JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 103, NO. D24, PAGES 31,695-31,707, DECEMBER 27, 1998
Variability of water vapor in the tropical upper troposphereas
measuredby the Microwave Limb Sounderon UARS
H.L. Clark and R.S. Harwood
Department
of Meteorology,
University
of Edinburgh,
Scotland,
UnitedKingdom
P.W. Mote t
NorthwestResearchAssociates,Bellevue,Washington
W.G. Read
JetPropulsion
Laboratory,
CaliforniaInstituteof Technology,
Pasadena
Abstract.TheMicrowave
LimbSounder
(MLS), aninstrument
ontheUpperAtmosphere
ResearchSatellite(UARS), measures
watervaporin the uppertroposphere,
with best
sensitivityat the standardUARS level at 215 hPa. In this paper,we analyzethe MLS
observations
with a view to characterizing
the temporalandzonalvariations
of upper
tropospheric
watervaporbetween20øN and20øS.Time seriesof watervaporthroughout
the tropicsshowa strongannualcyclewith maximumamplitudeat 20øN and 90øE.
An intraseasonal
cycle with a periodof 30-85 daysis evidentoverthe WesternPacificat
latitudesfrom 10øNto 20øS. Thecycleis associated
witheastward
propagating
disturbances
of zonalwavenumbers1-2, suggesting
thatthisintraseasonal
cycleis relatedto theMaddenJulian oscillation.
1. Introduction
Water vapor is a significantabsorberand emitter of infrared radiationand is the most dominantgreenhousegas
[e.g., Houghton et al., 1990; Jonesand Mitchell, 1991]. The
responseof the climateto increasesin anthropogenic
greenhousegasesdependsuponthe watervaporfeedback,which
is generallybelievedto bepositive;increased
globaltemperaturesleadto an increasein watervapor,in turncontributing
to furtherwarming.Lindzen[ 1990] suggested
thatincreased
global temperaturesand consequentincreasedconvection
may lead to a drying of the uppertropospherethroughsubsidence,which could offset somewarming. Some debate
still surrounds
thisidea asmorerecentpaperssuggest[e.g.,
Chou, 1994; Soden,1997; Spencerand Braswell, 1997] and
the primaryreasonfor the continuinguncertaintyis the lack
of adequateglobalwatervapormeasurements
in the upper
troposphere.
Historically,measurements
of the water vaporfield have
relieduponradiosonde
profiles;theaccuracyandspatialextent of which are limited [e.g., Elliot and Gaffen, 1991; Sodenand Lanzante,1996]. Radiosonde
profilesare confined
largely to northernhemispherelandmasseswith the result
that the water vapor field in the upper troposphereof the
tropicalregionhasbeenpoorlyobserved.Satellitesprovide
a better way to achievegreaterspatialand temporalcoverage. The Microwave Limb Sounder(MLS) on the Upper
AtmosphereResearchSatellite (UARS) is sensitiveto water vapor in the upper troposphere.With a 3 km field of
view in the vertical, it has greatervertical resolutionthan
that of infrared instruments such as Meteosat [Schmetzand
Turpienen, 1988], the U.S. GeostationaryOperationalEnvironmental Satellite (GOES) [Soden and Bretherton, 1993],
and the TIROS OperationalVertical Sounder[Salathgand
Chesters, 1995], which are sensitiveto water vapor in a
broad layer of the upper troposphere• 300hPa thick. The
MLS measurementof upper tropospherichumidity (UTH)
is relatively insensitiveto cirrus clouds,giving it additional
advantagesover infrared techniques.The temporalresolution of MLS
is better than that of solar occultation
instru-
mentssuchasthe Stratospheric
AerosolandGasExperiment
2 (SAGE II) from which measurementsare limited to • 30
per day.For severalyears,MLS hasprovideddaily coverage
of the tropicalregionsinceits launchin September1991.
The MLS water vapordata revealsynoptic-scale
features
• Also at Joint Institutefor the Studyof the Atmosphereand and detrainmentstreamsextendingfrom tropicalconvective
Ocean,University of Washington,Seattle
regions[Readet al., 1995], as well asthe grossannualvariationsof the zonalmean[Elsonet al., 1996]. In the tropics,
Copyright1998 by the AmericanGeophysicalUnion.
Newell et al. [1996] have shown that the observeddistribution of water vaporis consistentwith the Walker circulation,
Papernumber98JD02702.
0148-0227/98/98JD-02702509.00
andNewell et al. [ 1997] haveshownthatuppertropospheric
3i,695
31,696
CLARK ET AL.: VARIABILITY OF UPPER TROPOSPHERIC WATER VAPOR
watervaporiscloselyrelatedto seasurface
temperature
vari- method used by Jacksonet al. [1990]. A searchradius
ationsin the easternPacific. In this paper,the MLS mea- of 2150 km was used with a scalingdistanceof 1000 km
surementwill be used to examinethe temporaland zonal so that footprints1000 km away from the givengrid point
variabilityof uppertropospheric
watervaporin thetropical have a relative weighting of 1/e whereasany that fall diregionandto account
for theobserved
variabilityby relating rectlyon the grid pointhavea relativeweightingof 1. Zonal
andmeridionalwinds and verticalvelocitywere takenfrom
it to known processes.
theEuropeanCentrefor Medium RangeWeatherForecast's
(ECMWF) initializedreanalysisdata,whicharegriddedwith
2. Data
a 2.5ø by 2.5ø resolutionon the standardpressurelevels.
UARS is in an almost circular orbit at an altitude of
585 km and an inclinationof 57ø to the equator[Reber,
1993]. It makes,,,, 14 orbitseachday with adjacentorbits 3. Results
beingseparated
by ,,,,2670km attheequator,
therebyallowing a zonalmeanandsixwavenumbers
to beresolved.MLS
makesa limb scanperpendicular
to the orbitpathat tangent
heightsfrom 90 km to the surface.Eachscantakes65.5 s
and consistsof a profileof measurements
whichare usually retrievedonto15 pressure
levels.MLS provides
a 3 km
field of view in the vertical. Latitudinal coveragechanges
from between 80øN and 34øS to 34øN and 80øS because the
satelliteperformsa yaw maneuver
aboutevery36 days,but
thetropicalregionis observed
daily,enablinga nearlycontinuous time series to be constructed. The MLS instrument
WhereasElson et al. [1996] focusedon the variability
of zonal meanwater vaporin latitude-timeplots,we focus
on the variability of zonal crosssectionsin longitude-time
plots. First, we discussthe variabilityat low frequencies
that is associated
with the annualcycle. Next, we discuss
intraseasonal
variability and the connectionbetweenwater
vaporandmeteorologicalfields.
To createlongitude-timeplots, the MLS footprintswere
firstinterpolatedontogrid pointsas describedin section2.
Data gapswere then filled in time usinga Kalmanfilter.
Most data gapsare of only 1 day, but a gap of ,,• 2 weeks
occurredin June 1992. The resultinglongitude-timeplots
for the time periodfrom December1, 1991,to May 3, 1993,
are shownin Plate 1, with the five regionsshowingthe five
latitude bins between20øN and 20øS. Gaps longerthan 3
dayshavebeenmasked.
is describedin more detailby Barath et al. [1993], andthe
measurement
techniqueis described
by Waters[ 1993].
The 205 GHz channelon MLS is principallyusedto measurechlorinemonoxidebut it is sensitiveto watervaporin
the uppertropospherewhen concentrations
are in the range
of 100 to 300 ppmv [Read et al., 1995]. The best sensitivity occurswhen the water vapor concentrationis about
3.1. The Annual Cycle
150 ppmv;of the standardUARS pressure
levels,it is at 215
In Plate 1 and particularlyin Plates la, lb, ld, and le,
hPa (,,,, 12 km at low latitudesand ,-• 7 km at highlatitudes)
that this concentration most often occurs. We therefore focus
an annualcycle in mixing ratio is apparent.Mixing ratios
are high in local summerovercontinentswhenconvectionis
our studyon the 215 hPa level.
Retrievalsin the upper tropospheremay be affectedby strongand low in local winter when convectionis weaker.
The northernhemispherelatitudes,20øN and 10øN, are
thick cirrus clouds. In the tropical region between6 and
12 km, the retrievalsare not significantlyaffected,but at shownin Platesl a and lb respectively.The highestwater
latitudespoleward40ø, a significantfractionof the mea- vapormixing ratiosoccurbetween50øE and 120øEoverthe
suredradiancesmay comefrom scatteringby cirrusclouds Indian Oceanand SE Asia from May to Septemberwhen
[Bond, 1996]. Ice crystalsin cirruscloudsat a concentra- convectionis strong.The annualcyclewas isolatedby least
in theretion of 0.1 g m-3 overa horizontal
distance
of 120 km squaresfittingto a sinewave.It is mostpronounced
could contributeto 20% of the absorptioncoefficientat gioninfluencedby the Indian monsoon.At 20øN and90øE,
215 hPa but will usuallybe less,and at concentrations
less over the Bay of Bengal, it accountsfor 80% of the variance
than0.01g m-3 theeffectis negligible
[Readetal., 1995]. in watervapor.Local maximain the annualcycleoccurjust
The data are derived from the initial MLS UTH retrieval
westof PanamaasNewell et al. [ 1997] havenoted,andminas describedby Read et al. [1995]. A new and improved ima over the cold eastern Pacific and Atlantic Oceans.
At the equator(Plate l c), the range of mixing ratios is
MLS UTH retrieval has recently been developed,which
among other featureshas a formal error estimationcalcu- smallerthanat the othertropicallatitudes.The smallerrange
Zone
lationandhasbeencomparedwith Vaisalathin-filmcapaci- is probablybecausethe IntertropicalConvergence
tive radiosondemeasurements.
Comparisons
with thenewer (ITCZ) is rarely situatedat the equatoritself [Philanderet
productshowthecurrentanalyzedproductat 215 hPa,which al., 1996]. The maximummixing ratiosare locatedoverthe
is usedin this paper,to be biasedhigh by 50-60 ppmv(ev- Indian Ocean and Indonesia. The minimum mixing ratios
occuroverthe easternPacificandare subjectto an east-west
erywhere)andto havea precisionof 5 ppmv.
Footprintswere interpolatedontopointsspacedevery5ø migrationon annualtimescales,resultingin lower valuesexin longitudearoundlatitudecirclesat 20øN, 10øN,the equa- tendingfurther west in June-Augustand further east from
tor,10øS,and20øS.Footprints
thatfell withina given"searchNovember
to January.
As notedbyNewellet al. [1997],the
radius"ofthesepointsweregivena distance
weighting
based amplitude
of theannualcycle,like therange,is smallest
at
uponthe regression
retrievalandspace-time
interpolationtheequator,
andprobably
forthesamereason.At theequa-
CLARK ET AL.: VARIABILITY OF UPPER TROPOSPHERIC WATER VAPOR
31,697
zonalwind andsurfacepressure,hasa periodof 30-60 days
andcanbe detectedacrossthetropics.The convective
signal
hasa broaderrangeof 30-95 days[e.g.,Kiladis and Weickmann, 1992; Salby and Hendon, 1994] and is strongestin
6O
5O
the Indian Ocean and western Pacific.
•
Eastwardmovingmoist featuresare apparentin the water vaporfields(Plate 1) andaremostprominentat 10øSin
•o
southern summer.
The moist anomalies at 10øS cover lon-
gitudesfrom 90øE to 245øE (115øW). One suchfeaturebeginsat 105øEandtravelsto 190øEin •22 days,thusmoving
lO
witha speedof ,-•5m s-1. At theequator
duringsouthern
summer,thereare eastwardmovingfeaturessimilarto those
at 10øS, but they are less intense;during southernwinter,
eastwardpropagationis lesscommon. Similar propagation
o
0
100
200
300
Iongi[ude(degrees)
Figure 1. Amplitudeof theannualcyclein watervapormixing ratio (ppmv) at 10øSand215 hPa.
of moist anomalies occurs at 10øN in northern summer with
a more limited longitudinalextent, being confinedmostly
to the Indian Ocean. There is also evidence for the eastward
movementof dry featuresbetween130ø and280øEin northtor,it exhibitsits maximumamplitudeat 90øW,off thecoast ern hemispherewinter at 10ø and 20øN (Platesl a and lb).
of South America, where it accounts for 53 % of the variance. In this section,the eastwardmovingmoistfeaturesandtheir
Newell et al. [ 1997] showedthat water vapor variationsin relationshipto theMJO will be discussed.
the region 80ø-90øW and 0ø-10øS are related to variations
The periodicityof the disturbances
at a fixed longitude
in seasurfacetemperature(SST) causedby E1Nifio andalso may be investigated
from thepowerspectrumor from thelag
to anomalouschangesin SST; the latter accountfor •56%
correlogram.For a data set like that in Plate 1, the lag corof the water vaporvariance.
relogramhassomeadvantages,
as it is insensitiveto change
In the southernhemisphere,Platesl d and le, mixing ra- of phase in the waves from one winter to the next and to
tios are greater over Indonesiaand South America during whetherthe waveperiodis a properharmonicof the length
summer
and lower
over the eastern Pacific
in winter.
In
of the data set. Furthermore, it can reveal some information
the longitude-timesection(Plate l d), strongannualcycles evenwhenthe wavesexistfor little morethanoneperiod.
can be seen over the African continent and South America.
Figure 2 showsthe lag correlogramfor 160øE and 10øS
Figure 1 showsthe variationof the amplitudeof the annual basedon the 520 days of data usedfor Plate 1. Here, 160øE
cycle with longitudeat 10øSand revealsfurtherlocal max- is illustratedas the longitudewhich exhibitsthe greatestinima over Indonesia and the mid-Pacific. The influence of the
traseasonalvariability. The annualcycle and low-frequency
cold easternPacificin suppressing
convection
and leading variabilitywasremovedfrom eachlongitudeusinga 150 day
to a drier uppertroposphere
andof the warm westernPacific Butterworthfilter. Also shownis the red noisebackground
leadingto a moisteruppertroposphere
is evidentin Plate spectrumcomputedfrom lag-oneautocorrelation[Gilman et
1d. This constitutes
the Walkercirculationwith risingair in al., 1963] and the 95% confidence limits. The data have anthe west and subsidingair in the east. The influenceof the ticorrelationlessthan -0.2, at lags of 20-35 days, well beWalkercirculationon uppertropospheric
watervaporfrom low thered spectrumandsignificantat the95% level. This is
MLS has been noted by Newell et al. [1996] in two short suggestiveof wavesof period40-70 days,an interpretation
periodsof datafrom September17 to October22, 1991,and supportedby the existenceof positivecorrelationsat lagsof
February7 to March 14, 1994. Here, the presenceof the 40-70 days. Inspectionof Plate 1 suggeststhat thesecorWalkercirculationcanbe seento be a morepersistent
fea- relationsarisefrom featureswhich are only apparentin the
ture. Mixing ratios over the westernPacificshoweast-west southernsummermonths. Only slightly over one cycle is
migrationsimilarto that at the equator,with low humidity apparentin each year and so very high correlationscannot
foundfurtherwestduringwinter,whenthe oceanis colder. be expected.Moreoverthe magnitudeof the correlationat
Eastwardmovingfeaturesduringthe summermonthsorigi- the waveperiodwill be necessarily
lessthanthat at the half
nateoverthe warm IndianOceanandarequicklyterminated period. By cross-correlating
the time seriesat each longioncetheyreachthe easternPacific.Thesehavea higherfre- tude with the time seriesat 160øE usingdifferentlags, the
quencythanthe annualcycleandareinvestigated
by analyz- estimated
propagation
speed
wasconfirmed
tobe4-5m s-x.
ing the time serieswith a focuson intraseasonal
timescales.
Figure 3 showshow the power at eachfrequencyvaries
3.2. Intraseasonal Variability
with longitude for 10øS. The strongestsignal corresponds
to a frequency
of 0.014days
-• or a periodof 70 daysand
with the convecOne of the main modesof variability in the tropicsis falls within the frequencyrangeassociated
the Madden-Julianoscillation(MJO) [Madden and Julian, tive signalof the MJO. The powerspectrumwastestedfor
1971]. The MJO compriseslarge-scalecirculationanoma- significanceat 95% by computingthe signalto red noiseralies associated
with convective
anomalieswhichpropagate tio at eachlongitudeand comparingit with the chi-squared
eastward
at 3-6 m s-1. The dynamical
signalof MJO, in distributionfollowing the methodof Gilman et al. [1963].
31,698
CLARK ET AL.' VARIABILITY
'G
•_
OF UPPER TROPOSPHERIC WATER VAPOR
0.5
c)
._
o
o
50
100
150
200
Lag (days)
Figure 2. Auto-correlationof the time seriesof 215 hPa water vapormixing ratios at 160øE and 10øS
(solid line), red noise(dashedline), and 95% confidencelimits (dash-dottedlines).
At 70 daysand160øE,values
greater
than112ppm
2 aresig- This contrastswith the signalin watervaporwhich is dominificant. The signalis confinedin longitudefrom 140ø to
180øE. This area, over the western Pacific, has been shown
to exhibit strongintraseasonal
variabilityin outgoinglongwaveradiation(OLR) which is often usedto infer deepconvection. Salby and Hendon [1994] examined11 years of
OLR datafrom the advancedvery high resolutionradiometer (AVHRR) and found that intraseasonalbehaviorwas coincidentwith centersof climatologicalconvectionandwarm
seasurfacetemperature,with maximaovertheIndianOcean
and westernPacific and secondarymaxima over the eastern
Pacific,Africa, and SouthAmerica. They notedthat mostof
the intraseasonal
variance occurred over the Indian Ocean.
nant over the westernPacific. The absenceof a signalin the
water vapor field over the Indian Ocean may indicatethat
convectivemoisteningis not reachingthe 215 hPa level.
In order to focus on the intraseasonalfrequencyrange
characteristic
both of the MJO
and of the eastward
I
I
I
0.030
33
0.020
0.010
IO0
0.000
520
1 O0
200
300
Longitude(deõrees)
50.0
100.0
mov-
ing disturbancesseenin Plate 1, we filteredthe water vapor
data with a 30-85 day Butterworthband-passfilter similar
to that used previouslyin studiesof OLR [e.g., Salby and
Hendon, 1994; Zhang and Hendon, 1997]. When applied
to 520 daysof data at 10øSand 160øE,the 30-85 day filter
band accountedfor 35% of the total power comparedwith
19% for the annualcycle. Mixing ratiosin this bandvary by
150.0
200.0
250.0
[ppmv']
Figure 3. Powerspectrum,
asa functionof longitudeandfrequency,
of deseasonalized
watervaporat
215hPa.Contour
intervals
are50 ppmv2.
CLARKET AL.: VARIABILITYOFUPPERTROPOSPHERIC
WATERVAPOR
31,699
Power spectrumat 215 hPa, 10S
24 -
-- -
1.6
305---'-----•'--'-------•
• ........
• 37
• .-••••-•:•
••
1.2
m 104
0.4
-
130 -
-
173 -
-
260 -
-
520
•
-4
,
•
,
•
,
•
-2
0
Wavenumber
2
0.0
,
4
Figure 4. Powerspectrum,as a functionof wavenumberandfrequency,of de-seasonalized
watervapor
at 215 hPa and 10 øS.
+ 60 ppmv.At the equator,intraseasonal
activityoccurring 1-3 with periodsfrom 30 to 60 days.Modesat k = 0 havea
in the30-85 day bandaccounts
for 35% lesspowerthanat smallshareof thepower,mostlyat 37 and74 daysandrep10øS.
resenta zonally symmetricmode. Becausethe zonal struc-
At 10øS the water vapor anomalies,consistentwith the
OLR signal[SalbyandHendon,1994],possess
a strongseasonalitythat is apparentin the longitude-timesection(Plate
l d). The eastwardpropagatinganomaliesare strongwhen
the ITCZ is near the latitude in question(local summer)
but virtually absentwhen the ITCZ is not (local winter).
When the eastwardpropagatinganomaliesare strong(December1991 to March 1992), intraseasonal
variabilityac-
tureandfrequencyof the watervaporanomaliesare similar
to thoseusuallyidentifiedwiththeconvective
component
of
theMJO (e.g.,OLR), we tentatively
identifytheseanomalies
with the MJO.
To separatethe modesof variabilitysuggested
by Figure4, we calculateextendedempiricalorthogonal
functions
(EEOFs) [Weareand Nasstrom,1982, Wanget al., 1995].
Theirmorecommoncousins,
empiricalorthogonal
functions
counts for as much as 63% of the total variance, but, when (EOFs) , identifycoherent
variations
by findingeigenvecthey are weak (August-November1992), intraseasonal
vari- torsof the(symmetric)
covariance
matrixandrankingthese
ability only accountsfor 7% of the total variance.
eigenvectors
in descending
orderof variance
explained.For
Furtherinsightsaboutthe natureof the intraseasonalvari- thelongitude-time
arrayof watervaporat 10øS,thecovariability comefrom a wavenumber-frequency
spectralanaly- ancematrix for ordinaryEOFs wouldbe formedby sum-
sis. The low-frequencyvariability was removedfrom 520 ming(intime)thecovariance
qi(t)qj(t) of watervaporqi(t)
days of data using the 150 day band-passfilter as before, at everycombination
of longitudinalgridpoints.The EOFs
and the filtered data havebeenregressedagainstzonal and wouldbe one-dimensional
functionsof longitudeandwould
temporalharmonics
of theformcos(k:r-cot) to identifythe yield a clearerpicture of zonally coherentvariations. Expower at discretewavenumbersk and frequenciesco. Fig- tendedEOFs are found by calculatingcovarianceof water
ure 4 showsthe resultsof this analysis.There is very little vaporat each grid point not just with water vaporat other
powerat the lowestfrequencies,owingto the filtering;most gridpointsbutalsoat differentlag timesl, i.e., qi,!, where
power is concentratedin positive(eastward)wavenumbers the lag variesin this casebetween-60 and +60 daysat 5
31,700
CLARKET AL.: VARIABILITYOF UPPERTROPOSPHERIC
WATERVAPOR
Extended EOF
Extended EOF 2
1
60
20
0
-2O
-60
0
6,0
120
180
240
300
0
360
60
Longitude
1.20 1.80 240
300
360
Longitude
Extended EOF 4
Extended EOF 3
6:0
,•
20
20
•o
0
0
• -2O
-20
..
-60
0
60
120
180
240
300
360
0
60
120
180
240
300
360
Longitude
Longitude
Figure 5. Spatiotemporalstructureof the first four extendedEOFs of deseasonalized
water vaporat
215 hPa and 10 øS.
day intervals.EEOF analysisyieldsa clearerpictureof zonally and temporallycoherentvariations.
The first four EEOFs are shownin Figure 5. EEOFs (and
EOFs) sometimesoccurin conjugatepairs,identifiedby the
closeness
of theireigenvalues,by theirassociated
time series
which are in quadratureand by their spatiotemporal
structure which is alsoin quadrature.The significanceof conjugatepairsis thatthey describea quasi-periodicoscillationof
somesort.The first pair of EEOFs in figure5 havethe characteristicsof a conjugatepair, as do the nextpair of EEOFs.
The next four EEOFs (not shown)form two moreconjugate
pairs, but their interpretationis more difficult. The oscillation indicatedby the first pair of EEOFs showseastward
propagationbetweenabout75øE and200øE(asin Plate 1d);
and the time lag betweenmaxima is 55 days. The phase
anceduringsouthernwinter as notedabove. Spectralanalysis confirmswhat a visual inspectionsuggests:The dominant spectralpeak for PC 1 and PC2 is at 50 days,while the
dominantspectralpeakfor PC3 andPC4 is at 36 days.(Note
that the temporalresolutionof the PCs is I day, while the
temporalresolutionof the EEOFs is 5 days.) Basedon the
spatiotemporal
structurerevealedin Figure5 andthecharacteristicsof the time seriesjust discussed,
it seemsreasonable
to associatethe firstpair of EEOFswith thebandof variance
at 50-60 days and wavenumbers1-4 in Figure 4 and to associatethe secondpair of EEOFs with the k - 0 modeat 37
daysin Figure 4.
Figure7 is a reconstruction
of the longitude-timesection
in plate l d usingcombinations
of the PCs. The original
longitude-time
section
is shown
forcomparison
alongside
speedof anomalies
is 3-4 m s-•. Highvariance
is con- reconstructions
usingPC 1 and2 only,PC 3 and4 only,and
fined to the Indian Ocean and westernPacific(75øE-200øE).
By contrast,the oscillationindicatedby the secondpair of
EEOFs is somewhatmore broadlydistributedin longitude,
showsno identifiableeastwardor westwardpropagation,and
the time lag betweenmaxima is 35 days.
The time seriesof the coefficientsof the EEOFs, or principal components(PC) (Figure 6), revealboth the intraseasonal cycles and the seasonalenvelope,with smallervari-
PC 1, 2, 3 and4 combined.Figure7 highlightstheeastward
movinganomaliesin theoriginaldataanddemonstrates
how
theeastwardpropagation
is capturedby theEEOFs.Most of
the eastwardpropagationwestof the datelineis pickedout
by EEOF 1 and2. EEOF 3 and4 accountfor lessof thevariance, but their inclusion increasesthe eastwardextent of the
anomalies.
3.2.1. Relationship with other meteorological fields.
CLARKET AL.' VARIABILITYOFUPPERTROPOSPHERIC
WATERVAPOR
time series of PC 1
i
60
40
20
0
-20
,
.
I
,
ß
i
.
,
31,701
time series of PC 2
!
,
,
6O
':
4O
.
20
I'
!
.
.
i
.
,
i
.
.
i
•
i
i
•
•
i
,
•
"
f'
ß
-20
-40
-40
-60
.
.
i
,
Jan92 Apr
,
i
i
i
Jul
I
i
i
I
i
i
Oct Jan93 Apr
-60
Jan 92
Apr
time series of PC 3
•, •,
•
.:
• 3. ':
.. •.
•
,.
•
•
Oct Jan93 Apr
6ø
f'
•
,
,
time series of PC 4
40
20
Jul
i
ß
20
.
-20
-40
-60
'
............
Jan92 Apr
-40
-60
Jul
Oct Jan93 Apr
,
,
,
Jan 92 Apr
Jul
Oct Jan93 Apr
Figure 6. Time seriesof the coefficientsof theEEOFs (principalcomponents
(PC)) in thepreviousfigure.
In orderto strengthen
the proposedidentification
between over the westernPacific and vertical velocity is still strong
thewatervaporanomalies
andMJO,thewatervaporfield
is compared
withverticalvelocityandzonalwindsobtained
fromECMWFreanalysis.
Weexamine
thelifecycleof water
vaporduringthe eastwardmovingeventin December1991
to February1992identified
in section
3.2, andcompare
this
tomidtropospheric
verticalvelocity
at 500hPa.Bothquantitiesareplottedas5 dayaverages
in Figure8 wherewater
vapormixingratios> 170ppmvareshaded
in greyandupwardverticalvelocitylessthan-0.05Pas- • iscoloredblack.
The datesindicatethemiddleof eachpentad.
From December17-22, mixingratios> 170 ppmvare
spread
outin a bandalongtheequator.Severalsmallpocketsof verticalvelocityarestrungoutalongtheequator
from
Africa as far as 135øW and there is indication of the South
PacificConvergence
Zone(SPCZ).On theDecember27, the
distribution
of highwatervaporbeginsto shiftfrom Africa
andtheIndianOceantowardIndonesia.
Similarly,thedistributionof pocketsof verticalvelocityhasshiftedfromAfrica
and the Indian Ocean toward Indonesia. Both the SPCZ and
theITCZ arewellpronounced
in theverticalvelocityfield.
OntheJanuary1, thereis a largeverticalvelocitypocket
at this time. On the January11, the pocketof verticalvelocity beginsto breakup, but watervapormixing ratiosremain
high andcontinueto migrateeastwardandto intensifyalong
the SPCZ. The SPCZ remainsa strongcharacteristic
in both
the verticalvelocityand watervaporfieldsuntil February5,
but lingersas a featurein water vaporfor a further 5 days.
On the February10, the small pocketsof verticalvelocity
havereturnedto Africa andthe IndianOceanandmixing ratios have begunto increasehere also. Finally, water vapor
dispersesfrom the SPCZ.
This correspondence
betweenwatervaporandverticalvelocity illustratesa strongrelationshipbetweenhigh mixing
ratios and convectionand indicatesthat the convectivesystem as a whole moveseastwardratherthanthe water vapor
beingadvectedalongat the 215 hPa level. The patternsof
water vapor mixing ratio are similar to those in OLR during the MJO in March-April 1988 describedby Matthewset
al. [1996]. Both begin as a low intensityband along, and
mostly southof, the equatorwhich becomesorganizedover
Indonesiaand the west Pacificand is followedby development in a southeastdirectionalong the SPCZ. Again, the
SPCZ remainsprominentin the watervaporfield for longer
overthewestPacificanda corresponding
areaof highmixing ratiois beginningto develop.By January6, the moist than in OLR. Knutsonand Weickmann[1987] noted the dearea has moved eastward from Indonesia to become centered
velopmentof OLR f¾omthe Indian Oceanto the SPCZ. Rui
31,702
CLARK ET AL.: VARIABILITY OF UPPER TROPOSPHERIC WATER VAPOR
M LS data
PC 1 2
PC 3 4
PC I 2 3 4
jUnl,,,,,,,,l,,,,,,,,,,,,,,,,,,,,,,,,,,l,,,,,,
.....
,,....,....']l'....I.....
;• ........
J•
•?;•:.:...
Apr
Feb •:....ß-•'•
•
;:.,:.:...::•:•.::•':-;•...:.:.::?;:•.:•.::•:::½•,•.:;•;....:•.•---•:•::•:.;::•-•:c.:•.:..
:'•:•:•:•;;;::•.•"?•7•;';:"-:::"-'--•."•;•
':.•':•::•:•"':::•:•'%-.'"'•
'
::7•::.•::..
•.•-':•:.'•-':•.
::;:•'"'-:
..........
......
•..%'.'•'.---.:
•
..•7...:•:
.•:'
.. .•:...:.:;.."
::Z•:'(•::.•.:•
c?..---.;;'•
::.d.•,.•.•,•:•:•:.:.:;.....•:•.•::.:::,;
.. • ..,......:.:...
ß.:.• .'...,
;•-•7.:¾-.--:.
•.'27..-.:.:•:•:..:...•....::•:
.-'•:..
•e
Dec
................
.
-- _
:;•:•':•:.•'.....•...7
•'½'•:"" '...•::•z..• ....... •-:..... •:• ..... •.,,•:.....:....:...•,............
........-• ....... ,.•,:•.:;.:.;...............
..•:.:•:•e::•.....
.:::•:•
...... '•.d,•:-:•
.... .... ,::.•:::.....::...:.
.;:::.•. .;•.:-•:•:;•:•:•-;•::%:.•?:•../•,'•
:.:
•.•.:.•.-.-.-:
....
Oct
.....
•.;•
...... -;•:•.•.......
•::•.•:.
•;
.•.:::.:•.•,.:::•
• ...•-•.?:'17•7;
• •..
..•:..:......:
..... .•.•.... :.•
:.....
,.
:....•.:::;•--•:,•
..-'•:•'•..•
-•;.....,•?'-'-:.%•,.•:'::
-...... ?..•....
Aug'•'"'
.•?•½':':?
:.:.?:•::•::::?:•::•??--';-::
.......
ß
.':"•
.......
'::'"'-•'"":%•::;•:5:•,.•?•
f'....
........
Aug
:.:........•.
•-•,
:.:.::•;•½•.:
....•:
..............
:..•:.::.....:•:•...•.Z•
..½..:,:.:'•
...
•un '.......................
•'•
.............
:.':•;"•.:,.•.;•..•:?:•;;.".:?;•':'/"•'•....
.....d;;:?::::•e..;.....•
.... •.. '•'..•::.•:.::•.....•.
.... ...•.•:•..:..:c;•.•,:•..•.:•Z;•:::?.
..•.." • ß ..,'
.•:-:...•.."
.g?•'•.:
•:'•::-.:;:....•:•.:•::.::::•--:';.:::•::•:•-•:?,•,•..:...
:•.
......
Jun
.... :•::•::•;.•:
........:..
'.';:.•:.?:•? ....>•e•. ..•.::•.•:.
:..••:-' ....
..:,:•;•:•:.:;..":.::-'-•4•"•:;
':""•"
-.-,...
Apr ...'
........:•:..................
ß
.....
'-:•,%
....
Apr
•..•...:...•:;•.:..:.:.•....•
.....•.:;..:..?:.•:•::•;.;:::;:.•:.::•:•:;•;.•...•.•
.•.•;•:.7'•
ß
.,•:•i-.•.•i•-::.i•.:?
•
:.•.•:•,:?;:-:•:.:
ß
::•::•;
.•?-"-•:::•'•:.•:?•?•7:"::'"'-:':':'•?•"•:
....
7,"•
..........
•;•(½•:•:::•::?•:':
............
•-•.
•::::-•:•:.•.-.-
'•:.;•:•.•
Feb
Feb.• .......
..-•:?
• ...............
?:--.?•?•:::½:•.-.
..........
".':•-•:•'•:•::•:½--•
.......
.(.'7:•--.:::;:•:•:•?•:.•
-,•::••-:;:.•:
.?
•.•¾
'•c•,•.....•
' :•:...•
.........
.•:'"•':-'-•:•
.....
?7.
.......
7'"'%:::::.•::•?"• .........
...........
•=•
.......
..•?:':'•'•-•-•:•:½•'•'•:?.?('":-•:':'-'"
..:.::;..•.;•:.:•
.-.:,:.'-.•-
•c 91 z•?•'v•.'•',
:.•:-.:';•::•:::•"
.•::•:-•'-(;
'•':'•"'
.........
:'"•::':•
.•'•:•
0
120 180 2403•3•
•ngimde
•
ß•:'•½
........•'•..........
120 180 2403•3•
•ngimde
•
::•'•:•
........•:=•':
.....•c 91
120 1802•3•3•
•ngimde
•
120 180 2403•360
•ngimde
Figure 7. Reconstruction
of the longitude-timesectionusingcombinationsof the principalcomponents(PC).
and Wang [1990] found that developmentalong the SPCZ
occurredin strongMJO events,and Matthews et al. [1996]
found that enhancementor excitation of convectionalong
the SPCZ is observedin virtually all MJOs from 1979 to
analysisshowedthat the maximum intraseasonalvariancein
zonal wind was off the equatorandconcentratedin the western Pacific. This is an importantreasonfor the mixing ratios
being higher at 10øSthan at the equatorand further indica1988.
tion thatthe high mixing ratiosare a resultof enhancedevapThe high water vapormixing ratios also appearto be re- oration and convection. The fact that the largest variances
lated to westerlywind burstsin the lower troposphere.Wang do not lie on the equatorappearsto be in conflictwith those
[1988] showedthat althoughthe mean surfacewinds in the theories which attribute the MJO to moist Kelvin waves, as
tropics are easterly,they are often westerly in the region remarkedby Zhang and Hendon [ 1997].
3.2.2. Interannual variability. There have been sugwhere convectiveanomaliesassociatedwith the MJO originate.Zhang [ 1996] foundthatdeepconvectiontendedto oc- gestionsthat the MJO varies in strengthfrom year to year
cur in connectionwith 850 hPa westerlywind perturbations, and that this may be related to the E1 Nifio-SouthernOsciland Hendon and Glick [1997] showedthat westerly wind lation phenomenon[e.g., Lau and Chan, 1986]. Accordanomalieswere relatedto enhancedevaporation.A time se- ingly, we have investigatedthe behaviorof the MLS upriesof zonalwindsat 850 hPaand 10øSis shownin Figure9. per tropospherichumidity data for other years. There are
Regionsof westerlywind are evidentduringsouthernsum- some practical difficulties with this as data gapsprevent a
mer in both 1992 and 1993 and are well correlated with the
detailedinvestigationof the southernsummermonths. The
eastwardmovingmoistfeaturesin the longitude-timeplot of data gapshave increasedin lengthas the spacecraftand inwatervapor(Plate 1d). The longitudinalextentof the west- strumenthave aged and make later years more difficult to
erly windsin 1992 (from 40ø-200øE) corresponds
well with study.Longitude-timesectionsfor southernsummerperiods
the more evident eastwardpropagationin the water vapor are shown for 10øS in Plate 2 for all the available MLS data.
Data have been treatedin the sameway as for Plate 1. Befield of this year.
In contrast to the zonal winds at 10øS, westerly wind cause there were not sufficient data for the southern summer
burstsat the equator(not shown) are weaker and have a of 1994-1995, that periodis omitted.
limited longitudinalextent. Similarly,Zhang and Hendon
[ 1997] who also used850 hPa zonal winds from the ECMWF
Signsof eastwardpropagation
aremostapparentin 19911992 whenmixing ratiosbetween170 and 195 ppmv reach
CLARK
ETAL.'VARIABILITY
OFUPPER
TROPOSPHERIC
WATER
VAPOR
1991
Dec
17
1992
Jen 21
1991
Dec 22
1992
Jan 26
1991
Dec 27
1992
Jan
.31
1992
Jon
1
1992
Feb
5
1992
Jon
6
1992
Feb
10
1992
Jon
11
1992
Feb
15
1992
Jan
16
31,703
Figure8. The 5 dayaveraged
watervapormixingratios,at 215hPaand> 170ppmv(grey)andECMWF
verticalvelocityat500hPaand<-0.05Pas-• (black).Datesindicate
themiddleof thepentad.
as far as 240øE(120øW).This is furthereastthanin any 1992, andin 1992-1993 eastwardmovementis lessevident.
other year, and it shouldbe notedthat this was an E1 Nifio In the summerof 1993-1994(Plate2c), totalvarianceis
year [Trenberth,1997]. In all otheryears,mixingratios the sameas in 1991-1992, but that associatedwith east> 170ppmvneverextendfurthereastthan210øE(150øW). wardpropagation
is less. In 1995-1996,a La Nifiayear
Knutsonand Weickmann[1987] noticedsimilar behaviorin
[Trenberth,
1997],thelowestmixingratiosof anyof the
OLR andproposed
thatlow seasurfacetemperatures
or descending
motionassociated
withtheWalkercirculation
suppressfurthereastwarddevelopment.
Hence,the greaterextent of the anomaliesin 1991-1992is probablya resultof
years,<70 ppmv,arereachedovertheeasternPacific,and
themeanandmaximummixingratiosarealsolowerthanin
otheryears,withthemaximum
mixingratiosneverexceed-
ing245ppmv.Although
thedatain 1996-1997
aresparse,
the eastward shift of convection that is a fundamental com-
mixingratioscanbe seento be ashighasthosein 1991-
ponent of ENSO.
1992,andtheremayevenbesigns
of eastward
propagation
In 1992-1993 (Plate 2b), the 30-85 day filter band ac- in lateFebruary
andearlyMarch.Highmixingratiosdonot
countsfor 52% of thevariancecomparedwith 63% in 1991- havethesame
eastward
extent
asin 1991-1992.
Salbyand
31,704
CLARK ET AL.: VARIABILITY
OF UPPER TROPOSPHERIC WATER VAPOR
May.g3
May.93
Dec.92
a
v
Dec.g2
b
---- Aug.92
Aug.92
_
E
i
Apr.g2
Apr.g2
•,
Dec.91
Dec.91
1O0
0
200
1 O0
0
3O0
300
May.g3
May,g3 -
Dec,92
c
200
longitude(degrees)
longitude(degrees)
d
Dec.92
• Aug.g2
'-- Aug.g2
E
E
Apr.g2
Apr.g2
Dec.91
Dec.91
0
1oo
200
1 O0
0
300
200
300
longitude(degrees)
longitude(degrees)
May.93
30
20
e Dec,92
ß
10
õ
o
,_o -lO
v
-20
-30
Aug.g2
o
E
45
go
155
180
225
270
315
560
longitude
Apr.g2
•
Dec.91
0
1 O0
200
longitude
(degrees)
500
70
95
120 145 170 195 220 245
WaterVapour(ppmv)
Plate 1. Longitude-time
sections
of watervapormixingratios(ppmv)at 215 hPaandlatitudesof (a)
20øN, (b) 10øN,(c) 0ø, (d) 10øS,and(e) 20øS.Data havebeenKalmanfilteredanda maskappliedto
gapsof morethan3 days.
CLARK ET AL.: VARIABILITY OF UPPER TROPOSPHERIC WATER VAPOR
,-
29,Mar,92
•
'
31,705
30.Mar,93
:
-
a
b
-
28. Feb.92
28,Feb,95
o
-,-'
o_
---'
29.Jan.92
29.Jan,g3
E
E
30.Dec,g2
30.Dec.91
01 .Dec.92
01. Dec.91
1O0
0
2OO
300
1 oo
o
300
29.Mor.96
50.Mar.94
d
c
28.Feb.96
28.Feb.94
---
200
longitude(degrees)
longitude(degrees)
•
29.Jan,94
2g,Jon.96
._
o_
30. Dec.95
30.Dec.93
01 .Dec.95
01 .Dec.93
100
0
200
1 oo
0
300
30.Mot.g7
200
300
longitude(degrees)
longitude(degrees)
--
e
30•
20!
28.Feb.97
10
-•o
-2o:
-oi }) '
---' 29,Jan.97
E
0
45
f
90
\'
155
180
225
270
315
560
longitude
50.Dec.96
•
J,
01 ,Dec.96
0
1oo
200
longitude(degrees)
.300
70
95
120
145
170
195
220
245
WaterVapour(ppmv)
Plate 2. Longitude-timesectionsof watervapormixingratios(ppmv)at 215 hPaand 10øSfor 120 days
of southernsummerfrom December-Marchin (a) 1991-1992, (b) 1992-1993, (c) 1993-1994, (d) 19951996,(e) 1996-1997. Data havebeenKalmanfiltered,anda maskhasbeenappliedto gapsof morethan
3 days.
31,706
CLARK ET AL.' VARIABILITY OF UPPER TROPOSPHERIC WATER VAPOR
Apr.g3
Jan.g3
.--
0ct.92
=
Jun.92
Mar.g2
Dec.91
0
1 O0
200
300
Longifude(degrees)
0.00
6.00
[rn,-•]
Figure9. Longitude-time
section
ofECMWFzonalwinds(ms-•) at 10øSand850hPa.
Hendon [1994] found that the convectivesignalwas absent extentof the intraseasonal
signalwas limited to the western
duringthe strongENSO cycle of 1982-1983 The relation- Pacificwhere it dominatedoverthe annualcycleduringthe
ship betweenthe MJO and ENSO warrantsfurtherinvesti- time periodfrom December1991 to May 1993.
gation.
In additionto the eastwardpropagatingmodesinterpreted
as the MJO, zonally symmetricmodeswith periodsof •37
4. Conclusions
and 74 dayshave alsobeenidentifiedfrom the EEOF analysis, combinedwith spectralanalysisof the principalcomWe haveshownthatwatervaporin thetropicaluppertro- ponents(Figures5 and 6) and from power-spectrum
analypospherevarieson bothannualandintraseasonal
timescales. sis and wavenumber-frequency
spectralanalysis(Figures3
The annualcyclecanbe clearlyseenin longitude-time
sec- and 4). These may be relatedto standingcomponentsof
tionsof watervapor,butduringsouthern
hemisphere's
sum- tropicalconvectionwhich havebeen observedin OLR and
mer, variabilityat intraseasonal
timescalesexceedsvariabil- 150 hPa divergence[e.g.,Hsu et al., 1990; Zhu and Wang,
ity at the annualtimescale.The annualcycleis •nostpro- 1993;Zhangand Hendon,1997].
nouncedover the landmassesof South America and Africa,
A sequenceof pentadmaps showedthe movementof
in agreementwith the studyby Newell et al. [1997]. Al- moist anomalies from Africa across the Indian Ocean, inthoughthe annualcycleis generallysmallerovertheoceans, tensification over Indonesia and the West Pacific and devel-
localmaximaoccurin someoceaniclocations,
notablyin re- opmentalongthe SPCZ. The verticalvelocityfield hassimgionsinfluencedby the Asian monsoonandoverIndonesia
ilar characteristics,
lendingsupportto our assertion
thatthe
and the central Pacific.
moistanomaliesat 215 hPaareassociated
with slowlytransLongitude-timesectionsrevealedeastwardmovingmoist
lating convection.The origin of the moistanomalieswould
anomalies
thatweremostprominentat 10øSfromDecember
appearto be enhancedsurfaceevaporationduringwesterly
1991 to March 1992. It is significantthatthisperiodcoin- wind bursts,which were seento be stronglycorrelatedwith
cideswith an E1Nifio. The featuresoriginatein the Indian
watervaporandareknownto play an importantpartin genOceanandpropagate
atspeeds
of 3-4m s- • untiltheyreach eratingandmaintainingthe MJO.
theeastern
Pacific.
TheEEOFanalysis
andwave-frequency
analysis
revealeastward
propagating
modesat 30-60days.
atthe
Thesespectralcharacteristics
in spaceandtime (Figures. Acknowledgments.Thewatervapordatawasproduced
Laboratory,
California
Institute
of Technology,
un4, 5, and6) togetherwith the longitudinal
lifecycleof the JetPropulsion
der
contract
with
NASA
and
funded
through
its
UARS
Prqject.
The
moistanomalies
observed
in longitude-time
sections,
resemworkwassupported
by NERCin theUK andby NASAcontracts
ble thoseof the MJO. We thereforeidentifythe anomalies NAS1-96071
andNAS5-32862.
Theauthors
thankH.C.Pumphrey
withtheconvective
component
of theMJO.Thelongitudinal for useful discussions.
CLARK ET AL.' VARIABILITY OF UPPERTROPOSPHERIC
WATERVAPOR
References
Barath,F., et al., The Upper AtmosphereResearchSatelliteMicrowaveLimb SounderInstrument,J. Geophys.Res.,98, 1075110762, 1993.
31,707
sphericwatervaporfrom UARS MLS, Bull.Am.Meteorol.Soc.,
76, 2381-2389, 1995.
Reber,C.A., The upperatmosphere
researchsatellite
(UARS), Geophys.Res.Lett., 20, 1215-1218,1993.
Rui, H., and B. Wang, Developmentcharacteristics
and dynamic
structureof tropicalintraseasonal
convectionanomalies,J. At-
Bond, S.T., The potential effect of cirrus on Microwave Limb
mos. Sci., 47, 357-379, 1990.
Sounderretrievals,Ph.D. thesis,pp.161-173, Univ. of EdinSalath6,E.P. and D. Chesters,Variabilityof moisturein the upper
burgh,Scotland,UnitedKingdom,1996.
troposphere
asinferredfromTOVS satelliteobservations
andthe
Chou, M.-D., Coolnessin the tropicalPacific during an E1 Nifio
ECMWF modelanalysisin 1989, J. of Clim., 8, 120-132, 1995.
episode,J. of Clim., 7, 1684-1692,1994.
behaviourof clouds,
Elliot, W.P.andD.J. Gaffen,On theutility of radiosonde
humidity Salby,M.L., andH.H. Hendon,Intraseasonal
archivesfor climate studies,Bull. Am. Meteorol. Soc., 72, 1507temperatureandmotionin the tropics,J. Atmos.Sci., 51, 22071 •;9fl
1 c•c• 1
999zl
100A
Schmetz,J., and O.M. Turpeinen,Estimationof the uppertropospherichumidity field from METEOSAT water vapourimage
ley, andR.S. Harwood,Space-timevariationsin water vaporas
observedby the UARS MicrowaveLimb Sounder,J. Geophys.
dataJ. Appl. Meteorol.,27, pp 889-899, 1988.
Soden,B.J., Variationsin the tropicalgreenhouse
effect duringE1
Res., 101, 9001-9015, 1996.
Nifio, J. of Clim., 10, 1050-1055, 1997.
Gilman,D.L., EJ. FuglisterandJ.M. MitchellJr., On the power
Elson, L.S., W.G. Read, J.W. Waters, P.W. Mote, J.S. Kinners-
spectrumof red noise,J. Atmos. Sci., 20, 182-184, 1963.
Hendon, H.H. and J. Glick, Intraseasonalair-seainteractionin the
Soden,B.J.andEP. Bretherton,
Upper-tropospheric
relativehumidityfromtheGOES6.7/zmchannel'
Method
andclimatology
for July 1987,J. Geophys.
Res.,98, 16669-16688,1993.
tropicalIndianandPacificoceans,J. of Clim.,10, 647-661,1997.
Houghton,
J.T.,G.J.Jenkins,
andJ.J.Ephraumus,
ClimateChange: Soden,B.J., and J.R. Lanzante, An assessmentof satelliteand raThe IPCC ScientificAssessment,
CambridgeUniv. Press,New
diosonde
climatologies
of upper-tropospheric
watervapor,J. of
York, 1990.
Clim., 9, 1235-1250, 1996.
Hsu, H-H., B.J. Hoskins and F-E Jin, The 1985/86 intraseasonal Spencer,
R.W., andW.D. Braswell,How dry is thetropicalfree
oscillationand the role of the extratropics,
J. Atmos. Sci., 47,
troposphere?
Implications
for globalwarmingtheory,Bull.Am.
823-839, 1990.
Meteorol. Soc., 78, 1097-1106, 1997.
Jackson,D.R., R.S. Harwood, and E. Renshaw,Testsof a scheme Trenberth,K.E., The definitionof E1 Nifio, Bull. Am. Meteorol.
Soc., 78, 2771-2777, 1997.
for regression
retrievaland space-timeinterpolationof stratospherictemperaturefrom satellitemeasurements,Q. J. R. Me- Wang,B., Replyto comment
by B. Wangon"Anair-seainteraction
teorol. Soc., 116, 1449-1470, 1990.
modelof intraseasonal
oscillations
in thetropics",
J.Atmos.Sci.,
Jones,R.L. and J.EB. Mitchell, Is watervaporunderstood?,
Nature, 353, 212, 1991.
Kiladis, G.N. and K.M. Weickmann, Circulation anomalies as-
sociatedwith tropicalconvectionduringnorthernwinter,Mon.
Weather Rev., 120, 1900-1923, 1992.
Knutson,T.R., and K.M. Weickmann,30-60 day atmospheric
oscillations: Compositelitecycles of convectionand circulation
anomalies,Mon. WeatherRev., 115, 1407-1437, 1987.
Lau, K.M. andP.H. Chan,The 40-50 dayoscillationandENSO: A
new perspective,Bull. Am. Meteorol. Soc.,67, 533-534, 1986.
Lindzen,R.S., Somecoolnessconcerningglobalwarming,Bull.
44, 3521-3525, 1988.
Wang,R., K. Fraedrich,
andS.Pawson,
Phase-space
characteristics
of thetropicalstratospheric
quasi-biennial
oscillation,J. Atmos.
Sci., 52, 4482-4500, 1995.
Waters,J.W.,Microwave
LimbSounding,
in Atmospheric
Remote
Sensingby MicrowaveRadiometry,editedby M.A. Janssen,
383-496,JohnWiley, New York, 1993.
Weare,B.C., andJ.S.Nasstrom,
Examples
of extended
empirical
orthogonalfunctionanalysis,Mon. WeatherRev.,110, 481-485,
1982.
Zhang,C., Atmospheric
intraseasonal
variabilityat the surfacein
Am. Meteorol. Soc., 71,288-299, 1990.
the tropicalwesternPacificOcean,J. Atmos.Sci., 53, 739-756,
1996.
Madden,R.A. andP.R.Julian,Detectionof a 40-50 day oscillation
in the zonal wind in the tropicalPacific,J. Atmos. Sci., 28, 702- Zhang,C., andH.H. Hendon,Propagating
andstanding
compo708, 1971.
nentsof the intraseasonal
oscillationin tropicalconvection,J.
Atmos. Sci., 54, 741-752, 1997.
Matthews, A.J., B.J. Hoskins, J.M. Slingo, and M. Blackburn,
seesawbetween
Developmentof convectionalongthe SPCZ within a Madden- Zhu, B., andB. Wang,The 30-60 dayconvection
Julian oscillation,Q. J. R. Meteorol. Soc., 122,669-688, 1996.
the tropicalIndian and westernPacificOceans,J. Atmos. Sci.,
Newell,R.E., Y. Zhu, W.G. Read,andJ.W.Waters,Relationship
betweentropicaluppertropospheric
moistureand easterntrop-
50, 184-199, 1993.
ical Pacific sea surfacetemperatureat seasonaland interannual
H. L. ClarkandR. S. Harwood,
Department
of Meteorology,
timescales,Geophys.Res.Lett., 24, 25-28, 1997.
University
of Edinburgh,
WestMainsRoad,Edinburgh,
U.K. EH8
Newell, R.E., Y. Zhu, E.V. Browell, W.G. Read, and J.W. Waters, 3JZ. (e-mail: H.Clark@ed.ac.uk;r.harwood@ed.ac.uk
Walkercirculationandtropicaluppertropospheric
watervapor,
P. W. Mote, NorthwestResearchAssociates,
P.O. Box 3027,
J. Geophys.Res., 101, 1961-1974, 1996.
Bellevue,WA 98009. (e-mail:mote@nwra.
com)
Philander,S.G.H., D. Gu, D. Halpern,G. Lambert,N.-C. Lau T.
W. G. Read,JetPropulsion
Laboratory,
4800 Oak GroveDrive,
Li, andR.C. Pacanowski,Why the ITCZ is mostlynorthof the Pasadena,
CA, 91109-8099.
(e-mail:bill@mls.jpl.nasa.gov)
equator,J. of Clim., 9, 2958-2972, 1996.
Read, W.G., J.W. Waters,D.A. Flower, L. Froidevaux,R.E Jarnot, (Received
January28, 1998' revisedJuly17, 1998'
D.L. Hartmann,R.S. Harwood,andR.B. Rood, Upper tropo- acceptedAugust14, 1998.)