GEOPHYSICALRESEARCHLETTERS,VOL. 25, NO. 9, PAGES1363-1366,MAY 1, 1998 Ice nucleationprocesses in uppertropospheric wave-cloudsobservedduring SUCCESS E. J. Jensen •, O. B. TooI12, A. Tabazadeh •, G. W. Sachse s, B. E. Anderson • K. R. Chan•, C. W. Twohy 4, B. Gandrud 4, S. M. Aulenbach 4, A. Heymsfield 4, J. Hallett5, andB. Gary6 Abstract. We have compared in situ measurements near the leading-edgesof wave-cloudsobservedduring the SUCCESS experimentwith numericalsimulations. Observationsof high supersaturationswith re- with sporadicand localizednucleation[e.g.,Sassenet al., 1989]. It is often impossible to tell what stageof evolutiona sampledcloudelementwasin. Wave-clouds that form at the cold crests of mountain waves are ideal systems to studycloudphysical processes because the spect toice/•" 50%) near the leading edge ofavery cold wavecloud < -60øC) are approximatelyconsistent wind field is laminar and well defined, and the cloud with recenttheoreticaland laboratory studiessuggest- is stationary.Ice crystalsnucleateat the leading(uping that large supersaturations are requiredto homoge- wind) edgeof the cloud,and growasthey are advected neouslyfreezesulfateaerosols.Also,the peakicecrystal downwind.Heymsfield andMiloshevich [1993,1995]exnumberdensities observed in thiscloud(about4 cm-s) ploitedobservations in wave-clouds to exploreice nucleaxe consistent with the number densities calculated in ation mechanisms at temperaturesrangingfrom -31 øC our model.In the warmerwave-cloud (T _ -37øC) rel- to-56øC. Although updraft velocitiesin wave-clouds ativelylargeice numberdensitieswereobserved(20-40 may be larger than in typical cirrus, the wave-clouds cm-S). Ourmodelcalculations suggest that theselarge are still usefulto study cloudprocesses. In this study we have combinedobservationsfrom the number densitiesare probably causedby activation of sulfate aerosolsinto liquid dropletsfollowedby subse- SUCCESSmissionwith numericalmodelingto evaluate quent homogeneousfreezing. If moderate numbersof ice nucleationprocesses in two wavecloudsat very dif- effective heterogeneous freezing nuclei(0.5-1cm-s) had ferenttemperatures(-64øC and-37øC). We will show been presentin either of theseclouds,then the number that the observed critical humidities for ice nucleation densitiesof ice crystalsand the peak relative humidities and ice crystal number densitiesare consistentwith calshould have been lower than the observed values. culationsassuming that homogeneous freezingof sulfate aerosols is the dominant nucleation mechanism. The first sectionof the paper briefly summarizesthe SUCCESSwave-cloud observations usedin thisstudy.Next, Introduction Over the past few decades,severalstudieshaveshown we describes the wave-cloud formation simulations. Fithat cirrus cloudsplay an important role in the earth's nally, we comparethe model resultswith observations radiationbudgetand climate[e.g.,Liou, 1986].Hence, and evaluatethe implicationsfor ice nucleationpro- it is important to understandthe processesresponsi- cesses. ble for their formation. In particular, the temperatures and humidities required for ice nucleationmay control SUCCESS Wave-cloud Measurements cirrus frequency. Theoretical analyseshave been used For comparisonwith nucleation simulations,we reto predict the ice supersaturationrequired for homoquire measurementsof peak relative humidities and geneous freezingof sulfateaerosols[Sassen and Dodd, ice crystal number densities at the wave-cloudlead1989;Heymsfieldand Sabin,1989;Jensenet al., 1994], ing edges. Relative humiditiesare calculatedfrom the and it has been suggestedthat this is probablythe dom- laserhygrometerwatervapormeasurements [Collinset al., 1995], and the meteorological measurement system ever, these predictions have not been extensivelyconon boardthe NASA firmed with atmosphericobservations. Nor is it clear (MMS) temperaturemeasurements inant ice nucleation mechanism in cirrus clouds. How- DC-8. Given the uncertaintiesin these measurements, what role heterogeneousice nucleationplays. The processesresponsiblefor cirrus cloud formation the uncertainty in the derived relative humidity with (i.e., ice nucleationand growth)are typicallyvery dif- respectto ice (RHI) is about 10%. During the flight ficult to study with in situ observations. Midlatitude on May 2, the leadingedgeof a wave-cloudsystemover continental cirrus cloudsare usually highly structured Colorado was crossedseveraltimes. In this study, we will focuson the leading-edgecrossings at 20:18:28UT and 20:29 UT, when the dynamicsseemedto approxiNASA AmesResearch Center,MoffettField, CA mately match the idealizedwave-cloudmodel. At these times, the DC-8 flew approximatelydownwindinto the 2Universityof Colorado,Boulder,CO SNASALangleyResearch Center,Hampton,VA leading-edge ofthewave-cloud (20:18:28 UT cross. in•) (20:29UT crossing). 4NationalCenterforAtmospheric Research, Boulder,CO andupwindout of the leading-edge 5DesertResearchInstitute,Reno,NV 6JetPropulsionLaboratory,Pasadena,CA Copyright 1998bytheAmerican Geophysical Union. Papernumber98GL00299. 0094-8534/98/98GL-00299505.00 The measuredtemperature, relative humidity, and ice crystal number density are plotted versustime for the first crossingin Figure 1. The peak number densities of ice crystalslarger than about 3/•m radius measured by both the counterflowvirtual impactor (CVI) and the video ice particle sampler(VIPS) were about 27 cm-s, and the peak RHI wasabout 150-160%.Ice 1363 1364 JENSEN ET AL.' ICE NUCLEATION IN WAVE CLOUDS Wove-cloud simulation: results at ! 1.9 km DC-8 20.48 0 data: 2 4 - 20.491 [•ce •n) I0 12 14 Toro!N• !o.ooo .... N• (r>3 •) :.•0 140 .r.., 1,o00 100 ~ .., .., 4, i 140 O.lOO 12o Figure 1. DC-8 measurements for the leading-edge crossing at 20:18:28UT, May 2 are plottedversustime. At this time, the DC-8 flew downwindinto the leadingedgeof the wave-cloud at 11.9km. The measuredtem- o.o,o I 0,001 L, ........ o • ......... t., .•, ..... • ....... 1 2 3 ß • ........ 4 •, -. ...... n 100 5 6 -7O Ti.• (.•.) ½ ..-c-oC• N•, (era•) : : : : SlmuloUo, perature(MMS), relativehumidity(laserhygrometer), and ice crystalnumberdensity(CVI and VIPS) are Figure 2. The model RHI, ice crystalnumberdensity, plotted versustime. The magentaasterisksand trian- and temperature are plotted versus time at 11.9 km glesarethe VIPS datafromthe lowresolution andhigh for the May 2 case. The total number density of ice resolution cameras, respectively. crystals(solidblackcurve)and the numberdensityof ice crystalswith radii largerthan about 3/•m (dashed curve) that would be detectedby the CVI are shown. The DC-8 RHI and CVI data from the passat 20:29 UT number densities largerthanabout10cm-3 werenever havealsobeenplotted versustime (convertedto parcel observed in this wave-cloud. time advectedby the wind) for comparison. Duringthe wave-cloud sampling flighton April 30, we werenot ableto fly legsdirectlyupwindof the cloud due to restrictedair-space.However,severallegswere flownalongthe leadingedgedippingin andout of the sizedistributionwith N - 100 cm-•, rocloud. The peak RHIs measured just upwindof this and er = 2.3. 0.02/•m, Tabazadehet al. [1997]usedrecentlaboratorydata cloud were on the order of 120-130%, and the peak ice for the physicalquantities numberdensitiesindicatedby the CVI rangedfrom 20 to developnew expressions fleezingof sulfateaerosols. to 40 cm-3. The actual maximum RHI may not have that control homogeneous We have used the expressions recommendedby Tababeen detecteddue to the cross-windflight path. The temperatureat the cloudleadingedgefor the zadeh et al. to calculate aerosolfleezing rates in this May2 casewasabout-64øC.FortheApril30 casethe work. In someof the simulations,we alsoincludedhet- fleezingnuclei(seeJensenandToon[1997] temperature wasabout-37øC.The slopeof the poten- erogeneous fleezingcalculations). tial temperaturesurfaces (measured by the microwavefor detailsof the heterogeneous We initialized the model atmospheric temperature temperature profiler(MTP)) onthe upwindsideof the (MTP and wave crestindicated a vertical wind speedof about 1.2 profilesbasedon SUCCESSmeasurements MMS). We used an initial RHI vertical profile with a m-s-1 for the May 2 case.This estimateapproximately agrees withthe verticalwindspeedindicated by MMS. Gaussianshapepeakedwherethe wave-cloudwas obThe verticalwind speedsindicatedby MMS in the April served(11.9 kin). The peak RHI was initially set at 30 wave-cloudweremuchstronger(updraftsup to 5 m- just below100%.The updraftvelocityin the modeloscillatedsinusoidallywith time; however,overthe short s-l). Theseverticalwind speedmeasurements have largeuncertainties; the sensitivity of the modelresults to verticalwind speedis discussed below.The temperatures,humidities,ice numberdensities, and plausible Wove-cloud simulation: results at 8.75 km 1•0 -3 15,o - rangesfor verticalwindspeeds for the twowave-cloud casesare given in Table 1. Wave-Cloud Ice Nucleation lO.OOo Simulations 1.000 To simulate ice nucleation at the wave-cloudleading edges,weuseda one-dimensional versionof the detailed 0.100 ice cloudmodeldescribedby Jensenet al. [1994].Previously,we haveusedthis modelto evaluatethe sensitivity of cirruscloudsto sulfateaerosolnumberdensity, temperature, andcoolingrate [Jensen andToon,1994]; and heterogeneous ice nuclei[Jensenand Toon, 1997]. 0.001 -1100 Threeparticletypesare included:puresulfatesolution 1.50 1.60 1.70 1.80 1.90 2.0,0 2. I0 2.20 aerosols(H2SO4/H20), activatedwater droplets,and Time(min) ice crystals.For eachparticletype, 60 radiusbinsare includedspanningthe ranges0.005- 100/•m for sulfate Figure 3. SameasFigure2, but forthe April 30 wave- 1 aerosols anddroplets, and0.01- 232/•mfor icecrys- cloudcase.We alsoshowthe liquiddropletnumber tals. For the sulfateaerosols,we assumea log-normal densityversustime in this case. JENSEN ET AL.' ICE NUCLEATION Table 1. Summary of SUCCESS Wave-CloudMeasurements Date May 2, 1996 April 30 Time (UT) Temp 20:29:12 -64øC 21:12:00 -36.7øC Updraft 0.8-1.6m-s-• 3-5 m-s- Peak RI-II Peak Ice Conc. 150-160% 2-7 cm-3 _•130% 20-40 cm IN WAVE CLOUDS 1365 Table 1, if we use our old aerosolfreezingparameterization, then not only is the peak RHI decreasedto about 126%,the peakice crystalnumberdensityis 17 cm-• (muchhigherthan the observedvalues). For the rapid cooling rate and low temperature of this wave-cloud, the peak crystal number density is probably controlled exclusivelyby homogeneousfreezing. However, heterogeneousnuclei could modify the crystal number density in wave-cloudsif a substantial fraction of the sulfate aerosols contained effective het- erogeneousfreezingnuclei. Simulationsincludingheterogeneous ice nucleiindicate that the number density of ice crystalsgeneratedcould be reducedcomparedto freezingof pure time-period duringwhichnucleation occurred, thever- the numbergeneratedby homogeneous sulfateaerosols(seeTable 2). Also,if substantialnumtical windspeeddid not varysubstantially. bers of effectiveheterogeneousice nuclei are included, then the peak RHI at the cloud leading edge is much Results lower than the observedvalues. The consistencybeMay 2, 1996 Case tw.een the observedpeak ice number densitiesand RHIs We are assuming herethat our 1-D modelrepresents and the results from modeling with only homogeneous a columnof air passingthroughthe wave-cloud.For freezingsuggeststhat relatively large numbersof effec- icenuclei(•_ 0.5-1 cm-s) werenot comparison with aircraftdata,we extractethe model tive heterogeneous resultsversustime at the DC-8 flight level. Figure 2 present. However,this doesnot rule out the possibility ice nucleiwere present. showsthe modelRHI, ice crystalnumberdensity,and that fewer heterogeneous effective radius versus time at 11.9 km for the May 2 case.Icecrystalsdonotbeginnucleating in thissimula- April 30, 1996 Case total numberof ice crystals,we alsoshowthe number The results of the simulation for the April 30 wavedensityof crystalswith radii largerthan about3 pm cloud case are shownin Figure 3. In this simulation, that wouldbe detectedby the CVI. The crystalstake the temperatureis just high enoughsuchthat sulfate about28 seconds to growto 3 pm, sothe numberof aerosolsare activated into water dropletsbeforethe suldetectablecrystalsdoesnot rapidlyincreaseuntil just fate aerosolscan freeze directly. Large numbersof liqtion until the RHI reaches145-150%. In additionto the after the RHI reachesits maximum. Usingthe aircraft uid droplets are activated sincethere is no nucleation speedandambientwindspeed,wehaveconverted the barrier to overcome. The liquid droplets grow, dilute, time-seriesdata from the passat 20:29 UT to a data and freeze within 5-10 seconds.The droplets never get seriesversustime as an air parcelwouldhavebeenen- larger than about 1-2 •umradius. The peak ice crypteringtheleadingedgeandoverlaid thedataonFigure tal numberdensityin the simulation(about 30 cm-•) for this case. The 2. ThepeakRHI andicenumberdensityfromthesim- agreeswell with the measurements ulationagreereasonably wellwiththe observed values. Theobserved peakRHI wasabout10%largerthanpredictedby ourmodel,andicecrystals weredetected by Table 2. Wave-CloudSensitivityTests(May 2 case) the CVI furtherupwindthat predictedby the model. The sensitivityof the modelresultsto key environ- Test Vz Nc, Nh•t • Ni• RHI. mental conditions assumedin the model is summarized in Table2. As discussed previouslyby JensenandToon I baseline 1.2 100 0 2.3 6.9 150.4 [1994], changes in theupdraftvelocity cansubstantially2 low vz 0.8 100 0 2.3 4.0 149.0 distributionhaveonlymoderateimpacton the number 5 highvz highN•, low No, 1.6 100 0 1.2 300 0 2.3 2.3 9.7 8.2 151.3 148.9 1.2 100 0 2.8 6.1 149.1 highScrit 1.2 100 0 2.3 2.3 5.7 9.8 160.0 151.1 changethe peakice crystalnumberdensity;whereas 3 changes in the sulfateaerosol numberdensityandsize 4 of icecrystalsnucleated.A reduction of the updraft 6 higha•, speed to 0.8m-s-• reduces thepeakicecrystalnumber 7 low ac, densityto lessthan 4 cm-3 andimproves the agree- 8a low Scrit ment with the observations.Ice crystalgrowthratesare 9a 1.2 1.2 1.2 30 100 100 0 0 0 anotherimportantsource of uncertainty in the model. 10b slowgrow 1.2 100 0 For our baselinesimulation,we assumed that the depo- 11 sitioncoefficient (or massaccommodation coefficient)12 wasunity. Decreasing thisparameter to 0.5 resultsin 13 anincrease in thepeakicenumber density to 9.8cm-3. 14 The peak ice saturationin thesesimulations is al- 15 bet. her. bet. her. bet. IN IN IN IN IN 1.2 1.2 1.2 1.2 1.2 100 100 100 100 100 0.1 0.5 0.7 I 2 2.3 1.8 2.3 2.3 2.3 2.3 2.3 2.3 5.2 7.8 17.0 6.3 2.4 0.71 0.90 1.8 151.1 151.9 126.6 149.9 147.1 140.7 136.8 131.7 mostentirelycontrolled by the assumed freezingproperties of the sulfate aerosols.Changesin coolingrate The sensitivitiesof the peak ice crystal number densi- -60øC the sulfate aerosolsshouldfreezewhen the RHI erogeneous freezing icenuclei,Nh•t (cm-S),andthe width (cm-S),andpeakRHI (%) to updraftvelocity, vz or crystalgrowthratedonot affectthe peakRHI. We ties,Nice 1 3 havepreviously suggested that at temperatures below (m-s-), sulfateaerosolnumberdensity,No, (cm-), het- exceeds onlyabout125%(ratherthanabout150%in- of the aerosolsize distribution,a•,, are given. aThe aerosolfreezingrate was increased(decreased) redicated by ourcurrentfreezing parameterization). The May 2 wave-cloud measurements suggest that the ice suitingin a decrease(increase)in the peakRItI. coefficient wasdecreased to 0.5,resulting supersaturation required forhomogeneous freezing may bThedeposition be ashighas 160%.As indicated by the lastentryin in a 30% reductionin crystalgrowthrates. 1366 JENSEN ET AL.: ICE NUCLEATION ß ß ß i - ß ß i ß ß ß i ß - ß i - ß IN WAVE CLOUDS possiblyresultingin a decreasein the icecrystalnumber density. If a very large fractionof the sulfateaerosols contained effective freezing nuclei, then the observed high ice number densitiescould have been generated without intermediate liquid droplet activation. However, given the high crystal number densitiesin this case, it seemslikely that homogeneousfreezing domi- ß r'e,lce nated the ice nucleation process. lO 1 Summary Our model simulationsincludingonly homogeneous freezingnucleationpredict peak humiditiesand icenumber densitiesthat agreewith the SUCCESS wave-cloud -42 -40 -38 -36 -34 -32 observations(within the rangesof measurementand model uncertainties). The observationssupportthe Figure 4. The pe•k liquiddrople••nd ice crystal recent predictions that large supersaturationsare renumberdensity•re plo•ed versusminimum•emper•- quired to initiate homogeneous freezingof sulfateaero•ure for simulationswi•h • r•nge of ini•i•1 •emper•ures. sols. Sensitivity tests suggestthat no more than about The effec•iYer•dius of ice crystalsin •he simulations•e 0.5-1 cm-3 effective heterogeneous icenucleicouldhave •1soshown. The maximum number densityof ice crys- been presentgiven the observedhigh supersaturations tals forms• •emper•ures jus• low enoughfor liquid and ice number densities.We cannot rule out the possiwaterdropletsto form (near-37.5ø0). bility that smaller numbersof heterogeneous ice nuclei were present. In other types of cirrus cloudswith lower coolingrates, smallnumbersof effectiveice nucleicould peak RHI in the simulation(147%) is somewhatlarger dominate the ice nucleation process. than the observedpeak valuesof about 135%;however, Acknowledgments. This researchwas supported by the locations where the peak RHIs occurred may not NASA's SubsonicAssessment Program, directedby Howard have been sampledon this flight. Wesoky. The peak number of ice crystals is very sensitiveto the formation of activated liquid droplets prior to ice References nucleationin our simulations.Figure 4 showsthe peak ice crystal and liquid droplet number densitiesversus Collins, J. E., et al., A novel external path water vapor sensor,presentedat the AtmosphericEffects of Aviation the minimum temperature. If the temperatureis just Program5th annualmeeting,April 23-28, 1995. low enoughsuchthat the sulfatehazeaerosols freezebefore liquid dropsform, then the peak ice numberdensity Heymsfield,A. J., and R. M. Sabin,Cirruscrystalnucleation , , , i i i i i ß shouldbe lessthan 10 cm-3; whereasthe observations by homogeneousfreezingof solutiondroplets, J. Airnos. Sci., J6, 2252, 1989. onApril 30 showed icenumberdensities of 20-40cm-3. The largestnumberof ice crystalsis generatedif the temperature is just cold enoughsuchthat liquid drops are activated. Above this temperature, the ice number density decreaseswith increasingtemperature becausethe liquid dropletsget larger beforethey freeze. Hence,when the first few dropletsfreeze,the resulting large ice crystals deplete the vapor rapidly, resulting in rapid evaporationof the remaining dropletsbefore they can freeze. If the temperature drops just below -37.5øC suchthat the sulfate aerosolsfreezebefore liquid droplets are activated, then the peak ice crystal number density drops dramatically. This lower number density resultsfrom the fact that the aerosolfreezing rate is strongly dependent on their size and composition, and the freezing temperature varies substantially acrossthe aerosolsize range. Hence, the largest few aerosolsfreeze first, and the resulting crystalsdeplete the vapor before more aerosolscan freeze. The transition temperature of about -37.5øC predicted by our calculationsis consistentwith previous Heymsfield,A. J., and L. M. Miloshevich,Homogeneous ice nucleationand supercooledliquid water in orographic wave clouds,J. Atmos. Sci., 50, 2335, 1993. Heymsfield,A. J., and L. M. Miloshevich,Relativehumidity and temperature influenceson cirrus formation and evolution: Observationsfrom wave clouds and FIRE II, J. Atmos. S½i., 5œ,4302, 1995. Jensen,E. J., et al., Microphysicalmodelingof cirruspart I: Comparisonwith 1986 FIRE IFO measurements,J. Geophys. Res.,, 99, 10421, 1994. Jensen, E. J., and O. B. Toon, Ice nucleation in the upper troposphere: Sensitivity to aerosolnumber density, temperature, and coolingrate, Geophys.Res. Lett., , 21, 2019, 1994. Jensen,E. J., and O. B. Toon, The potential impact of soot particlesfrom aircraft exhauston cirrusclouds,Geophys. Res. Lett.,, œ•, 249, 1997. Liou, K.-N., S.C. Ou, and C. Koenig,An investigationon the climatic effect of contrail cirrus, Lecture Notes in En- grg., Vol. 60, Springer,Berlin, 138-153,1990. Sassen, K., et al., Mesoscaleand microscale structure of cirrus clouds: Three case studies, J. Atmos. Sci., •6, wave-cloud measurements [Heyrnsfield andMiloshevich, 371, 1989. 1995]. Their measurements indicateddropletswith Sassen,K., and G. C. Dodd, Haze particle nucleationsimulations in cirrus clouds, and applicationsfor numerical radii up to a few •m in a cloud leading-edgewith a and lidar studies,J. Atmos. S½i.,•6, 3005-3014,1989. temperature of-35øC. In a leading edge with a temperature of-37øC, dropletswere marginally detectable Tabazadeh,A., E. Jensen,and O. B. Toon,A modeldescriptionforcirruscloudnucleation fromhomogeneous freezing with radii of only 1-2 urn. In a penetration at -44øC, of sulfateaerosols,submittedto J. Geophys.Res., 1997. no dropletsformed. Also, the earlier wave-cloudmeasurementsshowedlowerice numberdensities(about 10 Eric. J. Jensen,NASA Ames ResearchCenter, MS 245-4, cm-3) in the penetration at -44øCthanin the warmer Moffett cases(about20 cm-3). Field, CA 94035. Heterogeneous freezingnucleicouldpotentially cause (ReceivedJune25, 1997; revisedNovember20, 1997; aerosolfreezingbeforeactivationinto dropletsoccurred, acceptedNovember25, 1997.)