JOURNAL OF GEOPHYSICAL
RESEARCH,
VOL. 105, NO. D13, PAGES 17,457-17,470, JULY 16, 2000
Intraseasonal variations of water vapor in the tropical upper troposphere and tropopause region 1 Robert S. Philip W. Mote,1'2HannahL. Clark,3 Timothy J. Dunkerton, Harwood, 3 and HughC. Pumphrey 3
Abstract. We showthe signatureof the tropicalintraseasonal oscillation(TIO) in upper troposphericmoistureand dynamical fields, roughly between200 and 100 hPa. Relationshipsamongthesefields are examinedusinglag-correlationanalysis
and usingmultivariateextendedempiricalorthogonalfunctions(MEEOFs), which maximize the sharedexplainedvarianceamong severalfields for both spatial and temporal variations. The MEEOFs show that all of the fields respond to the TIO and that the TIO is the dominant factor influencingeach of the fields on these timescales.Convectionassociatedwith the TIO moistensthe upper troposphereup to about 150 hPa, as expected; the behavior at 100 hPa is more complex. Over the longitude range where the TIO is associatedwith convection,roughly 60ø-180øE, 100-hPa temperature and water vapor tend to be reduced above convectionon TIO timescales. East of 180ø, though, the temperature and water vapor variations at 100 hPa become decoupled. The water vapor variations, like those of 200-hPa velocity potential, appear to speedup at about 180øE. 1.
Introduction
reviewarticleof MaddenandJulian[1994]).SinceMadden and Julian [1972]pointedout the signatureof the
In recent years, becauseof its importance as a green- TIO on 100-hPa temperature, less attention has been house gas, water vapor in the upper troposphere has paid to the upper troposphereand tropopauseregion been the focus of much interest. This interest is driven than to the lower troposphere. Deep convection, which by the need to reduce uncertainty in one of the great reaches the upper troposphere, is often identified usunknownsin our understandingof Earth's climate and ing outgoir•g longwave radiation (OLR)andhasbeen of the possiblerole of water vapor in a changingclimate, used in several studies of the TIO [e.g., Kiladis and and this interest is also stimulated by a recent explosion Weickmann, 1992; Wheeler and Kiladis, 1999]. Brown in the number and coverageof high-quality measureand Zhang [1997] showed that high-level cloudiness and ments of water vapor in the upper troposphere. Chief humidity throughout the troposphere had fundamenamongtheseis the MicrowaveLimb Sounder(MLS) tally different characteristicsduring the moist and dry
on the Upper AtmosphereResearchSatellite (UARS),
which has proven to be a valuable resource in filling phasesof the TIO. Recently, Clark e! al• [1998]examined longitude-time sectionsof MLS water vapor in out our knowledgeof water vapor in the tropical upper the upper troposphere at tropical latitudes and identitroposphere[e.g.,Readet al., 1995;Elsonet al., 1996]. fied eastward moving moist and dry features that varA curiousfeature of tropical climate is the appearance ied on intraseasonal timescales. Spectral characterisof coherent, eastward moving variations in winds and diagramsand convection with a timescale of 30-60 days; this feature tics obtainedfrom wavenumber-frequency from analysis using extended empirical orthogonal funcis known as the Madden-Julian oscillation or the trop-
ical intraseasonal oscillation(TIO) and has been doc- tions(EEOFs) revealeda zonallysymmetriccomponent at 70 days and an eastward traveling component at umentedin numerousmeteorological studies(see the
30-60 days. The spectral characteristics in time and spacewere consistentwith those expectedfor the TIO. •Northwest ResearchAssociates,Bellevue,Washington Enhancedwater vapor coincidedwith strong 850-hPa 2Also at Climate Impacts Group, Joint Institute for the westerly wind bursts from the Indian Ocean to the westStudy of the Atmosphere and Oceans, University of Washern Pacific, which are known to play an important role ington, Seattic. in TIO events and possibly also in the initiation of E1 3Department of Meteorology, University of Edictburgh, UK
Copyright2000 by the AmericanGeophysicalUnion. Paper number 2000JD900158.
0148-0227/00/2000JD90015850 9.00
Nifio events[e.g.,McPhaden,1999]. Clark e! al. [1998]focusedon one particular moist phase of the TIO in early 1992 and examined sequences of 5-day averagedfields of water vapor and vertical very. The water vapor 'life cycle' was characteristic 17,457
17,458
MOTE
ET AL.: INTRASEASONAL
VARIATIONS
IN THE UPPER
TROPOSPHERE
of TIO events and compared well with the life cycle in OLR described in many studies. However, Clark et al.
now available, and V490 is of higher quality than either
This paper explores and quantifies the relationships among upper troposphericwater vapor, meteorological fields, and OLR, the last of which providesa good indication of convectiveactivity. We also investigatethe extent to which the TIO may affect conditions at 100 hPa near the tropical tropopause, where Mote et al.
the highestlevelfor which MLS upper troposphericdata are retrieved, and found the V422 data again very simi-
V422 or V5 (W. Read,personalcommunication, 2000), [1998]did not quantifythe connections betweenwater but the variations in V422 and V490 are very similar vapor and other fields, nor did they show how water (correlation0.95 for daily variationsin the 0ø-10øSlatvapor was linked to OLR or other measuresof the TIO. itude band). We alsoexaminedvariationsat 147 hPa,
lar to V490 (correlation0.94). Onedrawback with using V490 and V5 for intraseasonal
studies is that the UARS
yaw cycle,whoseperiod(averageabout 36 days)falls within the range of periodsof interest here, is more ap-
[1998]noteda 30- to 60-dayspectralpeakin water va- parent in these later versions. por measured by MLS. We look at both the life cycle The other MLS radiometer that retrieveswater vapor and the space-time variations of the TIO, objectively usesthe 183-GHz band, which is most sensitiveto walinking the variations among severalfields using multi- ter vapor in the stratosphere and mesosphere. A non-
variateextendedempiricalorthogonalfunctions(MEE- linear retrieval, describedby Pumphrey[1999],offers OFs), which are explainedin section4. MEEOFs, many advantagesover the standard linear retrieval and which were previously used by Dunkerton and Baldwin has beenusedin a few previousstudies[e.g., Mote et [1995],offeran efficientwayto studycoherentvariations al., 1996, 1998]. It extendsthe range of the retrieval amongseveralfields,whereasother methods(e.g., co- down to 100 hPa with a higher vertical resolutionthan herenceor singularvaluedecomposition) are usefulfor the standard retrieval; the next highest level is at 68 linking variations between only two fields.
hPa. In this paper we use the 100-hPa data from the
2.
nonlinearretrieval(V0104),whichexhibitvariationson intraseasonal timescales[Mote et al., 1998].
Data
The 100-hPa data have someshortcomingsthat must The meteorologicaldata used in our analysescome be discussed.First, the retrieval reports that the data from the European Centre for Medium-Range Weather are often strongly influencedby the a priori profiles,so Forecasts (ECMWF) Reanalyses (ERA)[Gibsonet al., that the reported variations are probably smaller than 1997]. Data are availableat 2.5øx2.5ø horizontalres- the observedspatial variations. Despite the weaknessof olution on several standard pressure surfaces. Fields the signal, the variations that are derived are broadly used here are 100-hPa temperature and 200-hPa veloc- consistentwith both expected seasonalvariations in waity potential, which is derived from the wind field. ter vapor[Mote et al., 1996]and higher-frequency variUpper tropospheric water vapor data come from the ations [Mote et al., 1998]. The relianceon monthlyMicrowaveLimb Sounder(MLS), an instrumenton mean, zonal-meanclimatology,however,meansthat the the UpperAtmosphere Research Satellite(UARS). The retrieved water vapor values jump abruptly from the UARS is in a polar orbit with an inclination of 570 to end of one calendarmonth to the beginningof the next. the equator [Reber,1993],making 15 orbits eachday, It is straightforwardto removethis effectby subtracting with adjacent orbits being separated by about 2670 km the zonal mean, as was done in the previous studies. at the equator. The satellite performs a yaw maneuver Second,the vertical resolutionat 100 hPa is nearly 4 about every 36 days, but the tropical regionis observed km, so one might assumethat the retrieval of 100-hPa daily, enabling a nearly continuous time series to be water vapor in the tropicsfeelssomeinfluencefrom upconstructed. The MLS instrument is described in more per troposphericwater vapor. In fact, the upper trodetail by Barath et al. [1993],and the measurement posphericwater vapor probably has little influence on
techniqueis describedby Waters[1993]. Two MLS radiometers are used for measuring water vapor, correspondingto spectral bands at 205 and 183 GHz.
The
205-GHz
radiometer
is sensitive to wa-
the retrieved100-hPadata, for two reasons.(1) The retrieval codeignoresradianceswith a tangent pressure below about 100 hPa, with the result that the averaging kernel for 100 hPa actually peaks near 68 hPa
ter vapor in the upper troposphere when concentrations
[Pumphrey, 1999],and (2) the retrievalhasto fit radiare in the range of 100-300 ppmv [Readet al., 1995]. ancesfrom severalcloselyspacedtangent heightsbe-
The best sensitivity occurs when the water vapor concentration is about 150 ppmv; of the standard UARS
tween
100 and 46 hPa.
While
the bottom
one or two
may contain radiance from the upper troposphere,the
pressurelevels,it is at 215 hPa (•12 km at low lati- next two or three will contain less of this. The retrieval tudes) that this concentration mostoftenoccurs.For has to fit them all, which it can do best by choosing consistency with Clark et al. [1998],who usedVersion a profile that is not influencedmuch by the upper tro422 data, and (moreimportant)becauseV422 remains posphere. In short, the 100 hPa water vapor product the only version for which a validation paper has been
is influencedprimarily by water vapor in a layer in the published[Readet al., 1995],we useV422 in this paper lower stratosphereand not much by water vapor in the aswell. Newerversions(V490 andV5) of MLS data are upper troposphere.
MOTE
ET AL.'
INTRASEASONAL
VARIATIONS
IN THE
UPPER
TROPOSPHERE
17,459
Outgoinglongwaveradiation(OLR) data comefrom shown).That is, large-scale divergence (generallyindicated by negativevelocitypotential)is locally associ-
National Oceanicand AtmosphericAdministration polar orbiting satellitesand have been subjectedto quality control and then interpolated in time and spaceas
ated with enhancedwater vapor, and convergenceis locally associatedwith reducedwater vapor. Sincelargedescribed by LiebmannandSmith[1996].The interpo- scale upper level divergenceis broadly associatedwith lated data are available as daily averageson a 2.5øx convection,this negative correlation reflects the well2.50 latitude-longitudegrid and have no missingvalues. known fact that convectiontends to moisten the upper Daily ERA and OLR data are availableover a period troposphere(as wasemphasized in Figure 1). of recordspanningdecades,but for this study we confine Many studies[e.g., Hendon and Salby, 1994; Knutattentionto a periodof 19 months(whichincludetwo sonand Weickmann,1987;MatthewsandKiladis, 1999] northernwinters) when the 183-GHz radiometerwas have investigated the life cycle of the TIO through operating: October 1, 1991, to April 30, 1993. This spatial regressionmaps of dynamical fields and OLR. time period, which includes the Tropical Ocean-Global In a similar manner, we use a reference time series Atmosphere/Coupled Ocean-Atmosphere Response Ex- of 200-hPa ECMWF velocity potential at 11.25øS and periment(TOGA/COARE) IntensiveObserving Period, 168.75øE, the point at which it has maximum variance, is fairly representativeof the climatologicalTIO behav- to calculate one-point correlation maps of water vapor ior [Chen and Yanai, 2000]. For the time-longitude at variouslags (Plate 1). The correlationsare calcusectionsused in the EEOF and MEEOF analysis, we lated using data from December 1, 1991, until March averagedata between 10øSand the equator, taking ad- 31, 1992, a period when the TIO wasstrong. A positive vantage of the large spatial scale of the TIO and the time lag means the water vapor field lags the velocity coherenceof features acrossthis latitude range on this potential field, and a negative time lag meansthe water timescale. MLS profiles between 10øS and the equator vapor field leads the velocity potential field. In contrast to Figure 1, evidenceof eastward propaare binned in 240 longitude bins, consistentwith the gation in Plate i is more subtle. A dry feature (negalarge-scalenature of the TIO. tive correlation)can be seenmovingeastwardthrough the referencepoint from about day-15 to day +15. The 3. Spatial Relationships Among Fields plate revealstwo noteworthy aspectsof the water vapor 3.1. Outgoing Longwave Radiation responseto the TIO' first• the spatial pattern, and secClark et al. [1998]studieda sequence of 5-day aver- ond, a strong projection onto a 50-day timescale. The aged maps of water vapor, taken during a TIO event spatial pattern associatedwith maximum velocity poto between December 1991 and February 1992. The se- tential (lag 0) at the referencepoint (corresponding maximum convergence and minimum convection) conquence corresponded to one of the eastward moving moist events seen in the longitude-time section at 10øS sists of a large area of reduced water vapor surround(their Plate l d). The water vaporfield was compared ing the referencepoint and extending into the Northern with vertical velocity from ECMWF reanalysis to iden- Hemisphere. Ringing this is an arc of enhanced water tify regions of convection. Strong upward motion in vapor extending from the Bay of Bengal to Hawaii. At the middle troposphere coincidedwith enhanced water lags of +25 days the pattern is almost the same as at lag 0, only in reverse, strongly pointing to a 50-day vapor in the upper troposphere. Here we compare water vapor with outgoing long- timescale. These features suggestthat the TIO influwave radiation for the same sequenceof days. Figure 1 showsmixing ratios of water vapor at 215 hPa which are greater than 170 pprnv, and OLR emitting at lessthan
ences the circulation
and moisture
distribution
of the
upper troposphere over a wide area and that the re-
sponseis largely through a Hadley-Walker type circula225W m-a the samethresholdvalueusedby Matthews tion with subtropical subsidencein the Northern Hemiet al. [1996]to indicateenhancedconvectionduringthe sphereringing tropical convectionat, and just south of, TIO event of March 1988. It is clear from this figure the equator. that the water vapor and OLR fields are similar, and 4. EEOF and MEEOF Analysis that convectionmoistens the upper troposphere within Procedure a fairly large area around the convection. It is also clear that some of these features propagate eastward, In the previoussectionwe used maps and correlations for example, the convectiveactivity north of Australia to demonstrate the relationships between two fields. on January 1 that travels east of the date line and then However, consideringmore than two fields becomesexdies out by February 15. Note also that the features tremely difficult, and the spatial detail in the maps can span a wide range of latitudes but are generally cen- obscurethe basic relationships. In order to simplify the tered south of the equator. TIO-related fluctuations in the fields under considera3.2. Velocity Potential and Water Vapor
tion here(watervaporat two levels,temperature,OLR, and velocitypotential), we describea techniquecalled
In general, at most tropical locations, water vapor MEEOF analysis, an extension of the commonly used
and 200-hPavelocitypotentialare anticorrelated(not empiricalorthogonalfunction (EOF) analysis. EOF
17,460
MOTE ET AL.' INTRASEASONAL
VARIATIONS
1991 Dec 17
....:...... •&
IN THE UPPER TROPOSPHERE 1992 Jan 21
.•
1991 Dec 22
1992 Jan 26
1991 Dec 27
1992 Jan 31
1992 Jan 1
1992 Feb 5
1992 Jan 6
1992 Feb 10
1992 Jan 11
1992 Feb 15
II'll I
1992 Jan 16
Figure 1. Five-day averagesof water vapor mixing ratios at 215 hPa greater than 170 ppmv
(shaded regions) andofoutgoing longwave radiation (OLR)lessthan225W m-2 (solidregions). Dates indicate the middle of the pentad.
analysisis usefulfor finding coherentspatialpatternsin noisy data. In this section we explain extended EOFs
(EEOFs), whichareusefulfor identifyingcoherentpatternsof spatiotemporalvariations,and MEEOFs (multivariate EEOFs), which are usefulfor identifyingco-
x(A= 1,t - NAt), x(A = 1,t - (N - 1)At), ..., x(A = 1,t), ...x(A= 1,t + NAt), x(A=2,t-NAt),..., =
t + vzxt).
herent spatiotemporalvariations amongseveralfields.
are then calculated,leadingto a square EEOFs [e.g., WeareandNasstrom,1982;Clarket al., The covariances matrix with dimension L(2N + 1). EOFs are 1998]for longitude-timesectionsx(•,t) are calculated covariance by first forming a vector at eachtime t made up of the then calculatedin the usual manner, leading to patterns time seriesx at each of L longitudesand eachof 2N + 1 (EEOFs) that maximizethe explainedspatiotemporal lag times:
variance. The EEOFs calculated here are fairly insen-
MOTE
ET AL.- INTRASEASONAL
VARIATIONS
IN THE
UPPER
TROPOSPHERE
water vapour
-30days
water vapour
5days
water vapour
-25days
water vapour
10doys
water vapour
15days
17,461
--
--
water vapour
-20days
l.lS,• •' water vapour
- 15days .....
iL
water vapour
20days
woter vapour
25days
IL
'55-.I (I all
water vapour
- 10days
.
--
water vapour
-Sdoys
30doys
water vapour
!
f I
;I
ß
,1 , water vapour
-0.50
-0.44
-0.38
-0..52
Odoys
-0.26
-0.20
-0.14
-0.08
-0.02
0.04
0.10
0.16
0.22
0.28
0.34
0.40
Plate 1. One-point correlation maps at various lead and lag times, showing the correlation
betweenMicrowaveLimb Sounder(MLS) water vapor at 215-hPaand a referencetime seriesof 200-hPaEuropeanCentrefor Medium-RangeWeatherForecasts (ECMWF) velocitypotentialat 11.25øS, 168.75øE, indicated by the large asterisks.
17,462
MOTE
ET AL.- INTRASEASONAL
VARIATIONS
IN THE
sitiveto the choiceof lag interval(the resultshereused an intervalof 5 days) and to longerwindows(30 days here). We presentthe resultsof the EEOF analysisin
TROPOSPHERE
EEOF
1
2
..... '"'"":'"•" :•"• .•...&,..:•:• :; •;•...:
10
10 120 180 240 300 360
.•,'•:•. •_.,:•.t.•...:,,.,, •-,.
0
60
120 180 240 300 360
Q•oo -30 ...... -20
-10 0 10
Relationship
Betwee.-:. MEEOFs
and
EEOFs
30
120 180 240 300 360
By construction, MEEOFs maximize the combined variance of all the fields. Is it possible, however, that the MEEOFs are artificially selecting variations that are not the dominant
variations
for one or more
of the
fields? We can checkfor such artificial selection by com-
paring the EEOF• with the EEOFi, in particular calculating the correlation in the longitude-lag space of EEOFs. If they are very similar, then the variation identified by the MEEOFs is the dominant variation For velocity potential, the leadingpair of EEOF• looks
very similarto the leadingpair of EEOFi (Figure 2),
20
Longitude
6.1.
-30 ,,•. "-•-.--:•---.•..-:. *-'O•:::•,: , •,•-•:•g-•':•
......
60
For all variables the first two EEOF• clearly represent a conjugatepair. That is, they have similar wavy spatial structure and are about a quarter of a cycle apart. Thereforeeachof the fieldsis strongly affectedby a propagating disturbance; its characteristic timescale is of the order of 40-50 days. The fact that the phases line up so well indicates common causality.
';:• ...... '-':•,•g -2o•--
60
less coherent.
30
60 120 180 240 300 360
-30 *-'•.
height crosssectionsreveals that the change in zonal phase speed is probably related to a change in vertical phasespeed(G. Kiladis, personalcommunication, 2000). A changein phase tilt could explain these changes. For water vapor at 100 hPa, the EEOF• also showsomeeastwardpropagation,though the pattern is
0
60
120 180 240 300 360
Longitude
except that the phasingis somewhatdifferent. Lag correlations of the principal component time series reveal that the first EEOF• is 5 daysaheadof the first EEOFi, with a correlation
Figure 4. Structure of the two leading EEOF,• for each of the five fields, derived for all five fields simultaneously.Left columnshowsfirst MEEOF; right column shows second MEEOF.
between the two time series of 0.94.
Table i shows,for each field, the correlationsbetween the EEOFi and the EEOF,•. The phase shift of the first two EEOF,• for velocity potential reducesthe correlations, but they are fundamentally the same as the EEOFi. For OLR and 215-hPa water vapor, the first two EEOF• are nearly the same as the EEOFi, as in-
dicatedby the fairly high (above0.9) absolutecorrefieldswere all derivedtogether, and together the EEOFs explain the largest fraction of the combined variance,
lations. The variations identified by the MEEOF approach are indeed the dominant story for those fields. but are shown separately. These EEOFs are distinct The third and fourth EEOF,• are quite similar to the from the EEOFs discussedearlier, which were derived third and fourth EEOFi only for velocity potential. For for eachfield independently;like the earlier EEOFs, the 100-hPa temperature, the weak correlations between two shownhere form a conjugatepair. For clarity in the the first two EEOF,• and the first two EEOFi are only subsequentdiscussionwe refer to the EEOFs calculated partly becauseof phaseshifting. Despite the weak reindependently as EEOFi and we refer to the EEOFs exlationship between the EEOF• and EEOFi for temtracted from the MEEOFs as EEOF•. perature, water vapor at 100 hPa has a fairly strong The first two EEOF• for all five variables show evsimilarity between the EEOF• and the EEOFi, indiidence of eastward propagation, and the tropospheric cating that it is better connectedto the tropospheric
fields(top three) and 100 hPa temperaturealsoshow fields(aswasapparentin Plate 3).
MOTE
ET AL.'
INTRASEASONAL
VARIATIONS
EEOF
1
3
4
17,467
ably indicating that the V422 retrieval of 147-hPa data
reliestoo heavilyon valuesat 215 hPa), 0.62 for V490,
•(.200
1
0.76
0.63
0.16
0.03
2
0.64
-0.76
0.00
-0.14
3
0.26
0.16
-0.90
0.13
4
-0.10
0.30
-0.12
-0.90
1
0.98
2
-0.19
3
0.11
4
0.07
0.17
OutgoingLongwaveRadiation -0.18
TROPOSPHERE
ter vapordata havecorrelations of 0.83for V422 (prob-
Number
2
UPPER
the data as described above, the 147- and 215-hPa wa-
Table 1. CorrelationsBetweenEEOFmand EEOFi EEOF
IN THE
0.07
-0.02
-0.97
-0.01
-0.11
-0.15
-0.73
0.56
-0.62
-0.68 -0.24
Q215 1
0.90
0.02
0.24
2
-0.01
-0.93
0.23
0.16
3
0.17
-0.22
-0.40
-0.34
4
0.14
0.15
0.11
0.37
Tjoo
and 0.54 for V5. Prominent eastward moving features in the 215-hPa data also appear in the 147-hPa data. (• ......... tl•r it •pp•ar• that. hc•t.hin the timo moan and on intraseasonaltimescales, there is a level somewhere
between
147 and
100 hPa
at which
the
water
vapor response to convection changessign. That is, convection moistens the upper troposphere up to that level and dries the upper troposphere and tropopause region above that level. 6.3. The 100-hPa Vapor
Temperature
and Water
It is surprising that the two fields that are least cor-
1
0.43
-0.57
0.01
-0.66
related(both the EEOFsandthe originaldata) are the
2
0.52
0.38
0.70
-0.06
two fields at 100 hPa.
3
-0.78
0.26
-0.08
-0.26
related with those of X200. This is partly because of
4
0.40
0.71
-0.13
0.04
QIoo 1
-0.83
-0.53
-u.u2
0.09
2
0.53
-0.81
0.17
-0.02
3
0.33
0.02
-0,40
0.71
4
0.21
-0.33
-0.67
-0.40
EEOF, extendedempirical orthogonalfunction.
6.2. Relationships Among Fields From Figure 4, it is apparent that some of the fields have fairly coherent relationships with other fields. To quantify these relationships, we calculate the correlation betweenthe leading EEOFs of eachfield with those of the other
fields.
Table
2 shows these correlations.
Water vapor at 215 hPa is 1800 out of phase with all the other fields, as indicated by the negative correlations in its rows and columns: In the convective phase
of the TIO, velocitypotentialis low, OLR is low (high clouds),215-hPawater vaporis high, 100-hPatemperature is low, and 100-hPa water vapor is low. Velocity potential generally has the highest correlations with other fields, and 100-hPa temperature has the lowest. Velocity potential, which has been used in many dynamical studies of the TIO, is evidently the best indicator variable for these other upper tropospheric fields especially when considering the connec-
the different
scale vertical
motions.
3. The characteristics (e.g.,verticalresolution)of the two data
sets are different.
The first possibility seemsplausible from looking at Table
2.
Correlations Between EEOF,•
of Different
Fields Field
X2oo EEOF
OLR
OLR
Q2xs
T•oo
Qloo
0.82
-0.76
0.47
0.71
1.00
-0.76
0.42
0.47
-0.07
-0.57
1.00
0.28
1
1.00
Q•5
1.00
T•oo
Q•oo
mean distribution of upper tropospheric water vapor
EEOF
(300 and 150hPa) had maximain the tropicalconvec-
X2oo
confirm this antiphase relationship; spatially binning
of the two fields in the Western
ent.
The fact that water vapor at 215 and 100 hPa behaves oppositely in responseto the TIO is consistent with
mumin the samelocations.(The SAGEII data arevery sensitiveto clouds,unlikeMLS.) MLS data at 147hPa
are each cor-
2. The vertical gradients of the two fields are differ-
X•oo
tive areas, while the 100-hPa distribution had a mini-
EEOFs
Hemisphere, where water vapor behavesmuch like X200. As a consequenceof this behavior in the Western Hemisphere, the TIO signature is more apparent in 100-hPa water vapor than in 100-hPa temperature. There are several possible factors that may combine to explain the difference between the temperature and water vapor fields. 1. The two fields are related through different processesin the Eastern and Western Hemispheres. For example, convection in the Eastern Hemisphere may produce local saturation and drying, while free Kelvin waves in the Western Hemispheremay produce large-
tion to 100 hPa.
Stratospheric Aerosoland Gas Experiment(SAGE) II climatology[Rind et al., 1993],whichshowedthat the
behavior
Yet their
OLR
Q2•s Txoo
Qxoo
1.00 2
1.00
0.81
-0.75
0.45
1.00
-0.76
0.38
0.46
-0.06
-0.57
1.00
0.23
1.00
0.71
1.00
17,468
MOTE
ET AL.: INTRASEASONAL
VARIATIONS
IN THE UPPER TROPOSPHERE
Plate 3, which shows that temperature and water va- temperatureat 100hPa), althoughcoherentwith OLR por tend to vary in the same sensebetween about 60o and Q2•5 in the Eastern Hemisphere, have very differand 120o. Aircraft observationsof tropical water vapor ent behavior east of 210øE. Variations in velocity poshow saturation more commonly in convectiveregions tential show the familiar increasein phase speed, while than in the tropicaleasternPacific[Kelly et al., 1993]. the temperature at 100 hPa showsan odd reduction in Consequently,the vertical gradientsare differentin the horizontal phasespeed,which may in fact be related to two regionstoo, which could provide part of the expla- a changein the vertical phase speed. nation for the difference. Water vapor at 100 hPa clearly respondsto the TIO as well, though the variations are rather subtle and at The third possibility is also plausible. The two data first glance unrelated to the variations in temperature setshave rather differentcharacteristics;one is remotely at the same level. Its similarity to velocity potential in sensed,and the other is generatedby global reanalyses the Western Hemispheresuggeststhat there, at least, using a general circulation model. The issueof vertical the origin of the anomaliesis purely dynamical and not resolution is particularly important. As noted in secinduced by radiative anomaliesabovecold cloud tops. tion 2, the 100-hPa water vapor measurement actually One possible explanation for the differencebetweenthe representsa layer probably some 4 km thick, mostly at water vapor and temperature fields is simply that the altitudes above 100 hPa. By contrast, the reanalyses water vapor data are more smeared out in the vertical have finer vertical resolution and do almost as well as
and that the verticalwavelengthof the TIO signalnear ationsin temperaturenear the tropopause(G. Kiladis, the tropopause is substantially different in the Western personalcommunication, 1999, 2000). There is ample Hemispherethan in the Eastern Hemisphere. high-resolutionradiosondedata at revealingTIO vari-
Our analysis demonstrates,for the first time on a global scale,that the responseof 100-hPa water vapor et al. [1998]stated that the spectralpower of water to the TIO is out of phasewith the water vaporresponse vapor variations in the 30- to 70- day band drops by at 215 and 147 hPa. While the convectivelyactivephase an order of magnitude between 100 and 68 hPa, the moistensthe upper troposphere,the tropopauseregion next higher MLS level. Analysis of radiosondetemper- becomesdryer. The tendencyof the tropopauseto beaturesby G. Kiladis (personalcommunication, 1999) come higher, colder, and dryer during active convection is well known from observationalcampaignsusreveals that the vertical scale of the TIO is so shallow in the lower stratospherethat the phaseundergoesa com- ing radiosondesequippedwith frost-pointhygrometers plete reversal between 100 and 70 hPa, in agreement [V6'melet al., 1995]. Unclearfrom suchlocal obserwith the earlierstudyof Hendonand œiebmann[1990]. vations, however, is the horizontal extent of the water For disturbancesthat are sharply attenuated above the vapor responsein relation to convection.Tropical contropopause and for disturbances with shallow vertical vection is organized within cloud clusters and superstructure in the vicinity of 100 hPa, the variations in clusters[Hendonand Liebmann,1994; Dunkertonand Crum, 1995],but circulationanomaliesassociated with temperature could be well representedby ERA data but the intraseasonal oscillation are much broader in lonthe variations in water vapor could be strongly diminished in the MLS retrieval. Taken together, the attenu- gitudinal extent. Our water vapor resultsat 100 hPa, although consistentwith ground-basedobservationsof ation and small vertical scale of the TIO combined with tropopause behavior, do not adequately resolvethe horthe poor vertical resolution of MLS in the lower stratoizontal structure of mesoscalecloud clusters;therefore sphere imply that the true variation in water vapor at we cannot see the local responseto tropical convection. 100 hPa associatedwith the TIO could be substantially Instead, these results suggestthat water vapor anomastronger than indicated by the MLS instrument. As for lies visible from satellite extend over a wide range of the behavior of 100-hPa temperature, it is possiblethat longitudes, including regions where the TIO is decouthe discrepancyin the Western Hemisphere is related evidence that the TIO is both sharply attenuated and of small vertical scale in the lower stratosphere. Mote
to the paucity of observationsthat could constrain the ECMWF analysis. 7.
Discussion
and
Conclusions
We have shownthat the well-known tropical intraseasonal oscillationhas a fairly robust and important signature on upper tropospheric water vapor at 215 hPa
pledfromconvection (i.e.,in the WesternHemisphere). Interpretation of the 100-hPa water vapor data is complicated by uncertainty as to which levels are actually responsiblefor this signaland to what extent the data at this level are spuriouslyaffectedby water vapor
from a slightlyhigherlevel (68 hPa). Until this un-
certainty is resolved,and a connectionto nearby temperature anomalies is made, we cannot be certain of (and, not shown,at 147 hPa) and that the TIO signa- the causeof the 100-hPawater vapor signal,whether it ture in OLR is very coherentwith the signaturein wa- involvesupward displacement of the water vaporgraditer vapor. Convection associated with the TIO moist- ent, horizontaltransport, changeof saturationmixing ens the upper troposphere,an effect that operates only ratio, or some combination of these effects. where a strong signature of the TIO in convectionis Clearly,the tropopause regionis affectedby the TIO, found: between about 60øE and 210øE. By contrast, both directlyby convection andindirectlyby the slow,
the dynamicalfields(velocitypotentialat 200 hPa and shallow Kelvin waves that it generatesin the lower
MOTE
ET AL.:
INTRASEASONAL
VARIATIONS
IN THE
UPPER
TROPOSPHERE
17,469
tropical intraseasonal oscillation, J. Geophys.Res., 100, stratosphere. More work is needed to understand how 25,781-25,790, 1995. these phenomenainteract and how they vary zonally, Elson, L.S., W.G. Read, J.W. Waters, P.W. Mote, J.S. Kinespecially in the eastern Pacific. It would be useful nersley,and R.S. Harwood, Space-timevariations in water to checkthese results against station data, but in the vapor as observedby the UARS MicrowaveLimb Sounder, eastern Pacific there are hardly any stations. Satellite J. Geophys. Res., 101, 9001-9015, 1996. measurementsof water vapor there are plentiful, but Gibson, J.K., P. Kallberg, S. Uppala, A. Hernandez, A. instruments
with
better
vertical
resolution
than MLS
in the vicinityof the tropopause (e.g., the HalogenOccultationExperiment,alsoaboardUARS) do not have adequatetemporal coverageto study the TIO. Two new instruments currently scheduledfor launch on the EOS CHEM satellite in 2002 may shed light on these questions. The High-Resolution Dynamics
Limb Sounder(HIRDLS) is an infraredinstrumentthat will
achieve
1-kin
vertical
resolution
and is intended
to be suitable for studying the lower stratosphere and tropopauseregion. "Observationsof the lower stratosphere are improved through the use of special nar-
row and more transparentspectral channels"(from http://www.eos.ucaroedu/hirdls/home.html). The EOS MLS [Waters et al., 1999] is an improvedversionof
Nomura, and E. Serrano, ERA description, Re-analysis Project Rep. Set. 1, Eur. Cent. for Medium-Range Weather Forecasts,Reading, England, 1997. Hendon, H.H., and J. Glick, Intraseasonal air-sea interaction in the tropical Indian and Pacific oceans, J. Clim., 10, 647-661, 1997.
Hendon,H.H., and B. Liebmann,The intraseasonal(30-50 day) oscillationof the Australian summer monsoon,J. Atmos. $ci., 4{7, 2909-2923, 1990. Hendon, H.H., and B. Liebmann, Organization of convection within the Madden-Julian oscillation, J. Geophys. Res., 99, 8073-8083, 1994. Hendon, H.H., and M.L. Salby, The life-cycle of the MaddenJulian oscillation, J. Atmos. Sci., 51, 2225-2237, 1994. Kelly, K.K., M.H. Proffitt, K.R. Chan, M. Loewenstein,J.R. Podolske, S.E. Strahan, J.C. Wilson, and D. Kley, Water vapor and cloud water measurementsover Darwin during the STEP 1987 tropical mission, J. Geophys. Res., 98,
8713-8723, 1993. the MLS instrument discussedhere, and has improved Kiladis, G.N., and K.M. Weickmann, Circulation anomalies bandwidth, which should improve the accuracy in the associatedwith tropical convectionduring northern winlower stratosphere and upper troposphere. Its closely ter, Mort. Weather Rev., 120, 1900-1923, 1992. spaced scan pattern should give improved vertical res- Knutson, T.R., and K.M. Weickmann, 30-60 day atmo-
olution of 1.5-2 km. Both HIRDLS
and EOS MLS will
spheric oscillations: Composite life cycles of convection and circulation anomalies, Mort. Weather Rev., 115, 1407-
have more frequent scans,giving better horizontal res1437, 1987. olution as well. Together, these instruments will reveal Liebmann, B., and C.A. Smith, Description of a complete in better detail the TIO responseof water vapor in the (interpolated)outgoinglongwaveradiation dataset, Bull. important layer between 200 and 100 hPa. Am. Meteorol. Soc., 77, 1275-1277, 1996. Acknowledgments. We are grateful to Brant Liebmann for providing the OLR data and to Chidong Zhang for helpful comments on the manuscript. George Kiladis very kindly showedus unpublished work that was very helpful in our study. The water vapor data were produced at the Jet Propulsion Laboratory, California Institute of Technology, under contract with NASA and funded through its UARS Project. The work was supported by NASA contracts NAS5-98078 and NAS1-96071 and by NERC in the UK.
References Barath, F., et al., The Upper Atmosphere Research Satellite Microwave Limb Sounder instrument, J. Geophys. Res.,
Madden, R.A., and P.R. Julian, Detection of a 40-50 day oscillation in the zonal wind in the tropical Pacific, J. Atmos. Sci., œ8,702-708, 1971. Madden, R.A., and P.R. Julian, Observations of the 40-50 day tropical oscillation--A review, Mort. Weather Rev., 1œœ,814-837, 1994.
Matthews, A.J., and G.N. Kiladis, The tropical-extratropical interaction between high frequency transients and the Madden-Julian oscillation, Mort. Weather Rev., 1œ7,661677, 1999.
Matthews, A.J., B.J. Hoskins, J.M. Slingo, and M. Blackburn, Development of convectionalong the SPCZ within a Madden-Julian oscillation, Q. J. R. Meteorol. Soc., 1œœ, 669-688,
1996.
McPhaden, M.J., Genesis and evolution of the 1997-98 E1 98, 10,751-10,762, 1993. Ni•o, Science, œ83, 950-954, 1999. Brown, R.G., and C. Zhang, Variability of midtropospheric Mote, P.W., K.H. Rosenlof, M.E. Mcintyre, E.S. Carr, J.C. Gille, J.R. Holton, J.S. Kinnersley, H.C. Pumphrey, moisture and its effect on cloud-top height distribution during TOGA COARE, Jo Atmos. Sci., 5•, 2760-2774, J.M. Russell III, and J.W. Waters, An atmospheric tape 1997. recorder: The imprint of tropical tropopause temperaChen, B., and M. Yanai, Comparison of the Madden-Julian tures on stratosphericwater vapor, Y. Ceophys. Res., 101,
oscillation(MJO) duringthe TOGA COARE IOP with a 15-year climatology, J. Geophys.Res., 105, 2139-2149, 2000.
Clark, H.L., R.S. Harwood, P.W. Mote, and W.G. Read, Variability of water vapor in the tropical upper troposphere as measured by the Microwave Limb Sounder on UARS, J. Geophys.Res., 103, 31,695-31,707, 1998. Dunkerton, T.J., and M.P. Baldwin, Observation of 3-6 day meridional wind oscillations over the tropical Pacific, 1973-1992: Horizontal structure and propagation, J. Atmos. Sci., 5œ,1585-1601, 1995. Dunkerton, T.J., and F.X. Crum, Eastwardpropagating•2to 15-day equatorial convection and its relation to the
3989-4006,
1996.
Mote, P.W., T.J. Dunkerton, and H.C. Pumphrey, Subseasonal variations in lower stratospheric water vapor, Geophys. Res. Lett., œ5,2445-2448, 1998. Pumphrey, H.C., Validation of a new prototype water vapor retrieval for the UARS Microwave Limb Sounder, J. Geophys. Res., 10•, 9399-9412, 1999. Read, W.G., J.W. Waters, D.A. Flower, L. Froidevaux, R.F. Jarnot, D.L. Hartmann, R.S. Harwood, and R.B. Rood, Upper tropospheric water vapor from UARS MLS, Bull. Am. Meteorol. Soc., 76, 2381-2389, 1995. Reber, C.A., The Upper Atmosphere Research Satellite
(UARS), Geophys.Res. Left., œ0,1215-1218,1993.
17,470
MOTE ET AL.: INTRASEASONAL
VARIATIONS IN THE UPPER TROPOSPHERE
Rind, D., E.-W. Chiou, W. Chu, S. Oltmans, J. Lerner, J. Larsen, M.P. McCormick, and L. McMaster, Overview of the Stratospheric Aerosol and Gas Experiment II water vapor observations:Method, validation, and data characteristics, J. Geophys. Res., 98, 4835-4856, 1993. VSmel, H., S.J. Oltmans, D. Kley, and P.J. Crutzen, New evidence for the stratospheric dehydration mechanismin the equatorial Pacific, Geophys.Res. Lett., 22, 3235-3238, 1995.
Waters, J.W., Microwave Limb Sounding, in Atmospheric Remote Sensing by Microwave Radiomerry, edited by M.A. Janssen, pp. 383-496, John Wiley, New York, 1993. Waters, J.W., et al., The UARS and EOS Microwave Limb
Wheeler, M., and G.N. Kiladis, Convectively coupled equatorial waves: Analysis of clouds and temperature in the wavenumber-frequencydomain, J. Atmos. Sci., 56, 374399, 1999.
H. L. Clark, R. S. Harwood, and H. C. Pumphrey, Department of Meteorology, University of Edinburgh,
Edinburgh EH9 3Jg, Scotland, UK. (
[email protected];
[email protected];
[email protected]) T. J. Dunkerton and P. W. Mote, Northwest Research Associates, PO Box 3027, Bellevue, WA 98009.
tim@nwra. com) Sounder(MLS) experiments,J. Atmos.Sci., 56, 194-218, (
[email protected], 1999.
Weare, B.C., and J.S. Nasstrom, Examples of extended empirical orthogonal function analysis, Mon. Weather Rev., 110, 481-485, 1982.
(ReceivedAugust24, 1999; revisedFebruary16, 2000; acceptedFebruary28, 2000.)
