JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 102, NO. D16, PAGES 19,575-19,584, AUGUST 27, 1997 The susceptibility of ice formation in upper tropospheric clouds to insoluble aerosol components PaulJ. DeMott, David C. Rogers,andSoniaM. Kreidenweis Departmentof Atmospheric Science,ColoradoStateUniversity,FortCollins Abstract. Ice may form by bothhomogeneous andheterogeneous freezingnucleationprocessesin cloudsat temperatures below-35øC.Most investigations havefocusedon the former process.This paperpresentsresultsfrom adiabaticparcelmodelcalculations that includethe effectsof bothfreezingprocesses in unactivatedsolutiondroplets.Uncertainties in predicting thehomogeneous freezingratesare discussed andusedto selectsolutiondropcomposition and freezingcharacteristics thatbracketthoseexpectedin theuppertroposphere. The heterogeneousfreezingratesof insolubleatmospheric aerosolsareparameterized basedonpublished freezingratesof carbonaceous particles.Processmodelsimulations showthatthepotential variabilityin ice formationin cirruscloudsis muchgreaterif heterogeneous freezingnucleationis considered in additionto homogeneous freezing.The impactof insolubleaerosolson ice formationis inferredto increasewith insolubleparticlesizeandwith thefractionof soluble aerosolscomabringinsolublecomponents. The maximram impactof heterogeneous nucleation isindicated forvertical motions lessthan0.2ms-•andforinsoluble components beingassociatedwith at least 10% of all solubleaerosols.The wide rangeof ice crystalconcentrations observedin cirrusis mostconsistent with the occurrenceof bothheterogeneous andhomogeneous ice formationprocesses. Theseconclusions are partiallysupported by existingobservations of aerosolsandcloudmicrophysical characteristics in uppertropospheric cloudsbut requirenew measurements for confirmation. 1. Introduction 2. Ice Formation Processesin Upper Tropospheric Clouds Cirruscloudsgenerallyform at temperatures below -30øC andat relativehumidity(RH) fromwell belowsaturationwith 2.1 HomogeneousFreezing Nucleation respectto water (at coldertemperatures) to water saturation Homogeneous freezingnucleationrefersto the spontaneous [e.g., Heymsfieldand Miloserich, 1995]. Widespreadcirrus formationof ice within the metastableliquid phase.The form midlatitudesare producedby gentlelifting processes with mationof ice in pure or highly dilute cloud dropletsis exlarge-scale updraft velocities of 1to20cms•. Morelocalized pectedto occur over a narrow range of temperaturesfrom cirrus-generating regions maycontain updrafts upto 1.5m s-• about -35ø to -38øC [see, e.g., Sassenand Dodd, 1988; [Gulteppeet al., 1995]. At the warm end of the cirrustemHeymsfietdand $abin, 1989].Thispredictionandthe validity peraturespectrum,any existingliquid clouddropletswhich of a classicalapproachare supportedby a large numberof areloftedand cooledareexpected to freezehomogeneously at laboratorystudiesof the freezingof purewaterandverydilute around-36øC. At lower temperatures,unactivatedsolution solutiondroplets[see, e.g., œruppacher,1995]. At temperadroplets(hazeparticles)areexpected to freezehomogeneously tures below the fleezing transitionzone for activatedcloud at a RI-I below water saturation.In contrast,particlesthat act droplets,unactivatedsolutiondropletscan freezein a manner as heterogeneous ice nuclei(IN) are thoughtto be presentin akin to the pure homogeneous process.However, elevated concentrations much lower than cloud condensation nuclei solute concentrations lower the freezing rates of solution (CCN). Consequently,heterogeneous processeshave been droplets,suchthat the transitiontemperaturesat which nuthoughtto be of lessimportance in cirrus.Thispapercritically cleation rapidly ensues are depressed in proportion evaluatesthispremisebasedon availableknowledge. (nonlinearly)with increasedsolutionmolality.TheoreticalinThis paper examinesthe potentialfor heterogeneous ice vestigationshave elucidatedthe dynamicsof homogeneous nucleationby immersionfreezingof solutiondropletsat trofreezingof hazeparticlesandthe microphysical consequences pospherictemperatures below -38øCto impactice formation within upper tropospheric(UT) clouds [e.g., $assen and in cirrusclouds.This problemis analyzedusing theoreti- Dodd, 1988, 1989;HeymsfieMand Sabin, 1989;HeymsfieM cal/empiricalcalculations of homogeneous andheterogeneous and Mitosevich, 1993, 1995; DeMott et al., 1994; densen et freezingnucleation.The expectedsensitivityof cloudmicro- at., 1994a,b]. physicsto naturalaerosolpropertiesis considered. Data on For thispaperthe adiabaticparcelmodelwithin whichcalaerosolandcloudmicrophysical characteristics arethenexam- culationsof homogeneous freezingwere madeis the one de- inedconsidering thesecalculations to determine whichproc- scribedby DeMott et at. [1994]. Solutiondropletsof a given essor combination of processes bestfit the observations. size and molalityare assumedto have effectivefreezingtemperatures [SassenandDodd, 1988]givenby Copyright1997 by the AmericanGeophysical Union. Papernumber97JD01138 0148-0227/97/97JD-01138509.00 Teff= T + )• ZSTm 19,575 (1) 19,576 DEMOTT ET AL.' ICE FORMATION IN UPPERTROPOSPHERICCLOUDS where T is haze droplettemperatureand ATtois the equilibrimn meltingpointdepression, bothin degreesCelsius.Effective freezingtemperature is thenusedas solutiondroplettemperaturefor substituting into the expression for the classical nucleationrate for pure water [cf. DeMott et al., 1994, Eq. (3)]. Anotherapproachwould be to explicitlycalculatethe effect of soluteon the classicalfreezingnucleationrate by considering the theoreticaltemperaturedependence of water activity[e.g.,Jensenet al., 1991]. However,equation(1) reflectsthat solutiondropletstypicallysupercool an amountin additionto that predictedby their equilibriumfreezingpoint depression[e.g., Hoffer, 1961' Pruppacherand Neiberger, 1963,Rasmussen,1982]. For ammoniumsulfate,Sassenand Dodd [1988] suggested that 3, = 1.7. A polynomialexpressionfor ATtoof (NH4)2SO4 as a functionof solutionmolality (34) was determinedas AT, n= •f'•i cij•i (2) The coefficients c,are givenin Table 1. Thesecoefficients differ fromthoseusedby DeMott et al. [1994] because theirsecond-orderpolynomialwas onlyvalid to aboutM = 2. Higher solutionconcentrations occurat the lower temperatures and humiditiesconsideredin this paper.The new expression for ATto was derived based on a combination of tabulated data [Weast,1981] andequilibriumfreezingpointscalculatedas in Robinsonand Stokes[1965], usingrecentwateractivitydata for (NH4)2SO4from Tang and Munkelwitz[1994]. Calculations otherwisefollow Deg/Iottet al. [1994], includingexpressionsfor othercritical quantities.Solutiondropletswere assumedto be in equilibrium.Droplet sizesand molalities were inferred from the well-known Kohler equation.This givesthe maximumgrowthand dilutionof solutiondroplets expectedat anyRH, a reasonable assumption for slowvertical is extremelyrapid.The particlesfreezeas solutiondropletsat a relativehumidityof 94.9%. Depletionof watervaporby ice crystalgrowththenexceedsthe production rateby expansion cooling so that nucleation ceases to occur. Thus onlyabout0.5 -3 cm of theCCN solutiondropletsfreeze. Theoretical studieshavealsodemonstrated thathomogeneous freezingof solutiondropletsis characteristically quite sensitive to temperature andverticalvelocity.Coldertemperaturescausefasternucleationrates.Increasedverticalvelocity increasesthe maximumparcelhumidityleadingto greater droplet dilution and increasedfreezing rates. The dashed "fence"in Figure2 outlinesmaximumrelativehumidityand maximumice crystalconcentration observedin 25 numerical simulationsfor (NH4)2SO4aerosols(3, = 1.7). Simulations were initializedat temperatures rangingfrom -38ø to -54øC andconstant vertical velocities ranging from1 to50cms4. It is seenthat solutiondropletsare predictedto freezehomogeneouslyat progressively lowerrelativehumidities(belowwater saturation)as temperaturedecreases.Sassenand Dodd [ 1989]usedthisapparentrelationshipto proposea simplelinearregression(for climatemodeluse)for the thermodynamic conditions requiredfor formingcirrusclouds. Uncertainties in the amountsof ice formedby homogeneous freezingand the relativehumidityrangeat which the processensuesmay be gleanedfrom previousliterature.It is importantto note theseuncertainties becausethey affectthe frequency,persistenceand radiative properties of cirrus clouds.For givenverticalmotionsand thermodynamic conditions, at least three factorslead to uncertaintiesin the amounts of ice formedby homogeneous fleezing and the humidity rangeat which this occursin the atmosphere. Theseare the nonequilibriumeffects of different solutes on freezing [Pruppacherand Neiberger, 1963], the actual solutecommotions and RH less than 95%. positionand the aerosolsizedistribution.The valueof 3,used To compareand contrastto casesincludingheterogeneous for ammoniumsulfate(3, = 1.7) is basedon the observednunucleation,we review resultsthat characterizethe sensitivities cleationtemperatures, or nonequilibriumfreezingpoint deand uncertainties of the homogeneous freezingprocess.The pressions,of a numberof both ionic and nonionicsolution curvesin Figure1 demonstrate theevolutionof thermodynam- emulsions(Rasmussen, 1982). Nevertheless, rigorousexperiics andice formationthat typifyrisingair parcelscontaining mentalinferencesto 3, havenot beenmadefor (NH4)2SO4nor solubleaerosolsat low temperatures. The three simulations for othersubstances of relevenceto the atmosphere. To demshownwereinitializedat -46øC.The initial relativehumidity onstratethe impactof this uncertainty, a simulationassuming was 81% (assumeddeliquescence point). Vertical velocity ideal ammoniumsulfatesolutions(3, = 1) is also shown in was constantat 10 cm s'. Dry aerosolsizewas basedon the Figure 1 (long-dashed curves).This assumption lowersthe CCN watersupersaturation spectrum ([CCN]= 200 S1'5) peak relativehumidityin the air parcelsbut only lowersby given by Heymsfield and Sabin [1989] for UT conditions. 10% themaximumice crystalconcentrations nucleated. Simulationresultsfor the caseof homogeneous freezingonly The potentialuncertaintyintroducedinto calculationsdue by (NH4)2SO4aerosols(3, = 1.7), shownby the thick solid to uncertaintyin solutecompositionis investigatedby also curvesin Figure 1, demonstrate thatthe onsetof ice formation performing calculations of homogeneousfreezing for Table1. Polynomial Coefficients forSome Critical Parametem Solute Parameter (NH4)2SO4 ATm H2SO4 ATm 1-12SO4 aw co +0.02199 . H2SO4 H2SO4 I• +0.99901 * Coeflidents fromChen[1994] c• C2 C3 C4 +4.12 -1.33055 +0.66822 -0.12445 +3.513627 +0.471638 +0.033208 +0.02505 +2.561806x10'2 +1.076046 x 10'2 +6.184438 x 104 +4233388x 10'5 +0.18993 +0.31201 -0.07029 +0.004896 +0.06212 -0.00268 C5 +0.00832 DEMOTT ET AL.' ICE FORMATION IN LIPPER TROPOSPHERIC CLOUDS -46 lOO 95 -47 ? = 85 -48 E ..,. *• 80 75 -49 7O o 1000 500 1500 2000 2500 3000 Time (s) 500 a • 450 • 400 .o 350 • 300 u 250 o o 200 • 150 o lOO _.u 50 500 1000 1500 2000 2500 3000 19,577 pressionin degreesCelsius was determinedfrom Weast [1981] and a fit to (2) was determinedwith the coefficients listed in Table 1. The expressions for p, and AT,, accurately representtabulateddatafor H2SO4weightpercentsfrom 0 to about35%. This coversmostof the rangerelevantto the upper troposphere. Much more guidanceis availableon the choiceof L for sulfuricacid solutions.Ohtake [1993] mademeasurements of actualfreezingpoint depression of bulk volumesof sulfuric acid solutions.He found that the freezingpoint depression was almostindistinguishable from ATto(k.: 1) for solution weightpercentsbetweenabout30 and37% but exceeded AT,, by up to 1.5timesat lowerweightpercents (L: 1.5).Freezing temperaturesof 5 gL H2SO4 solution drops measuredby Beyeret al. [1994] from about25 to 35 wt. % H2SO4are also in excellentagreementwith the sum of the freezingtemperature of pure dropsof this size plus publishedequilibrium fleezingpointdepression (e.g.,L: 1). Their freezingtemperaturesat lower solutionweightpercents,suggesting L < 1, may indicateheterogeneous effectsunder those conditions.More recently,Bertram et al. [1996] measuredthe freezingratesof solutiondropletsizesof H2SO4.A. Tabazadeh et al. (A model descriptionfor cirrus cloud nucleationfrom homogneous freezingof sulfateaerosols, submitted to Joumalof Oeophysical Research,1997) inferthat the Bertramet al. measurements areconsistent with a loweringof thefreezingpointdepression belowthe equilibriumvalueof only3øC,constant acrossthe rangeof solutionconcentrations measured. Therefore we will assumeL: 1 for sulfuricacidthroughout the restof this pahumidity andicecrystal per.Thisshouldgivetheminimumimpactof solutioneffects Time (s) Figure1. Temporal evolution of•ture, concentration in threenumerical simulations of adiabatic air FarceIs on freezingof solubleatmospheric particles. rising at10cms4.Onlyhomogeneous freezing nucleation isallowed Resultsof a simulationassumingsulfuricacid (L: 1) as in simulations of (a) (NH4)2SO4 CCN fieezir•attheirestimated nonequilibrium fi'eeh• points(Z = ].7 in (1)), (b) (NH4hSO4CCN at theirequmbrium freeingpoints½ = and(c) H=SO CCNfiew_Jng attheirequilibrium freehng points. T•mres varied slightly in thedifferent simulations, butonlyoneprofileis shownfor thesakeofclarity. theCCN composition arepresented in Figure1 (short-dashed curves).It is seenby comparingthesecurvesto the 3,: 1 curvesfor ammoniumsulfate,that the degreeof sulfateneutralizationis inherentlya minor factorin homogeneous freezlOO H2804/H20 solutiondroplets.The sameinitial dry size distribution is assumed as for ammonium -381 sulfate. For calculations of dropletsize, interfacialsurfacetensionat the solution/air interfacewas determinedas a function of temperatureand molalityfor pureH2SOa/H20as (3) CoefficientsCi are given in Table 1. This parameterization followsPruppacherand Klett [1978] and usesdata givenby Weast[1981]. It is valid for data compiledat T > -40øC and M < 4.2 but is usedoutsidetheseboundsin this study.Water activity(a,•)versusmolalitywasbasedon a formsuggested by Homogeneous-4:2•_. -- -- 95 -r ot/a= 76.10 - 0.155T+ Y',,c,M' 5 10 ,..21L•:50 90 n• E •0 E ß• 85 -5( 80 Chen [1994]. This is a,•= l/(l + •-]iciJ•') (4) This semiempirical formulationof a,• was foundto agreequite well with the data of Rard et al. [1976]. The temperaturedependenceof water activitywas not considered. Molality effectson solutiondensity(p,) weredeterminedbasedon Weast [1981] as P,= •']iCiM i (5) with the c, coefficients givenin Table 1. For computation of effectivefreezingtemperature, equilibriummeltingpoint de- 0.1 I 10 100 1000 10000 MaximumIce CrystalConc(L'•) Figure2. Maximtan icecrystal concentration versus relative humidity in 50 simulations ofairparcels initiated atthe deliquescence humidity 1.5 ofammonium sulfate CCN ([CCN]= 200& ) fork = 1.7in Eq. (1). Initialtemperatures (degrees Celsius) areindicated bythickcmwes and vertical motions (centimeters persecond) aregivenbythethincurves. Dashedcurvesassume ice formation by homogeneous freezhag of CCNsolution &opsonly,whilesolidcurves assrune alsothatallsolutiondrops contain heterogeneous fieez• nuclei whichrepresent 10ø,4 of CCN mass. 19,578 DEMOTT ET AL.' ICE FORMATION IN UPPERTROPOSPHERIC CLOUDS ing nucleation.For the equivalentassumption on 3, sulfuric aciddropletsare shownto freezeat 91.2% RH anda temperature of-47.5øC, while ammoniumsulfatedropletsfreezeat 91.3% RH at -47.6øC. Jensenet al. [1994a] statedthat they observeda similarresultusingtheirmodel. The solutecompositionmay be an issueas concernsthe deliquescence humidity.In particular,the deliquescence humidityof ammoniumsulfateis not knownfor cirrustemperatures. Chen [1994] has theoreticallyprojectedthat smaller (Nm4)2SO 4 aerosolsmay not even deliquescebelow water saturationat -40øC. Nevertheless,Chen did note that if these of maximumice crystalconcentrations by factorsof up to 3.5 at higher updrafts (100cms'l).Thiscouldonlybe accomplishedby varyingCCN spectraoverextremerangesthathave neverbeenobserved.A somewhatlargerrangeof maximum concentrations could be achieved if it was assumed that CCN concentrations wereon occasiondepletedto valueslessthan 1 cm'3 [Heymsfield andMilosevich, 1995].Jensen andToon [1992] calculatedthat the 2 ordersof magnitudeincreasein aerosolslargerthan 0.5 [tm that can occurin the uppertropospherefollowingmajor volcaniceruptionsmight enhance ice concentrations by a factorof 5. Theypredictedthisto occur aerosolshavebeenthrougha humidificationcyclepreviously, onlyfortemperatures below-55øCandupdrafts near5 cms4. theywill existas supersaturated solutiondropletsat RH val- We noteherethat conditionspresentin the casestudyof Sasues below ice saturation. Sulfuric acid aerosols have no such sen et al. [1995], in which the authorsspeculated that vollimitation,sotheyshouldbe morediluteat RH below90%. canicCCN increasedice concentrations in a cirruscloudby a Since the exact compositionof UT sulfateaerosolsand factor of 3 or more, did not meet thesecriteria. To summarize,we notefrom previousliteratureand calcutheir variabilityin spaceand time are still uncertain,we consider computationsmade for ammonium sulfate solution lationspresented that 1) uncertainties in the composition and dropletswith 3, = 1.7 and sulfuricacid solutiondropletswith nonequilibrium freezingbehaviorof solubleatmospheric par• = 1 to bracketthe potentialhomogeneous freezingbehavior ticlesaffectmainlythe onsethumidityat whichcirrusformby freezingprocess,and 2) the expecteduncerof atmosphericaerosols.The dashed"fence" in Figure 3 a homogeneous showsthe result of homogeneous freezingcalculationsfor tainty in the concentrationof ice crystalsformed due to sulfuricacid solutionsin the sameparameterspaceas shown changesin aerosolsizedistributionis < 50% exceptin very for ammoniumsulfatehazedropletsin Figure2. It is seenthat specialsituations.Theseresultswill be comparedand conice formation is extended to lower RH values for the same trastedwith the impactof considering the addedprocessof simulation conditions. heterogeneous freezingnucleation. The sensitivityof ice formationby homogeneous freezingto solubleaerosolsize distributionhas beenextensivelystudied, 2.2 HeterogeneousIce Nucleation sincethis is an importantaerosol-climate interactionissue. While thepotentialroleof heterogeneous icenucleiON) in Previousstudiesare consistentin predictingthat veryparticuice formationin UT cloudshasbeenstated[Sassenand Dodd, lar changesin the sizedistributionof the solubleaerosolspe1988;Detwiler, 1989;HeymsfieMand Sabin, 1989;Heymsciesdeliquescing arenecessary to haveany significantimpact fieM andMilosevich,1995;Jensenet al., 1994a,b], derailed on the maximum ice crystalconcentrations formed.For exanalysesof IN effectson cirruscloudformationhavenot been ample,Jensenand Toon[1994] haveshownthat increases of presented. One reasonfor this is that heterogeneous ice fortypicalCCN concentrations at all sizesby a factorof 20 inmationprocesses havenot shownthemselves to be readily creasedmaximumice crystalconcentrations by abouta factor tractable through the application of classical theory 1.5 in the extremecase.DeMott et al. [ 1994] notedsimilarre[Pruppacherand Klett, 1978]. The surfacephysicaland sultswhenaerosolconcentrations largerthan0.5 lamwereinchemicalrequirements for IN are complexand poorlyundercreasedby an orderof magnitude.In both instances,the exstood.It hasbeenrecognizedthatIN maybe typedaccording treme caseswere for updraftvelocitiescharacteristic of the lowervalues expected in cirrusclouds (<10cms'l).Heyms- to a numberof hypotheticalseriesof processesor mechanisms:deposition, condensation freezing,immersionfreezing fieM and Sabin [ 1989] simulatedCCN-influencedincreases andcontactfreezing[see,e.g., Vali, 1985].In considering ice formation for conditions existing in most UT clouds, the lOO deposition andimmersionfreezingprocesses appearto be the 5 10,• 20,,•5•. onesof mostimportance. Deposition,the formationof ice di-38 •' 95 rectlyfromthe vaporontoa particlesurface,is onlyrelevant for "dry"andinsolubleparticles.Immersionfreezing,thenu-42 "-- "-- •' ' cleationof a liquidparticleby an insolubleparticlewithin it, Homogeneous ' • •" -r 9O -46 • is relevantfor liquid solutiondroplets.With deepconvective E -50 '. transportof cloud water into the uppertroposphere, other E processes might becomeimportant,althoughbelow -38øC ß85 homogeneous freezingwill efficientlyconvertany remaining -3 0 waterto ice in strongupdrafts. 10 80 DeMott et al. [1994] treatedthe caseof depositionIN in -4(• wlHeterogeneous the uppertroposphere. If presentin concentrations that have -5 been measured in boundary layer air, then ice formation could 75 be dominated by deposition nucleiin cloudswith verticalmo0.1 I 10 100 1000 10000 , ß• . , ß . ........ i ........ i , , , ..... ! , ....... i MaximumIceCrystalConc.(L4) , ..... ,, tionslessthanabout 0.5m s4. If deposition IN wereassumed to be depletedat higheraltitudes,thena complexcompetition processand homogeneous freezFigure3. AsinFigure2,butformlfia/caeidAvater aerosols asstaned betweenthis heterogeneous tofieezeassolution droplets attheirequilibrium fieefi• point&pres- ing was inferred.The effectof the presenceof depositionIN sions(L = 1 in (1)) wasto decrease thenumbersof nucleated icecrystals by up to DEMOTT ET AL.' ICE FORMATION IN LIPPERTROPOSPHERIC CLOUDS an order of magnitude.densenand Toon [1994] found the same result when nonspecificheterogeneous ice formation was startedat 5% ice supersaturation in numericalmodelcalculations.The decreases in ice crystalconcentrations are due to the vapordepletioneffectsof ice crystalgrowthat low ice supersaturations in coolingand humidifyingair parcels.The existence of evenlow numbersof deposition nucleiin the upper tropospherehas not been validated. Also, observations that particlesin many regionsof the uppertroposphere and lower stratosphere are primarilyliquid in regionsof higher humidity[Pueschelet al., 1994] suggesta weakpotentialrole of depositionIN. This studywill assumethe lack of deposition IN in the uppertroposphere and focuson the potential roleof immersionfreezingnucleationin cirrusclouds. Some proportion,perhapsa large proportion,of atmosphericaerosolparticlesconsistof a mixture of water soluble and insolublecomponents. Thesemixed aerosolswill become solutiondropletsandmay becomethe nucleifor naturalcloud formation.Laboratorystudieshavebeenperformedon freez- 19,579 In (6),a = 1.04x 10'4(notethatthere wasa typographical error in the exponentof a givenby DeMott [1990]), b = 7.767, rpis equivalent radius,andT, is supercooling below0øC.This expression quantifiedresultsto -34øC.Homogeneous freezing of thefreelysuspended 5 •tm clouddropletsoccurred at lower temperatures. The resultsimply that the populationof active sitesincreaseswith increasingsootsurfaceareaand decreasing temperature.The freezingfractionwas not found to be time dependent in the experimental temperature range.These resultsare typicalof mostobservations of freezingnuclei.In contrast,classicaltheorypredictsthat all particlesof a given size and chemistrywill nucleateat the sametemperature if given enough time. Thereforean empirical approachto parameterizing freezingby sootaerosolsis preferredfor this studybecausea classicalcalculationwould overestimatethe heterogeneous freezingcontribution. The heterogeneous freezingefficiencies givenby (6) were extrapolated for unactivatedsolutiondropletsat coldertem- peratures by settingT, = -Teff.An exampleof resultsfor an ing initiatedby insolubleparticlesin larger(10 to 2000 •tm) equivalent simulation to theonegivenin Figure1 is shownin solutiondroplets[Hoffer, 1961;Pruppacherand Neiberger, Figure4. The CCN composition and sizedistributionin these 1963; Reische!and Vali, 1975]. These studiesdemonstrated numerical simulations were the same as for the simulations that dropswould freezeat higher temperaturesthan without for homogeneous freezingnucleationalone. It was also asinsolublecomponents. Thishasbeena majorcomplication in sumedthat every CCN containedan insolubleparticlethat studyingthe freezingof larger volumesof "pure" water represented10% of its mass. It is seen that this liberal as[Langhamand Mason, 1958; Pruppacherand Neiberger, sumptionon the availabilityof heterogeneous freezingnuclei 1963]. Certainclaymineralshavebeenidentifiedas particu- leadsto a dominance of ice formationby the heterogeneous larly efficientfleezingnuclei [Hoffer, 1961]. Althoughthe process,independentof the assumptions made on solution concentrations of highlyefficientfreezingnucleimaybe small composition andZ. A factorof 10 fewerice crystals arepro- in theuppertroposphere dueto precipitation scavenging, any ducedin the simulations includingheterogeneous freezing. insolubleparticlemayactas a freezingnucleusat highlysu1 O0 -46 percooled temperatures. For example,DeMott [1990] showed 95 that sootparticlesbeginto act as heterogeneous freezingnuclei in cloud dropletsat temperatures below about -25øC. -47 o O Thusit is reasonable to assumethat someinsolubleparticles '• 90 will also cause unactivated(mostly submicron)solution s5 dropletsto freezeat a highersolutionconcentration thanthey .>_ wouldfreezehomogeneously. • 80 The possibleinterplaybetweenhomogeneous and heterogeneousfreezingprocesses in cirrusconditionswas examined 75 by incorporating a parameterization of heterogeneous freezing nucleationinto numericalcalculations. The parameterization 7O o is basedonDeMott's[1990]dataon thefreezingefficiencyof -48 E -49 500 lOOO sootparticles immersed in clouddroplets. Sincesootis a very inefficientfreezingnucleus,this parameterization is a conservativeonefor representing theactionof atmospheric insoluble species by heterogeneous freezing.Nevertheless, it is a potentially relevantonesincesootis a knowninsolublespecies in the uppertroposphere and hasan anthropogenic sourcethere in theformofjet aircraftemissions. Theactualcomposition of fleezingnucleiin the uppertroposphere is unknown.We describetheparameterization for heterogeneous freezingin this sectionand characterize its impacton ice formationin the adiabaticparcelmodel. The followingsectionthen investigatesthe impacton ice formationof associating different amountsand sizesof insolubleaerosolcomponents with the CCN distribution. 2000 2500 3000 2000 2500 3000 Time (s) 5OO • 450 --. 400 ._o350 ß• 300 • 250 o (.) 200 '• 150 r,.) lOO b _.u 50 o , o In theexperiments onheterogeneous freezingnucleation by 15oo 500 1000 1500 Time (s) sootparticles,DeMott [1990] fit the fractionsF of immersed Figure4. As in Figure1, butheterogeneous freefir•nucleation of sootparticlesfrozento therelationship, solutionsis includedin simulationsas •bed F= a 4Irrp 2(Ts) b (6) in section2.2. The in- soluble heterogeneous fieehr•nuclei areassaaned torepresent 10ø,4 of themassofeveryCCNparticle. 19,580 DEMOTT ET AL.' ICE FORMATION 1NUPPERTROPOSPHERIC CLOUDS sults from 45 selectedsimulationsfor (NH4)2SO4 (•, '- 1.7) hazeparticlesthat includebothheterogeneous andhomogeneousfreezingprocesses. The fractionsof CCN of all sizescontainingan insolublecomponent(representing 10% of particle mass)were varied at decadalvaluesfrom 0.0001 to 1. Maximum ice crystalconcentrations formedfor initial parceltemlowerRH at whichheterogeneous freezingensues. Furtherresultsfor the maximum influenceof the heteroge- peratureof-38 ø, -46ø and -54øC and verticalvelocitiesof 1, in Figure5. Simulations were neousfreezingprocesson ice formationas a functionof com- 10 and50 cms'• areshown conditions but arenot shown.The right position,temperature, and verticalvelocityare indicatedby donefor intermediate resultsdominatedby heterogethe thick solidline "fences"in Figures2 and 3. Theseresults borderof the figurerepresents whereasthe left sidedepictsresults demonstrate that if all aerosolsin the uppertroposphere were neousfreezingnucleation, freezing.It is seenthat the presmixed aerosols,then the actionof a heterogeneous freezing dominatedby homogeneous nucleationprocessin additionto homogeneous freezinglow- ence of 1 insolubleparticlein every 100 CCN can have a nucleated,but erstherangeof maximumicecrystalconcentrations in all cir- markedinfluenceon ice crystalconcentrations per secruswithvertical motions lessthan50 cms-]. Forupdraftsonlyfor cirrusascentratesbelowa few centimeters lessthan20cms-] therangeofmaximum icecrystal concen- ond.Fractions_>0.1 of CCN containinginsolubleparticlesare trations is about10to 2000L-• fora homogeneous freezing necessaryto affectice formationat all ascentratesbelow 50 These form soonerduring parcel ascent,suppressing the maximumhumidityachieved.The onsetof ice formationis earliestand at the lowesthumidityfor the assumption of sulfuric acid CCN composition. This occursbecauseH2SO4solution dropletsgrow more quicklythan (NH4)2SO4at the that,foranyascent rate,thelowering of processactingalone.In contrast,ice crystalconcentrationscms-].It isalsoseen by heterogeneous nucleation is greatrange fromabout1to500L-•whena heterogeneous process is ice crystalconcentration Homogeneous freezingnucleation included.Theprimaryimpactof uncertainties in thechemical est at highertemperatures. composition (solublecomponent of aerosols) andfreezingbe- becomesthe more dominantprocesswhen ascentratesexceed arebelow-50øC. hahor (•,) is on the humidityrequiredfor the onsetof ice for- 50cms-•andtemperatures An equivalent series of simulationsto thosein Figure 5, mation. The differentpositions,breadth,and slopesof the of sulfuric solidfencesin Figures2 and 3 demonstrate that ice formsat but assumingthat the hazeparticlesare composed the lowestRH, and over the widesthumidityrangethe more acid (•, = 1), are shownin Figure6. Resultsare quitesimilar hygroscopic is the soluteand the closeris its freezingpoint to thosefor (NH4)2SO4aerosols,althoughthe impactof heterogeneous freezingnucleationis strongerfor H2SO4.If 10% depression to theequilibriumvalue. of theparticlescontainan insolublecomponent thatrepresents just 10% of theirmass,thenthe heterogeneous processlowers 3. Susceptibilityof Ice Formation to Mixed the maximumice crystalconcentrations nucleatedby up to an Particle Composition To address the uncertainties orderof magnitude.This occursfor verticalmotionslessthan in the abundance and frac- about 20cms-•andtemperatures higher than-50øC. Sensitivity to thesizesof insoluble aerosolcomponents was tionationof mixedparticlesin the atmosphere, the fractionof CCN containinginsolubleparticlesand the sizesof insoluble examined in two other series of numerical simulations for components were variedin numericalsimulations at several sulfuricacid solutiondroplets.Figure7 showsmodelresults updraftsand severalinitial temperatures. Figure5 showsre- for the assumptionthat eachparticlecontains50% insoluble mass.This assumption impliesmuchlargerinsolubleparticle sizeacrossthe CCN sizespectrum. The fractionof all aerosols 10000 containingan insolublecomponentneededto causethe transitionfrom a homogeneous to a heterogeneously dominated ice formation process was just a few hundredths for thiscase. 1000 In simulationsresultsnot shown here, a soot diameter of 0.08 pm was assumedfor all aerosolsizesfor which sucha size .10cms-1......... 100 10000 G o, o n 10 1000 • I '- -.., 100 0.0001 0.001 0.01 0.1 l'O'cm,s ß -- ..:.:2222::... 1 Fraction with insoluble component Figure5. Maximumicecrystal concentration inparcel modelsimulationsversus thenumberfraction of ammonium sulfate CCN (L = 1.7) containing aninsoluble particle, asa function ofvertical velocity and initial parcel •ture. Vertical velocities of1,10and50cms-]are 0.1 , , ill 0.001 i I I I lllll I 0.01 I I I .I I II i 0.1 I I I ! I Ill 1 dmoted byalternate solidordashed linegroupir• goingfrombottom Fraction with insoluble component totopinthefigure. Linesbetween datapoints arelinearextrapolations. The'initial temtxommres were-38(squares), -46(triangles), and-54øC acidwatersolution droplets (circles). Thesizesof thesoluble andinsoluble components werethe Figure6. As in Figure5, butfor sulfufic •=1). sameasforFigure4. 0.0001 DEMOTT ET AL.' ICE FORMATION IN UPPER TROPOSPHERIC CLOUDS nation for observations 10000 • of increased ice concentrations 19,581 in the volcanicallyaffectedcirrusis a switchto a morepredominant homogeneous freezingprocesscomparedto surroundingUT cirrus.That is, the lack of insolubleparticlesin the exchanged stratospheric air resultedin a hypothetical shiftfrom the right to theleft sideof Figure7. n 1000 •................... .• 50 cm s-1 10 4. Comparison to Atmospheric Observations of Aerosols 1 and Ice Formation Observations of the microphysics and thermodynamics of UT cloudssupportthat bothhomogeneous andheterogeneous nucleationprocessesplay roles in differentsituations.The 0.0001 0.001 0.01 0.1 1 generalimportanceof the homogeneous freezingprocessis Fraction with insoluble component clearlysupported by therareobservation of liquidwatercolder Figure7. As inFigure6, butasstnm•thattheinsoluble coinFonentthan about -38øC. However,it is possiblethat the predomiof mlfi,dcacidCCN represent 50ø/6of aerosol massacross thesize nance of this mechanism for ice formation is overstated due to thenatureof the cloudssampledby variousauthors.Most observationswhich clearlydocumentice initiationat temperarepresented lessthan 10% of the total mass.In this case,a tures below -35øC are from orographicwave clouds. These heterogeneous freezingimpact on cirrus ice formationwas cloudsprovideexcellentoutdoorlaboratories for studyingthe only predictedif the fractionof all aerosolscontainingin- formationof ice due to their relativelylaminar flow fields. solublematterapproached a valueof 1. This is dueto the low- Nevertheless, the typicalupdraftsin thesecloudsrangefrom eredfreezingefficiencyof thesesmallerinsolubleparticulates. about 0.5m s-•to 10m s'•.Asshown inFigure 2, anyheteroThis sizeis characteristic of the largerparticleswithin the ex- geneousfreezingprocessis predictedto be maskedby a haustfromjet aircraft.Therefore,a minimumpotentialimpact dominanthomogeneousprocessfor vertical motionsexceedof widelydispersed aircraftexhaustaerosolson heterogeneous ingabout 0.5m s4. Thisisconsistent withtheobservations of ice formationprocesses in cirrusmightbe inferred.However, HeymsfieMandMilosevich[ 1993, 1995] in orographicwave this resultcoulddependon chemicalprocessing and surface cloudsand $assenand Dodd [1988] in an altocumuluscloud. activationof aerosols at latertimesaftergeneration. Withinjet Vertical cloudmotionsexceeded1 m s-• in both studies.Both contrails,it is quitepossiblethat evensmallersootparticles studiesalsonotedice formationin agreement with the expeccouldact as heterogeneous freezingnucleidue to the higher tationsof a homogeneous freezingnucleationprocess.Howhumiditiesgenerated. Karcher et al. [1996] provideevidence ever,SassenandDodd alsonotedthatheterogeneously formed that sootaerosolsas smallas 0.02 pm are responsible for ice ice crystalscouldhavebeenpresentin concentrations as high 0.1 I , llll , , I I , ,Ill , I , , 11111 , I I I , Ill as50 L-• without beingdetected. Suchconcentrations of het- formation in some contrails. In additional simulations not shown here the numbers of CCN containinginsolublematter (10% by mass) at sizes greaterthan0.5 gm were increased by a factorof 20. This was doneby alteringthe CCN spectrumto the form [CCN] = 80 erogeneous ice nucleicouldplay an importantrole in clouds withupdrafts below0.5 m s'i, a category thatprobably in- cludesmanywidespread cirrusclouds. The mechanisms of ice formationin cirrusformedby a 5ø'6.Thisbhange waspreviously shown toincrease icecrystal gentlelifting processshouldin principlebe distinguishable. concentrations nucleatedby homogenous freezingby a maxi- Ice crystalconcentrations predictedfor a homogeneous ice mum of 30% [DeMott et al,. 1994]. The effecton the aerosol formationprocessare in the rangefrom a few hundredto a fractioncontainingan insolublecomponent neededto causea fewthousand perliterforanupdraft rangeof 10to 20 cms'• transitionto a heterogeneous freezingnucleationprocesswas (Figures5 to 7). In contrast,the ice crystalconcentrations nearlyequivalentto the effectof raisingthe insolublemass predictedwhenincludinga heterogeneous ice formationprocpercentto 50% (i.e., Figure7). The presenceof an insoluble essare only 1 to a few hundredper liter for theseconditions. particulatewithin just a few percentof an enhancedpopula- Dowling and Radke [1990] haveestimateda typicalrangeof tion of larger CCN lowered ice crystal concentrations by ice crystal concentrations for cirrus cloudsto be between nearlyan orderof magnitude.Aerosolsfrom volcanicerup- about 1 and100L'•. These concentrations arequitelowcomtionsare known to enhancethe largerpart of the UT aerosol paredto the theoreticalpredictions for homogeneous freezing, size distribution. The influence of volcanic aerosols on cirrus but agreewell with thoseexpectedwith the inclusionof a hetcloudscouldthereforestronglydependon the freezingnuclea- erogeneousfreezing process.However, we must note that tion activityof containedinsolublecomponents and changes measurements using improvedin situ methodsfor detecting in composition of the volcaniccloudovertime. Volcanicaero- concentrations of small ice crystalsin cold UT clouds[e.g., sols in the stratosphere are known to transformover time, He?msfieldand Milosevich, 1995; Strom and Heintzenberg, suchthat the larger insolublematteris depositedto the tro- 1994] suggestthat averageice crystal concentrations have posphereand the remainingaerosolsare pure sulfuricacid probablybeen underestimated in the past. Nevertheless,the [Zhaoet al,. 1995]. Thereforeit is likely that a pure sulfuric estimatedrange of concentrations presentin cirrus (1 to acid aerosolcloudexistedin the stratosphere at the time that 10,000 L4) hasnotchanged. Observations of crystal concenSassenet al. [1995] inferredexchangeof Pinatuboaerosols trationsbelowa few hundredper liter in recentlyformedcirinto the troposphereand subsequent modificationof cirrus rus canbe only be accounted for by a singularhomogeneous cloudpropertiesby increasedCCN. Thus an alternateexpla- freezinghypothesis if updraftsaretypicallyjust a few centime- 19,582 DEMOTTET AL.' ICE FORMATIONIN UPPERTROPOSPHE•CCLOUDS terspersecond or if wideregionsof depleted CCN concentra- positionof CCN which couldfreezeat 70% RH at -50øC.A tionsare presentin the uppertroposphere [Heymsfield and possibleexplanationis nitric aciduptakeby solutiondroplets Milosevich, 1995]. Unfortunately,the measurement of small vertical motionsremainshighly uncertainand coincident measurements of CCN with in situ microphysicalobservationsin theuppertroposphere arefew in numberat thetimeof thiswriting. The relativehumidityregimeof cirruscloudsshouldalso [Lamb et al., 1996]. However,the saturationratios of nitric acidin the uppertroposphere in midlatitudesappearto be too low for such an effect to occur. The potentialimportanceof heterogeneous freezingnuclei mightalsobe inferredfrommeasurements of the development of ice crystalconcentrations in cloudswarmerthantheh6moindicate their formative mechanisms. The definition of this geneousfreezingtemperature regime;for example,cloudsat regimeis somewhat complicated by theverticalmicrophysical around-30øC.HeymsfieldandMilosevich[ 1993, 1995]noted structure of cirrusthatplacesnucleation at thecloudtopwith a generalabsenceof ice particleswarmerthan -34øC in the wavecloudstheystudied.Thismightindicatethe regions of ice crystal growth and evaporationbelow supercooled in [HeymsfieM andMilosevich, 1995].Nevertheless, Heymsfield absenceof freezingnuclei. However,other observations and Milosevichwere able to definea minimumRH required wave clouds(W.A. Cooper,Ice formationin wave clouds: duringevaporation, preprint,Ameriforcirrusformation by usingthemaximumrelativehumidities Observedenhancement on CloudPhysics,15observedin clearair conditionsduringNASA's FIRE II pro- canMeteorologicalSocietyConference gram[of.Heymsfield andMilosevich,1995,Eq. (1)]. Their 20 January,Dallas,Texas,pp. 147-152,1995)andorographic RH limit is displayed as a functionof temperature in Figure8. capclouds[Cooperand Vali, 1981]haveshownthe presence attemperatures downto-34øC. Thethreshold values (arbitrarily > 0.001L'l) required forho- ofupto 10L'1icecrystals Direct informationon the abundance of heterogeneous IN mogeneous and heterogeneous freezingof differentsolution couldresolvetheissueof theirimpordropletsare also shownin Figure8. For the assumptions in theuppertroposphere made,it is seenthat the calculations includingheterogeneous tancethere,but few measurements of IN existanywhereother freezingnucleation for sulfuricacidsolution droplets and3,= thanat the Earth'ssurface.Bigg [1967] collectedaerosolson 1 are mostcloselyin accordwith observations. Heymsfield filtersduringlong latitudinalaircrafttransectsat 4 to 12 km and Milosevichstatedthat their observations might be ex- andprocessed thesein a diffusionchamber.He foundIN con- centrations ranging from0.05to > 1 L4 whenprocessed at plainedif a heterogeneous ice formationprocess occurred at an RH of 10% lowerthanhomogeneous freezing.This is approximately thevaluewe predictfor thetemperature rangeof their measurements. The lowestcirrustemperaturesampled by Heymsfieldand Milosevichwas about-47øC,and the minimumRH requiredfor cloud formationwas near 80%. Thereforeit is possiblethat the slopeof their extrapolated curvemaybe toosteepandmayunderestimate required RH at coldertemperatures. Thisis supported by ourmodelingcalcu- havebeenprocessed for the lower cloudtemperatures that existedat the samplingaltitudes.D.C. Rogersand P.J. DeMott (Measurementsof natural ice nuclei, CCN, and CN in winter clouds,preprint,AmericanMeteorological SocietyConference on CloudPhysics,15-20 January,Dallas,Texas,pp. 139-144, 1995) collectedbulk bag samplesof aerosolsnear wave lationsand the fact that it is difficult to hypothesize a tom- clouds(4 to 9 km altitudes)for processing to detectIN in the water saturationat -15øC. It seemspossiblethat higher IN concentrations would have been detected if the filters could lOO homogeneouslimit • 3: 13::90 I:: E• i• 85 (NH4)2SO4 ............... '-• • .................... .................. / -' homogeneous limit .e 80 75 -65 Observed Min. RH vs.Ttoform cirrus H2SO4/H20 • _ / heterogeneouslimit ./ ' ß• I:Z:: • / heterogeneous limit H2SO4/H20 / -60 -55 ,, -50 . -45 -40 -35 Temperature of Min. RH (øC) Figure8. Threshold humidity versus temperature fortheformation oficeinnumerical s'andations including onlyhomogeneous freehng (homogeneous limit)andhomogeneous plusheterogeneous fieezhzg nucleation (heterogeneous limit)of (NH4)2SO4 (• = 1.7)and H2SO4/•-I20 ½= 1.0)solution droplets. Cloud model iceconcentrations > 0.001 L4arepredicted above eachthreshold curve. Theheterogeneous nucleation thresholds areforthecase represented inFigure 7 (soluble fraction = 0.5,with10% ofaerosols conta'lni• aninsoluble particle). Thehomogeneous limiting curves areforanutxtmff of1cms4 andwould beoffset toward higher humidifies forhigher updrafts. Theheterogeneous limiting curieisonlyslightly sensitive to uv•ft since heterogeneous fieehng rates werepmmnetmz• asa fimction oftem•uae alone. Alsoplotted areobserved threshold humidities required forcinuscloud formation estimated byHeymsfield andMilosevich [1995]. DEMOTT ET AL.' ICE FORMATION [N UPPER TROPOSPHERIC CLOUDS 19,583 laboratory.Aerosolswere processed over a rangeof satura- sizes of insoluble matter. It also dependson the relative freezingefficiencyof differentinsolubleaerosolsin comparisimulations ofwavecloudascent (1 to8 m s-1)in a controlledsonto the sootspeciesassumedin this paper.Other,moreefexpansion cloudchamber.IN concentrations measuredat or ficient, freezing nuclei may be presentin the upper troabove watersaturation at-30øCranged from1 to 100L'•. posphere.Resultswill remain speculativeuntil appropriate tion conditions in a continuous flow diffusion chamber and for These concentrations represented fromabout1 partin 105to greater than1 partin 103ofCCNconcentrations measured in new measurements are obtained. 3. The rangeof concentrations of ice crystalsnucleatedand the samebag samples.Since,basedon DeMott [ 1990], only the humidityregimefor the onsetof ice formationin cirrus about 1% of available freezingnuclei might be active at a predictedwhen includinga heterogeneous nucleationprocess temperatureof-30øC, the observedIN concentrations could agree with existing atmosphericobservations.In contrast, indicatefrom 0.01 to 0.1 fractionsof CCN containinginsolu- predictionsthat consideronly a homogeneous freezingprocess ble freezingnuclei.This is within the rangeof predictedhet- actingin cirrusdo not agreewith most existingobservations erogeneous effectsbasedonFigures5 to 7. of ice crystalconcentrations and the humidityrequiredfor the Existingmeasurements of UT aerosolproperties lend some onset of ice formation. supportto the potentialfor a heterogeneous ice formation 4. Existingmeasurements of aerosolsand ice nucleiin the processin UT clouds.Hagen et al. [1994] foundthat UT par- upper tropospheresuggestthat the numbers and sizes of ticlessmallerthan 0.1 gm were usuallymixed particleswith mixed aerosolscapableof acting as heterogeneous freezing solublefractionsof from 0.3 to 0.5. Particleslarger than 0.1 nucleiare sufficientto reducecirruscrystalconcentrations and grnwerealsopredominately mixedparticleswith 5 to 20% of lowerthe RH requiredfor cloudformation. the particlescontaininginsolublespeciessuch as carbon, The resultspresentedin this paperunderscore the needfor aluminum,and titanium. ?ueschelet al. [1994] found Si and in situ measurementsof ice nucleatingaerosolsin the upper includingidentifyingnaturaland anthropogenic Mg in 10 to 20% of "liquid" particlesin backgroundUT air. troposphere, Consideringour numericalcalculations,suchhigh percent- sourcesof theseaerosols.Suchan effortwas initiatedduring ages and large sizes of insolublespeciesassociatedwith the springof 1996 [Thompsonet al., 1996, (pp.175-199)]. In solubleaerosolssuggesta strongpotentialfor the actionof addition,laboratorystudiesare neededto deœme solutionefheterogeneous ice nucleationprocesses.However, we must fectson the freezingratesof solubleand insolublespeciesat alsonotetwo casespresentedby Sheridanet al. [ 1994] indi- or neartheir actualdeliquesced sizesat coldtemperatures. Theresultspresented alsohavesignificance to theradiative catingthatUT aerosols werepredominantly composed of sulfates and did not contain detectable amounts of insoluble propertiesof cirrus.Sincelargenumbersof smallice crystals matter. Sheridanet al. also notedthat the severalpercentof absorband scatterradiationmore effectivelythan a smaller aerosolsthey sampledcontaininginsolublecrustalmaterials numberof largeice crystals[e.g.,Kinne and Liou, 1989], the and metals were not coated with sulfates at a detectable level. sensitivitiesof ice formationto temperatureand verticalvelocityareveryimportant.It was shownthat the functionalde5. Summary pendenciesof ice crystal concentrations versustemperature and verticalvelocityare quite differentdependingon the naA numericalanalysisof the impactof homogeneous and ture of the ice fonuationprocess(homogeneous versusheteroheterogeneous ice formationprocesses on initial cirruscloud geneous)in the uppertroposphere. A variablepopulationof composition has beenpresented. The salientconclusions are freezingnucleusaerosolscouldlead to a switchin the prethe following. dominanceof heterogeneousversus homogeneousfreezing 1. A numberof uncertainties existwhichimpactthe pre- and a large difference(orderof magnitude)in maximumice dictedhomogeneous freezingof unactivated solutiondroplets crystal concentrationsnucleated.This sensitivityto aerosol in theuppertroposphere. Solutecomposition doesnot appear number concentrations doesnot theoreticallyexist, exceptin to be a criticalparameter,with two notableexceptions: the veryspecialcircumstances, for cirrusformedonlyby homogehigh deliquescence humidityof ammoniumsulfateand the neous freezing of CCN. Ice crystal concentrations are ineffectof nitricaciduptakeby solution droplets. Uncertainty in verselyrelatedto effectiveice crystalradius[e.g.,Jensenet the freezingconditions(nonequilibrium freezingpoint deal., 1994b].Thereforelower maximumice crystalconcentrapression)of relevantsolutesis a moreimportantfactor.The tion for any set of conditionsimpliesa smalleropticaldepth impactof this uncertainty is primarilyan indirectoneon the andlowernet radiativeforcingin cloudsinfluencedby heteroapparentrelativehumidityneededfor ice formationto occur. geneousnucleation.At the sametime a heterogeneous freezConsidering the expected rangeof atmospheric solutionefing processfavorsthe morefrequent(at lowerRH values)and fectsonfreezingpointdepression, it was shownthatan uncerwidespreadformationof ice clouds. taintyof severalpercentRH existsfor predicting cirruscloud In conclusion,heterogeneous ice nucleicouldexerta much formation by homogeneous freezing.Variations in icecrystal strongerandmorevariableclimate-forcing influencein cirrus concentrations are only about30% due to this factor.Varicloudsthanthat which is exertedby solubleaerosolsacting abilitiesin solubleaerosolsizedistribution mayleadto simisolelyby a homogeneous freezingprocess.This studyprolar uncertainties in icecrystalconcentrations formed. videsa quantitative exampleof theimportance of considering 2. Insolublematerialactingas heterogeneous freezingnu- ice nuclei in the overall aerosol-cloud-climatescenario,as cleiwithinsolution droplets in UT air maydecrease icecrystal statedby Rogers[ 1994]. concentrations in someckruscloudsby up to a factorof 10. Theapparent humidityrequiredfor ice formation mayalsobe decreased by upto 10%.Theimportance of thisheterogeneous Acknowledgments.This researchwas primarilysupportedby the freezingprocesscompared to the competing homogeneousNationalAeronauticsand SpaceAdministrationundergrantNAG-2freezingnucleation process depends ontheconcentrations and 924. Partialfundingwas providedby the U.S. Departmentof Energy 19,584 DEMOTTETAL.' ICEFORMATION 1NUPPERTROPOSPHERIC CLOUDS undergrantDE-FG03-95ER61958andby theNationalScience Foun- Lamb, D., A.M. Moyle, and W.H. Brune,The environmentalcontrol of individualaqueousparticlesin a cubicelectrodynamic levitation dationundergrantATM-9311606.The contents of thispaperevolved system,AerosolSci. Technol.,24, 263-278, 1996. from materialpresentedat the 1995 NASA Atmospheric Effectsof and homogeneous AviationProgramAnnualMeeting,April 23-28, VirginiaBeach,Vir- Langham,E.J.,and B.J.Mason,The heterogeneous nucleationof supercooled water,Proc. Phys.Soc.A, 247, 493-505, ginia.We thankthereviewersfor theirusefulcomments. 1958. Ohtake,T., Freezingpointsof H2SO4aqueoussolutionsand formation References ofstratospheric iceclouds,Tellus,45, 138-144, 1993. Pruppacher,H.R., A new look at homogeneous ice nucleationin suBertram,A.K., D.D. Patterson,and J.J.Sloan.,Mechanismsand tempercooledwaterdrops,or.Aimos.Sci., 52, 1924-1933.,1995. peraturesfor the freezingof sulfuricacid aerosols measured by Pruppacher,H.R., and J.D. Klett, Microphysicsof Cloudsand PreFTIR spectroscopy, J. Phys.Chem.,100, 2376-2383, 1996. cipitation,714 pp., D. Reidel,Norwell, Mass., 1978. Beyer,K.D., S.W. Seago,H.Y Chang,andM.J. Molina,Composition Pruppacher,H.R., and M. Neiberger,The effectof water solublesubandfreezingof aqueous H2SO4/HNO3 solutions underpolarstratostanceson the supercooling of waterdrops,or.Atmos.Sci., 20, 376sphericconditions,Geophys.Res.Lett., 21, 871-874, 1994. 385, 1963. Bigg, E.K., Crosssectionsof ice nucleusconcentrations at altitude Pueschel,R.F., J.M. Livingston,G.V. Ferry, and T.E. DeFelice,Aerooverlongpaths,J. Aimos.Sci., 24, 226-229., 1967. solabundances and opticalcharacteristics in the PacificBasinfree Chen, J.P., Theory of deliquescence and modifiedKohler curves,or. troposphere, Atmos.Environ., 28, 951-960, 1994. Aimos. Sci., 51, 3505-3516, 1994. Cooper,W.A., and Gabor Vali, The origin of ice in mountaincap clouds,or.Aimos. Sci., 38, 1244-1259., 1981. Rard, J.A., A. Habenschuss,and F.H. Spedding,A review of the osmotic coefficientsof aqueousH2SO4at 25øC, or.ChemEng. Data, 21,374-379, 1976. DeMott, P.J.,An exploratorystudyof ice nucleationby sootaerosols, Rasmussen,D.H., Thermodynamicand nucleationphenomena:A set J.Appl. Meteorol., 29, 1072-1079, 1990. of experimental observations, or.Cryst.Growth,56, 56-66, 1982. DeMott, P.J., M.P. Meyers, and W.R. Cotton, Numerical model Reischel,M.T., and G. Vali, Freezingnucleationin aqueouselectrosimulationsof cirruscloudsincludinghomogeneous and heterogelytes.Tellus, 27, 414-427, 1975. neousice nucleation,or.Aimos. Sci., 51, 77-90, 1994. Robinson,R.A., and R.H. Stokes,ElectrolyteSolutions,571 pp., Detwiler, A., Commentson "Homogeneous nucleationrate of highly ButterworthsLondon, 1965. supercooled cirrusclouddroplets"by K. SassenandG.C. Dodd,or. Rogers,D.C., Commentson "Quantifyingand minimizinguncertainty Aimos. Sci., 46, 2344-2345, 1989 Dowling,D.R., andL.F. Radke,A summaryof thephysicalproperties of cirrusclouds,or.Appl. Meteor., 29, 970-978, 1990. Gulteppe,I., D.O'C. Starr,A.J. Heymsfield,T. Uttal, T.P. Ackerman, and D.L. Westphal, Dynamicalcharacteristics of cirrus clouds from aircraftand radar observations in micro and meso-7scales,or. Aimos. Sci., 52, 4060-4078, 1995. Hagen, D.E., J. Podzimek,A.J. Heymsfield,M.B. Trueblood,and C.K. Lutrus, Potentialrole of nuclei in cloud elementformationat high altitudes,Aimos.Res., 31, 123-135, 1994. Heymsfield,A.J., andR.M. Sabin,Cirruscrystalnucleation by homogeneousfreezingof solutiondroplets,or.Aimos.Sci., 46, 22522264, 1989. of climateforcingby anthropogenic aerosols"by J.E. Penneret al., Bull. Am. Meteorol. Soc., 75, 2313-2314, 1994. Sassen,K., and G.C. Dodd, Homogeneous nucleationrate for highly supercooled cirrusclouddroplets,or.Aimos.Sci., 45, 1357-1369, 1988. Sassen,K., and G.C. Dodd, Haze particle nucleationsimulationsin cirrus clouds,and applicationsor numericaland lidar studies,or. Aimos. Sci., 46, 3005-3014, 1989. Sassen,K., D.O'C. Starr, G.G. Mace, M.R. Poellot, S.H. Melfi, W.L. Eberhard,J.D. Spinhirne,E.W. Eloranta,D.E. Hagen, and J. Hallett, The 5-6 December1991 FIRE IFO II jet streamcirrus case study:Possibleinfluencesof volcanicaerosols, or.Aimos.Sci., 52, 97-123, 1995. Heymsfield,A.J., and L.M. Milosevich,Homogeneous ice nucleation Sheridan,P.J., C.A. Brock, and J.C. Wilson, Aerosolparticlesin the and supercooled liquidwaterin orographicwaveclouds,or.Aimos. uppertroposphere and lower stratosphere: Elementalcomposition Sci., 50, 2335-2353, 1993. and morphologyof individualparticlesin northernmidlatitudes, Heymsfield,A.J., and L.M. Milosevich,Relativehumidityand temGeophys.Res.Lett., 21, 2587-2590, 1994. peratureinfluenceson cirrus formationand evolution:Observa- Str6m, J., and J. Heintzenberg,Water vapor, condensedwater and tions from wave clouds and FIRE II, or.Aimos. Sci., 52, 4302crystalconcentration in orographically influencedcirrusclouds,or. 4326, 1995. Hoffer, T.E., A laboratoryinvestigationof dropletfreezing,or.Meteorol., 18, 766-778, 1961. Jensen,E.J., O.B. Toon, and P. Hamill, Homogeneous freezingnucleationof stratospheric solutiondroplets,Geophys.Res.Lett., 18, 1857-1860, 1991. Aimos. Sci., 51, 2368-2383, 1994. Tang, I.N., and H.R. Munkelwitz,Water activities,densities,and refractiveindicesof aqueoussulfatesand sodiumnitratedropletsof atmospherici•nportance,or. Geophys.Res., 99, 18801-18808, 1994. Thompson,A.M., R.R. Friedl,and H. Wesoky,Atmosphericeffectsof Jensen,E.J., and O.B. Toon,The potentialeffectsof volcanicaerosols aviation:First report of the SubsonicAssessment Project,NASA on cirrus,Geophys.Res.Lett., 19, 1759-1762, 1992. Ref. Publ., 1385, 1996. Jensen,E.J., and O.B. Toon, Ice nucleationin the uppertroposphere: Vali, G., Atmospheric ice nucleation: A review,or.Rech.Aimos.,19, Sensitivityto aerosolcomposition and size distribution,tempera105-115., 1985 ture, and coolingrate, Geophys.Res.Lett., 21, 2019-2022, 1994. Weast,R.C. (Ed..), CRC Handbookof Chemistryand Physics,61st Jensen,E.J., O.B. Toon, D.L. Westphal,S. Kinne, and A.J. Heymsed.,2463 pp., CRC Press,BocaRaton,Fla., 1981. field., Microphysical modelingof cirrus1. Comparisonwith 1986 Zhao,J., R.P. Turco,andO.B. Toon,A modelsimulation of Pinatubo FIRE IFO measurements,or. Geophys.Res., 99, 10421-10442, volcanic aerosols in the stratosphere, or. Geophys. Res., 100, 73151994a. Jensen,E.J., O.B. Toon, D.L. Westphal,S. Kinne, and A.J. Heymsfield., Microphysicalmodelingof cirrus2. Sensitivitystudies,J. Geophys.Res., 99, 10443-10454, 1994b. Karcher, B., T. Peter, U.M. Biermann, and U. Schumann,The initial compositionof jet condensation trails, or.Aimos.Sci., 53, 30663083, 1996. Kinne, S., and K.-N. Liou, The effectsof the nonsphericityand size distributionof ice crystalson the radiativepropertiesof cirrus clouds,Atmos.Res., 24, 273-284, 1989. 7328, 1995. P.J. DeMott. S.M. Kreidenweis,and D.C. Rogers,Departmentof At- mosphericScience,ColoradoStateUniversity,Fort Collins,CO 80523.(email:[email protected]) (ReceivedDecember18, 1995;revisedJanuary17, 1997; acceptedApril 10, 1997.)
© Copyright 2026 Paperzz