H igh sea surface temperature (SST) is generallyrequired for deep convection that reaches thetropopause. In the current climate, the SST
threshold for deep convection is somewhere around26°–27°C, depending upon region and season (Gra-ham and Barnett 1987; Waliser et al. 1993). Over suchwarm oceans, SST changes cause deep convective ad-justment, which in turn excites the dynamic responseof predominantly first baroclinic mode structure withstrong surface wind signals. This robust relation be-tween SST, deep convection, and wind has led to therapid advance in understanding and modeling the
tropical ocean–atmosphere interaction that gives riseto phenomena ranging from the El Niño–SouthernOscillation (ENSO; Neelin et al. 1998) to the northwarddisplacement of the intertropical convergence zone(ITCZ; Xie and Philander 1994; Philander et al. 1996).Atmospheric general circulation models (GCMs) showconsiderable skills in simulating this deep adjustmentto SST anomalies associated with ENSO (e.g., Alexanderet al. 2002).
Over cool oceans where deep convection does notoccur,1 the atmospheric adjustment to changing SSTdiffers markedly from that over warm oceans. Coolocean–atmosphere interaction is poorly understood,and this lack of understanding is a stumbling blockin the current effort to study non-ENSO climate vari-ability. Unlike their success in simulating the South-ern Oscillation, atmospheric GCMs disagree amongthemselves in their atmospheric response to SSTanomalies in the extratropics (see Kushnir et al. 2002for a review).
*International Pacific Research Center Contribution Number 239and School of Ocean and Earth Science and Technology Contribu-tion Number 6261.AFFILIATION: XIE—International Pacific Research Center andDepartment of Meteorology, University of Hawaii at Manoa,Honolulu, HawaiiCORRESPONDING AUTHOR: Shang-Ping Xie, IPRC/SOEST,University of Hawaii at Manoa; 1680 East West Road, Honolulu, HI96822E-mail: [email protected]: 10.1175/BAMS-85-2-195
New observations from space reveal surprisingly robust patterns of air–sea coupling over cool
oceans where such coupling has been thought to be weak.
1 While the SST threshold is a convenient way to divide the warmand cool regimes for air–sea interaction, local SST is not theonly factor for deep convection, which is also influenced byother factors such as large-scale subsidence. It may be morephysically appropriate to divide the warm and cold regimes ac-cording to whether there is significant deep convection or not.
196 FEBRUARY 2004|
Observational efforts so far have failed to yield con-clusive evidence of SST effects on the planetary-scaleatmospheric circulation over cool oceans because of ahigh level of weather noise in the extratropics on onehand, and short and sparse in situ observations on theother. Most often negative correlations between SSTand surface wind speed variability are observed in theextratropics for seasonal means and on the basin scale(Fig. 1). Researchers have viewed this as being indica-tive of one-way forcing from atmosphere to oceanthrough wind-induced changes in surface turbulenceheat flux (sidebar 1).
To achieve an adequate sample size in a grid box,climate datasets typically have resolutions of severalhundred to 1000 km and of a month to a season.Major ocean currents like the Kuroshio and GulfStream are only around 100–200 km wide and formsharp SST fronts that are poorly represented in thesedatasets. It is on these narrow oceanic fronts, however,that ocean dynamics become important and causelarge SST variations. Thus, conventional climatedatasets may severely underrepresent such dynami-cally induced SST anomalies and their atmosphericeffect.
The vast oceans can only be sparsely observed bytraditional ship-based measurements, but the rapidadvance in space-based microwave remote sensing isrevolutionizing ocean observations, providing globalfields of key ocean–atmospheric variables at unprec-edented resolutions in space and time. Unlike visibleand infrared remote sensing, microwave measure-
ments are unaffected by clouds except by those withsizable precipitation. Based on microwave remote-sensing data, several recent studies have identified SSTvariations induced by ocean dynamics, and they havemapped the effects of these variations on the atmo-sphere (Xie et al. 1998; Wentz et al. 2000; Cheltonet al. 2001; Xie et al. 2001; Liu 2002; White and Annis2003; O’Neill et al. 2003; Vecchi et al. 2004). Thesenew satellite observations reveal a rich variety of pat-terns of ocean–atmosphere coupling in the Indian,Pacific, and Atlantic Oceans, from the equator to themidlatitudes. Because individual papers tend to focuson one particular phenomenon in one particular re-gion, the full spectrum of cool ocean–atmosphereinteraction has not been identified.
The present paper synthesizes these recent stud-ies and compares atmospheric response over differ-ent ocean conditions in different parts of the WorldOcean. By assembling these observational facts, wewish to see whether there is a common atmosphericresponse pattern, and if so, by what mechanisms andunder what conditions this response takes place. Asynthesis of similarities and differences in the atmo-spheric adjustment over different regions of theWorld Ocean can help shed light on the dynamics ofcool ocean–atmosphere interaction and stimulate fu-ture studies.
Over cool oceans, the direct effect of SST variationsis likely to be trapped within the planetary boundarylayer (PBL), the lowest 1–2 km of the atmosphere,because this layer is capped by a stable layer, often in
FIG. 1. SST–wind relation in the North Pacific and Atlantic Oceans. (left) COADS SST (color shade),surface wind vectors, and SLP regressed upon the Pacific decadal oscillation index (Mantua et al. 1997).(right) COADS SST (color in °C) and NCEP surface wind (m s-----1) composites in Jan–Mar based on across-equatorial SST gradient index (Okumura et al. 2001).
197FEBRUARY 2004AMERICAN METEOROLOGICAL SOCIETY |
the form of a temperature inversion (e.g., Norris1998). In the present review, we focus on the satelliteobservations of SST, surface wind velocity, andboundary layer clouds—variables most relevant toair–sea interaction. We begin with a description of thewind responses to SST variations, and then continuewith a survey of cloud responses to these variations.Finally, we discuss how SST-induced anomalies in theboundary layer might lead to a deep response thatreaches the upper troposphere.
Unless we mention otherwise, the following mi-crowave measurements were used in the studies weare reviewing: sea surface vector wind by theSeaWinds scatterometer on the QuickSCAT satellite(Liu 2002); and SST, cloud water, and scalar windspeed by the Tropical Rain Measuring Mission(TRMM) satellite’s microwave imager (TMI; Wentzet al. 2000). QuickSCAT data are available since July1999 on a 0.25° grid, and TMI data since December
1997 at 0.25° resolution. QuickSCAT provides a dailycoverage over 90% of the World Ocean while the TMItakes about 2 days to cover the global Tropics.
WIND RESPONSE. This section examines the SSTeffect on winds. We organize our discussion accord-ing to the sign and size of correlation between SSTand wind speed [denoted as r(T, W) hereafter], which,as will become clear, is a useful indicator of whetherSST variations are merely forced by the atmosphereor whether the SST changes exert an influence on sur-face wind.
Positive correlation between temperature and windspeed. In the eastern equatorial Pacific and Atlantic,there is a SST front centered along 1°–2°N and sepa-rating the equatorial cold tongue from the warmerwater to the north. Tropical instability waves (TIWs),due to hydrodynamic instabilities of equatorial ocean
With its huge heat content andslow dynamic adjustment, theocean is generally considered asimportant for climate variabilitywith time scales longer than aseason. On interannual tointerdecadal time scales, SSTvariability is generally caused bythe following mechanisms: surfaceeffects due to surface heat flux andEkman advection that involvesonly the ocean mixed layer, andsubsurface effects by horizontaland vertical advection due tothermocline variability. Forexample, the subsurface effectsdominate the equatorial upwellingzone as seen in ENSO whereocean wave adjustment sets thetime scale (Neelin et al. 1998).
Surface effects become domi-nant over open off-equatorialoceans. In particular, surfacelatent and sensible heat flux isgenerally of first-order importanceover regions with a strong nega-tive correlation between SST andwind speed as in Fig. 1. In theTropics, because atmosphericinternal variability is generallytransient, two-way interactionbetween the ocean and atmo-sphere is necessary to generate
modes of SST variability withpreferred spatial patterns. Thecoupled SST–wind pattern inFig. 1b, for example, can bemaintained by a so-called wind–evaporation–SST feedback withthe help of weather noise (Changet al. 1997; Xie and Tanimoto1998). Surface heat flux is thecentral agent for such a thermo-dynamical feedback, in contrast tothe Bjerknes feedback that givesrise to ENSO with ocean dynamicsas a key element. (The SST–windrelation in the equatorial Pacificnear the date line in Fig. 1a isbroadly consistent with theBjerknes feedback.)
Matters become very differentin the extratropics, where atmo-spheric internal variability is oftenorganized in space formingstationary modes such as thePacific–North American andNorth Atlantic Oscillation pat-terns. The temporal characteris-tics of such atmospheric chaoticvariability are not well understoodbut often modeled as white noise.By simply integrating chaoticweather noise with its large heatinertia, the ocean can generateslow SST variability. Therefore,
OFF-EQUATORIAL SST VARIABILITYthe negative correlation in Fig. 1in the extratropics can be inter-preted as ocean passively respond-ing to wind-induced latent andsensible heat flux, with SSTdecreasing as the prevailingwesterlies intensify (Frankignoul1985; Barsugli and Battisti 1998).
Thus, understanding two-wayocean–atmosphere interaction inthe extratropics requires answersto the following questions. First,do the ocean dynamics matter andwhere? Recent ocean GCM studiessuggest that much of SST variabil-ity over the Kuroshio–OyashioExtenstion (KOE) east of Japan isdue to ocean dynamic effects. Thedeep winter mixed layer thereallows the ocean memory of pastwind changes to have a markedeffect on SST (Xie et al. 2000) bythe Rossby wave (Schneider andMiller 2001) or gyre–boundaryadjustment mechanism (Seageret al. 2001). To complete thefeedback loop, the second set ofquestions asks whether theatmosphere responds to extratro-pical SST variations and whatfeedback it provides. They are thefocus of the main body of thispaper.
198 FEBRUARY 2004|
currents, cause large meanders in this equatorialfront, which propagate westward at typical wave-lengths of 1000 km and typical periods of 30 days(Fig. 2a). The coupling between wind and SST is veryclear in longitude–time sections (not shown; Xie etal. 1998) and discernible even on snapshot images(Fig. 2b; Chelton et al. 2001). Figure 3a shows TIW-induced SST anomalies and associated anomalouswind vectors, based on a regression analysis ofHashizume et al. (2001). Here wind speed and SSTare positively correlated, with the prevailing south-easterly trade winds being accelerated over warm SSTanomalies and decelerated over cold anomalies. Thisis opposite to the negative SST–wind correlation inFig. 1 and represents a clear SST influence on theatmosphere. In fact, the wind speed anomalies act todampen SST anomalies by means of surface latent-and sensible heat fluxes (Zhang and McPhaden 1995;Thum et al. 2002).
Such a positive SST–wind speed correlation is gen-erally thought to be due to a vertical shear adjustmentof the atmosphere near the sea surface (Wallace et al.1989; Hayes et al. 1989). The increase in SST reducesthe static stability of the near-surface atmosphere,causing intensified turbulent mixing that brings downfast-moving air from aloft and accelerates the surfacewind. Indeed, a transect of shipboard radiosondemeasurements along 2°N shows a tendency for stron-ger vertical shear in the lowest few hundred metersover the colder regions of the TIWs (Fig. 4).
This vertical momentum–mixing mechanism isquite ubiquitous and appears to dominate the windadjustment to SST changes on the major ocean fronts,where we expect ocean dynamics to be the most im-portant mechanism in causing SST variability (see
sidebars 2 and 3). The characteristic positive SST–wind correlation has been detected from satellites in dif-ferent regions of the World Ocean: the Atlantic equa-torial front (Hashizume et al. 2001), the western northIndian Ocean (Vecchi et al. 2004), the Kuroshio andits extension (Nonaka and Xie 2003), the SouthernOcean (O’Neill et al. 2003), and the Gulf Stream rings(Park and Cornillon 2002). Marked wind speed re-duction due to the same mechanism is also observedin cold wakes to typhoons that are about 100 km wideand only last for a few days (Lin et al. 2003).
Zero correlation between wind speed and temperature.Now we turn our attention to a region with a weakSST gradient. As the trade winds impinge on theHawaiian Islands, they are blocked by the high moun-tains, an effect that creates a complex sequence of in-teractions casting a long shadow in the atmosphereand on the ocean (Fig. 3b). In this figure, the effectsof the islands are isolated using a meridional filter thatremoves the large-scale trade wind system. The filter-ing reveals an anomalous band of warm SST west ofHawaii that results from the advection of warmer wa-ter from the west by an eastward current, the currentitself being forced by an island-induced wind curl (Xieet al. 2001). Anomalous surface winds converge ontothis warm band, indicating the in situ formation of asea level pressure (SLP) low as a result of warming ofthe atmospheric boundary layer. The anomalouszonal winds are due to the bending by the Coriolisforce that acts on the anomalous meridional winds orcan equivalently be interpreted as geostrophic windassociated with the meridional SLP gradient.
Under the SLP-driving mechanism, the anoma-lous SST and winds are 90° out of phase, with the
anomalous winds strength-ening the prevailing north-easterly trades to the north,and weakening them to thesouth of the warm band. Asa result, the spatial correla-tion between SST and windspeeds is nearly zero. Ingeneral, this lack of corre-lation may also result fromthe coexistence of one-way,basin-scale forcing of theocean by the atmosphere[r(T, W) < 0] and the oceanfront–induced wind cova-riation [r(T, W) >)], or itmay simply reflect no rela-tionship between SST and
FIG. 2. TMI SST on 2–4 Sep 1999 (a) color and (b) contours. QuickSCAT windcurl with the meandering equatorial front. From Chelton et al. (2001).
199FEBRUARY 2004AMERICAN METEOROLOGICAL SOCIETY |
wind variability. A correlation analysis between SSTand wind divergence or curl will help reveal theSST-induced SLP mechanism for wind variability.(Wind convergence and curl calculations are also aspatial filter that emphasizes small-scale features suchas the banded SST structure west of Hawaii.)
Thus, the SST–wind speed correlation, if statisti-cally significant, is indeed useful in determining thecausal relationship be-tween the two. A signifi-cant positive correlationindicates that the ocean in-fluences surface winds; asignificant negative corre-lation in the extratropicsindicates a passive SST re-sponse to changes in theatmosphere (which couldbe remotely forced by SSTvariations elsewhere, how-ever); and a statistically in-significant correlation re-quires further analysis thatconsiders lags in space ortime to determine whetherthis lack of correlation isdue to the effects of the SLPdriving mechanism or issimply a reflection of ran-dom independent varia-tions in the ocean andatmosphere.
Vertical structure of theatmosphere. Cool oceanswithout deep convectionare often capped by a tem-perature inversion. Thevertical displacement of theinversion results in largetemperature anomalies andis important for boundarylayer pressure adjustmentto changing SST. In the ra-diosonde transect in Fig. 4,the capping inversion var-ies in height by as much as500 m, from 1.5 km overthe warmer regions of theTIWs to 1 km over thecolder regions. A rise in theinversion height causescold temperature anoma-
lies in the 1- to 1.5-km layer that ride directly abovethe warming effect of SST anomalies. This verticaltemperature dipole reduces hydrostatic pressure atthe sea surface. Hashizume et al. (2002) suggest thatthis reduction in SLP due to the adjustment of the in-version height is the reason for the puzzling absenceof pressure-driven wind anomalies over TIWs, an in-ference first made by Hayes et al. (1989) and subse-
FIG. 4. Longitude–height section of virtual potential temperature (contours forqqqqqv > 300 K and color shade for qqqqqv < 300K) and zonal wind velocity (vectors inm s-----1) based on a radiosonde transect along 2°N. The survey took place dur-ing 24–25 Sep 1999.
FIG. 3. Surface wind response to SST changes. (top) TMI SST (color in °C)and QuickSCAT wind (vectors in m s-----1) regressed upon TIW-induced SSTanomalies at, 1.5°N, 105°W (Hashizume et al. 2001). (bottom) SST and windvariations induced by the presence of the Hawaiian Islands (Xie et al. 2001).Basin-scale background fields as represented by 8° meridional running meanshave been removed. Note the difference in SST–wind correlation from thatin Fig. 1.
200 FEBRUARY 2004|
Wind–SST coupling analogous tothat on the equatorial front hasrecently been detected fromsatellite observations around theworld. Some examples follow.
During June–August, thesouthwesterly monsoonal windsoff the coast of Somaliaand Arabia, called theFindlater Jet, inducestrong coastal up-welling that is oftenassociated with coldfilaments extendingoffshore. Two such coldfilaments, south andnorth of Socotra Island,respectively, arepermanent features ofsummer SST climatol-ogy and act to disruptthe Findlater Jet, whichhas until now beendepicted as a broad andcontinuous flow (leftpanel of Fig. SB1). Thealong-jet variations inthe climatological-mean wind speed andSST are positivelycorrelated in space andare consistent with theSST modulation ofvertical mixing andwind shear (Vecchi et al. 2004).
In the extratropical NorthPacific, weather noise is muchgreater than in the Tropics. Eventhere, the wind speed maximum isobserved to follow the warmKuroshio Current that takes anoffshore path south of Tokyo inthe right panels of Fig. SB1(Nonaka and Xie 2003). The coldpool between Japan and theoffshore Kuroshio and the coldring farther to the east are bothassociated with a reduction inlocal wind speed.
Chelton et al. (2001) develop aquantitative procedure to test thevertical mixing mechanism forTIW-induced wind variability andshow that the perturbation windcurl (divergence) is linearlyproportional to the crosswind(downwind) gradient of perturba-
tion SST. O’Neill et al. (2003)show that this proportionalityholds in the circumpolar SouthernOcean (Fig. SB2), a rather surpris-ing result considering that it is oneof the stormiest oceans in theworld. They average data for
3 months to suppress synopticwind variability.
Together, these satellite studiesstrongly suggest that the verticalmomentum mixing and theassociated shear adjustment arethe dominant mechanism for
ATMOSPHERIC ADJUSTMENT NEAR OCEAN FRONTS
FIG. SB1. (left) TMI SST (color in °C) and wind speed (contours in m s-----1) averaged forJul 2000 over the western North Indian Ocean. Note the cold wedges due to coastalupwelling and their decelerating effect on wind. (top right) TMI SST and (bottom right)wind speed over the Kuroshio Current south of Japan for Apr–Jun 2001 (Nonaka andXie 2003). The Kuroshio appears as a stream of warm water in the SST imagery.
FIG. SB2. Surface wind adjustment to SST variations inthe Southern Ocean. Relationship between anomaliesof (a) downwind SST gradient and wind divergence and(b) crosswind SST gradient and wind curl. Spatial varia-tions with wavelengths longer than 10° lat and 30° lonare removed. From O’Neill et al. (2003).
201FEBRUARY 2004AMERICAN METEOROLOGICAL SOCIETY |
quently supported by satellite measurements of sur-face wind2 (Xie et al. 1998; Liu et al. 2000; Cheltonet al. 2001).
Thus, the vertical structure of the atmosphere isvery important for understanding the mechanisms bywhich it adjusts to SST variations. Because satellitesdo not measure vertical structure well, numericalmodeling, carefully validated against observations, isa valuable tool to help us understand what is happen-ing. For example, Fig. 5 shows the vertical structureof the atmosphere in a high-resolution regional modelas it adjusts to a pair of zonal bands of positive andnegative SST anomalies west of Hawaii (Hafner andXie 2003). The air temperature in the boundary layer
follows the SST anomalies imposed in the model, butat the inversion level there are considerable tempera-ture anomalies due to the vertical displacement of theinversion. Generally, low sea level pressure forms overthe warm SST band, onto which surface winds con-verge as in satellite observations. Two anomalousmeridional circulations form with the updraft over thesurface wind convergence, and they are trappedwithin the PBL that is capped by an inversion 2 kmhigh. Although this circulation is shallow—only one-tenth of the tropopause height—its structure is remi-niscent of the Hadley circulation and is dominated bythe first baroclinic mode.
Vertically integrated, barotropic boundary layermodels are often used to model surface wind velocity(e.g. Lindzen and Nigam 1987; Battisti et al. 1999),and they are successful when the boundary layer flowis connected to deeper tropospheric circulation, suchas that feeding the rising branch of the Hadley cell atthe ITCZ. In such cases, winds tend to flow roughlyin the same direction over the depth of the boundarylayer. The baroclinic response of the inversion-toppedboundary layer to SST anomalies over cool oceans,
ocean front–PBL interactionaround the world. The disparity inhorizontal scale between atmo-spheric adjustment and oceanicfronts seems to be the mainreason for the prevalence ofpositive SST–wind speed correla-tion there. Consider an idealizedcase of a step function oceanicfront (Fig. SB3a). Dynamicadjustment of the atmosphere tothis infinitesimally narrow front islikely to yield a much smootherprofile of surface air temperaturewith finite gradients. See Warneret al. (1990) and Rogers (1989) forexamples of such cross-frontaldynamic adjustment. If we assumeneutral surface stability andvanishing cross-frontal wind in thebackground for simplicity, theresultant sea–air temperaturedifference is such that the warmerside of the SST front destabilizeswhile the colder side becomesstable. Increased mixing on theunstable warmer side reduces thenear-surface wind shear, leadingto an acceleration of surface
winds. Such a wind accelerationhas been observed in in situmeasurements on the warmerflank of the Pacific equatorial front(Wallace et al. 1989) and frontsover the Gulf Stream (Sweet et al.1981) and Kuroshio (Y. Tanimotoand H. Tokinaga 2002, personalcommunication).
Detailed stability distributionacross the front also depends onthe background cross-frontalwind. Wallace et al. (1989)propose that the thermal advec-tion by cross-frontal mean flow isimportant. By this lateral advec-tion mechanism, the verticalmixingadjustmentoccurs only onthe downwindside of thefront(Fig. SB3b).Regardless ofthe details ofadjustmentinvolved, largechanges in
stability seem common acrossocean fronts.
Across ocean fronts, advectionby the background winds isgenerally an important term inthe PBL thermodynamicalequations, displacing the tempera-ture and humidity responsedownstream of SST anomalies.Small et al. (2003) show that suchdisplacements lead to perturba-tion SLP fields with resultant windanomalies that are positivelycorrelated with underlying SST.This is an alternative explanationfor the ubiquitous positive r(T, W)correlation near fronts.
FIG. SB3. Adjustment of surface air temperature (Tain dashed line) to a sharp SST front (solid) due to (a)gravity wave adjustment and (b) advection by the back-ground wind (arrow). SST – Ta is positive and hencethe near-surface atmosphere is more unstable on thewarmer than the colder flank of the front.
2 Based on a high-resolution simulation, Small et al. (2003) sug-gest that SLP may still be an important mechanism for windadjustment to TIW-induced SSTAs because with a small Co-riolis effect near the equator, modest SLP variations can causea large wind response. Significant SLP anomalies are indeed de-tected from buoy measurements, but their cause and dynamicconsequences are still open to different interpretations (Croninet al. 2003).
202 FEBRUARY 2004|
however, calls for including baroclinic modes for sur-face wind simulation.
CLOUD RESPONSE. Low-level boundary layerclouds play an important role in the energy budgetof the climate system. They act to cool the atmosphereby emitting longwave radiation into space as well asthe ocean by reflecting solar radiation. The interac-
tion of these low clouds with SST is complicated andpoorly understood because it involves not only stabil-ity and moisture convergence but also cloud microphys-ics. This section presents examples of the differentways that low clouds can respond to SST anomalies.
Deser et al. (1993) first noted a correlation betweenclouds and TIW-induced meanders of the SST front.A detailed regression analysis of TMI data indicates
It is well known that land shapeaffects weather and climate. Forexample, the Pacific Coast ofSouth America is desert, whileAmazonia on the other side of theAndes is rich in rainfall and hoststhe largest rain forest of theworld. On the global scale, theTibetan Plateau is a controllingelement in Northern Hemisphereclimate. Even the tall mountains ofthe tiny Hawaiian Islands exertfar-reaching effects on the PacificOcean and atmosphere.
It may sound improbable thatsubmerged bottom topographycan change winds and clouds, butsatellite observations reveal abathymetric effect over theYellow and East China Seas (Xieet al. 2002). These seas, locatedbetween China, Korea, and Japan,together form one of the largestshelf seas of the world. Most of
these seas are shallower than100 m and their bottom topogra-phy is uneven with deeper andshallower tonguelike regions. Inwinter, riding on the northerlymonsoon, the frigid and drycontinental air blows over the seasand cools the surface. Heat istransferred from the oceanbottom upward through convec-tion that can reach more than100 m. The deeper the area, themore heat it contains, and theslower it cools. In this manner, thecooling rate of the water columnis determined by its height. Thismechanism, in combination withthe advection of warm Kuroshiowater by bathmetry-followingshelf currents, generates warmand cold tongues over deepchannels and the shallow bank,respectively.
The bathymetric effect of the
seas does not stop with causingSST variations. QuickSCAT andTRMM measurements revealremarkable spatial covariations inwind and clouds with SST.Convergent wind and increasedcloudiness are found over thebathymetric-induced warmtongues. In particular, one suchband of covariation between theocean and atmosphere meandersthrough the Yellow Sea betweenChina and Korea, following a deepchannel for the amazing distanceof 1000 km (Fig. SB4). Windconvergence is also found on thewarmer flank of the Kuroshio,supporting increased CLW/precipitation there. The mecha-nism for this ocean effect on thewinds appears to be due to thestability and wind shear adjust-ment similar to that over oceanicfronts (sidebar 2).
FIG. SB4. Jan–Mar SST climatology (contours in °C) over the Yellow and East China Seas, along with (a)bottom depth (m), (b) velocity (vectors in m s-----1) and divergence (color in 10-----6 s-----1) of QuickSCAT wind,and (c) TMI cloud liquid water (10-----2 mm). The QuickSCAT and TMI climatologies are Jan–Mar aver-ages for 2000–02. From Xie et al. (2002).
OCEAN DEPTH AFFECTS CLIMATE
203FEBRUARY 2004AMERICAN METEOROLOGICAL SOCIETY |
that anomalies of cloud liquid water (CLW) contentare roughly but not exactly in phase with SST anoma-lies, with the former consistently displaced northeastof the latter (Fig. 6a). The in-phase relationship arisesbecause intensified verticalmixing deepens the bound-ary layer over positive SSTanomalies (Fig. 4) andtransports humid surfaceair above the condensationlevel. The phase differencebetween CLW and SSTanomalies seems to be dueto surface moisture conver-gence; anomalous surfacewinds converge north ofthe warm SST center anddiverge south of it (Fig. 3).The moisture convergenceand SST effect on verticalmixing cooperate to thenorth of but oppose eachother south of the centers ofanomalous SST.
Thus, the surface wind’sadjustment plays a role inhow clouds respond to SSTchanges, and the phasing ofthe cloud response dependson wind response mecha-nisms. The vertical mixingmechanism produces windconvergence that is 90° outof phase with SST, resulting
in cloud anomalies that aredisplaced north of SSTanomalies in TIWs (Fig. 6).The SLP mechanism, onthe other hand, causes windconvergence and cloudanomalies that are in phasewith the SST anomalies. Inthe Hawaiian wake, for ex-ample, anomalous windsconverge onto the warmband, leading to the forma-tion of a cloud band collo-cated with the warm SSTtongue west of Hawaii(Fig. 6b). Since wind veloc-ity and CLW are measuredindependently, the formerby QuickSCAT and the lat-
ter by TRMM satellites, the physical consistency oftheir covariations gives us confidence in the results.The model simulation shows that CLW increases inthe anomalous updraft near the inversion height and
FIG. 5. Latitude–pressure section of (left) temperature (10-----1 K), (right) cloudwater content (10-----5 kg kg-----1), and meridional circulation anomalies simulatedin a regional atmospheric model west of HI, zonally averaged in 167°–162°W.The green contours encompass the inversion layer. SST anomalies are im-posed zonally uniformly west of HI as in satellite observations (Fig. 3b) andtheir profile is plotted in the lower panels.
FIG. 6. Boundary layer cloud response. (top) TMI SST (contours in °C) andCLW (color in 10-----2 mm) regressed upon TIW-induced SST anomalies at 1.5°N,105°W (Hashizume et al. 2001). Only the 0.4°C and –0.4°C contours are plot-ted for SST. (bottom) TMI SST (contours in °C) and CLW (color in 10-----2 mm)in the central subtropical North Pacific are averaged for Aug–Nov 1999(Xie et al. 2001).
204 FEBRUARY 2004|
decreases in the two anomalousdowndrafts (Fig. 5b).
In both TIWs and the Hawaiianwake, CLW anomalies are roughlypositively correlated with and act todampen the SST anomalies thatcaused the cloud changes in the firstplace.3 The associated greater radia-tive cooling at the top of the bound-ary layer increases vertical turbulentmixing in the boundary layer, acloud effect that needs further inves-tigation. This cloud-top cooling alsoaffects SLP and hence surface winds(Nigam 1997). (TMI measures col-umn-integrated CLW and we haveassumed that over cool oceans witha strong inversion, SST-inducedCLW variability is mostly associatedwith low clouds.)
Over many regions of the sub-tropical and midlatitude oceans,however, increased cloudiness isfound not over warm but over coolSST anomalies (Norris and Leovy1994; Norris et al. 1998). Figure 7 shows an exampleof this negative SST–cloudiness correlation in thetropical Atlantic, based on composites of ship obser-vations with reference to a cross-equatorial SST gra-dient index. Associated with an anomalous basin-scaleSST dipole that changes sign across the equator arefour anomalous zonal cloud bands with alternatingsigns. The equatorial pair of cloud bands results froma warm ocean–atmosphere interaction and reflects ashift of the ITCZ rainband to the anomalously warmside of the equator. This cloud band pair is associatedwith strong anomalous surface wind convergence thatsupplies the moisture for convection. Poleward ofthese high-cloud changes, cloud anomalies arebroadly distributed and negatively correlated with theSST anomalies underneath. Unlike the deep clouds inthe ITCZ, these low-level cloud anomalies are unre-lated to significant changes in surface wind conver-gence (Tanimoto and Xie 2002).
Klein and Hartmann (1993) found a positive cor-relation between low-level stratus cloud cover and the
static stability of the lower troposphere. Noting that adecrease in SST increases the static stability, Philanderet al. (1996) showed that this positive feedback betweenSST and stratus clouds helps to maintain the equato-rial asymmetry of the Pacific climate. Norris (1998)suggested that this feedback is accomplished by changesin cloud types—a decrease in SST favors persistent andextensive stratocumulus clouds to more spotty tradewind cumulus clouds, thereby increasing the cloudcover. Cold advection in the boundary layer, a domi-nant mechanism for synoptic variability in cloudiness(Klein 1997) and large-scale subsidence above the in-version may also play a role.
Why does a rise in SST lead to an increase in low-cloud amount in one region but a decrease in an-other? We propose that the horizontal scale of theanomalous SST might be the key. Surface moistureconvergence is inversely proportional to the spatialscale of the anomalies. Convergence is strong forsmall-scale SST anomalies, where it dominates othercloud-changing mechanisms, giving rise to a positivecorrelation with CLW. For large, basin-scale SSTanomalies, moisture convergence is of secondary im-portance, and mechanisms such as changes in staticstability take over, resulting in a negative correlationwith cloudiness. This scale-dependence of moistureconvergence explains why significant SST–cloudi-ness correlations are almost always negative around
3 This negative feedback results from the reflection of incom-ing solar radiation by clouds. The presence of low clouds alsoincreases downward infrared radiation, an effect to warm thesea surface, which Wang and Enfield (2003) suggest is impor-tant in the eastern Pacific warm pool.
FIG. 7. COADS SST (contours in °C) and cloudiness (color in %) com-posites in Jan–Mar based on a cross-equatorial SST gradient index.Note that their correlation in the subtropics is opposite to that inFig. 6.
205FEBRUARY 2004AMERICAN METEOROLOGICAL SOCIETY |
the World Ocean in Norris and Leovy’s (1994) coarse-resolution (10°) analysis of ship observations. PositiveSST–CLW correlations, however, exist in high-reso-lution satellite observations over cool oceans onsmaller spatial scales, as demonstrated in this review.
DISCUSSION. We have surveyed recent studiesthat use new satellite observations to address the ques-tions of whether and how the atmosphere respondsto SST changes over cool oceans. Our focus is on sur-face wind and low-level clouds, both of which are theimportant forcing mechanisms for SST. SST-inducedvertical momentum mixing emerges as a ubiquitousmechanism by which surface wind adjusts to SST gra-dients at oceanic fronts in regions ranging from theequatorial cold tongue to the midlatitude North Pa-cific and the Southern Ocean. This vertical mixingmechanism dominates near ocean fronts because thewidth of these fronts is smaller than the atmosphericadjustment scale and the static stability of the near-surface atmosphere varies rapidly. It is unclear, how-ever, how high this shear adjustment reaches. Thewind speed in the lowest 100 m of the atmospherefollows a log profile that varies with static stabilitynear the surface. Is this log-profile adjustment themain mechanism for the surface wind variations re-ported here or does the wind adjustment occur overthe whole depth of the PBL? The radiosonde obser-vations of TIWs in Fig. 4 suggest the latter, but moremeasurements and modeling studies are needed toconfirm this idea.
The correlation betweenSST and wind speed thatresults from the verticalmomentum mixing is posi-tive, in contrast to the nega-tive correlations thoughtto result from one-way at-mospheric forcing. Thus,the SST–wind correlationturns out to be a useful in-dicator of the causal direc-tion in the interaction be-tween a cool ocean and theatmosphere. SST-inducedSLP changes are anotherimportant mechanism forthe generation of surfacewind anomalies resulting ina zero correlation betweenwind speed and local SST.In such a situation, addi-tional analysis of lagged
characteristics in space and time is necessary to deter-mine whether a causal relation between SST and windexists.
Regarding low-level cloud response, both positiveand negative correlations with SST are seen in obser-vations. We suggest that the horizontal scale of theSST anomaly pattern is key to determining the signof the cloud response. At small scales (a few hundredkilometers) the moisture convergence is important forcloud formation and a positive SST–cloudiness cor-relation tends to appear. On the basin scale, on theother hand, convergence is negligible, and the SSTeffect on the capping of the boundary layer becomesdominant, leading to a negative correlation betweenSST and cloudiness. This scale-dependence of theSST–cloudiness correlation has recently been demon-strated in a high-resolution regional atmosphericmodel over the southeastern tropical Pacific (H. Xu2002, personal communication).
A natural question is how these wind and cloudresponses feed back on the ocean. With increasedwind speed and cloudiness over positive SST anoma-lies on ocean fronts, the thermal feedback is likely tobe negative. The dynamic feedback is hard to assessat this time because of nonlocal wave adjustment andnonlinear processes in the ocean. The wind stressanomalies resulting from ocean front–atmosphereinteractions amount to 20%–30% of the climatologi-cal mean, but because of the small spatial scale offronts (~100 km), the wind stress curl anomalies canbe as large as the mean. Figure 8 shows the 4-yr-mean
FIG. 8. Gulf Stream’s effect on surface wind. Ekman pumping velocity (10-----6 m s-----1)derived from QuickSCAT wind stress, averaged for 4-yr period of Aug 1999–Jul 2003.
206 FEBRUARY 2004|
Ekman pumping velocity derived from QuikSCATobservations, on which the effect of the warm GulfStream is obvious without any filter. A narrow andintense upwelling zone is found off the east coast ofthe United States, roughly collocated with the warmcurrent.4 Similar wind curls are found in the Kuroshiofront in the subtropical East China Sea (sidebar 3) andalong the Pacific equatorial front (Chelton et al. 2001).Such frontal-scale wind curl is neither resolved inother wind stress products nor in global GCMs, butmay contribute to the recirculation of the Gulf Streamby Sverdrup dynamics (Behringer et al. 1979). Chenet al. (2003) suggest that the interaction between theSST front and the wind curls is an important elementof ocean-front dynamics that strongly affects verticalnutrient transport and bioproductivity.
All the examples discussed in this paper concernthe local and shallow atmospheric response that isconfined to the vicinity of SST changes in the hori-zontal and to the boundary layer in the vertical.Important questions left unaddressed are whetherthis shallow boundary layer response can lead to adeep response over the whole depth of the tropo-sphere and by what mechanism this is accomplished.Kushnir et al. (2002) review the progress and diffi-culty in addressing this problem of deep response.The difficulty that current globe GCMs—at typicalresolutions of 2.5° in the horizontal and 500 m in thevertical—have in simulating the shallow boundarylayer response may be partially responsible for theirdivergent results, given the strong scale dependenceof this response and the importance of the inversiondiscussed in this paper. Most previous studies havefocused on the response of stationary waves in themid- to upper troposphere and viewed transient ed-dies as a feedback that amplifies and modifies theexisting wave patterns. Several recent studies suggestan alternative scenario with boundary layer anoma-lies playing a central role. SST-induced temperature
anomalies change the distribution of baroclinicity inthe planetary boundary layer, which modulates thegrowth of transient eddies (Xie et al. 2002) and givesrise to a robust response in the storm track(Nakamura and Shimpo 2004). Transient eddies thenforce stationary waves by their effect on precipitationin the storm track (Inatsu et al. 2003), rather than bytheir sensible heat and vorticity fluxes on which manyprevious studies have focused.
ACKNOWLEDGMENTS. I would like to thankD. Chelton for his helpful discussion and comments, manycolleagues for their contributions to the collaborativeresearch that this paper draws on, especially,M. Cronin, J. Hafner, H. Hashizume, M. Inatsu, W. T. Liu,H. Mukougawa, M. Nonaka, Y. Okumura, J. Small,Y. Tanimoto, and G. Vecchi. My discussion with B. Mapeswas very helpful for interpreting the cloud response.Thanks are also extended to G. Speidel and anonymousreviewers for their helpful comments. This work is sup-ported by NASA, NOAA, NSF, NSFC, and the FrontierResearch System for Global Change through its supportfor IPRC.
REFERENCESAlexander, M. A., I. Blade, M. Newman, J. R. Lanzante,
N.-C. Lau, and J. D. Scott, 2002: The atmosphericbridge: The influence of ENSO teleconnections onair–sea interaction over the global oceans. J. Climate,15, 2205–2231.
Barsugli, J. J., and D. S. Battisti, 1998: The basic effectsof atmosphere–ocean thermal coupling on mid-latitude variability. J. Atmos. Sci., 55, 477–493.
Battisti, D. S., E. S. Sarachik, and A. C. Hirst, 1999: Aconsistent model for the large-scale steady surfaceatmospheric circulation in the tropics. J. Climate, 12,2956–2964.
Behringer, D., L. Regier, and H. Stommel, 1979: Ther-mal feedback on wind-stress as a contributing causeof the Gulf Stream. J. Mar. Res., 34, 699–709.
Chang, P., L. Ji, and H. Li, 1997: A decadal climatevariation in the tropical Atlantic Ocean from ther-modynamic air-sea interactions. Nature, 385, 516–518.
Chelton, D. B., and Coauthors, 2001: Observations ofcoupling between surface wind stress and sea surfacetemperature in the eastern tropical Pacific. J. Climate,14, 1479–1498.
——, M. G. Schlax, M. H. Freilich, and R. F. Milliff, 2004:Satellite radar measurements reveal short-scale fea-tures in the wind stress field over the world ocean.Science, in press.
4 QuickSCAT measures the stress on the ocean surface.Changes in either wind or ocean currents can result in varia-tions in QuickSCAT stress. The wind curl pattern in Fig. 8may result from the Gulf Stream’s effect on surface dragrather than its effect on surface winds (Chelton et al. 2004).Kelly et al. (2001) report a similar ocean current effect in ana-lyzing QuickSCAT measurements over the equatorial Pacific.Taking advantage of the abundant shipboard observationsover the North Atlantic east of the United States, H. Tokinagaand Y. Tanimoto (2003, personal communication) compilea new surface wind dataset and obtain an Ekman pumpingfield very similar to that in Fig. 8, including the narrow streakof intense upwelling along the Gulf Stream.
207FEBRUARY 2004AMERICAN METEOROLOGICAL SOCIETY |
Chen, D., W. T. Liu, W. Tang, and Z. Wang, 2003: Air–sea interaction at an oceanic front: Implications forfrontogenesis and primary production. Geophys. Res.Lett., 30, 1745, doi: 10.1029/2003GL017536.
Cronin, M. F., S.-P. Xie, and H. Hashizume, 2003: Baro-metric pressure variations associated with easternPacific tropical instability waves. J. Climate, 16,3050–3057.
Deser, C., J. J. Bates, and S. Wahl, 1993: The influenceof sea surface temperature on stratiform cloudinessalong the equatorial front in the Pacific Ocean. J.Climate, 6, 1172–1180.
Frankignoul, C., 1985: Sea surface temperature anoma-lies, planetary waves, and air-sea feedback in themiddle latitudes. Rev. Geophys., 23, 357–390.
Graham, N. E., and T. P. Barnett, 1987: Sea surface tem-perature, surface wind divergence, and convectionover tropical oceans. Science, 238, 657–659.
Hafner, J., and S.-P. Xie, 2003: Far-field simulation ofthe Hawaiian wake: Sea surface temperature and oro-graphic effects. J. Atmos. Sci., 60, 3021–3032.
Hashizume, H., S.-P. Xie, W. T. Liu, and K. Takeuchi,2001: Local and remote atmospheric response totropical instability waves: A global view from thespace. J. Geophys. Res., 106, 10 173–10 185.
——, ——, M. Fujiwara, M. Shiotani, T. Watanabe, Y.Tanimoto, W. T. Liu, and K. Takeuchi, 2002: Directobservations of atmospheric boundary layer responseto slow SST variations on the Pacific equatorial front.J. Climate, 15, 3379–3393.
Hayes, S. P., M. J. McPhaden, and J. M. Wallace, 1989:The influence of sea surface temperature on surfacewind in the eastern equatorial Pacific. J. Climate, 2,1500–1506.
Inatsu, M., H. Mukougawa, and S.-P. Xie, 2003: Atmo-spheric response to zonal variations in midlatitudeSST: Transient and stationary eddies and their feed-back. J. Climate, 16, 3314–3329.
Kelly, K. A., S. Dickinson, M. J. McPhaden, and G. C.Johnson, 2001: Ocean currents evident in satellitewind data. Geophys. Res. Lett., 28, 2469–2472.
Klein, S. A., 1997: Synoptic variability of low-cloud prop-erties and meteorological parameters in the subtropi-cal trade wind boundary layer. J. Climate, 10, 2018–2039.
——, and D. L. Hartmann, 1993: The seasonal cycle oflow stratiform clouds. J. Climate, 6, 1587–1606.
Kushnir, Y., W. A. Robinson, I. Bladé, N. M. J. Hall, S.Peng, and R. Sutton, 2002: Atmospheric GCM re-sponse to extratropical SST anomalies: Synthesis andevaluation. J. Climate, 15, 2233–2256.
Lin, I.-I., W. T. Liu, C.-C. Wu, J. C. H. Chiang, and C.-H. Sui, 2003: Satellite observations of modulation of
Lindzen, R. S., and S. Nigam, 1987: On the role of seasurface temperature gradients in forcing low levelwinds and convergence in the tropics. J. Atmos. Sci.,44, 2418–2436.
Liu, W. T., 2002: Progress in scatterometer applications.J. Oceanogr., 58, 121–136.
——, X. Xie, P. S. Polito, S.-P, Xie, and H. Hashizume,2000: Atmospheric manifestation of tropical insta-bility waves observed by QuickSCAT and TropicalRain Measuring Mission. Geophys. Res. Lett., 27,2545–2548.
Mantua, N. J., S. R. Hare, Y. Zhang, J. M. Wallace, andR. C. Francis, 1997: A Pacific interdecadal climateoscillation with impacts on salmon production. Bull.Amer. Meteor. Soc., 78, 1069–1079.
Nakamura, H., and A. Shimpo, 2004: Seasonal variationsin the Southern Hemisphere storm tracks and jetstreams as revealed by reanalysis dataset. J. Climate,in press.
Neelin, J. D., D. S. Battisti, A. C. Hirst, F. F. Jin,Y. Wakata, T. Yamagata, and S. Zebiak, 1998: ENSOtheory. J. Geophys. Res., 103, 14 261–14 290.
Nigam, S., 1997: The annual warm to cold phase transi-tion in the eastern equatorial Pacific: Diagnosis of therole of stratus cloud-top cooling. J. Climate, 10, 2447–2467.
Nonaka, M., and S.-P. Xie, 2003: Covariations of seasurface temperature and wind over the Kuroshio andits extension: Evidence for ocean-to-atmosphericfeedback. J. Climate, 16, 1404–1413.
Norris, J. R., 1998: Low cloud type over the ocean fromsurface observations. Part I: Relationship to surfacemeteorology and the vertical distribution of tempera-ture and moisture. J. Climate, 11, 369–382.
——, and C. B. Leovy, 1994: Interannual variability instratiform cloudiness and sea surface temperature. J.Climate, 7, 1915–1925.
——, Y. Zhang, and J. M. Wallace, 1998: Role of lowclouds in summertime atmosphere–ocean interac-tions over the North Pacific. J. Climate, 11, 2482–2490.
Okumura, Y., S.-P. Xie, A. Numaguti, and Y. Tanimoto,2001: Tropical Atlantic air–sea interaction and its in-fluence on the NAO. Geophys. Res. Lett., 28, 1507–1510.
O’Neill, L. W., D. B. Chelton, and S. K. Esbensen, 2003:Observations of SST-induced perturbations of thewind stress field over the Southern Ocean on seasonaltimescales. J. Climate, 16, 2340–2354.
Park, K.-A., and P. C. Cornillon, 2002: Stability-inducedmodification of sea surface winds over Gulf Stream
Philander, S. G. H., D. Gu, D. Halpern, G. Lambert, N.-C. Lau, T. Li, and R. C. Pacanowski, 1996: The roleof low-level stratus clouds in keeping the ITCZmostly north of the equator. J. Climate, 9, 2958–2972.
Rogers, D. P., 1989: The marine boundary layer in thevicinity of an ocean front. J. Atmos. Sci., 46, 2044–2062.
Schneider, N., and A. J. Miller, 2001: Predicting west-ern North Pacific Ocean climate. J. Climate, 14,3997–4002.
Seager, R., K. Kushnir, N. H. Naik, M. A. Cane, and J.Miller, 2001: Wind-driven shifts in the latitude of theKuroshio–Oyashio extension and generation of SSTanomalies on decadal timescales. J. Climate, 14,4249–4265.
Small, R. J., S.-P. Xie, and Y. Wang, 2003: Numericalsimulation of atmospheric response to Pacific tropi-cal instability waves. J. Climate, 16, 3723–3741.
Sweet, W., R. Fett, J. Kerling, and P. La Violette, 1981:Air–sea interaction effects in the lower troposphereacross the north wall of the Gulf Stream. Mon. Wea.Rev., 109, 1042–1052.
Tanimoto, Y., and S.-P. Xie, 2002: Inter-hemisphericdecadal variations in SST, surface wind, heat flux andcloud cover over the Atlantic Ocean. J. Meteor. Soc.Japan, 80, 1199–1219.
Thum, N., S. K. Esbensen, D. B. Chelton, and M. J.McPhaden, 2002: Air–sea heat exchange along thenorthern sea surface temperature front in the east-ern tropical Pacific. J. Climate, 15, 3361–3378.
Vecchi, G. A., S.-P. Xie, and A. S. Fischer, 2004: Ocean–atmosphere covariability in the western Arabian Sea.J. Climate, in press.
Waliser, D. E., N. E. Graham, and C. Gautier, 1993:Comparison of the highly reflective cloud and out-going longwave radiation datasets for use in esti-mating tropical deep convection. J. Climate, 6, 331–353.
Wallace, J. M., T. P. Mitchell, and C. Deser, 1989: Theinfluence of sea surface temperature on surface windin the eastern equatorial Pacific: Seasonal andinterannual variability. J. Climate, 2,1492–1499.
Wang, C., and D. B. Enfield, 2003: A further study of thetropical Western Hemisphere warm pool. J. Climate,16, 1476–1493.
Warner, T. T., M. N. Lakhtakia, and J. Doyle, 1990:Marine atmospheric boundary layer circulationsforced by Gulf Stream sea surface temperature gra-dients. Mon. Wea. Rev., 118, 309–323.
Wentz, F. J., C. Gentemann, D. Smith, and D. Chelton,2000: Satellite measurements of sea surface tempera-ture through clouds. Science, 288, 847–850.
White, W. B., and J. L. Annis, 2003: Coupling of extrat-ropical mesoscale eddies in the ocean to westerlywinds in the atmospheric boundary layer. J. Climate,16, 1095–1107.
Xie, S.-P., and S. G. H. Philander, 1994: A coupled ocean-atmosphere model of relevance to the ITCZ in theeastern Pacific. Tellus, 46A, 340–350.
——, and Y. Tanimoto, 1998: A pan-Atlantic decadal cli-mate oscillation. Geophys. Res. Lett., 25, 2185–2188.
——, M. Ishiwatari, H. Hashizume, and K. Takeuchi,1998: Coupled ocean-atmospheric waves on theequatorial front. Geophys. Res. Lett., 25, 3863–3866.
——, T. Kunitani, A. Kubokawa, M. Nonaka, and S.Hosoda, 2000: Interdecadal thermocline variabilityin the North Pacific for 1958–97: A GCM simulation.J. Phys. Oceanogr., 30, 2798–2813.
——, W. T. Liu, Q. Liu and M. Nonaka, 2001: Far-reaching effects of the Hawaiian Islands on the Pa-cific Ocean-atmosphere. Science, 292, 2057–2060.
——, J. Hafner, Y. Tanimoto, W. T. Liu, H. Tokinaga,and H. Xu, 2002: Bathymetric effect on the winter seasurface temperature and climate of the Yellow andEast China Seas. Geophys. Res. Lett., 29, 2228,doi:10.1029/2002GL015884.
Zhang, G. J., and M. J. McPhaden, 1995: The relation-ship between sea surface temperature and latent heatflux in the equatorial Pacific. J. Climate, 8, 589–605.
