Nov 20, 1995 - format as illustrated by the dial in the upper right corner. A vector pointing horizontally to the right denotes a diurnal phase maximum (that is ...
JOURNAL
OF GEOPHYSICAL
RESEARCH,
VOL. 100, NO. Dll,
PAGES 22,999-23,008, NOVEMBER
20, 1995
Daily variability of lower troposphericwinds over the tropical western Pacific David S. Gutzler1 NOAA AeronomyLaboratory,Boulder,Colorado
Leslie M. Hartten Cooperative Institutefor Research in Enviromnental Sciences, Universityof Colorado,Boulder NOAA AeronomyLaboratory,Boulder,Colorado
Abstract. Daily variabilityin lowertropospheric windsoverthe tropical westernPacificis examinedusingdatafrom a networkof wind-profilingradars deployedduringthe TropicalOcean-GlobalAtmosphere CoupledOceanAtmosphere Response Experiment(TOGA COARE). The amplitudeof the diurnalcyclein both u andv at thethewesternmost sites,locatedjustto theeast
of thelargeislandof NewGuinea, isof theorderof 0.5ms-1, significantly largerthanthe diurnalcycleobservedeastof the datelinein previousstudies. The largerdiurnalvariabilityis attributedto thepronounced nocturnal maximumin deepconvectionovertheislandof New Guinea;the influenceof thisdiurnalcirculationextends(with smalleramplitude)to themoredistant profilersites10ø-15ø E of New Guinea.The semidiurnalcycleof the zonalwind is smallerin amplitudebut coherentin phaseat all sites;the meridionalwind exhibitsa negligiblesemidiurnalcycle.Theseresultsare consistentwith the solarsemidiurnaltidal oscillationin pressureandentirelyconsistentwith previousstudiesof easterntropicalPacificwinds.Two of the profilerswere deployedon shipsratherthanislands,andwe find no systematic differences betweenthe shipdataandthe smallislanddataattributableto thecomplete absenceof landat the shipsites. 1. Introduction
Daily variationsarean importantcomponent of coupledoceanatmospherevariability in the tropics,particularly acrossthe westernPacific warm pool. Convectiveactivity and rainfall, whichdisplayprominent diurnalcycles[Hendonand Woodberry, 1993;MapesandHouze, 1993],playa first-orderrole in boththe troposphericheat budget (throughlatent heat release) and the oceanicsalinitybudget(throughinjectionof freshwaterontothe oceansurface). We would expectto find daily variationsin windsassociated with thecirculations of convectivesystems, but suchfluctuations arepoorlyresolvedby operational measurement systems(suchas radiosondes), whichgenerallysampleatmosphericconditions onlyonceor twiceperday. The few observational studies in the literature have been based
iment. Williams et al. [1992] used data from a 50-MHz wind profiler to documentthe diurnal variability of winds above 2.5 km at Christmas
Island
in the central Pacific
Ocean.
Like
most
upperair data from the tropical Pacific, the observationsused for thesestudiescome from island sites,so at least someof the daily variabilitycould be due to local topographiceffects or land-sea heatingcontrasts. A recentstudy of daily variability in surfacewinds east of the dateline,basedon datafrom openoceanmoorings,was published by Deser [1994]. The data used in that study should be truly representativeof the open ocean,but as we will show there are some important differences between the daily variability describedin that studyand the daily variability observedover the western Pacific. The convective conditions east of the dateline,
featuring large sea surface temperature (SST) gradients and on specialobservingcampaigns or on systemsthat furnishmore dominatedon the large-scale by the nearly linear, open ocean frequentdata for limited periodsof time. Hastenrath[1972] IntertropicalConvergenceZone north of the equator,are signidocumented diurnal and semidiurnal fluctuations from several ficantly different from those over the western Pacific, where sitesjustnorthof theequatorin thetropicalwesternPacific,using intraseasonaloscillations and large, scatteredislands provide 4-times-dailyradiosondeascentsduringa specialfield exper- substantialtemporal and spatialorganizationfor deep convective systemsin the absenceof large-scaleSST gradients. Several competingprocessescould be responsiblefor daily variability in the lower troposphereover the tropical western •NowatUniversity ofNewMexico. Pacific. The responseof the wind field to direct solar thermal forcing can be calculatedfrom first principles[ Chapmanand Copyright1995by theAmericanGeophysical Union. Lindzen,1970; Haurwitz and Cowley, 1973; Forbes and Gillette, 1982]. Observationsof the semidiurnal(12-hour) cycle in the zonalwind at sitesin the tropicalPacificeastof the datelinehave Papernumber95JD01879. 0148-0227/95JD-01879505.00 been shown to follow closely the fluctuationsexpected to be 22,999
23,000
GUTZLER AND HARTFEN: DAILY VARIABILITY OF LOW-LEVEL TROPICAL WINDS
associated with the thermallyforcedsemidiurnal pressurewave [Williams et al., 1992; Deser, 1994]. Indirect forcing for a diurnal (24-hour) cycle comes from deep convective activity, which exhibits a prominentnocturnalmaximum in the tropics [Gray and Jacobson, 1977; Hendon and Woodberry, 1993; Janowiaket al., 1994].
The recentCoupledOcean-Atmosphere ResponseExperiment (COARE) in the westerntropicalPacificincludeddeploymentof six continuously operated, 915-MHz boundary layer windprofiling radars[Ecklundet al., 1988; Carter et al., 1995] which allow diurnal and semidiurnalvariationsin lower tropospheric windsto be resolvedmuchmore accuratelythan is possiblewith radiosondeascents. These radars are smaller and operate at a highertYequency thanthe profileron ChristmasIslandutilizedby Williams et al. [1992] and are designedto furnish data in the lowestfew kilometersof the atmosphere. This manuscript describes the diurnal and semidiurnal variations
in zonal and meridional
winds derived
from
lower
Two of the profilers were deployed on Chinese research vesselsduringthe IOP. Thesesystemswere stabilizedand calibrated as describedby Carter et al. [1992]. The ship-based profilesprovide,to ourknowledge,thefirst directobservations of diumal variabilityabovethe oceansurfacenot subjectto contaminationby landeffects(all previousstudiesof diumal variability havingbeenderivedfrom soundingstakenfrom islands). Thus the half-hourly data from the shipsallow us to comparatively
assessthe variabilityobservedat islandsitesfor the possible influenceof topographyor land-seaheatingcontrasts. Time seriesof daily averagedzonal wind from theseprofilers wereshownby Gutzleret al. [ 1994], who usedthe profilerdatato form the lowest 4.1 km of daily zonal wind profiles extending into the stratosphere(radiosonde data from the radiosonde componentof the ISS systemswere usedfor altitudesabove4.1 km). Gutzler et al. [1994], however,useda "quick-look"version of the ISS data, consistingof hourly 5-minute averagestransmitted in real time from each site via satellite over the Global
System. troposphericwind profilersduringthe 4-monthCOARE Intensive Telecommunications For the presentstudywe use an enhancedversionof the ObservingPeriod(IOP). By examiningthe spatialdistribution andverticalstructure of theamplitude andphaseof thediurnal profiler data, derivedfrom spectralmomentsrecordedin situ on and semidiurnal cycles in the winds, we attempt to infer the optical disks and post-processedat the NOAA Aeronomy relative importance of direct (solar) and indirect (convective) Laboratory[Miller and Riddle, 1994; Riddle and Carter, 1994]. forcing,and we commenton the possibleeffectsof diurnal varia- Duringeachhalf-hourperiod,wind profileswerecollectedfor 25 bility on someaspectsof large-scale,coupledocean-atmosphere min at two different verticalresolutions(238 m or 98 m between the centroidsof the volumessampled),yielding up to 48 halfvariabilityin the warm pool region. hour profiles per day. The different vertical resolutionsof the data are denoted"coarse"and "fine", respectively. This pro2. Wind Profiler Data cedureis designedto take into accountthe normal decreasein The 915-MHz profilerswere an integralpart of the networkof atmosphericreflectivity with height; smallervolumesof air can be samplednearthe groundto obtainfiner verticalresolution,and Integrated Sounding Systems (ISSs) [Parsons et al., 1994] resolution aloftallowstheprofileto reachhigheraltitudes deployed to the northeastof the large island of New Guinea coarser In thisstudywe usefine during the COARE IOP (Figure 1). (In this paper, unless thanwouldbe possibleat fine resolution. explicitly statedotherwise,we use the term "New Guinea" in its resolutiondata up to 1.6 km altitude and coarseresolutiondata geographical,not its political, sense. The countryof PapuaNew between 1.4 km and 4 km, with the diurnal and semidiurnal Guinea or PNG includesthe easternhalf of the large island of cyclesfrom the two profilesblendedin theoverlaplayercentered schemealso includesa signiNew Guinea, eastof 141øE, as well as the islandsof New Britain near 1.5 km. The post-processing algorithmfor removingthe effectsof andNew Ireland and many smallerislands. Two of the profilers- ficantlymore sophisticated at Manus and Kavieng-were located on PNG territory, but no falling hydrometeorsfrom the horizontalwinds, and generates profilerswere situatedon the large island of New Guinea.) Four quality control flags for each datum basedon vertical and temchecks[ Riddleand Carter, 1994]. of thesesystemsdefine a quadranglecenteredjust southof the poralconsistency The total periodof recordfor our studyis the 120-dayCOARE equator,boundingthe IntensiveFlux Array (IFA), within which many additionalinstrumentplatformswere deployedduringthe IOP (November1, 1992 to February28,1993), but eachstation's recordis somewhatdifferent,as summarizedin Table 1. The lOP [Webster and Lukas, 1992]. Two additional ISSs were deployedon islandsto the west and east of the IFA. The sites shortestrecordsare from the researchvessels,partially because range in latitude from 1.1øN to 4øS, and in longitude from we excludedperiodsof time whenthe shipswere in transitto and 147.4øE to 166.9øE. More detailed discussion of the locations of from their assignedpositionsin the IFA (seeTable 1 and Figure 1). The shipprofilershave a slightlylower heightrangethan do the siteson islandswill be presentedin section4. the land-basedsystemsbecauseof the effects of ship motion [Carter et al., 1992]. Wind directionsare also more variable than 4N Kapingamarangi
FQ
4S
Nauru
........
.::• Kav•_og'• / IFA"'?,• RV3 (Ship •3)
New;;;;';;"",.., Manus ..,;;'" '7;•:•----•1 ,..... (ShpC1)
'"•...
......i%%2? '"•'•"•d•in "....,::::. i"...... ""'-...i '::2',., "::::::' ,"1ß
i
i
i
i
140E
150E
160E
170E
at land stationsdue to uncertaintiesin ship heading[Riddle and Miller, 1994a,b]. In addition,a faulty chip in the profiler on RfV Kexue (RV3) led to the intermittentappearanceof false echos duringmuchof the IOP. Quality controlflagsandspectralwidths wereusedto screenthe affecteddatain the work presentedhere. The record at Nauru is also shortbecauserecordingproblems causedseveral gaps in the data record and preventedthe application of the initial stagesof the post-processing algorithm for much of the record.
Figure 1. Map of ISS wind profiler sitesduringthe COARE Figure2 showsvectorprofilesof thelOP-meanwindsat each lOP. The Intensive Flux Array is denotedby the dashed of the six sites. The boreal winter season of 1992/1993 was quadrangle.Characteristics of the settinganddataat eachsiteare characterizedby enhancedconvectioncenteredjust southof the summarized in Table 1. equatorto the eastof the IFA, so the windsshownin Figure2 are
GUTZLER
AND HARTFEN:
DAILY
VARIABILITY
OF LOW-LEVEL
TROPICAL
WINDS
23,001
Table 1. Characteristics of the Six ISS 915-MHz ProfilersDeployedfor the COARE IntensiveObservingPeriod Numberof Days Profiler Site
Abbreviation
Kapingamarangi Naum
KAP NAU
Marius
Kavieng RIV Kexue#3 R/V Shiyan#1
Location
Local Time.
Elevation,
Coarsea
Finea
0000 UTC
m ASL
Resolution
Resolution
I.IøN, 154.8øE 0.5øS,166.9øE
1020 1110
3 38
115 86b
I 16 88b
MAN
2. IøS, 147.4øE
0950
4
119
120
KAV RV3 RV 1
2.6øS, 150.8øE -2øS, - 158øE -4øS, - 156øE
1000 1030 1020
4 7 5
116 76c.d 94d.e
80c.d
115
92d,e
COARE IntensiveObservingPeriodof November1, 1992 throughFebruary28, 1993( 120days). The sitesaremappedin Figure1. aMaximumnumberof daysusedto calculatelOP meanhalf-hourlyaverages.The 30-min consensus averageswereretainedonly if they were computedfrom at leasttwo triads,hadan assigned qualityflag < 4, andspectralwidth < 299 c•n [ Miller andRiddle, 1994].
bNodatafromJanuary 7-28, 1993,andJanuary 30-February 4, 1993. CLastday of operationis February18, 1993;portcalls from November27-30, 1992, December12-18, 1992, andJanuary23-30, 1993.
dLastdayof operation is February23, 1993;portcallsfromDecember12-17,1992,andJanuary23-30, 1993. eln additionto standardqualitycontrol,only datacollectedwithin 1øLatitudeand 1øLongitudeof the listedpositionare included.
somewhatmore westerlythan normal [Gutzler et al., 1994]. The
NAU, RV3, and R/V Shiyan (RV1). This geographicaldistri-
strongest meanwesterlies (about6 ms-l at a heightof 2 km) are butionof variability may be associatedwith the large fluctuations observedat NAIl, the easternmost site. The zonal componentof the wind is quite coherentwith heightin the lowest4 km. The meridional wind is northerly near the surface and tends to decreasewith heightin the lowest2 km. There is no significant differencebetweenthe meanwind profileat shipRV3 relativeto the profiles at island sitesMAN and KAV, located at about the same latitude as RV3.
3. Daily Variability Figure 3 showsvariance-conserving power spectra(i.e., power spectraldensity times frequencyversuslog{frequency}) of the zonal wind, u, from eachsite at approximately1.5 km altitude (a heightat which the varianceis large). The total varianceof u at eachstationis listedin Table 2; it is leastat KAP andgreatestat N
75ms
-t
IOP-mean w•nds (m s-t) from TOGA-COARE s
KAP
NAU
MAN
KAV
RV3
RV1
of convection that occurred near the dateline during the lOP [Gutzler et al., 1994]; however, it is alsothe casethat the spectra from thesethree siteswere computedon the basisof fewer continuousdaysthanwerethe spectrafrom the othersites,andwithin thoserecordswere gapsof severaldays(Table 1). Superimposed on each spectrumin Figure 3 is a "red-noise" spectrumderivedfrom the observedvarianceand autocorrelation [Gilmanet al., 1963], from which it is seenthat all sitesdisplay considerablepower at the low-frequencyend of the spectrum. The low-frequencyvariance maxima would almostcertainly be even morepronouncedif the datarecordwere longerand allowed betterresolutionof low frequencyvariability. All of the stations also showa relative peak in variancenear 5 days,as observedin otherCOARE soundingsby Numagutiet al. [1995]. The 5-day peak exceedsred noiseat the land stations.The only site with a clear narrowband spectral peak at both the diurnal and semidiurnalfrequenciesis KAP, which exhibitsa very small amplitudebackgroundspectrum. (The samepower at diurnalor semidiurnalfrequenciesat the other siteswould not standout from neighboringfrequencies.)A diurnalpeak is seenat KAV, while RV 1 hasa semidiurnalpeak. Table 2. Variance Statistics for Zonal Wind u at the Six ISS
3
Sitesat Three Different Heights. - 750 '•
Fluctuations with
- 800 •
- 850 • -
Station
latitude
decreases
Total Variance, km
Period < 1 day, km
2.5
1.5
2.5
KAP
30
20
14
4
4
3
NAU
70
53
35
14
13
7
MAN KAV RV3 RV1
37 46 85 66
27 33 50 39
16 18 18 15
5 7 25 12
5 7 15 8
3 4 7 6
Profiler Site
900
to the mght
Figure 2. IOP mean vector wind profiles for each of the six
sites.A 7.5 ms-1reference vectoris shown in theupperright comer. The right-hand ordinate scale is determined from the climatologicalpressureprofile at Koror (7øN, 134øE),which is a closeapproximationto the averagepressureprofile at thesesites: (a) KAP, (b) NAU, (c) MAN, (d) KAV, (e) RV3, and (f) RV 1.
0.5
1.5
0.5
Total variance,and variance of fluctuationswith Period < 1 day, are determinedfrom spectralanalysis. The valuesfor 1.5 km are associated
with spectra shownin Figure3. Unitsof variancearem2/s2.
23,002
GUTZLER AND HARTFEN: DAILY VARIABILITY
Power spectral
density
a) KAP
50-
OF LOW-LEVEL TROPICAL WINDS
of zonal wind
50-
40
40
• 30
• 30
•'
• zo
•o lO
lO
o lOOO
lOO
lO
1
.1
. Ol
o lOOO
lOO
lO
50-
1
.1
. Ol
.1
. Ol
.!
. Ol
Period (days)
Period (days)
c) MAN
50
40
d) K•v
4O
• 30
30
•-
•o
2O
lO
lO
o lOOO
o
lOO
lO
1
.1
. Ol
lOOO
lOO
lO
50-
i
Period (days)
Period (days)
e)RV3
5O
40
f) RV1
4O
• 30
•o lO
lO
o !o( )o
/
o
lOO
lO
1
.1
.01
Period (days)
lOOO
lOO
lO
1
Period (days)
Figure 3. Variance-conserving powerspectraof the zonalwind at approximately1.5 km altitudeat eachof the six profiler sites.Small arrowsin eachpaneldenoteperiodsof 5 days,24 hours,and 12 hours. The dashedline in each plot is a red noisespectrumderivedfrom the observedvarianceandlag 1 autocorrelation:(a) KAP, (b) NAU, (c) MAN, (d) KAV, (e) RV3, and (f) RV1.
Figure 4 showsanalogousspectrafor the meridionalcomponentof thewind, v; corresponding variancesarelistedin Table 3. In generalthereis much lessvariancein v thanin u. The total varianceat NAU is about50% higherthanat the othersites. The diurnal harmonicis quite pronouncedat KAV and MAN, but thereis no noticeablediurnalpeak at the otherstations.A semidiurnalharmonicis evidentonly at NAU. The spectralpeaknear 5 days,seenat all sitesexceptMAN and KAV, departsconsiderablyfrom a red spectrum.The presence of thispeakin both u and v winds is consistentwith the interpretationof 4-5 day
variability as resulting from near-equatorial mixed Rossbygravity waves [Liebmannand Hendon, 1990; Nurnagutiet al., 1995]. In order to describe the periodic daily variability at each station, we first examined the lOP mean zonal and meridional
wind components for eachhalf-hourduringthe day. Thesevalues are averagedover all availabledata for eachparticularhalf-hour period,sothe individualIOP meanhalf-hourvaluesmay be based on slightlydifferentdaysat eachsiteandaltitude.Figure5 shows time-heightvectorplotsof the IOP meandaily variability,that is,
GU'I•LER
AND HARTtEN:
DAILY VARIABILITY
Power spectral
density
of meridional
a) KAP
50-
OF LOW-LEVEL
50-
TROPICAL WINDS
23,003
wind
b) NAU
40
"• 30 •'
2o
•
lO
20 lO
0
o
1000
100
10
1
lOOO
lOO
Period (days) c) MAIN
50
lO
1
.1
. Ol
.1
.01
Period (days) 5o
d) KAV
4o
• 30
• 30
•-
•-
2o lO
2o lO
o............. ••k.•.••:! uL,,•, .... 1000
100
10
1
I
o lOOO
.1
100
Period (days) e) RV3
5O
10
1
Period (days) 50-
f) RV1
4O
"• 30 •lO
20 •o
o
lOOO
lOO
lO
1
.1
. Ol
Period (days)
lOOO
lOO
lO
1
.1
Period (days)
Figure 4. Like Figure 3, but showingspectraof the meridionalwind: (a) KAP, (b) NAU, (c) MAN, (d) KAV, (e) RV3, and (f) RV 1.
half-hourlydeviationsaboutthe lOP mean daily average. It is apparentin theseplotsthat the temporalandverticalcoherenceof the profilerdata decreasesmarkedlyabove4 km, so we restrict our discussion to data from altitudes less then 4 km.
At KAP, the principalfeatureis the tendencyin the lowest3 km for easterlydeviationsat about0800 LT and westerlydeviations at about 1500 LT (all times refer to local time (LT); see Table
1 for the difference
between local time and Coordinated
Universal Time (UTC) at each station). There is almost no periodic daily variability in v. Daily variations at NAU are
noisier,but a tendencyfor daily u variability similar to that at
KAP is discemible.At the two ships,thedaily variabilityis again mostlyzonal; the largestdaily deviationsoccur after noon when the zonalwind exhibitswesterlydeviations. The daily variability at MAN and KAV is much larger in amplitude,with meridionaldeviationsfrom the southduringthe day and from the north at night. At MAN this diurnal cycle is concentrated in v, whereas at KAV the diurnal variations are
southeasterly andnorthwesterly.As KAV is locatedfartheraway from thelargeislandof New Guinea,to the eastof MAN, thereis a suggestion thatthe diurnalcyclesat thesestationsare relatedto their proximity and location relative to the large land massof
23,004
GUTZLER AND HARTYEN: DAILY VARIABILITY
Table 3. Variance Statistics for Meridional
Wind v.
Fluctuations with
Total Variance, k•n
Period< 1 day,km
2.5
0.5
2.5
Profiler Site 1.5
1.5
0.5
KAP
13
14
12
2
2
2
NAU MAN KAV RV3 RV1
20 10 13 13 13
20 12 13 12 12
22 13 13 9 6
5 2 3 5 3
4 2 3 4 2
5 3 3 3 2
The valuesfor 1.5 km are associated with spectrashownin Figure4.
New Guinea. Furthermore, the nocturnal onshoreflow maximum
occursslightlyearlierat MAN comparedto KAV, suggestive of a forcedcirculationpropagating offshorefromNew Guinea.
OF LOW-LEVEL
TROPICAL WINDS
remarkablyconsistentfrom site to site. The semidiurnalcyclein v is very small at all sitesand is not shown. Theseresultsreproduce the observedsemidiurnalwind variability over the eastern near-equatorial Pacific[Williams et al., 1992; Deser, 1994], and they correspondclosely to the cycles predictedby solar tidal theory [Chapman and Lindzen, 1970; Haurwitz and Cowley, 1973]. The consistencyof our resultswith thoseof Williamset al. [1992] and Deser [1994] providesfurther confirmationthat solarforcingis responsiblefor the semidiurnalcycle.
4. Discussion
Daily variabilityin lower tropospheric windsover the tropical westernPacific has been examinedusing half-hourly profiles from a network of wind-profiling radarsdeployedduring the TOGA COARE field experimentin the westerntropicalPacific. The amplitudeof the diurnal cycle in both u and v in the western
partof theCOAREdomainis of theorderof 0.5 msq . Thisis
significantlylarger than the diurnal cycle observedeast of the datelinein previousstudies. The semidiurnalcycle of the zonal mean winds shown in Figure 5 were extracted via harmonic wind is smallerin amplitude,and its phaseis consistentwith the analysis,definingthe averagediurnaland semidiurnalcyclestbr solar tidal oscillation in pressureand entirely consistentwith the zonal and meridional wind at each site and altitude. Vertical easterntropicalPacificresults. We find no systematicdifferences profilesof thesecycles,displayedin vectortbrmat,are shownin betweenthe shipdata and the small islanddataattributableto the Figures6-8. completeabsenceof land at the shipsites. The possibility that sea breeze/land breeze circulations conProfilesof the amplitudeand phaseof the diurnal (24-hour) harmonicof the zonal wind u are shownin Figure 6. The sites tributeto the diurnalcycle mustbe consideredfor the four land closestto the equator, KAP at IøN and NAU at 0.5øS, exhibit stations,but several factors argue againstsuch a contribution. small but coherent diurnal cycles with phase maxima (i.e., There are strongdiurnal circulationsat KAP, NAU, MAN, and maximum westerly deviation from the daily mean) late in the KAV (Figures 6 and 7). Each extendsto altitudes of 2 km or afternoon.Themaximum amplitude is about0.4 ms-l at KAP more, and each is such that flow from a particular compass andabout0.3ms4 atNAU, bothmaxima occurring about1 km directionpeaksat about the sametime throughoutthat depth. abovethe surface. There is no clear evidenceof verticalphase This flow is roughlybarotropic;thereis no apparentreturn flow propagation in thelowertroposphere.Theseharmonics arequali- at higher altitudes. In contrast, the depth of onshore flow tatively similarto the corresponding diurnalcyclesof surface observedduring sea breezesis typically severalhundredmeters zonal wind at buoysjust north of the equatorat 170øW [Deser, [Bantaet al., 1993;Finkeleet al., 1995],althoughdepthsof 1 km 1994]. or more are possible in the tropics [Wexler, 1946]. Sutcliffe The westernmoststations,MAN and KAV, exhibit large [ 1937] detectedboth the sea/land breeze and the return flow aloft; diurnalcyclesextendingthroughoutthelowest4 km. The ampli- he found the average vector difference between afternoonand The 24-hour and 12-hour harmonics of the time series of IOP
tudeof thediurnalcyclereaches 0.4 ms-I atMAN and0.5 ms4 at
moming soundings tobeof theorderof 1 ms4, extending about1
KAV, andthe phaserepresentswesterlyflow duringthe daytime and easterlyflow at night, with MAN leadingKAV. Fartherto the east,theamplitudeof the diurnalcycledecreases to about0.3
km abovethe top of the discerniblesea breeze. The deviations abovethe deepunidirectionalflow shownin Figures6 and 7 are considerably smallerthan Sutcliffe's. Furthermore,the directionsand/orthe timing of the maximum
ms-l inthelowest 1.5kmatRV1 andRV3whilethephase ofthe harmonicis notsignificantly differentfromthephaseat KAV. Figure7 showsprofilesof the diurnalharmonicof the meridionalwind,v. All siteshavevery smallharmonics exceptMAN and KAV, where a coherent diurnal harmonic extends to about 3
winds are inconsistent with a sea breeze at all four land stations.
The ISS station at KAP
was located on the southern end of
Taalin, a small island (diameter less than 0.4 km) along the easternborder of KapingamarengiAtoll. If the shallow water
km.Thediurnal cycleatMAN hasanamplitude of about0.7ms-1 within the atoll were to heat and act as "land", at KAP a sea in the lowest1.5 km with maximumsoutherlywindsat about 1100 LT; KAV exhibits a somewhatsmalleramplitudeand a slightlylater phase. The amplitudeof the diurnalcycle in v at MAN and KAV in the lowest2 km is substantially largerthan any of the corresponding amplitudesfoundby Deser [ 1994]east of the dateline; however, all of the ISS siteseast of KAV exhibit
smallerdiurnalcyclesthanmostof the buoysitesconsidered by Deser. The diurnalcyclesat MAN and KAV in the lowest2 km reveal phaselags consistentwith the plotsof daily variability shownpreviouslyin Figure5, that is, the diurnalcycleat MAN leadsthecycleat KAV by aboutan houranda halfi Figure 8 showsprofiles of the semidiurnalharmonicof the zonal wind. The amplitude and phaseof this harmonicare
breeze would be expected to blow from the east during the daytime. Instead,Figures6 and 7 showa westerlymaximumlate in the afternoon.The islandof Nauru is shapedsomewhatlike a rectanglewhosemajor axis, about 5.3 km in length, runs northeastto southwest; the maximumwidthof theislandperpendicular to this major axis is about4.7 km. The profiler at NAU was locatedmidwayalongthe southwestendof Nauru, with the large plateauthat coversmost of the island extendingnortheastfrom the station.The observedafternoonmaximumagainoccursrather late in the day for a sea breeze (about 1800 LT) and is almost entirely from the west (the meridionalcomponentof the diurnal cycle is not coherentwith height). This is a highly unlikely directionfor a seabreezegiventheislandgeometry.
GUTZLER ANDHARTYEN: DAILYVARIABILITYOFLOW-LEVEL TROPICAL WINDS a) KIP
23,005
d) KAV •2.0
6
500
m s-1
-
55O
....
-.
--
•
•
...- -
600
500
-
55O
-
600
_
ß • .,. .. .... _
i
.... . .
t
.
•
..
',,
t
.
-.
x
•
......
. .- .. - ....•..,., ..-/
650.-•
,, - .
t
700 •
.
750 800
' ' • "-- '"
'
'
'
• '-" "- • "' ......
t
.
0
_
.......
•
-
12
.......
850
900
900
950
18
•
- 800•
850 -
.
- 750 2
950
24
i
0
b) NAU
i
i
[
i
i
•
]
i
i
i
[
i
i
12
i
!
i
i
[
[
18
]
1000
i
e) RV3 ,
-1--• -
500
2.0 m s-1
55O
55O
/IX
600
- 750
-
\ .. • \
t,
I
•
•
.-----
.
'•'•"'-•..--•
-
,
,
..,.,•,'•--.--....
..
.,. - - . ,,,%,,,.- . x • •
850
x
,
,
900
/
-
,
\ ......
,, • -
-
\,
•
•
"-
•
-...,.....--
,
,
_ /
,. -- ,, , '
,., .
.
... •
.
,
t •
-
t
-
/%
,,..
- • .-'•,,' I
.-
'"
"'
'-
, • .-..,,..
900 -
1000
6
12
18
i
0
24
c) MAN
i
i
i
•
i
•
i
i
i
!
i
i
i
12
i
i
i
i
i
i
18
i
•
•
850
950
0
700
750
800 •
/ / ,, ,.•._ 1•---/
"x
, /
, i
_•
600
- 650-•
'"'•"'- • ; ,,•'"''----...---.. / / .. / \ j ,, !.,,_..
•
800 • -
i ._.
/' 1.. -/ •'"---"•--- .
650 •
700
'-"*'-'•,•--""-'- ....
950
-1000
i
24
f) RV1 > 2.0 m s-1
--•-
_
-
'--'.-',/'/ /.,- x • '"'- t'-.. t t • ,,
55O
- 1• -"' ' 2.--' •//-' '--•,,,''- /. •
600
J-Ill'
--'"""'"'"''\
''•
....
Ix",,
"- •' %1// '- • • ....
- - x ' /-"
750 / -800
• t / • • /' /' /',.. / . • / /,/'//
-
t t t • I t t I / .. -,- ' •//// • / /z'.." /1
, t lt• i
i
i
,, It
i
i
i
i
12
i
i
i
i
i
18
i
i
i
i
850
500
2.0 m s-1
, •
, x
.
ß
I
"- '
"
ß
•
.-
'
i •
....
'"'•'
. -..--
\
--.. ---•.•,---• ...•,
\
ß
ß
ß
.x
• ,, x , ,
-.--•-
.--
- "" "" " '
'
- / ....
95O
ß
\
•
•
,,
,, •
.
.
,
•
.
800
....
, /'-- --- i
• '
.
--.•-
.
ß '
.
7oo E
750
9OO
/
- x•"' - 650'-•
, ... - .,-- ,.-. . •
'"
•
,,
•'"-"'- 600
I I 'X "'*"'-"'•-" /-'-'
• ..
,,- ,
/ - /
/
. -,
,.,
• ---'•
•
24
Local Time (hours)
0
i
•
i
i
i
•
6
i
i
i
i
i
i
1'2
i
i
!
i
!
i
i
i
850
\-
18
i
• •
•- 900
1000
i •
55O
950 - 1000
24
Local Time (hours)
Figure 5. Vector plots ofthelOP-mean daily variability ofwinds atthesixsites asafunction ofaltitude (i.e., halfhoufiy deviations from thelOPmean winds shown inFigure 2). Forclarity thetemporal resolution oftheplotis
onlyhalftheresolution of thedata(i.e.,every other vector, centered onthehalf-hour, isplotted). A 2 ms4
reference vector isshown intheupper-right comer ofeach panel: (a)KAP,(b)NAU,(c)MAN,(d)KAV,(e) RV3, and (f) RV1.
NewIreland isa fewhundred kilometers inlength; the diurnal cycles inFigures 6and 7show asouthwesterly maximum southernmost hundred kilometers lienorth-south, and from thereduring the early afternoon, clearly atodds with any expected sea theisland extends northwestward. Atitsnorthwestern end the breeze. The same southwesterly afternoon maximum isobserved bulk ofthe island swings westward forabout 25km; north ofthis atMAN.This station issimilarly located atthenorthwestern
land lies Balgai Bay, formed byapeninsula which juts out tothe terminus ofapeninsula which provides the northeastern boundary northwest along the main axis ofthe island. The ISS site isatthe ofaharbor; Seeadler Harbor isbounded tothesouth byLos
northern tipofthis peninsula; due south ofthestation lies marsh,Negros Island and tothewest bythelarger land mass ofManus thebay, and thewestward-directed island mass. AtKAVa sea Island itself. Asea breeze atMAN would beexpected toblow
breeze would beexpected tobein somesense northwesterly or fromtheeast.
northerly, with theexact direction dependent onwhether the
Instead ofbeing driven bydifferential sensible heating asso-
marshy peninsula onwhich thestation issituated could heat ciated with local land/sea temperature contrasts, webelieve the
enough todrive asea breeze circulation orifasea breeze wouldpronounced diurnal variability atMANand KAVisprimarily a betied toeither thelarger solid land area south ofthebayorthe manifestation ofthestrong diurnal cycle indeep convection over
bulk ofNewIreland tothesoutheast. Inanycase, theobserved theland areas ofNewGuinea andNewBritain, located tothe
23,006
GUTZLER AND HARTTEN: DAILY VARIABILITY
OF LOW-LEVEL TROPICAL WINDS 1.0 m s-•
1.0 m s-• 24-hr
harmonic
in
KAP
NAU
zonal
(m s-•)12000
wind
MAN
KAV
RV3
12-hr harmonic in zonal wind (m s-•)06000 09
RV1
MAN
4
KAV /
RV 3
! 3
600 •
.•
850•
•:
•V1
/
t
/
/
/
!
!
;
!
!
/
!
750 .
800 •
2
850 • !
900
Northern Hemmphere/
Southern
Near-equator
9OO
1
Hemmphere
Figure 8. Like Figure 6, but for the semidiurnalcycle (12-hour harmonic)of the zonal wind.
Figure 6. Vertical profiles of the diurnal cycle (24-hour harmonic)of zonal wind at the six ISS sites,representedin vector format as illustrated by the dial in the upper right corner. A vectorpointing horizontally to the right denotesa diurnal phase maximum (that is, maximum westerlywind) at 0000 local solar time, and phaseprogressescounterclockwiseas shown on the dial. A vectorwhoselengthis the radiusof the dial representsan
amplitude of I ms-1 The tail of eachvectoris plottedat the altitudecorresponding to the centroidof the layer sampledby the radar to obtain the data.
south and west of the IFA. Hendon and Woodberry [1993] showedthat the diurnal cycle of tropical deep convectionis stronglytied to land areas, with areas of intenseopen ocean convectiveactivitydisplayingmuchweakerdiurnalmodulation.
The daily maximum in deep convection(as representedby minimum values of satellite-observedbrightnesstemperature,
indicativeof cold cloud tops) over New Guinea was found to occurat about2200 LT or about 1200 UTC (H. Hendon,personal
communication,1994). This is approximatelyan hour before the time of maximumnortheasterlyflow observedin the lowest 2 km at MAN, and 2-3 hoursbefore the weakernortheasterlymaximum at KAV
and the research vessels. The diurnal flow at 1200 UTC
at 1.5 km altitude at the six ISS sites is shown in Figure 9, illustratingthe lower troposphericinflow toward New Guinea at this time.
The phaseof the diurnal cycle of deepconvectionover New Guinea can be comparedwith the correspondingdiurnal cycle in open ocean rainfall measurements. danowiak et al. [1994] recentlyshowedthe daily variabilityof rainfall from five optical gauges deployed on moored open ocean buoys during the COARE IOP. The gaugeswere located between 154øE and 165øEon the equatoror at 2øS,that is, betweenKAP and NAU, far to the northeastof New Guinea. Heavy rain eventstendedto occur between0300 and 0600 LT (broadly consistentwith the estimateof Mapes and Houze [ 1993]), and considerablylater in the night than the island-forcedconvection. In the absenceof preferredgeographicalanchorsfor open oceanrain events,there is no reasonto expectconvectivelyforceddiurnalwind cyclesin climatologicalstatisticsat individualopen-oceansites. Althoughthe phasesof the diurnal cyclesof wind near New Guinea and convection over New Guinea seem consistent,
06
24-hr harmomc •n memd•onal w•nd (m s-i)
10ms
-r
12 (f• 00 18
MAN
KAV
modelingstudies(beyondthe scopeof our investigation)of the responseto anchored, diurnally varying convection will be requiredto make this explanationfully convincing. Sucha study shouldfirst seek to reproducethe observedamplitudeand horizontal and vertical scalesof the forced responsein the winds. Another feature of the observationsworth examining is the apparentoffshorepropagationof diurnalsignalin the winds(that
750 '•
4N -•
• 0.5 m/s
KAP,.•
600 •
2-
EQ
850 • _
9OO
4S
.¾...AN KAV
NAU