Multidecadal Climate Variability in the Greenland ... - Semantic Scholar
GEOPHYSICAL RESEARCH LETTERS, VOL. 24, NO. 3, PAGES 257-260, FEBRUARY 1, 1997
Multidecadal climate variability in the Greenland Sea and surroundingregions:a coupledmodel simulation ThomasL. Delworth,SyukuroManabeandRonaldJ. Stouffer Geophysical FluidDynamicsLaboratory/NOAA, Princeton, New Jersey
Abstract. Pronounced oscillationsof oceantemperature temperature (SST) and climate in the Greenland Sea (see and salinityoccurin the GreenlandSea in a 2000 year Fig. 1) andsurroundingregions. integrationof a coupledocean-atmosphere model. The oscillations,involving both the surfaceand subsurface Model Description and Experimental Design oceanlayers, have a timescaleof approximately40-80 years, and are associatedwith fluctuationsin the intenThe coupledocean-atmosphere model usedin this studyis sity of the East GreenlandCurrent.The GreenlandSea identical to that describedin detail in Manabe et al. (1991). temperatureand salinity variations are precededby The model is global in domain,includingthe Arctic, with relarge-scalechangesin near-surface salinityin the Arctic, alistic geographyconsistentwith resolution.The model is which appearto propagateout of the Arctic throughthe forcedwith an annualcycle of insolationat the top of the atEastGreenlandCurrent.Theseanomaliesthenpropagate mosphere.The atmosphericcomponentnumerically intearoundthe subpolargyre into the LabradorSea andthe gratesthe primitiveequationsof motionusinga semi-spectral central North
Atlantic.
These oscillations
are coherent
with previouslyidentifiedmulti-decadalfluctuationsin the intensityof the North Atlantic thermohalinecirculation. The oscillationsin the G•:eenlandSea are related to atmosphericvariability. Negative (cold) anomaliesof surface air temperatureare associatedwith negative
techniquein which the variablesare representedby spherical harmonicsandby corresponding grid pointswith a spacingof 7.5ø longitudeand 4.5ø latitude.There are 9 unevenlyspaced levels in the vertical. The oceaniccomponentof the model uses a finite differencetechniquewith 12 unevenly spaced levelsin the vertical, and a horizontalresolutionof approximately 3.75ø longitudeand 3.75ø latitude.Sea ice is predicted
(cold)seasurface temperature (SST)anomalies in the Greenland Sea,with amplitudes up to 2øCnearGreen- usinga simplemodeldevelopedby Bryan (1969), which allanddecliningto severaltenthsof a degreeC overnorth-
lows sea ice less than 4 m thick to move with the ocean cur-
western Europe.Thec01dSSTanomalies andintensified rents. The
East
Greenland
Current
are
also
associated
with
model atmosphereand ocean interact through
fluxes of heat, water, and momentum at the air-sea interface.
enhancednortherlywindsoverthe GreenlandSea.
The coupledmodelemployedfor thisstudyhasa relatively low spatialresolutionand simplifiedphysicalparameterizations.In spiteof thesesimplifications,the model has successIntroduction fully reproducedmany aspectsof observedinterannualto decadalvariability (Manabe and Stouffer,1996). We feel that The Great Salinity Anomaly (GSA; Dickson et al, 1988) importantinsightscan be gainedfrom suchmodelsregarding was a spectacular exampleof decadalscaleoceanicvariability the fundamentalworkingsof the coupledclimatesystem. in thehighlatitudes of theNorthAtlantic.Suchevents areof In orderto reduceclimatedrift, adjustmentsto the heat and potentiallyprofoundclimatic importancebecauseof the cenwaterfluxesare appliedat the air-seainterfaceoverbothopen tral role of the high latitudesof theNorth Atlanticin the global and sea-icecoveredregions.These flux adjustmentsare dethermohalinecirculation(Walsh and Chapman,1990). rivedfrom preliminaryintegrationsof the separateatmospher-
AagaardandCarmack(1989)suggest thattheGSAoriginat- ic and oceaniccomponents (see Manabeet al., 1991, for
ed throughan enhancement of thefreshwaterandseaice flux-
details).The adjustments vary seasonallyandspatiallybut are constantfrom one ye• to the next. Since the adjustments experiment(H'fikkinen,1993) supportthis. The fresh water were derivedprior to the startof the coupledmodel integraanomalymoveddownthe EastGreenlandCurrent(EGC), fol- tion they are not correlatedwith sea surfacetemperatureand lowedthe subpolargyrethroughtheLabradorSeaandinto the salinity anomalies.They are thereforeunlikely to systematicentralNorth Atlantic,eventuallyreturningto the Greenland cally amplify or dampensuchanomalies.In addition,the ades from the Arctic in the late 1960s. The results of a numerical
Sea in the early 1980s.Evidencethat a similareventmay have justments helpto ke6pthemodelneara realisticmeanstateso that variousfeedbackprocesses,suchas ice-albedofeedback, Dicksonet al. (1988; seetheirFig. 8). canhavea realisticinfluence onthemodelvariability. In this paper we describefluctuationsof surfacesalinity After preliminaryintegrationsto achievean initial condiwhich resemblethe GSA in a 2000 year integrationof a coution in approximateequilibrium,the modelwastime integratpled ocean-atmosphere model.The salinityfluctuationsare ased for a periodof 2000 years.The outputfrom thisintegration sociated with substantial multidecadal variations of sea surface formsthe datasetfor the analysespresentedbelow.
occurred in theearlypartof the20thcentury ispresented by
Thispaperis notsubjectto U.S. copyright. Published in 1997 by theAmericanGeophysical Union.
Descriptionof Model Variability
Papernumber96GL03927.
multidecadal variationsof seasurfacesalinity(SSS)andSST
Analysesof the integrationdescribedabovehave revealed 257
258
DELWORTH ET AL.: MULTIDECADAL
CLIMATE VARIABILITY
IN THE GREENLAND
SEA
2
.., .... ooo -2
0
250
500
750
1000
'
, 1250
1500
1750
2000
Year
Figure1.Timeseries of annual meanmodelSSTin theDenmark Strait(15øWto26øWat70øN).Priortoplotting,thetimeseries wasfilteredsuchthattimescales shorter than10yearswereeffectively removed. Thevaluesareanomaliesfrom a longtermmeanof 0.62. UnitsareøC. in the GreenlandSea region which are reminiscentof the GSA. Someaspects of thiswereseenin ManabeandStouffer (1996), who notedthat substantial,low frequencyvariations of SST andnear-surface air temperature occurredin thevicinity of the GreenlandSea(seetheirFig. 7b). The time series of annual mean model SST from a similar
regionof the currentintegration(shownin Fig. 1) is characterizedby multidecadal variationsof temperature, with peakto-peakdifferences of morethan3øC.The SST variations are tightly coupledto SSS variationsin the sameregion(linear
fy andspreadto covernearlythe entireGreenlandSeaand portions of thecentralNorthAtlanticby Lag0 (Fig.4). In addition, an area of positiveSST anomaliesdevelopsoff the North Americancontinent.This patternappearswhenthe intensityof themodelthermohaline circulation is weakest(Delworthet al., 1993;seeFig. 6a), andis reminiscent of the SST anomaliesseenin observations (Kushnir,1994). In additionto advection,the SST and SSS anomaliesin the GreenlandSea are stronglyinfluencedby convection. Freshwateranomalies in the near-surface layer caninhibitconvection, therebysup-
the heat and saltexchange with sub-surface layers correlation= 0.92), with typical amplitudesof 0.2-0.3 parts pressing per thousand. The spectrumof SST shownin Fig. 2 reveals andamplifyingthe surfaceanomalies. Atmosphericvariationsare also associated with the SST enhancedvarianceat 40-80 years,with a spectralpeakat 5060 years.Visualinspection of Fig. 1 showsthat thereis a fluctuationsshownin Fig. 1. Surfaceair temperatureanomamodulationof the amplitudeof this 40-80 year variability. lies over the Greenlandand NorwegianSeaswith amplitudes with SST in theDenmark Waveletanalysisand singularspectrumanalysiswere per- up to 2øCarepositivelycorrelated formed on this time series(not shown) and reveal that this Strait, with anomaliesof smaller magnitudeextendinginto modulation has a time scale of several centuries. northwestern Europe.Thereare alsoassociated variationsin In orderto describethe space-time structure of thisoscilla- surfacepressure,as shownin Fig. 4. The surfacepressure 2 years tion, linear regressions were computedat variouslagsbe- anomalieshavetheir largestamplitudeapproximately tween the time seriesof severalvariables(oceantemperature, beforethe largestmagnitudeof the SST variations.The spato the anomalous pressure patsalinity,andcurrents) at eachgridpointversus thetimeseries tial patternhasa resemblance of SST shownin Fig. 1. The regression analyses, computed
usingannualmeandata,revealthechanges in thesevariables associated with
SST anomalies in the Denmark
PERIOD (years)
Strait.
200
Figure3 showsthe progression of SSSanomalies from the Arctic to the North Atlantic. Conditionsat Lag -15 indicate fresh anomaliesdominatingthe Arctic. At Lag -10, these
•
:
100
50
20
10
5
3
2
_'._.... .....
10
fresh anomalies still dominate the Arctic, and have extended
southwarddown the EGC and to the regionoff Newfound-
land. The superimposed currentanomaliesshowa stronger thannormalEGC enhancingthe freshwatertransportout of theArctic.At Lag 0 the freshwater anomalies extendaround
the subpolar gyreto thecentralNorthAtlantic.At thistime, positiveSSSanomalies areseento develop in theArctic,a harbinger of thedeveloping opposite phaseof theoscillation. The increasein Arctic SSS is consistentwith the anomalously
1
ol
o01
ol
FREQUENCY (yr-1)
strongtransport of freshwateroutof theArctic.At thesame Figure2. Spectrumof the time seriesof annualmeanmodel time that the surface currentsindicate enhancedflow from the Arctic into the Greenland Sea, there is an anomalousflow at
SST in the Denmark Strait (15øW to 26øW at 70øN; this is the time seriesfrom Fig. 1 prior to filtering).The logarithmof the
estimates (thick,solidline)is plottedversusthelogadepthfrom the Greenland Seainto the Arctic(not shown). spectral Additionalregression analyses (notshown)revealthatposi- rithm of frequency.The thin, solidline denotesthe spectrum tiveseaice anomalies aregenerallyin phasewiththenegative of a backgroundfirst orderMarkov process,andthe dashed SSS and SST anomalies.
lines denote the 95% confidence interval about that back-
The spatialpatternof SST anomalies is shownin Fig. 4. groundspectrum.The spectrumwasestimatedby takingthe Startingfrom small negativeanomaliesat Lag -15 (not Fourier transform of the autocovariancefunction using a shown),negativeSSTanomalies in theDenmarkStraitampli- Tukeywindowwith a maximumlag of 200.
DELWORTH ET AL.: MULTIDECADAL CLIMATE VARIABILITY IN THE GREENLAND SEA
?L- TM ...
(
-1•
-•
ß
-o.•
• .:
-0.2
,
-o.s
',.
o.•
259
, ..... ß....
....
•;,_
/
.,'
..... ,
0.2
Figure4. SameasFig. 3 usingthe regressions of SST at each grid point versusthe time seriesof annualmean SST in the Denmark Strait (15øW to 26øW at 70øN). The regressioncoefficientsof SST are indicatedby the color shading.Superimposedusing red contourlines are the regressions of surface pressure(mb per-2øC SST anomaly).Resultsare shownfor Lag 0 for SST, and Lag -2 for surfacepressure(i.e., surface pressureleadingSST by 2 years). ternprecedingthe GSA (WalshandChapman,1990), although smaller in magnitude.The associatedcirculationindicatesa weakeningof the meansoutherlywindsover the Nordic Seas when SSTs are negative,coincidentwith an enhancedEGC. This association suggests thepossibilityof activeatmospheric participationin this modeloscillation.A similarpossibility wassuggestedby WohllebenandWeaver(1995). Discussion
Previous analyses(Delworth et al., 1993) of this model have revealed multidecadal
variations of the THC in the North
Atlantic with a time scaleof 40-80 years,generallysimilarto thetimescaleof the variabilitydescribedabove.In orderto investigatethe relationshipbetweentheseTHC variationsand the GreenlandSea SSS and SST oscillations,the squaredcoherencywas computedbetweena time seriesof the THC intensity and the SST time series in Fig. 1. The squared coherencyreachesa maximumof 0.50 at a periodof approxi-
Figure3. Regressions of annualmeanSSSat eachgrid point versus the time series of annual mean SST in the Denmark
Strait (15øW to 26øW at 70øN). The contouredvaluesare the
slopesof the regressionlines multipliedby -2 (to indicate conditionsassociatedwith an SST anomalyof-2øC). The fieldat Lag -15 indicatesconditions 15 yearspriorto a-2øC SST anomalyin theDenmarkStrait.Unitsarepartsperthousandper-2øCSSTanomaly.In addition,theregressions of the -0.
-0.!
-0.
-0.02
0
0.02
0.05
0.1
0.2
surface currents areindicated bythevectors. Unitsarecms-1 per-2øCSST anomaly.For clarityof plotting,notall current vectorsareshownat veryhighlatitudes.
260
DELWORTH ET AL.: MULTIDECADAL CLIMATE VARIABILITY IN THE GREENLAND SEA
References mately 57 years,with the THC time serieslaggingthe SST (and SSS) in the Denmark Strait by approximately10 years. Aagaard,K., andE.C. Carmack,The role of seaice andfreshwater The THC is weakestwhen SSS over the sinkingregionof the in the Arctic circulation,J. Geophys.Res., 94, 14,485-14,498, THC (definedas the areafrom 50øN to 70øN in the North At1989. lantic)is at a minimum(seeFig. 9a of Delworthet al., 1993). Briffa, K.R., P.D. Jones, T.S. Bartholin, D. Eckstein, This relationship leadsusto speculate thatenhanced transport F.H. Schweingruber, W. Karlen, P. Zetterberg,and M. Eronen, of relativelyfreshwaterand seaice from the Arctic, through Fennoscandian summersfrom AD 500: temperaturechangeson the East Greenland Current and Denmark Strait, and into the shortandlongtimescales, ClimateDynamics,7, 111-119,1992. LabradorSearegionsmay weakenthe THC in the NorthAtBryan, K., Climate and the oceancirculation.Part III: The ocean model, Mon. Wea. Rev., 97, 806-827,1969. lantic. The increaseof relatively fresh, low densitywater in the near-surface layersof the LabradorSea and neighboring D'Arrigo, R.D., E.R. Cook, and G.C. Jacoby,Annual to decadalscale variations in northwest Atlantic sector temperatures regions("fresh water capping")inhibits the convectiveexinferred from treerings,CanadianJournalof ForestResearch, changeof heatbetweenthe cold near-surface layersandthe 26,143-148,1996. relativelywarmersub-surface layers(Lazier, 1980, provides observationalevidencefor this). This reducedsupplyof heat Delworth, T., S. Manabe, and R.J. Stouffer,Interdecadalvariations of the thermohalinecirculationin a coupledocean-atmosphere to the surfacelayer inhibitsair-seaheatexchangeby cooling model, J. Climate, 6, 1993-2011, 1993. the surfacewaters,therebyreducingthe supplyof negative Dickson, R.R., J. Meinke, S.A. Maimberg, and A.J. Lee, The buoyancyto the near-surface layer andweakeningtheTHC. "Great Salinity Anomaly"in the northernNorth Atlantic 1968The abovediscussionsuggeststhat the GreenlandSea os1982,Progressin Oceanography, 20, 103-151,1988. cillationcouldhavea role in generatingthemultidecadalvari- H•ikkinen,S., An Arctic sourcefor the great salinityanomaly:a ationsof the THC throughfreshwater cappingof its sinking simulationof the Arctic ice-oceansystemfor 1955-1975, J. region.However,there is also the possibilitythat the mulGeophys.Res.,98, 16397-16410. tidecadalvariationsof theTHC may havea role in generating Kushnir, Y., Interdecadal variations in North Atlantic sea surface the GreenlandSea oscillation.It is difficult to distinguishbetemperatureand associated atmosphericconditions,J. Climate, 7, 141-157, 1994. tweenthesetwo possibilities basedon onenumericalexperiment, in which all processesare inherently coupled. Lazier, J.R.N., Oceanographicconditionsat oceanweathership Bravo 1964-1974,Atmosphere-Ocean, 18, 227-238, 1980. Therefore,additionalnumericalexperiments will be required to understand both the interactions between these two oscillations and the factors that determine their time-scales.
One possiblemechanismfor decadalscale(10-20 year) variabilitywasput forwardby Mysak et al. (1990), whosuggestedthatanomalous riverrunoffintotheArcticcreatedthe anomalousfreshwater and seaice pulseknownas the GSA. In our model, however,the time seriesof river runoff into the
Arctichasa whitespectrum (i.e.,approximately uniformvarianceat all time scales).This suggests that an explicitforcing
throughanomalous riverrunoffis notpresent in thismodel. Tree ring records(D'Arrigo, 1995;Briffa et al., 1992)provide substantialevidence of multidecadalvariability in the
Manabe,S., and R.J. Stouffer,Two stableequilibriaof a coupled ocean-atmosphere model,J. Climate,1, 841-866,1988. Manabe, S., R.J. Stouffer,M.J. Spelman,and K. Bryan, Transient responseof a coupled ocean-atmosphere model to gradual changesof atmosphericCO2. Part I: annualmean response,J. Climate, 4, 785-818, 1991.
•,
and R.J. Stouffer,Low-frequencyvariabilityof surface air temperature in a 1000-year integration of a coupled atmosphere-ocean-land surfacemodel, J. Climate, 9, 376-393, 1996.
Mann, M.E., J. Park, and R.S. Bradley,Global interdecadaland century-scale climateoscillationsduringthe pastfive centuries, Nature, 378, 266-270, 1995.
Mysak, L.A., D.K. Manak, and R.F. Marsden,Sea-iceanomalies observedin the GreenlandandLabradorSeasduring1901-1984 and their relation to an interdecadalArctic climate cycle, data from the last five centuries(Mann et al., 1995) provide ClimateDynamics,5, 111-133, 1990. evidence for distinct multidecadalvariability in the North Atlantic and Arctic. It is certainlyplausiblethat interactions Walsh, J.E., and W.L. Chapman,Arctic contributionto upperoceanvariabilityin theNorthAtlantic,J. Climate,3, 1462-1473, betweenthe Arctic and the North Atlantic may contributeto
highlatitudesof theNorthAtlantic,with spectralpeaksranging from 30 to 80 years.Analysesof instrumental andproxy
1990.
thisvariability,particularlyby alteringthesurfacefreshwater Walsh,J.E., W.L. Chapman,andT.L. Shy,Recentdecreases of sea exportfrom the Arctic (as in the modelvariabilitydescribed level pressurein the Central Arctic, J. Climate, 9, 480-486, here). A promisingpath to enhancingour understanding of 1995. suchvariabilitylies in a combinationof observational and Wohlleben, T.M.H., and A.J. Weaver, Interdecadal climate proxydataanalysis combined witha widevarietyof modeling variabilityin the subpolarNorth Atlantic, ClimateDynamics, studies.
Acknowledgements.We are very gratefulto JerryMahlman,the Director of GFDL, for his wholeheartedsupport. We thank Drs. Steve Griffies, Tertia Hughes, LawrenceA. Mysak, and Michael Winton for very helpful reviewsof an earlierversionof thismanuscript,aswell astwo anonymous reviewers.
11,459-467, 1995.
T.L. Delworth,S. ManabeandR.J. Stouffer,Geophysical Fluid Dynamics Laboratory/NOAA, P.O. Box 308, Princeton,NJ 08542. (e-mail:[email protected]) (Received September 23, 1996; revised December 6, 1996; acceptedDecember12, 1996.)
Multidecadal climate variability in the Greenland Sea and surroundingregions:a coupledmodel simulation ThomasL. Delworth,SyukuroManabeandRonaldJ. Stouffer Geophysical FluidDynamicsLaboratory/NOAA, Princeton, New Jersey
Abstract. Pronounced oscillationsof oceantemperature temperature (SST) and climate in the Greenland Sea (see and salinityoccurin the GreenlandSea in a 2000 year Fig. 1) andsurroundingregions. integrationof a coupledocean-atmosphere model. The oscillations,involving both the surfaceand subsurface Model Description and Experimental Design oceanlayers, have a timescaleof approximately40-80 years, and are associatedwith fluctuationsin the intenThe coupledocean-atmosphere model usedin this studyis sity of the East GreenlandCurrent.The GreenlandSea identical to that describedin detail in Manabe et al. (1991). temperatureand salinity variations are precededby The model is global in domain,includingthe Arctic, with relarge-scalechangesin near-surface salinityin the Arctic, alistic geographyconsistentwith resolution.The model is which appearto propagateout of the Arctic throughthe forcedwith an annualcycle of insolationat the top of the atEastGreenlandCurrent.Theseanomaliesthenpropagate mosphere.The atmosphericcomponentnumerically intearoundthe subpolargyre into the LabradorSea andthe gratesthe primitiveequationsof motionusinga semi-spectral central North
Atlantic.
These oscillations
are coherent
with previouslyidentifiedmulti-decadalfluctuationsin the intensityof the North Atlantic thermohalinecirculation. The oscillationsin the G•:eenlandSea are related to atmosphericvariability. Negative (cold) anomaliesof surface air temperatureare associatedwith negative
techniquein which the variablesare representedby spherical harmonicsandby corresponding grid pointswith a spacingof 7.5ø longitudeand 4.5ø latitude.There are 9 unevenlyspaced levels in the vertical. The oceaniccomponentof the model uses a finite differencetechniquewith 12 unevenly spaced levelsin the vertical, and a horizontalresolutionof approximately 3.75ø longitudeand 3.75ø latitude.Sea ice is predicted
(cold)seasurface temperature (SST)anomalies in the Greenland Sea,with amplitudes up to 2øCnearGreen- usinga simplemodeldevelopedby Bryan (1969), which allanddecliningto severaltenthsof a degreeC overnorth-
lows sea ice less than 4 m thick to move with the ocean cur-
western Europe.Thec01dSSTanomalies andintensified rents. The
East
Greenland
Current
are
also
associated
with
model atmosphereand ocean interact through
fluxes of heat, water, and momentum at the air-sea interface.
enhancednortherlywindsoverthe GreenlandSea.
The coupledmodelemployedfor thisstudyhasa relatively low spatialresolutionand simplifiedphysicalparameterizations.In spiteof thesesimplifications,the model has successIntroduction fully reproducedmany aspectsof observedinterannualto decadalvariability (Manabe and Stouffer,1996). We feel that The Great Salinity Anomaly (GSA; Dickson et al, 1988) importantinsightscan be gainedfrom suchmodelsregarding was a spectacular exampleof decadalscaleoceanicvariability the fundamentalworkingsof the coupledclimatesystem. in thehighlatitudes of theNorthAtlantic.Suchevents areof In orderto reduceclimatedrift, adjustmentsto the heat and potentiallyprofoundclimatic importancebecauseof the cenwaterfluxesare appliedat the air-seainterfaceoverbothopen tral role of the high latitudesof theNorth Atlanticin the global and sea-icecoveredregions.These flux adjustmentsare dethermohalinecirculation(Walsh and Chapman,1990). rivedfrom preliminaryintegrationsof the separateatmospher-
AagaardandCarmack(1989)suggest thattheGSAoriginat- ic and oceaniccomponents (see Manabeet al., 1991, for
ed throughan enhancement of thefreshwaterandseaice flux-
details).The adjustments vary seasonallyandspatiallybut are constantfrom one ye• to the next. Since the adjustments experiment(H'fikkinen,1993) supportthis. The fresh water were derivedprior to the startof the coupledmodel integraanomalymoveddownthe EastGreenlandCurrent(EGC), fol- tion they are not correlatedwith sea surfacetemperatureand lowedthe subpolargyrethroughtheLabradorSeaandinto the salinity anomalies.They are thereforeunlikely to systematicentralNorth Atlantic,eventuallyreturningto the Greenland cally amplify or dampensuchanomalies.In addition,the ades from the Arctic in the late 1960s. The results of a numerical
Sea in the early 1980s.Evidencethat a similareventmay have justments helpto ke6pthemodelneara realisticmeanstateso that variousfeedbackprocesses,suchas ice-albedofeedback, Dicksonet al. (1988; seetheirFig. 8). canhavea realisticinfluence onthemodelvariability. In this paper we describefluctuationsof surfacesalinity After preliminaryintegrationsto achievean initial condiwhich resemblethe GSA in a 2000 year integrationof a coution in approximateequilibrium,the modelwastime integratpled ocean-atmosphere model.The salinityfluctuationsare ased for a periodof 2000 years.The outputfrom thisintegration sociated with substantial multidecadal variations of sea surface formsthe datasetfor the analysespresentedbelow.
occurred in theearlypartof the20thcentury ispresented by
Thispaperis notsubjectto U.S. copyright. Published in 1997 by theAmericanGeophysical Union.
Descriptionof Model Variability
Papernumber96GL03927.
multidecadal variationsof seasurfacesalinity(SSS)andSST
Analysesof the integrationdescribedabovehave revealed 257
258
DELWORTH ET AL.: MULTIDECADAL
CLIMATE VARIABILITY
IN THE GREENLAND
SEA
2
.., .... ooo -2
0
250
500
750
1000
'
, 1250
1500
1750
2000
Year
Figure1.Timeseries of annual meanmodelSSTin theDenmark Strait(15øWto26øWat70øN).Priortoplotting,thetimeseries wasfilteredsuchthattimescales shorter than10yearswereeffectively removed. Thevaluesareanomaliesfrom a longtermmeanof 0.62. UnitsareøC. in the GreenlandSea region which are reminiscentof the GSA. Someaspects of thiswereseenin ManabeandStouffer (1996), who notedthat substantial,low frequencyvariations of SST andnear-surface air temperature occurredin thevicinity of the GreenlandSea(seetheirFig. 7b). The time series of annual mean model SST from a similar
regionof the currentintegration(shownin Fig. 1) is characterizedby multidecadal variationsof temperature, with peakto-peakdifferences of morethan3øC.The SST variations are tightly coupledto SSS variationsin the sameregion(linear
fy andspreadto covernearlythe entireGreenlandSeaand portions of thecentralNorthAtlanticby Lag0 (Fig.4). In addition, an area of positiveSST anomaliesdevelopsoff the North Americancontinent.This patternappearswhenthe intensityof themodelthermohaline circulation is weakest(Delworthet al., 1993;seeFig. 6a), andis reminiscent of the SST anomaliesseenin observations (Kushnir,1994). In additionto advection,the SST and SSS anomaliesin the GreenlandSea are stronglyinfluencedby convection. Freshwateranomalies in the near-surface layer caninhibitconvection, therebysup-
the heat and saltexchange with sub-surface layers correlation= 0.92), with typical amplitudesof 0.2-0.3 parts pressing per thousand. The spectrumof SST shownin Fig. 2 reveals andamplifyingthe surfaceanomalies. Atmosphericvariationsare also associated with the SST enhancedvarianceat 40-80 years,with a spectralpeakat 5060 years.Visualinspection of Fig. 1 showsthat thereis a fluctuationsshownin Fig. 1. Surfaceair temperatureanomamodulationof the amplitudeof this 40-80 year variability. lies over the Greenlandand NorwegianSeaswith amplitudes with SST in theDenmark Waveletanalysisand singularspectrumanalysiswere per- up to 2øCarepositivelycorrelated formed on this time series(not shown) and reveal that this Strait, with anomaliesof smaller magnitudeextendinginto modulation has a time scale of several centuries. northwestern Europe.Thereare alsoassociated variationsin In orderto describethe space-time structure of thisoscilla- surfacepressure,as shownin Fig. 4. The surfacepressure 2 years tion, linear regressions were computedat variouslagsbe- anomalieshavetheir largestamplitudeapproximately tween the time seriesof severalvariables(oceantemperature, beforethe largestmagnitudeof the SST variations.The spato the anomalous pressure patsalinity,andcurrents) at eachgridpointversus thetimeseries tial patternhasa resemblance of SST shownin Fig. 1. The regression analyses, computed
usingannualmeandata,revealthechanges in thesevariables associated with
SST anomalies in the Denmark
PERIOD (years)
Strait.
200
Figure3 showsthe progression of SSSanomalies from the Arctic to the North Atlantic. Conditionsat Lag -15 indicate fresh anomaliesdominatingthe Arctic. At Lag -10, these
•
:
100
50
20
10
5
3
2
_'._.... .....
10
fresh anomalies still dominate the Arctic, and have extended
southwarddown the EGC and to the regionoff Newfound-
land. The superimposed currentanomaliesshowa stronger thannormalEGC enhancingthe freshwatertransportout of theArctic.At Lag 0 the freshwater anomalies extendaround
the subpolar gyreto thecentralNorthAtlantic.At thistime, positiveSSSanomalies areseento develop in theArctic,a harbinger of thedeveloping opposite phaseof theoscillation. The increasein Arctic SSS is consistentwith the anomalously
1
ol
o01
ol
FREQUENCY (yr-1)
strongtransport of freshwateroutof theArctic.At thesame Figure2. Spectrumof the time seriesof annualmeanmodel time that the surface currentsindicate enhancedflow from the Arctic into the Greenland Sea, there is an anomalousflow at
SST in the Denmark Strait (15øW to 26øW at 70øN; this is the time seriesfrom Fig. 1 prior to filtering).The logarithmof the
estimates (thick,solidline)is plottedversusthelogadepthfrom the Greenland Seainto the Arctic(not shown). spectral Additionalregression analyses (notshown)revealthatposi- rithm of frequency.The thin, solidline denotesthe spectrum tiveseaice anomalies aregenerallyin phasewiththenegative of a backgroundfirst orderMarkov process,andthe dashed SSS and SST anomalies.
lines denote the 95% confidence interval about that back-
The spatialpatternof SST anomalies is shownin Fig. 4. groundspectrum.The spectrumwasestimatedby takingthe Startingfrom small negativeanomaliesat Lag -15 (not Fourier transform of the autocovariancefunction using a shown),negativeSSTanomalies in theDenmarkStraitampli- Tukeywindowwith a maximumlag of 200.
DELWORTH ET AL.: MULTIDECADAL CLIMATE VARIABILITY IN THE GREENLAND SEA
?L- TM ...
(
-1•
-•
ß
-o.•
• .:
-0.2
,
-o.s
',.
o.•
259
, ..... ß....
....
•;,_
/
.,'
..... ,
0.2
Figure4. SameasFig. 3 usingthe regressions of SST at each grid point versusthe time seriesof annualmean SST in the Denmark Strait (15øW to 26øW at 70øN). The regressioncoefficientsof SST are indicatedby the color shading.Superimposedusing red contourlines are the regressions of surface pressure(mb per-2øC SST anomaly).Resultsare shownfor Lag 0 for SST, and Lag -2 for surfacepressure(i.e., surface pressureleadingSST by 2 years). ternprecedingthe GSA (WalshandChapman,1990), although smaller in magnitude.The associatedcirculationindicatesa weakeningof the meansoutherlywindsover the Nordic Seas when SSTs are negative,coincidentwith an enhancedEGC. This association suggests thepossibilityof activeatmospheric participationin this modeloscillation.A similarpossibility wassuggestedby WohllebenandWeaver(1995). Discussion
Previous analyses(Delworth et al., 1993) of this model have revealed multidecadal
variations of the THC in the North
Atlantic with a time scaleof 40-80 years,generallysimilarto thetimescaleof the variabilitydescribedabove.In orderto investigatethe relationshipbetweentheseTHC variationsand the GreenlandSea SSS and SST oscillations,the squaredcoherencywas computedbetweena time seriesof the THC intensity and the SST time series in Fig. 1. The squared coherencyreachesa maximumof 0.50 at a periodof approxi-
Figure3. Regressions of annualmeanSSSat eachgrid point versus the time series of annual mean SST in the Denmark
Strait (15øW to 26øW at 70øN). The contouredvaluesare the
slopesof the regressionlines multipliedby -2 (to indicate conditionsassociatedwith an SST anomalyof-2øC). The fieldat Lag -15 indicatesconditions 15 yearspriorto a-2øC SST anomalyin theDenmarkStrait.Unitsarepartsperthousandper-2øCSSTanomaly.In addition,theregressions of the -0.
-0.!
-0.
-0.02
0
0.02
0.05
0.1
0.2
surface currents areindicated bythevectors. Unitsarecms-1 per-2øCSST anomaly.For clarityof plotting,notall current vectorsareshownat veryhighlatitudes.
260
DELWORTH ET AL.: MULTIDECADAL CLIMATE VARIABILITY IN THE GREENLAND SEA
References mately 57 years,with the THC time serieslaggingthe SST (and SSS) in the Denmark Strait by approximately10 years. Aagaard,K., andE.C. Carmack,The role of seaice andfreshwater The THC is weakestwhen SSS over the sinkingregionof the in the Arctic circulation,J. Geophys.Res., 94, 14,485-14,498, THC (definedas the areafrom 50øN to 70øN in the North At1989. lantic)is at a minimum(seeFig. 9a of Delworthet al., 1993). Briffa, K.R., P.D. Jones, T.S. Bartholin, D. Eckstein, This relationship leadsusto speculate thatenhanced transport F.H. Schweingruber, W. Karlen, P. Zetterberg,and M. Eronen, of relativelyfreshwaterand seaice from the Arctic, through Fennoscandian summersfrom AD 500: temperaturechangeson the East Greenland Current and Denmark Strait, and into the shortandlongtimescales, ClimateDynamics,7, 111-119,1992. LabradorSearegionsmay weakenthe THC in the NorthAtBryan, K., Climate and the oceancirculation.Part III: The ocean model, Mon. Wea. Rev., 97, 806-827,1969. lantic. The increaseof relatively fresh, low densitywater in the near-surface layersof the LabradorSea and neighboring D'Arrigo, R.D., E.R. Cook, and G.C. Jacoby,Annual to decadalscale variations in northwest Atlantic sector temperatures regions("fresh water capping")inhibits the convectiveexinferred from treerings,CanadianJournalof ForestResearch, changeof heatbetweenthe cold near-surface layersandthe 26,143-148,1996. relativelywarmersub-surface layers(Lazier, 1980, provides observationalevidencefor this). This reducedsupplyof heat Delworth, T., S. Manabe, and R.J. Stouffer,Interdecadalvariations of the thermohalinecirculationin a coupledocean-atmosphere to the surfacelayer inhibitsair-seaheatexchangeby cooling model, J. Climate, 6, 1993-2011, 1993. the surfacewaters,therebyreducingthe supplyof negative Dickson, R.R., J. Meinke, S.A. Maimberg, and A.J. Lee, The buoyancyto the near-surface layer andweakeningtheTHC. "Great Salinity Anomaly"in the northernNorth Atlantic 1968The abovediscussionsuggeststhat the GreenlandSea os1982,Progressin Oceanography, 20, 103-151,1988. cillationcouldhavea role in generatingthemultidecadalvari- H•ikkinen,S., An Arctic sourcefor the great salinityanomaly:a ationsof the THC throughfreshwater cappingof its sinking simulationof the Arctic ice-oceansystemfor 1955-1975, J. region.However,there is also the possibilitythat the mulGeophys.Res.,98, 16397-16410. tidecadalvariationsof theTHC may havea role in generating Kushnir, Y., Interdecadal variations in North Atlantic sea surface the GreenlandSea oscillation.It is difficult to distinguishbetemperatureand associated atmosphericconditions,J. Climate, 7, 141-157, 1994. tweenthesetwo possibilities basedon onenumericalexperiment, in which all processesare inherently coupled. Lazier, J.R.N., Oceanographicconditionsat oceanweathership Bravo 1964-1974,Atmosphere-Ocean, 18, 227-238, 1980. Therefore,additionalnumericalexperiments will be required to understand both the interactions between these two oscillations and the factors that determine their time-scales.
One possiblemechanismfor decadalscale(10-20 year) variabilitywasput forwardby Mysak et al. (1990), whosuggestedthatanomalous riverrunoffintotheArcticcreatedthe anomalousfreshwater and seaice pulseknownas the GSA. In our model, however,the time seriesof river runoff into the
Arctichasa whitespectrum (i.e.,approximately uniformvarianceat all time scales).This suggests that an explicitforcing
throughanomalous riverrunoffis notpresent in thismodel. Tree ring records(D'Arrigo, 1995;Briffa et al., 1992)provide substantialevidence of multidecadalvariability in the
Manabe,S., and R.J. Stouffer,Two stableequilibriaof a coupled ocean-atmosphere model,J. Climate,1, 841-866,1988. Manabe, S., R.J. Stouffer,M.J. Spelman,and K. Bryan, Transient responseof a coupled ocean-atmosphere model to gradual changesof atmosphericCO2. Part I: annualmean response,J. Climate, 4, 785-818, 1991.
•,
and R.J. Stouffer,Low-frequencyvariabilityof surface air temperature in a 1000-year integration of a coupled atmosphere-ocean-land surfacemodel, J. Climate, 9, 376-393, 1996.
Mann, M.E., J. Park, and R.S. Bradley,Global interdecadaland century-scale climateoscillationsduringthe pastfive centuries, Nature, 378, 266-270, 1995.
Mysak, L.A., D.K. Manak, and R.F. Marsden,Sea-iceanomalies observedin the GreenlandandLabradorSeasduring1901-1984 and their relation to an interdecadalArctic climate cycle, data from the last five centuries(Mann et al., 1995) provide ClimateDynamics,5, 111-133, 1990. evidence for distinct multidecadalvariability in the North Atlantic and Arctic. It is certainlyplausiblethat interactions Walsh, J.E., and W.L. Chapman,Arctic contributionto upperoceanvariabilityin theNorthAtlantic,J. Climate,3, 1462-1473, betweenthe Arctic and the North Atlantic may contributeto
highlatitudesof theNorthAtlantic,with spectralpeaksranging from 30 to 80 years.Analysesof instrumental andproxy
1990.
thisvariability,particularlyby alteringthesurfacefreshwater Walsh,J.E., W.L. Chapman,andT.L. Shy,Recentdecreases of sea exportfrom the Arctic (as in the modelvariabilitydescribed level pressurein the Central Arctic, J. Climate, 9, 480-486, here). A promisingpath to enhancingour understanding of 1995. suchvariabilitylies in a combinationof observational and Wohlleben, T.M.H., and A.J. Weaver, Interdecadal climate proxydataanalysis combined witha widevarietyof modeling variabilityin the subpolarNorth Atlantic, ClimateDynamics, studies.
Acknowledgements.We are very gratefulto JerryMahlman,the Director of GFDL, for his wholeheartedsupport. We thank Drs. Steve Griffies, Tertia Hughes, LawrenceA. Mysak, and Michael Winton for very helpful reviewsof an earlierversionof thismanuscript,aswell astwo anonymous reviewers.
11,459-467, 1995.
T.L. Delworth,S. ManabeandR.J. Stouffer,Geophysical Fluid Dynamics Laboratory/NOAA, P.O. Box 308, Princeton,NJ 08542. (e-mail:[email protected]) (Received September 23, 1996; revised December 6, 1996; acceptedDecember12, 1996.)
