Hypercanes: A possible link in global extinction scenarios

JOURNAL

OF GEOPHYSICAL

RESEARCH,

VOL. 100, NO. D7, PAGES 13,755-13,765, JULY 20, 1995

Hypercanes: A possible link in global extinction scenarios Kerry A. Emanuel,• Kevin Speer,2 Richard Rotunno,3 Ramesh Srivastava, 4 and Mario

Molina

•

Abstract. Bolide impactsand large-scalevolcaniceruptionshave been proposedas possiblecausesof the massiveextinctionof life that has occurredepisodicallyin Earth's history.In spite of the catastrophicdisruptionof the local environmentthat accompanies bolide impactsand volcaniceruptions,it hasbeen difficultto explainwhy theseevents sometimeslead to global extinctionof species.We proposethat, in some cases,the missinglink may be providedby hypercanes,runawayhurricanesthat are capableof injectingmassiveamountsof water and aerosolsinto the middle and upper stratosphere, where they may have profound effectson atmosphericchemistryand radiative transfer. Hypercanesare theorizedto occurwhen the sea surfacetemperatureexceedsa critical threshold,which may occurwhen seawater is locally heated by bolide impact, shallow-sea volcanism,or possibly,by overturningof superheatedbrine pools formed by underwater volcanicactivity.Simulationsusinga convection-resolving nonhydrostatic,axisymmetric numericalmodel showthat hypercanescan indeed developwhen the sea surface temperature is high, and that they inject large amountsof massinto the stratosphere. 1.

Introduction

Evolution of life has been punctuated by events during which large numbers of specieshave become extinct. Largest among the known events was the Late Permian extinction, about 245 m. y. ago, during which about 96% of all species disappeared.The most recent major extinctionoccurredat the end of the Cretaceous,between 64 and 66 m. y. ago, and was marked by the disappearanceof about three fourths of all speciesthat existedat that time. Two hypotheseshaveemerged to explainthe end-Cretaceousextinction.Alvarezet al. [1980] proposedthat the impact of a large asteroidled to the massive extinction,while Officerand Drake [1983]and othersarguethat the extinction was related to massivevolcanic eruptions that resultedin the formation of the DeccanTraps. It hasalsobeen suggestedthat thesetwo geophysicaleventsare related [Altet al., 1988]. Considerableevidencehas accumulatedin recent years in supportof the idea that the Earth was struck by one or more bolides at the end of the Cretaceous.Foremost among such evidenceis the findingbyAlvareze! al. [1980] that iridium and other platinum group metalsknown to be more concentrated in extraterrestrial bodies than on Earth, are enriched in a thin clay layer at the Cretaceous-Tertiaryboundary worldwide. Other evidenceincludesthe findingthat this claylayer contains extraterrestrialaminoacids[Zahnleand Grinspoon,1990],soot attributed to global wildfires [Wolbache! al., 1985], and an Osmium isotopicratio with a distinctlyextraterrestrialsignal

[Luck and Turekian,1983]. Many Cretaceous-Tertiary(K/T) boundary sequencesalso contain abundant shockedmineral grains that suggestan impact, although it has also been proposed that this may result from explosivevolcanism.Recent attention

has focused on the Chixslub

feature

on the northern

Yucatan coastas a possibleimpact crater, thoughseveralother craterscan also be dated to the K/T transition, suggestingthat multiple impactsmay have occurredin a short time. There is alsoevidencein K/T sedimentsof a large tsunamievent in the Caribbean region. Although the evidencefor the impact of at least one large bolide at the end of the Cretaceousis mounting,there remains considerablecontroversyabout just how suchan event might lead to massiveextinction of life. The original Alvarez et al. [1980] paper suggestedthat the hypothesizedasteroid impact would have thrown up a large quantity of dust into the stratosphere,where it would have reduced substantiallythe incoming solar radiation, leading to a dramatic decreaseof photosynthesisand concomitantdisruptionof the food chain. Other means by which the impact might have led to mass extinctionsinclude:(1) dramatic,directwarmingof the atmosphere and/or ocean by the impact [Hst2,1980]; (2) global coolingowingto sootproducedby globalwildfires[Wolbachet al., 1985]; (3) chemical contaminationof the atmosphere and/or upper ocean [Hst2,1980]; (4) globalwarmingby large increasesin atmosphericCO2 owingto the impact on carbonate terrain [O'Keefeand.4hrens,1989];and (5) globalacidrain owing to large increasesin atmosphericsulfur and nitrates [Prinnand Fegley,1987].As an alternativeto the bolide impact hypothesis,large-scalevolcanismhas alsobeen cited as a pos•Centerfor Meteorologyand PhysicalOceanography, Massachusiblecauseof massextinctions.Of particularinterestare flood settsInstitute of Technology,Cambridge. 2Institut Frangais deRecherche pourl'Exploitation dela Mer, Brest, basaltfissureeruptions,which are capableof producingindi-

viduallavaflowswith volumesgreaterthan 100 km3 at very

France.

-3National Centerfor Atmospheric Research, Boulder,Colorado. 4Department of Geophysical Sciences, Universityof Chicago,Chicago,Illinois.

Copyright 1995 by the American GeophysicalUnion. Paper number 95JD01368. 0148-0227/95/95JD-01368505.00

high eruption rates. It is speculatedthat suchlarge eruptions would inject massiveamountsof debrisinto the stratosphere where it might lead to reduced sunlight and depletion of ozone,and that immensequantitiesof acid rain would reduce the alkalinity and p H of the upper ocean. The massextinctionswere evidentlyselective.The weight of

13,755

13,756

EMANUEL

ET AL.: HYPERCANES

AND GLOBAL EXTINCTIONS

evidencesuggests that the most dramaticextinctionsoccurred

among theplanktonic foraminifers [D'Hondt andKeller, 1991] and calcareousnannoplankton[Thierstein,1981], and a major molluskgroupdisappearedentirely[Ward,1988].But diatoms, dinoflagellates,and other dwellersof the upper oceansuffered only modestreductions.The planktonicflora and fauna appeared to undergoa rather suddendecline.There were also large reductionsin land animals,particularlyamongthe larger species(e.g., dinosaurs),but considerablecontroversysurroundsthe questionof the timingof suchreductions,and many smaller animals survived the K/T transition.

Many, although not all, of the proposedmechanismsby whichbolide impactsand/orlarge-scalevolcanismlead to mass extinctionsrely on massiveinjections of material into the stratosphere,where it may remain for manyyears.By contrast, injectionsof large quantitiesof aerosolsinto the troposphere shouldhave only transientand local effects,sinceexcessmaterial iswashedout of the troposphereon a timescaleof weeks. The original Alvarez et al. [1980] thesiscontendsthat large amountsof material from the collidingbolide and from the crustwould find its way to the stratosphere,perhapsby being mechanicallyinjected through the "hole" punchedin the atmosphere by the incoming bolide. To our knowledge, this conjecturehas never been subjectedto rigoroustests. Another conjectureholds that significantamountsof mass entered the stratospherethrough hot plumes generated directly throughthe heat of impact or the combustionof surface material.The volcanichypothesisalsoreliesheavilyon thermal plumesto loft material into the stratosphere,whereit canhave long-term effects.But the altitude to which thermal plumes from areally limited surface sourcesascendin the stratosphere 1

Figure 1. The Carnot cycleof a mature tropicalcyclone.Air spiralsinward from the environmenta to near the center c, while acquiringentropyfrom the seasurface.Air then ascends adiabaticallyand flowsout to large radius,where it eventually losesheat by radiation to spacealong legs3 and 4.

hydrostaticmodel, lending credenceto the idea that hypercanesare physicallypossible.We then discussplausiblescenariosin which hypercanesmight have occurredas a resultof bolide impact or underseavolcanism.We concludeby specu-

latingaboutthe effectsof hypercanes on the stratosphere and how theseeffectsmay have helped causemassextinctions.

increases onlyasthe• powerof thesurface heatflux[Morton 2.

et al., 1956].It is thushighlyunlikelythat globalwildfirescould affect the stratospherein a significantway, even if there were manyof them, sincethe surfaceheat flux per unit area is vastly insutficientto achievestratosphericaltitudes.Moreover, even the very largestof historicaleruptions(e.g.,Krakatoain 1883) managedto achieveonly limited penetrationinto the stratosphere,suchthat the material settledout in a matter of a few years(see,for example,Budykoet al., [1988]). Moreover,the settlingtime of aerosolsin the stratospheremayhavean upper bound, owing to the fact that the very smallestparticleshave diameterscomparableto or smallerthan the meanfree path at thosealtitudes[Reitmeijer,1993]. It is not obviousthat either large-scalevolcanismor bolide impactscould loft enoughmaterial into the stratospherefor long enough periods to have globaleffectsof a magnitudenecessary to explainmassextinctions.

Hypercanes

2.1. Theory

The energycycleof a mature hurricanecan be characterized as that of a Carnot heat engine[Emanuel,1986], as illustrated in Figure1. In its normalstate,the tropicalatmosphereis quite far from beingin thermodynamicequilibriumwith the underlying ocean.This disequilibriumis necessaryto supportthe large rates of heat lossby evaporationneededto balancethe incomingsolarradiation.(Little heat is lostdirectlyby infrared radiation,owingto the opacityof the moistatmosphereabove, leaving evaporationas the principal means of disposingof heat.) The disequilibriumis reflectedin the subsaturationof air just above the sea surface;typical relative humiditiesare about 80%.

As the ambient air spiralsinto the storm center in a thin boundarylayer, the wind speedincreasesand so too doesthe rate of evaporation.In the strongesthurricanes,the air near the corecomescloseto a stateof thermodynamic equilibrium with the ocean,reflectinga large heat gain on its inwardjourney. This heat gain is observedto occur at nearly constant temperature.Having reachedthe core, the air spiralsupward in a ring of convectionknown as the "eyewall," which surroundsa circularregionof nearlycloud-freeair knownas the "eye." This upward motion representsa nearly adiabaticexpansionof the working fluid, which is a mixture of moist air and suspendedcondensedwater droplets.These dropletsultimately combineinto particleslarge enoughto fall out of the systemas precipitation.The air then flowsout at high altitude to large distancesfrom the storm center, after which it loses the heat it gainedfrom the oceanby radiatingit to space,or in

Our presentpurposeis to suggestthat in at leastsomecases of oceanicmassivevolcanismor bolide impact, hypothetical atmosphericstormsknown as hypercanesmay have playedan essentialrole in injectingmaterial into the stratosphere.Hypercanesare extraordinarilyintensehurricaneswhoseenergy productionis so large that it cannotbe balancedby surface dissipation,resultingin stormsthat are sointensethat internal dissipationbecomesimportant.These stormsare hypothesized to occurwhen the degree of air-sea thermodynamicequilibrium exceedsa theoretically defined threshold value. Their circulationswould penetrate to high altitudes in the stratosphere, where they could deposit large quantities of mass, includingwater vapor, condensedwater, and volcanicash. After reviewingthe theory of hypercanes,we presentthe first numerical simulationsof these stormsusing a fully non- an open-cycle system, by •xportingit to otheratmospheric

EMANUEL

ET AL.' HYPERCANES

AND GLOBAL EXTINCTIONS

13,757

RH=75% -100 -90

•

-80

To

Hypercanes

-7O

{ o C ) -60 -50 -40 -30 -20

CONTOUR

20

25

30

35

FROM

40

700

TO

1010

BY

10

45

50

T s (øC) Figure 2. Minimum centralpressure(mbar) of tropicalcyclonescalculatedfrom (1), with ambientsurface relative humidity of 75%. The asteriskdenotesconditionstypical of the tropical Pacific.Normal sea level pressureis 1013 mbar, while the lowestpressureever recordedin a tropical cycloneis 870 mbar.

circulationsystems.Finally, the air descendsback to the surface, losingmore heat by electromagneticradiation.But since the vertical profile of air temperaturein the environmentof the storm is nearly that of adiabatic ascent of moist air with suspendedwater droplets, this additional heat loss is almost exactlythat whichwould have occurredif the condensedwater had been retained and was now reevaporating. If the working fluid is consideredto consistof a mixture of dry air, water vapor, and condensedwater, the energy cycle may be closelyapproximatedby isothermalexpansionof air spiralinginward alongthe oceansurface,followedby adiabatic expansionof air ascendingin the eyewall,followedby isothermal compressionof air losingheat at the storm top, and followed finally by adiabaticcompressionto the starting point. The thermodynamicefficiencyof the engineis proportionalto the temperaturedrop from the seasurfaceto the lower stratosphere,where the outflow occurs.For typical conditionsthis temperature drop is of the order of 100 K, while the inflow temperatureis roughly300 K, givingan efficiencyof aboutone third.

The total increasein enthalpyof the inflowingair is limited by the degreeof air-seathermodynamicdisequilibriumof the ambient air: When this air reachesthermodynamicequilibrium, no further heat input is possible.(Suchan equilibriumis characterizedby equal sea and air temperaturesand air relative humidityof 100%.) This providesan upper bound on the energyinput to the system. In the mature hurricanethe energyinput balancesdissipation. Most of this dissipationoccursin the thin, turbulent boundarylayer abovethe seasurfaceand is proportionalto the total pressuredrop from the ambientenvironmentto the storm center.A smalleramountof energyis lost at large radii in the outflow, where the low angular momentum of the outflow (whichhasbeenpreviously lostto the seain the inflow)must be restoredto environmentalvaluesby turbulence.A yet much smaller,but poorly known, amount of energyis lost in turbulence in the interior of the system. Neglectingthis last sink, the energybalancemay be written as an equation for the central surfacepressureof the storm, which is a direct measureof its intensity.Given the aforementioned bound on the heat input, there is a corresponding

boundon the surfacepressuredrop from the ambientenvironment to the center. Emanuel [1988] developed an implicit equation for this drop, in the form

ln(x) = -(A/x)

+ B,

(1)

wherex is the ratio of the centralvalue of the partial pressure of dry air to its ambientvalue, andA andB are coefficientsthat dependon the ambient degree of air-seathermodynamicdisequilibrium, the thermodynamicefficiency,and the storm radius, which determineshow much energy is dissipatedin the outflow.

Solutionsto (1) are graphedin Figure 2 as a functionof the sea surfacetemperatureand the outflow temperature,assuming a fixed environmental surface relative humidity of 75%. (The surfacerelative humidity is observedto vary little with local climate conditions over the ocean; then the sea surface

temperaturedeterminesthe degreeof air-seathermodynamic disequilibrium.)The asteriskindicatesconditionstypicalof the warmest regions of the tropics. Computationsof this same bound for climatologicallynormal conditions(see, for example, Emanuel, [1986]) showthat a smallfractionof real hurricanescome very close to their energy bound. On the other hand, accuratenumericalsimulationsof hurricanes(see, for example,Rotunnoand Emanuel, [1987]) seemto alwaysproduce storms at the theoretical limit of intensity. Nonaxisymmetric interactions

of real storms with their environments

and

local reductionof sea surfacetemperature,owingto turbulent mixingof surfacewaterswith deeper,colderwater, limit most, but not all, storms to intensities well below the theoretical limit.

Figure 2 also showsthat there are no solutionsto (1) for certainvaluesof the coefficientsAandB. Closeanalysisof the systemshowsthat in the supercriticalregime where no solutionsto (1) are possible,the Carnot cycleis unstable,owingto the the componentof heat input by isothermalexpansion.The lower the central surfacepressure,the more heat is added to the inflow by isothermalexpansion;this addedheat intensifies the vortex, leading to further reductionof the central surface pressure,more heat input by isothermalexpansion,and so on. This effect is so strong in the supercriticalregime that the additionalsurfacedissipationcannotbalanceit, and the vortex

13,758

EMANUEL

ET AL.: HYPERCANES

AND GLOBAL EXTINCTIONS

intensifiesuntil internal dissipationbecomessutficientto balance the energyinput. The resultingsystemis what we have called a hypercane. To date, the only numericalsimulationof a hypercanewas that of Emanuel[1989],who useda very simplemodelconsisting of only three layersof atmosphereand whichmade several restrictingassumptionsabout the force balancesin the flow. This neverthelessshowedthat in the hypercaneregime of parameterspace,the vortexcontinuesto intensifyuntil numerical stabilitycriteria are exceeded,and the integrationhad to be stopped.Here we present results of integrationsin the hypercaneregimeusinga much more sophisticated model.

equationsthat governthe propagationof soundare integrated on a smaller time step than the other terms. This may be considereda specialway of calculatingmasscontinuitythat avoidsthe solutionof an elliptic equationat the expenseof integratingsomeof the terms on a smallertime step. In the presentcasethe effectivemasscontinuityequationis the standard anelasticequation, in which the backgrounddensity, rather than full density,is employed.This can be shownto be a goodapproximationas long asthe Mach numberof the flow is small.In the caseof hypercanes,however,the Mach number is of orderunity,andthe anelasticapproximationbreaksdown. The two approximationsdiscussedaboveare probablynot seriousproblemsbefore and just after the transition of the 2.2. Numerical Simulations model storm to the hypercaneregime, in which the central We performedseveralintegrationsof the numericalhurri- surfacepressureis lower than the critical transitionvalue, but cane model developedby Rotunnoand Emanuel [1987]. It is they become problematicas the hypercaneapproachesfull designedto simulate the axisymmetriccirculationof hurri- intensity.Thuswhile we feel that the model in its presentform canes, including the circulation within convective clouds, showsthat the transitionto the hypercaneis indeeda physical though the latter are forced to be axisymmetricrings. This possibility,the characteristicsof the mature storm shouldbe model integratesthe fully nonhydrostatic, compressible equa- viewed as preliminary. tions on an axisymmetricdomain 1000 km in radiusand 45 km ' With one exceptionthe initial conditionis identicalto that in altitude, with radial and vertical resolutionsof 5 km and 1.25 usedby Rotunnoand Emanuel [1987],consisting of a tropical km, respectively.The model alsointegratesconservationequa- atmosphereand sea surfacetemperaturethat has been adtionsfor heat andwater,but condensed wateris representedby justed so that the atmosphereis effectivelyneutral to moist only one category,the total liquid water mixing ratio. The convection.This mimicsthe real state of the atmospherein terminalvelocityof the falling liquid water is taken to be 7 m whichmoistconvectionis in statisticalequilibriumwith larges-• if thewatercontentexceeds 1 g kg-• andzerootherwise; scaleprocesses,such as radiative coolingof the atmosphere thisis the crudestway of handlingrain while retainingonlyone and surface heat fluxes, that act to destabilize the flow to categoryof condensate.One drawbackto this approachis that convection.Such a state is nearly but not exactlyneutral to evaporationof falling rain mustproceedat a pacesufficientto convection.It is an exampleof self-organizedcriticality.The keepthe air saturated;thisexaggerates the rate of evaporation. stratospherein the initial stateis isothermal. A spongelayeroccupyingthe top 5 km of the domainservesto Onto this statewe superimposea bell-shapedaxisymmetric partially absorbinternal wave activity,which in naturewould seasurfacetemperatureanomalywith a decayscaleof 100 km. radiateinto the highatmospherewithoutmuchbackreflection. The maximumsea surfacetemperatureat the centeris 50øC, All turbulencein the model is representedby an eddy vis- decayingto the ambient value of 27øC at large radii. This cositythat dependson the local rate of deformationand the anomaly is intended to represent the result of local ocean local Richardson number, the latter of which is defined with a warming by bolide impact or underseavolcanism;since its dry or moist static stability,dependingon whether the air is scale is assumedto be much smaller than an atmospheric unsaturated or saturated. Surface fluxes of heat, moisture, and deformationradius,the ambient atmosphereas a whole will momentum are representedusing the classicalbulk aerody- not adjustappreciablyto the anomaly. namic formulae, but with wind-dependentexchangecoeffiAs in the work by Rotunnoand Emanuel [1987], an initial cients (all of which are equal). Radiative coolingis crudely vortexwith maximum wind speedsof 12 m s-• is usedto mimickedby a Newtonian relaxationback to the initial condi- initiate the finite-amplitudeinstabilitythat resultsin tropical tion, with maximumvalueslimitedto 2 K d-•. cyclones.(Later we shallseethat the large temperaturegradiA few approximationsto the equationsthat are considered ents generatedby the sea surfacetemperatureanomalyare minor for simulationsof ordinaryhurricanesmay prove to be sufficientto initiate a storm from very small amplitudeinitial problematicin the hypercaneregime. In the constructionof conditions.) the model equationsthe pressureand potential temperature Figure 3 showsthe evolutionof the centralsurfacepressure are separatedinto partsrepresentingthe backgroundstateand and maximum azimuthal wind with time. After a short time the deviationsfrom it, the former beingfunctionsof altitudealone. initialvortexundergoesrapid intensification, with centralpres-

The full pressure gradientacceleration, givenby -CpOvV•r, is sures lower than 300 mbar at around 25-40 hours. Afterward, replacedin the modelby --½pOv•7'lT , where½pis the heat the vortexweakenssomewhatand approachesa slowlydecaycapacityat constantpressure,Ov is the virtual potential temperature, and •r is the nondimensionalpressureperturbation. The overbarindicatesthe backgroundstatevalue.This approximationneglectsthe quadratictermsin the pressureandbuoyancyperturbations.In normal hurricanestheseperturbations are of the order of 10% of the meanvalues,so the neglectof the quadraticterms is justified. In hypercanes,however,the perturbationsare of the sameorder asthe meanvalues,and so this approximationbreaksdown. Although sound waves are meteorologicallyunimportant, they are explicitlycalculatedin the model. The terms in the

ing statewith very strongwinds.Experimentsin whichthe cap on the radiative coolingrate was removed showthat the resultingstormsreach a nearly steadystate. The slow decayin the control experimentat long time is a result of the inability of the systemto lose heat at a sufficientrate. The spatialdistributionsof variousquantitiesin the mature hypercaneat 50 hoursinto the simulationare shownin Figure 4. Thesefieldshavebeen averagedover a 5-hour period. Surfaceradialinflowin the boundarylayerreachespeakvaluesin

excess of 50 m s-•, whileoutflownearthe tropopause hasa maximum valueof 60 m s-•. Althoughmostof the outflow

EMANUEL

ET AL.: HYPERCANES

280 lOOO

240

Vma x 200 (m s-• )

700 (mb) 160

120

AND GLOBAL EXTINCTIONS

13,759

the transition from hurricane to hypercane,suggestingthat hypercanesare physicallypossible.The experimentssuggest that sea surfacetemperaturesof greater than about 45øCoccurringon scalesgreaterthan about 50 km may be capableof supportinghypercanes. It remainsto be seenwhetherseasurface temperatureanomaliesof this size and magnitudeare capableof spontaneousignitionof hypercanes,or whether an independentdisturbanceor "trigger" is necessary,as is certainly the casewith ordinaryhurricanes.To help answerthis question,we ran an experimentidentical to the control but beginningwith a warm core vortex with maximumwinds of

only2 m s-•. Thisinitialvortexamplified to thesameintensity

80

asthe controlsimulationwith only a minor delay,showingthat the atmospherictemperaturegradientsassociated with the sea 20O surfacetemperature anomaly are sufficientfor spontaneous generationof storms. 0 0 40 50 60 We have assumedthat the ambient atmosphereis calm and Time (hours) that it has the thermodynamiccharacteristicsof the present Figure 3. Evolutionof the maximumazimuthalvelocity(sol- tropical atmosphere.But significantvertical wind shear is id line) and minimumpressure(dashedline) in the main ex- known to inhibit or prevent the developmentof hurricanes, and there is no reasonto suspectthat this is not the casewith periment. hypercanes,whoseearly evolutionis similarto that of hurricanes.At the very least,the presenceof meanwind, even if it occursnear the tropopause,someconvectionin the eyewall is does not vary with altitude, will move the storm acrossthe reachesaltitudesgreaterthan 30 km, and weak outflowcan be ocean surface.In order for hypercanesto develop,the mean detectedthere. The large penetrationdepthsare achievedbe- wind would have to be small enoughrelative to the size of the causeof the very high specificentropycontentsreachedby air sea surfacetemperature anomalyto permit the storm to despiralinginto the core. Figure 4d showsthat the equivalent velop in the time it takesto crossthe anomaly.Given a develpotentialtemperature(a quantitywhosenatural logarithmis opment time of 40 hours,mean windswould have to be less proportionalto the total specificentropy) of air rising in the thanabout1 m s-• to allowhypercane development overa eyewallhasvaluesin excessof 700 K. Suchlarge valuesresult 100-km-scalesea surface temperature anomaly. Such condifrom the nearly isothermalexpansionof air flowing into re- tionsare unusual,but not unknownin the deep tropics.But it gionsof very low pressurein the eye. is unlikely that hypercaneswould developover localizedsea The azimuthalwind componentis shownin Figure 4b. The surfacetemperatureanomaliesof the kind likely to havebeen maximum value of 220 m s- • occurs at the surface at a radius producedby bolide impactor volcanismif the meanwindsare of 6 km, and the cycloniccirculationextendsup to about 35 km much strongerthan this. Given weak but nonvanishingtropoaltitude. sphericwinds, we conjecturethat a sequenceof hypercanes At 50 hoursthe verticalvelocity(Figure 4c) reachespeak would develop,with eachmemberdevelopingrapidlyover the valuesin excessof 35 m s-•, thoughat other timesvalues sea surfacewarm anomaly,moving away, and decayingwhile 40

greaterthan 70 m s-• occurred.Theseare comparable to vertical velocities

achieved in the most severe middle

another

forms.

latitude

thunderstorms,and are about an order of magnitude greater than are typical in normal hurricanes.Descent occursat the 3. Environmental Effects of Hypercanes innermostgrid point (the eye),andweakdescentcharacterizes The most significantcharacteristicof hypercanes,from the standpointof environmentalimpact, is their ability to inject most of the outer region of the simulatedstorm. The water vapor mixing ratio (Figure 4e) showsextreme large amountsof massinto the middle stratosphere,where it valuesof over50 g kg-• in the core,reflecting the lowpartial may remain for manyyears.If we assume,on the basisof the pressureof dry air there. Condensedwater mixing ratioswell numericalsimulations,that a ring of air between5 and 35 km

over10gkg-• extend wellupintothestratosphere (Figure4f). Severalexperimentswere performedto testthe sensitivityof the hypercaneevolutionto the size and magnitudeof the sea surfacetemperatureanomaly;theseare summarizedin Figure 5. As the geometricsizeof the seasurfacetemperatureanomaly decreases,the maximumwind speeddecreasesbut is still very substantialwhen the decayscaleis only 25 km. The mature circulation

is somewhat

steadier when the scale of the sea

surfacetemperatureanomalyis small.When the peakvalue of the ocean temperatureis reduced,the maximumwind speed falls but is still substantialwhen the peak sea surfacetemperature is 42øC.

radiusis ascending at a rateof 5 m s- • at an altitudeof 20 km (seeFigure4c), then the net massflux into the middlestrato-

sphereisabout10•økgs-•. Thisisenough to replacetheentire massof the atmospherebetween the 100- and 50-mbar pressure surfacesin about 6 months. Two aspectsof this mass transportare of particularinterest:(1) the flux of water substance,which has the potential of radicallyaltering radiative transferthroughthe stratosphereand of affectingatmospheric chemistryand (2) the flux of aerosols,which also may alter radiative

3.1. The

transfer.

Stratospheric Water

critical issue here is how much water ascends in the Although the approximationsin the model physicsrender suspectsome features of the mature hypercanecirculation, main hypercaneupdraft without precipitatingout of the systheseapproximationsshouldnot be of great importancenear tem. The minimumabsolutetemperaturein the updraft occurs

13,760

EMANUELET AL.:HYPERCANES AND GLOBALEXTINCTIONS

5O

TIME 3o .........

__:_-:_--:::::::::::_-_-_-•_-•i';;;

b

a 15.0

,'

,.'

30,

d

Iv

0

rodius

300

0

radius

300

Figure4. Distributions ofvariables at50hours intothemainexperiment ina domain extending radially to

300kmandvertically to30km.(a)Radial velocity inmeters persecond withcontour interval of10ms-•. (b)

Azimuthal velocity inmeters persecond withcontour interval of 10ms-•. (c)Vertical velocity withcontour interval of2 ms-•. (d)Equivalent potential temperature withcontour interval of25K. (e)Watervapormixing ratioin grams perkilogram withcontour interval of 3 g kg-•. (f) Liquid watermixing ratioin grams per kilogram withcontour interval of 3 g kg-•. condensed water wouldbe in the ice phase;a criticalissueis how much of the icewouldbe in the form of verysmallcrystals tropopause, whereoutflowin normalhurricanes is concenlongenough to affectthe trated,thistemperaturecorresponds to a saturation waterva- thatmightremainin thestratosphere por mixingratioof about30 ppm,but at 26-kmaltitudethe globalheat balance. near the altitude of the outflow; this is about -70øC. At the

In anairparcelundergoing rapidverticallifting(say,greater saturation mixingratioiscloserto 100ppm,owingto thelower than 30 m s-•), the condensate doesnothavetimeto growto total pressure. Thusassuming that the updraftis saturated, somewhatmore water vapor is presentin the hypercaneup- precipitationsizes;therefore,almostall the water vapor at thecloudbasewillbedeposited in thestratosphere draft.Onepeculiarity of thehypercane environment isthatthe present saturation mixingratioactuallyincreases with altitudein the in the formof smalliceparticles.Also,in a strongupdraftthe with respectto waterpreventnearlyisothermal stratosphere; thusanyhypercane effluent air mayremainsupersaturated and depositing on to ice thatbegins to sinkisothermally through thelowerstratosphereingwaterdropsfrom evaporating a process characteristic of the Bergeron-Findeisen underthe influenceof radiativecoolingwill experiencea de- particles, of precipitation initiation.In an air parcelundercreaseof saturationmixingratio with time.Thusfurthercon- mechanism goinga slower rateof lifting,someof thecondensate mayhave densationof watermayoccurin the descending air. tofalloutof theparcel; therefore only Of potentiallyfar greaterimportance is the flux of con- timeto growbigenough densed phasewaterat highaltitudes in thehypercane updraft. a fractionof thewatervaporpresentat thecloudbasemaybe at highlevelsin theformof smalliceparticles. At the low temperatures of the outflowregion,all of the deposited

EMANUEL

ET AL.: HYPERCANES

AND GLOBAL EXTINCTIONS

10s I[

13,761

i • i •,=5 i 10i6 • •i

o'[

10 s 10

o

20

40

60

80

lOO

Tlrne (hours)

Figure 5. Evolution of the maximum azimuthal velocity in five experiments.Curve A, the main experimentwith a maximum temperatureof 50øCand decayscaleof 100 km; curveB, maximum temperature of 47øC and decay scale of 100 km; curveC, maximumtemperatureof 42øCand decayscaleof 100 km; curveD, maximumtemperatureof 47øCand decayscaleof 50 km; and curveE, maximumtemperatureof 47øCand decay scale of 25 km.

What happensto small ice crystalsdepositedin the middle stratosphere? Let usbeginby estimatingthe averagesizeof the ice particlesdepositedin the stratosphere.Cloud microphysical calculationsshowthat when an unmixedair parcel is lifted beyond the saturation level, the air becomessupersaturated; the supersaturationreaches a maximum at a short height above,and thereafter it decreasesslowlywith continuedlifting of the parcel (see,for example,Rogersand Yau [1989]). In an unmixedair parcel, in which condensationis the only process operating, the concentrationof cloud drops (expressedas numberof particlesper unit massof air) is determinedat the point of the maximum supersaturation,and thereafter it remains constantwith height; furthermore, with continuedcondensationthe size distributionof the cloud dropstends to be monodisperse.In this approximationthe particle radiusr, the mixingratio M, and the particleconcentrationn are related by

M = (4rrp,/3)nr3,

(2)

where Pi is the densityof the ice particles.For a rapidly lifted parcelthe ice mixingratiomaybe takenequalto the saturation water vapor mixingratio at the cloud baselevel (considering that there is no fallout and that the saturationmixing ratio is smallin the stratosphere); for a slowlylifted parcelthe "average" radiusmay be calculatedfrom (2), providedthat M is taken to be the actual mixing ratio of the condensatein the stratosphereand that we have an estimateof the particle con-

l

• • • • •,•••

lOO

VERTICAL

AIR VELOCITY•

M/S

Figure 6. Cloud particle concentrationas a function of vertical air velocityfor three nucleusspectra.

sation nuclei; that is, he assumed that the concentration of

nucleiactivatedat supersaturation cris ao•3,wherea and/3are constantsthat depend on the characteristicsof the cloud condensationnuclei. In (3), f• dependsupon the cloud basetemperature and pressurewhile f2 is a function of/3. The power law is in fair agreementwith observations;a and/3 are found

to rangefrom about20 to 3000cm-3 and0.4 to 1.2,respectively, when cr is expressedin percent [Braham, 1976]. Maritime aerosols have smaller values of a, while continental and

polluted aerosolshave highervalues.There is also a tendency for/3 to be larger for maritime aerosols.For our purposesthe predominant effect is the dependenceof N on a and the vertical air velocityw. Figure 6 showsthe cloud particle concentration as a function of the vertical air velocity for three nuclei spectra.We see that the cloud particle concentration

variesfrom aboutl0 6 g-• to 107 g-•. Oncethe particlesare depositedin the stratospherethey can settleslowly,and if the ambientair is unsaturatedwith respectto the particlestheycan evaporatewhile settling. We now considerthe settlingtime of the particles.First let us consider the settling time in the absenceof vertical air motion and particle evaporation.Except for the very largest particlesconsideredabove,the particle terminal fall speedv is given by the Stokesformula:

v = (2pi#/9rl)r2,

(4)

where # is the gravitationalaccelerationand r• the dynamic viscosityof air. Calculationsshowthat particlesof radii 3, 10, 30, and 60 tzm will take about 2216 hours (92 days), 199.4 hours(8.3 days),22.16 hours(0.92 days),and 5.55 hours(0.23 days),respectively,to settle through 10 km. Thus these particles can reside in the stratospherefor time periods ranging centration. from a day to months if sedimentationis the only process Twomey(1959) derivedthe followingapproximateequation operatingto remove the particles. for the concentrationN of cloud drops: However, as updraft air comes to rest in the stratosphere and mixeswith the ambientair, or as the particlessettle out of N = a2/(13 + 2)f, f2(13)w313/2(13 + 2), (3) the parcel into the ambient air, they will tend to evaporateif where w is the vertical velocityof the air. Twomey assumeda the ambient air is unsaturated.The rate of changeof particle power law spectrumof the critical supersaturationof conden- massis m, given by

13,762

EMANUEL dm

4.n'rDPs

ET AL.: HYPERCANES

LD

--=--S,1+3'

•-- X- p;,

(5)

where D is the diffusivityof water vapor in air, Ps is the saturationvapor densityin the ambientair (with respectto

ice), L is the latentheat of sublimation, k is the thermal conductivity of air, p'sis the derivativeof Pswithrespectto the

AND GLOBAL EXTINCTIONS

I os 1 04 -

1 03 -

temperature,and S is the supersaturationwith respectto the

particle. (Equation (5) involvescertain approximations; it is accurate only for small magnitudesof S. However, for our purposesit shouldbe adequateevenfor considerablesubsatu-

102 -

rations.) Foranapproximate calculation weassume constant S

1 01 -

and constantD, Ps, and 3';for the isothermalstratospherethe latter is a goodapproximationexceptfor D, whichis inversely proportionalto the pressure.With theseapproximations, (5) gives

10ø 1 01

=

2Dps

1 02 INITIAL

RADIUS

(Micrometer)

Figure 7. Fall distanceto completeevaporationasa function

St,

r2 r•+ (1+ 3') Pi

(6)

where r o is the initial radius.For subsaturatedconditionsthe time *e to completelyevaporatea particle of radiusr o is given by

(1 4- 3')lOi

of initial radius for three relative humidities.

middle stratospheremightbecomecloudy,with profoundconsequences for globalclimate.Even if the condensedwater were removedfrom the stratosphereon fairly short timescales,the

re= 2Dps1- •i

(7)

saturated

h•=18r• Dps 1- •i'

(8)

radicals.

air would

remain.

The injectionof largeamountsof water into the stratosphere for the chemistryof that where •i is the ambient relative humiditywith respectto ice. may have significantconsequences A particle of initial radius less than 60 /xm will evaporate region.Water vapor is the sourceof the free radicalsOH and completelyin lessthan 5.6 hoursin an ambientrelativehu- HO2, which contribute to stratospheric ozone depletion midity of 99%, assumingT = 220 K andp = 25 mbar. Since throughthe followingcatalyticcycle: the diffusioncoefficientis inverselyproportionalto the presOH + 03 --• HO2 + 02 sure,the time to completeevaporationwouldbe doubled(quadrupled)if we had usedthe value of D for p = 50 mbar (100 HO2 + 03 • OH + 2 02 mbar), keepingT - 220 K. The time to completeevaporation Net: 2 03 -• 3 02 is alsoinverselyproportionalto Ps; thereforeusinga temperature lower than 220 K will tend to increasethe evaporation This is the reaction couplet that contributesmost to ozone time, not only becauseof a decreasein Ps, but also because destructionin the lower stratosphere;it is followedin imporwith pressurefixed,D decreaseswith decreasingtemperature. tanceby catalyticcyclesinvolvingnitrogenoxidesand halogen If we consider sedimentation and evaporation simultafree radicals.The OH radical plays also another important neously,(4) and (6) can be used to derive the followingex- role: It activateschlorine (by convertingthe relativelystable pressionfor the distanceof fall in whichthe particleevaporates HC1speciesto C1atoms)and deactivates nitrogen(by convertcompletely(assumingno vertical air motion): ing nitrogendioxideto nitric acid,a more stablespecies),the net effect being enhanced ozone depletion by chlorine free Pig (1 + 3')p, r• Figure 7 showsthe evaporationdistanceas a functionof the initial radiusfor three selectedrelativehumidities.(Again,we haveused T = 220 K, p = 25 mbar; the aboveremarkson the

effectsof changing pressure andtemperature on % applyto hœ aswell.) It shouldbe notedthat the Stokesfall velocitylaw on which (8) is basedbreaksdown for particle radii exceeding about50/xm. We seethat a particleof initial radius50/xm can evaporatecompletelyin a fall through 2 km in a relative humidity of 99%. Thuswe seethat mostof the condensatelofted high into the stratosphereis likely to evaporatein the stratosphere and increaseits vapor content, until and unlessthe stratospherebecomessaturated.

Assuming that1 g kg- • of watersubstance exists in thering of convectionas it entersthe middle stratosphereand that the updraft characteristics are thosedescribedabove,then the flux

of waterintothemiddlestratosphere is about107kgs-•. This is sufficientto completelysaturatethe layer between 100 and 50 mbar in about 20 days.Were this to happen,the lower and

An important separate effect of water on stratospheric chemistrycouldresultfrom the formationof clouds:Chemical reactionson cloud droplets activate chlorine and deactivate nitrogenoxides,in a manneranalogousto that describedabove for the OH radical.Sucha mechanismexplainsthe formation of the Antarctic ozone hole. Furthermore, over Antarctica, precipitation of the cloud particles leads to the irreversible removalof nitrogenoxides,becausethey scavengenitric acid. This removal setsthe stagefor efficientozone destructionby halogenfree radicals,a processthat slowsdown in the presence of nitrogenoxides. Stratosphericclouds are, however, rather scarce and are normallypresentonly over the poles in the winter or spring months,which is where ozone depletion has taken place in recentyears.These observationsindicatethat the presenceof stratosphericcloudsat low latitudescouldhave profoundconsequencesfor ozone depletion. The main natural sourcefor chlorinein the stratosphere is methylchloride,producedin the biosphere;if additional amountsof chlorine were to be in-

EMANUEL

ET AL.: HYPERCANES

AND GLOBAL EXTINCTIONS

13,763

jected, for example,from seasalt, the effectson stratospheric ten rock to be insulated from seawaterby a thin solid crust, ozone could be dramatic. effectivelyreducingthe heat transfer rate. However, they also describethe probable intimate mixing of hot debriswith sea3.2.

Aerosol Injection

water as the ocean rushes back into the crater. Furthermore,

althoughthey discounta large-scale,ocean-wideheating, becauseof the tendencyfor rotational constraintsto confinethe hot column locally, it is preciselythis confinementwhich encouragesthe high sea surfacetemperature(SST) needed for hypercaneformation. To investigatethe natureof this rotationalconfinementwith strongheating,we considera less dramatic event: a volcanic eruptionwithin the deep Red Sea hot brine pools.This providesuswith a real, existingnatural laboratoryfor testingideas aboutthe mixingand confinementof superheatedseawater. The brine pools are reservoirsof hot, salty fluid which accumulate around geothermalsourcesat the floor of the Red Sea. Their extremesaltinessturns them into huge heat storage 4. Bolide Impact, Undersea Volcanism, devicesin which temperaturecan build up, sincetheir density and Hypercane Formation remainshigh enoughfor them to rest on the bottom.Typically, The numerical simulations described in section 3 suggest the observedtemperature is above 60øCand salinityis above that hypercanesmight occur over localized regionsof water 250 practicalsalinityunits (psu) [Ross,1983], givinga density deepwater,at 22øCand heatedto peak temperaturesin excessof about 45øCon scales of about1200kg m-3. The overlying greater than about 50 km. Such regions could only initiate 40.5 psu,hasa densitycloseto 1030kg m-3. To makethe hypercanesif the ambientatmosphericconditionswere other- dense brine buoyant, its temperature must be raised above wisefavorablefor hurricaneformation;that is, the atmosphere 300øCby volcanicactivity.An estimateof brine volume in the would have to have nearly moist adiabaticlapseratesthrough Atlantis Deep Basin,Chain Basin,DiscoveryBasin,and nearby is 7 km3 [Blancet al., 1990;J.-L.Charlou,permost of the troposphereand could not contain much vertical smallerbasins wind shear.In addition,the meanwind speedwould haveto be sonal communication,1994]. The possibilityfor a substantial small, particularly in the case of small-scalesea surfacetem- volume of water to become superheatedbefore becoming that even after dilutionwith the surrounding perature anomalies,so that the developingstorm would not buoyantsuggests move rapidly awayfrom the seasurfacetemperatureanomaly. seawaterin the rising plume, the final temperature anomaly All of these conditionssuggestthat hypercaneswould be may still be very high. To explore the dynamical possibilitythat this hot brine extremely rare, occurring only when the ocean surface is heated to an extraordinarydegree somewherein the tropics, reachesthe surfacein somediluted form, we adapted a threeand then only under ideal conditions. Solar heating is not dimensionalnonhydrostaticnumericalmodel. This model has capableof raising sea surfacetemperaturesto the necessary been describedand applied to convectionin the Mediterranean Sea by Jonesand Marshall [1993] and to hydrothermal degree,and thus other heat sourcesmust be considered. In the hypotheticalimpactscenarioenvisionedbyEmiliani et plumesby Speerand Marshall [1995].A nonhydrostaticmodel al. [1981],a 14-km-diameter bodyofmass2.5x 10•skghitsthe is necessaryin our caseto resolveexplicitlythe processeswhich Earthwitha speed2 x 104m s-•. Onlya tinyfraction(0.1%) producedilutionand mixingwith strongverticalmotion and to aboutentrainment.Earth's rotation of the initialkineticenergy,5 x 1023J, is lost asthe bolide avoidmakingassumptions movesthrough the atmosphere;the principal transfer to the constrainsthe lateral expansionof the plume in the sea and is atmosphereoccursvia the plume thrown up by the impact, expectedto influencethe long-termevolutionof the volcanic estimatedat 25% of initial kinetic energy.The remaining75% plume. The domainis a periodicbox 32 km x 32 km x 2000 m, with is availableto excavatethe impactcrater in the oceanand solid earth and to heat the rock and seawater.It is thoughtthat such heatingdistributedover a disk8 km in diameterat the bottom, an impact would create a crater roughly 100 km in diameter representinga volcaniceruptionin the basins.The grid is 132 and 35 km deep at its center, exposingunderlyingmantle. As x 132 x 20, for a grid size of 250 m in the horizontal and 100 the ocean rushes back over the rim to fill in the crater over a m in the vertical. Initially, there is a layer of hot brine one grid period of hours, it is heated mainly by contactwith hot or box thick on the bottom with neutral buoyancy;that is, it is molten rock. This direct heatingdependscruciallyon the de- assumedthat heatinghasalreadyraisedthe temperatureof the gree of heat exchangebetween the hot debris, upwelling brine to the point of convectiveoverturning.The horizontal mixingcoefficients are all set to 5 m2 s-•, and the vertical magma, and seawater. Subtractingthe energygoinginto the plume, the total energy coefficients are 0.2 m 2 s-•. These small values ensure that available fromthe collisionis 4 x 1023J, including thatliber- mixing occursmainly by advection. To estimate the forcing strength, an eruption documented ated from the warm crust and mantle [Emiliani et al., 1981]. on the Juan de Fuca Ridge cleft segmentis usedas a reference Using this energyto heat seawaterin a layer 4 km thick (4 x

The largeverticalvelocitiesand massflux ratessustainedin hypercanesare capableof injectinglarge amountsof aerosol into the middle stratosphere.Bolide impact might generate large amountsof fine dust by excavation,while underseavolcanismproducesash. The injectionof volcanicash or terrestrial material excavatedby bolide impact into the middle stratospherecan be expectedto influenceclimate for several years to a decade, given the observedpresenceof volcanic material in the stratospherefor severalyears following major historicaleruptions.The largest of historicaleruptionshave led to measurableglobal cooling[Budykoet al., 1988].

10•3m3),theoceantemperature in thecrater100km in diam- [Foxet al., 1992].In thiscase,0.05km3 of magmaerupted covering approximately 2.1km2 of seafloor. A totalheattrans-

eter could rise by 1000øC,that is, create a column of boiling water with sea surface temperature near 100øC (higher at depth owingto pressure). Emiliani et al. [1981] noted that direct heatingof the ocean by lava outflowis attenuatedbecauseof the tendencyfor mol-

fer of about 1.5 x 10TMJ m -2 occurred as this lava flow cooled

to ambient temperature.For a thicknessof order d = 10 m,

the timescale for coolingis of orderd2/4•(, giving107 Sfor a conductivity K = 10-6 m2 s-•. Then the heat flux into the

13,764

EMANUEL

ET AL.: HYPERCANES

AND GLOBAL

EXTINCTIONS

integration.Runs with similarstratificationand sourcegeometry showedsimilar behavior. The buoyancyflux from the source,thoughlarge compared to inactiveridge crest fluxes,is simplytoo weak to drive the plume acrossthe thermocline.(Note that the largestvolcanic plumes in the atmosphereeasily penetrate the stratosphere [Sparks,1986].The analogousthermocline,the tropopause,is only roughlydouble the stratificationin the troposphereand doesnot appear to form much of a barrier.) A penetration scalebasedonbuoyancy fluxis (B/Ntherm) 3 1/2, whichis only 12 m. A source1000 times strongerwould appear to be strong enoughto penetratethe thermocline;the total sourcestrength

wouldthen be 3.2 x 10is W, placingit amongthe largest Figure 8. Simulationof a Red Sea hot brine plume created by a volcanicheat sourceon the bottom. The domainis 32 km on a sideand 2 km deep.Simulationshowscontoursof salinity (solid lines) at an interval of 2.5 psu from 60-70 psu and density(dashedlines) at an interval equivalentto iøC. The correspondingmaximum temperature anomaly is roughly 40øC.The brine risesoff the bottom as a plume up to the base of the thermocline,where it spreadslaterally.After a period of about 1 day the plume is circulatinganticyclonically (clockwise). After a few daysthe plume reachesthe edge of the

eruptionsobservedon land [Settle,1978;Wilsonet al., 1978].In the presentmodel configuration,suchlarge forcingis impossible to simulate

as it leads to numerical

instabilities.

We con-

clude that the dynamicalconstraintof rotation is important to form a stronganomalyand that theseanomaliescan attain the surfacein the caseof very powerful, but not unprecedented, volcanic

sources.

We note parenthetically that isopycnalsat intermediate depthsoutcrop at the northern end of the Red Sea, so if the subsurfacehot plume is carried north in the general circuladomain. tion, it is conceivablethat it may eventuallyreach the surface followingisopycnals.It is interestingthat even modesteruptionsare capableof forminga hot brine plume at intermediate oceanisof order104W m-2 onaverage. To represent a strong depths. On the other hand, a straightforwardway for volcanismto eruption, a heatfluxvalueofH = 50 kW m-2 waschosen. The heat the surfacelayer of the oceanis for it to occurin shallow equivalent buoyancy fluxisB = #aH/pcp-- 4.5 x 10-s m2 s-3, andthe totalbuoyancy inputintegrated overthe source water, as in the first stagesof islandformation.To raise the temperatureof an oceansurfacelayer50 m thick and 50 km on area is F = 2700 m4 s-3.

The stratificationof the Red Seais relativelyweakbelow500

mdepth, where thebuoyancy frequency Ndeep ---7 x 10-4 S-1, but there is a strongthermoclinein the upper 500 m, where

a side(125km3) from25øCto 50øCrequires1.3 x 1019J. To supplythisheat over an eruptiontimescaleof 10 hours[Wilson

etal., 1978]impliesa thermalsource of 3.4 x 10TM W. At 100%

Nther m ----7 x 10-3 S-1. Usinga Coriolisparameter appropri- efficiencyof heat transferfrom magmato seawater,the corremagmaflowrateis about6 x 104 m3 s-• (assuming atefor midbasin (latitude20øN),f = 5 x 10-s s-1, thelarge sponding pCp -3 x 106J m3øC,aninitialtemperature of 1000øC, and ratioNtherm/f ----140 or Ndeep/f: 14 shows thatstratification and not rotation controlsthe height of penetration.The pen-

an equal heating contributionfrom latent heat of freezing).

implyhigherflowrates,but evenhigherflow etrationheightis thereforeZ•v = 3.8(F/N3)TMabovea vir- Lowerefficiencies tual source several kilometers

below the real seafloor. For the

aboveforcingandNdeep , Z N -- 6400 m, sufficient to reachthe baseof the thermocline.At the baseof the thermoclinea point sourceapproximation is no longerappropriatefor scalingpurposes.

A numericalsimulationof an eruptioninto the brine pools showsthe generationof a hot brine plume at intermediate depthsin the Red Sea (Figure 8). Numericalconstraintsprecluded the use of realisticbrine temperatureand salinity,so salinitywas used as a proxy tracer for mixing;the final temperatureof the equilibriumplume as determinedby salinities of 40-100 psuis roughlyin the range25-50øC.Over the rising and spreadingperiod, most of the heat in the plume comes from the brine pools and not the source. This situation is similar to that of the megaplumeswhich have been observed abovethe midoceanridge in the Pacific Ocean [Bakeret al., 1989].After one day of integrationthe radial motionof water toward the sourceat depth and away from the sourceat the spreadinglevel has generateda vortex, anticyclonicat intermediatedepthsand cyclonicin deep water, owingto rotation. The vortex holds together a lens of warm brine above the source. Despite this confinement and recirculation of the plume, it still doesnot reach the surfaceafter severaldaysof

rates are not unknownin existingvolcanoes[Settle,1978]. The natureof the eruptionwill determinewhetheror not the heat transferefficiencyis high. Explosiveeruptionsare more effectiveat mixing pyroclastwith surroundingfluid; slower eruptionshave to cover a large area with thin layersof lava in

orderto generate fluxesof the order104W m-2, capableof adequatelyheatingthe overlyingwater columntensof meters deep in a period of days.Thus high efficiencyis not necessary for a stronganomaly,but the quicker the anomalyis formed the lesschancethat it is dissipatedinto the backgroundbefore a stormdevelops.The aboveestimatesshowonly the plausibility of a volcanicallyproduced SST anomaly adequateto generatea hypercane.The actual probabilityof satisfyingthe various conditions

is difficult to determine.

Finally, it shouldbe emphasizedthat even the large air-sea heat fluxesinducedby hypercanegenerationare nevertheless far too smallto erasethe sea-surfacetemperatureanomalyin a short time. The maximum surfacefluxes producedby the numerical

simulation

would cool a 100-m column of seawater

at the rate of no more than IøC per day. The averagefluxes over the warm pool would cool the same column by about 0.1øCper day. Thus it would take severalweeksto reducethe seawatertemperatureto below the hypercanethreshold,even

EMANUEL ET AL.: HYPERCANES AND GLOBAL EXTINCTIONS

13,765

if therewere no continuingheatfluxesinto the oceanfrom the volcanicor bolide impact-inducedsource.

Emiliani, C., E. B. Kraus, and E. M. Shoemaker, Sudden death at the end of the Mesozoic, Earth Planet. Sci. Lett., 55, 317-334, 1981.

5.

J. Geophys.Res., 97, 11,149-11,162, 1992. Hsii, K. J., Terrestrialcatastrophecausedby cometaryimpact at the

Summary

Fox, C. G., W. W. Chadwick,andR. W. Embley,Detectionof changes in ridge-crestmorphology usingrepeatedmultibeamsonarsurveys,

end of Cretaceous,Nature, 285, 201-203, 1980. The twoprincipalscenarios for rapidglobalextinctions, bolide impactand extensive volcanism, both rely to somedegree Jones, H., and J. Marshall, Convection with rotation in a neutral ocean:A studyof open-oceandeepconvection, J. Phys.Oceanogr., on stratospheric effectsto produceworldwideecosystem stress. 23, 1009-1039, 1993. We regardwith skepticismclaimsthat the mechanicaleffectsof Luck,J. M., andK. K. Turekian,Osmium-187/osmium-186 in mangabolideimpactwoulddirectlyleadto large-scale contamination nesemodulesand the Cretaceous/Tertiary boundary,Science,222, 613-615, 1983. of the stratosphere or that the heatingeffectsof suchan impact or the wildfiresthat result from it could lead to significant Morton, B. R., G.I. Taylor, and J. S. Turner, Turbulentgravitational convection from maintained and instantaneous sources,Proc. R. injectionof massinto the stratosphere.Similarly,the weak Soc. London A, 234, 1-23, 1956. dependence of the maximumpenetrationaltitudeof turbulent O'Keefe, J. D., and T. J. Ahrens,Impact productionof CO2 by the

convectiveplumes on the surface heat flux casts doubt on whether even unusuallylarge volcaniceventscould disrupt climate much beyond what has been observedin historical eruptions.We proposeinsteadthat someinstancesof undersea volcanismor oceanicbolideimpactcouldleadto the formation of hypercanescapableof injectinglarge amountsof material into the middle stratosphere.The contaminationof the normally dry stratospherewith water may have large transient effectson climate through the direct absorptionof infrared

Cretaceous/Tertiary extinctionbolide and the resultingheatingof the earth, Nature, 338, 247, 1989.

Officer, C. B., and C. L. Drake, The Cretaceous-Tertiary transition, Science,219, 1383-1390, 1983.

Prinn, R. G., and B. Fegley,Bolide impacts,acidrain and biospheric traumas at the Cretaceous/Tertiaryboundary,Earth Planet. Sci. Lett., 83, 1-15, 1987.

Reitmeijer,F. J. M., Volcanicdustin the stratosphere between34 and 36 km altitude during May, 1985,J. Volcanol.Geotherm.Res.,55, 69-83, 1993.

Rogers,R. R., andM. K. Yau,A ShortCoursein CloudPhysics, 293pp.,

Pergamon,Tarrytown, N.Y., 1989. radiationby water vapor and throughthe radiativeeffectsof large-scalestratosphericcloudiness,and may lead to serious Ross,D. A., The Red Sea,in Estuariesand EnclosedSeas,editedby B. H. Ketchum, pp. 293-307, Elsevier, New York, 1983. depletionof ozone and thus to increasedultraviolet radiation Rotunno, R., and K. A. Emanuel, An air-seainteractiontheory for at the surface.The injectionof large quantitiesof aerosolinto tropicalcyclones,II, J. Atmos.Sci., 44, 542-561, 1987. the middle stratospheremay also have strongeffectson cli- Settle, M., Volcaniceruptioncloudsand thermal power output of mate.

The necessaryconditionsfor the formation of normal hur-

explosiveeruptions,J. Volcanol.Geotherm.Res.,3, 309-324, 1978. Sparks,R. S. J., The dimensionsand dynamicsof volcaniceruption columns,Bull. Volcanol.,48, 3-15, 1986.

ricanesleadusto concludethat hypercaneformationwouldbe Speer,K., and J. Marshall,Convectionfrom seafloorhot springs,J. limitedto the tropics,evenin the caseof very large anomalies Mar. Res., in press,1995. of seasurfacetemperature.If our hypothesis hasany merit, it Thierstein,H. R., Late Cretaceousnannoplanktonand the changeat suggeststhat global extinctionswill be somewhatmore frequentin the caseof oceanicbolide impact or underseavolcanismin the tropics.Thispredictionmayprovidesomemeansof empiricallytestingthe theory.

the Cretaceous-Tertiary boundary,in TheDeepSeaDrillingProject: A Decadeof Progress, editedbyJ. Wamme,E. L. Winteres,andR. G. Douglas, Spec. Publ. Soc. Econ. Paleontol. Mineral., 32, 355-394, 1981.

Twomey,S., The nuclei of natural cloudformation:The supersaturation in naturalcloudsandthe variationof clouddropletconcentration, Geofis.Pura Appl., 43, 243-249, 1959. References Ward, P., Maastrichtianammoniteandinoceramidrangesfrom Bayof BiscayCretaceous/Tertiary boundarysections,in Paleontology and Alt, D., J. M. Sears,and D. W. Hyndman,Terrestrial maria: The Evolution:ExtinctionEvents,edited by M. Lamolda, E. Kauffman, originsof largebasaltplateaus,hotspottracksand spreadingridges, and O. Walliser, pp. 119-126, Soc. Espan.de Paleontol.,Madrid, J. Geol., 96, 647-662, 1988. Alvarez, L. W., W. Alvarez, F. Asaro, and H. V. Michel, Extraterres-

trial cause for the Cretaceous-Tertiaryextinction, Science,208, 1095-1108, 1980.

1988.

Wilson, L., R. S. J. Sparks,T. C. Huang, and N. D. Watkins, The controlof eruptioncolumnheightsby eruptionenergetics and dynamics,J. Geophys.Res.,83, 1829-1836, 1978.

Baker, E. T., J. W. Lavelle, R. A. Feely, G. J. Massoth,and S. L. Wolbach, W. S., R. S. Lewis, and E. Anders, Cretaceous extinctions: Walker, Episodicventingof hydrothermalwatersfrom the Juan de Evidence for wildfires and search for meteoric material, Science, Fuca Ridge,J. Geophys.Res.,94, 9237-9250, 1989. 230, 167-170, 1985. Blanc, G., J. Boul•gue, and J.-L. Charlou, Proills d'hydrocarbure 16gersdansl'eau de mer, les saumureset les eauxinterstitiellesde la Zahnle, K., and D. Grinspoon,Comet dustas a sourceof aminoacids at the Cretaceous/Tertiaryboundary,Nature, 348, 157, 1990. fosseAtlantis II (mer Rouge),Oceanol.Acta., 13, 187-197, 1990. Braham,R. R., CCN spectrain C-k space,J. Atmos.Sci.,33, 343-345, K. A. EmanuelandM. Molina, Centerfor MeteorologyandPhysical 1976. Oceanography,Massachusetts Institute of Technology,Cambridge, Budyko,M. I., G. S. Golitsyn,andY. A. Izrael, GlobalClimateCatas- MA 02139. trophes,99 pp., Springer-Verlag,New York, 1988. R. Rotunno, P.O. Box 3000, National Center for AtmosphericReD'Hondt, S., andG. Keller,Somepatternsof planktonicforaminiferal search, Boulder, CO 80307. assemblage turnoverat the Cretaceous/Tertiary boundary,Mar. MiK. Speer,Institut Frangaisde Recherchepour l'Exploitationde la cropaleontol.,17, 77, 1991. Emanuel,K. A., An air-seainteractiontheoryfor tropicalcyclones,I, J. Atmos. Sci., 42, 1062-1071, 1986.

Emanuel,K. A., The maximumintensityof hurricanes,J. Atmos.Sci.,

Mer, B.P. 70, 29280 Plouzan6, France.

R. Srivastara,Departmentof GeophysicalSciences, The University of Chicago,Chicago,IL 60637.

45, 1143-1155, 1988.

Emanuel,K. A., The finite-amplitudenatureof tropicalcyclogenesis, J. Atmos. Sci., 46, 3431-3456, 1989.

(ReceivedSeptember8, 1994;revisedMarch 8, 1995; acceptedMarch 22, 1995.)