JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 103, NO. D24, PAGES 31,763-31,774, DECEMBER 27, 1998
A numericalstudyon the couplingbetweenseasurface temperatureand surfaceevaporation Dimitris
Tsintikidis
HydrologicResearchCenter, San Diego, California
GuangJ. Zhang ScrippsInstitutionof Oceanography, Universityof Califomia, SanDiego, La Jolla
Abstract.The feedbackbem'eenseasurfacetemperature (SST)andsurfaceevaporation is an importantissuein the studyof climatechange. To understand thisfeedbackandits interaction with surfacewind in thetropicalPacificOcean(30øN- 30øS),andin particularoverthe wam• pool region,a dynamictropicalatmosphericcirculationmodel is used. The model consistsof a twolayerfree troposphere m•da well-nfixedboundarylayer. It involvesactiveinteractions bctxveen the boundarylayerflow, forcedby an SST gradient,andthefree atmospheric flow, forcedby SST. VariousSST fields (representing climatology., E1Nifio, andLa Nifia conditions)areusedto drive themodel. It is foundthatthebim•edaverages of evaporation andwind speedincreasewith SST for up to about300-301K. Fromthatpointon theydecrease with SST. In addition,negative SSTanomalies correspond to excesslatentheatflux andwind speed.Theseresultsarein agreement with relevantobservations. To understand thethermodynamic versusdynamiceffectsof SST on surfaceevaporation, in oneof the experiments we imposea suddenpositiveSST pem•rbafionon the climatologicalSST field duringthe modelintegration.It is shownthatwhile surfaceevaporation is initiallyenhanced in response to the SST change,as theatmospheric circulation gradually"feels"theSSTperturbation, itsdynamiceffecttlu'ough thecirculation changebecomesmoreapparentoverthe SST perturbationregion. Overall, the resultsof our study showthatin the low SST regimethebehaviorof evaporation is dictatedby thermodynamics, whereasin the high SST regimeit is dictatedby atmospheric dynamicconsiderations.
1.
Introduction
layer leading to cooling of the ocean surthce. Convection
camesthe excessenergyupwardto the uppertroposphere, and The feedback between evaporation and sea surlhce fromthere it is transportedby large-scalecirculation to the temperature (SST), as well as the SST regulation, in general,in easternPacific.By assuminga constantwind speedand relative the equatorial Pacific Ocean are important issues in climate humidityin thebulk aerodynamic Ibmmlafor latent heat flux in change. Several studies exist that try. to elucidate on the their simpleanalytic model,Hartmann and 31icheL•en[1993] interaction between SST and evaporation. In an attempt to also concludethat evaporativecooling is the mostimportant explain their interaction Ramanathan aml Collins [1991] factorin limiting the tropicalSST. proposedtJ•e"thermostat hypothesis." According to their A cormnonassumptionin fl•e above studies is that surlhce hypothesis, convection starts to develop as SST exceedsa evaporationshouldincreasewith risingSST on the basisof the thresholdvalue. As SST keepsincreasing,so does convection. Clausius-Clapeyron eftk•ct.This assumptionignoresthe role of More intenseconvection results in the generationof optically thick cirrus clouds that cover a large area. The clouds act as shields and keep the solar radiation from reaching the sea surface.Hencetheyresultin the coolingof the seasurthce. The "them•ostathypothesis" hasprovokedmany interesting discussionsand debateson what processescontrol SST and
the large-scale circulation as a ventilation thctor and its dependenceon the SST. In fact,.the large-scalecirculation is
tightlyrelatedto the SST in the tropics. For instance,Bony et al. [1997b] show that the m•ioritv of the large-scaleupward motionover the tropicaloceansoccursat SSTsabove300 K. In a recentstudy, Zhang and •lcPhaden [1995] (hereinafter
especially whatlimitsSSTto theobserved upperlinfitof ahnost referred to asZM95)perfonned anextensive analysis of thedaily 305 K. Among themis one by IVallace [1992], accordingto surfacedata fromthe tropicalatmosphere-ocean (TAO) moored which evaporativecooling is a morecompetitivemechanism buoysdeployedin the equatorialPacific (137ø E to 95ø W and regulatingthe tropical westernPacific SST. "Hot patches"of 8ø N to 8ø S) from July 1991 to June 1993. Using bulk SSTresttitin moreheat exchange(throughsensibleand latent aerodynmnictbnnulasto computesurfaceheat, moisture, and heat fluxes)betweenthe oceanand the atmosphericboundary momentran fluxes,they found that surfaceevaporationis not monotonicallyrelated to SST as was assumedin previous studies. Surthceevaporation, instead, increaseswith SST Ibr Copyright1998 by theAmericanGeophysical Union. SSTsup to approximately301 K and decreases thereafteras SST keeps increasing. Similar behavior was observedbetween the Papernumber1998JD200027. 0148-0227/98/1998JD200027509.00 wind speedand SST as well. 31,763
31,764
TSINTIKIDIS
AND ZHANG: SST AND SURFACE EVAPORATION COUPLING
The variationof evaporationwith SST and, in pa.rticular,the 2. fact that evaporationdecreasesin the high SST regimecan result
Model
from differentprocessesinvolving the atmosphere-ocean Themodelusedh• thisstudyis a modifiedversionof Wang coupling. Bettsand Ridgway[1989] studiedthe coupling andLi [1993].In thevertical,it is composed of a two-layerfree between theoceansurfaceandthe atmospheric boundarylayer. troposphere anda well-nfixed boundarylayer. Thecirculation They used a one-dimensional thermodynanfic modelIbr the in the free troposphereof tl•e model is driven by the atmospheric boundarylayeranda one-dimensional radiative thenn0dynamic forcingof theSSTthroughconvection andby modelto esthnate the radiativefhrxestlu:ough the boundary thelatentheatreleasefromconvection followingGill's [1980] layer,together withtherequirement of theenergy balance at the dynamic model.Theboundary layerflowis drivenby theSST ocean surface. Theyfoundthatwhentheatmospheric modelis gradientfollowingLindzenand Nigam [1987]. The free
decoupled fromthe SST?surfaceevaporation increases atmosphere andtheboundary layerarecoupled bymatching the continuously withSSTtbra givenwindspeed.However, when flowfieldattheirinterface. A desired feature of thisapproach is SSTis coupled with theiratmospheric model,thatis, SSTis thatit resultsin a morerealisticsimulation of the tropical determined by the energybalanceat the oceansurfbee,circulation [WangandLi, 1993]. evaporation decreases with increasing SST(seetheirFigure Themodelislinearized tothebasicatmospheric stateat rest 20). Thisisbecause lessevaporation leadstomorenetheatflux that is consistent with a uniforn•SST distribution.The intotheoceansurface, resulting h•higherSST. linearized governing equations onan equatorial [• planeare Sui et al. [1991] furtherinvestigatedthe coupling between
the atmosphere and the ocean boundary layers using Betts and c)u c)(p _exU +rV2u Ridgway's [1989] atmospheric model and a one-dimensional -•-=/Syv--•-
(1)
ocean-mixed-layermodel. The authors emphasizedthe upper
oceanthem•odynamic processes in affecting the relationships
between the SST and the surface heat fluxes. They tbund that when the two modelsare coupled,with SST changesdetemfined
•
&p
"•'=-/Syu--•--eyV+ rV2v
(2)
bythe ocean-mixed-layer heat budget, sur12ace evaporation 1&pd(b'B 1)V V•(16I)•.V....... decreaseswith SST.Also, surfiace evaporationchanges,which
affects theSST, in response to theprescribed windspeed
c02 &
change. Inaddition, the ocean mixed layer depth and the
(3)
CEIWI 2½ h
upwelling at thebottom of themixed layercanalsochange,
•-Nrp+r2V
therebyaffecting the SSTas well. While the abovework
providesan importantinsightinto the local couplingbetween
theatmosphere andtheocean andtheunderstanding oftheair-
ø•u•
&P
o•rs
sea interaction, the one-dimensional model cannot address the Ot=/Syv•--•--e!•xU• +Ac)X +rV2u•
(4)
nonlocalcouplingbetweenthe atmosphereand the ocean. The
atmospheric circulation (and hence the surface winds) ishighly ø•(P ø7I's dependent on theSST and itshorizontal distribution, and •• --fiYU•-"•--eI•yV•+A • +rV2vB
(5)
consequently, prescribingsurfacewind speedslimits the degree
of thecoupling between theatmosphere andtheocean. where V = (u,v)andVs = (us,rs) arethelowerfreetroposphere In theirobservational study,ZM95suggested a mechanism andboundary layerwindvelocities, respectively, andq0is the thatcanexplain theobserved relationship betxveen theSSTand lower-troposphere geopotential perturbation; œ= (•;•,œy), and thestirface evaporation: in a largeregionanSSTincrease will œs= (ess-, œs•.) aretheRayleigh I¾iction coefficients forthelower
result in a moreconvectively unstable atmosphere. Theensuingfreetroposphere andboundary layer,respectively; d = (p•-p,) / convection leadsto large-scale upwardmotion. By massApisthenondimensional depth of the,boundary layer;B,!, and continuity,the upwardmotiongivesrise to the low-level F areheatingcoefficients (seebelowformoredetailson their convergence. At thecenter oftheconvergence regionthewind tornrelation); andTsis theSST.Thereader is alsodirected to speed is low,leading to lowevaporation. Whiletheworksby Table1 foradditional explanations andvalues of constants. bothBettsandRidgnvay [1989]andSttietal. [1991 ] locuson Theprecipitation P,.isestimated bythesumofthevertically theSSTresponse to thesurtSace andupperoceanheatbudget, integrated moisture convergence andlocalevaporation. It is thismechanism emphasizes theatmospheric response to theSST represented by theõ termsin equation (3), whereõ represents a forcingandpertainsto long thnescales. To investigate the role of the SST forcingon the surfacewind and evaporation, and its contribution to the observed
SST-dependentconditional heating' 1,
Ts > 301.5 K and P,. > 0
dependence of surfaceevaporationon SST, in this study, a d = • (Ts- 298.5)/3,298.5K < Ts< 301.5K, andP,.> 0 (6) dynamictropical atmosphericcirculationmodel is used,with 0, otherwise emphasison the coupling amongSST, surfaceevaporation, convection,and large-scalecirculation. We will focuson the Latent heat is assumed to be releasedin precipitatingregions abilityof themodelto reproduce the observed relationships of only. Thelatentheatreleaseis parameterized usinga simplified SST with evaporationand wind speed of ZM95, and the Kuo [1974] scheme(in which the rate of latent heating is responseof the systemto SST anomalies.In section2 the model proportional to total precipitation) and the observational is briefly described.Section3 describesthe simulationsand evidenceof Wallset et al. [1993]. The latter authorsshowed results. Finally, section 4 provides a brief sintartary and the thatwhenSST is lesstlmn25.5 øC deepconvection rarely takes
conclusions of the study.
placewhereas for SSTsbetween 25.5øCand28.5øCthe amount
TSINTIKIDIS AND ZHANG: SST AND SURFACEEVAPORATION COUPLING
31,765
Table 1. Parametersand their valuesduring simulations ,
Symbol
Value
Parameter ,
p, pe pu p2 tip
pressureat sea surface pressureat the top of the boundarylayer pressureat the top of the free troposphere pressureat the middle of the fi'eetroposphere half-pressuredepthof the free troposphere
1000 hPa 900 hPa 100hPa 500 hPa 400 hPa
R
cp
specific heat atconstant pressure gas constantfor d•y air
287 J deg' kg'
L½
latentheatof condensation
2.5x 10'6Jkg4
L• Co
latent heat ofevaporation dry'gravity wave speedof the baroclinicmode
2.5 x1.?'6 Jkg4 50 m s
r
horizontalmomentumdiffusioncoefficient
1061112 s'•
So
staticstability parameter atseastuTace
7.85x 10'6m2s'2Pa'2
1004 Jdeõ '•k.• '•
b
efficiency of condensationheating
0.75
N
NewtonJan cooling coefficient
2 x 10-• s-•
d
nondimensionalboundarylayer depth
0.25
C, œ,.œ•,
heatexchange coefficient tYeetroposphere friction coefficients
1.5x 10'3 10'6
H Hi h
air densityscaleheight water vapor densityscaleheight ball depth of fi'ee troposphere
7.8 km 2.2 km 3.9 km
q½
vertical meanspecific lmmidity inthe
1.17x 10'2
lower troposphere coefilcientof SST gradientforcing
5.55 x 10'2
of highly reflective clouds is a linear fitnction of SST. The frequencyof occurrenceof deepconvectionreachessaturation when SST is higher than 28.5 øC. Surfaceevaporationis parameterized usinga bulk aerodynamic Ibnnula(seesection3). The various heating intensity coefficientsare defined as: I=q3/q•,heating coefficientdue to wave convergence,B=q½/q•, heating coefficientdue to Ihctional convergence,and F=(q.•q0)/q•,heatingcocflicientdueto evaporation.The quantib,q•,is the specificlmmidity at the sea surliqcetemperature,it can be obtainedby the Clausius-Clat•eyron equation,and it is a linear
For the values of the constantsthat appear in equations (7)(10), seeTable 1. As can be seen,the thermodynamic equation (3) includesthree mechanisms represented throughthe heating intensitycoefficients:convection-wave convergence throughI, evaporation-wind feedback through F, and convectionfrictional convergencefeedbackthrough B. In addition, the moisturecontents q3,qc,q•, q0arefi•nctions of SSTand therefore the coefficients I, F, B are directlycontrolledby SST. The above formulationis based on a three-forcebalance: Coriolis,pressuregradient,and frictional Ibrces. Li amt lVang
functionof SST [seeIFaug amt Li, 1993]; q• and q3 are the (1994a) concludedthat the use of anisotropic,latitudeverticalmeanspecifichumidities in the boundary' laycrandthe dependentfriction coefficientsare necessapy to successfully lowertropospheric layer,respectively.lFcmg[1988] showed shnulate surfacewind fieldsin a three-force balanceboundarythat they can be calculatedbv
layer model. Consequently,their meridional, œey, and zonal, œex,coefficients were used in this study. For the free (7) troposphereconstantvalue coefficientswere used(see Table 1). The model has been used successfidlyto study a variety of issues relevant to the tropical atmosphericcirculation. For (8) instance,Li aml IFang [1994a] used it to study the annual cycle of the sealevel pressure,surfiqce winds and precipitation. wherem is theratioof the air densityscaleheightH to the They also usedit to model the circulation anomaliesassociated watervapordensityscaleheightH•. The surface air specific with the E1 Nifio/La Nifia events. Also, Li a,d IFaug [1994b] humidity qo is assumedto be a fi•nction of SST based on the fitrtherappliedthis model to study the influenceof the SST on originalCOADS(Comprehensive Ocean-Atmosphere DataSet) the tropical intraseasonaloscillations.
datasetandis givenby 3.
qo = (0.972 Ts- 8.92)xl 0-3.
Numerical
Simulations
and
Results
(9)
A seriesof simulations were perlbnnedusing various SST
Formoredetails onthederivations of(7) - (9) seelFcmg [1988] fieldsto drivethe1nodel: (1) annual climatological SSTfield,
and[Fang andLi [1993].Finally, q• stands forthevertical(2)SSTfieldforJuly1987representing anE1Nifiocondition,
mean specific humidity in the lower-tropospheric layerand(3)SSTfieldIbrJanuary 1989representing a LaNifia necessary toproduce a vanishing effective static stability in the condition. These threeSSTfields werechosen to exambin the
presence ofwave-convergence-induced heating. It isprovided relationship between theevaporation and theSST distribution ß The SST fields are shown in Figure 1. The climatological SST qc=2Cpp 2 CO 2/(bRApLc ) (10)field isthesame asthe one used byLiamt IFcmg [1994a] and is
by
31,766
TSINTIKIDISAND ZHANG:SSTAND SURFACEEVAPORATIONCOUPLING Climatological SST field
120
140
160
180
200
220
240
260
280
Longitude, degrees
July 1987 SST field b
6
.1
120
140
29.1
160
180
200
220
240
260
280
Longitude, degrees January 1989 SST field c
30 24
18 12
-12 -18 -24 -30
120
140
160
180
200
220
Longitude, degrees
240
260
280
TSINTIKIDIS AND ZHANG: SST AND SURFACEEVAPORATIONCOUPLING 320
31,767
Figure 2 showsthe relationshipbetweenlatentheat flux and SST for each of the SST fields.
The latent heat flux F is binned
into SST intervals of 1 K each. Overall, latent heat flux
240
increaseswith increasingSSTup to about 300-301 K. From thenon it decreases as SST keepsincreasing.TheE1 Nifio and La Nifia monthpatterns(panels2b and 2c, respectively)are similar. However,the latentheatfitix duringtheE1Nhio month is larger than whereasthe La Nifia month latent heat fitix is smaller,than of the climatologicalSST. The relationships shown in Figure 2 qualitatively resemblethe resultsof ZM95 (seetheirFigure6) with panel2a exhibitingthe largestdegree
+
+
160 80
+
+
+
+
+
+
+
+
320
,
+
ß
of sinfilarity. The peaklatent heat flux of about 140 W/m2 occurs at about300.5K compared to ahnost120W/m2at 301 K
,
+
of ZM95. Our results also appearto be consistentwith the resultsof Bon), et al. [1997a], who comparedthe atmospheric reanalysesfromthe Data Assimilation Office (DAO) of the GoddardLaboratory.for Atanospheres and the National Center lbr Enviromnental Prediction(NCEP). They foundthat in the
b +
240 +
+
160 +
+
+
+
80
+
+ + +
20
e
15
320 E
+
+
+ +
e 10
+
240
+
.__
160
•
+ +
+
+
+
5
+
+
+ +
+ +
+
0
294
'
+
+
+
+
+
+
2•6
+
20
298
300
302
304
SST, degrees K
Figure 2. Latentheatflux, F, asa fi•nctionof SST tbr eachof the
SSTfieldsshownrespectively in Figtire 1. The solid line represents themeanand• signsrepresent (1 o fromthemean
+
ineach bin.Notethelarge o values indicating largevariability in latent heat flux.
+ +
,+
+
+
+
based on the COADS
data set. The SST fields for E1 Nifio
and
La Nifia conditionsare obtainedfrom the ReynoldsSST data set [Rey,oMs amt Smith, 1994]. It can be seenthat the maxbriton SST valuesin the E1Nifio monthare locatedeast(by about30ø35ø)comparedto the climatologicalm,xxhmun location. The La
2o
Nifiamonth SSTn•x, chntun is located southxvest (byabout10ø) with respectto the climatology. The model domain
is from 30øN
+
to 30øS and
120øE
+
+
+
* •0
+
+
to
100øW, with a horizontalresolutionof 2ø x 2ø. The integration
.__.
thne step is15min, and notime smoothing isapplied tothe • 5 solutions. Each simulationis mn for 30 days, and it takes less than 10 days of thne integration for the model circulation to reachan equilibriumstate. ThereIk)re the results shoxvnbeloxv, which are the last 5-day averagesunless otherwise stated, representthe equilibriun•statefor a givenSST lk)rcing.
+
+
0294
+
+
+ 296
+ 298
+ 300
+
•+
302
SST, degrees K
Figure 3. As in Figure2 but for wind speed.
Figure 1. The SST fieldsusedhi this studyto drive the simulations. Top panel: Climatologicalcase. Middle panel: July 1987 case(El Nifio month). Bottompanel: January1989 (La Nifia month). All fields are degrees Centigrade.
+
3O4
31,768
TSINTIKIDIS AND ZHANG: SST AND SURFACE EVAPORATION COUPLING
300
understoodfrom the bulk equationfor latentheat flux:
25O
F = pLvCeU[q s(Ts)-q01 2OO
+
(11)
+
xvherep is the surt•ce air density, L,, is the latent heat of
150
evaporation, qs*(Ts)isthe saturation watervaporspecific humidity, U is the wind speed,and Ce is the heat exchange coefficient,which is setto a constantin the model(seeTable 1 for values of constants). The teml in the brackets is a linear functionof SST (seeequation(9)). Equation(11) indicatesthat the latentheat flux dependson both SST and wind speed. Note that for a given SST,the variability of latent lieat tlux is very significant(Figure2). This hnpliesthat factorsother than SST, i.e., atmosphericdynamicstlu:oughlow-level winds, can have a significantinfluenceon surfaceevaporation. T1]eresults of a linear regressionof the points of Figures 2 and 3 are shownin Table 2. The computationof the regressionbasedslopeswas doneafterBonyet al. [1997b], who computed a "sensitivityindex"I, to aid themin the study of interannual variationsin atmosphericand radiative variables. The index is given by
100 50
3OO -t-
25O
2OO
+/•+ +
150 100
+ +•.••+ + +
+ +
+
50
=
Z(axaSST)
3OO 25O c
2OO
150 100 50
o
15
E(ASST) 2
(]2)
wherex is a genericvariable(in our caselatent heat flux F and wind speed W). The above formula essentially provides the linear regressioncoefficientof the least squaresfit to (x = IxASST). The ZM95 slopes(seetheir Table 4) fall within the range of our coefficients. Overall, it can be seen that the negativeslopes(for SSTslargerthan 300 K) vary muchmore than the positive slopes (for SSTs less fi•an 300 K). In addition,the negativeslopesare steeperthanthe observedones of ZM95 by up to a factor of 4. The discrepancy,at least partially, can be attributedto the fact that the calculationswere
performed for eachmonthseparately, as opposedto averaging
Wind speed, m/s
Figure 4. Latent heat fluxasafunction ofwind speed foreach of overa period of 2 years (which include E1Nifioeffects).
thecases shown inFigure 1. Forsolid lineand'+' explanation Finally,it is expected that modellimitations mayhave seeFigure2 caption.
contributed to the discrepancy as well (.seeDiscussionsection
abornmodel limitations). Figure 5a shows the differentfields of SST, evaporation NCEP analysis,variation of surfaceevaporationwith SST is (Figure5b), and wind speed(Figure 5c) betweenE1 Niiio and similar to the observations of ZM95. clhnatology.In the regionsof negativeSSTanomaliesover the The wind speedpatterns(Figure 3) exhibit shnilarbehavior tropicalwesternPacific,surfacewind speedanomalies arelarge to their latent heat flux counterparts shownin the previous and positive. On the otherhand,in the positive SSTanomaly figure,indicating thatlatentheatflux is stronglyinfluenced by regionseastof the dateline,surfacewind speedanomaliesare
windspeed.Indeed,in Figure4, whichshowsthebehaviorof negativeor relativelysmall. The surface evaporation in the thelatentheatflux asa fimctionof wind speed,the dependencemodeldomainis increased duringthe E1Nifio monthwith the of latent heat flux on wind speedis clear. This can be increase beh•gthelargesthi the negativeSSTanomaly regions
Table 2.
LinearRegression SlopesWith Respectto SST SST
i•F/i•T,,W m'2K4, climatology i•F/i•T,, W m'2K4, July1987 •F/•Ts,W m'2K'•, January 1989 i•U/i•T,, ms'l K'l, Climatology •U/•Ts,m s'l K4, July1987 •U/•Ts,ms'l K'l, January 1989 (F -- LatentHeatFlux,U --
Wind Velocity)
< 300 K
>300 K
10.5
-40.0 -33.6 -11.2
18.9
7.0 0.7
-3.3
1.3
-2.6
0.5
-1.9
(July1987SST)- (Climatological SST)
31,769
30-
24 18 12
0 -12 -18 -24 -30-
120
140
160
180
200
220
240
260
280
260
280
Longitude, degrees
(July 1987 Evaporation) - (Climatological Evaporation)
120
140
160
180
200
220
240
Longitude, degrees (July 1987 Wind Speed) - (Climatological Wind Speed) c
•
6
•
-6 -12 _.
-3O 120
• 140
160
180
• 200
220
240
260
280
Longitude, degrees
Figure 5. Panela: Differencefield betweenJuly 1987(El Nifio) SSTandclinmtologySST. Panelb: asin pemcl a but for latentheatflux. Panelc: as in panela but Ibr wind speed.
31,770
TSINTIKIDISAND ZHANG:SSTAND SURFACE EVAPORATION COUPLING (January 1989 SST) - (Climatological SST) 3024-
1812-
-12 -
-18 -24 -
-30 .20
140
160
180
200
220
240
260
280
Longitude, degrees
(January 1989Evaporation)- (ClimatologicalEvaporation)
-12 -18 -24 -
-30 -
120
140
160
180
200
220
240
260
280
Longitude, degrees (January 1989 Wind Speed) - (Climatological Wind Speed)
24
18
o •
-6 -12 -18
-30
120
140
160
180
200
220
240
260
280
Longitude, degrees
Figure 6. As in previousfigurebut behveenJanuary1989(La Nifia) andclhnatology.
TSINTIKIDIS AND ZHANG: SST AND SURFACE EVAPORATION COUPLING
31,771
(Gaussian Perturbation Evaporation) - (Climatological Evaporation) 30 24 18
•
6
o•
•
. -12
-24 -•0
120
140
160
180
200
220
240
260
280
Longitude,degrees
(GaussianPerturbationWind Speed)- (Climatological Wind Speed) b 3024-
1.6
0.7-------• 2.5
18
3.
1.•
12
o -12 -18
-24
120
140
160
180
200
220
240
260
280
Longitude, degrees
Figure 7. As in Figure4 but for the artificialGaussian perturbation andclimatology.
the interconnectionsalnong SST, latent heat flux, and wind speed, additional simulations were conducted. The new simulationswere lnotivatedby' the work of [Valiser [1996]. In an attelnptto" . understandwhy in the contextof high SST the evaporativetEedbackgivesway to the shortwavefizedbackin a local sense,yet why it is necessaryto considerthe lar•,e-scale and remote effects of the atmosphere " I'Faliser [ 1996] Consider an idealized sign, +' ' proposed a ...... •,,• exi•rhnent: m,xmca•t of the equator and •,,,,n•es "•'" '• situationin which the atmosphereoverlies an oceanof uniform behaviorof the othertwo fields is opposite. The differences in the aforementioned fields tbr the La Nifia temperature,it' we "hiseft" a large-scalepositive temperature monthare shown in Figure 6. The pattenis are similar to the anomaly in the ocean-surthcemixed layer, how does the previouscase,but the spatial extentand locationsare altered. atmosphererespond? In the region of warmerSST the systeln In the latter case the differences in latent heat flax and wind will initially' try. and equilibrate through enhanced surthce speedare smaller,resultingt¾oma generallycoolerSST tbrcing. fluxes of latent and sensible heat. These enhanced fluxes will So fi•r, we have shown that latent heat flux is strongly initiatea low-levelatlnospheric convergence over the regionof SST "which would reducefile surfacewind speed influencedby surl}•cewind speed,which hi turn is highly maxilnmn dependent on the SST distribution.To gain furtherinsightinto and thus the surfiaceevaporation there. This would be
andthe smallestin the positiveSST anomalyregions. Figure 5 presents, essentially,the spatialefibctsof the variousSSTfields on surfacewind and evaporation.All three panels suggestthat negativeSST anomalycorresponds to excesslatentlieat flux and wind speed, i.e., note the general reduction of SST deficit southwest of the equatorand the generalincreaseof latent heat flux andwind speeddiftZrences.As the SST anomalybecomes
31,772
TSINTIKIDIS
AND ZHANG: SST AND SURFACE EVAPORATION
COUPLING
Waliser's Hypothesis,220, O, 20.5, 5.5 10
i
i
, o (E)
(C)
•,._...•_•(W)
0
10
15
20
25
3O
Time, days
Figure8. Thneseriesof evaporation at 5 selectedgrid points(N(North),S(South),C(Center),E(East), W(West)).Thehorizontal andverticalwidthsoftheGaussian SSTperturbation were20.5 and5.5 degrees, respectively.
of fl•e SST perturbationare consistent with the results of ZM95. To quantitatively test at suchlocationswhere the effiacts Waliser's meclmnism,an artificial Gaussian SST perturbation quite pronounced. hmnediately after the SST perturbation is was superimposedon the climatological SST field sho•vn in inserted, latent lieat flux is increaseddue to the SST et'l•ct on the Figure la. The perturbationvariedin magnitude(1ø or 2øC), saturationhumiditydeficit of the surfaceair. On the other hand, spatialsize, and locationand was abruptly "h•scrted" afterthe the surfacewind speed "I&ls" the SST perturbation more systemreachedan equilibrium, i.e., after about 20 days. The gradually through the response of the circulation to the model was mn until the new equilibriumwas reached. pammeterizeddiabatic heating. After a few days the system The different fields in surfacewind and evaporationresulting beginsto approachthe new equilibrium state,and the changes from a GaussianSST of 2øCat peak, centeredat the equator and of the latent heat flux and wind speed from the previous 140ø W, are shown in Figure 7. Surfiqcewind speed and equilibrium values are governed mostly by the dynamic evaporation are increasedat the center ,'redto the west of the response. For instance,at the center of the SST perturbation, positive SST perturbation but are decreasedat the periphery evaporationis significantly lower at the nexvequilibrimn due and to the east of the SST perturbation. Thesechangescan be to the reduced surface wind speed in response to the SST explained by the response of the tropical circulation to a perturbationeven though SST is 2øC warmer. Wallace [1992] perturbation heating source [e.g., Webster, 1972; Gill, 1980]. suggestedthat when a "hot patch" appears,due to the SST When a SST perturbationis superimposed on the climatological effect on the saturationdeficit of the surfaceair, the surfacelatent SST field, tlu'oughconvective paramctcrization,an anomalous lieat flux would hacrease locally. This is not supportedby our convectivelatent heat releaseis assumedover the region of the numerical experiment. •Ille temporal behavior of surlYace positive SST perturbation,which would result in a convergent evaporationand wind speedis in support of the mechanismof flow toward the SST perturbation center in the loxver IFaliser [1996], and suggeststhe hnportanceof the coupling troposphere. When this perturbation flow pattern i s between the surface latent heat flux and the large-scale superimposedon the backgroundeasterly 11owcorresponding circulation. to the climatological SST field, it results in a decreasedtotal wind at the center and to the west of the positive SST 4. Sun]mary and Conclusions perturbationand an increasedtotal wind elsewhere. It should A dynamictropicalatmospheric circulationmodelwas used be mentionedthat qualitatively, similar results are obtained as the Gaussian SST perturbation varied in magnitude, location, to investigatethe relation amongthe sea surfiacetemperature, the large-scalecirculation,and the surt2•ce evaporation. The and spatial extent.
Figures8 and9 showthe temporal behaviorof theresponsemodelconsistsof a two-layertroposphere and a well-mixed of thesystem at fiveselected gridpoints.Thepointsarechosenboundary layer. Interactions areallowedbetween theboundary
TSINTIKIDIS AND ZHANG: SSTAND SURFACEEVAPORATIONCOUPLING
31,773
Waliser's Hypothesis,220, O, 20.5, 5.5 16
i
i
i
220, lO, (N)
240, 0 (E)
12
220,-10 (S) 220, 0 (C)
200, 0 (W)
0
5
10
15
20
25
30
Time, days
Figure 9. As in previousfigurebuttbr windspeedandfor the same5 points.
layer and the freetroposphericflows (both of themdepend on clhnatologicaldata. For E1Nifio or La Nifia conditions,it may SST). Various SST fields (climatology,E1 Nifio, and La Nifia not representthe bestfit to the data. Thus the sinrelatedstirface months)were used to tbrce the model. Also, an artificial evaporationin these casesmay be affected. Secondly,the Gaussianperturbation was superimposed on the climatological precipitationswitch-onschemeis rather simple,and it maynot
SST field in orderto investigatethe system'stemporaland representaccuratelythe rainfall genesisprocesses.Also, the spatialresponseto SST anomalies. modelignoresthe barotropicmodein the freetroposphereand The results indicatethat for loxv SSTs (_301K) the atbrcmentioned B,F). Finally, theterrain effectisalsovery. crudely represented,
quantities decrease asSSTincreases. Themagnitude ofSSTi.e.,tlu'ough tlu-ee values representing high,]ncdium, andno (i.e.,cold versus warm months) affects only themagnitude ofthe future. landelevations. These issues willbeclosely examined inthe
pattern of the aforementionedquantities and not the pattern itself. In addition, the binned patternsmatchthe observations
Acknowledgments. We wish to thank Bin Wang for providingus with
ofZM95. Agreement also withthedata reanalyses findings of his tropical atmosphere circulation model. DTalso thm3cs Konstantine
Bony etal. [1997a, b]gives credence toourmodeling approach. Georgakakos formany stinmlating discussions during thecourse ofthe
The results in the loxv SST regime can be explainedby project. This researchwas financiallysupported by a CalSpace
them•odyn,'mficconsiderations. The results in the high SST minigrant, NSFgrantATM95-25800,andNASAgrantNAG 5-2794. regime appear to be counterintuitive to thermodynamic
considerations, butthelattercanbeexplained if oneconsiders References atmosphericdynanficarguments,as proposedby ZM95.
Whenanartificial Gaussian SSTperturbation is suddenly Betts, A.K.,andW.Ridgway, Climatic equilibrium oftheatmospheric
appliedto the coupledocean- atmosphere system,it tries to
convective boundary layerovera tropical ocean, d. Atmos. Sci.,46,
account forthe changetlu-ough enhancement of latentheatflux
2621-2641, 1989.
initially mainly inthearea oftheSSTperturbation. However, Bony, S.,Y.assessment Sud, K.M. Lau,J.Susskind, andS.Saha, Comparison and ' satellite of NASA/DAOandNCEP-NCARreanalyses over
thedynanfic effect tlu-ough theresponse ofthecirculation tothe tropical ocean: Atmospheric hydrology and radiation, d.Clim., 10(6), SSTperturbationbecomessignificantafterward. Overall, the 1441-1462, 1997a. behaviorof the evaporationevolutionat and aroundthe center Bony,8., K. M. Lau,andY. Sud,Seasurface temperature andlarge-scale
oftheSSTperturbation supports l•raliser's [1996]mechanism. circulation influences on tropical greenhouse effectandcloud radiativeforcing,J. Clim., 10(8), 2055-2077,1997b.
Finally, it should bementioned thatthere areseveral factors Gill, A.E.,Some simple solutions forheat-induced tropical circulation,
limitingthemodelperformance. First,theempirical tbmmla tbr
Q.J.R. Meteorol. Soc., 106,447-462, 1980.
the surfaceair humidity(equation(9)) is basedon the COADS Hartmann, D. L., andM. L. Michelsen, Large-scale effectson the
31,774
TSINTIKIDISAND ZHANG:SSTAND SURFACEEVAPORATIONCOUPLING
regulation of tropicalseasurface temperature, J. Clim.,6, 2049-2060,Waliser,D. E., N. E. Graham, andC. Cautier, A comparison of thehighly 1993. reflectivecloudandoutgoing longwave radiation datasets for use in Kuo, H.-L., Furtherstudies of the parameterization of the influenceof estimating tropicaldeepconvection, J. Clim.,6, 331-353,1993.
cunmlus convection onlarge-scale flow,J.Atmos. Sci.,31, 1231-1240,Wallace,J. M., Effectof deepconvection ontheregulation of tropical 1974. seasurface temperattire, Nature,357.230-231.1992.
Li, T., andB. Wang,A thermodynamic equilibrium climate model for Wang, B.,andT. Li.A simple tropical atmosphere model of relevance to monthlymeansurfacewindsand precipitation over the tropical
pacific, J.Atmos. Sci.,51,1372-1385, 1994a.
Li, T., andB. Wang,Thei•ffluence of seasurfacetemperature on the
short-term climate variations, J. Atmos. Sci..50.260-284,1993.
Wang, B.,Thedynamics oftropical lowfi'equency waves: Ananalysis of moistKelvinxvaves, d.Atmos. Sci.,45,2051-2065, 1988.
tropical intraseasonal oscillation: A numerical Study,_•lon. WeatherWebster, P. J..Response of thetropical atmosphere to localsteady
Rev.,12,2349-2362, 1994b.
forcing, 31on.I4.,'•ather Rev.,101,803-816, 1972.
Lindzen, R. S.,andS,Nigam, Ontheroleof seasurface temperature Zhang, G.J.,,'red M.J.McPhaden, Therelationship between seastirface gradients inforcing low-level winds andconvergence inthetropics, ,/. temperature andlatent heatfluxintheequatorial pacific, J. Chin., 8, Atmos.Sci., 45, 2440-2458, 1987.
589-605,
1995.
Ramanathan, V., andW. Collins,Thernmdynamic regulation of ocean warming by cirrus cloudsdeducedfi'om observationsof the 1987 El Nifio, Nature, 351, 27-32, 1991.
Reynolds, R. W., and T. M. Smith,Improvedglobalsea stirface temperature m•alyses usingoptimuminterpolation, J. Ctim.,7, 929-948,
D. Tsh•tikidis,IlydrologicResearchCenter, 12780 ttigh Bluff Drive, Ste250, SanDigo,CA, 92130. (e-mail:
[email protected]. edu) Sui, C.-H., K.-M. Lau, and A. K. Betts. An euilibriummodel for the G.J. Zbang, Scripps h•stitutionof Oceanography,University of coupled ocean-atmosphere boundary layerin thetropics. J. Geophys.California.SanDiego, La Jolla,CA, 92093. 1994.
Res., 96, 3151-3163, 1991.
Waliser,D. E., Someconsiderations onthe thermostat hypothesis, Bull. (Received November 1•, 1997: revised June 17, 1998: Am. )lYeteorol.Soc., 77, 357-360, 1996.
acceptedJune 24, 1998.)
