23 December 1983, Volume 222, Number 4630

Nuclear Winter: Global Consequenof Multple Nuclear Explosio

R. P. Turco, 0. B. Toon, T. P. AckenJ. B. Pollack, Carl Sa

Concern has been raised over theshort- and long-term consequences ofthe dust, smoke, radioactivity, and toxicvapors that would be generated by anuclear war (1-7). The discovery that

quantities of sooty smiattenuate sunlight andmate. These developmencalculate, using new datmodels, the potential glol

Summary. The potential global atmospheric and climatic consequfwar are investigated using models previously developed to studsvolcanic eruptions. Although the results are necessarily impreciserange of possible scenaros and uncertainty in physical paramiprobable first-order effects are serious. Significant hemispherical atsolar radiation flux and subfreezing land temperatures may be cautraised in high-yield nuclear surface bursts and by smoke from cityignited by airbursts of all yields. For many simulated exchanges of smegatons, in which dust and smoke are generated and encircle the e4weeks, average light levels can be reduced to a few percent of artemnperatures can reach -15° to -25°C. The yield threshold for mclimatic consequences may be very low: only about 100 megatonsmajor urban centers can create average hemispheric smoke opticathan 2 for weeks and, even in summer, subfreezing land temperaturea 5000-megaton war, at northern mid-latitude sites remote from tarcfallout on time scales of days to weeks can lead to chronic mean dcrads from external whole-body gamma-ray exposure, with a likely finternal dose from biologically active radionuclides. Large horizortemperature gradients caused by absorption of sunlight in smokemay greatly accelerate transport of particles and radioactivity froHemisphere to the Southern Hemisphere. When combined with thetion from nuclear blast, fires, and fallout and the later enhancement oiradiation due to ozone depletion, long-term exposure to cold, dark,;could pose a serious threat to human survivors and to other specie4

dense clouds of soil particles may haveplayed a major role in past mass extinc-tions of life on Earth (8-10) has encour-aged the reconsideration of nuclear wareffects. Also, Crutzen and Birks (7) re-cently suggested that massive fires ignit-ed by nuclear explosions could generate23 DECEMBER 1983

tal effects of dust and(henceforth referred to

SCIE:NCE

exchange and would inherit the postwarenvironment. Accordingly, the longer-term and global-scale aftereffects of nu-

clear war might prove to be as importantas the immediate consequences of thewar.

To study these phenomena, we used a

series of physical models: a nuclear warscenario model, a particle microphysics

model, and a radiative-convective mod-man el. The nuclear war scenario model spec-

ifies the altitude-dependent dust, smoke,tgan radioactivity, and NO. injections for

each explosion in a nuclear exchange(assuming the size, number, and type ofdetonations, including heights of burst,

oke that would geographic locales, and fission yieldperturb the cli- fractions). The source model parameter-its have led us to ization is discussed below and in a moreLa and improved detailed report (15). The one-dimension-bal environmen- al microphysical model (15-17) predicts

the temporal evolution of dust andsmoke clouds, which are taken to be

ences of nuclear rapidly and uniformly dispersed. They the effects of one-dimensional radiative-convectivedue to a wide model (1-D RCM) uses the calculated

eters, the most dust and smoke particle size distribu-ttenuation of the tions and optical constants and Mie the-sed by fine dust ory to calculate visible and infrared opti-and forest fires cal properties, li,ht fluxes, and air tem-;everal thousand peratures as a function of time andarth within 1 to 2 height. Because the calculated air tem-mbient and land peratures are sensitive to surface heatmajor optical and capacities, separate simulations are per-detonated over formed for land and ocean environ-

il depths greater ments, to define possible temperaturezs for months. In contrasts. The techniques used in our 1-gets, radioactive D RCM calculations are well docu-Dses of up to 50 mented (15, 18).equal or greater Although the models we used can pro-ital and vertical vide rough estimates of the average ef-and dust clouds fects of widespread dust and smokem the Northem clouds, they cannot accurately forecastprompt destruc- short-term or local effects. The applica-f solar ultraviolet bility of our results depends on the rateand radioactivity and extent of dispersion of the explosions. clouds and fire plumes. Soon after a

large nuclear exchange, thousands of in-dividual dust and smoke clouds would be

smoke clouds distributed throughout the northern mid-as nuclear dust latitudes and at altitudes up ta 30 km.

and nuclear smoke) generated in a nucle-ar war (11). We neglect the short-termeffects of blast, fire, and radiation (12-14). Most of the world's populationcould probably survive the initial nuclear

R. P. Turco is at R & D Associates, Marina delRey, California 90291; 0. B. Toon, T. P. Ackerman,and J. B. Pollack are at NASA Ames ResearchCenter, Moffett Field, California 94035; and CarlSagan is at Cornell University, Ithaca, New York14853.

1283

on

Janu

ary

24, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

Horizontal turbulent diffusion, verticalwind shear, and continuing smoke emis-sion could spread the clouds of nucleardebris over the entire zone, and tend tofill in any holes in the clouds, within 1 to2 weeks. Spatially averaged simulationsof this initial period of cloud spreadingmust be viewed with caution; effectswould be smaller at some locations andlarger at others, and would be highlyvariable with time at any given location.The present results also do not reflect

the strong coupling between atmosphericmotions on all length scales and themodified atmospheric solar and infraredheating and cooling rates computed withthe 1-D RCM. Global circulation pat-terns would almost certainly be alteredin response to the large disturbances inthe driving forces calculated here (19).Although the l-D RCM can predict onlyhorizontally, diurnally, and seasonallyaveraged conditions, it is capable of esti-mating the first-order climate responsesof the atmosphere, which is our intentionin this study.

Scenarios

A review of the world's nuclear arse-nals (20-24) shows that the primary stra-tegic anld theater weapons amount to- 12,000 megatons (MT) of yield carriedby - 17,000 warheads. These arsenalsare roughly equivalent in explosive pow-er to I million Hiroshima bombs. Al-

though the total number of high-yieldwarheads is declining with time, about7000 MT is still accounted for by war-heads of > I MT. There are also

30,000 lower-yield tactical warheadsand munitions which are ignored in thisanalysis. Scenarios for the possible useof nuclear weapons are complex andcontroversial. Historically, studies of thelong-term effects of nuclear war havefocused on a full-scale exchange in therange of 5000 to 10,000 MT (2, 12, 20).Such exchanges are possible, given thecurrent arsenals and the unpredictablenature of warfare, particularly nuclearwarfare, in which escalating massive ex-changes could occur (25).An outline of the scenarios adopted

here is presented in Table 1. Our base-line scenario assumes an exchange of5000 MT. Other cases span a range oftotal yield from 100 to 25,000 MT. Manyhigh-priority military and industrial as-sets are located near or within urbanzones (26). Accordingly, a modest frac-tion (15 to 30 percent) of the total yield isassigned to urban or industrial targets.Because of the large yields of strategicwarheads [generally -100 kilotons(KT)], "surgical" strikes against individ-ual targets are difficult; for instance, a100-KT airburst can level and burn anarea of - 50 kmi2, and a 1-MT airburst,

5 times that area (27, 28), implyingwidespread collateral damage in any"countervalue," and many "counter-force," detonations.

Table 1. Nuclear exchange scenarios.

Percent of yield

Ur- TotalTotal ban Warhead number

Case* yield Sur- or yield of(MT) face indus- range explo-bursts trial (MT) sionstar-

gets

1. Baseline exchange 5,000 57 20 0.1 to 10 10,4002. Low-yield airbursts 5,000 10 33 0.1 to I 22,500t9. 10,000-MTt maximum 10,000 63 15 0.1 to 10 16,160

10. 3,000-MT exchange 3,000 50 25 0.3 to 5 5,43311. 3,000-MT counterforce 3,000 70 0 I to 10 2,15012. 1,000-MT exchange§ 1,000 50 25 0.2 to 1 2,25013. 300-MT Southern 300 0 50 1. 300

Hemispherell14. 100-MT city attack¶ 100 0 100 0.1 1,00016. Silos, "severe" case# 5,000 100 0 5 to 10 70018. 25,000-MTi "future war" 25,000 72 10 0.1 to 10 28,300t*Case numbers correspond to a complete list given in (15). Detailed detonation inventories are notreproduced here. Except as noted, attacks are concentrated in the NH. Baseline dust and smoke parametersare described in Tables 2 and 3. tAssumes more extensive MIRVing of existing missiles and somepossible new deployment of medium- and long-range missiles (20-23). tAIthough these larger total yieldsmight imply involvement of the entire globe in the war, for ease of comparison hemispherically averagedresults are still considered. §Nominal area of wildfires is reduced from 5 x I0s to 5 x 104km2. llNominal area of wildfires is reduced from 5 x 105 to 5 x 103 km2. ¶The central city burden ofcomnbustibles is 20 g/cm2 (twice that in the baseline case) and the net fire smoke emission is 0.026 g per gramof material burned. There is a negligible contribution to the opacity from wildfires and dust. #Includes asixfold increase in the fine dust mass lofted per megaton of yield.

1284

The properties of nuclear dust andsmoke are critical to the present analy-sis. The basic parameterizations are de-scribed in Tables 2 and 3, respectively;details may be found in (15). For eachexplosion scenario, the fundamentalquantities that must be known to makeoptical and climate predictions are thetotal atmospheric injections of fine dust(s 10 p.m in radius) and soot.Nuclear explosions at or near the

ground can generate fine particles byseveral mechanisms (27): (i) ejection anddisaggregation of soil particles (29), (ii)vaporization and renucleation of earthand rock (30), and (iii) blowoff andsweepup of surface dust and smoke (31).Analyses of nuclear test data indicatethat roughly I x I0 to 6 x I10 tons ofdust per megaton of explosive yield areheld in the stabilized clouds of land sur-face detonations (32). Moreover, sizeanalysis of dust samples collected innuclear clouds indicates a substantialsubmicrometer fraction (33). Nuclearsurface detonations may be much moreefficient in generating fine dust than vol-canic eruptions (15, 34), which havebeen used inappropriately in the past toestimate the impacts of nuclear war (2).The intense light emitted by a nuclear

fireball is sufficient to ignite flammablematerials over a wide area (27). Theexplosions over Hiroshima and Nagasakiboth initiated massive conflagrations(35). In each city, the region heavilydamaged by blast was also consumed byfire (36). Assessments over the past twodecades strongly suggest that wide-spread fires would occur after most nu-clear bursts over forests and cities (37-44). The Northern Hemisphere has- 4 x 107 km2 of forest land, whichholds combustible material averaging- 2.2 g/cm2 (7). The world's urban andsuburban zones cover an area of

1.5 x 106 km2 (15). Central cities,which occupy 5 to 10 percent of the totalurban area, hold -10 to 40 g/cm2 ofcombustible material, while residentialareas hold - 1 to 5 g/cm2 (41, 42, 44,45). Smoke emissions from wildfires andlarge-scale urban fires probably lie in therange of 2 to 8 percent by mass of thefuel burned (46). The highly absorbingsooty fraction (principally graphitic car-bon) could comprise up to 50 percent ofthe emission by weight (47, 48). In wild-fires, and probably urban fires, - 90percent of the smoke mass consists ofparticles < 1 p.m in radius (49). Forcalculations at visible wavelengths,smoke particles are assigned an imagi-nary part of the refractive index of 0.3(50).

SCIENCE, VOL. 222

on

Janu

ary

24, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

Simulations

The model predictions discussed heregenerally represent effects averaged overthe Northern Hemisphere (NH). The ini-tial nuclear explosions and fires wouldbe largely confined (51) to northern mid-latitudes (300 to 60°N). Accordingly, thepredicted mean dust and smoke opacitycould be larger by a factor of 2 to 3 atmid-latitudes, but smaller elsewhere.Hemispherically averaged optical depthsat visible wavelengths (52) for the mixednuclear dust and smoke clouds corre-sponding to the scenarios in Table 1 areshown in Fig. 1. The vertical opticaldepth is a convenient diagnostic of nu-clear cloud properties and may be usedroughly to scale atmospheric light levelsand temperatures for the various scenari-Os.

In the baseline scenario (case 1, 5000MT), the initial NH optical depth is = 4,of which 1 is due to stratospheric dustand 3 to tropospheric smoke. After 1month the optical depth is still = 2.Beyond 2 to 3 months, dust dominatesthe optical effects, as the soot is largelydepleted by rainout and washout (54). Inthe baseline case, about 240,000 km2 ofurban area is partially (50 percent)burned by 1000 MT of explosions(only 20 percent of the total exchangeyield). This roughly corresponds to onesixth of the world's urbanized land area,one fourth of the developed area of theNH, and one half of the area of urbancenters with populations > 100,000 inthe NATO and Warsaw Pact countries.The mean quantity of combustible mate-rial consumed over the burned area is

1.9 g/cm2. Wildfires ignited by theremaining 4000 MT of yield burn another500,000 km2 of forest, brush, and grass-lands (7, 39, 55), consuming = 0.5 g/cm2of fuel in the process (7).

Total smoke emission in the baselinecase is - 225 million tons (released overseveral days). By comparison, the cur-rent annual global smoke emission isestimated as - 200 million tons (15), butis probably < 1 percent as effective asnuclear smoke would be in perturbingthe atmosphere (56).The optical depth simulations for cas-

es 1, 2, 9, and 10 in Fig. 1 show that arange of exchanges between 3000 and10,000 MT might create similar effects.Even cases 11, 12, and 13, while lesssevere in their absolute impact, produceoptical depths comparable to or exceed-ing those of a major volcanic eruption. Itis noteworthy that eruptions such asTambora in 1815 may have producedsignificant climate perturbations, even

23 DECEMBER 1983

with an average surface temperature de-crease of S 1 K (57-60).Case 14 represents a 100-MT attack on

cities with 1000 100-KT warheads. In theattack, 25,000 km2 of built-up urban areais burned (such an area could be ac-counted for by : 100 major cities). Thesmoke emission is computed with fireparameters that differ from the baselinecase. The average burden of combustiblematerial in city centers is 20 g/cm2 (ver-sus 10 g/cm2 in case 1) and the average

smoke emission factor is 0.026 gram ofsmoke per gram of material burned (ver-sus the conservative figure of 0.011 g/gadopted for central city fires in the base-line case). About 130 million tons ofurban smoke is injected into the tropo-sphere in each case (none reaches thestratosphere in case 14). In the baselinecase, only about 10 percent of the urbansmoke originates from fires in city cen-ters (Table 3).The smoke injection threshold for ma-

Table 2. Dust parameterization for the baseline case.

Materials in stabilized nuclear explosion clouds*Dust mass Dust size H20

Type of burst (ton/MT): distributiont (ton/MT):

Land surface 3.3 x 105 0.25/2.0/4.0 1.0 x 105Land near-surface 1.0 x 105 0.25/2.0/4.0 1.0 x l0s

Dust composition: siliceous minerals and glassesIndex of refraction at visible wavelengths4: n = 1.50 - 0.001 iStabilized nuclear cloud top and bottom heights, Zt and Zb, for surface and low-air bursts§:

z, = 21 yO2; Zb = 13 yO.2; where Y = yield in megatonsMultiburst interactions are ignored

Baseline dust injectionsTotal dust = 9.6 x 108 tons; 80 percent in the stratosphere; 8.4 percent < I pLm in radiusSubmicrometer dust injection is - 25 ton/KT for surface bursts, which represents - 0.5 percent

of the total ejecta massTotal initial area of stabilized fireballs = 2.0 x 106 km2

*Materials are assumed to be uniformly distributed in the clouds. tParticle size distributions (number/cm3 - pm radius) are log-normal with a power-law tail at large sizes. The parameters rm and a are the log-normal number mode radius and size variance, respectively, and a is the exponent of the r-" dependence atlarge sizes. The log-normal and power-law distributions are connected at a radius of 1_Im (15). fTherefractive indices of dust at infrared wavelengths are discussed in (10). §The model of Foley andRuderman (87) is adopted, but with the cloud heights lowered by about 0.5 km. The original cloud heights arebased on U.S. Pacific test data, and may overestimate the heights at mid-latitudes by several kilometers.

Table 3. Fire and smoke parameterization for the baseline case.

Fire area and emissionsArea of urban fire ignition defined by the 20 cal/cm2 thermal irradiance contour (= 5 psi peak

overpressure contour) with an average atmospheric transmittance of 50 percent:A (kM2) = 250 Y, where Y = yield in megatons detonated over cities; overlap of fire zones isignored

Urban flammable material burdens average 3 g/cm2 in suburban areas and 10 g/cm2 in citycenters (5 percent of the total urban area)

Average consumption of flammables in urban fires is 1.9 g/cm2Average net smoke emission factor is 0.027 g per gram of material burned (for urban centers it is

only 0.011 g/g)Area of wildfires is S x 105 km2 with 0.5 g/cm2 of fuel burned, and a smoke emission factor of

0.032 g/gLong-term fires burn 3 x 10'4 g of fuel with an emission factor of 0.05 g/g

Fire plume heights (top and bottom altitudes)Urban fires: I to 7 kmFire storms (5 percent of urban fires): Zb - 5 km; z, c 19 kmWildfires: I to 5 kmLong-term fires: 0 to 2 km

Fire durationUrban fires, 1 day; wildfires, 10 days; long-terrh fires, 30 days

Smoke propertiesDensity, 1.0 g/cm3; complex index of refraction, 1.75 - 0.30 i; size distribution, log-normal

with rm(p.m)/a = 0.1/2.0 for urban fires and 0.05/2.0 for wildfires and long-term firesBaseline smoke injections

Total smoke emission = 2.25 x 108 tons, 5 percent in the stratosphereUrban-suburban fires account for 52 percent of emissions, fire storms for 7 percent, wildfires

for 34 percent, and long-term fires for 7 percentTotal area burned by urban-suburban fires is 2.3 x l0' km2; fire storms, 1.2 x 104 km2; and

wildfires, 5.0 x 105 km2

1285

on

Janu

ary

24, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

jor optical perturbations on a hemispher-ic scale appears to lie at 1 x 108 tons.From case 14, one can envision the re-lease of 1 x 106 tons of smoke fromeach of 100 major city fires consuming- 4 x '107 tons of combustible materialper city. Such fires could be ignited by100 MT of nuclear explosions. Unex-pectedly, less than 1 percent of the exist-ing strategic arsenals, if targeted on cit-ies, could produce optical (and climatic)disturbances much larger than those pre-viously associated with a massive nucle-ar exchange of = 10,000 MT (2).

Figure 2 shows the surface tempera-ture 'perturbation over continental landareas in the NH calculated from the dustand smoke optical depths for severalscenarios. Most striking are the extreme-ly low temperatures occurring within 3 to4 weeks after a major exchange. In thebaseline 5000-MT case, a minimum landtemperature of - 250 K (-23°C) is pre-dicted after 3 weeks. Subfreezing tem-peratures, persist for several months.Among the cases shown, even the small-est temperature decreases on land are- 5° to 10°C (cases 4, 11, and 12),enough to' turn summer into winter.Thus, severe climatological conse-quenceg might be expected in each ofthese cases. The 100-MT city airburstscenario (case 14) produces a 2-monthinterval of 'subfreezing land tempera-

tures, with a minimum again near 250 K.The temperature recovery in this in-stance is hastened by the absorption ofsunlight in optically thin remnant sootclouds (see below). Comparable ex-changes with and without smoke emis-sion'(for instance, cases 10 and 11) showthat the tropospheric soot layers cause asudden surface cooling of short duration,while fine stratospheric dust is responsi-ble for prolonged cooling lasting a yearor more. [Climatologically, a long-termsurface cooling of only 1°C is significant(60).] In all instances, nuclear dust actsto cool the earth's surface; soot alsotends to cool the surface except whenthe soot cloud is both optically thin andlocated near the surface [an unimportantcase because only relatively small tran-sient warmings s 2 K can thereby beachieved (61)].

Predicted air temperature variationsover the world's oceans associated withchanges in atmospheric radiative trans-port are always small (cooling of s 3 K)because of the great heat content andrapid mixing of surface waters. Howev-er, variations in atmospheric zonal circu-lation patterns (see below) might signifi-cantly alter ocean currents and upwell-ing, as occurred on a smaller scale re-cently in the Eastern Pacific' (El Ninio)(62). The oceanic heat reservoir wouldalso moderate the predicted continental

land temperature decreases, particularlyin coastal regions (10). The effect isdifficult to assess because disturbancesin atmospheric circulation patterns arelikely. Actual temperature decreases incontinental interiors might be roughly 30percent smaller than predicted here, andalong coastlines 70 percent smaller (10).In the baseline case, therefore, continen-tal temperatures may fall to = 260 Kbefore returning to ambient.

Predicted changes in the vertical tem-perature profile for the baseline nuclearexchange are illustrated as a function oftime in Fig. 3. The dominant features ofthe temperature perturbation are a largewarming (up to 80 K) of the lower strato-sphere and upper troposphere, and alarge cooling (up to 40 K) of the surfaceand lower troposphere. The warming iscaused by absorption of solar radiationin the upper-level dust and smokeclouds; it persists for an extended periodbecause of the long residence time of theparticles at high altitudes. The size of thewarming is due to the low heat capacityof the upper atmosphere, its small infra-red emissivity, and the initially low tem-peratures at high altitudes. The surfacecooling is the result of attenuation of theincident solar flux by the aerosol clouds(see Fig. 4) during the first month of thesimulation. The greenhouse effect nolonger occurs in our calculations because

10810 a 1i0 7

Time after detonation (sec)

20ato

0

Co

E0

0

100 200 300

Time after detonation (days)

20

10 0.0

O IT

E0

-10 na0

-20 0

0co-30

-40

Fig. 1 (left). Time-dependent hemispherically averaged vertical opti-cal depths (scattering plus absorption) of nuclear dust and smokeclouds at a wavelength of 550 nm. Optical depths 5 0.1 are negligible,

1 are significant, and > 2 imply possible major consequences.Transmission of sunlight becomes highly nonlinear at optical depths- 1. Results are given for several of the cases in Table 1. Calculatedoptical depths for the expanding El Chich6n eruption cloud are shownfor comparison (53). Fig. 2 (right). Hemispherically averagedsurface temperature variations after a nuclear exchange. Results are

shown for several of the cases in Table 1. (Note the linear time scale,unlike that in Fig. 1). Temperatures generally apply to the interior ofcontinental land masses. Only in cases 4 and 11 are the effects of firesneglected.

SCIENCE, VOL. 222

Ambienttemperature-

1 0

Freezing9 point of

pure water

0

C 1.00

0-t0)

1286

on

Janu

ary

24, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

solar energy is deposited above theheight at which infrared energy is radiat-ed to space.

Decreases in insolation for several nu-clear war scenarios are shown in Fig. 4.The baseline case implies average hemi-spheric solar fluxes at the ground S 10percent of normal values for severalweeks (apart from any patchiness in thedust and smoke clouds). In addition tocausing the temperature declines men-tioned above, the attenuated insolationcould affect plant growth rates, and vigorin the marine (63), littoral, and terrestrialfood chains. In the 10,000-MT "severe"case, average light levels are below theminimum required for photosynthesis forabout 40 days over much of the NorthernHemisphere. In a number of other cases,insolation may, for more than 2 months,fall below the compensation point atwhich photosynthesis is just sufficient tomaintain plant metabolism. Because nu-clear clouds are likely to remain patchythe first week or two after an exchange,leakage of sunlight through holes in theclouds could enhance plant growth activ-ity above that predicted for averagecloud conditions; however, soon thereaf-ter the holes are likely to be sealed.

Sensitivity Tests

A large number of sensitivity calcula-tions were carried out as part of thisstudy (15). The results are summarizedhere. Reasonable variations in the nucle-ar dust parameters in the baseline sce-nario produce initial hemispherically av-eraged dust optical depths varying fromabout 0.2 to 3.0. Accordingly, nucleardust alone could have a major climaticimpact. In the baseline case, the dustopacity is much greater than the totalaerosol opacity associated with the ElChich6n and Agung eruptions (59, 64);even when the dust parameters are as-signed their least adverse values withinthe plausible range, the effects are com-parable to those of a major volcanicexplosion.

Figure 5 compares nuclear cloud opti-cal depths for several variations of thebaseline model smoke parameters (withdust included). In the baseline case, it isassumed that fire storms inject only asmall fraction (- 5 percent) of the totalsmoke emission into the stratosphere(65). Thus, case I and case 3 (no fire-storms) are very similar. As an extremeexcursion, all the nuclear smoke is in-jected into the stratosphere and rapidlydispersed around the globe (case 26);large optical depths can then persist for ayear (Fig. 5). Prolongation of optical23 DECEMBER 1983

effects is also obtained in case 22, wherethe tropospheric washout lifetime ofsmoke particles is increased from 10 to30 days near the ground. By contrast,when the nuclear smoke is initially con-tained near the ground and dynamicaland hydrological removal processes areassumed to be unperturbed, smoke de-pletion occurs much faster (case 25). Buteven in this case, some of the smoke stilldiffuses to the upper troposphere andremains there for several months (66).

In a set of optical calculations, theimaginary refractive index of the smokewas varied between 0.3 and 0.01. Theoptical depths calculated for indices be-tween 0.1 and 0.3 show virtually nodifferences (cases 1 and 27 in Fig. 5). Atan index of 0.05, the absorption opticaldepth (52) is reduced by only - 50 per-cent, and at 0.01, by - 85 percent. Theoverall opacity (absorption plus scatter-ing), moreover, increases by = 5 per-

Fig. 3. NorthernHemisphere tropo-sphere and strato-sphere temperatureperturbations (in Kel-vins; 1 K = 1C) afterthe baseline nuclearexchange (case 1).The hatched area in-dicates cooling. Am-bient pressure levelsin millibars are alsogiven.

40

j 30

0

= 20

10

0

cent. These results show that light ab-sorption and heating in nuclear smokeclouds remain high until the graphiticcarbon fraction of the smoke falls belowa few percent.One sensitivity test (case 29, not illus-

trated) considers the optical effects in theSouthern Hemisphere (SH) of dust andsoot transported from the NH strato-sphere. In this calculation, the smoke inthe 300-MT SH case 13 is combined withhalf the baseline stratospheric dust andsmoke (to approximate rapid global dis-persion in the stratosphere). The initialoptical depth is =1 over the SH, drop-ping to about 0.3 in 3 months. Predictedaverage SH continental surface tempera-tures fall by 8 K within several weeksand remain at least 4 K below normal fornearly 8 months. The seasonal influenceshould be taken into account, however.For example, the worst consequencesfor the NH might result from a spring or

0

Time after detonation (days)

1000

Fig. 4. Solar energyfluxes at the groundover the NorthernHemisphere in the af-termath of a nuclearexchange. Results aregiven for several ofthe cases in Table 1.(Note the linear timescale.) Solar fluxesare averaged over thediurnal cycle and overthe hemisphere. Incases 4 and 16 firesare neglected. Alsoindicated are the ap-proximate flux levelsat which photosyn-thesis cannot keeppace with plant respi-ration (compensationpoint) and at whichphotosynthesis ceas-es. These limits varyfor different species.

N

cm

E

c:

VD 100c

0a0

ox

20

0 10

co0co,

Time after detonation (days)

1287

on

Janu

ary

24, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

summer exchange, when crops are vul-nerable and fire hazards are greatest.The SH, in its fall or winter, might thenbe least sensitive to cooling and darken-ing. Nevertheless, the implications ofthis scenario for the tropical regions inboth hemispheres appear to be seriousand worthy of further analysis. Seasonalfactors can also modulate the atmospher-ic response to perturbations by smokeand dust, and should be considered.A number of sensitivity tests for more

severe cases were run with exchangeyields ranging from 1000 to 10,000 MTand smoke and dust parameters assignedmore adverse, but not implausible, val-ues. The predicted effects are substan-tially worse (see below). The lower prob-abilities of these severe cases must beweighed against the catastrophic out-comes which they imply. It would beprudent policy to assess the importanceof these scenarios in terms of the productof their probabilities and the costs oftheir corresponding effects. Unfortu-

nately, we are unable to give an accuratequantitative estimate of the relevantprobabilities. By their very nature, how-ever, the severe cases may be the mostimportant to consider in the deploymentof nuclear weapons.With these reservations, we present

the optical depths for some of the moresevere cases in Fig. 6. Large opacitiescan persist for a year, and land surfacetemperatures can fall to 230 to 240 K,about 50 K below normal. Combinedwith low light levels (Fig. 4), these se-vere scenarios raise the possibility ofwidespread and catastrophic ecologicalconsequences.Two sensitivity tests were run to de-

termine roughly the implications for opti-cal properties of aerosol agglomerationin the early expanding clouds. (The sim-ulations already take into account con-tinuous coagulation of the particles in thedispersed clouds.) Very slow dispersionof the initial stabilized dust and smokeclouds, taking nearly 8 months to cover

the NH, was assumed. Coagulation ofparticles reduced the average opacityafter 3 months by about 40 percent.When the adhesion efficiency of the col-liding particles was also maximized, theaverage opacity after 3 months was re-duced by - 75 percent. In the mostlikely situation, however, prompt ag-glomeration and coagulation might re-duce the average hemispheric cloud opti-cal depths by 20 to 50 percent.

Other Effects

We also considered, in less detail, thelong-term effects of radioactive fallout,fireball-generated NOR, and pyrogenictoxic gases (15). The physics of radioac-tive fallout is well known (2, 5, 12, 27,67). Our calculations bear primarily onthe widespread intermediate time scaleaccumulation of fallout due to washoutand dry deposition of dispersed nucleardust (68). To estimate possible exposure

1 week 1 month 4 months 1 year 1 week 1 month 4 months 1 year

100

10

a)

*0.2

C)a

0.1

0.1

0.0 1 . , I

105 143 107 108Time after detonation (sec)

105 106107Time after detonation (sec)

Fig. 5 (left). Time-dependent vertical optical depths (absorption plus scattering at 550 nm) of nuclear clouds, in a sensitivity analysis. Opticaldepths are average values for the Northern Hemisphere. All cases shown correspond to parameter variations of the baseline model (case 1) andinclude dust appropriate to it: case 3, no fire storms; case 4, no fires; case 22, smoke rainout rate decreased by a factor of 3; case 25, smoke initial-ly confined to the lowest 3 km of the atmosphere; case 26, smoke initially distributed between 13 and 19 km over the entire globe; and case 27,smoke imaginary part of refractive index reduced from 0.3 to 0.1. For comparison, in case 4, only dust from the baseline model is considered(fires are ignored). Fig. 6 (right). Time-dependent vertical optical depths (absorption plus scattering at 550 nm) for enhanced cases ofexplosion yield or nuclear dust and smoke production. Conditions are detailed elsewhere (15). Weapon yield inventories are identical to thenominal cases of the same total yield described in Table I (cases 16 and 18 are also listed there). The "severe" cases generally include a sixfold in-crease in fine dust injection and a doubling of smoke emission. In cases 15, 17, and 18, smoke causes most of the opacity during the first I to 2months. In cases 17 and 18, dust makes a major contribution to the optical effects beyond 1 to 2 months. In case 16, fires are neglected and dustfrom surface bursts produces all of the opacity.

108

SCIENCE, VOL. 2221288

on

Janu

ary

24, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

levels, we adopt a fission yield fractionof 0.5 for all weapons. For exposure toonly the gamma emission of radioactivedust that begins to fall out after 2 days inthe baseline scenario (5000 MT), thehemispherically averaged total dose ac-cumulated by humans over severalmonths could be = 20 rads, assuming noshelter from or weathering of the dust.Fallout during this time would be con-fined largely to northern mid-latitudes;hence the dose there could be - 2 to 3times larger (69, 70). Considering inges-tion of biologically active radionuclides(27, 71) and occasional exposure to local-ized fallout, the average total chronicmid-latitude dose of ionizing radiationfor the baseline case could be - 50 radsof whole-body external gamma radia-tion, plus -50 rads to specific bodyorgans from internal beta and gammaemitters (71, 72). In a 10,000-MT ex-change, under the same assumptions,these mean doses would be doubled.Such doses are roughly an order of mag-nitude larger than previous estimates,which neglected intermediate time scalewashout and fallout of tropospheric nu-clear debris from low-yield (< 1-MT)detonations.The problem of NO, produced in the

fireballs of high-yield explosions, and theresulting depletion of stratosphericozone, has been treated in a number ofstudies (2-4, 7, 73). In our baseline casea maximum hemispherically averagedozone reduction of = 30 percent isfound. This would be substantially small-er if individual warhead yields were allreduced below 1 MT. Considering therelation between solar UV B radiationincreases and ozone decreases (74), UV-B doses roughly twice normal are ex-pected in the first year after a baselineexchange (when the dust and soot haddissipated). Large UV-B effects couldaccompany exchanges involving war-heads of greater yield (or large multi-burst laydowns).A variety of toxic gases (pyrotoxins)

would be generated in large quantities bynuclear fires, including CO and HCN.According to Crutzen and Birks (7),heavy air pollution, including elevatedozone concentrations, could blanket theNH for several months. We are alsoconcerned about dioxins and furans, ex-tremely persistent and toxic compoundswhich are released during the combus-tion of widely used synthetic organicchemicals (75). Hundreds of tons ofdioxins and furans could be generatedduring a nuclear exchange (76). Thelong-term ecological consequences ofsuch nuclear pyrotoxins seem worthy offurther consideration.23 DECEMBER 1983

Meteorological Perturbations

Horizontal variations in sunlight ab-sorption in the atmosphere, and at thesurface, are the fundamental drivers ofatmospheric circulation. For many of thecases considered in this study, sizablechanges in the driving forces are implied.For example, temperature contrastsgreater than 10 K between NH continen-tal areas and adjacent oceans may inducea strong monsoonal circulation, in someways analogous to the wintertime pat-tern near the Indian subcontinent. Simi-larly, the temperature contrast betweendebris-laden atmospheric regions and ad-jacent regions not yet filled by smokeand dust will cause new circulation pat-terns.Thick clouds of nuclear dust and

smoke can thus cause significant climaticperturbations, and related effects,through a variety of mechanisms: reflec-tion of solar radiation to space and ab-sorption of sunlight in the upper atmo-sphere, leading to overall surface cool-ing; modification of solar absorption andheating patterns that drive the atmo-spheric circulation on small scales (77)and large scales (78); introduction ofexcess water vapor and cloud condensa-tion nuclei, which affect the formation ofclouds and precipitation (79); and alter-ation of the surface albedo by fires andsoot (80). These effects are closely cou-pled in determining the overall responseof the atmosphere to a nuclear war (81).It is not yet possible to forecast in detailthe changes in coupled atmospheric cir-culation and radiation fields, and inweather and microclimates, which wouldaccompany the massive dust and smokeinjections treated here. Hence specula-tion must be limited to the most generalconsiderations.Water evaporation from the oceans is

a continuing source of moisture for themarine boundary layer. A heavy semi-permanent fog or haze layer might blan-ket large bodies of water. The conse-quences for marine precipitation are notclear, particularly if normal prevailingwinds are greatly modified by the per-turbed solar driving force. Some conti-nental zones might be subject to continu-ous snowfall for several months (10).Precipitation can lead to soot removal,although this process may not be veryefficient for nuclear clouds (77, 79). It islikely that, on average, precipitationrates would be generally smaller than inthe ambient atmosphere; the major re-maining energy source available forstorm genesis is the latent heat fromocean evaporation, and the upper atmo-sphere is warmer than the lower atmo-

sphere which suppresses convection andrainfall.

Despite possible heavy snowfalls, it isunlikely that an ice age would be trig-gered by a nuclear war. The period ofcooling (: I year) is probably too shortto overcome the considerable inertia inthe earth's climate system. The oceanicheat reservoir would probably force theclimate toward contemporary norms inthe years after a war. The CO2 inputfrom nuclear fires is not significant cli-matologically (7).

Interhemispheric Transport

In earlier studies it was assumed thatsignificant interhemispheric transport ofnuclear debris and radioactivity requiresa year or more (2). This was based onobservations of transport under ambientconditions, including dispersion of de-bris clouds from individual atmosphericnuclear weapons tests. However, withdense clouds of dust and smoke pro-duced by thousands of nearly simulta-neous explosions, large dynamical dis-turbances would be expected in the af-termath of a nuclear war. A rough analo-gy can be drawn with the evolution ofglobal-scale dust storms on Mars. Thelower martian atmosphere is similar indensity to the earth's stratosphere, andthe period of rotation is almost identicalto the earth's (although the solar insola-tion is only half the terrestrial value).Dust storms that develop in one hemi-sphere on Mars often rapidly intensifyand spread over the entire planet, cross-ing the equator in a mean time of = 10days (15, 82, 83). The explanation appar-ently lies in the heating of the dust aloft,which then dominates other heat sourcesand drives the circulation. Haberle et al.(82) used a two-dimensional model tosimulate the evolution of martian duststorms and found that dust at low lati-tudes, in the core of the Hadley circula-tion, is the most important in modifyingthe winds. In a nuclear exchange, mostof the dust and smoke would be injectedat middle latitudes. However, Haberle etal. (82) could not treat planetary-scalewaves in their calculations. Perturba-tions of planetary wave amplitudes maybe critical in the transport of nuclear wardebris between middle and low latitudes.

Significant atmospheric effects in theSH could be produced (i) through dustand smoke injection resulting from ex-plosions on SH targets, (ii) throughtransport of NH debris across the mete-orological equator by monsoon-likewinds (84), and (iii) through interhemi-spheric transport in the upper tropo-

1289

on

Janu

ary

24, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

sphere and stratosphere, driven by solarheating of nuclear dust and smokeclouds. Photometric observations of theEl Chich6n volcanic eruption cloud (ori-gin, 14'N) by the Solar Mesosphere Ex-plorer satellite show that 10 to 20 percentof the stratospheric aerosol had beentransported to the SH after - 7 weeks(85).

Discussion and Conclusions

The studies outlined here suggest se-vere long-term climatic effects from a5000-MT nuclear exchange. Despite un-certainties in the amounts and propertiesof the dust and smoke produced by nu-clear detonations, and the limitations ofmodels available for analysis, the follow-ing tentative conclusions may be drawn.

1) Unlike most earlier studies [for in-stance, (2)], we find that a global nuclearwar could have a major impact on cli-mate-manifested by significant surfacedarkening over many weeks, subfreezingland temperatures persisting for up toseveral months, large perturbations inglobal circulation patterns, and dramaticchanges in local weather and precipita-tion rates-a harsh "nuclear winter" inany season. Greatly accelerated inter-hemispheric transport of nuclear debrisin the stratosphere might also occur,although modeling studies are needed toquantify this effect. With rapid inter-hemispheric mixing, the SH could besubjected to large injections of nucleardebris soon after an exchange in theNorthern Hemisphere. In the past, SHeffects have been assumed to be minor.Although the climate disturbances areexpected to last more than a year, itseems unlikely that a major long-termclimatic change, such as an ice age,would be triggered.

2) Relatively large climatic effectscould result even from relatively smallnuclear exchanges (100 to 1000 MT) ifurban areas were heavily targeted, be-cause as little as 100 MT is sufficient todevastate and burn several hundred ofthe world's major urban centers. Such alow threshold yield for massive smokeemissions, although scenario-dependent,implies that even limited nuclear ex-changes could trigger severe aftereffects.It is much less likely that a 5000- to10,000-MT exchange would have onlyminor effects.

3) The climatic impact of sooty smokefrom nuclear fires ignited by airbursts isexpected to be more important than thatof dust raised by surface bursts (whenboth effects occur). Smoke absorbs sun-light efficiently, whereas soil dust is gen-

1290

erally nonabsorbing. Smoke particles areextremely small (typically < 1 ,um inradius), which lengthens their atmo-spheric residence time. There is also ahigh probability that nuclear explosionsover cities, forests, and grasslands willignite widespread fires, even in attackslimited to missile silos and other strate-gic military targets.

4) Smoke from urban fires may bemore important than smoke from collat-eral forest fires for at least two reasons:(i) in a full-scale exchange, cities holdinglarge stores of combustible materials arelikely to be attacked directly; and (ii)intense fire storms could pump smokeinto the stratosphere, where the resi-dence time is a year or more.

5) Nuclear dust can also contribute tothe climatic impact of a nuclear ex-change. The dust-climate effect is verysensitive to the conduct of the war; asmaller effect is expected when loweryield weapons are deployed and air-bursts dominate surfke land bursts.Multiburst phenomena might enhancethe climatic effects of nuclear dust, butnot enough data are available to assessthis issue.

6) Exposure to radioactive falloutmay be more intense and widespreadthan predicted by empirical exposuremodels, which neglect intermediate fall-out extending over many days andweeks, particularly.when unprecedentedquantities of fission debris are releasedabruptly into the troposphere by explo-sions with submegaton yields. AverageNH mid-latitude whole-body gamma-raydoses of up to 50 rads are possible in a5000-MT exchange; larger doses wouldaccrue within the fallout plumes of radio-active debris extending hundreds ofkilometers downwind of targets. Theseestimates neglect a probably significantinternal radiation dose due to biological-ly active radionuclides.

7) Synergisms between long-term nu-clear war stresses-such as low lightlevels, subfreezing temperatures, expo-sure to intermediate time scale radioac-tive fallout, heavy pyrogenic air pollu-tion, and UV-B flux enhancements-aggravated by the destruction of medicalfacilities, food stores, and civil services,could lead to many additional fatalities,and could place severe stresses on theglobal ecosystem. An assessment of thepossible long-term biological conse-quences of the nuclear war effects quan-tified in this study is made by Ehrlich etal. (86).Our estimates of the physical and

chemical impacts of nuclear war are nec-essarily uncertain because we have usedone-dimensional models, because the

data base is incomplete, and because theproblem is not amenable to experimentalinvestigation. We are also unable to fore-cast the detailed nature of the changes inatmospheric dynamics and meteorologyimplied by our nuclear war scenarios, orthe effect of such changes on the mainte-nance or dispersal of the initiating dustand smoke clouds. Nevertheless, themagnitudes of the first-order effects areso large, and the implications so serious,that we hope the scientific issues raisedhere will be vigorously and criticallyexamined.

References and Notes

1. J. Hampson, Nature (London) 250, 189 (1974).2. National Academy of Sciences, Long-Term

Worldwide Effects of Multiple Nuclear-WeaponDetonations (Washington, D.C., 1975).

3. R. C. Whitten, W. J. Borucki, R. P. Turco,Nature (London) 257, 38 (1975).

4. M. C. MacCracken and J. S. Chang, Eds.,Lawrence Livermore Lab. Rep. UCRL-516S3(1975).

5. J. C. Mark, Annu. Rev. Nucl. Sci. 26, 51 (1976).6. K. N. Lewis, Sci. Am. 241, 35 (July 1979).7. P. J. Crutzen and J. W. Birks, Ambio 11, 114

(1982).8. L. W. Alvarez, W. Alvarez, F. Asaro, H. V.

Michel, Science 208, 1095 (1980); W. Alvarez,F. Asaro, H. V. Michel, L. W. Alvarez, ibid.216, 886 (1982); W. Alvarez, L. W. Alvarez, F.Asaro, H. V. Michel, Geol. Soc. Am. Spec.Pap. 190 (1982), p. 305.

9. R. Ganapathy, Science 216, 885 (1982).10. 0. B. Toon et al., Geol. Soc. Am. Spec. Pap.

190 (1982), p. 187; J. B. Pollack, 0. B. Toon, T.P. Ackerman, C. P. McKay, R. P. Turco, Sci-ence 219, 287 (1983).

11. Under the sponsorship of the Defense NuclearAgency, the National Research Council (NRC)of the National Academy of Sciences has alsoundertaken a full reassessment of the possibleclimatic effects of nuclear war. The presentanalysis was stimulated, in part, by earlier NRCinterest in a preliminary estimate of the climaticeffects of nuclear dust.

12. Office of Technology Assessment, The Effectsof Nuclear War (OTA-NS-89, Washington,D.C., 1979).

13. J. E. Coggle and P. J. Lindop, Ambio 11, 106(1982).

14. S. Bergstrom et al., "Effects of nuclear war onhealth and health services," WHO Publ. A36.12(1983).

15. R. P. Turco, 0. B. Toon, T. P. Ackerman, J. B.Pollack, C. Sagan, in preparation.

16. R. P. Turco, P. Hamill, 0. B. Toon, R. C.Whitten, C. S. Kiang, J. Atmos. Sci. 36, 699(1979); NASA Tech. Pap. 1362 (1979); R. P.Turco, 0. B. Toon, P. Hamill, R. C. Whitten, J.Geophys. Res. 86, 1113 (1981); R. P. Turco, 0.B. Toon, R. C. Whitten, Rev. Geophys. SpacePhys. 20, 233 (1982); R. P. Turco, O. B. Toon,R. C. Whitten, P. Hamill, R. G. Keesee, J.Geophys. Res. 88, 5299 (1983).

17. 0. B. Toon, R. P. Turco, P. Hamill, C. S.Kiang, R. C. Whitten, J. Atmos. Sci. 36, 718(1979); NASA Tech. Pap. 1363 (1979).

18. 0. B. Toon and T. P. Ackerman, Appl. Opt. 20,3657 (1981); T. P. Ackerman and 0. B. Toon,ibid., p. 3661; J. N. Cuzzi, T. P. Ackerman, L.C. Helme, J. Atmos. Sci. 39, 917 (1982).

19. Prediction of circulation anomalies and atten-dant changes in regional weather patterns re-quires an appropriately designed three-dimen-sional general circulation model with at least thefollowing features: horizontal resolution of 10'or better, high vertical resolution through thetroposphere and stratosphere, cloud and precip-itation parameterizations that allow for excur-sions well outside present-day experience, abili-ty to transport dust and smoke particles, aninteractive radiative transport scheme to calcu-late dust and smoke effects on light fluxes andheating rates, allowance for changes in particlesizes with time and for wet and dry deposition,and possibly a treatment of the coupling be-tween surface winds and ocean currents andtemperatures. Even if such a model were avail-able today, it would not be able to resolvequestions of patchiness on horizontal scales ofless than several hundred kilometers, of local-

SCIENCE, VOL. 222

on

Janu

ary

24, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

ized perturbations in boundary-layer dynamics,or of mesoscale dispersion and removal of dustand smoke clouds.

20. Advisors, Ambio 11, 94 (1982).21. R. T. Pretty, Ed., Jane's Weapon Systems,

1982-1983 (Jane's, London, 1982).22. The Military Balance 1982-1983 (International

Institute for Strategic Studies, London, 1982).23. World Armaments and Disarmament, Stock-

holm International Peace Research InstituteYearbook 1982 (Taylor & Francis, London,1982).

24. R. Forsberg, Sci. Am. 247, 52 (November 1982).25. The unprecedented difficulties involved in con-

trolling a limited nuclear exchange are discussedby, for example, P. Bracken and M. Shubik[Technol. Soc. 4, 155 (1982)] and by D. Ball[Adelphi Paper 169 (International Institute forStrategic Studies, London, 198 1)].

26. G. Kemp, Adelphi Paper 106 (International In-stitute for Strategic Studies, London, 1974).

27. S. Glasstone and P. J. Dolan, Eds., The Effectsof Nuclear Weapons (Department of Defense,Washington, D.C., 1977).

28. The areas cited are subject to peak overpres-sures - 10 to 20 cal/cm2.

29. A 1-MT surface explosion ejects - 5 x 106 tonsof debris, forming a large crater (27). Typicalsoils consist of = 5 to 25 percent by weight ofgrains I ,um in radius [G. A. D'Almeida andL. Schutz, J. Climate AppI. Meteorol. 22, 233(1983); G. Rawson, private communication].However, the extent of disaggregation of the soilinto parent grain sizes is probably s 10 percent[R. G. Pinnick, G. Fernandez, B. D. Hinds,Appl. Opt. 22, 95 (1983)] and would depend inpart on soil moisture and compaction.

30. A I-MT surface explosion vaporizes = 2 x 104to 4 x 104 tons of soil (27), which is ingested bythe fireball. Some silicates and other refractorymaterials later renucleate into fine glassyspheres [M. W. Nathans, R. Thews, I. J. Rus-sell, Adv. Chem. Ser. 93, 360 (1970)].

31. A 1-MT surface explosion raises significantquantities of dust over an area of - 100 km2 by"popcorning," due to thermal radiation, and bysaltation, due to pressure winds and turbulence(27). Much of the dust is sucked up by theafterwinds behind the rising fireball. Size sortingshould favor greatest lifting for the finest parti-cles. The quantity of dust lofted would be sensi-tive to soil type, moisture, compaction, vegeta-tion cover, and terrain. Probably > I x 105 tonsof dust per megaton can be incorporated into thestabilized clouds in this manner.

32. R. G. Gutmacher, G. H. Higgins, H. A. Tewes,Lawrence Liv,ermore Lab. Rep. UCRL-14397(1983); J. Carpenter, private communication.

33. M. W. Nathans, R. Thews, I. J. Russell [in (30)].These data suggest number size distributionsthat are log-normal at small sizes (! 3 pum) andpower law (r-a) at larger sizes. Considering datafrom a number of nuclear tests, we adopted anaverage log-normal mode radius of 0.25 ,um,a = 2.0, and an exponent, a = 4 (15). If allparticles in the stabilized clouds have radii in therange 0.01 to 1000 p.m, the adopted size distribu-tion has = 8 percent of the total mass in parti-cles I p.m in radius; this fraction of thestabilized cloud mass represents 0.5 percentof the total ejecta and sweep-up mass of asurface explosion and amounts to = 25 tons perkiloton of yield.

34. Atmospheric dust from volcanic explosions dif-fers in several important respects from thatproduced by nuclear explosions. A volcaniceruption represents a localized dust source,while a nuclear war would involve thousands ofwidely distributed sources. The dust mass con-centration in stabilized nuclear explosion cloudsis low (5s g/m3), while volcanic eruption col-umns are so dense they generally collapse undertheir own weight [G. P. L. Walker, J. Volcanol.Geotherm. Res. 11, 81 (1981)]. In the densevolcanic clouds particle agglomeration, particu-larly under the influence of electrical charge,can lead to accelerated removal by sedimenta-tion [S. N. Carey and H. Sigurdsson, J.Geophys. Res. 87, 7061 (1982); S. Brazier et al.,Nature (London-) 301 115 (1983)]. The size distri-bution of volcanic ash is also fundamentallydifferent from that of nuclear dust [W. I. Rose etal., Am. J. Sci. 280, 671 (1980)], because theorigins of the particles are so different. Theinjection efficiency of nuclear dust into thestratosphere by megaton-yield explosions isclose to unity, while the injection efficiency offine volcanic dust appears to be very low (15).For these reasons and others, the observedclimatic effects of major historical volcaniceruptions cannot be used, as in (2), to calibratethe potential climatic effect of nuclear dustmerely by scaling energy or soil volume. How-

23 DECEMBER 1983

ever, in cases where the total amount of submi-crometer volcanic material that remained in thestratosphere could be determined, climate mod-els have been applied and tested [J. B. Pollack etal., J. Geophys. Res. 81, 1071 (1976)]. We usedsuch a model in this study to predict the effectsof specific nuclear dust injections.

35. E. Ishikawa and D. L. Swain, Translators, Hiro-shima and Nagasaki: The Physical, Medicaland Social Effects of the Atomic Bombings(Basic Books, New York, 1981).

36. At Hiroshima, a weapon of roughly 13 KTcreated a fire over - 13 km2. At Nagasaki,where irregular terrain inhibited widespread fireignition, a weapon of roughly 22 KT caused afire over = 7 km2. These two cases suggest thatlow-yield (5 1-MT) nuclear explosions canreadily ignite fires over an area of = 0.3 to 1.0km2/KT-roughly the area contained within the

1O cal/cm2 and the = 2 psi overpressure con-tours (27).

37. A. Broido, Bull. At. Sci. 16, 409 (1960).38. C. F. Miller, "Preliminary evaluation of fire

hazards from nuclear detonations," SRI (Stan-ford Res. Inst.) Memo. Rep. Project IMU4021-302 (1962).

39. R. U. Ayers, Environmental Effects of NuclearWeapons (HI-518-RR, Hudson Institute, NewYork, 1965), vol. 1.

40. S. B. Martin, "The role of fire in nuclear war-fare," United Research Services Rep. URS-764(DNA 2692F) (1974).

41. DCPA Attack Environment Manual (Depart-ment of Defense, Washington, D.C., 1973),chapter 3.

42. FEMA Attack Environment Manual (CPG 2-1A3, Federal Emergency Management Agency,Washington, D.C., 1982), chapter 3.

43. H. L. Brode, "Large-scale urban fires,' Pac ificSierra Res. Corp. Note 348 (1980).

44. D. A. Larson and R. D. Small, "Analysis of thelarge urban fire environment," Pacific SierraiRes. Corp. Rep. 1210 (1982).

45. Urban and suburban areas of cities with popula-tions exceeding 100,000 (about 2300 worldwide)are surveyed in (15). Also discussed are globalreserves of flammable substances, which areshown to be roughly consistent with knownrates of production and accumulation of com-bustible materials. P. J. Crutzen and I. E. Gal-bally (in preparation) reach similar conclusionsabout global stockpiles of combustibles.

46. Smoke emission data for forest fires are re-viewed by D. V. Sandberg, J. M. Pierovich, D.G. Fox, and E. W. Ross ["Effects of fire onair," U.S. Forest Serv. Tech. Rep. WO-9(1979)]. Largest emission factors occur in in-tense large-scale fires where smoldering andflaming exist simultaneously, and the oxygensupply may be limited over part of the burningzone. Smoke emissions from synthetic organiccompounds would generally be larger than thosefrom forest fuels [C. P. Bankston, B. T. Zinn, R.F. Browner, E. A. Powell, Combust. Flame 41,273 (1981)].

47. Sooty smoke is a complex mixture of oils, tars,and graphitic (or elemental) carbon. Measuredbenzene-soluble mass fractions of wildfiresmokes fall in the range = 40 to 75 percent [D.V. Sandberg et al. in (46)]. Most of the residue islikely to be brown to black (the color of smokeranges from white, when large amounts of watervapor are present, to yellow or brown, when oilspredominate, to gray or black, when elementalcarbon is a major component).

48. A. Tewarson, in Flame Retardant PolymericMaterial, M. Lewin, S. M. Atlas, E. M. Pierce,Eds. (Plenum, New York, 1982), vol. 3, pp. 97-153. In small laboratory bums of a variety ofsynthetic organic compounds, emissions of"solid" materials (which remained on collectionfilters after baking at 100°C for 24 hours) rangedfrom = I to 15 percent by weight of the carbonconsumed; of low-volatility liquids, 2 to 35percent; and of high-volatility liquids, I to 40percent. Optical extinction of the smoke gener-ated by a large number of samples varied from= 0.1 to 1.5 m- per gram of fuel burned.

49. In wildfires, the particle number mode radius istypically about 0.05 p.m [D. V. Sandberg et al.,in (46)]. For burning synthetics the numbermode radius can be substantially greater, but areasonable average value is 0.1 p.m [C. P. Bank-ston et al., in (46)]. Often, larger debris particlesand firebrands are swept up by powerful firewinds, but they have short atmospheric resi-dence times and are not included in the presentestimates (C. K. McMahon and P. W. Ryan,paper presented at the 69th Annual Meeting, AirPollution Control Association, Portland, Ore.,27 June to I July 1976). Nevertheless, becausewinds exceeding 100 km/hour may be generatedin large-scale fires, significant quantities of fine

noncombustible surface dust and explosion de-bris (such as pulverized plaster) might be liftedin addition to the smoke particles.

50. This assumes an average graphitic carbon massfraction of about 30 to 50 percent, for a purecarbon imaginary refractive index of 0.6 to 1.0[J. T. Twitty and J. A. Weinman, J. Appl.Meteorol. 10, 725 (1971); S. Chippett and W. A.Gray, Combust. Flame 31, 149 (1978)]. The realpart of the refractive index of pure carbon is1.75, and for many oils is 1.5 to 1.6. Smokeparticles were assigned an average density of Ig/cm3 (C. K. McMahon, paper presented at the76th Annual Meeting, Air Pollution Control As-sociation, Atlanta, Ga., 19 to 24 June 1983).Solid graphite has a density = 2.5 g/cm3, andmost oils, s I g/cm3.

51. A number of targets with military, economic, orpolitical significance can also be identified intropical northern latitudes and in the SH (20).

52. Attenuation of direct sunlight by dust and smokeparticles obeys the law IIIo = exp(-T/p.0), whereT is the total extinction optical depth due tophoton scattering and absorption by the parti-cles and p.0 is the cosine of the solar zenithangle. The optical depth depends on the wave-length of the light and the size distribution andcomposition of the particles, and is generallycalculated fronm Mie theory (assuming equiva-lent spherical particles). The total light intensityat the ground consists of a direct component anda diffuse, or scattered, component, the latterusually calculated with a radiative transfer mod-el. The extinction optical depth can be written asT = XML, where X is the specific cross section(m2/g particulate), M the suspended particlemass concentration (g/m3), and L the path length(m). It is the sum of a scattering and an absorp-tion optical depth (r = rT + Ta). Fine dust andsmoke particles have scattering coefficientsXs _ 3 to 5 m2/g at visible wavelengths. Howev-er, the absorption coefficients Xa are very sensi-tive to the imaginary part of the index of refrac-tion. For typical soil particles, Xa : 0.1 m2/g.For smokes, Xa can vary from - 0.1 to 10 m2/g,roughly in proportion to the volume fraction ofgraphite in the particles. Occasionally, specificextinction coefficients for smoke are given rela-tive to the mass of fuel burned; then X implicitlyincludes a multiplicative emission factor (gramsof smoke generated per gram of fuel burned).

53. R. P. Turco, 0. B. Toon, R. C. Whitten, P.Hamill, Eos 63, 901 (1982).

54. J. A. Ogren, in Particulate Carbon: Atmospher-ic Life Cycle, G. T. Wolff and R. L. Klimisch,Eds. (Plenum, New York; 1982), pp. 379-391.

55. To estimate the wildfire area, we assume that 25percent of the total nonurban yield, or 1000 MT,ignites fires over an area of 500 km2/MT-approximately the zone irradiated by 10 cal/cm2-and that the fires do not spread outsidethis zone (39). R. E. Huschke [Rand Corp. Rep.RM-5073-TAB (1966)] analyzed the simulta-neous flammability of wildland fuels in the Unit-ed States, and determined that about 50 percentof all fuels are at least moderately flammablethroughout the summer months. Because = 50percent of the land areas of the countries likelyto be involved in a nuclear exchange are coveredby forest and brush, which are flammable about50 percent of the time, the 1000-MT ignitionyield follows statistically.

56. Most of the background smoke is injected intothe lowest I to 2 km of the atmosphere, where ithas a short lifetime, and consists on the averageof : 10 percent graphitic carbon [R. P. Turco,0. B. Toon, R. C. Whitten, J. B. Pollack, P.Hamill, in Precipitation Scavenging, Dry Depo-sition and Resuspension, H. R. Pruppacher, R.G. Semonin, W. G. N. Slinn, Eds. (Elsevier,New York, 1983), p. 1337]. Thus, the averageoptical depth of ambient atmospheric soot isonly ! 1 percent of the initial optical depth ofthe baseline nuclear war smoke pall.

57. H. E. Landsberg and J. M. Albert, Weatherwise27, 63 (1974).

58. H. Stommel and E. Stommel, Sci. Am. 240, 176(June 1979).

59. 0. B. Toon and J. B. Pollack, Nat. Hist. 86, 8(January 1977).

60. H. H. Lamb, Climate Present, Past and Future(Methuen, London, 1977), vols. I and 2.

61. Notwithstanding possible alterations in the sur-face albedo due to the fires and deposition ofsoot (15, 80).

62. S. G. H. Philander, Nature (London) 302, 295(1983); B. C. Weare, Science 221, 947 (1983).

63. D. H. Milne and C. P. McKay, Geol. Soc. Am.Spec. Pap. 190 (1982), p. 297.

64. 0. B. Toon, Eos 63, 901 (1982).65. The stratosphere is normally defined as the

region of constant or increasing temperaturewith increasing height lying just above the tropo-

1291

on

Janu

ary

24, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

sphere. The residence time of fine particles inthe stratosphere is considerably longer than inthe upper troposphere, because of the greaterstability of the stratospheric air layers and theabsence of precipitation in the stratosphere.With large smoke injections, however, the ambi-ent temperature profile would be substantiallydistorted (for instance, see Fig. 3) and a "strato-sphere" might form in the vicinity of thie smokecloud, increasing its residence time at all alti-tudes (15). Thus the duration of sunlight attenua-tion and temperature perturbations in Figs. I to6 may be considerably underestimated.

66. Transport of soot from the boundary layer intothe overlying free troposphere can occur bydiurnal expansion and contraction of the bound-ary layer, by large-scale advection, and bystrong localized convection.

67. F. Barnaby and J. Rotblat, Ambio 11, 84 (1982).68. The term "intermediate" fallout distinguishes

the radioactivity deposited between severaldays and - I month after an exchange from"prompt" fallout (: I day) and "late" fallout(months to years). Intermediate fallout is ex-pected to be at least hemispheric in scale andcan still deliver a significant chronic whole-bodygamma-ray dose. It may also contribute a sub-stantial internal dose, for example, from 1311.The intermediate time scale gamma-ray doserepresents, in one sense, the minimum averageexposure far from targets and plumes of promptfallout. However, the geographic distribution ofintermediate fallout would still be highly vari-able, and estimates of the average dose madewith a one-dimensional model are greatly ideal-ized. The present calculations were calibratedagainst the observed prompt fallout of nucleartest explosions (15).

69. There is also reason to believe that the fissionyield fraction of nuclear devices may be increas-ing as warhead yields decrease and uraniumprocessing technology improves. If the fissionfraction were unity, our dose estimates wouldhave to be doubled. We also neglect additionalpotential sources of radioactive fallout fromsalted "dirty" weapons and explosions overnuclear reactors and fuel reprocessing plants.

70. J.. Knox (Lawrence Livermore Lab. Rep.UCRL-89907, in press) reports fallout calcula-tions which explicitly account for horizontalspreading and transport of nuclear debrisclouds. For a 53Q0-MT strategic exchange,Knox computes average whole-body gamma-raydoses of 20 rads from 40° to 60°N, with smalleraverage doses elsewhere. Hot spots of up to 200rads over areas of - 106 km2 are also predictedfor intermediate time scale fallout. These calcu-lations are consistent with our estimates.

71. H. Lee and W. E. Strope [Stanford Res. Inst.Rep. EGU2981 (1974)] studied U.S. exposure totransoceanic fallout generated by several as-sumed Sino-Soviet nuclear exchanges. Takinginto account weathering of fallout debris, pro-tection by shelters, and a 5-day delay beforeinitial exposure, potential whole-body gamma-ray doses c 10 rads and internal doses ; 10 to100 rads, mainly to the thyroid and intestines,were estimated.

72. These estimates assume normal rates and pat-terns of precipitation, which control the inter-mediate time scale radioactive fallout. In severe-ly perturbed cases, however, it may happen thatthe initial dispersal of the airborne radioactivityis accelerated by heating, but that intermediatetime scale deposition is suppressed by lack ofprecipitation over land.

73. H. Johnston, G. Whitten, J. Birks, J. Geophys.Res. 78, 6107 (1973); H. S. Johnston, ibid. 82,3119 (1977).

74. S. A. W. Gerstl, A. Zardecki, H. L. Wiser,Nature (London) 294, 352 (1981).

75. M. P. Esposito, T. 0. Tiernan, F. E. Dryden,U.S. EPA Rep. EPA-6001280-197 (1980).

76. J. Josephson, Environ. Sci. Technol. 17, 124A(1983). In burning of PCB's, for example, re-lease of toxic polycyclic chlorinated organiccompounds can amount to 0.1 percent byweight. In the United States more than 300,000tons of PCB's are currently in use in electricalsystems [S. Miller, Environ. Sci. Technol. 17,llA (1983)].

77. C.-S. Chen and H. D. Orville [J. Appl. Me-teorol. 16, 401 (1977)] model the effects of finegraphitic dust on cumulus-scale convection.They show that strong convective motions canbe established in still air within 10 minutes afterthe injection of a kilometer-sized cloud of sub-micrometer particles of carbon black, at mixingratios 5 50 ppb by mass. Addition of excesshumidity in their model to induce rainfall resultsin still stronger convection; the carbon dust israised higher and spread farther horizontally,while - 20 percent is scavenged by the precipi-tation. W. M. Gray, W. M. Frank, M. L. Corrin,and C. A. Stokes [J. Appl. Meteorol. 15, 355(1976)] discuss possible mesoscale (- 100 km)weather modifications due to large carbon dustinjections.

78. C. Covey, S. Schneider, and S. Thompson (inpreparation) report GCM simulations which in-clude soot burdens similar to those in our base-line case. They find major perturbations in theglobal circulation within a week of injection,with strong indications that some of the nucleardebris at northern mid-latitudes would be trans-ported upward and toward the equator.

79. R. C. Eagan, P. V. Hobbs, L. F. Radke, J. Appl.Meteorol. 13, 553 (1974).

80. C. Sagan, 0. B. Toon, and J. B. Pollack [Sci-ence 26, 1363 (1979)] discuss the impact ofanthropogenic albedo changes on global climate.Nuclear war may cause albedo changes by burn-ing large areas of forest and grassland; by gener-ating massive quantities of soot which can settleout on plants, snowfields, and ocean surfacewaters; and by altering the pattern and extent ofambient water clouds. The nuclear fires in thebaseline case consume an area = 7.5 x I0W km2,or only = 0.5 percent of the global landmass; itis doubtful that an albedo variation over such alimited area is significant. All the soot in thebaseline nuclear war case, if spread uniformlyover the earth, would amount to a layer -0.5p.m thick. Even if the soot settled out uniformlyon all surfaces, the first rainfall would wash itinto soils and watersheds. The question of the

effect of soot on snow and ice fields is underdebate (J. Birks, private communication). Ingeneral, soot or sand accelerates the melting ofsnow and ice. Soot that settles in the oceanswould be rapidly removed by nonselective filter-feeding plankton, if these survived the initialdarkness and ionizing radiation.

81. In the present calculations, chemical changes instratospheric 03 and NO2 concentrations causea small average temperature perturbation com-pared to that caused by nuclear dust and smoke;it seems unlikely that chemically induced climat-ic disturbances would be a major factor in anuclear war. Tropospheric ozone concentra-tions, if tripled (7), would lead to a small green-house warming of the surface [W. C. Wang, Y.L. Yung, A. A. Lacis, T. Mo, J. E. Hansen,Science 194, 685 (1976)]. This might result inmore rapid surface temperature recovery. How-ever, the tropospheric 03 increase is transient(- 3 months in duration) and probably second-ary in importance to the contemporaneoussmoke and dust perturbations.

82. R. M. Haberle, C. B. Leovy, J. B. Pollack,Icarus 50, 322 (1983).

83. During the martian dust storm of 1971-1972, theIRIS experiment on Mariner 9 observed thatsuspended particles heated the atmosphere andproduced a vertical temperature gradient thatwas substantially subadiabatic [R. B. Hanel etal., Icarus 17, 423 (1972); J. B. Pollack et al., J.Geophys. Res. 84, 2929 (1979)].

84. V. V. Alexandrov, private communication; S.H. Schneider, private communication.

85. G. E. Thomas, B. M. Jakosky, R. A. West, R.W. Sanders, Geophys. Res. Lett. 10, 997 (1983);J. B. Pollack et al., ibid., p. 989; B. M. Jakosky,private communication.

86. P. Ehrlich et al., Science 222, 1293 (1983).87. H. M. Foley and M. A. Ruderman, J. Geophys.

Res. 78, 4441 (1973).88. We gratefully acknowledge helpful discussions

with J. Berry, H. A. Bethe, C. Billings, J. Birks,H. Brode, R. Cicerone, L. Colin, P. Crutzen, R.Decker, P. J. Dolan, P. Dyal, F. J, Dyson, P.Ehrlich, B. T. Feld, R. L. Garwin, F. Gilmore,L. Grinspoon, M. Grover, J. Knox, A. Kuhl, C.Leovy, M. MacCracken, J. Mahlman, J. Mar-cum, P. Morrison, E. Patterson, R. Perret, G.Rawson, J. Rotblat, E. E. Salpeter, S. Soter, R.Speed, E. Teller, and R. Whitten on a variety ofsubjects related to this work. S. H. Schneider,C. Covey, and S. Thompson of the NationalCenter for Atmospheric Research generouslyshared with us preliminary GCM calculations ofthe global weather effects implied by our smokeemissions. We also thank the almost 100 partici-pants of a 5-day symposium held in Cambridge,Mass., 22 to 26 April, for reviewing our results;that symposium was organized by the Confer-ence on the Longterm Worldwide BiologicalConsequences of Nuclear War under a grantfrom the W. Alton Jones Foundation. Specialthanks go to Janet M. Tollas for compilinginformation on world urbanization, to May Liufor assistance with computer programming, andto Mary Maki for diligence in preparing themanuscript.

SCIENCE, VOL. 2221292

on

Janu

ary

24, 2

008

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om