JOURNAL OF GEOPHYSICAL
RESEARCH, VOL. 102, NO. C9, PAGES 20,937-20,955, SEPTEMBER 15, 1997
Temporal and spatial variabilities of the South China Sea surface temperature anomaly PeterC. Chu, Shihu• Lu,• •nd YuchunChen• Department of Oceanography,Naval PostgraduateSchool,Monterey, California
Abstract.
In this study we use the National Centers for Environmental Prediction
(NCEP) monthlyseasurfacetemperature(SST) fields(1982-1994)to investigate the temporaland spatialwriabilities of the South China Sea (SCS) warm/cool anomalies. Three steps of analysis were performed on the data set: ensemble
mean (T), compositeanalysisto obtain the monthlymean anomalyrelativeto the ensemblemean (T), and empiricalorthogonalfunction(EOF) analysison the residuedata relativeto (T) +(T). The ensemblemean SST field (T) has a rather weak horizontal gradient: 29øC near the Borneoco•st to 25ø-26øCnear the southeast
Chin•
coast.
Two areas of evident
SST anomalies
were found in the
monthlyT variation:westof Borneo-Palawan Islands(WBP) and southeastof the southernVietnam coast(SVC). Four patterns,monsoonand transitioneachwith two out-of-phasestructures, were found. During the spring-to-summertransition
(Marchto May) the warm anomalyis formedin the northernSCSwith T > 1.8øC locatedat 112ø-119ø30•E, 15ø-19ø30•N. Duringthe fMl-to-wintertransition(October to November)the northernSCS(northof 12øN)coolanomalyisformedin November with T < -0.6øC located at 108ø-115øE,13ø-20øN.We performed an EOF analysis on the residue data relative to T + T in order to obtain transient and interannual
variationsof the SST fields. EOF1 accountsfor 47% of the varianceand represents
the northernSCS warm/coolanomalypattern. EOF2 accountsfor 14% of the variance and representsthe southern SCS dipole pattern. Strong northern SCS
w•rm •nomMy (IøC w•rmer) •ppe•rs duringOctober-November 1987•nd J•nu•ryFebruary1988, and strongnorthernSCS coolanomaly(IøC cooler)occursduring March 1986 and November 1992. Furthermore, • strong crosscorrelation between wind stress curl and SST anomalies, computed from the European Centre for Medium-RangeWeather Forecastanalyzedwind stressdata and the N CEP SST 'data for different lags, showsthe existenceof an air-seafeedbackmechanismin the SCS deep basin. Based on limited data sets, both cool and warm
1. Introduction
anomaliesweredetectedin SCS.Dale [1956]reporteda The SouthChinaSea(SCS)hasa bottomtopography coolanomalyoff •he centralVietnamesecoastin sum(Figure1) that makesit a uniquesemienclosed ocean mer. Nitani [1970]founda coolanomalylocatedat the
basin that is overlaid by a pronouncedmonsoon sur-
northwest of Luzon. Reports from the South China Sea
facewind. Extendedcontinentalshelves(lessthan 200 Institute of Oceanology [1985]indicatethat in the cenm deep)are foundon the northernand the southwest- tral South China Sea, a warm anomaly appears in both ern parts, while steepslopeswith almost no shelvesare summer and winter but closer to Vietnam in summer found in the eastern part of SCS. The deepestwater is confinedto a bowl-type trench. The maximum depth is around 4700 m.
at the surface. Recently, a warm anomaly was reported
in the central SCS during the late springseason[Chu and Chang,1995a,b, 1997; Chuet al., 1996, 1997]and
a cool anomaly was detected in the central SCS during •On leave from the Institute of Plateau Atmospheric December 29, 1993 to January 5, 1994 from the analyPhysics,Academia Sinica, Lanzhou, China.
sis of TOPEX/ POSEIDON data [Soonget al., 1995].
This paper is not subjectto U.S. copyright. Publishedin 1997 by the American GeophysicalUnion.
surfacetemperature(SST) anomaly?We usedthe National Centersfor EnvironmentalPrediction(NCEP)
Paper number 97JC00982.
SST data set to investigatetheseproblems.
What are the temporal and spatial variabilities of sea
20,937
20,938
CHU ET AL.' SOUTH CHINA SEA SURFACETEMPERATURE VARIABILITIES
Geographyand isobaths(unit'm) 25
2O
15 .o
10
1!0
115
120
125
Longitude(E),Depth--0(m) Figure1. Geography andisobaths showing thebottom topography ofthe South China Sea.
The SST data (1982-1994)for this study wereob- The satellite SST data were obtained from analysis of tained from the NCEP monthly real-time SST global National Environmental Satellite Data and Information data set, which was establishedby blendingin situ Service data by the University of Miami's Rosentiel and satelliteSST data [Reynolds,1988;Reynoldsand Schoolof Marine and AtmosphericSciences.We usethe Marsico, 1993]. The monthly optimuminterpolation monthly SST for the period 1981-1994for this study.
(OI) fieldsarederivedfromlinearinterpolation of the weeklyOI fieldsto daily fieldsthenaveraging the daily 2. Seasonal Variation of the SCS Wind valuesover a month [Reynoldsand Marsico,1993]. Stress The monthly fields have the samespatial resolution SCS experiences two monsoons, winter and summer, (10 x 1ø) as the weeklyfields.Beforethe analysis is every year. During the winter monsoon season,a cold completed, the satellitedata are adjustedfor biasesusnortheast wind blows over SCS (Figure 2a) as a reing the methodof Reynolds [1988]and Reynolds and sult of the Siberian high pressure system located over Marsico[1993].A description of the OI analysis isgiven the east Asian continent. Radiative cooling and persisby Reynoldsand Smith[1994].The biascorrection imtent cold air advection maintain c61d air over SCS. The provesthe large-scale accuracyof the OI. The data pronortheast-southwest oriented jet stream is positioned at cessing procedure described herewasdoneby NCEP. In situ SST data were obtained from the Comprehen-
the central SCS. Such a typical winter monsoonpattern
to April). Duringthe sive Ocean-AtmosphereData Set for the period 1981- lastsnearly6 months(November summer monsoon season, a warm and weakersouthwest 1989, and from radio messagecarried on the Global wind blows over SCS (Figure 2b). Such a typicalsumTelecommunicationSystemfor the period 1990-present.
CHU ET AL.' SOUTH CHINA SEA SURFACE TEMPERATURE 25
25
2(
20
VARIABILITIES
20,939
Z15
•10
110
115
120
125
110
Longitude(E), Depth=0(m)
115
120
125
Longitude(E), Depth=0(m)
Figure 2. Mean surfacewind stressvectorsin the SouthChina Seafor (a) Decemberand (b) June(computedfromthe ECMWF data set).
mermonsoon patternlastsnearly4 months(mid-May
2.Wintermonsoon pattern:Duringthe wintermon-
to mid-September). The meansurfacewind stressover
soonseason, a northeast-to-southwest orientedline sep-
SCSis nearly0.2 N/m2 andreaches 0.3 N/m2 in the aratesthe SCSinto two parts:the westernpart uncentralportionin December (Figure2a) and is nearly der anticyclonic wind stresscurl and the easternpart 0.1N/m2 in June(Figure2b). is undercyclonic windstresscurl, that is, the AC-C UsingtheEuropean CentreforMedium-Range Weatherpattern(Figure3c). The maximumabsolute valuesof
Forecast (ECMWF)analyzed winddata,Trenberth et
bothcurlsreach4x10-7N/m3. Sucha patternpersists
al. [1989]calculatedglobalmonthlymeanwind stress
from November
to March.
3. Spring-to-summer transition pattern shows an opposite dipole pattern to the fall-to-winter transition period. The dipole curl is characterized by a WBP cy-
curl ( on a 2.5ø x 2.5ø grid for the period 1982-1989. The drag coefficientwascorrectedfor atmosphericstability, with sea-air temperature differenceand the relative humiditytakento be meanmonthlyclimatological cloniccurl and a SVC anticycloniccurl (Figure 3d).
values.The ensemble mean(1982-1989)( field (Figure 3a) showsa dipolepattern' cycloniccurlsoverthe eastern SCS and anticyclonic curls over the western SCS.
This coincideswith the prevailingnortheastwindsover the central
SCS.
Four patterns, monsoonand transition each with two
out-of-phase structures,were found (Figures3b-3e). The monsoonpattern is featured by a northwest an-
ticyclonic (cyclonic) curlanda southeast cyclonic (anticyclonic)curlin winter(summer).The transitionpat-
The strength of the dipole curl is much weaker than the winter monsoonperiod. 4. Summer monsoon pattern is 1800 out-of-phase with the winter monsoonpattern. A northeast-to- southwest oriented line separatesthe SCS into two parts: the western part under cyclonic wind stress curl and the
easternpart is under anticyclonicwind stresscurl (Figure 3e). The strengthof the wind stresscurl during summermonsoonseasonis about half that during winter
monsoon
season.
tern isfeaturedby a southernSCScyclonic/anticyclonic dipole. 1. Fall-to-winter transition pattern showsa southern
3. Composite Analysis of SST 3.1.
Ensemble
Mean
SST
Field
SCS (southof 12øN)dipolecurlstructurelocatedat the We now examine the data to see if we can obtain a warm westof Borneo-Palawan Islands(WBP) and the southanomaly signal.SST isrepresented by T(xi, Yi, X/•,tt), eastof the southernVietnamcoast(SVC). The typical where (xi, yj) is the horizontal grids, Xk = 1982,1983, pattern is featured by a WBP anticycloniccurl and a ... 1994 is the time sequence in years, and tt -1, 2, ... SVC cycloniccurl (Figure 3b). The strengthof the curl is quite weak, and the maximum absolute value is 12 is the monthly sequencewithin a year. Before in1.2x10-7N/maforthecyclonic curland0.6x10-7N/ma vestigatingthe monthlyvariationof SST, we definethe followingtwo temporal averages: for the anticyclonic curl.
20,940
CHUET AL.' SOUTHCHINASEASURFACE TEMPERATURE VARIABILITIES ensemble
September Mean
mean field
25
25 ':i ß"
'-.••••'•.%•:-i;, / •
ß,
ß
20
•::•;•'•]'•I• .....
Z15
"'
........ -•,••• • • '•10
•
ß •
•:•.... •'.
. .• ........ :. • .•. •',•'
5b.-•-.-•6: ........ •'••7'"h ...... 110
115
120
110
125
115
120
125
Longitude(E),Depth=O(m)
Longitude(E), Depth=0(m)
April Mean
January Mean 25
25
20
2O
Z15 J
.
•10
z
110
115
120
125
Longitude(E), Depth:0(m)
5
110
115
120
125
Longitude(E), Depth=0(m)
Figure $. Surfacewindstresscurl: (a) ensemble mean,(b) September, (c) January,(d) April, and (e) July (computed from Trenberth et al. 1080). The unit is 10-a N/m a. 1994
T(•, v•, x•, tt), T(zi, yj,tt)- AXk=1982
12
-
(2)
mean. The ensemble mean(1982(1) whichis the ensemble 1994)SSTfieldovertheSCS(Figure4) shows a pattern whichis thelong-termmeanvalueforthemonthtt, and of northeast-southwestoriented isotherms with a pos-
A x - 13
(1004-1982 + 1).
CHUET AL.- SOUTHCHINASEASURFACE TEMPERATURE VARIABILITIES
July Mean
20,941
(T > 1.0øCin July) in the southernSCS.The summer
25
T pattern is out-of-phase of the winter pattern.
4. Fall-to-wintertransition(Octoberto November) pattern showsthe northward expansion of a WBP low
anomalyand the formationof a northernSCS (north of 12øN)low anomalyin November.The low anomaly
2O
with T < -0.6øC is located at 108ø-115øE, 13ø-20øN Z15
(out-of-phaseof the spring-to-summer transitionpattern). A smallhigh anomalyformsnear WBP.
rolo
mean SST anomalyfields (Figure 6)' During winter (Decemberto February) in the north SCS (north of 12øN) isothermsare nearlyparallelnortheast-southwest
These features can be easily seen from a seasonal
oriented curveswith T decreasingnorthward, and in the
south SCS (south of 12øN) there existsa WBP high anomalyand a SVC low anomaly (Figure 6a). During spring-to-summertransition, the SVC low anomaly
weakens,and the high anomaly (originalwinter WBP high anomaly)movestowardwest,expandsnorthward,
and finallyformsa largeSCShighano•maly. A WBP 110
115
120
low anomalyforms(Figure6b). After T formsthe typ-
125
ical pattern shown in Figure 6b, the SCS high anomaly
Longitude(E), Depth=0(m)
starts
to weaken
and retreats
toward
SVC.
The
north
Figure 3. (Continued)
SCS has nearly parallel isotherms with T increases northward, and the WBP low anomaly strengthens.
itive temperature gradient toward southeastnear the
Gradually,it shiftsto the summerpattern (Figure6c),
equator. The ensemblemean has a rather weak horizon-
tal temperaturegradient,decreasing from 29øCnearthe Borneo coast to 25øC near the southeast China coast.
3.2. Mean MonthlySSTAnomalies (•)
which features a WBP low anomaly and a SVC high anomaly in the south SCS. Such a summer pattern lasts
around 4 months (June to September). At the end of summer,the low anomaly (original summerWBP low anomaly)moveswestwardand expandsnorthward,
The long-term monthly mean valuesrelative to the
ensemble mean,T(xi, yj),
T(xi,Yj,tl)--•(xi, Yj,tl)--•(xi, Yj)
ensemble
mean of SST field
25
(3)
leadsto the composite featuresof the monthlymean
SSTanomalies, whichhavethefollowing features (Fig-
2O
ure 5).
1. Thetypical winter(December toFebruary) • pattern contains northeast-to-southwest oriented isotherms
in the northernSCS(northof 12øN),a WBP high anomaly(T • -1.2øC in January),and a SVC low anomaly(T ( -2.4øC in January)in the southernSCS (southof 12øN). 2. Spring-to-summer transition(Marchto May) pattern showsthe northwardexpansionof the WBP high anomalyand the formationof a northernSCS(northof 12øN) high anomaly.The isolineof T - 1øC encloses almost all the SCS in May. The high anomaly with T • 1.8øC is located at 112ø-119ø30•E,15ø-19ø30•N.A small low anomaly forms near WBP.
3. Summer(June to September)T pattern is char-
110
acterized by northeast-to-southwestoriented isotherms
in the northern SCS (north of 12øN), a WBP low anomaly(T ( 0.4øCin July), anda SVC highanomaly
115
120
125
Longitude(E), Depth=0(m) Figure 4. The ensemblemean of the SST field during 1982-94.
20,942
CHU ET AL.- SOUTH CHINA SEA SURFACETEMPERATURE VARIABILITIES
monthly mean SST anomaly 25
-• ••' '•••x •••'.•,--•; ..•e, '•-•" "---x' •,•
monthy mean SST anomaly
--:.• • //.-.
...... ..,-.--•.,%.•.. . :!•• x,:,xx',x-:.:-:'..,, .••••-•,•..'::"::, ,'•. ::4•. :,,,x-:-:. ß ,•.x,x:. -,
••,•
.%',..*:..: /.
/,
25 .
E.','-x.--, •.•••••,,.%,x...%%•, •'%%•,.,:.... •. •','x':, .... .. ••""
:i:..x.'.-::.•-..--:•-L"i•:' •. :- .....
-,,,•.4 /. ./
.
/
•///;
-.
--- .,.
r'*-..-"•,•: ,.'•:"• = •:::.:..?..:. :.:,?•..•..". •:..•.,:. 7.:
2O
ßß .....: ,ß 20
i=..4:,':,½:.,:i•. •"h.-'.•..• •'.'.:%%'•'::" . •:½.•,.'.•.,,.,., / -q./ ..-. ---" -.-I.-..'•.?•.,:.:.: ......
["**'*,•.-x •••••.' : / /.
/ . .,1-,•
.
:t- &-.;
[•4,,x.,•,-,,,.I :/ /. ½/ :,4- '"• •: _,,'•::½:..,. ,..*,:
b,',.• .... /.}./-.•...j: ...... ;,: ....:.... .•::...?..
•?'* / I • •/-t-/ :.'".y-':' : • "'2_'
I'":•'-, .•.:' ', ' ,: ':"-:-
. •-1½8i,', ,,, •'.'-0.4 '- -'i" ---'-?'• I•
,; ///½
.....
,. ,.•
/.. 2'...........
• ./ / ,
110
•.-%,-%%',,•. : -0.2
5 ..... I"• .- . ,
'
120
:
--'----
•,•,,•, :' • •- -:-• ..,•%•- ß•
•
• "A_•'
- ...-. /
! :l -
...... . . .•.
I '
125
•..'.¾•-
._,•,
..• •_•
½ .
•
-,.
'•:.... &,•...,•-•' 9' :•'•(:•':
.... '•'L''• •'i'//">'=" •
1 15
Longitude(E), Depth=0(m)
monthlymean SST anomaly
•.:.•..• /
.--,.,. ,..,
,,,; 110
125
25
Z15
.••ff.='.='•..: ....
_,',..•,....•:•:..-,. -•'•'-, •", , .
•.,,,,,, •,•, ,,,:,.,,.,,,,• k-
115
Longitude(E), Depth=O(m)
2O
I '• ..b.';:':,', *:• p
•. '-;..../../ ..... .,.... :.*•"::.." 5.'.. :•:½',., /'/ V':i !. •'.. ß ß ß .'.."•, 7/,.;'>.,, .'.... :. ! ..... :,,:: "'-'•., .•
•. •
.• •:.../. •-
•1•4 / .4-''
.•:•/
'
-•"'
E'/.
'•,/.,' • ...... :.•.... _:/-.-:..... ff..•,:.,.
_/...d'•: .. •
:,, ._,
:
z"'""•.
ß : ....... :......... 110
115
ß': 120
Longitude(E), Depth=0(m)
Figure 11. Cross-correlationcoefficientbetweenthe surfacewind stresscurl
and SST anomalies for variousSST lags:(a) 1 month,(b) 2 months,(c) 3 months,(d) 4 months,(e) 5 months,and (f) 6 months.
125
20,952
CHUET AL- SOUTHCHINASEASURFACE TEMPERATURE VARIABILITIES 25
. L
20
-:
/.
/
2.
./
Z15
_j 10•'.•
.z_ -• .-..... 2....... /.'. ....... •.,',,•.:
'•1o
ß.> ..,..: ....
110
115
120
110
125
115
125
120
Longitude(E), Depth=0(m)
Longitude(E), Depth--0(m) 25
25
/ J
2O
20
I /
/
2'.
0.4 Ill ß
Z15
Z15
•10
•1o
O
,
,
,
110
115
110
125
120
115
120
125
Longitude(E),Depth=0(m)
Longitude(E),Depth--0(m)
Figure 11. (continued)
regions(deepbasinandshelf)indicatesdifferentphysics pumpingeffect' cyclonicwind stresscurl inducesupwellingwhichsucksthe deepcoldwaterto the surface, to be involved in the air-sea feedbackprocesses. and anticyclonicwind stresscurl generatesdownwelling whichpreventsthe deepcoldwater to the surface.The
5.3. Possible the Northern
Air-Sea Feedback SCS Basin
Mechanism
in
cold(warm)SSTanomalytendsto produce high(low) surfacepressureperturbationwhichleadsto the genera-
(cyclonic)windstresscurlanomaly. We may suggest an air-sea feedback mechanism in tion of anticyclonic positive in thebasin the SCS deep basin from the evident crosscorrelation Thusthe values of •{•) become between( and T for varyinglag r. The negativevalues as-2 > r > -7. The wind effectto generatethe cen-
hasbeenverified of•(r) (0< r < 3) in thebasin indicate theEkman-tral SCSwarm/coolSSTanomalies
CHU ET AL.: SOUTH CHINA SEA SURFACE TEMPERAT •RE VARIABILITIES
110
115
120
125
110
Longitude(E), Depth=O(m)
115
120
20,953
125
Longitude(E), DePth=O(m)
25
25
2O
2O
Z15
.__
'•1o
110
115
120
Longitude(E),Depth=O(m)
125
110
115
120
125
Longitude(E),Depth=O(m)
Figure 12. Cross-correlation coefficientbetweenthe surfacewind stresscurl
and SST anomalies for variouswindstresscurllags:(a) 1 month,(b) 2 months, (c) 3 months,(d) 4 months,(e) 5 months,and (f) 6 months.
by a numericalmodel (Chu et al., Generationof the (SST) fields(1982-1994).Compositeand EOF analySouth China Sea deep basin warm and cool anomalies seswereusedfor the study.We obtainedthe following by wind-topography forcing,submittedto Journal of resultsfrom this observationalstudy: Geophysical Research,1997). 1. The ensemble meanSST field (T) wasestablished with a rather weakhorizontalgradient(29øCnear the 6. Conclusions Borneocoastto 26øCnearthe southeastChinacoast). Thegoalof ourstudywasto detectwhetherwarm/cool Thisresultagreedquitewellwith the earlystudyusing anomalyexistsin the SCSusingthe NationalMeteoro- the Navy'sMasterOceanographic Observation Data Set logicalCenter(.rr) monthlyseasurface temperature [Chuet al., 1996,1997].
20,954
CHU ET AL.: SOUTH CHINA SEA SURFACETEMPERATURE VARIABILITIES 25
25
/
/
/
2O
2O
-/0..4
o.6•
Z15
Z15
1.8øC located at 112ø-119ø30•E,15ø-19ø30•N.Dur-
is formed in November
with T