Aug 27, 2001 - Naval Research Laboratory, Monterey, California. G. L. Stephens and R. T. ... [Kaufman, 1998]) and National Polar-orbiting Opera- tional Environmental ..... loted Aircraft Studies (CIRPAS) Twin Otter, a manned aircraft which ...
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 106, NO. D16, PAGES 17,981-17,995,AUGUST 27, 2001
GOES 10 cloud optical property retrievals in the context of vertically varying microphysics S. D. Miller Naval Research Laboratory, Monterey, California
G. L. Stephensand R. T. Austin Department of Atmospheric Science,Colorado State University, Ft. Collins, Colorado
Abstract. An optimal estimation approach is applied to the physical retrieval of single-layercloudopticalproperties(optical depth and effectiveradius) usingmultispectral imager channelson the western Geostationary Operational Environmental
Satellite (GOES 10). The retrieval includesdiagnosticinformation pertaining to uncertainty and dependenceon a priori assumptionsrequired by the forward model.
Satelliteretrievalsof 0.65-/•m cloudopticaldepth and effectiveradius (micrometers) for marine stratocumulus(in both drizzle and drizzle-freeconditions)during the CloudSat Antecedent Validation Experiment and tropical cirrus during the Atmo-
sphericRadiation Measurement(ARM)-Unmanned AerospaceVehiclespringflight seriesare examinedtogetherwith data from the NASA/Jet PropulsionLaboratory AirborneCloudRadar (ACR), the ARM Cloud DetectionLidar, and the Colorado State UniversityScanningSpectralPolarimeter(SSP) instruments.Optical depths are found to be consistentbetween GOES and SSP after taking into account the
degradationof responsivityin GOES channel1. ColocatedACR/GOES observations support past evidence that passive satellite detection of drizzle size droplets (in terms of a significantpositive bias in the retrieved effective radius) may be
possibleundercertainconditions.The ability of a dual lidar/radar activeobserving systemto provide independentinformation over different cloud particle size regimes is illustrated and speaksto the vertical variability of cloud microphysics.The GOES estimate of effective radius was found to be more representative of the upper 0.5 km of the cirrus
1.
clouds examined.
Introduction
Clouds play an integral, dynamic role in the climate systemin terms of distributing both radiative and latent heating in the horizontaland vertical, mechanisms that in turn drive the atmospheric circulations responsible for cloud formation. As we become increasingly
tern. The implication is that indirect methods,basedon the identification of certain understood and predictable feedback processes,may be our only recourse to addressingthe questionspertinent to the role of anthropogenicallyaltered cloudsin climate stability/change. A soundphysicalrepresentationof the entire systemoperating in concert is therefore required, and hence our common interest is to improve the ability of global climate models (GCM) to capture these processes.The
cognizantof our ability to alter the environment(e.g., through such actionsas fossilfuel burning), a correspondingawarenessof the need to understandour partreatment of cloud and cloud feedbacks, in which many
ticular role in climate change has emerged. This reof the aforementioned climate responsesare thought to quires the ability to separate the natural, or internal, be manifest, is recognized as one of the most glaring variability of a highly nonlinear systemfrom the signal
arisingfrom external (e.g., anthropogenic)forcing. To confusethis issuefurther, we recognizethat these exter-
nal forcing signalsare not alecoupledfrom, but rather convolvedintrinsically within, the greater climate sysCopyright 2001 by the American GeophysicalUnion. Paper number 2000JD000057.
0148-0227/01/2000JD000057509.00
weaknessin our current ability to simulate/predictre-
alistically the present and future climatic states. Satellite remote sensingprovides the most efficient platform for gathering the quantity and diversity of cloud information needed to both observe and improve upon the numerical modeling of global climate. As we enter into an era of atmospheric sciencewhere numeri-
cal weatherprediction(NWP) and global-scalesatellite observations beginto converge(e.g., direct assimilation 17,981
17,982
MILLER
ET AL.: GOES CLOUD RETRIEVALS
of satelliteradiances[seeJoinerandDee,2000]),impor- ciliary data. For simplicity,only single-layercloudprotant questionsconcerningthe extent to which satellite- files are considered here. Passive data from the National providedinformation can addressexistingmodel needs OceanicandAtmosphericAdministration(NOAA) Geohave cometo the fore. A diversearray of new environ- stationaryOperationalEnvironmental Satellite10 (curmentalsatellites(e.g., Earth ObservingSystem(EOS, rently,GOES West) imagerare usedtogetherwith two [Kaufman,1998]) and National Polar-orbitingOpera- active sensors:the National Aeronauticsand SpaceAdtional EnvironmentalSatelliteSystem(NPOESS [e.g., ministration (NASA)/JetPropulsion Laboratory(JPL) LockheedMartin Missiles and Space,1995; $chueler and AirborneCloudRadar (ACR) and the AtmosphericRaBarnes,1998]))carryinginnovativepassiveinstruments diationMeasurement (ARM) [e.g.,Pattinoset al., 1991; operating acrossboth the optical and microwave por- Stokesand Schwartz,1994]UnmannedAerospace Vehitions of the spectrum will soon be complementedby cle (UAV) program[Stephens et al., 2000b]CloudDeactive cloud and aerosolprofiling devices.For example, tection Lidar (CDL). Section2 describesthe passive a profiling cloud radar on CloudSat (G. L. Stephens and active instrumentation. Section 3 provides a brief et al., The CloudSat mission: A new dimension to space-basedobservationsof cloud in the coming millennium, submitted to Bulletin of the American Me-
overview of the forward model, estimation theory, and setup for the retrievals. Sections4 and 5 present results from selectedcase studiesexamining the method
teorologicalSociety,2000) and lidar on PathfinderIn-
applied to marine stratocumulusand cirrus. Section
struments for Cloud and Aerosol SpaceborneObserva-' 6 discussesthe potential pitfalls of cloud property in-
tions/ClimatologieEtenduedesNuageset desAerosols terpretationwhen observed/retrieved by the different (PICASSO/CENA) [Winker and Wielicki, 1999] are remote sensorsindependently,and section 7 concludes scheduledto launch in 2003. These new observingsys- the paper with a summaryof the findings. tems embody a coming of age in our ability to provide numerical weather predication with the quality of data Instrumentation required to refine current microphysicalparameteriza- 2. Passive and Active tions and assimilate cloud information operationally. 2.1. Passive Instruments: GOES Imager and To accomplishthese tasks, our thinking must turn to- CSU SSP ward a "multisensor"mind-set where the many differRetrievals presented in this work emphasizeshortent measurements at our disposal are used in concert to wave and near-infrared radiance data provided from producebest estimates(by supplementingthe weakness GOES 10. Operated by NOAA National Environmenof one observingsystemwith the strengthsof another) of the desired environmental parameters. tal Satellite Data and Information Service(NESDIS) Motivation for spaceborne multisensor cloud remote and situatedon the equatorat (0øN, 135øW),GOES 10 sensingapplications stems from many recent studies. is the third satellite in a seriesof new generation geo-
Uttal et al. [1995]employtwo millimetricradarsand a
stationaryplatforms[seeMenzel and Putdom,1994],
10.6-/•m lidar in developinga cloud boundary statisti-
achievingoperational status in 1997. The fundamen-
cal database.Baum et al. [1995]presenta multispec- tal advantageof the geostationaryorbit is its ability tral/multiresolutionmethodologyfor analysisof mul- to provide continuoustemporal coverageof cloud evolution. Among the instruments aboard GOES is a advancedvery high resolutionradiometer(AVHRR), li- scanningradiometer capable of imaging the reflected dar, and radar. Intrieri et al. [1993]retrievecirruspar- and emitted radiation from the Earth's surface and ticle sizesusing a dual lidar/radar technique. $assen atmosphere. Patterned after the NOAA AVHRR inet al. [1989]usethis combinationto examineboth the strument, the next generationGOES imager is a filter mesoscaleand microscale structure of cirrus, while ac- radiometer (its responseis dependenton the amount tive and passiveinstrumentation are combinedby Spin- of radiation receivedwithin the filter bandwidth) emhirne and Hart [1990]and $pinhirneet al. [1996]in ploying a 12.25in.-diameteraperture and dichromatic addressingthis problem. Heymsfieldet al. [1991]use beam splitters that divide tile sceneradiation into five radiometer and lidar data in their analysisof convection spectralpassbands (hereinafter,"channels")with bandassociated with severe storms in the Midwest United widths situatedat 0.52-0.74 (channel1, visible), 3.79States, and Minnis et al. [1990]combinelidar with 4.04 (channel2, shortwaveIR), 6.47-7.06(channel3, satellite imager data to study cirrus properties. Active water vapor), 10.2-11.2 (channel4, longaveIR), and and in situ measurementsare combinedby $pinhirne et 11.6-12.5(channel5, water vapor window)/•m across al. [1989]in their analysisof cloudtop liquid water in the optical spectrum. marine stratocumulus. These studies represent a very GOES images the Earth using a dual-axis system small subset of multisensor observing systemsapplied which performseast-westhorizontal scansof the detecto cloud research. tor array followedby north-south incrementsbetween This paper examinesthe information content of satel- scan lines. The systemis capableof completinga fulllite cloudopticalpropertyretrievals(opticaldepth and Earth disk imagein roughly25 min. Smallerarea scans effectiveradius) using airborne active and in situ an- (e.g.,of the continentalUnitedStates(CONUS)) which tilayered cloud sceneswhich includes information from
MILLER
ET AL.: GOES CLOUD
RETRIEVALS
17,983
require less time are included in the GOES schedule, and 3-min "rapid scan" mode is operated during priority interrupt periodsfocusingon a particular mesoscale event. Spatial resolution of imager data used in this study are 0.57 x I km for visible imagery and 2.3 x
andthe ARM/UAV CloudDetectionLidar (CDL). The ACR is a W band(94 GHz, or roughly3 ram)radarthat
well below
ing volumeis proportionalto the sixth momentof the particle size distribution, and so the largest particles
is sensitiveto clouddroplet sizehydrometeorswhich be-
haveas Rayleighscatterers(i.e., smallsizeparameters)
and henceis able to providea profile of cloudbackscat4 km for IR channels2, 4, and 5 (4 x 8 km for the ter in addition to definingcloud top/base boundaries water vapor channel 3). Cloud heterogeneityexists in multiple layer profiles. The return from a scatterthese scales and must
be considered
when
conducting retrievals using these data. Together with
GOES 8 (0øN, 75øW), a coveragebetween80øN-80øS of the volume tend to dominate the backscatter return. and 195ø-20ø W (at the equator) is available,with a Owing to systemsensitivity (minimum detectablesigroutine scanning schedule providing two views of the nal (MDS) of-38 dbZ definedat a range of 5-kin and Continental/Pacific(CONUS, PACUS) and Northern 150-m resolution(G. Sadowy,JPL, personalcommuniHemisphericscan every half hour. The current GOES cation, 2000)), volumescomprisedsolelyof very small seriesare alsoequippedto operatein rapid scan (four droplets or ice crystals and at lower water contentsmay CONUS sectorsat 7.5-rainintervalsin a half hour) and be missedaltogether by the ACR. The instrument has super rapid scan (ten 1-min scansof 1000 X 1000 km flown on a number of research aircraft including the sectors)modesfor the purposeof monitoring rapidly DC-8 and Twin Otter in support of various cloud redevelopingweather phenomena. mote sensingcampaigns(e.g., ARM and the Cloud ExThe ColoradoState University(CSU) ScanningSpec- perimentseries(CLEX) [e.g.,Fleishaueret al., 2000]). tral Polarimeter(SSP) utilizesa rotating optical bandThe Nd:YLF laser diode-pumped CDL, operating at passfilter to measurepolarized/unpolarizedradiances 1.053 /•m, is designedfor a maximum unambiguous and fluxesovera spectralregionspanning[0.4, 1.1/•m] range of 20 km with selectableresolutiondown to 50 (and henceis designedfor daytimeoperations).A de- m (100 m typically selected). It is very sensitiveto tailed description of the SSP, its radiometric calibra- backscatterfrom the smallercloudparticles(which oftion (performed both at the Sandia National Labora- ten play a substantial role in cloud reflectivity of suntory (SNL) and the Los Alamos National Laboratory light) and can often detect the portion of cloudmissed (LANL) Optical and Infrared Laboratory),and sample owingto MDS constraintson the ACR. This enhanced resultsare presentedby Stephenset al. [2000a]. The detection is afforded at the expense of very high scathalf-powerbandwidthsof the channelsare narrow (of tering extinctionof the beam within the cloudmedium, the orderof 20 nm), and the accuracyof the calibration leadingultimately to completeattenuationof the signal. of radiancesis approximately3%. In normal operation, The CDL therefore has difficulty penetrating more turlight passesthrough the opticshead and is focusedonto bid (e.g.,visibleopticaldepths> 3.0) mediaand cannot a rotating circular variable filter and then upon a six- provide a full profile of cloud or multiple cloud layers. detectorassembly(flux, unpolarizedradiance,and four Operating together, cloud radar and lidar have the popolarized radiances:parallel, perpendicular,right-hand tential to complementeachother and thereby provide a circular, and left-hand circular). For this work, only greaterbreadthof informationabout the cloudprofile. the unpolarizedradiancefield (detector 3, with measured half-angle field of view of 20 mrad) was used. 3. Cloud Property Retrieval Overview
This corresponds roughlyto a 0.02-kin2 spot upon a
cloud viewed at nadir from a range of 4 kin. Keeping the caveat of spatial resolutiondifferencesin mind, it has been included here as a first-order independent source of validation
for the GOES
10 retrievals
of cloud
The general principlesof inversiontheory in remote
sensingare presentedby Twomey[1977]and discussed in terms of cloudmicrophysicsretrievalsin particular by
TwomeyandSeton[1980].Applicationsto the retrieval
optical depth. Its light-weight and compact design allows for its use on a variety of airborne platforms such as the General Atomics Altus II Unmanned Aerospace Vehicle(UAV), flown on variousoccasionsin support of
of cloud optical properties from multispectral imagery
the ARM program[e.g.,Crowleyand Vitko,1994].For
of the current work are founded on this rich body of
retrievals done with the SSP in the current study, the
literature• and the specificmethod employedis outlined
instrument
by Miller et al. [2000].Herewe providethe readerwith
was oriented
in the nadir to measure cloud
includestudiesby Arking and Childs [1985], Twomey and Cocks[1989],Nakajimaet al. [1991],and Minnis et al. [1992a,1992b],amongmany others. Retrievals
a brief architectural
reflectance.
this latter
2.2. Active Instruments: UMass/JPL and
ARM
overview and defer the details to
reference.
ACR
CDL
3.1.
Cloud Optical Properties
The active instruments consideredin this study were
The retrieved cloud optical properties of interest to
the Airborne Cloud Radar (developedand operatedby the University of Massachusetts(UMass) and JPL),
this study were the optical depth and effectiveparticle radius. The optical depth is defined as
17,984
MILLER
r-
ET AL.: GOES CLOUD
RETRIEVALS
that maps parameter space to measurement space, and
aext(z)dz, •zZt
(1) an errorterm eycontainingthe uncertaintiesassociated
b
where zb,zt define cloud base and top boundaries and
with both the forward model assumptionsand measurement noise. It is worth noting that this method is com-
O'ex t is the height-dependentextinctioncoefficient(scat- pletely generalin allowing for any number of measuretering plus absorption). FollowingHansen and Travis ments (y) from variousplatformsto be included,each [1974],the effectiveradiusis givenas possiblywith its own forward model F. Likewise, any number of parameters may comprisethe state vector x, with successof retrieval hinging ultimately on the degree of independentinformation provided in y. For the
-
(2)
wheren(r) is the particle sizedistributionand r is the radius of a particle in that distribution. For marine stratocumulus particle size distributions without a drizzle mode, r• typically varies between 5 and 15 •m depending on position and the diurnal state of cloud evolution
[e.g., Greenwaldand Christopher,1999]. The physi-
retrievals of this work, y contains radiancesfrom either GOES 10 or the SSP, and the cloud optical properties
(r and re) are containedin x. Assuming Gaussian error statistics, the solution to the maximum likelihood estimator is equivalent to the minimization of a scalar cost function expressedas
q• = [.•-- Xa]TS•-I[• -- Xa]q-
cal basis for retrieval of r and r• using reflectedsolar radiation
in the visible
and shortwave
infrared
"win-
[y-.F(•)]Ts•-I[y-F(•)].
(4)
dow" channels(e.g., • m 0.6 and 3.9 •m, respectively) the combined measurement (e.g.,inhas been addressedextensivelyin the literature (e.g., HereSy represents strument noise) and forward model (e.g., truncation of Twomeyand Cocks[19891,Nakajima et al. [1991],and polynomial expansion, scattering phase function, plane Kleepsies[1995]to name a few). The GOES 10 retrievals presentedin this study were based on this solar reflection 3.2.
method.
Forward
Model
parallel)uncertainties that defineeyin (3), xa is a vec-
tor containingany prior information about the desired state vector(e.g., as providedfrom climatologyor independentmeasurement),and $a containsthe uncertain-
ties attached to x•. For the retrieval of cloud optical properties, these prior constraints are necessarilyvery model in theseretrievals is describedelsewhereby Miller loose and are only enforced when the forward model et al. [2000]and will not be repeatedin detail here. demostrates little sensitivity to x. Estimates for the This discrete ordinate doubling and adding code is remaining forward model uncertaintiesare provided by based on the interaction principle and the plane-parallel Miller et al. [2000]. approximation. It was designed to be applicable to Retrieving the optimal estimate of x is accomplished the simulation of passiveradiometers operating across usinga Newtonian method outlined directly by Rodgers the visible and infrared portion of the spectrum by [1976]. Pioneeringwork in this area is developedby including instrument filter responsefunction weightThe
radiative
transfer
model
used as the
forward
Turchinet al. [1971]and WestwaterandStrand[1968].
ing. Required inputs to the model includeinformation With the further assumptionof moderately linear beabout the atmosphericprofile (e.g., temperature,preshaviorof F(x) in the neighborhoodof the currentguess sure/height,gaseousabundanciesfor applicationto a xi, the Newtonian method provides an iterative solution
correlatedK absorptionfollowingKratz [1995]),clouds (e.g., level(s),single-scatterand extinctionproperties), surfaceproperties(e.g., temperature,albedo), and solar/observergeometry. Its correctcomputationof reflectance
and transmittance
in turbid
media
of the form
•i+1-- •i -- (S•-1q-KiT•-IKi)-1 X [KiTS•-I(y - El)q-S•-•(Xa-- •i)],(5)
has been
verifiedagainstDISORT [e.g., Stamneset al., 1988], where K describesthe sensitivity of the forward model tabular resultsfrom van de Hulst [1980],and Monte to changesin the retrieved parameters: Carlo stochastic
3.3.
Estimation
radiative
transfer
simulations.
OF
K- 05'
Approach
(6)
of (5) is usuallyachievedwithin four iterThe quantity a remote sensingdevice measuresis a Convergence ations, with criteria, error covarianceassignment,and function of the physical medium and can be described additional diagnostic variables associatedwith the remathematically in the following way: trieval
y = F(x, Here the set of measurementsy is expressedin terms of the desired retrieval state vector x, all system parameters b that y is sensitive to but are not currently
beingretrieved,a forwardmodelof the systemF(x, b)
outlined
in detail
for the current
cloud retrieval
applicationby Miller et al. [2000]. 3.4.
Retrieval
Formulation
The case studies to follow include observations from active sensors flown aboard research aircraft. Cloud
MILLER
ET AL.: GOES CLOUD
RETRIEVALS
17,985
heights(tops and bases)for this study were provided cusedon GOES 10 observationswhich provided continfromthe activesensor(e.g.,lidar or radar) profilesasan uouscoveragethrough the 2 to 3 hour flight durations. additional constraint to the retrievals. GOES 10 data The marine stratocumulusobservedduring this exwere collectedfrom archivescollectedby the Coopera-
periment provided an opportunity to examine GOES
tive Institute for Researchof the Atmosphere(CIRA) retrievals in both drizzle and drizzle-free conditions. Earth Station and were extracted along the aircraft These cloudshave been examined by severalresearchers, flight tracks at 15- to 30-min temporal resolution(de- includingAlbrechtet al. [1995],Frischet al. [1995],and pendingon availability). To assurethat observations Yuteret al. [2000].Herea drizzle/no-drizzle reflectivity correspondedto the closesttime match with the satel- thresholdof-15 dbZ has been adopted following Frisch lite imagery(therebyminimizingerrorsincurredowing et al. [1995](basedon observations of stratocumulus to intraimagecloud advectionor evolution),the time- cloudsin the Azores). It is important to note that owstamped aircraft observationswere compared against ing to the high sensitivity of radar backscatterto parthe GOES image times, and pixel data fi'o•nthe closest ticlediameterD (i.e., Z crD 6) onlya fewdrizzle-sized
set were extracted accordingly. Rawinsondesoundings droplets in the radar volume will dominate the return.
wereusedto definethe temperature/moistureprofiles, Thus drizzle need not occur at the surface for the cloud and solar/satellitegeometrywere computedfor each to be considered as a "drizzle case" in the examples to point alongthe flight track. Combiningthis information follow. Only in those instanceswhere the refiectivities with cloud heights derived from active sensorsallowed exceed-15 dbZ and display these returns well beneath for the undesiredthermal component of the GOES 10 3.9-/•m measurementsto be simulated in the forward model instead of being removedby other methods. Using a multiprocessorcluster, a gridded lookup table of
the local cloud base can we postulate that drizzle is
reachingthe surface(even here, surfaceclutter in the radar return masks its observation within I to 2 range
gatesfrom the surface). forward model simulations(the most computationally Presented here are results from two CAVEX flights expensiveand time-consumingcomponent of these re- conductedon adjacentdays (July 2 and 3, 1999). In trievals)wascreatedfor eachcasefollowingthe example both cases the aircraft were forced to conduct their obof Nakajimaand Nakajima [1995].Resultsand accom- servations over waters considerably south of the base panying diagnosticinformation as describedabove were airport in Monterey Bay owingto the lack of cloud cover written in files at the conclusionof each retrieval point. in the immediate vicinity. The flight track, GOES 10
4. CAVEX
Monterey
Stratocumulus 4.1.
Case
Marine
Retrieval
Overview
The MontereyCoastalStratusExperiment(MCSE), funded by the Office of Naval Research(ONR), involved an extensive study of persistent marine stratus cloudsoff the California coastlineduring June-July 1999. An enhancementto this experiment, supported
by the Departmentof Energy (DOE) ARM/UAV program, was the CloudSat Antecedent Validation Exper-
iment (CAVEX). It includedthe DOE Twin Otter, a manned aircraft equipped with the ACR oriented in the nadir, and the Center for InterdisciplinaryRemotely Pi-
lotedAircraft Studies(CIRPAS) Twin Otter, a manned aircraft whichcarrieda suiteof in situ particle sizinginstrumentsincludingthe Forward ScatteringSpectrometer Probe(FSSP) [e.g.,Knollenberg, 1976].The instrument is sensitive to cloud particles having diameters betweenroughly 2 and 30/•m and therefore cannot detect the presence of a drizzle mode. The two aircraft wereorganizedto fly in formation over selectedportions of their legs such that colocatedradar and in situ observations would be collected. Relationships between the radar reflectivity crosssectionsand retrieved op-
tical properties/in-itu observations matchedto these data are the subject of a separate paper currently in preparation. While some of the CAVEX flights were coordinated with AVHRR overpasses,this researchfo-
channel I observations,and radar profile for the July 2 caseare shownin Plate 1, and those for July 3 are shown in Plate 2. While the visible imagery might suggestsimilar cloudcoverageon both days,the cloudradar reveals that a significant drizzle mode was present in the ob-
servedcloudson July 2 (notingrefiectivitiesreachingup -5 dbZ and extendingto the oceansurface)compared with July 3, when relatively no drizzle was inferred either by radar or observedby the CIRPAS in situ probes. 4.2.
Results
Retrieved •- and re for July 2 and 3 are given in Figures 1 and 2, respectively. Uncertainties in the retrieved
parametersare shownas error bars (1 sigma)about the points and are of the order of 10% and 15% for r and re, respectively. These values are due largely to forward model uncertainty accountedfor by the retrieval approachoutlined above. Additional diagnostics(not shown)indicatedthat both retrievalsrelied very little on the a priori guesses(i.e., the forward model possessedhigh sensitivityto the satellitemeasurements). The larger clouddropletsand drizzle phasepresentduring the July 2 flights produceda significantincreasein GOES 10 retrievedre (valuesexceeding18/•m) as compared with July 3, where re more typical of those expectednear cloudtop (e.g., of the order of 10-12/•m) were found. Satellite pixels containing broken cloudiness(e.g., over 1848-1906 UTC of the July 2 leg and 1636-1642UTC of the July 3 leg) can also result in anomalouslyhigher valuesof retrievedre (and low val-
17,986
MILLER
ET AL.- GOES CLOUD RETRIEVALS
for a givenremotelyretrieved (•-,re) pair. The adjusted data are indicated by asterisksin Figures 3 and 4. While
OES-10 Ch 1 990702 1800 124 W
122 W
the corrected effective radii approach values closer to the in situ measurements,significant disparities remain ß
'" '
particularly for the drizzle cases. No adjustment for horizontal inhomogeneityof the cloud field was made in comparing the in situ data with GOES retrievals Partly cloudy GOES pixels lead to overestimatesof effectiveradiusand underestimatesof cloudoptical depth
36N
as mentioned
,34•
4.3. 2.0
Monterey 07/02/99
above.
Discussion
The differences observed between in situ and remote
sensing estimates of re warrant some discussion. We note first that our forward model accounts for neither '
-30.0-25.0-20.0-15.0--10.0-5.0
0.0
ACRReflectivity (dBZo) I
1.0
.
I
I
,
I
,I
• •l,,l'•l i• '
0.5
a drizzle phasenor a spectrumof growingparticles beyondthe specifiedunimodal (modifiedgammawith dispersionfactorp = 2 [e.g.,Stephens, 1978])dropletsize distribution. The model therefore attempts to account for instancesof observedlow 3.9-/•m cloud refiectivities (characteristicof largerdropletsleadingto enhancedsolar absorption)by shifting the assumeddroplet population (via an increasein re) to larger values.It is here
0.0 i ....
17.0
I'•
17.5
' ' ' I '"""'
18.0
' I ß ß ' ', t ....
18.5
Ti.
19.0
GOES-10 Ch.1 990703 16(X)Z
19.5
20.0
124W
t_
(UTC)
oce•
su•f•ce h•e
120W
M•ntemy Bay
P]a•e 1. (top) O0•S 10 imagewi•h fiigh• •Ac• superimposed•d (bottom) ACR b•c•sc•e• p•ofi]e CAV•X July 2, 1999. R•d•
122W
36 N
sidelobe•e•u•s f•om •he
bee• removed.
F!t•h• Track
ues of •-) owing to the dark ocean surfaceand a decreasedpixel-averagedreflectivity. Comparisons of GOES 10 retrieved effective radii
,34. N •
2.0
most terrestrial clouds,a head-to-headcomparsionbetweenGOES 10 (whichdetects,to greateror lesserextent dependingon •-, a cloudtop re) and in situ data at varying positionswithin the cloud is expectedto differ significantly. To accountin part for the vertical variability of cloud droplet sizes,-retrievedvaluesfrom
GOES 10 wereadjustedto the depthbelowcloudtop (as definedby the ACR and the CIRPAS altitude during colocatedportionsof the CAVEX flights)usingthe procedure outlinedby NakajimaandKing [1990](and
adaptedto 3.7 pm by NakajimaandNakajima[1995]). The methodassumesthat cloudliquid water and re are linearlyincreasingfunctionsof heightand adoptsthe vertically inhomogeneousvertical stratification model of Albrechtet al. [1988](corresponding to resultsob-
I
.
.
,
.
!
!
,
Monterey07/05/99
with in situ observationsmade by the FSSP for driz-
zle and drizzle-freeconditionsare shownin Figures3 and4. Becauseverticalinhomogeneity is significantfor
•
1.5 -30.0-25.0-20.0-15.0-10.0-5.0
0.0
ACRReflectivity(dBZo) 1.0
ii ,
!
0.5
0.0 - "
15.5
'
'
'
I
16.0
....
I
16.5
½ ß '
ß I
17.0
....
i
'
17.5
Time (UTC)
Plate 2. (top) GOES 10 imagewith flight track sutainedoff the coastof southernCaliforniaduringthe perimposedand (bottom) ACR backscatterprofile for First ISCCP RegionalExperiment(FIRE)) to adjust CAVEX July 3, 1999. Radar sidelobereturns from the the retrievedvalueto an arbitrary levelwithin the cloud ocean surface have been removed.
MILLER
ET AL-
GOES CLOUD RETRIEVALS 50
CAVEX 6/24/9; Drizzle'
40
20
17,987
40
GOES 10Retrievals + Uncertainty GOES 10Adjusted toCIRPAS Altitude
30
CIRPAS FSSP InSitu
'
20
41] 2
10
30
._2 0.8
o
20
.:::•,,•,• 'i'iii•i•.Pp:•.•: •i,,:.•.•" '•'•!1111 ":•' ':'• .... i,
.•:
•0
_
• •
ACR-ob ..... d cloud bound
0.6
._• ß
F_•a---• CIRPAS Altitude ,.•
..""!' "•;i:': % •'!!.
0.4
..:• :••, :•:
C2
0•l
0.0
17.0
17.5
18.0
18.5
19.0
19.5
20.0
18.0
17.b
Time (UTC)
•igure
18.5
19.0
Time (UTC)
1. Retrieved optical depth and effectiveparticle
3O
radii for GAVBX July 2, 1999 (dri•le conditions).
CAVEX7/2/99 25
Drizzle
..•:..•.,-GOES 10 Retrievals + Uncertainty
GOES 10 Adjusted to CIRPASAltitude
that the poor initial assumption for particle size dis-
• 2o
tribution becomesan importantfactor in the retrieved re. In simulatingthe effectsof large dropletson cloud
•, • 15
absorption, Wiscombe et al. [1984]findthat boththe
•
75
secondand third momentsof the drop distribution must be known in order to definethe scatteringpropertiesof clouds. Knowledge of such moments is not included in
q0
•
this analysis,and this leadsto unrealisticretrievalsof re
1•
in connection with an incorrect assumption of the size
•. •.•
distribution. Whilenolonger representing a physical• •.o 08
resultin the contextof the assumedunimodalparticle distribution, the larger valuesof retrieved r• may prove
usetiffin somecasesas a flagfor identifyingthe presence of a significant drizzle mode in the true distribution.
f• o.•
._
•- o 4
o.• GO 18 65
18 70
18 75
18.80
18.85
18.90
18.9.5
Time (UIC)
Figure 3. Comparison of GOES 10 retrieved effective radius and in situ FSSP
data for two drizzle
cases from
CAVEX. (top) Time seriescomparisonwhen CIRPAS and Twin Otter were colocated.(bottom) CIRPAS altitude with respect to ACR-observed cloud boundaries.
The findingssupport previouswork by Rosenreidand
Gutman [1994], who in their AVHRR retrievalsfind =
30
'
._0
020,..
that optically thick cloudswith cloud top effective radii greater than about 14/•m correspondto precipitating regions. From in situ observationsof stratocumulus during the Atlantic Stratocumulus Transition Experi-
ment (ASTEX), Gerber [1996]found r, exceeding16 /•m near cloud top (defining this as a threshold for strong coalescenceand a rapid increase in the drizzle mode), alsosupportingthe current CAVEX results. 15.5
16.0
16.5
Time (UTC)
17.0
17.5
AVHRR retrievals carried out by Nakajima and Naka-
jima [1995]for drizzleand drizzle-freemarinestratocuFigure 2. Retrieved optical depth and effectiveparticle mulus during ASTEX display a positive shift (of the radii for CAVEX July 3, 1999 (drizzle-freeconditions). order of 8 pm, similar to the current results) in re-
17,988
MILLER i
i
CAVEX6/19/99 30
ET AL.: GOES CLOUD
i
i
of the diurnal cyclein marine stratacumulus re [e.g., Minnis et al., 1992b;Greenwald and Christopher, 1999]
No Drizzle
:'::.:...•:.•..4•::. GOES 10 Retrievals + Uncertainty
rendersdifficult a completesupport or refutation of either set of findings. Perhapsmost importantly, these
• •-•_GOES 10 Adjusted to OlRPASAltitude •
•
ClRPAS
FSSP
RETRIEVALS
In Situ
contradictions
2o
underscore
the need for continued
inves-
tigation of this radiatively significantclouds. ._
•
lO
5.
Cirrus
Retrieval:
ARM-UAV
SFS
Kauai .-.
•-•-•
0.8
• o--
ACR-observed
0.6
c•
I
5.1.
CIRPAS Altitude
E
0.4
0.2 0.0
17.5
18.0
18.5
Time (UTC)
c•wx •/•/'•• •izz,• '
altitudes near 3 km. The main objective of the April 30 flight was to collectcirrus data. The UAV departed
•
from the westerncoastof Kauai at 1915 UTC (0915 local military time) and landedat 0236 UTC, and the Twin Otter takeoff and landing times were 2050 UTC
i':::•-..!:::.'...•:c:• GOES 10Retrievals + Uncertainty • • • GOES 10Adjusted toCIRPAS Altitude
E 15 ._•
10
-•
CIRPAS FSSP InSitu
•
' : q"""•:':'i'•'""•:i-•,, •..... •:'•"'-d
and 0207 UTC, respectively. The retrievals presented in this study used observations made between 2130
._>
"• '-
• • •
and 0030 (May 1) UTC (or 1130-1430local military time). The two aircraft performedseveralcoincident legs(betweenroughly(22.1øN,160.8øW)and (22.8øN, 159.9øW))overthis period. The Twin Otter's altitude and headingwerechosento match as closelyas possible
5
, .-. 0.8 E
•
during April and early May 1999, consistedof two aircraft in formation flight carrying an array of active and passiveremote sensingdevices.On April 30, the Altus II UAV flew the SSP (orientedat nadir) at high altitude (near 16 kin) abovetropicalcirrus,whilethe DOE
Twin Otter carriedthe ACR (orientedin the zenith) at
20
o
Overview
TheARM-UAVspring flightseries (SFS),conducted
cloud
17.0
Case
•
• ,
• I
,
,
,
,
I
,
,
,
CIRPAS Altitude
the track of the UAV
o.6
such that the same clouds were
sampled simultaneously. For comparisonsbetweenthe UAV and the ACR, only
c• 0.4
:[: 0.2
pointswithin -4-2min and -4-4km (with respectto the ground track) were consideredas being colacaredobTime (UTC) servations(i.e., all realignmentmaneuversand en route Figure 4. Same as Figure 3, but for two drizzle-free portionsof the flight legswere discardedfrom theserecasesobservedduring CAVEX. trievals). Thesethresholdswereconsideredas sufficient in light of the upper troposphericwind speedsobserved from rawinsondes. Included in the UAV payload was trieved re for the drizzle case. Although cloud radar the CDL, alsooriented in the nadir, suchthat the cirrus was not available to that study, in situ probes were of interest were illuminated both above and below by able to confirmthe presence/absence of drizzle along active instruments. The CDL data were used to identhe flight tracks. tify cloud top altitude and the ACR data were used to Despitecorroborationwith somestudies,otherprevi- define the cloud base with accuracyof the order of 0.1 ousresearchwouldappearto refutethis detectability,at km. leastin a generalsense.Forexample,Hah et al. [1995] Imagerdata from GOES 10 (channelsI and 2) were (usingproductsderivedfrom the InternationalSatellite obtained from CIRA archives for this case, and pixel CloudClimatologyProject(ISCCP)) and Greenwald et data were extracted along the Altus flight track. The 0.0
1 6.30
1 6.35
1 6.40
16.45
1 6.50
al. [1999](usingGOES9 imagerdata) do not observe data were available at 15- to 30-min intervals througha markedchangein satellite-retrieved re for drizzling out the duration of the Altus/Twin Otter formation versuscloudsdevoidof detectabledrizzle. Larger re flights. The GOES 10 view of this case showingthe were inferred by passivemicrowaveremote sensorssensitive to the liquid water path. This indicatesthat the
flight track and correspondingACR observationsis given in Plate 3. The cirrus observedat the onsetof the experlarger dropsresidedin the lowerregionsof the cloud iment occurred between 11.0 and 13.5 km and were opsystems,wherethe near-infraredsatellite radiancessuch tically thin. Over the courseof the day, the layer thickas GOES 3.9 •m are insensitive.The additionalfactor enedgeometricallyto a rangebetween7.0 and 14.0 km.
MILLER
ET AL-
GOES CLOUD
AlthoughLagrangiansampleswere not made, it could be inferredfrom severaltrack legsthat the mesoscale cloudsystemevolvedfrom a singlelayer near 12.5 to 14 km to a two-layerstructure(with the lowercloudlayer formingbetween7 and 9 km) and finally into a thick, single-layeredcloud that closedthe gap and spanned
RETRIEVALS
17,989
ble channel. For the GOES and SSP physical retrievals of cirrus, a hollow column ice crystal phase function
(randomlyoriented,with associatedasymmetryparameter and single-scatteralbedosspecifiedas a function of
wavelengthand effectiveradius) was assumed.The as-
sumption was basedon previous observationsof pristine hollowcolumncrystalsoverthe temperature range (-45ø 1 time seriesrevealed that the cirrus layer, which was to -50ø C), correspondingto the cirruscloudtops of the asseciatedsynopticallywith an approachingbaroclinic current study (A. Heymsfield,personalcommunication, wave,advectedinto the regionfrom the west/southwest 2001). Unfortunately, no in situ data were available in and deepened(as inferredin part by coolercloudtops the Kauai experiment to verify this assumption. in the infraredchannels)overtime. Identifiedfrom rawIn their analysis of cirrus observed during FIRE, insondeobservationswas a deep, dry layer below 7 km, Wielickiet al. [1990]find largeerrorsbetweenobserved coincidingclosely with the ACR-observed base of the and theoreticalreflectanceswhen scatteringphasefunccirrus cloud. tions for spherical particles are assumed. The phase over 7 to 14 km in the vertical.
A GOES
10 channel
Available in addition to rawinsonde data were short-
range forecastsproduced by the European Centre for Medium-RangeWeather Forecasts(ECMWF). These data supportedthe forwardmodelrequirementsby providing temperature and moistureprofilescloserto where the actual observationswere being recorded.Clear-sky GOES 10 imagery was used as a forward model Lambertian estimate of the ocean surface albedo, which was observedto vary between 3 and 6% for the visiGO•S 10 Ch. 1 gg04..,'O2•
Z .'•/.--UAV S.rS
function
data
used for these cirrus retrievals
are based
on Monte Carlo ray-tracing simulations for idealized
mediaby Yangand Liou [1998].They wererepresented numerically in the forward radiative transfer model by a Legendre polynomial expansion. While their use reduces the forward model uncertainty associated with the scattering phase function when compared with a double Henyey-Greenstein approximation, uncertainty still exists owing to the many variants of possibleice crystal habits and distributions known to occur in cirrus. However, the cirrus examined by the current study were of sufficient optical thickness that multiple scattering plays a significant role in smoothing the finescale details particular to the phase function of any pristine habit. Retrieved optical depths using these ice phase functions tend generally to be smaller than their best fit double Henyey-Greensteincounterparts, owing to their relatively strong backscatter peaks at visible wavelengthsthat a.llowfor a smaller r to produce the same observed
5.2. 15
solar backscatter.
Results
The GOES 10 retrievals of r and r• for the colocated
-30.0
-20,0
- 10.0
Reflectivity(dBZ,)
0.0
portions of the Kauai April 30, 1999, flight legs are shown in Figure 5. Visible optical depths range from 0 to 4.5, and effective radii are between 15 and 80 pm, with 1-sigmauncertainty bars included. The larger uncertainties in r• compared with the CAVEX retrievals are conservative,reflectingthe difficultyin both approximating and constrainingan "effectiveradius" for an equivalent ice spherewhen dealing with highly complex crystal geometriesand at a single viewing angle. Forward model sensitivity to suchlarge particle sizesis also small, leading to larger uncertainties in the retrieved r•.
New instruments such as the Polarization and Di-
rectionality of the Earth Reflectances(POLDER) and the Along-TrackScanningRadiometer(ATSR 2) [e.g., 21 22 23 24 Baran et al., 1999a]are capableof observingscenes from Time (UTC) more than one viewing direction (each corresponding Plate 3. (top) GOES 10 imagewith flight track su- to a differentscatteringangle) typically over the range perimposedand (bottom) ACR backscatterprofile for from 60ø to 180ø. This enablesa sampling of the phase ARM-UAV SFS April 30, 1999. function that is sufficient to rule out particular habits
MILLER
17,990
ET AL.:
GOES
8
CLOUD
RETRIEVALS
A scatterplotcomparingcorrectedGOES 10 against SSP retrievalsof visible cloud optical depth (an SSP channelcenteredon the GOES channel I responsefunc-
tion was selected)is shownin Figure 6. The comparisonswere made in terms of optical depth as opposedto
reflectance(the latter beinga more appropriatemetric for vicariouscalibrationanalysis)owingto the potential for differences in the observerzenith angles(exceeding 38ø betweenGOES and SSP) producinglarge varia-
•o8 8O
tions in reflectance for cirrus. Owing to its spectral
rangecutoffat 1.1/•m, the SSP doesnot providesuffi-
6O
cient sensitivity for the retrieval of cloud particle size. 4O
The retrieved effective radii from GOES were therefore
usedas a priori valuesto the SSP retrievalsof r. The small dependenceof visible reflectanceon particle size 0 , maintainsthe independenceof theseoptical depth com21.5 22.0 22.5 23.0 23.5 24.0 24.5 Time (UTC) parisons.The one-to-onecorrespondence is includedas a diagonal line on the scatterplot. The visible optical Figure B. ARM-UAV Kauai cirrus optical depthsasredepths retrieved by both instruments were found to be trieved by GOES 10 usingACR and CDL cloudheights for the periods where the Altus and Twin Otter were similar, with someunderestimationby GOES owingto colocated. subpixelvariability and to a lesserextent differences between the filter responsefunctions of the two instru(e.g., hexagonalplatesor columns)as beingthe dom- ments. The scatter of points is explained by recalling inant species.Baran et al. [1999b]take advantageof the •2 order of magnitudedifferencein resolutionbetween GOES 10 and the SSP. this ability to constrainthe phasefunctionof cirrus(in termsof identificationof a "dominantcrystalhabit") in 2O
their cloud retrievals.
Such near-instantaneous
multidi-
6.
Passive
Versus
Active
Cloud
rectionalimagery is not availableto GOES alone (ex- Sensitivity cept in the local casesof coincidentAVHRR overpasses, When working with different observingsystems,it asdemonstrated by Turk et al [1998]),necessitating an is important to recognizethe inherent strengthsand a priori assumption on the crystal habit and a correlimitations unique to each device. The two active inspondingincreaseduncertaintyassignment.A fore/aft configurationof two SSPinstruments(or a singleinstru- struments(cloudradar and lidar) includedin this exment mountedon a pivotingplatform) wouldsufficeto perimentwereby no meansredundant. While the millimetric radar hasthe capabilityto penetrateand provide capture this sameinformation but was not implemented during the ARM-UAV SFS observations. An advantageto multisensorcampaignsis the ability to evaluatethe retrievals againstindependentestimates. This is particularly relevant to GOES, whosevisible detector suffersa decreasedthroughput and hencechanging calibration with time owing to radiation darkening of the scanmirror coating[e.g.,Knapp and Yonder
Haar, 2000]. ChannelI was not designedfor quanti-
ARM-UAV K i 4/50/99
R= 0.96aua
ß".'ß •,.:,•1" m
' :' .i;"",-'."%-'.
$
ß
where R is the responsivity(- I at launch), A is the degradation rate (per day), and 5t is the•ime (in days)
ßß
'"'::
., m•,."•"
ß • ,, _,,,•,;.•/•,;" '
,
1
0
since launch. Responsivitiesfor GOES 10, assuming comparable degradation to that observedfor GOES 8
and 9 (insufficientdata for GOES 10 were availableat the ti•ne of this writing), wereof the orderof 0.9 for the
•.•
""•I
' .,.' - ..-',2F ....
form
(7)
:.- :
..,....:;% " ." ..½•,"
2
et al. [1998]to estimatea responsivity correctionof the
R = e-ASt,
ß
. .;;,,:-'."/i
Instead, a vicarious calibration based on the
brightnessvariation of stars has been used by Bremer
m mm•
.dl' m ß,,•
tative applications, and there exists no onboard calibration.
' '" ""ß"""""• '/•/' t
1
2
..3 GOES
4
...5
'T
Figure 6. Comparisonbetween cirrus optical depths retrieved by GOES 10 and SSP radiances. Correlation
springand early summerretrieval casestudiespresented coefficient(R) and root-mean-square difference(RMS) in this work.
are indicated.
MILLER
ET AL.:
GOES
CLOUD
10
17,991
encesexist betweenthe two observingsystems.At the
ARM-UAV Kauai 4/50/99 15 ß
RETRIEVALS
..
. : ß,I ..... . ..... .ß..... • . ••.•': ......... . •...... ••:•..... ½. ..:..: ß2-.: ..... •,; •...:: •:." , .... ß:.•,. ß ..•:•:.-. '•...... -.. ,.:.-•:-.•t•j:. '-:.•.......•! ':. . ..•:.•-.. •........... •'•-•:::• .......,, :..:•..,,..F-.:t? .:•. •.:-.. :•..,..•. •.,.,:.•;.: .•.-.•..,%................. "'•':.. ;-•'-.',.•.: :"':'• ß -:•..::':' .".•'..•' '• ...... ": •-'•:'• "..'-%•:. ':.:--?'•.. •a.:•:.--t ..•i:':-!• :.•-:.:•..?.•
.. •."?• .•ß.•.....:,..•...•...::• . l•,}.•... .. ". .....ß ....... :.-, '"•-•...: ......
levelof formation(nucleation,nearcloudtop for cirrus) particlesare small, growinglarger (e.g., by diffusional growth within the regimesupersaturatedwith respect to ice) as they sedimentdownwardtowardcloudbase. In the frameworkof the radar/lidar system,the radar will morereadilyseethe larger(lower)particlesand the lidar will seethe smaller (upper) particles. The large
50.0 150.0 250.0 350.0 450.0 C.DL BackScatter (Arbitrary Units)
COLOCATED
-.30.0
-20.0
-10.0
0.0
Reflectivity(dBZ•) 22
25
24
Time (UTC)
Kauai observations of cirrus Figure 7. ARM-UAV cloudsby the CDL and the ACR. Note differencesin sensitivity between the two instruments. Colocation times are indicated between the two panels. From
Stephenset al. [2000b]
particlesnear cloud basetypically feature very strong forward scatteringlobes,suchthat off-beamlidar photons (diffusedby smallercrystalsand correspondingly smallerasymmetryparametersnear cloudtop) will be transportedmore readily outsideof the lidar detector field of view. The lidar signal is extinguishedcompletelyby thesemultiple scatteringprocesses and cloud base is undetected. Meanwhile, the radar transmits almost completelythrough the small formationlevel cirrus particleswith backscatteredenergybelowthe minimum detectablesignalat that range. Hencevery different mechanismsare responsiblefor the detection constraints
of the two instruments.
Shownin Figure 8 are effectiveradii retrievedby the the GOES 10 and empiricalradar methods. To compute
r•, icewatercontent(IWC, gramsper cubicmeter)was first calculatedaccordingto a relationshipdevelopedby Sassenand Liao [1996]for W bandradar:
detailed crosssectionsthrough optically thick clouds,it suffersfrmn insensitivity to clouds comprisedof small particles and aerosol. Becausethe lidar transmits at a
IWC - 0.0217Z/ø'8•,
(8)
visiblewavelength(3 ordersof magnitudesmallerthan whereZi is the radar reflectivity(mm6/ms) for ice. the cloud radar), the size parametersassociatedwith This quantityis incorporatedinto a relationshipfor cirby Platt [1997]whichcalculatesan effecsmallerparticles are large enoughto produce detectable rus proposed tive radiusfor eachgate basedon the radar-derivedice backscatter returns. The drawback to the lidar is that the signal is quickly attenuated within optically thick
water
content:
media. The combinedcloud radar/lidar systemcap-
=
tures information at both ends of the particle size distribution a.nd thereby provides an improved means to observingthe cloud profile. Figure 7 exemplifiesthis conceptfor the Kauai ARMUAV SFS April 30 cirrus case. The CDL data are shown in the upper panel, and the ACR data are provided in the lowerpanel (colocationbetweenthe Twin Otter and
Altus, as definedby the space/timecriteria discussed earlier, is shownin betweenthesepanelsas a segmented
line). In the CDL data, a sharpupperboundarynear 13 km is observed,and the backscatterfades in a striated fashionto a highly variable and at times indistinguishable base. Complete attenuation of the lidar beam in optically thicker regionsof cloud is indicated by the occasionaldisappearanceof the otherwise strong surface
140
3iWC 1-• 2jp
,
(9)
GOES-10
Radar (Upper 0.5-km mean) Rodor (Full-column meon)
120
'•_ lOO •
80
ßc
60
40
return at 0 kin. The ACR, on the other hand, detects 20 thesesame cirrus as having a sharply definedbasenear 0 .... I .... I , , , , I , , , , I , • , • I , , , , I 7 km and a similar striated fading of the signaltoward 2 .5 22.0 22.5 25.0 25.5 24.0 24.5 a ragged cloud top altitude varying between 10 and 12 Time (UTC) km. At first glance, it might appear that two entirely different cloud systemswere being observed! Figure 8. Comparisonof radar and GOES 10 retrieved The physicsbehindcirrusformation/dissipationpro- effectiveparticle radii, vertical mean and adjusted to vides us with a clue as to why such marked differ- GOES observations.
17,992
MILLER ET AL- GOES CLOUD RETRIEVALS
wherethe coefficients j - 3.48,k - 0.679 wereselected retrievals. The sensitivityof the GOES re retrievalsto the small cirrus crystalsin the formation zone between 11and13km (particlesdetectedby the CDL but missed in largerpart by the ACR) exemplifies the important considerationof detectability when comparingdifferent
for ice columns and p is an assumeddensity for the ice crystalsconsidered.An active sensorcolumn mean effective radius was thereby computed for comparison againstthe singleGOES 10 values. Figure 8 showsthe active-retrieved re, which include range gates that contain values exceeding 100/•m. The column mean re is significantlylarger than the GOES results. The large differences were due to contributions
to the mean radar-
derived reflectivity near cloud base. An important point to remember with passive retrievals based on measurements
of solar radiation
is that
the reflectedenergyis not a function of re averageduniformly throughout the cloud depth, but rather of some value closerto the incidentboundary(i.e., cloudtop). The depth to which solar radiation will penetrate the cloudis a function of the water/ice absorptionat the wavelength in question. The 1.65-/•m water vapor window channel, while offering sensitivity to particle size, has weaker absorption than that at 3.9/•m. As a result, reflection measured at the former wavelength is representativeof a deeper penetration within the cloud, whereas the measurement at 3.9/•m is more representative of particles closer to cloud top. Nakajima and King [1990]investigatethis problemfor the caseof marine stratocumulus observedduring the First ISCCP Regional Experiment conductedoff the southernCalifornia coastlinein July of 1987. They showthat for marine stratocumulus(where particle size increaseswith height in the cloud) the effectiveradiusretrievedfrom solar reflectionat 3.9/•m is roughly 90% of the cloud top re when r •_ 5.0. A. Benedettiet al. (Application of eigen matrix techniqueto compute global reflection, transmission and Green's function, part I, Theory, submitted to Journal of Quantitative Spectroscopy and Ra-
observing systems.
7. Summary and Conclusions This paper enlisted active and in situ observationsto examinethe utility and limitations of GOES retrievals of opticaldepth and effectiveradiusfor marinestratocumulus and cirrus, clouds that are thought to play significant roles in radiative feedback processesin climate. To better understandthesefeedbacks,an observingsys-
tem capableof providinga detaileddescriptionof physi-
cal/opticalpropertiesthroughoutthe depthof the cloud is required. The active sensorsincludedhere in a limited capacitywere demonstratedas usefulin providing informationabout cloudheight, the presence/absence of a drizzle mode in stratocumulus,and profilesof particle size and water content for cirrus. We regard these multisensorobservingsystemsas essentialto furthering our existingknowledgebaseand therebyimprovingour
ability to characterizethe effectsof cloudin numerical weather prediction models. The findingsof this studysupportthe premisethat in some cases a discrimination between drizzle and drizzlefree marine stratocumulus clouds can be made on the
basis of GOES 10 daytime retrievals of effective ra-
diususing0.65-and3.9-•umchannels(with re exceeding about -,•16•umas a proposedthresholdingvaluefor the cloudsconsidered). On the basisof two drizzle/drizzlefree casepairs (one shownhere) from CAVEX, these diative Transfer,2000) and Miller and Stephens[2000] findingswere affordedby the availabilityof colocated model the sensitivity of simple cloud media illuminated ACR measurements.Owing to the limited casestud-
from above and show a similar decreasedsensitivity to optical properties at cloud base that vary accordingto cloud optical thickness. In an attempt to compensatefor these effects, a new radar-derived effective radius comprisinga mean of the uppermost0.5 km of cloudwas computed. The adjusted
re are shownin Figure 8 as small triangles. These data fell much
closer to the GOES
10 retrievals
and were
ies examinedhere and previousstudiesin which sucha discriminationwas not found, however,we stop short of claiminga generalpassive-only capabilityfor warmrain processdetection. As is more often the case,in nature a spectrumof possibilitiesexiststo confoundthe absolute rule, and a true understandinglies buried amidst the exceptions.Compilationof additionalcasestudies
observedto detrend the increaseof active-computedre
(e.g.,in connection globaldata setsas providedby future spaceborneactive sensormissions)shouldaid in
overtime (as the clouddeepened).Remainingdiscrep-
this research.
anciesfor the optically thin clouds are explained in part by the lack of radar sensitivity to a majority of the pro-
GOES retrievals of cirrus cloud optical depth during the ARM-UAV SFS experiment were found to be consistent with those estimated by the CSU SSP after a
file (therebycomputinga meanbasedon the largerparticles that were detected while still missingthe smaller
correctionto degradationin GOES channel1 respon-
contributions) andby the empiricalnatureof theradar sivity was applied. It is important to determinesuch estimate for re. The result is supported by the work
of Platnick[2000],who,in examiningthe effectsof vertical structure on the retrieval of droplet effective ra-
dius, finds that a simple weighting function based on the maximum penetration of reflected photonsprovides a good representation of standard near-infrared band
consistency(or bias) for considerationin future comparisonsbetweentheseplatformsaswell asstand-alone experiments.While the GOES imagerdesignwas not intendedoriginallyfor quantitativeapplications,countlessresearchershave proceededalong these lines, since
onlygeostationary platformsprovidea highspatialand
MILLER
ET AL.:
GOES
temporal resolution necessaryto monitor cloud activity on the global scale. Finally, cirrus as observedby GOES 10, SSP, ACR, and CDL were compared. It was found that GOES retrievals of re were biased toward the smaller ice particlesnear cloud top which went undetectedby the ACR but were detected by the CDL. The complementarydetection capabilities of the two active sensorsillustrate the motivation
for the CloudSat
and PICASSO-CENA
CLOUD
RETRIEVALS
17,993
Baran, A. J., S. J. Brown, J. S. Foot, and D. L Mitchell, Retrievalof tropicalcirrusthermal opticaldepth, crystal size and shape using a dual view instrument: A tropical cirrus anvil case, J. Atmos. Sci., 56, 92-110, 1999a. Baran, A. J., P. D. Watts, and P. N. Francis, Testing the coherenceof cirrus microphysicaland bulk properties retrieved from dual-viewing multispectral satellite radiancemeasurements,J. Geophys.Res., 10,{,31,673-31,683, 1999b.
Barker,H. W., G. L. Stephens,and Q. Fu, The sensitivityof domain-averagedsolar fluxesto assumptionsabout cloud
missionsflying in formation. The current study speaks geometry, Q. J. R. Meteorol. Soc., 125, 2127-2152, 2000. to the vertical variation of true cloud properties and Baum, B. A., J. M. Alvarez, T. Uttal, J. Intrieri, M. Poellot, the caveats associated
with
results derived
from non-
E. Clothiaux, T. P. Ackerman,D. O. Starr, J. Titlow, and V. Tovinkere, Satellite remote sensingof multiple cloud layers, J. Atmos. Sci., 52, 4210-4230, 1995. here, horizontal heteorogeneityalso plays a significant Bremer, J. C., J. G. Baucom, H. Vu, M.P. Weinreb, and N. role in the disparity betweentrue cloud properties and Pinkine, Estimation of long-termthroughputdegradation thoseretrievedundera plane-parallelassumption[e.g., of GOES 8 & 9 visible channelsby statistical analysisof star measurements,Proc. SPIE Int. Soc. Opt. Eng., 3J39, Barker et al., 2000;Fu et al., 2000].
profiling passive-only methods. Although notaddressed
With spaceborneactive sensors,the passive/active retrievals shown here in limited capacity will become readily available on a regular basis. In addition to CloudSat tracking of PICASSO-CENA and the EOS
Aqua (and hencecontinuouscolocationwith MODIS), opportunitiesfor comparisonsagainstGOES, AVHRR, and DMSP
satellite
retrievals
will also abound.
An
important immediate outcome of the current research is one illustration of how active data may potentially be used in conjunction with passiveradiometer data to qualify the information content of passive-onlycloud property retrievals. However, it is aknowledgedthat
145-154, 1998.
Crowley, P. A., and J. Vitko, Jr., The AtmosphericRadiation MeasurementUnmannedAerospaceVehicleprogram: An overview, in Proceedingsof the Third Atmospheric Radiation Measurement (ARM) Science Team Meeting, CONF-9303112,U.S. Dep. of Energy,Washington,D.C., 1994.
Fleishauer,R. P., V. E. Larson, D. L. Reinke, and T. H. Von-
der Haar, ComplexLayeredCloudExperiment(CLEX-5): Preliminary phenomenologyof four case studies, paper presentedat BattlespaceAtmosphericand Cloud Impacts on Military Operations Conference,Fort Collins, Colo., April 25-27, 2000. Frisch, A. S., C. W. Fairall, and J. B. Snider, Measurement of stratus cloud and drizzle parameters in ASTEX with a Ka-band Doppler radar and microwave radiometer, J. Atmos. Sci., 52, 2788-2799, 1995.
much work is to be done in determining how to apply this information optima.lly. This will hold especially true when attention is turned to multiple-layeredclouds Fu, Q., M. C. Cribb, H. W. Barker, S. K. Krueger, and A. Grossman, Cloud geometry effectson atmosphericsolar and their role in modifyingthe ra.diative heating profile absorption, J. Atmos. Sci., 57, 1156-1168, 2000. of the atmosphere.
Gerber, H., Microphysics of marine stratocumulus clouds with two drizzle modes, J. Atmos. Sci., 53, 1649-1662,
Acknowledgments. Fundingfor this researchwas provided by NOAA GIMPAP grant NA67RJ0152 AMEND 25 and CG/AR grant DAAL01-98-0078. CAVEX was supported under California Institute of Technology/JPLcontract 1212032, and the Cooperative Institute for Research of the Atmosphere at Colorado State University provided GOES 10 data under DOD Center for Geosciences,Phase II grant DAAH04-94-G-0420. The authors thank Phil Parrain, Tom Greenwald, Sundar Christopher, and Teruyuki Nakajima for their assistancewith data, references,and insightful discussion,and Hafiidi Jonssonand the staff of CIRPAS for access to and assistance with
the FSSP data from
MCSE/CAVEX.
References Albrecht, B. A., D. A. Randall, and S. Nicholls, Observations of marine stratocumulusclouds during FIRE, Bull. Am. Meteorol. Soc., 69, 618-626, 1988. Albrecht, B. A., C. S. Bretherton, D. Johnson, W. H. Schubert, and A. S. Frisch, The Atlantic Stratocumulus Transition Experiment-ASTEX, Bull. Am. Meteorol. Soc., 76, 889-904, 1995. Arking, A., and J. D. Childs, Retrieval of cloud coverparameters from multispectral satellite images, J. Clim. Appl. Meteorol., 2J, 322-333, 1985.
1996.
Greenwald,T. J., and S. A. Christopher,Daytime variation of marine stratocumulusmicrophysicalpropertiesas observedfrom geostationarysatellite, Geophys.Res. Left., 26, 1723-1726, 1999.
Greenwald,T. J., S. A. Christopher,J. Chou, and J. C. Liljegren, Intercomparison of cloud liquid water path derived from the GOES 9 imager and ground-basedmicrowave radiometersfor continental stratocumulus, J. Geophys. Res., 10J, 9251-9260, 1999.
Hah, Q., W. Rossow, R. Welch, A. White, and J. Chou, Validation of satellite retrievalsof cloudmicrophysicsand liquid water path using observationsfrom FIRE, J. Atmos. Sci., 52, 4183-4195, 1995. Hansen,J. E., and L. D. Travis, Light scatteringin planetary atmospheres,SpaceSci. Rev., 16, 527-610, 1974. Heymsfield,G. M., R. Fulton, and J. D. Spinhirne, Aircraft overflightmeasurementsof Midwest severestorms: Implications on geosynchronoussatellite interpretations, Mon. Weather Rev., 119, 436-456, 1991. Intrieri, J., G. L. Stephens,W. L. Eberhard, and T. Uttal,
A methodfor determiningcirruscloudparticlesizesusing lidar and radar backscatter technique, J. Appl. Meteorol., 32, 1074--1082, 1993.
Joiner• J.• and D. P. Dee, An error analysisof radiance and suboptimal retrieval assimilation, Q. J. R. Meteorol. Soc., 126, 1495--].514, 2000.
17,994
MILLER ET AL.- GOES CLOUD RETRIEVALS
Kaufman,Y. J., The Earth ObservingSystem(EOS) AM-1 proving radiative forcing and feedbackin general circulaplatform, J. Geophys.Res., 103, 32,139-32,140,1998. tion models, in Proceedingsof the 1991 Annual ConferKing, M.D., Y. J. Kaufman, W. P. Menzel, and D. Tanr•, ence of the American MeteorologicalSociety, pp. 83-86, Remote sensingof cloud, aerosoland water vapor propAm. Meteorol. Soc., Boston, Mass., 1991. erties frotn moderate resolution imaging spectrometer Platnick, S., Vertical photon transport in cloud remote sens(MODIS), IEEE Trans. Geosci.RemoteSens.,30, 2-26, ing problems, J. Geophys.Res., 105, 22,919-22,935, 2000. 1992. Platt, C. M. R., Sizespectra,extinctionand ice/water conKleepsies, T. J., The retrieval of marine stratiform cloud
propertiesfrom multiple observationsin the 3.9-•umwindow under conditionsof varying solar illumination, J. Appl. Meteorol., 3•, 1512-1524, 1995. Knapp, K. R., and T. H. Vonder Haar, Calibration of eighth Geostationary ObservationalEnvironmentalSatellite (GOES-8) imager visible sensor,J. Atmos. Oceanic Technol., 17, 1639-1644, 2000. Knollenberg,R. G., Three newinstrumentsfor cloudphysics measurements,part I, Model descriptionand sensitivity •neasurements,Preprints, in International CloudPhysics Conference,pp. 554-561, Am. Meteorol. Soc., Boulder,
tent of frontal and capping cirrus clouds,J. Atmos. Sci., 5J, 2083-2098, 1997.
Rodgers, C. D., Retrieval of atmospherictemperature and composition from remote measurementsof thermal radiation, Rev. Geophys.,1•, 609-624, 1976. Rosenfeld, D., and G. Gutman, Retrieving microphysical propertiesnear the tops of potential rain cloudsby multispectral analysis of AVHRR data, Atmos. Res., 3•, 259283, 1994.
Sassen,K., and L. Liao, Estimation of cloud content by Wband radar, J. Appl. Meteorol., 35, 932-938, 1996. Sassen,K., D. O. Starr, and T. Uttal, Mesoscaleand miColo., 1976. croscalestructure of cirrus clouds: three case studies, J. Kratz, D. P., The correlatedK-distributiontechniqueas apAtmos. Sci., •6, 371-396, 1989. plied to the .AVHRR channels,J. Quant. Spectrosc.Ra- Schueler,C. F., and W. L. Barnes, Next-generationMODIS diat. Transfer. 53, 501-517, 1995. for polar operational environmental satellites, J. Atmos. LockheedMartin Missilesand Space,SystemSpecification Oceanic Technol., 15, 430-439, 1998. for the National Polar-Orbiting OperationalEnvironmen- Spinhirne, J. D., and W. D. Hart, Cirrus structure and ratal Satellite System (.NPOESS), LMMS-P•31903, condiative parameters from airborne lidar and spectral ratract F04701-91-C-0068,CDRL sequenceA018, 183 pp., diometer observations:The 28 October 1986 FIRE study, Bethesda, M.D.,
1995.
Mon. Weather Rev., 118, 2329-2343, 1990. Spinhirne, J. D., R. Boers, and W. D. Hart, Cloud top liquid water from lidar observationsof marine stratocumulus, J. Appl. Meteorol., 28, 81-90, 1989. 75, 757-781, 1994. Spinhirne, J. D., W. D. Hart, and D. L. Hlavka, Cirrus Miller, S. D., and G. L. Stephens,A multisensorapproach infrared parameters and shortwave reflectance relations to the retrievaland modelvalidationof globalcloudiness, from observations,J. Atmos. Sci., 53, 1438-1458, 1996. Ph.D. dissertation,442 pp., Dep. of Atmos. Sci., Colo. Stamnes, K., S.-C. Tsay, W. Wiscombe, and K. State Univ., Fort Collins, 2000. Jayaweera, Numerically stable algorithm for discreteMiller, S. D., G. L. Stephens,C. K. Drummond, A. K. Heiordinate-methodradiative transfer in multiple scattering dinger,and P. T. Partain, A multisensordiagnosticsateland emitting layered media, Appl. Opt., 27, 2502-2509,
Menzel, W. P., and J. F. YV. Purdom, IntroducingGOESI: The first of a new generationof GeostationaryOperational EnvironmentalSatellites,Bull. Am. Meteorol.Soc.,
lite cloud property retrieval scheme,J. Geophys.Res., 105, 19,955-19,971, 2000.
Minnis, P., D. F. Young, K. Sassen,J. M. Alvarez, and C. J. Grund, The 27-28 October 1986 FIRE IFO cirrus casestudy: Cirrus parameterrelationshipsderivedfrom satellite and lidar data, Mon. Weather Rev., 118, 2402-2425, 1990.
Minnis, P., P. W. Heck, and D. F. Young, Inferenceof cirrus cloud properties using satellite-observedvisible and infrared radiances,part II, Verification of theoretical cirrus radiative properties, J. Atmos. Sci., 50, 1305-1322, 1992a.
Minnis, P., P. W. Heck, D. F. Young, C. W. Fairall, and J. B. Snider,Stratocumuluscloudpropertiesderivedfrom simultaneous
satellite
and island-based
1988.
Stokes,G. M., and S. E. Schwartz, The AtmosphericRadia-
instrumentation
during FIRE, J. Appl. Meteorol., 31, 317-339, 1992b.
Nakajima,T. Y., and M.D. King, Determinationof the optical thicknessand effectiveparticle radius of cloudsfrom
reflectedsolarradiationmeasurements, part I, Theory,J. Atmos. Sci., •7, 1878-1893, 1990. Nakajima, T. Y., and T. Nakajima, Wide-area determination of cloud microphysical properties from NOAA
AVHRR measurements for FIRE and ASTEX regions,J. Atmos. Sci., 52, 4043-4059, 1995.
Nakajima, T. Y., M.D. King, and J. D. Spinhirne, Determination of the optical thicknessand effectiveparticle radius of clouds from reflected solar radiation
measure-
ments, part II, Marine stratocumulus observations,J. Atmos. Sci., •8, 728-750, 1991.
Patrinos, A. A., D. Renne, G. Stokes,and R. G. Ellingson, AtmosphericRadiation Measurement-A programfor im-
tion Measurement(ARM) program:Programmaticbackground and design of the cloud and radiation test bed, Bull. Am. Meteorol. Soc., 75, 1201--1221, 1994. Stephens,G. L., Radiation profilesin extendedwater clouds, part II, Parameterization schemes, J. Atmos. Sci., 35, 2123--2132, 1978.
Stephens,G. L., R. F. McCoy Jr., R. B. McCoy, P. Gabriel, P. T. Partain, S. D. Miller, and S. P. Love, A multipurpose ScanningSpectralPolarimeter(SSP):Instrumentdescription and sample results, J. Atmos. Oceanic Technol., 17, 616--627, 2000a.
Stephens, G. L., et al., The Department of Energy's At-
mosphericRadiation Measurement (ARM) Unmanned AerospaceVehicle (UAV) program, Bull. Am. Meteorol. Soc., 81, 2915-2937, 2000b. Turchin, V. F., V. P. Kozlov, and M. S. Malkevich, The use of mathematical
statistics
methods
in the solution
of in-
correctly posedproblems, Soy. Phys., Usp., Engl. Transl.,
•3(6), 681-703, 1971. Turk, J., J. Vivekanandan, T. Lee, P. Durkee, and K. Nielsen, Derivation and applications of near-infrared cloud reflectances from GOES-8 and GOES-9, J. Appl. Meteorol., 37, 819-831, 1998. Twomey, S., Introduction to the Mathematics of Inversion 2n Remote Sensing and Indirect Measurements, 243 pp., Dover, Mineola, N.Y., 1977. Twomey, S.: and T. Cocks, Remote sensingof cloud parameters from spectral reflectancein the near-infrared, Beitr. Phys. Atmos., 62, 172-179, 1989.
MILLER
ET AL.: GOES CLOUD
RETRIEVALS
17,995
Twomey,S., and K. J. Seton,Inferencesof grossmicrophysi- Wiscombe, W. J., R. M. •relch, and W. D. Hall, The effects cal propertiesof cloudsfrom spectral reflectancemeasureof very large drops on cloud absorption, part I, Parcel ments, J. Atrnos. Sci., 37, 1065-1069, 1980. models, J. Atraos. Sci., 41, 1336-1355, 1984. Uttal, T., J. M. Intrieri, and W. L. Eberhard, Cloud bound- Yang, P., and K.-N. Liou, Single-scatteringproperties of ary statistics during FIRE II, J. Atrnos. Sci., 52, 4276complex ice crystals in terrestrial atmosphere, Contrib. 4284, 1995.
van de Hulst, H. C., Multiple Light Scattering Tables, Formulas, and Applications, vol. 2, 739 pp., Academic, San Diego, Calif., 1980. Westwater, E. R., and O. N. Strand, Statistical intbrmation content
of radiation
measurements
used in indirect
sensing,J. Atmos. Sci., 25, 750--758, 1968. Wielicki, B. A., J. T. Suttles, A. J. Hcymsfield, R. M. Welch, J. D. Spinhirne, M.-L. C. Wu, D. O. Starr, L. Parker, and R. F. Arduini, The 27-28 October 1986 FIRE IFO cirrus case study: comparison of radiative transfer theory with observationsby satellite and aircraft, Mon. Weather Rev., 118, 2356--2376, 1990. Winker, D. M., and B. A. Wielicki, The PICASSO-CENA Mission, in SPIE Sensors,Systems, and Next Generation
Satellites V, Int. Soc. for Opt. Eng., Bellingham, Wash., 1999.
Atraos. Phys., 71, 223-248, 1998. Yuter, S. E., Y. L. Serra, and R. A. Houze, Jr., The 1997 Pan American Climate Studies Tropical Ea•stern Pacific ProcessStudy, part II, Stratocumulus region, Bull. Am. Meteorol. Soc., 8i, 483-490• 2000.
R. T. Austin and G. L. Stephens, Department of Atmospheric Science. Colorado State University, Fort Collins, CO 80523. (austin't•atmos.colostate.edu; stephens•at mos.colostate.edu) S. D. Miller, Naval Research Laboratory, Marine Meteorology Division. Satellite Meteorological Applications Section, 7 Grace Hopper Avenue, MSS2, Monterey, CA 939435502. (miller(t_•nrhnry. navy.rail)
(Received October 17, 2000: revisedMarcix 21, 2001; acceptedMarch 21, 2001.)