Microphysical variability of Amazonian deep convective cores observed by CloudSat and ...

18 Pages·2017·2.13 MB·English

Atmos.Chem.Phys.,18,6493–6510,2018
https://doi.org/10.5194/acp-18-6493-2018
©Author(s)2018.Thisworkisdistributedunder
theCreativeCommonsAttribution4.0License.
Microphysical variability of Amazonian deep convective cores
observed by CloudSat and simulated by a multi-scale
modeling framework
J.BrantDodson1,PatrickC.Taylor2,andMarkBranson3
1ScienceSystemsandApplications,Inc.,Hampton,VA,USA
2ClimateScienceBranch,NASALangleyResearchCenter,Hampton,VA,USA
3DepartmentofAtmosphericScience,ColoradoStateUniversity,Ft.Collins,CO,USA
Correspondence:J.BrantDodson([email protected])
Received:19September2017–Discussionstarted:6October2017
Revised:5March2018–Accepted:12March2018–Published:8May2018
Abstract.Recentlylaunchedcloudobservingsatellitespro- in the spatiotemporal distribution of precipitation (includ-
vide information about the vertical structure of deep con- ing floods and drought), radiation, and other climate sys-
vection and its microphysical characteristics. In this study, tem elements (Arakawa, 1975). Simulating convection re-
CloudSatreflectivitydataisstratifiedbycloudtype,andthe lies on our understanding of the physics controlling and
contoured frequency by altitude diagrams reveal a double- modulatingitsbehavior,includingcloudmicrophysics,cloud
arc structure in deep convective cores (DCCs) above 8km. scaledynamics(updraft/downdrafts),entrainment,andother
Thissuggeststwodistincthydrometeormodes(snowversus large-scale atmospheric interactions, cloud–surface interac-
hail/graupel) controlling variability in reflectivity profiles. tions,andcloud–radiationinteractions(Randalletal.,2003;
Theday–nightcontrastinthedoublearcsisaboutfourtimes Arakawa,2004).
larger than the wet–dry season contrast. Using QuickBeam, Atmospheric convection exhibits variability on multiple
the vertical reflectivity structure of DCCs is analyzed in timescales, including diurnal and seasonal. The convective
two versions of the Superparameterized Community Atmo- diurnalcycle(CDC),awell-documentedandimportantmode
spheric Model (SP-CAM) with single-moment (no graupel) ofvariability,isparticularlypronouncedoverland(Yangand
and double-moment (with graupel) microphysics. Double- Slingo, 2001; Nesbitt and Zipser, 2003; Tian et al., 2004;
momentmicrophysicsshowsbetteragreementwithobserved KikuchiandWang,2008;Yamamotoetal.,2008).TheCDC
reflectivityprofiles;however,neithermodelvariantcaptures is characterized by a rapid insolation-driven transition from
the double-arc structure. Ultimately, the results show that shallow to deep convection in the early afternoon, followed
simulating realistic DCC vertical structure and its variabil- byeitheraslowdecaythroughtheeveningandearlymorn-
ity requires accurate representation of ice microphysics, in ing,ortransitionintomesoscaleconvectivesystems(MCSs),
particularthehail/graupelmodes,thoughthisaloneisinsuf- persistingintothenextmorning(Machadoetal.,1998;Nes-
ficient. bitt and Zipser, 2003). Geostationary satellite observations,
inparticular,haveprovidedaglobalviewofthespatialcom-
plexityoftheCDC(YangandSlingo,2001).However,most
1 Introduction satellitedatasetsonlyobserveconvectivecloudtopproper-
tiesusingpassivelysensedvisibleandinfraredradiancesand
As a driver of the hydrological cycle, the frequency and in- cannot sense the cloud interior. With passive sensors, it is
tensityofatmosphericdeepconvectioninfluencesspatialand difficulttoclearlydistinguishbetweendeepconvectivecores
temporalcharacteristicsofprecipitation.Ourabilitytosimu- (DCCs) and related anvils, the latter having a diurnal cycle
lateconvectivebehavioronshort(diurnal)andlong(climate offset from DCCs by approximately three hours (e.g., Fu et
change) timescales significantly modifies projected changes al.,1990;Linetal.,2000;Zhangetal.,2008).
PublishedbyCopernicusPublicationsonbehalfoftheEuropeanGeosciencesUnion.
6494 J.B.Dodsonetal.:MicrophysicalvariabilityofAmazoniandeepconvectivecores
Asmallnumberofsatellites,suchasCloudSat(Stephens naturesofdifferentcloudtypes,includingthosegeneratedat
et al., 2008), carry radars that penetrate cloud tops, sensing differenttimesintheconvectivelifecycle,creatingaclearer
theinteriorofdeepconvection.Spaceborneradarsallowthe viewoftheday–nightcontrast.
examination of deep convection invisible to most satellites, The Amazon Basin is an ideal location for studies of the
especiallyinternalstructureandmicrophysics.CloudSatcar- CDC for multiple reasons. First, the Amazon has a well-
ries a W-band radar which is specifically attuned to ob- defined,high-amplitudecontinentalCDC,peakingregularly
servesmallercloudliquidandicehydrometeors.Othersatel- in the mid-afternoon (Yang and Slingo, 2001). Amazonia
lites,suchasTropicalRainfallMeasuringMission(TRMM) has a prominent recurring propagating coastal squall line,
(Kummerowetal.,1998)andtheGlobalPrecipitationMea- and various secondary local effects related to orography
surement (GPM) mission (Hou et al., 2014) are attuned and the Amazon River (Janowiak et al., 2005; Burleyson et
to larger precipitation-sized hydrometeors. A sizable body al., 2016). However, aside from the aforementioned squall
of literature describes observations of tropical convection line, it lacks the major diurnally propagating signals ob-
using these satellites (e.g., Petersen and Rutledge, 2001; servedinothercontinentalconvectiveregions(e.g.,thecen-
Schumacher et al., 2004; Nesbitt and Zipser, 2003; Liu et tralUnitedStatesandsouthernChina)(Wallace,1975;Daiet
al.,2007;LiuandZipser,2015;LiuandLiu,2016).Inaddi- al.,1999;Zhouetal.,2008)thatmakegeneralizingregional
tiontoradars,therearelidarssuchasthatcarriedbyCloud- CDCstudiesdifficult.Second,Amazoniahasawell-defined
Aerosol Lidar and Infrared Pathfinder Satellite Observation wet and dry season, allowing a seasonal examination of the
(CALIPSO) (Winker et al., 2009), which has been used to CDC including distinct meteorological forcing for convec-
examinethepropertiesofconvectionandanvils(e.g.,Sassen tion:locallyforced(commoninthedryseason)versusnon-
et al., 2009; Riihimaki and McFarlane, 2010; Del Genio et locallyforced(commoninthewetseason).Finally,theAma-
al.,2012). zonian CDC alters the top-of-atmosphere radiative diurnal
Spaceborne cloud radars, precipitation radars, and lidars cycle(Taylor,2014a,b;DodsonandTaylor,2016),represent-
offer complementary views of tropical convection. Cloud ing an influence on regional and global climate that meteo-
radars sample the higher altitudes and anvils of DCCs, as rologicalreanalysesandclimatemodelsstruggletosimulate
well as associated stratiform clouds, but are ineffective at (Yang and Slingo, 2001; Itterly and Taylor, 2014; Itterly et
lower altitudes where precipitation-sized hydrometeors at- al.,2016).
tenuate the radar beam. Precipitation radars examine the This paper documents and describes a detailed view of
lower-andmid-levelstructureofDCCs,butcannotseecloud thediurnalvariabilityoftheconvectiveverticalstructureob-
topsandanvils.Lidarsaresensitivetothetopsofthickclouds servedbyCloudSat.Oneofthekeymethodsusedtoaccom-
and can measure their altitude with high precision; how- plishthisistoseparatethevariabilityinconvectivefrequency
ever,theyareunabletopenetratethetopsofDCCsandmost from the variability in radar reflectivity. This new perspec-
anvils.Inthisstudy,wewillfocusonthehigh-levelfeatures tive not only clarifies previous findings, but also reveals a
ofDCCsandanvils,whereCloudSatismostuseful. unique,previouslyunreporteddouble-arcstructureintheav-
CloudSat’s polar orbit, crossing the Equator during early erageradarreflectivityprofileofdeepconvection.Thisnew
afternoon and early morning, provides two views of the finding relates to the ice microphysical structure relevant to
CDC, near the beginning and end of the mean precipita- convective dynamic and thermodynamic properties, includ-
tion diurnal cycle over land. This ability has not been well- ingprecipitationrate,downdraftand coldpoolstrength(af-
exploited in the literature. The study by Liu et al. (2008) fectedbyevaporationandsublimationofhydrometeors),la-
represents one of the few analyses of observed day–night tentheatingverticalprofile(andassociatedwarminganddry-
differences using CloudSat. They surveyed day–night con- ingoftheconvectiveenvironment),anddetrainedwatermass
trastsbetweenreflectivityprofilesoverbothtropicallandand (McCumberetal.,1991;Grabowskietal.,1999;Gilmoreet
ocean, finding that high reflectivity clouds occur more fre- al.,2004;Lietal.,2005).Basedonourresults,thesimulation
quentlyatnightthanduringthedayatallaltitudesexceptat of deep convective characteristics, and the comparison be-
cloudtops(13km).Theauthorsinterpretthisdifferenceasa tween simulation and observations, benefits from a detailed
consequenceoftheCDC,inwhichthepeakindeepconvec- representationoficemicrophysics.
tivefrequencyoccursafterthe13:30LSTCloudSatoverpass
time,whiletherearestillfrequentlingeringMCSsduringthe
01:30LST overpass. However, previous efforts muddle the 2 Dataandmethodology
physicalinterpretationbymixingthefrequencyofbothshal-
lowanddeepconvection,theverticalconvectivereflectivity 2.1 Observations
profile, and the properties of other cloud types. Resolving
theseissues,weexaminetheday–nightcontrastinthereflec- CloudobservationsaretakenfromCloudSat,acloudobserv-
tivity profile of mature deep convection after stratifying by ing member of the A-Train (Stephens et al., 2008), orbiting
cloudtype.Thismethodologyallowstheseparationofcon- at 705km altitude, 98◦ inclination, and with an equatorial
vectivefrequencyfromtheindividualreflectivityprofilesig- crossing time of 01:30am/pm local time. The primary in-
Atmos.Chem.Phys.,18,6493–6510,2018 www.atmos-chem-phys.net/18/6493/2018/
J.B.Dodsonetal.:MicrophysicalvariabilityofAmazoniandeepconvectivecores 6495
strumentistheCloudProfilingRadar(CPR),a94GHzradar parameterization(amongotherthings)oftheCommunityAt-
witha1.1kmwideeffectivefootprintand480mverticalres- mosphericModel(CAM)withacloud-resolvingmodel(the
olution,oversampledtocreatea240meffectiveverticalres- System for Atmospheric Modeling, described by Khairout-
olution. CloudSat operated as designed from June 2006 to dinov and Randall, 2003), coupled within each GCM grid
March2011,untilsufferingabatterymalfunction.Thistime point.Thisisastudyofopportunity,usingdatamadeavail-
periodservesasthetemporaldatadomain. able from past work, and so the time domain is limited to
For cloud-type stratification, the CPR cloud mask and an Amazonian dry season in the early 21st century. Only
radar reflectivity fields from the 2B-GEOPROF product datafrom02:00and14:00LSTareincludedinthisanalysis,
(Marchand et al., 2008) are used. The cloud-type stratifica- which are closest to the CloudSat overpass times of 01:30
tion identifies the four following cloud types: DCCs, anvils and13:30LST.WeusetwoversionsofSP-CAM,employing
(AVN), clouds attached contiguously with DCCs (CLD-D), single-moment (SPV4) and double-moment (SPV5) micro-
and other clouds (CLD). DCCs are identified using a CPR- physics.Hydrometeormixingratiosforcloudice,cloudwa-
based methodology on a profile-by-profile basis (i.e., with ter,rain,snow,andgraupel(doublemoment-only)takenfrom
noconsiderationtoneighboringcolumns)accordingtothree the cloud-resolving model (CRM) component of SP-CAM
criteriafromDodsonetal.(2013): areusedtosimulatetheassociated94GHzreflectivityprofile
usingtheQuickBeamradarsimulator(Haynesetal.,2007).
1. mustbeatleast10kmtall
The CRM is based on the System for Atmospheric Model-
2. musthaveacontinuousverticalregionofreflectivityof ing,andisrunintwo-dimensionalmode,with4kmhorizon-
atleast−5dBZbetween3and8kmaltitude tal spacing and approximately 100m–1km vertical spacing
(varyingbyaltitude).TheformulationofSPV5,andthedif-
3. the maximum reflectivity value at any altitude in the ferencesbetweenSPV4andSPV5,aredocumentedbyWang
middletroposphere(3to8km)mustbeatleast0dBZ. etal.(2011).TheonlymajordifferencesbetweenSPV4and
SPV5 of direct relevance to deep convection are the micro-
Theloweraltitudeboundincriterion2israisedto5kmwhen
physical and radiative parameterizations; we attribute pri-
heavyprecipitationisdetected(indicatedbylowsurfacere-
marydifferencesbetweenSPV4andSPV5tomicrophysics.
flectivity),accountingforattenuationeffects.Thelattertwo
A major difference between the SPV4 and SPV5 micro-
criteria restrict the set of profiles with deep cloud layers to
physicsisthetreatmentofprecipitatinghydrometeors.SPV4
thosewhichlikelycontainactive,vigorousDCCsonly.These
hasdiagnosticvariablesforsnowandgraupel.TheSPV4mi-
criteria are guided by the cloud definitions used in creat-
crophysics scheme predicts only non-precipitating and pre-
ing the 2B-CLDCLASS product (Wang and Sassen, 2001).
cipitating hydrometeors, which are partitioned into frozen
When the data are stratified by these criteria, CloudSat ob-
andliquidbytemperature.Snow andgraupelarediagnosed
served187457verticalprofilesofDCCsintheAmazonover
fromfrozenprecipitatingwater.However,snowandgraupel
the time domain, with just less than half (92071, or 49%)
asdistincthydrometeorspeciesdonotplayaroleintheprog-
occurringduringthedaytimeoverpasses.
nostic microphysical equations in SPV4. In contrast, snow
Figure1showsthespatialdomainoftheanalysisregion,
centered on northern and central South America (25–0◦S, andgraupelareprognosedhydrometeortypesinSPV5,and
70–50◦W).Inaddition,aCloudSatoverpassthroughthedo- aretreatedasadistinctspeciesintheprognosticmicrophys-
ical equations. So in addition to the reflectivity field calcu-
main displays the associated cloud presence, morphologies,
lated from precipitating ice in SPV4, we examine the in-
andradarreflectivityforasingleoverpass(Fig.1a–c),aswell
fluence of partitioning the precipitating ice into diagnosed
as the identified cloud types in the bottom panel (Fig. 1d).
snow graupel on the simulated radar reflectivity field. This
This particular overpass provides an example of the stratifi-
newvariantofSPV4isherebyreferredtoasSPG4.Notethat
cationmethod,displayingavarietyofdeepconvectivecloud
therearenodifferencesbetweenSPV4andSPG4otherthan
systems and associated anvils, including narrow single-cell
the reflectivity fields simulated by QuickBeam. This lets us
updrafts to the north, and wider multi-celled convection to
distinguish between the effects of adding graupel to the hy-
thesouth.Note,onlyasubsetoftheprofileswithindeepcon-
drometeorspecies,andtheeffectofswitchingfromsingle-to
vective cloud systems are labeled DCCs – this is deliberate
double-momentmicrophysics.Becausethehydrometeorpa-
toincludeonlytheprofileswhichlikelycontainactivecon-
rametersinQuickBeamaresimilarforprecipitatingiceand
vectiveupdrafts.
snow,themaineffectofthepartitioningistheincreasedre-
2.2 Modeling flectivityfromdiagnosedgraupel.
It is difficult to directly compare satellite-retrieved and
Toinvestigatetheabilityofmodelstosimulatetheobserved model-simulatedconvectivecloudice(Waliseretal.,2009).
DCCverticalstructureandtheinfluenceofmicrophysics,we Radarreflectivityservesasasubstitutebasisforcomparison,
usetheSuperparameterizedCommunityAtmosphericModel wheremodelreflectivityiscomputedwitharadarsimulator.
(SP-CAM)(Khairoutdinovetal.,2005).SP-CAMisamulti- However,icephasemicrophysicalpropertiesofDCCsarea
scale modeling framework (MMF) replacing the convective key component in determining the model-simulated reflec-
www.atmos-chem-phys.net/18/6493/2018/ Atmos.Chem.Phys.,18,6493–6510,2018
6496 J.B.Dodsonetal.:MicrophysicalvariabilityofAmazoniandeepconvectivecores
Figure1.(a–c)ExampleofaCloudSatcross-sectionalobservationofafternoonconvectionon11December2008atapprox.17:33UTC
(13:33LST). Left-to-right on the x axis corresponds with south-to-north. (a) is the cloud mask product from 2B-GEOPROF, with colors
representingthecloudmaskvaluecorrespondingwithcertaintyofcloudidentification;(b)isradarreflectivity;and(c)iscloudtype.Red
indicatesDCCs,blueindicatesanvils,darkgrayindicatescloudsattachedcontiguouslywithDCCs,andlightgrayindicatesotherclouds.
◦ ◦
(d)MapofnorthernSouthAmericawiththestudyregion(25–0 S,70–50 W)markedwiththeredbox.Theheavyblacklinecrossingthe
studyregionindicatesthepathoftheCloudSatswathshowninpanel(a).ThedominantcloudtypeobservedbyCloudSatalongthepathis
indicatedbythecoloreddots.
Atmos.Chem.Phys.,18,6493–6510,2018 www.atmos-chem-phys.net/18/6493/2018/
J.B.Dodsonetal.:MicrophysicalvariabilityofAmazoniandeepconvectivecores 6497
Figure2.Toprow:verticalfrequencyprofilesof(a)cloudoccurrence,(b)cloudtopheights,and(c)cloudbaseheights.Blacklinesarefor
day/night,andred(blue)isforday-only(night-only).Solid(dashed)lineisthemean(standarddeviation).Middlerow:sameasthetoprow,
butfor(d)DCC-onlyoccurrence,(e)topheights,and(f)baseheights.Bottomrow:sameasthetoprow,butfor(g)anvil-onlyoccurrence,
(h)topheights,and(i)baseheights.
tivityprofile.Thiscreatesachallengeforinterpretingsimu- 3 PropertiesofconvectionasobservedbyCloudSat
latedreflectivity,asitisdifficulttodetangletheinfluencethat
modelmicrophysicshasonthesimulatedreflectivityprofile 3.1 Meancloudproperties
and its relationship with other properties of simulated con-
vection (e.g., vertical updraft velocity). Observed reflectiv- Untanglingtheinfluenceofconvectivefrequencyonthedeep
ity profiles are affected by multiple ice hydrometeor types, convectiveverticalreflectivitystructurebenefitsfromanex-
including both snow and graupel/hail. In order for models amination of the mean CloudSat-observed cloud properties
to realistically simulate DCC reflectivity profiles, and thus (Fig. 2). First, we will look at the frequency of all cloud
allowforrobuststatisticalreflectivitymodel/observationin- types, and then subset the clouds into DCCs and anvils.
tercomparison (in the vein of Liu et al., 2008), the models Clouds occur in a layer between 1.5 and 12km in altitude,
mustsimulatebothicehydrometeorsrealistically(afunction with small maxima in cloud occurrence frequency (COF)
of the microphysics) and the relationships between the hy- at 11.5 and 2.5km (Fig. 2a). The vertical profile of cloud
drometeorsandotheraspectsofconvection(e.g.,verticalve- top heights (Fig. 2b) shows four regions of interest – a pri-
locity).Therefore,itisdifficulttostrictlyandsimplyattribute mary maximum at 13km, a secondary maximum at 1.2km,
model–observationaldifferencestospecificaspectsofthepa- a broad enhanced frequency region between 2.5 and 6km,
rameterizations. Nevertheless, it is still possible and useful and a small maximum at 7.5km. The cloud base heights
toshowtheaggregateeffectsthatthechoiceofmicrophysics (Fig.2c)showalargeprimarymaximumat1.2km,abroad
have on the simulated reflectivity field, and so (at least par- secondary maximum centered at 11km, and a small local
tially)accountformodel-observationdifferences. maximum at 5km. These features are consistent with the
identified tri-modal vertical cloud structure in convectively
www.atmos-chem-phys.net/18/6493/2018/ Atmos.Chem.Phys.,18,6493–6510,2018
6498 J.B.Dodsonetal.:MicrophysicalvariabilityofAmazoniandeepconvectivecores
active tropical regions (Johnson et al., 1999; Khairoutdinov
et al., 2009), with shallow convective clouds, cumulus con-
gestus,andDCCswithassociatedanvilscomprisingthebulk
ofAmazonianclouds.TheDCC-anvilandshallowcumulus
modesaremoreprominentinthedatathanthecumuluscon-
gestus mode; this might be related to the greater variability
oftopheightsforcongestusthantheothercloudtypes,which
areconstrainedbythelevelofneutralbuoyancy(forDCCs-
anvils)andtheatmosphericboundarylayertop(forshallow
cumulus).
Mean COF vertical profiles differ significantly between
day and night, differences at most altitudes have p values
(two-tailed t test) (cid:28) 0.01. High (low) level clouds are en-
hanced during night (day), and the opposite suppressed. In
addition,highaltitudeclouds(above12.5km)aremorefre-
quent during day than night. The cloud top height profiles
(Fig. 2b) show a contribution to high level cloud frequency
fromadaytimetopheightincrease(i.e.,tallerDCCs,likely
Figure3.Thecontouredfrequencybyaltitudediagrams(CFADs)
with overshooting tops), indicating that despite there being ofreflectivityforDCCsinAmazonia.Thecolorsrepresenttheprob-
fewerdaytimehighcloudstheyaretallerthanthoseatnight. ability density function (PDF, as percentage) of radar reflectivity
DCCs occur in 3% of CloudSat profiles over Amazonia ateach240mtalllayerobservedbyCloudSat.Thevertical(hori-
(Fig. 2d), with mean top heights near 14km (Fig. 2e). The zontal)blacklineindicates0dBZ(8km).Threedistinctfeaturesof
tallestDCCsreachanaltitudeof18km,whichpenetratethe theCFAD,discussedinthetext,arelabeledonthefigure.(a)indi-
tropopauseandlikelycontributetostratosphere–troposphere catesthedarkband,withthehorizontallinesegmentsshowingthe
meanaltitude.(b)and(c)markthelocationsofthehighandlow
interactions(JohnstonandSolomon,1979;Cortietal.,2008;
reflectivityarcs,respectively.Thediagonallinesegmentsshowthe
Averyetal.,2017).DCCsareonaverageabout0.5kmtaller
orientationofeacharc,andtheroughvaluesofthePDFmodes.
during the day than night (also significant at p (cid:28) 0.01).
Anvil cloud frequency (Fig. 2g) peaks at 12km, with anvil
top heights (Fig. 2h) reaching their maximum at 13.5km
(0.5km lower than DCCs). Anvil bases occur in a broad
layer between 5km (by definition the lowest altitude) and ing enhanced beam attenuation from melting hydrometeors
11km,diminishingwithheightabovethisaltitude(Fig.2i). at the freezing level. Reflectivity in the higher cloud alti-
The5kmlowerlimitofanvilbases(where5kmischosenas tudes(above7.5km)decreaseswithheightprimarilythrough
beingnearthefreezingline)isevidentiallyanartificiallimit reduction in hydrometeor size – this is because, assuming
imposedbythemethodology,andtheremaybenocleardis- Rayleigh scattering and ignoring phase changes, hydrom-
tinction between anvil clouds and deeper free-tropospheric eteor size dominates reflectivity (proportional to the sixth
cloudsinnature. power of diameter) (Battan, 1973). Large hydrometeors fall
out of the updraft more rapidly than small hydrometeors,
3.2 Deepconvectivereflectivityprofiles leadingtoverticalsizesorting.
The CFAD associated with the DCC vertical profile dis-
Thefrequencycomponentofthedataforvariouscloudtypes plays an interesting feature above 8km. While the CFAD
hasbeenisolatedanddescribed,sonowtheverticalstructure followsthecharacteristicarcshapeatloweraltitudes,inthe
variabilityisopenforexamination.Themeanverticalprofile upper troposphere the CFAD splits into two arcs. The low-
ofreflectivityinDCCsovertheAmazon,aswellasthetotal reflectivity arc decreases below 0dBZ at 11km, whereas
variability,arepresentedascontouredfrequencybyaltitude the high-reflectivity arc remains above 0dBZ at 14km.
diagrams(CFADs),wherethecoloredshadingrepresentsthe The double-arc structure most likely indicates two differ-
probability density function of reflectivity at each altitude entmodesofhydrometeors:alow-reflectivityarcassociated
(Fig.3).Previousresearch(e.g.,Bodas-Salcedoetal.,2008; withsnowandahigh-reflectivityarcassociatedwithgraupel
Satoh et al., 2010; Nam and Quaas, 2012) associates deep and hail. Cloud ice has typical reflectivity values below the
convection with a characteristic arc shape in the reflectivity minimum detection threshold of CloudSat (−28dBZ), and
CFAD, maximizing in the middle troposphere and decreas- doesnotcontributeasmuchtotheCFADastheothericehy-
ingatupperandloweraltitudes.Asimilarshapeisapparent drometeor species. This double-arc structure is not obvious
inFig.3.Reflectivityisreducednearthesurfacefromradar (andthusunreported)inpreviousstudiesexaminingradarre-
beamattenuationbyraindrops(Sassenetal.,2007).Thekink flectivityprofilesindeepconvection(e.g.,Bodas-Salcedoet
in the reflectivity profile at 5km is a ”dark band” mark- al.,2008;Satohetal.,2010;NamandQuaas,2012)largely
Atmos.Chem.Phys.,18,6493–6510,2018 www.atmos-chem-phys.net/18/6493/2018/
J.B.Dodsonetal.:MicrophysicalvariabilityofAmazoniandeepconvectivecores 6499
Figure4.(a)TheCFADforDCCs,sameasFig.3.Theblackcurvesnearthecenterofthedataaretheaveragereflectivityprofiles.Dashed
linesarethestandarddeviationbounds.(b)Sameas(a),butforanvils.
becausetheDCCsarenotcleanlyseparatedfromothercloud otherconvectiveproperties(e.g.,frequency,topheight,pre-
types,leadingtoablurredreflectivitystructure. cipitation,radiativeeffects.Liuetal.(2007)suggestthatthis
Howdoweknowthatthedoublearcsareassociatedwith metricmaybemoreusefulforcharacterizingconvectivein-
different hydrometeor species? Figure 4b shows the reflec- tensitythancloudtopheight,atraditionalconvectivemetric.
tivityCFADsforanvilclouds.Insteadofadouble-arcreflec-
tivitystructureabove8km,anvilshaveasinglearcwithre- 3.3 Dayversusnightandwetversusdryseason
flectivitywellbelow0dBZabove10km.TheCFADclosely variability
resembles those constructed by Yuan et al. (2011), specifi-
callyforthickanvils,wherereflectivityvaluesof0dBZare TheDCCreflectivityprofiles,inparticulartheCFADdouble-
frequentataltitudesof8–10km,anddecreasesrapidlywith arc structure, show significant day–night and wet–dry sea-
height. However, Yuan et al. (2011) show the reflectivity son variability. The day–night contrast results show the up-
maximum extending 1–2km higher in altitude than Fig. 5 per cloud reflectivity is larger during day than night, by up
does. to4.5dBZat12.5km(Fig.4j–l,solidline),whichiscaused
This single reflectivity arc corresponds with the low- byamoreprominenthigh-reflectivityarcduringthedaythan
reflectivity(i.e.,snow)arcobservedinDCCs.Thisresultis night.Thisfeaturealsosupportstheconclusionthatdaytime
consistent with the hydrometeors in anvils consisting of the updraftvelocitiesarehigherthannighttimevelocities.Thisis
snowandcloudicedetrainedfromDCCs(modifiedbycloud consistentwiththecontinentalCDC,asdescribedpreviously.
processes as the anvils age), while dense ice hydrometeors Nighttime convection (midway between the afternoon peak
eitherfailtobedetrainedintotheanvilorquicklysediment andmorninglull)islikelytobeweakeningand/ortransition-
from the anvil base immediately adjacent to the DCC. This ing to MCSs (Machado et al., 1998), and exists in an envi-
corresponds with in situ measurements of anvil hydrome- ronmentpartiallycontaminatedbyearlierconvectionandnot
teors from West African convection (Bouniol et al., 2010). beingrejuvenatedbyinsolation(Chaboureauetal.,2004).
Note that some graupel particles are likely detrained into These results clarify the day–night contrast presented by
theanvilsproducedbyDCCswithstrongupdrafts(Cetrone Liuetal.(2008).InCloudSatobservations,DCCsaremore
andHouzeJr.,2009).However,theseparticlesarenotlarge frequentduringnighttimethandaytime.However,theDCCs
and/ornumerousenoughtocreatethedouble-arcstructurein occurring during the daytime overpass have larger updraft
theanvilreflectivityCFADs. velocitiesthanthoseatnight.Thepositiveday–nightreflec-
Thepresenceoftwodistinctgroupsofhydrometeorsinthe tivitydifferenceintheuppercloudextendswellbelow12km
upper cloud indicates a fundamental mode of variability in (the altitude indicated by Liu et al., 2008), and represents a
theDCCreflectivityprofile.Higher(lower)reflectivityinthe day–nightcontrastinthemicrophysicalpropertiesoftheice
upper cloud indicate a larger (smaller) ratio of hail/graupel phaseinDCCs.Insummary,DCCsobservedduringtheday-
particlestosnow.Dense,large,reflectiveparticlesgenerated time overpass are less frequent, but taller, with larger verti-
inDCCswithhigherverticalvelocitiesareloftedhigherinto calvelocities,andmoreicehydrometeorsinthehail/graupel
the upper cloud (Liu et al., 2007), linking the upper cloud phases,thantheDCCsobservedatnight.
reflectivitytoupdraftvelocity.Thisrelationshipcanbeused The double-arc reflectivity structure in the upper tropo-
asaproxymetricofconvectiveintensity,andcomparedwith sphereexhibitsseasonaldifferences.Amazonianconvection
exhibitsstrongseasonalvariationsinthefrequencyandother
www.atmos-chem-phys.net/18/6493/2018/ Atmos.Chem.Phys.,18,6493–6510,2018
6500 J.B.Dodsonetal.:MicrophysicalvariabilityofAmazoniandeepconvectivecores
Figure5.(a–i)TheCFADsofreflectivityforDCCsinAmazoniaseparatedbytimeofdayandseason.Theleftcolumnisforallfourseasons
(wet,dry,wet-to-dry,anddry-to-wet);themiddlecolumnisforthewetseason(WET);andtherightcolumnisthedryseason(DRY).The
firstrowisresultsforbothtimesofday,thesecondrowisday-only,andthethirdrowisnight-only.(j–l)Thedifferencebetweenthedayand
nightmeanreflectivityprofilevalues–i.e,thesecondrowminusthethirdrow.
propertiesbecauseofchangesinforcingmechanisms(Fuet filesduringthewetanddryseasons.Season-specificresults
al.,1999;Marengoetal.,2001;RaiaandCavalcanti,2008), show the same qualitative pattern as the annual results –
also connected with variability in day–night contrasts. Fig- higher reflectivity during day (night) than night in the up-
ure 5 shows the day–night contrast in DCC reflectivity pro- per(lower)cloud.Theamplitudeofthedifferenceissimilar
Atmos.Chem.Phys.,18,6493–6510,2018 www.atmos-chem-phys.net/18/6493/2018/
J.B.Dodsonetal.:MicrophysicalvariabilityofAmazoniandeepconvectivecores 6501
inbothseasons,butthealtitudesofenhanceddaytimereflec- ingthatincludingdiagnosedgraupelenhancesthereflectiv-
tivity is limited to above 10km in the dry season. Overall, ity profile by no more than 4dBZ. SPV5 microphysics pre-
theday–nightradarreflectivitycontrastisfourtofivetimes dictsgraupel,andtheuppertropospherereflectivityislarger
larger than the wet–dry season contrast, underscoring day– thanSPG4,showingadditionalimprovementintheswitchin
nightcontrastsasamajormodeofdeepconvectivevariabil- microphysicalschemes.ButSPV5hasalesser(butstillno-
ity. ticeable) disagreement with observations. No model variant
TheCFADsshowadditionaldifferencesbetweenwetand reproduces the observed double-arc structure, suggesting a
dryseasons.WetseasonCFADsshowawell-defineddouble fundamentaldeficiencyinrepresentingthebehavioroflarge
arcreflectivitystructureintheuppercloud,whereasthedry icehydrometeorsinconvectiveupdrafts.
seasonCFADsdonot,particularlyatnight.Thisresultmight Figure 7 shows that the convective updraft velocities in
beaconsequenceofthedrierthermodynamicalenvironment both SPV4 and SPV5 never exceed 5ms−1, which is un-
andaerosolcharacteristicsofthedryseasonenvironment(in- realistically low for deep convection in the Amazon (Gian-
cluding anthropogenic aerosol from biomass burning) (e.g., grandeetal.,2016).Thisislikelyamajorcontributortothe
Andreaeetal.,2004;Linetal.,2006).However,itmayalso lowreflectivityinthesimulatedreflectivityabove10km,and
be a sampling artifact due to the smaller number of DCCs is likely related to the coarse resolution of the CRM (Petch
duringthedryseason(9616)versusthewetseason(78034). etal.,2002;Bryanetal.,2003;Khairoutdinovetal.,2009).
To test the sample size influence, we implement a Monte However, vertical velocity is not the sole contributor to the
Carlostylerandomsamplingmethodologytoreducethewet size of the simulated reflectivity. Surprisingly, the disagree-
season sample size to that of the dry season. Reducing the mentbetweenSPV4andSPV5doesnotdirectlycorrespond
sample size of the wet season to that of the dry season ob- with a proportionally large change in simulated updraft ve-
scures the double-arc reflectivity structure (not shown), so locity profile. Despite upper cloud reflectivity being higher
theinfluenceofseasonalityonthedouble-arcstructurecan- in SPV5 than SPV4, mean updraft velocity in the upper
notclearlybeattributedtoseasonalchangesintheconvective clouddecreases(andturnspositiveinthelowertroposphere).
environment. In addition, SPV4 DCCs have net negative velocity below
3km. This may not seem like an intuitive result initially,
andclosertothepropertiesofstratiformprecipitation.How-
4 Comparisonwithsimulatedcloudfromamulti-scale ever, it likely represents the thermal “bubble” nature of at-
modelingframework mosphericconvection(ScorerandLudlam,1953;Batchelor,
1954; Carpenter et al., 1998; Sherwood et al., 2013; Morri-
Thesefindingsraisequestionsaboutongoingmodelingstud- son, 2017). The DCC identification method favors columns
ies that use simulated radar reflectivity as a metric for con- withhighreflectivityinthemid-touppertroposphere.These
vectiveactivity.TheobservedCFADsdepictcomplexstruc- columnsusuallycontainstrongupdraftsatthesamealtitudes,
ture and variability in convective reflectivity. Can models whichistheconvectiveupdraftthermal.Incontrast,nearthe
replicatethis?Inthissection,wewillexaminetheabilityof surface, the high-buoyancy air has already been evacuated
SP-CAM to replicate the properties of DCCs observed by into the thermal aloft, leaving neutral or negatively buoyant
CloudSat, in particular the double-arc reflectivity structure. airinthelowertroposphere.Theselectionprocessisnotper-
Inaddition,becausethemodelprovidesadditionalinforma- fect, and Fig. 7a and b show that strong downdrafts are oc-
tionaboutthesimulatedatmospherethatcannotbeeasilyob- casionallyincludedinthesetofDCCprofiles.Nevertheless,
served,suchasverticalupdraftvelocity,wewilllookatthe thenetsinkingmotionbelow3kminSPV4isconsistentwith
relationships between the radar reflectivity fields and other deepconvection.
aspectsofthesimulatedconvection.
4.2 Simulatedreflectivitystratifiedbyupdraftvelocity
4.1 Simulatedreflectivityandverticalvelocityprofiles
Theargumentwepresentreliescriticallyontherelationship
Figure6displaystheCFADsofAmazonianDCCsforSPV4 betweenconvectiveupdraftvelocity,thegraupelphaseofmi-
and SPV5. Both versions produce reflectivity values more crophysics, and radar reflectivity. These are difficult to un-
than 10dBZ lower than observed above 5km in altitude. weave inthe observations, because ofa lack ofdirect verti-
Specifically, the observed graupel/hail branch of the reflec- calvelocityobservations,andthelimitationsofmicrophysi-
tivityarcismissinginbothmodelversions.SPV4ispartic- calretrievals(particularlyinsceneswithheavyprecipitation
ularlyunrealistic,asthemicrophysicsschemedoesnotrep- for W-band radars, Mace et al., 2007). However, it is pos-
resent graupel. SPG4 diagnoses graupel, but their effect on sible to separate them in the simulation by stratifying ver-
the reflectivity profile is minor. There is an enhancement of tical reflectivity and hydrometeor profiles by updraft veloc-
reflectivityof2dBZat6km,anda1–2dBZreductionofre- ity.InFigs.8–11,theCRM-levelverticalprofilesassociated
flectivityelsewhereintheprofile.Thisisaresultofswitch- withDCCsareconditionallysampledbythemaximumpos-
ingfromnon-precipitatingicetosnowinQuickBeam,mean- itiveverticalvelocity(denotedhereafterasW )occurring
max
www.atmos-chem-phys.net/18/6493/2018/ Atmos.Chem.Phys.,18,6493–6510,2018
6502 J.B.Dodsonetal.:MicrophysicalvariabilityofAmazoniandeepconvectivecores
Figure6.(a,b,c,d)CFADsofdryseasonDCCreflectivityinAmazoniafrom(a)CloudSat,(b) SPV4,(c)SPG4(i.e.,SPV4withdiagnosed
graupelincluded),and(d)SPV5,asinFig.4.(e,f,g)Thedifferenceofthemeanverticalreflectivityprofilebetween(f)SPG4andSPV4,
(f)SPV5andSPV4,and(g)SPV5andSPG4.
Figure7.(a,b)CFADsofDCCverticalupdraftvelocityfrom(a)SPV4and(b)SPV5.TheformatisthesameasforthereflectivityCFADs.
NotethatSPV4andSPG4verticalvelocities(notshown)areidentical.(c)ThedifferencebetweenSPV4andSPV5meanverticalvelocity
profiles.
Atmos.Chem.Phys.,18,6493–6510,2018 www.atmos-chem-phys.net/18/6493/2018/