Table Of Content1768 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME70
Sensitivity of Warm-Frontal Processes to Cloud-Nucleating Aerosol Concentrations
ADELEL.IGELANDSUSANC.VANDENHEEVER
ColoradoStateUniversity,FortCollins,Colorado
CATHERINEM.NAUD
ColumbiaUniversity,NewYork,NewYork
STEPHENM.SALEEBY
ColoradoStateUniversity,FortCollins,Colorado
DEREKJ.POSSELT
UniversityofMichigan,AnnArbor,Michigan
(Manuscriptreceived5June2012,infinalform24January2013)
ABSTRACT
AnextratropicalcyclonethatcrossedtheUnitedStateson9–11April2009wassuccessfullysimulatedat
highresolution(3-kmhorizontalgridspacing)usingtheColoradoStateUniversityRegionalAtmospheric
ModelingSystem.Thesensitivityoftheassociatedwarmfronttoincreasingpollutionlevelswasthenexplored
by conducting the same experiment with three different background profiles of cloud-nucleating aerosol
concentration.Totheauthors’knowledge,nostudyhasexaminedtheindirecteffectsofaerosolsonwarm
fronts.Thebudgetsofice,cloudwater,andraininthesimulationwiththelowestaerosolconcentrationswere
examined.Theicemasswasfoundtobeproducedinequalamountsthroughvapordepositionandriming,and
themeltingoficeproducedapproximately75%ofthetotalrain.Conversionofcloudwatertorainaccounted
fortheother25%.Whencloud-nucleatingaerosolconcentrationswereincreased,significantchangeswere
seeninthebudgetterms,buttotalprecipitationremainedrelativelyconstant.Vapordepositionontoice
increased,butrimingofcloudwaterdecreasedsuchthattherewasonlyasmallchangeinthetotalicepro-
ductionandhencetherewasnosignificantchangeinmelting.Theseresponsescanbeunderstoodintermsof
abufferingeffectinwhichsmallerclouddropletsinthemixed-phaseregionleadtobothanenhancedvapor
deposition and decreased riming efficiency with increasing aerosol concentrations. Overall, while large
changeswereseeninthemicrophysicalstructureofthefrontalcloud,cloud-nucleatingaerosolshadlittle
impactontheprecipitationproductionofthewarmfront.
1. Introduction etal.1989;Stewartetal.1998).Shouldaerosolsbeable
to alter the radiative, microphysical, and/or dynamical
Extratropicalcyclonesaretheprimarytransportersof
properties of extratropical cyclones, they could have
moisture,heat,andenergypolewardinthemidlatitudes
important consequences for global change scenarios.
(e.g., Trenberth and Stepaniak 2003; Schneider et al.
However,veryfewstudieshaveinvestigatedtheindirect
2006).Thefrontsassociatedwithextratropicalcyclones
effectofaerosolsonextratropicalcyclones,despitethe
contributeapproximately68%ofthetotalprecipitation
fact that both changes in pollution emissions globally
at midlatitudes (Catto et al. 2012). Additionally, their
andlocalseasonalenhancementsofaerosolconcentra-
deepandextensivecloudshieldshavelargeimpactsfor
tionsbecauseofburningcouldimpactthesesystems,as
radiationlocallyandonclimatescales(e.g.,Ramanathan
shownconceptuallybyWangetal.(2009).
Aerosols can affect cloud systems by changing the
microphysical structure of clouds. A change in cloud
Correspondingauthoraddress:AdeleIgel,ColoradoStateUni-
microphysicalstructurecanaffecttheradiativeproper-
versity,1371CampusDelivery,FortCollins,CO80523.
E-mail:[email protected] ties of clouds (Twomey 1977), change the lifetime of
DOI:10.1175/JAS-D-12-0170.1
(cid:2)2013AmericanMeteorologicalSociety
JUNE2013 IGEL ET AL. 1769
clouds(Albrecht1989),oraltertheratesofmicrophysical and Stewart 1997). Evaporational and sublimational
processes associated with precipitation production and coolingcanenhanceregionsofmesoscaledescent(Huang
latentheatrelease.Changesindiabaticheatingratescan and Emanuel 1991; Clough et al. 2000). Through these
then induce changes in vertical velocities and other dy- modifications of vertical velocity and circulation, as well
namicalfields.Understandingalloftheseindirecteffects as through local changes in the thermal structure, latent
andfeedbacksofaerosolsisanongoinganddifficulttask, heating has been seen to accelerate frontogenesis and
especially since the environment and dynamics of the frontalpropagationspeedforbothwarmandcoldfronts
systemitselfcanaltertheresponsetoaerosols(Matsui (Hsie et al. 1984; Szeto and Stewart 1997; Reeves and
et al. 2006; Fan et al. 2009; Khain 2009; Storer et al. Lackmann2004).
2010).Sincethemesoscaledynamicsofanextratropical Beforetheimpactsofaerosolsonwarmfrontscanbe
cyclonediffergreatlybetweenthecoldandwarmfronts, fully understood, the budgets of ice and liquid hydro-
a detailed study of all of the impacts of aerosols on meteormassinwarm-frontalcloudsmustbeexamined
extratropical cyclones is too extensive for one study; sinceaerosolindirecteffectsvarywithcloudphase.Such
therefore, thispaper will focus on just onepart of the budgetsofcloudiceandrain,butnotcloudwater(here
extratropicalcyclone,thewarmfront. andthroughoutthestudy,‘‘cloudwater’’means‘‘cloud
While there are no studies reported in the literature liquid water’’), have been studied before by Rutledge
that specifically examine the influence of aerosols on and Hobbs (1983) and Gedzelman and Arnold (1993)
warm fronts, aerosol impacts on other mixed-phase for warm fronts; however, both of those studies used
cloudsystemshavebeeninvestigated,andtheresultsof idealized, two-dimensional model setups with more
these studies are likely to be relevant to this study. primitive microphysical schemes than is used in the
Cheng et al. (2010) examined a mixed-phase front sys- present study. The current work will expand on these
tem in Taiwan and found that precipitation could be previousstudiesbypresentingbudgetsofice,rain,and
either increased or decreased with increasing cloud for the first time cloud water for warm-frontal clouds
condensationnuclei(CCN)dependingonchangesinthe fromathree-dimensional,high-resolutioncloud-resolving
depositional growth of ice and riming. CCN in mixed- simulation that employs sophisticated microphysics
phase clouds have been shown to result in decreased andaerosolparameterizationschemes.Thesensitivity
rimingefficiencyinobservations(Borysetal.2003)and ofthe budgetsandlatentheatingprofilesofthewarm
modeling studies of mixed-phase orographic clouds front will then be tested by altering the background
(Saleeby et al. 2009; Xue et al. 2012) because of de- profileofcloud-nucleating aerosolconcentration intwo
creased supercooled cloud droplets sizes. Both obser- additionalsimulations.
vations and modeling studies of these clouds show In section 2, the model will be described and a syn-
changesinthespatialdistributionofprecipitation(e.g., opticoverviewoftheselectedstormgiven.Insection3,
GivatiandRosenfeld2004;JirakandCotton2006;Lynn data analysis methods will be described. Section 4 in-
etal.2007;Saleebyetal.2009)butdonotalwaysagree cludesthebudgetsofice,cloudwater,andrain,inwarm-
on the magnitude or sign of the total precipitation frontalclouds.Insection5thesensitivityofthesebudgets
change,whichcandependonenvironmentalfactorsand andthewarmfrontasawholetoperturbationsinaerosol
theextentofthemixed-phaseregion(Muhlbaueretal. concentrationsisdiscussed.
2010; Zubler et al. 2011). Many results of mixed-phase
convectivecloudstudiesindicatethathigherCCNcon-
centrations can cause more liquid water to be lofted 2. Caseoverview
above the freezing level leading to additional freez-
a. Caseselectionandmodelsetup
ing and latent heat release (Khain et al. 2005; van den
Heeveretal.2006).Theenhancedbuoyancyandfeed- To approach this problem, an extratropical cyclone
backstotheupdraftareoftenreferredtoastheaerosol wassimulatedusingtheRegionalAtmosphericModeling
invigorationeffect,butthiseffecthasnotbeenstudied System(RAMS)(Pielkeetal.1992;Cottonetal.2003).
extensivelyinnonconvectivecloudtypes. Thecaseselectedwastheextratropicalcyclonethatde-
In addition, should aerosols be able to significantly velopedintheleeoftheRockyMountainsandtracked
alterlatentheatingrates,resultsofstudiesthatexamine acrosstheUnitedStatesfrom9to11April2009.Thiscase
thesensitivityofwarmfrontstolatentheatingmayalso wasselectedforseveralreasons.Mostimportantly,ithad
be relevant to this study. First, modeling studies have a long-lived and well-defined warm front. Additionally,
shown that latent heat processes, including condensa- no other extratropical cyclone was near enough to in-
tion and melting, lead to mesoscale banded features teractwiththisstorm,whichwouldfurthercomplicatethe
over the warm-frontal surface (Hsie et al. 1984; Szeto analysis. Third, the case was chosen because it did not
1770 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME70
move from water to land or vice versa until after the andprecipitationofthe warmfrontwithout beingtoo
simulationperiod.Inthiswaytransitionsbetweenwater memoryintensive.
andlandsurfacesdonotcomplicateinterpretationofthe RAMSexplicitlytrackstheratesofmanycloudpro-
results. cesses. Those that will be presented here (the names
The simulations were set up with an outer and inner used in the figures are given in parentheses) are total
two-waynestedgrid,thelocationsofwhichareshownin vapordeposition and sublimation of ice species (vapor
Fig. 1a. The outer grid has 250 3 300 grid points with to ice), total condensation and evaporation of cloud
15-kmhorizontalspacing(3750 km34500km)whilethe species(vaportocloud)andrain(vaportorain),cloud
inner grid has 697 3 1027 grid points with 3-km hori- droplet nucleation (cloud nucleation), conversion of
zontalspacing(2091km33081 km).Whileportionsof cloud water to rain through autoconversion of cloud
frontsandextratropical cycloneshave beenrunatfiner dropletsandaccretionofclouddropletsbyrain(cloudto
resolution (e.g., Garvert et al. 2005; Molthan and Colle rain),melting (melting),riming ofcloud droplets(rim-
2012), to our knowledge, no extratropical cyclone in its ing of cloud) and raindrops (riming of rain), and col-
entiretyhasbeensimulatedatsuchahighgridresolution lectionoficebyrainresultinginthemeltingoftheice
before.Bothgridshave45verticalsigmalevels,with75-m (ice to rain). These allow for the budgets of ice, cloud
spacingnearthesurfacestretchingto1-kmspacingatthe water, and rain to be explicitly calculated in the simu-
model top. The time steps on the inner and outer grids lations.Notethoughthatthetwoprocessescomprising
were 5 and 20s, respectively. A cumulus parameteriza- eachofthefollowingpairs—depositionandsublimation,
tion scheme based on Kuo (1974) and Molinari (1985) condensation and evaporation, and the cloud to rain
wasimplementedontheoutergridonly.TheHarrington processes—are not tracked separately and thus cannot
(1997) radiation scheme and a Smagorinsky (1963) tur- beexaminedindividually.
bulence scheme with modifications by Lilly (1962) and ThemodelwasinitializedusingGlobalForecastSystem
Hill(1974)wereimplementedonbothgrids. (GFS)analysisat0000UTC9April2009andintegrated
The simulations use a two-moment bin-emulating mi- for 48h. The lateral boundaries were nudged using ad-
crophysicalschemethatincludesfiveprognosticicespe- ditionalGFSanalyses.Theprofileofthoseaerosolsthat
cies:pristineice,snow,aggregates,graupel,andhail,and canserveascloudcondensationnuclei(CCN)wasmaxi-
threeprognosticliquidspecies:twoclouddropletmodes mized at the surface and decreased linearly from the
and rain (Meyers et al. 1997; Saleeby and Cotton 2004, surfacevalueupto4 kmabovewhichitwasconstantat
2008).Graupelandhailaredistinguishedbydensityrather 100cm23.Themodelwasrun3timeswithsurfaceaerosol
thansizeintheRAMSmicrophysicsscheme(Walkoetal. concentrations of 400, 800, and 1600 cm23; these con-
1995).Theirdensitiesaredefinedtobe0.3and0.9g cm23, centrationsarerepresentativeofclean,average,andpol-
respectively.Thesedensitiesforgraupelandhailallow luted conditions over continents (Andreae 2009). These
foratransitionindensityasiceparticlesbecomemore sensitivityrunswillbereferredtoasCCN400,CCN800,
heavily rimed or gradually melt and therefore allow andCCN1600,respectively.Apartfromthedifferencesin
the fall speeds of these particles to be more accurately initial aerosol concentrations, the model setups were
computed. Graupel and hail both have diameters of identical.Thecloud-nucleatingaerosolparticleswereal-
approximately 1–2 mm in these simulations; however, lowedtobeadvectedhorizontallyandverticallybutwere
thehailconcentrationisabout4timeslargerthanthatof notdepletedduringcloudnucleation.Topreventrunaway
graupel and dominates the mixing ratio because of its nucleation, aerosol particles were nucleated under ap-
higherdensity.Whilehailisnottypicallyassociatedwith propriateenvironmentalconditionsonlyiftheirnumber
warmfronts,the‘‘hail’’producedbythemodelismore concentrationdidnotexceedthenumberconcentrationof
appropriatelydescribedasheavilyrimediceorpartially cloud droplets present in the same grid box. Had the
melted ice, particularly below the melting level. Both cloud-nucleating aerosol been depleted, most particles
graupelandhailcanbe,butarenotnecessarily,partially wouldhavebeenrainedoutquickly,leavingaveryclean
melted.Therefore,wewillrefertothemasthe‘‘mixed- environmentinallthreecasessincenosurfacesourcesof
phase’’speciesthroughouttheremainderofthetext. aerosolwerepresent.Additionallytheaerosolswerenot
By using a bin-emulating scheme, each species is radiativelyactive,sonoassessmentoftheaerosoldirect
binned by size for the calculation of collection effi- effectismadeinthisstudy.
cienciesandfallspeeds;thesecalculationsgiveamore
b. Synopticoverview
accurate representation of riming and autoconversion
than can be simulated by using a traditional bulk mi- Figure 1 shows output from CCN400 depicting the
crophysics scheme. This parameterization scheme is evolution of the storm. At 1200 UTC 9 April, a broad
thereforewellsuitedforthestudyofthemicrophysics trough exists at both 300 and 500 hPa over the Rocky
JUNE2013 IGEL ET AL. 1771
FIG.1.(a),(c),(e),(g)Plotsof300-hPawindspeed(shaded;ms21)and500-hPageopotentialheight(contoured;m)
and(b),(d),(f),(h)surfacepotentialtemperature(shaded;K)andgeopotentialheight(contoured;m)outputfromthe
outergridat(a),(b)1200UTC9Apr,(c),(d)0000UTC10Apr,(e),(f)1200UTC10Apr,and(g),(h)0000UTC
11Apr.Theredandbrownlinesin(b),(d)arewarmfrontsanddrylines,respectively,andthebluelinein(f)isacold
frontanalyzedbyhand.Thethickpinklinein(f)istheobjectivelocationofthewarmfront.Frontsemanatingfrom
thestormcenterwerenotdetectedatthefinaltimeshownin(h).
1772 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME70
Mountains associated witha closed low at 850 hPabe-
ginning to formin eastern Colorado and westernKan-
sas.Thetroughat500 hPahasformedaclosedlow,the
downstreamridgehasweakened,andthelowat850 hPa
hasdeepened12 hlater.Thereisanenhancedpotential
temperature gradient at 850 hPa extending eastward
from the low center, consistent with the presence of
confluentbackgroundflow(Schultzetal.1998).Thisis
thewarmfrontthatwillbethefocusofthisstudy.There
is a tongue of warm air reaching the center of the cir-
culationfromthesouthandwestthatisassociatedwith
adryline(Figs.1b,d).Thisboundarywasassociatedwith
a severe weather outbreak in the southeast on 9–10
Aprilbutwillnotbediscussedinthispaper.
At1200UTC10Aprilthesystemisnearitspeakin-
tensity.Thelowcentershavebecomeverticallystacked
and beginto weaken shortly afterthis time. The warm
frontofinterestisnowassociatedwithadistincttrough
in the 850-hPa field and its eastern end is beginning to
interactwiththeAppalachianMountains.Thejetstreak
at 300 hPa in the northeast has intensified but shifted
littleinthehorizontalandislikelyaidinginlarge-scale
liftoverthewarmfront.By0000UTC11Aprilthelow
centerisjusttothewestofthemountains.Asitisnotthe
objective of this study to examine the influence of the
mountainsonthewarmfront,thesimulationwasended
here, though the system did continue to the Atlantic
OceanandregainedstrengthovertheGulfStream.
Figure2showsatime-andspace-averaged,front-normal
crosssectionofcloudice,cloudwater,andrainfromthe
CCN400 run of the warm front. Details on how these
cross sections were created are described in section 3.
The isentropes indicate that the warm front is shallow
withaslopeofabout1:150andthatthefrontalcloudis
predominantlystratiform.Itcanbeseenthatthemixed-
phase region is very deep, extending from about 2 to
6 km and that the melting level located near 2–3 km
throughoutmostofthesystem.
c. Validation
Although we are simulating an actual extratropical
cyclone,thegoalofthisresearchisnottopresentanin-
depthanalysisofacasestudy.Rather,weareinterested
in the sensitivity of the warm front to perturbations in
cloud-nucleating aerosol concentrations. As a conse-
quence, the results are treated more like those of an
idealizedmodelingstudy,wherethecasestudydataare
simplyusedtosimulatethestormtypeofinterest,inthis
case an extratropical cyclone. Therefore, it is not im- FIG.2.Time-averagedcross-sectionsof(a)cloudice,(b)cloud
portant that the model accurately capture the time water, and (c) rain mixing ratio (gkg21). Thin black lines are
isentropes(K),andthethickblacklineisthe08Clineormelting
evolutionorexactstructureofthecyclone,onlythatthe
level.Thegray,pink,andredlinesshowthe0.005gkg21contours
model simulates an extratropical cyclone that shows
ofcloudice,cloudwater,andrain,respectively,forreference.
average characteristics that are representative of such
JUNE2013 IGEL ET AL. 1773
FIG.3.(a)Radar-derivedestimateofthe24-hprecipitationtotalending1200UTC10AprfromtheNational
WeatherService.(b)Precipitationtotalforthesametimeperiod,butfromRAMSoutput.
asystem.However,itisusefultoknowthatthemodel using the 900-hPa level for detecting surface fronts, we
physicsaretoafirstordercorrectandthatthemodelhas found that the surface fields were much less noisy and
capturedsomebasicfeaturesoftheactualsystem.Total produced cleaner results than other near-surface levels.
precipitationisagoodfieldtouseforthissinceitcon- ThetechniqueproposedbyHewson(1998)wasoriginally
tains information about the storm location as well as conceived for fields of about 100-km horizontal resolu-
rainfall. tion,butheretheresolutionis3 km,sowehadtomake
Figure3ashowstheNationalWeatherService(NWS) severaladjustments.Wefirstsubsampledu andV
surf G_surf
Multisensor Precipitation Estimator (Lawrence et al. every30kmandthenappliedasmoothingroutine.These
2003)rainfallandFig.3btheCCN400totalprecipitation twofieldswerethenfedintothefrontdetectionroutine.
for the 24-h period ending 1200 UTC 10 April on the Onlythosefrontsidentifiedbytheroutinewithpotential
inner grid. The general pattern of precipitation is well temperature gradients greater than 25 3 1026 K m21
captured,especiallyaroundthewarmfront.RAMScor- werekept,andthesignoftheproductofgeostrophicwind
rectly places the maximum in precipitation in northern velocity and potential temperature gradient delineated
Missouri andsouthernIllinoisand Indiana,thoughper- betweenwarmandcoldfronts.
hapsitisshiftedslightlynorthandeast.Comparisonsof Despite the upscaling and smoothing, multiple warm
modeloutputtoNWSsurfaceanalysesalsoshowashift frontsweredetected,andadditionalmaskswereappliedto
to the northeast of the modeled location of the warm objectivelyisolatetheprimaryfront.Warm-frontalpoints
frontandcyclonecenterrelativetotheirobservedloca- alongcoastlineswereremoved.Thesepointsexistedsim-
tions (not shown). The local maxima in precipitation in plybecauseofthelargecontrastinpotentialtemperature
RAMSarehigherthanthoseseenintheradarestimate, betweenlandandoceananddidnotrepresentthewarm
but the average totals of 1–4 cm seem to be consistent. front associated with the extratropical cyclone. Warm-
RAMSdoesunderpredictrainfallinthesoutheastasso- frontal points were also required to have cloud cover
ciated with the secondary warm front, but that is not within 60 km to the north in order to eliminate the
a major concern for the analysis that will be presented boundary oriented north–south in the warm sector of
hereasourfocusisontheprimarywarmfront. thestorm (Figs. 1b,d) that was associatedwiththese-
vereweatheroutbreakmentionedearlier.Next,warm-
3. Methods frontal points with four or fewer neighbors within
150 km, one or zero neighbors within 60 km, or zero
a. Warmfrontdetection
neighborswithin300 kminthenexthourwerethrown
Toobjectivelylocatethewarmfrontateachhour,we out. The effects of these conditions were to remove
usedmodeledpotentialtemperatureu andgeostrophic short-lived warm-frontal structures. Finally, a 108 poly-
surf
windvelocityV ,andappliedthefrontdetectional- nomial was fit to the remaining warm-frontal points in
G_surf
gorithmofHewson(1998)onequal-sigmasurfacesrather order to preserve the shape and location of the warm
than equal-pressure surfaces as in previous work (e.g., front while smoothing clusters of points and to recover
Naud et al. 2012). While Hewson (1998) recommended theoriginal3-kminnergridresolution.
1774 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME70
FIG.4.Hovmo€llerdiagramsofthemodelaverageu,y,andpotentialtemperatureatthesurfaceasafunctionofdistancefromthewarm
front.Seetextforfurtherdetails.
Thefinalobjectivelocationsofthewarmfrontateach precipitation are present in most of the subsetted do-
hour(see,forexample,theoneinFig.1f)showthatthe main (Fig. 2). In this way, the number of points being
detection algorithm effectively locates the warm front averaged is the same in each case and any differences
suchthatitiscontinuousintime.Itperformsbestfrom between simulations must be due to actual changes in
hours27–42ofthesimulationduringwhichtimealmost thequantitybeingshown,ratherthanchangesinsample
theentirelengthofthefrontpropagatesforwardateach size.Theexceptionsarehydrometeordiameters,which
time step. It is over this interval that the analyses pre- are averaged only where they are nonzero and are
sentedherewillbeconducted. weighted by the number concentration of the proper
species.Allverticalprofilesshowninthefigurestofol-
b. Averaging
lowareaveragedfromthesurfacelocationofthefront
Theobjectivewarm-frontlocationswereusedtocreate to525 kmnorthofthefront.Areasimmediatelysouth
averagecrosssectionsthroughthefront.Thedistanceto of the surface front did not receive substantial pre-
the storm center and length of the front varies in time. cipitation and thereforewere not included in theaver-
However,tomaintainaconsistentdatavolumeforeach ages.Byaveragingto525 kmaheadofthefront,nearly
time,warmfrontsweredefinedtobegin60 kmeastofthe the full extent of the warm-frontal cloud is captured,
lowpressurecenter(thelargestdistancebetweenthelow thoughatitsfarthestreaches,onlyiceexistsandliquid
pressurecenterandthefirstpointofanyobjectivefront) speciesarenotpresent(Fig.2).
andtobe573 kmlong(thelengthoftheshortestfrontin
the simulation). The warm fronts are nearly parallel to
4. Cloudbudgets
thewest–eastdirectionateachtimestep(Fig.1f,forex-
ample),sonorotationofthedatawasnecessary.Model Figure5showsaverageverticalprofilesofeachofthe
output was taken along a north–south transect at each mostimportantprocessescontributingtotheproduction
gridpoint alongthe front andaveraged alongthe east– or depletion of ice species, cloud water, and rain in
westdirectiontoobtainasinglecompositecrosssection CCN400. The ice budget (Fig. 5a) shows that vapor
perpendiculartothefrontateachhour.Themeridional depositionisthedominantgrowthprocessaboveabout
spancapturedthebulkoftheprecipitationandcloudfield 4.5 km whereas riming is dominant below that level.
associatedwiththewarmfront.Figure4showsHovmo€ller This being a warm spring case (Fig. 1), very little ice
diagramsofthezonalwindu,themeridionalwindy,and reachesthesurfaceinthesimulationandthereforemost
potential temperature at the surface as a function of ice that is producedmusteventuallymelt. Sublimation
distance from the front. It can be seen that the warm- and the collection of ice by rain are minor sinks of ice
frontdetectionandaveragingcorrectlyplacesthefront comparedtomelting.Theseresultsliebetweenthetwo
on the warm side of the baroclinic zone and that the cases examined by Rutledge and Hobbs (1983). In the
cross-front wind shifts from positive to negative as ex- feeder zone of their weak updraft case they found no
pected. This indicates that the front detection and av- contributiontotheicebudgetbyriming,whereasinthe
eragingiseffectivelycapturingthefeaturesofthefront. feederzoneoftheirmesoscaleupdraftcasetheyfound
Inmanystudiesofaerosolindirecteffects,quantities the riming rate to be 2–3 times greater than the de-
are averaged only over regions where some definition positionalgrowthrate.
of cloud is met. However, in this study no minimum NeitherthestudybyRutledgeandHobbs(1983)nor
thresholdforcondensateisusedinanyoftheaveraging that by Gedzelman and Arnold (1993) examined the
ofcloudprocessesorcloudpropertiessincecloudand/or cloudwaterbudget.Thecloudwaterbudgetinthisstudy
JUNE2013 IGEL ET AL. 1775
FIG.5.Verticalprofilesofeachofthemajorprocessescontributingtothebudgetof(a)ice,(b)cloudwater,and(c)rain.Notethemelting
levelisnear2km.Positive(negative)valuesindicategrowth(decay)ofthespecies.Seethetextforafulldescriptionofeachprocess.
(Fig. 5b) shows two peaks in production, one in the ofprecipitationinwarm-frontalrainbands(Houzeetal.
mixed-phaseregionofthecloudnear4 kmandonejust 1981;RutledgeandHobbs1983).Thoughnodistinction
belowthemeltinglevelnear2 km.Theupperpeakap- has been made between banded and nonbanded pre-
pears to be caused by updrafts creating supersaturated cipitationinthisstudy,theseresultsalsosuggestthatice
conditions while the lower peak may exist because of crystals falling through the mixed-phase region of the
supersaturated conditions being created by the latent cloudefficientlycollectcloudwaterthatwouldnototh-
cooling due to melting. The cloud nucleation rate is erwise be converted to rain; this collection contributes
slightly greater than the condensational growth rate at significantlytothetotalprecipitation.
both peaks and significantly greater in the region be-
tween the peaks. The main process contributing to the
depletionofcloudwaterateachpeakisdifferent.Atthe 5. Sensitivitytochangesincloud-nucleating
upper peak, cloud-droplet-to-rain conversion rates are aerosolconcentrations
smallandrimingratesdominate.Howeveratthelower
a. Microphysicalcloudproperties
peak, cloud droplet to rain conversion rates are larger
thantherimingrate,thoughrimingisstillsignificantat Thesensitivityofthemassbudgetsandcloudproper-
the melting layer. Cloud droplet sizes and rain mixing ties to increases in cloud-nucleating aerosol concentra-
ratiosarebothmuchsmallerattheupperpeak(Fig.6), tionswillnowbediscussedbycomparingresultsfromthe
which would reduce the rates of autoconversion and CCN400, CCN800, and CCN1600 simulations. Figure 6
accretion.Whenintegratedthroughthecolumn,riming shows average vertical profiles of hydrometeor number
contributes about 33% more to the depletion of cloud concentration,mixingratio,anddiameterinthetopthree
waterthandocloudtorainconversionprocesses. rows.ThepercentagechangefromCCN400toCCN1600
FinallytherainbudgetisshowninFig.5c.Meltingof inthesequantitiesisshowninthebottomrow.Asdis-
ice is by far the most important producer of rainwater, cussedinsection2a,pristineice, snow, andaggregates
a result found by both Rutledge and Hobbs (1983) and (PSA) have been averaged together, the mixed-phase
GedzelmanandArnold(1993).Theconversionofcloud species(GH)arecombined,asarethetwoclouddroplet
water to rain is of secondary importance. The budgets modes.Althoughtheresponsesofeachicespeciestothe
showthatabout75%ofrainmassisduetomeltingofice, sensitivitytestsdiffer,groupingtheminthiswayshows
wherehalfoftheicemassisformedthroughdepositional theoverallimpactofchangesinaerosolconcentrations.
growthandhalfthroughtherimingofcloudwater.The ThePSAnumberconcentration, diameter,and mixing
remainingroughly25%ofrainmassisproducedthrough ratio are dominated by aggregates below 8 km. It is
warm rain processes in this case. The seeder–feeder important to note that profiles of mixing ratio are not
processisthoughttoberesponsiblefortheenhancement expected to match the total budget profiles in Fig. 5
1776 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME70
FIG.6.Averageverticalprofilesof(toprow)numberconcentration,(secondrow)mixingratio,and(thirdrow)numberconcentration
foreachofthefourgroupsofhydrometeorsdescribedinthetext.Diameterswereaveragedonlywheretheywerenonzeroandweighted
bynumberconcentration,butnominimumthresholdwasusedinaveragingnumberconcentrationsormixingratios.(bottomrow)The
percentagechangeofCCN1600fromCCN400fornumberconcentration,mixingratio,anddiameterofeachhydrometeorgroup.
sinceadvectionandsedimentationwerenotincludedin nearlyzerobelowthemeltinglevel.Overallthereisless
thebudgetcalculations. massinthiscategoryabovethemeltinglevelinthemore
Withincreasingpollutionclouddropletsaremorenu- polluted cases, but again, below the melting level dif-
merous and smaller which is expected according to the ferences are very small. Throughout the column, rain
cloudalbedoeffect(Twomey1977),andtheoverallcloud exhibitsverylittlevariationinmixingratioamongthe
water path is increased. PSA mixing ratio is highest in threesimulations.Therearemonotonictrendsinnumber
CCN1600below8 km,thoughCCN400andCCN800are concentrationabove2 kmwheredropsarelessnumerous
similar;thetrendbecomesclearlymonotonicat4 kmand butbelow2 km,thetrendisnonmonotonic.Thereisalso
below. Fewer GHparticles exist with enhancedaerosol noappreciablevariationinraindropletsizeinthelowest
concentrations although the differences in number are 4 kmofthecolumn.
JUNE2013 IGEL ET AL. 1777
FIG.7.Verticallyintegratedandaveragedratesofeachofthemajorwaterbudgetprocesses.
The two numbers printed with each process are the percentage change of CCN800 and
CCN1600,respectively,fromtheCCN400value.
Insightintothesedifferencesincloudpropertiescanbe that show suppressed rain formation in polluted sce-
gainedbyexaminingthechangesinbudgetterms(Fig.7). narios (e.g., Albrecht 1989; Xue and Feingold 2006;
Cloud condensation rates increase, but not enough to Saleeby et al. 2010). Figure 8 shows average vertical
compensate for the decrease in cloud nucleation rates, profilesoftherainmassproducedthroughcloud–cloud
such that there is less production of cloud water with collisions (autoconversion, Fig. 8a) and the rain mass
increasing cloud-nucleating aerosol concentrations. The growththroughcloud–raincollisions(accretion,Fig.8b).
trendsinthesetwotermsopposeoneanothersincenu- Itcanbeseenthattheautoconversionofclouddroplets
cleation and condensation compete for the available toraindecreasesasexpectedbecauseofthesmallercloud
watervapor.Inthemorepollutedcases,totalcondensa- droplet size, but that the accretion of cloud droplets by
tional growth increases because of the increased total rain increases with increasing aerosol concentrations.
surface area of the more numerous but smaller cloud Sinceformationofraindropsinthiscaseoccursprimarily
droplets (not shown). This change effectively decreases throughmelting,theycangainlargesizeswithoutgrow-
the supersaturation and inhibits cloud nucleation. The ingthroughautoconversionofcloudwater.Collisionef-
overall increase in cloud water mixing ratio seen pre- ficiency between cloud droplets and raindrops is not
viouslyisultimatelyaresultofdecreasedrimingcaused usuallyverysensitivetoclouddropletsize,soraindrops
bythedecreasedsizeoftheclouddroplets. will be able to accrete more cloud water in the more
Therearemanysimilaritiesbetweenwarm-frontaland polluted cases despite decreases in cloud droplet size
orographicclouds—asloped surface,weaktomoderate sincemorecloudwatermassisavailablebecauseofthe
updraft speeds and the presence of a mixed phase to decreased autoconversion. It is unclear whether the
name three—such that a comparison with results from nearly exact cancellationbetween clouddroplet auto-
orographiccloudsensitivitystudiesisuseful.Thetrends conversion and cloud–rain collisions in this case is
inthethreetermsdiscussedaboveareconsistentwiththe fortuitousornot.Additionally,notethatabouttwiceas
resultsofSaleebyetal.(2009)wholookedattheimpacts muchrainmassisgainedthroughcloud–raincollisions
of CCN on an orographic snowfall case. In their study than through autoconversion. This reemphasizes the
though, theyfounddecreasedsensitivitytorimingrates factthatverylittlerainisbeingproducedindependent
forCCNconcentrationsabove500 cm23,whereasinthe oficeprocesses.
present study significant decreases in riming are seen In the more polluted simulations PSA mixing ratio
evenfortheCCN1600case(Fig.7). increaseswhilethatofGHdecreases,asdiscussedabove.
Interestingly, there is very little change in cloud-to- The budget terms in Fig. 7 show that vapor deposition
rain processes across the three simulations such that increases,whilerimingdecreaseswithincreasingaerosol
they do not contribute to changes in mixing ratio of concentrations.Vapordepositionoccursprimarilyonthe
cloudwaterorrain.Thesameresultwasseeninoneof PSA species, so the increase in that process alone can
thesimulationsofanorographicmixed-phaseraincase explaintheincreaseinPSAmass.Mostofthechangein
examined by Muhlbauer et al. (2010). However, this vapordepositionisconfinedtothemixed-phaseregionof
resultisincontrastwithotherpurelywarmrainstudies thecloud(Fig.8c).Thischangeinvapordepositionmay
