risk of short-term radiation illness. We assume that they might emerge at 2 weeks and that the average protection factor would be 3 thereafter for the entire population.

Ranges of Casualties Calculated for Attack on U.S. Strategic Nuclear Targets

We have calculated the fallout patterns for the attack on U.S. strategic nuclear forces described in Table 3, assuming typical February, May, August, and October winds. Figure 9 shows, as an example, the calculated pattern for typical February winds.

Figure 9. Fallout pattern in a February attack on U.

Figure 9

Fallout pattern in a February attack on U.S. strategic nuclear targets.

The large fallout patterns originate at the six Minuteman-MX missile fields, each of which would absorb the equivalent of hundreds of 0.5-Mt groundbursts. The long fallout pattern originating in mid-California is due to an attack on the 16 missile silos at Vandenburg AFB. The other smaller fallout patterns are each associated with a 1-Mt groundburst on a total of 66 bomber, aerial tanker, nuclear-navy, nuclear-weapons storage, and command and communications facilities. Three contours for the unshielded, peak residual air-dose are shown:

  • 3,500 rads. Within this region, given an LD50 of 350 rads and our assumed sheltering posture (half the population with a protection factor of 10 and half with a protection factor of 3), more than three-quarters of the population would die.
  • 1,050 rads. Within this region, given an LD50 of 350 rads, more than half of that part of the population with an effective protection factor of 3 (i.e., one-quarter of the total population) would die.
  • 300 rads. Outside this region, given a protection factor of 3 or more, few deaths would occur from radiation illness.

To get an adequate idea of the sensitivity of our results to the monthly winds, we have also done casualty calculations for May, August, and October winds.

Table 4 shows our estimated ranges of deaths and total casualties due to blast, fire, radiation sickness, and cancer. Shown explicitly is the sensitivity of these results to the choice of blast-bum casualty model, radiation sickness LD50, and cancer-risk coefficient. The ranges in each subcategory show the variation associated with the choice of winds. The range shown for the total deaths reflects the combined effect of the variability associated with the winds, the LD50s, and the cancer dose-risk coefficients. Table 5 shows ranges of deaths from blast and bum and from fallout for the components of the attack. The summed total of the upper ends of the ranges of casualties calculated for these component attacks exceeds the maxima given for the total attack in Table 4. In part, this is because some of the same people would be killed by different subattacks and in part because the winds that maximize the fallout casualties from the counter-silo attack tend to minimize the fallout casualties from the other component attacks.

Table 4. Deaths and Total Casualties from Large-Scale Attack on U.S. Strategic-Nuclear Targets (in millions).

Table 4

Deaths and Total Casualties from Large-Scale Attack on U.S. Strategic-Nuclear Targets (in millions).

Table 5. Deaths and Total Casualties from Subcomponents of the Large-Scale Attack on U.S. Strategic-Nuclear Targets (in millions).

Table 5

Deaths and Total Casualties from Subcomponents of the Large-Scale Attack on U.S. Strategic-Nuclear Targets (in millions).

Overall, taking into account the different estimates associated with the two blast-bum casualty models, we find that there would be 13-34 million deaths and 25-64 million total casualties from this "counterforce" attack. Figure 10 shows how these estimates vary with low and high assumptions about casualty rates and with different months of the year.

Figure 10. Major attack on counterforce targets.

Figure 10

Major attack on counterforce targets. The bars labeled ''1" correspond to assumptions that would cause the fewest deaths; bars labeled "2" correspond to the assumptions causing the most deaths.

How do we explain the fact that the lower end of our range of estimated deaths is approximately equal to the upper end of the range of 3.2-16.3 million deaths calculated by the DOD in 1975 for a major counterforce attack on the United States (U.S. Congress, Senate Foreign Relations Committee, 1975; pp. 12-24)?

The low end of the DOD's range of deaths is apparently associated with the optimistic sheltering posture that was criticized by the Office of Technology Assessment review panel. Using a less optimistic sheltering posture, somewhat like that used in this paper, and assuming a 550-kiloton airburst and groundburst over each silo, the DOD analysts estimated 5.6 'million deaths resulting from an attack on U.S. missile silos alone (assuming March winds). This is quite close to the 4.9 million deaths that we find for February winds if we, like the DOD analysts, assume an LD50 of 450 rads and neglect cancer deaths.

The upper end of the DOD range of estimated deaths is associated with a sheltering posture similar to that we have used. However, the DOD analysts made a number of other assumptions that correspond roughly to those that characterize the lower end of our uncertainty range:

  • The overpressure model was used to calculate blast and burn casualties.
  • An LD50 of 450 rads was used for calculating the number of deaths from radiation sickness.
  • Cancer deaths were ignored.

The DOD analysts did assume March winds for their calculations—an assumption that would tend to maximize the fallout casualties from the attacks on the missile silos. This assumption was offset, however, by their omission of most of the targets in three of the five subsets of targets considered in our attack: strategic command, communication, and early-warning facilities; nuclear storage sites; and the ports for naval ships that have recently been assigned a strategic role with the deployment of long-range, sea-launched cruise missiles. As may be seen from Table 5, these target sets account for a large fraction of the total number of deaths calculated for our counterforce attack.

As has already been noted, our own range of 13-34 million deaths and 25-64 million total casualties would be higher if we included other likely targets, such as potential bomber dispersal bases. Our casualty numbers would climb higher still if we added estimates of the numbers of deaths and illnesses from the economic and social collapse that could be expected following such an attack.


Some of our results show clearly the enormous casualties that only 1 percent of the current Soviet strategic arsenal could inflict on the United States, even if the targets were military-industrial or strategic rather than population per se. We have also found that, as counterforce attacks become more comprehensive, the distinction (in terms of casualties) between "counterforce" and "counterpopulation" targeting becomes increasingly blurred. We expect to find similar results for U.S. counterforce attacks on the USSR.

These casualty estimates have a critical beating on the debate over the possibility of "limited nuclear options" as part of a strategic doctrine. Either superpower contemplating such an attack should be well aware of the fact that such attacks—even if limited to military targets—could cause casualties that approach those from all-out attacks. We emphasize this point especially because the significant underestimates in the published DOD casualty estimates for counterforce attacks suggest that U.S. policy concerning counterforce attacks is not fully realistic. Certainly this appeared to be the case during the recent debate of the "window of vulnerability" of U.S. ICBMs. Virtually no attention was given to the casualties that would result from an attack on U.S. missile silos.

Other work has shown that, even after such a devastating attack, either superpower would retain a residual capability to destroy the cities and economic infrastructure of the other many times over (e.g., Feiveson and von Hippel, 1983). And it seems likely that the other superpower—after suffering tens of millions of deaths—would launch at least as horrendous an attack in response. Therefore, the only conceivable rationale that theoretically could be used to justify a strategic counterforce attack would be the certainty that the other side had already committed itself to a major nuclear attack—a certainty that could not be achieved in the real world.

It is our hope that national decision makers will develop a better understanding of the "collateral" consequences of hypothetical first strikes and of the enormous destructive capacity of the weapons that would survive. That understanding should make them less likely to seek counterforce capabilities or to fear such attacks from the other side.


  • Air Force Magazine. May 1985. Guide to USAF Bases at Home and Abroad, pp.170-181.
  • Allison, G. T. 1971. Essence of Decision: Explaining the Cuban Missile Crisis. Boston: Little, Brown.
  • Arkin, W. M., A. A. Burrows, R. W. Fieldhouse, T. B. Cochran, R. S. Norris, and J. I. Sands. 1985. Nuclear weapons. Pp. 41-74 in World Armaments and Disarmament: SIPRI Yearbook 1985. London and Philadelphia: Taylor and Francis.
  • Arkin, W. M., and R. W. Fieldhouse, 1985. Nuclear Battlefields: Global Links in the Arms Race. Appendix A. Cambridge, Mass.: Ballinger.
  • Ball, D. 1983. Targeting for Strategic Deterrence. Adelphi Paper no. 185. London: International Institute for Strategic Studies.
  • Bennett, B. 1977. Fatality Uncertainties in Limited Nuclear War. Report no. R-2218-AF. Santa Monica, Calif.: The Rand Corporation.
  • Blair, B. G. 1985. Strategic Command and Control. Washington, D.C.: The Brookings Institution.
  • Brode, H. L., and R. D. Small. 1983. Fire Damage and Strategic Targeting. Note 567. Los Angeles, Calif.: Pacific-Sierra Research Corp.
  • Committee for the Compilation of Materials on Damage Caused by the Atomic Bomb in Hiroshima and Nagasaki. 1981. Hiroshima and Nagasaki: The Physical, Medical, and Social Effects of the Atomic Bombings. New York: Basic Books.
  • Cronkite, E. P., and V. P. Bond. 1958. Acute Radiation Syndrome in Man. U.S. Armed Forces Med. J. 9:313-324. [PubMed: 13519610]
  • Duffield, J., and F. von Hippel. 1984. The short-term consequences of nuclear war for civilians. Pp. 19-64 in The Environmental Effects of Nuclear War, J. London, editor; and G. F. White, editor. , eds. Boulder, Colo.: Westview Press.
  • Federal Emergency Management Agency. 1980. Population Grid File Tape based on 1980 Census Data (1 minute grid). Washington, D.C.: Federal Emergency Management Agency.
  • Feiveson, H., and F. von Hippel. 1983. The freeze and the counterforce race. Physics Today, January, p. 37.
  • Glasstone, S., and P. J. Dolan. 1977. The Effects of Nuclear Weapons. Washington D.C.: U.S. Departments of Defense and Energy.
  • Greer, D. S., and L. S. Rifkin. 1986. The Immunological Impact of Nuclear Warfare. This volume.
  • Haaland, C. M., C. V. Chester, and E. P. Wigner. 1976. Survival of the Relocated Population of the U.S. after a Nuclear Attack. Report no. ORNL-5041. Oak Ridge, Tenn.: Oak Ridge National Laboratory.
  • Harwell, M. A. 1984. Nuclear Winter: The Human and Environmental Consequences of Nuclear War. New York: Springer-Verlag.
  • Loewe, W. E., and E. Mendelsohn. 1982. Neutron and gamma doses at Hiroshima and Nagasaki. Nuclear Science and Engineering 81:325. See also a preliminary Los Alamos reestimate of the yield of the Hiroshima bomb as 14-16 kt. quoted in W.J. Broad, 1985.
  • Lushbaugh, C. C. 1982. The impact of estimates of human radiation tolerance upon radiation emergency management. Pp. 46-57 in The Control of Exposure of the Public to Ionizing Radiation in the Event of Accident or Attack. Bethesda, Md.: National Council on Radiation Protection and Measurement.
  • National Academy of Sciences, Committee on the Biological Effects of Ionizing Radiation. 1980. The Effects on Populations of Exposure to Low Levels of Ionizing Radiation. Washington, D.C.: National Academy Press.
  • Nitze, P. H. 1979. A Method for Dealing With Certain Fallout Questions. Attachment to a Prepared Statement for Presentation before the Senate Foreign Relations Committee. P. S10080. Congressional Record, July 20.
  • Oughterson, A. W., and S. Warren. 1956. Medical Effects of the Atomic Bomb in Japan. National Nuclear Energy Series. Atomic Energy Commission. New York: McGraw-Hill.
  • Postol, T. 1986. Possible Fatalities from Superfires following Nuclear Attacks in or Near Urban Areas. This volume.
  • Quanbeck, A. H., and A. L. Wood. 1976. Modernizing the Strategic Bomber Force: Why and How. Washington, D.C.: The Brookings Institution.
  • Rotblat, J. 1986. Acute Radiation Mortality in a Nuclear War. This volume.
  • Schmidt, L. A., Jr. 1975. Methodology of Fallout-Risk Assessment. Paper P-1065. Arlington, Va.: Institute for Defense Analyses.
  • Science Applications Inc. 1984. Assessment of Potential Military-Industrial Targets in CONUS (Continental United States) for Soviet Nuclear Attack.
  • U.N. Scientific Committee on the Effects of Atomic Radiation. 1982. Ionizing Radiation: Sources and Biological Effects. New York: United Nations.
  • U.S. Congress, Office of Technology Assessment. 1975. Response of the Ad Hoc Panel on Nuclear Effects, in U.S. Congress, Senate Foreign Relations Committee, Analyses of Effects of Limited Nuclear Warfare. Washington, D.C.: U.S. Government Printing Office.
  • U.S. Congress, Office of Technology Assessment. 1979. The Effects of Nuclear War. Washington, D.C.: U.S. Government Printing Office.
  • U.S. Congress, Senate Foreign Relations Committee. March 1974. J. R. Schlesinger in U.S.-U.S.S.R. Strategic Policies. Hearing before the Subcommittee on Arms Control, International Law and Organization of the U.S. Senate Committee on Foreign Relations, March 4, 1974. Washington, D.C.: U.S. Government Printing Office, 1974.
  • U.S. Congress, Senate Foreign Relations Committee. September 1974. J. R. Schlesinger in Briefing on Counterforce Attacks. Hearing before the Subcommittee on Arms Control, International Law and Organization of the US Senate Committee on Foreign Relations, September 11, 1974. Washington, D.C.: U.S. Government Printing Office, 1974.
  • U.S. Congress, Senate Foreign Relations Committee. 1975. Sensitivity of collateral damage calculations to limited nuclear war scenarios. In Analyses of Effects of Limited Nuclear Warfare. Washington, D.C.: U.S. Government Printing Office.
  • U.S. Defense Communications Agency. 1981. Unclassified tapes EA 275 and EB 275. Washington, D.C.: U.S. Defense Communications Agency.
  • U.S. Defense Intelligence Agency. 1969. Physical Vulnerability Handbook: Nuclear Weapons (AP-550-1-2-69-INT). Washington, D.C.: U.S. Defense Intelligence Agency.
  • Wilton, W., D. J. Myronuk, and J. V. Zaccor. 1981. Secondary Fire Analysis. Report #8084-6. Redwood City, Calif.: Scientific Service Inc.


This paper is based on a much longer technical report that is available from Princeton University's Center for Energy and Environmental Studies as Report #PU/CEES 198.

Copyright © 1986 by the National Academy of Sciences.
Bookshelf ID: NBK219165


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (6.8M)
  • Disable Glossary Links

Related information