DTIC ADA513414: The Effects of SST-Induced Surface Wind Speed and Direction Gradients on Midlatitude Surface Vorticity and Divergence

5.9 MB·English

by  Defense Technical Information Center

15JANUARY2010 O’NEILL ET AL. 255
The Effects of SST-Induced Surface Wind Speed and Direction Gradients on
Midlatitude Surface Vorticity and Divergence
LARRYW.O’NEILL
MarineMeteorologyDivision,NavalResearchLaboratory,Monterey,California
DUDLEYB.CHELTON
CollegeofOceanicandAtmosphericSciences,andCooperativeInstituteforOceanographicSatelliteStudies,
OregonStateUniversity,Corvallis,Oregon
STEVENK.ESBENSEN
CollegeofOceanicandAtmosphericSciences,OregonStateUniversity,Corvallis,Oregon
(Manuscriptreceived2May2008,infinalform1June2009)
ABSTRACT
Theeffectsofsurfacewindspeedanddirectiongradientsonmidlatitudesurfacevorticityanddivergence
fields associated with mesoscale sea surface temperature (SST) variability having spatial scales of 100–
1000kmareinvestigatedusingvectorwindobservationsfromtheSeaWindsscatterometerontheQuick
Scatterometer(QuikSCAT)satelliteandSSTfromtheAdvancedMicrowaveScanningRadiometerforEarth
ObservingSystem(AMSR-E)Aquasatellite.Thewind–SSTcouplingisanalyzedovertheperiodJune2002–
August2008,correspondingtothefirst61yearsoftheAMSR-Emission.Previousstudieshaveshownthat
strongwindspeedgradientsdevelopinresponsetopersistentmesoscaleSSTfeaturesassociatedwiththe
KuroshioExtension,GulfStream,SouthAtlantic,andAgulhasReturnCurrentregions.MidlatitudeSST
frontsalsosignificantlymodifysurfacewinddirection;thesurfacewindspeedanddirectionresponsesto
typicalSSTdifferencesofabout28–48Care,onaverage,about1–2ms21and48–88,respectively,overallfour
regions.WindspeedperturbationsarepositivelycorrelatedandverynearlycollocatedspatiallywiththeSST
perturbations.Winddirectionperturbations,however,aredisplacedmeridionallyfromtheSSTperturba-
tions,withcyclonicflowpolewardofwarmSSTandanticyclonicflowpolewardofcoolSST.
Previous observational analyses have shown that small-scale perturbations in the surface vorticity and
divergencefieldsarerelatedlinearlytothecrosswindanddownwindcomponentsoftheSSTgradient,re-
spectively.Whenthevorticityanddivergencefieldsareanalyzedincurvilinearnaturalcoordinates,thewind
speed contributions to the SST-induced vorticity and divergence depend equally on the crosswind and
downwindSSTgradients,respectively.SST-inducedwinddirectiongradientsalsosignificantlymodifythe
vorticityanddivergencefields,weakeningthevorticityresponsetocrosswindSSTgradientswhileenhancing
thedivergenceresponsetodownwindSSTgradients.
1. Introduction circulationpatternschangesurfaceoceantemperatures
through modulation of surface heat fluxes and upper
On spatial scales spanning ocean basins, sea surface
ocean mixing (e.g., Cayan 1992). On smaller spatial
temperature(SST)perturbationsarefoundtobenega-
scalesbetween100and1000 km,however,surfacewind
tivelycorrelatedwithsurfacewindspeedperturbations
speed and SST perturbations have a strong positive
(e.g.,Mantuaetal.1997;Okumuraetal.2001;Xie2004).
correlation in regions of large SST gradients, which
Overthesebroadspatialscales,large-scaleatmospheric
occur near meandering ocean currents (see review by
Smalletal.2008).Contemporaneousnear-globalsatel-
litemeasurementsofsurfacevectorwindsandSSThave
Corresponding author address: Larry W. O’Neill, Naval Re-
shownthatthesesmall-scalefeaturesinthesurfacewind
search Laboratory, 7 Grace Hopper Ave., MS2, Monterey, CA
field are particularly prevalent near large midlatitude
93943–5502.
E-mail:[email protected] oceancurrents.Previousstudieshaveonlyinvestigated
DOI:10.1175/2009JCLI2613.1
(cid:2)2010AmericanMeteorologicalSociety
Report Documentation Page Form Approved
OMB No. 0704-0188
Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and
maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information,
including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington
VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if it

does not display a currently valid OMB control number.
1. REPORT DATE 3. DATES COVERED
15 JAN 2010 2. REPORT TYPE 00-00-2010 to 00-00-2010
4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER
The Effects of SST-Induced Surface Wind Speed and Direction Gradients
5b. GRANT NUMBER
on Midlatitude Surface Vorticity and Divergence
5c. PROGRAM ELEMENT NUMBER
6. AUTHOR(S) 5d. PROJECT NUMBER
5e. TASK NUMBER
5f. WORK UNIT NUMBER
7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION
Marine Meteorology Division,Naval Research Laboratory,Monterey,CA REPORT NUMBER
9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR’S ACRONYM(S)
11. SPONSOR/MONITOR’S REPORT
NUMBER(S)
12. DISTRIBUTION/AVAILABILITY STATEMENT
Approved for public release; distribution unlimited
13. SUPPLEMENTARY NOTES
14. ABSTRACT

15. SUBJECT TERMS
16. SECURITY CLASSIFICATION OF: 17. LIMITATION OF 18. NUMBER 19a. NAME OF
ABSTRACT OF PAGES RESPONSIBLE PERSON
a. REPORT b. ABSTRACT c. THIS PAGE Same as 28
unclassified unclassified unclassified Report (SAR)
Standard Form 298 (Rev. 8-98)
Prescribed by ANSI Std Z39-18
256 JOURNAL OF CLIMATE VOLUME23
TABLE1.SlopesofthelinearrelationshipsbetweenthewindstresscurlandcrosswindSSTgradient(aC)andthewindstressdivergence
anddownwindSSTgradient(a )listedinpreviousstudiesovervariousregionsinunitsofNm22per8C.Acronymsusedhereinclude:
D
NationalCenterforAtmosphericResearch(NCAR)CommunityClimateSystemModel3.0(CCSM3.0);ModelforInterdisciplinary
ResearchonClimate3.2(MIROC3.2);ScrippsCoupledOcean–AtmosphereRegional(SCOAR)model;andNavalResearchLaboratory
CoupledOcean–AtmosphereMesoscalePredictionSystem(COAMPS)model.Theratioa /a isshownintherightmostcolumn.
C D
Study Region Source a a a /a
C D C D
Cheltonetal.(2001) TropicalPacific QuikSCAT–TMI
38N–18S 0.0075 0.0135 0.56
18S–58S 0.0112 0.0247 0.45
O’Neilletal.(2003) SouthernOcean QuikSCAT–Reynolds 0.0068 0.0124 0.55
Cheltonetal.(2004) SouthernOcean QuikSCAT–AMSR 0.0117 0.0165 0.71
TropicalPacific 0.0084 0.0188 0.45
Kuroshio 0.0057 0.0096 0.59
GulfStream 0.0057 0.0109 0.52
Chelton(2005) TropicalPacific QuikSCAT–TMI 0.0087 0.0145 0.60
0.0080 0.0142 0.56
O’Neilletal.(2005) Agulhas QuikSCAT–AMSR 0.0213 0.0284 0.75
MaloneyandChelton(2006) Agulhas QuikSCAT–AMSR 0.0126 0.0173 0.73
ECMWF 0.0041 0.0077 0.53
NCARCCSM3.0 0.0041 0.0088 0.47
MIROC3.2 0.0074 0.0131 0.56
Kuroshio QuikSCAT–AMSR 0.0044 0.0070 0.63
ECMWF 0.0016 0.0040 0.40
NCARCCSM3.0 0.0032 0.0065 0.49
MIROC3.2 0.0021 0.0059 0.36
Seoetal.(2007) TropicalPacific SCOAR 0.009 0.015 0.60
CoastalCalifornia
368–438N 0.002 0.005 0.40
328–398N 0.006 0.008 0.75
Cheltonetal.(2007) Californiacurrentsystem QuikSCAT–AMSR 0.0213 0.0301 0.71
Haacketal.(2008) Californiacurrentsystem QuikSCATCOAMPS 0.0182 0.0260 0.70
0.0157 0.0193 0.81
SST-induced changes in wind speed across SST fronts 2005).Thesurfacewindstresscurlanddivergencefields
andnotwinddirection.Themainobjectivesofthisstudy arerelatedlinearlytothecrosswindanddownwindSST
are(i)todeterminehowsmall-scaleSSTperturbations gradients, respectively. These curl and divergence de-
affect both surface wind speed and direction from 61 pendenciesaresimulatedtovaryingdegreesinnumer-
years of satellite observations of vector surface winds ical weather predication and climate models (Maloney
and SST and (ii) to determine their cumulative effects and Chelton 2006; Haack et al. 2008) and in regional
on the surface curl and divergence fields over the SST mesoscalemodels(Seoetal.2007;Spall2007b).
fronts associated with four midlatitude ocean currents: Inallpaststudies,thecurlresponsetocrosswindSST
the Kuroshio Extension in the North Pacific, the Gulf gradientshasbeenfoundtobesignificantlyweakerthan
StreamintheNorthAtlantic,theBrazil–MalvinasCon- thedivergenceresponsetodownwindSSTgradients,as
fluence in the South Atlantic, and the Agulhas Return measuredbytheslopesa anda ,respectively,ofthe
C D
Current in the Indian Ocean sector of the Southern linear statistical relations (Table 1). The ratio of these
Ocean. slopes a /a varies between about 0.4 and 0.75. Addi-
C D
Spatial variations in surface wind speed associated tionally, satellite observations have shown strong sea-
withsmall-scaleSSTvariabilitydevelopsurfacecurland sonal variability in a and a (O’Neill et al. 2005;
C D
divergence perturbations with magnitudes comparable Maloney and Chelton 2006; O’Neill et al. 2009, man-
to the large-scale curl and divergence (Chelton et al. uscript submitted to J. Climate, hereafter OCE) and
2004; O’Neill et al. 2003, 2005; Chelton et al. 2007). strong geographical variability between ocean basins
Satelliteobservationsofthesecurlanddivergenceper- (Cheltonetal.2004;OCE)andwithinindividualregions
turbations are remarkably well correlated with small- (O’Neilletal.2005).Fromamesoscalemodelingstudy
scale perturbations in the crosswind and downwind over the Gulf Stream, Spall (2007b) noted that a has
C
componentsoftheSSTgradient,respectively(Chelton a significant quadratic dependence on the large-scale
etal.2001,2004,2007;Chelton2005;O’Neilletal.2003, geostrophicwindspeed.ThisisshownbyOCEtoresult
15JANUARY2010 O’NEILL ET AL. 257
fromthenonlinearitybetweenSST-inducedsurfacewind produce changes in the MABL hydrostatic pressure
speedandwindstressperturbations. gradientsandtotheverticalturbulentstressdivergence
Beyond what is known about the dynamics of the intheMABLasdescribedbelow,ultimatelyleadingto
near-surface wind response to small-scale SST pertur- changesinnear-surfacewindscoupledtothemesoscale
bations, as discussed in section 2, little is known about SSTperturbations.
the dynamics underlying the curl and divergence re- AcrossSSTfronts,variationsinsurfaceheatingcause
sponsestoSSTgradients.Thedifferencesinthecurland cross-frontal pressure gradients near the surface since
divergence responses reveal an opportunity to better acoolerMABLovercoolerwaterformshighersurface
understandhowthemarineatmosphericboundarylayer pressures relative to a warmer MABL over warmer
(MABL)respondstosmall-scaleSSTperturbations. water.Thermallyinducedpressuregradientssoformed
In this study, surface wind speed and direction are can therefore accelerate near-surface flow across SST
obtainedfromtheSeaWindsscatterometeronboardthe isothermsfromcoolertowarmerwaterandviceversa.
Quick Scatterometer (QuikSCAT) satellite and SST is Using an analytical model of the boundary layer,
obtainedfrom theAdvanced MicrowaveScanning Ra- Lindzen and Nigam (1987) attributed acceleration of
diometerforEarthObservingSystem(EOS)(AMSR-E) cross-equatorialsurfaceflownorthoftheequatorialcold
Aqua satellite, as discussed in section 3. The statistical tongue intheeastern tropicalPacificprimarilyto ther-
responses of spatially high-pass filtered surface wind mallyinducedpressuregradientsassociatedwiththecold
speed and direction as functions of SST are shown in tongue meridional SST gradient. Even though some of
section4.Arelativelysimplestatisticalrelationisfound theassumptionsoftheiranalyticalmodelarenotgener-
between surface wind speed and SST despite the com- ally valid (e.g., Battisti et al. 1999; Stevens et al. 2002;
plicated force balances involved in the SST-induced Small et al. 2003), SST-induced pressure gradients are
MABL response described by previous studies. In sec- indeed strong contributors to the SST-induced surface
tion 5, the QuikSCAT vorticity and divergence of the wind response (e.g., Wai and Stage 1989; Warner et al.
surfacewindfieldareshowntodependonthecrosswind 1990; Small et al. 2003; Cronin et al. 2003; Mahrt et al.
anddownwindSSTgradientsinamanneranalogousto 2004;Bourrasetal.2004;Songetal.2004;Smalletal.2005b;
thedependenciesofthewindstresscurlanddivergence Song et al. 2006). Using mesoscale numerical simula-
considered in our previous studies. While the vorticity tions over the equatorial Pacific, Small et al. (2003)
and divergence responses to SST gradients have been showed that near-surface pressure perturbations form
interpretedpreviouslyintermsofSSTeffectsonsurface inresponsetoSST-inducedMABLairtemperaturevar-
wind stress magnitude (or wind speed), we show that iations,whicharegovernedthermodynamicallybyabal-
SST also modifies the surface wind direction, thus also ance between horizontal temperature advection and
contributing to SST-induced changes in the surface surfacesensibleheatfluxes.Fromsurfacepressureobser-
vorticity and divergence. These SST-induced modifica- vations over the equatorial Pacific, Cronin et al. (2003)
tionsofwinddirectionareshowntoberesponsiblefor have shown that SST-induced surface pressure gradient
thedifferingvorticityanddivergenceresponsestoSST. perturbations associated with the northern equatorial
We close in section 6 with a discussion of the implica- cold tongue are of a magnitude similar to those gener-
tions of the results of this study for coupled mesoscale atedbymesoscalemodelsimulations.Hashizumeetal.
wind–SSTinteractions. (2002)havealsosuggestedthatdeepeningandshoaling
of the MABL associated with cross-frontal turbulent
mixingvariationscounteractsthisthermodynamiccon-
2. Dynamicalexplanationsformesoscale
tributiontotheSST-inducedsurfacepressurevariations
wind–SSTcoupling
inwhathasbeenreferredtoasa‘‘backpressure’’effect.
Many observational and modeling studies have now Finally, while MABL pressure gradients contribute to
been undertaken to better understand the MABL re- near-surface flow acceleration perpendicular to SST
sponsetoSSTperturbationsovervariousregionsofthe isotherms, it is not clear how they contribute to the
World Ocean. The response of surface winds to meso- generationofnear-surfacevorticityasairblowsparallel
scaleSSTperturbationsistheculminationofadjustment to SSTisotherms, afeature generallyobservedin scat-
processesextendingthroughoutthedepthoftheMABL terometerwindfieldsglobally.
andappearstobecontrolledmainlybymodificationof Wallace et al. (1989) and Hayes et al. (1989) argued
the surface heat fluxes as air flows across SST fronts. thatverticalturbulentmixingofmomentumfromaloft
Surfaceheatfluxesareenhancedorsuppresseddepend- to the surface was more consistent with the changes in
ingon whether the flow is from cold to warm water or verticalwindshearandtheaccelerationofsurfacewinds
vice versa. Cross-frontal surface heating perturbations occurring across the northern edge of the equatorial
258 JOURNAL OF CLIMATE VOLUME23
Pacificcoldtonguethanwiththermallyinducedpressure dragonthispressure-drivenflow.Inthis‘‘pressure-drag’’
gradients. Earlier aircraft observations over the north mechanism,turbulenceactstoreducethesurfacewinds
walloftheGulfStreambySweetetal.(1981)alsoshowed inanoppositemannertothathypothesizedbyWallace
that turbulent mixing of momentum from aloft to the etal.(1989)andHayesetal.(1989).Thispressure-drag
surfacealsoplayedasignificantroleinnear-surfacewind mechanism was found to operate in a numerical simu-
speedvariations.Withthismechanism,SST-inducedsur- lation overthe Agulhas Return Current (O’Neill et al.
faceheatingmodifiesthestaticstabilityoftheboundary 2010), but only above about 150 m from the surface;
layer,enhancingtheverticalturbulentmixingofmomen- belowthislevel,thenear-surfacewindswereinfluencedby
tumasairblowsfromcooltowarmwaterandreducing theturbulentmixingofmomentuminaccordancewiththe
it as air blows from warm to cool water. Observations Wallaceetal.(1989)andHayesetal.(1989)mechanism.
of the near-surface vertical turbulent momentum flux Samelsonetal.(2006)arguedthatchangesinboundary
throughouttheworld’soceanshavegenerallyshowncon- layer depth somewhat downwind of the sharpest SST
siderable variation upon crossing SST fronts and have gradients,ratherthantheverticalredistributionofmo-
supportedthishypothesis(MeyandWalker1990;Freihe mentum, could act to change the surface stress. This
etal.1991;Bond1992;Jury1994;RouaultandLutjeharms mechanismactsthroughabalanceofafixedlarge-scale
2000;Mahrtetal.2004;deSzoekeetal.2005;Tokinaga pressure gradient and the vertical turbulent stress di-
etal.2006). vergence integrated vertically over the depth of the
Modelingstudiesofthesurfacewindresponsetome- boundary layer such that the ratio of surface stress to
soscale SST perturbations have been less consistent boundarylayerheightt/Hisequaltoafixedlarge-scale
abouttherolethattheverticalturbulentstressdivergence pressuregradientintegratedoverthedepthofthebound-
plays in the SST-induced surface wind response. Some arylayer.Thisbalancewasfoundtobeconsistentwith
investigatorsfindasignificantrole(WaiandStage1989; thecross-frontalchangesinsurfacestressandboundary
de Szoeke and Bretherton 2004; Bourras et al. 2004; layer height away from the steepest SST gradients in
Song et al. 2004; Skyllingstad et al. 2006; Thum 2006; thetwo-dimensionalmodelsimulationofSpall(2007b).
Songetal.2009;O’Neilletal.2010)whileothersfound ThisbalanceassumesthattheSST-inducedresponsesof
less significant roles (Small et al. 2003; Bourras et al. the vertical turbulent mixing of momentum, thermally
2004; Small et al. 2005a,b; Song et al. 2006; Samelson induced pressure gradients, and horizontal advection
etal.2006;Spall2007b).Additionally,thecontribution arenotsignificanttermsintheMABLmomentumbud-
ofsurfaceoceancurrentstothesurfacestressmayplay getinregionsdirectlyoverthesharpestSSTfronts(see
asmallbutnonnegligibleroleinthenear-surfaceverti- detaileddiscussioninSmalletal.2008).
cal turbulent stress divergence (Song et al. 2006; Liu At midlatitudes, Coriolis accelerations have been
etal.2007).Idealizedlarge-eddysimulationsoftheflow shown to be important to the response of the surface
across sharp SST fronts by Skyllingstad et al. (2006) winds to small-scale SST perturbations (e.g., Wai and
showed that on spatial scales of 1–20 km, which are Stage1989;Bourrasetal.2004;Thum2006;Songetal.
somewhat smaller than considered here, near-surface 2006;Spall2007b;O’Neilletal.2010).OverasingleSST
windspeedvariationsareinfluencedpredominantlyby front,Spall (2007b)has foundthat SST-inducedturbu-
SST-inducedcross-frontalvariationsintheverticaltur- lent mixing perturbations induce Coriolis accelerations
bulentmixingofmomentum,consistentwiththeWallace analogoustoaninertiallydrivenleewave.O’Neilletal.
etal.(1989)andHayesetal.(1989)hypotheses.Finally, (2010) has found evidence of a similar mechanism in
weakcouplingofsurfacewindstoSSTintheEuropean a simulation over a part of the Southern Ocean. The
CentreforMedium-RangeWeatherForecasts(ECMWF) strong large-scale mean wind speeds typically found
modelovertheAgulhasReturnCurrent(Maloneyand over midlatitude ocean fronts also lead to significant
Chelton 2006) was shown by Song et al. (2009) to be horizontal advective accelerations that strongly influ-
attributable primarily to a weak response of the pa- ence the near-surface momentum budget (Song et al.
rameterizedverticalturbulentstressdivergencetoSST- 2006;Thum2006;Spall2007b;O’Neilletal.2010).
inducedsurfaceheatingperturbations. Some debate remains as to the physical mechanisms
Smalletal.(2005a)envisionedaquitedifferentroleof involved in the coupling between surface winds and
the vertical turbulent stress divergence in a numerical mesoscaleSSTperturbations.Itisbecomingclear,how-
simulation of the near-surface wind response over the ever, that both pressure gradients and the turbulent
easterntropicalPacific.There,itwasfoundthatpressure mixing of momentum are involved significantly in the
gradients accelerated the near-surface flow across the response. There are also likely significant regional dif-
cold tongue SST front, and after some distance, the ferencesinthelarge-scaleatmosphericforcingthatmay
verticalturbulentstressdivergenceactedasafrictional alter the force balance involved in the surface wind
15JANUARY2010 O’NEILL ET AL. 259
response.Weshowbelowthattherearecomparatively spatialgrid.Thisavoidscomplicationsarisingfromthe
simplestatisticalrelationshipsbetweenthesurfacewinds estimationofspatialderivativesnearswathedgeswhere
and SST despite the rather complicated detailsof these centeredfirst-differenceestimatesofderivativesarecom-
physicalmechanisms. putedusingmeasurementsfromneighboringswathssep-
aratedintimebyseveralhoursormore.Thevariouswind
and derivative wind fields of interest in this study were
3. Descriptionofsatelliteobservationsandmethods
then constructed on a 0.258 latitude by 0.258 longitude
This study utilizes monthly-averaged surface vector spatial grid by fitting in-swath measurements to a qua-
wind observations from the QuikSCAT scatterometer dratic surface using locally weighted regression (‘‘loess’’
and SST from the AMSR-E over the 75-month period smoothing;ClevelandandDevlin1988)withahalfspanof
June 2002–August 2008, coinciding with the first 61 80 km. Based on the filter transfer function of the loess
years of the AMSR-E geophysical observational data smoother (Schlax et al. 2001; Chelton and Schlax 2003),
recordthatbegan1June2002.Eachofthesedatasetsis the resulting gridded fields have a spatial resolution
summarizedinthissection. analogoustothatofapproximately50-kmblockaverages.
Thespatialderivativefieldswerefoundateachgridpoint
a. QuikSCATwindfields
directlyfrom theregressioncoefficientsofthequadratic
In nonprecipitating conditions, scatterometers infer surface. The wind fields were then averaged at monthly
surfacevectorwindsoverwaterfrommicrowaveradar intervalsfromthegriddedswathdata.
measurements of small-scale surface roughness caused
b. AMSR-ESSTfields
bysurfacewindstress.Forlackofabundantdirectsur-
facestressmeasurements,scatterometermeasurements Thedetailedresultsofthisstudywouldnotbepossible
of microwave radar backscatter are calibrated to buoy without the near-all-weather microwave observations of
anemometermeasurements of the so-called equivalent SST over middle- and high-latitude regions provided by
neutral stability wind at 10 m, which is the 10-m wind theAMSR-EsinceJune2002(CheltonandWentz2005).
thatwouldbeuniquelyrelatedtotheobservedsurface ThemainadvantageofmicrowavemeasurementsofSST
wind stress if the atmosphere were neutrally stratified is the ability to measure SST through nonprecipitating
(LiuandTang1996).Therelationbetweentheequiva- clouds, which are essentially transparent to microwave
lentneutralstabilitywindvectoruandthesurfacewind radiation. Cloud cover generally obscures the midlat-
stressvectortist5r CNVu,wherer isthesurfaceair itude ocean regions of interest in this study more than
0 D 0
density,V is themagnitudeof u,and CN istheneutral 70% of the time (Chelton and Wentz 2005). Large re-
D
stabilitydragcoefficient. gions of missing data often occur in SST fields con-
The computations throughout this paper using structed from infrared satellite measurements over the
QuikSCATwindsarebasedon10-mequivalentneutral sharpest ocean fronts and eddies where strong ocean–
stabilitywinds.Onaverage,anemometermeasurements atmosphereinteractionsareexpectedtooccur(see,e.g.,
ofthe10-mwindspeedareabout0.2 m s21lowerthan Fig. 6 of Chelton and Wentz 2005). Indeed, stratocu-
the corresponding neutral-stability wind speed at 10 m mulus and cumulus cloud bands often form over sharp
(Mears et al. 2001; see also Fig. 16 of Chelton and SSTfrontsasaconsequenceofsomeofthesameMABL
Freilich2005)becausetheatmosphericsurfacelayeris, adjustment processes affecting MABL winds of inter-
on average, slightly unstable over all oceans. Through estinthisstudy(e.g.,NorrisandIacobellis2005).While
comparisons with high-quality buoy anemometer mea- SST measurements by the Tropical Rainfall Measuring
surements,QuikSCATwindspeedmeasurementerrors Mission (TRMM) Microwave Imager (TMI) have been
havebeenestimatedtobeabout1.7 m s21(Cheltonand available since December 1997 and thus encompass the
Freilich 2005). The wind direction accuracy improves entireQuikSCATmission,theyareonly made equator-
withincreasingwindspeed;forwindspeedsgreaterthan wardof408latitude,leavingmostofthemidlatitudere-
about5 m s21,thedirectionaccuracyofindividualwind gionsofinterestinthisanalysisunsampled.
measurements is better than 158 (see Fig. 8 in Chelton TheAMSR-Emeasureshorizontallyandvertically
andFreilich2005).Additionally,thereisnoevidenceof polarized brightness temperatures at six microwave
SST-dependent measurement biases in the QuikSCAT frequencies across a 1450-km swath. From these 12 mi-
wind observations (Chelton et al. 2001; Ebuchi et al. crowave brightness temperatures, SST is estimated over
2002). mostoftheglobaloceansusingaphysicallybasedstatis-
The spatial derivatives of wind components, wind ticalregressionalgorithm(WentzandMeissner2000).The
speed, and wind direction analyzed in this study were footprintsizeofAMSR-EmeasurementsofSSTis56km
computed ‘‘in-swath’’ before interpolating to a regular andthermsaccuracyofindividualSSTmeasurementsis
260 JOURNAL OF CLIMATE VOLUME23
about0.48C (CheltonandWentz2005).TheSSTfields terest here, the effects of synoptic weather variability
weregriddedontothesame0.258gridastheQuikSCAT were mitigated by using monthly-averaged wind and
surface wind fields and then averaged at monthly in- SST fields. It is noted that this spatial filtering was ap-
tervals. plied on all variables after computations involving the
windandSSTfieldswereperformed,includingthenon-
c. CrosswindanddownwindSSTgradient
linear crosswind and downwind gradient computations.
computations
Thisispreferabletocomputingthecrosswindanddown-
Asdiscussedintheintroduction,previousstudieshave windgradientquantitiesfrommonthly-averagedfields.1
shown that the wind stress curl and divergence fields
are found to be related linearly to the crosswind and
downwindcomponentsoftheSSTgradient,respectively. 4. ObservationsofSST-inducedsurfacewind
Crosswind and downwind gradients in this study were response
computed within each QuikSCAT measurement swath
a. SurfacewindspeedresponsetoSST
usingcurvilinearnaturalcoordinates(s,n),wheresand
n are local along-wind and crosswind coordinates, re- Maps of the QuikSCAT perturbation surface wind
spectively(e.g.,HaltinerandMartin1957;Holton1992). speedandtheAMSR-ESSTfieldsscalar-averagedover
The unit vector ^s is in the same direction as u. The the75-monthanalysisperiodareshowninFig.1foreach
positiveunitvectorn^ points908counterclockwiserela- of the four regions considered here. Large horizontal
tive to^s. The crosswind and downwind components of variationsinsurfacewindspeedoccurnearmeandering
the SST field T(x, y) in natural coordinates are com- SSTfronts,withstrongerwindspeedsoverwarmerwater
putedfromCartesianderivativesby and weaker wind speeds over cooler water. Typically,
surface wind speed variations of more than 1–2 m s21
›T ›T ›T accompany SST variations of about 28–48C over cross-
5(cid:2)sinc 1 cosc and
›n ›x ›y frontal distances of O(100 km). This is an appreciable
›T ›T ›T fractionoftheunfilteredscalar-averagedwindspeedsin
5 cosc 1 sinc , (1)
›s ›x ›y theseregions(Fig.2),whicharetypically8–13 m s21in
these long-term time averages. For the time-averaged
where (x, y) are the zonal and meridional Cartesian perturbationfields,thespatialcross-correlationbetween
coordinates,respectively,andcisthecounterclockwise the perturbation wind speed and the perturbation SST
surfacewinddirectionrelativetothexaxis. varies between 0.7 over the Kuroshio to 0.85 over the
CrosswindanddownwindSSTgradientswerecomputed SouthAtlantic(Table2).Inadditiontothequasi-stationary
hereonaswath-by-swathbasisbycombiningwindmea- coupledpatternsintheperturbationwindandSSTfields
surementsfromeachQuikSCATswathwithSSTgradi- apparentintheselong-termtime-averagedmaps,there
ents computed from gridded 3-day averaged AMSR-E is also significant coupled variability associated with
SSTfieldscenteredoneachdayoftheQuikSCATmea- transientSSTfeaturesevidentfromsatellitedataoverthe
surementswath.SwathsofcrosswindanddownwindSST Kuroshio (e.g., Nonaka and Xie 2003), the Gulf Stream
gradientswerethenaveragedatmonthlyintervals.This (e.g., Park and Cornillon 2002; Park et al. 2006), and all
procedureminimizesuncertaintiesassociatedwithcom- four midlatitude regions considered here (White and
puting the nonlinear crosswind and downwindSST gra- Annis2003).Thewind–SSTcouplingintheKuroshiore-
dients from wind and SST measurements that are not gionismuchmoreapparentduringtheborealwintertime
collocatedspatiallyandtemporally.
d. Spatialhigh-passfiltering
1Crosswind and downwind gradients of SST in our previous
SST-induced mesoscale wind variability was isolated studies(Cheltonetal.2001,2004;O’Neilletal.2003,2005)were
by removing large-scale spatial variability using a mul- computed from gridded, vector-averaged wind stress and SST
gradientcomponentsratherthanbeingcomputedonaswath-by-
tidimensionalloesssmoothingfunctionwithhalf-power
swathbasisandthenaveraging,whichismoreaccuratebecauseof
filtercutoffsof108latitude3208longitude(Cleveland
thenonlinearityofthesequantities.Asaresult,theSST-induced
andDevlin1988).Thisfilteringeffectivelyremovesthe windresponseinthosestudieswassomewhatunderestimatedfrom
large-scalecomponentsofthewindandSSTfieldsthat thoseobtainedfromthemoreaccuratelycomputedquantitiesused
areunrelatedtothewind–SSTcouplingofinteresthere. here. Additionally, the definition of the crosswind SST gradient
used in those studies ($T 3u^) (cid:3) k^is equal to 2›T/›n and the
Wind and SST fields spatially high-pass filtered in this
downwindSSTgradient$T (cid:3) u^isequalto›T/›s,whereu^isaunit
wayarehereafterreferredtoasperturbationfields.To vectorinthedirectionofthesurfacewindvectorandk^isaunit
isolatetheSSTinfluenceonsurfacewindsthatisofin- vectorintheverticaldirection.
15JANUARY2010 O’NEILL ET AL. 261
FIG.1.MapsoftheperturbationQuikSCATwindspeed(colors)andAMSR-ESST(contours)scalar-averaged
overtheperiodJune2002–August2008forthefourregionsconsideredinthisstudy.Thecontourintervalforthe
perturbationSSTis0.58C,andthezerocontourhasbeenomittedforclarity.Solidanddashedcontourscorrespondto
positiveandnegativevalues,respectively.Thespatialhigh-passfilterremovesspatialvariabilitywithwavelengths
longerthan208longitude3108latitude,asdiscussedinthetext.
(e.g.,NonakaandXie2003;OCE)andisassociatedwith (seeTable2).Additionally,a ismuchlargerinthetwo
y
strongseasonalityoftheSSTfrontsinthisregion(e.g., SouthernHemisphereregionscomparedtothoseinthe
NonakaandXie2003). Northern Hemisphere. The larger error bars at large
The mesoscale response of the wind speed to SST is perturbation SST magnitudes in Fig. 3 are due to the
quantifiedstatisticallybybin-averagingmonthlyaverages smaller numbers of samples in these bins. This is par-
oftheperturbationQuikSCATwindspeedasafunction ticularly acute over the Kuroshio, where the monthly-
of the perturbation AMSR-E SST over the 75-month averaged SST perturbations tend, on average, to be
analysis period (Fig. 3). Wind speed perturbations are smaller in magnitude than in the other three regions.
related linearly to, and correlated positively with, me- The geographical differences in a between the four
y
soscale SST perturbations over all four regions. The regions are likely due to geographic differences in the
relationshipbetweenthemonthly-averagedspatiallyhigh- verticalstructureoftheboundarylayerandlarge-scale
pass filtered wind speed V9 and SST T9 can thus be ex- forcing.Seasonalityofa isinvestigatedfrom51years
y
pressedempiricallyas ofsatellitedataoverthesefourregions(andtheeastern
tropicalPacific)inOCE.
V95a T9, (2) Finally, we note that for SST perturbations warmer
y
than about 1.58, there is a subtle flattening of the bin-
where a is the least squares estimate of the slope of averaged V9 relative to the linear fit, particularly over
y
theselinearrelationsandrepresentsthechangeinwind theSouthAtlanticandAgulhasReturnCurrentregions.
speedperunitchangeinSST(i.e.,a 5›V9/›T9)andthe Interestingly, this flattening does not occur over cool
y
primes hereafter represent spatially high-pass filtered SST perturbations. This difference between the bin-
fields. Notably, a varies by nearly a factor of 2 geo- averagedV9andthelinearfitislessthanabout0.25m s21,
y
graphically,from0.27m s21per8CovertheGulfStream andthereareveryfewdatapointsintheseouterbins,as
to 0.49 m s21 per 8C over the Agulhas Return Current shownbythehistogramsinFig.3.Becausethisflattening
262 JOURNAL OF CLIMATE VOLUME23
FIG.2.MapsoftheunfilteredQuikSCATwindspeed(colors)andAMSR-ESST(contours)scalar-averagedover
theperiodJune2002–August2008.TheSSTcontourintervalis28C.
of V9 is relatively small and affects only a very small proacheszero,inwhichcasethewinddirectionbecomes
percentageoftheobservations,itdoesnotsignificantly undefined.Bothcasesoccurintermittentlyineachofthe
affect the analysis presented here. This observation is four regions under consideration. To avoid these diffi-
currentlythesubjectofongoinganalysis. culties, mesoscale wind direction perturbations are de-
fined here as the counterclockwise angle between the
b. WinddirectionresponsetoSST
unfiltereduandthespatiallylow-passfilteredu~,where
The wind direction over these four regions is shown thetilderepresentsspatiallow-passfiltering.Thesewind
inFig.4bythestreamlinesofthevector-averagedsur- direction perturbations were computed from vector-
facewindfromQuikSCATforthe75-monthperiodcon- averagedwindcomponentsatmonthlyintervalsonlyat
sidered here. The surface flow in all four regions has a points where the magnitude of the monthly vector-
strongwesterlycomponent,characteristicoftypicalmid- averagedwindspeedexceeded1 m s21.
latitudesurfaceflow.Theflowimmediatelydownstream Spatialmapsoftheso-definedmesoscalewinddirection
oftheAsianandNorthAmericancontinentshasasig- perturbations scalar-averaged over the 75-month anal-
nificant northerly component before gradually turning ysisperiodareshowninFig.5alongwithcontoursofthe
northwardfaroffshore.OvertheSouthernHemisphere
regions,theflowiswesterlypolewardofthelarge-scale
subtropicalanticycloneslocatednear308–358S. TABLE 2.The(left)couplingcoefficient ay and(right)spatial
correlationcoefficientcomputedfromthemonthly-averagedper-
Forquantitativepurposes,itisdesirableheretoconsider
turbationQuikSCATwindspeedandAMSR-ESSTfields.
wind directionperturbations analogousto thespatially
high-passfilteredwindspeedperturbationsinvestigated ay[ms21(8C)21] Correlationcoefficient
insection4a.However,scalarwinddirectioncannotbe Kuroshio 0.30 0.70
spatiallyhigh-passfilteredin thesamemanneraswind GulfStream 0.27 0.79
speed in regions straddling absolute jumps in wind di- SouthAtlantic 0.43 0.87
Agulhas 0.49 0.83
rection exceeding 3608 and where the wind speed ap-
15JANUARY2010 O’NEILL ET AL. 263
FIG.3.Binnedscatterplotsofthespatiallyhigh-passfilteredQuikSCATwindspeed(V9)asafunctionofthe
spatiallyhigh-passfilteredAMSR-ESST(T9).Thebinaverageswerecomputedfrommonthly-averagedQuikSCAT
windandAMSR-ESSTfieldsoverthe75-monthanalysisperiodusingallavailabledatapointsintheregionsen-
closedinFigs.1and2.Thepointsanderrorbarsrepresentthemeansandstandarddeviationswithineachbin,
respectively,andthedashedlinesarelinearleastsquaresfitstothemeanswithineachbin.Beloweachscatterplot,
ahistogramofthespatiallyhigh-passfilteredAMSR-ESSTisincludedforreference.
AMSR-E perturbation SST. Perhaps the most striking perturbationscannonethelessbeobtainedherethrough
featureofthesewinddirectionperturbationsisthepro- cross-spectral analysis of the wind direction and SST
nounced meridional displacement of the wind direction perturbations in the meridional wavenumber domain.
perturbationsrelativetothoseofSST.IntheNorthern Estimatesofcross-spectralstatistics(squaredcoherence,
Hemisphere,negativewinddirectionperturbationsare transferfunction,andphasespectraasfunctionsofme-
shifted north of warm SST perturbations and positive ridional wavenumber) were computed at monthly in-
winddirectionperturbationsareshiftednorthofcoolSST tervalsalongcontinuousmeridionalsegmentswithineach
perturbations.Conversely,intheSouthernHemisphere, ofthefourregionsandthenensembleaveragedoverall
positive wind direction perturbations are shifted south 75monthsandsegments(Fig.6).
ofwarmSSTperturbationsandnegativewinddirection The squared coherence is largest overthe South At-
perturbationsareshiftedsouthofcoolSSTperturbations. lanticandAgulhasReturnCurrent,whereperturbation
Meridional displacements of the wind direction per- SSTvariabilityaccountsfor10%–15%ofthemeridional
turbationsrelativetothoseofSSTthatarequalitatively variance in the perturbation wind direction. Over the
evidentin Fig. 5precludethe useof simple binned av- Gulf Stream and Kuroshio, the squared coherence is
eragestostatisticallyquantifytherelationshipbetween below0.05.Winddirectionperturbationsaretherefore
winddirectionandSSTsuchaswasperformedwiththe lesswellcorrelatedwithSSTcomparedwithwindspeed.
wind speed perturbations in section 4a (Fig. 3). Statis- Thetransferfunctionrepresentsanestimateofthecou-
ticalrelationships betweenthewind directionand SST plingcoefficientbetweentheperturbationwinddirection