Low-Frequency Pycnocline Variability in the Northeast Pacific

18 Pages·2005·1.42 MB·English

AUGUST2005 CAPOTONDI ET AL. 1403
Low-Frequency Pycnocline Variability in the Northeast Pacific
ANTONIETTA CAPOTONDI AND MICHAEL A. ALEXANDER
NOAA/CIRESClimateDiagnosticsCenter,Boulder,Colorado
CLARA DESER
NationalCenterforAtmosphericResearch,*Boulder,Colorado
ARTHUR J. MILLER
ScrippsInstitutionofOceanography,LaJolla,California
(Manuscriptreceived14January2004,infinalform23November2004)
ABSTRACT
Theoutputfromanoceangeneralcirculationmodel(OGCM)drivenbyobservedsurfaceforcingisused
inconjunctionwithsimplerdynamicalmodelstoexaminethephysicalmechanismsresponsibleforinter-
annualtointerdecadalpycnoclinevariabilityinthenortheastPacificOceanduring1958–97,aperiodthat
includesthe1976–77climateshift.After1977thepycnoclinedeepenedinabroadbandalongthecoastand
shoaled in the central part of the Gulf of Alaska. The changes in pycnocline depth diagnosed from the
modelareinagreementwiththepycnoclinedepthchangesobservedattwooceanstationsindifferentareas
oftheGulfofAlaska.AsimpleEkmanpumpingmodelwithlineardampingexplainsalargefractionof
pycnoclinevariabilityintheOGCM.ThefitofthesimplemodeltotheOGCMismaximizedinthecentral
partoftheGulfofAlaska,wherethepycnoclinevariabilityproducedbythesimplemodelcanaccountfor
(cid:1)70%–90%ofthepycnoclinedepthvarianceintheOGCM.Evidenceofwestward-propagatingRossby
wavesisfoundintheOGCM,buttheyarenotthedominantsignal.Onthecontrary,large-scalepycnocline
depth anomalies have primarily a standing character, thus explaining the success of the local Ekman
pumpingmodel.TheagreementbetweentheEkmanpumpingmodelandOGCMdeterioratesinalarge
band along the coast, where propagating disturbances within the pycnocline, due to either mean flow
advectionorboundarywaves,appeartoplayanimportantroleinpycnoclinevariability.Coastalpropaga-
tionofpycnoclinedepthanomaliesisespeciallyrelevantinthewesternpartoftheGulfofAlaska,where
localEkmanpumping-inducedchangesareanticorrelatedwiththeOGCMpycnoclinedepthvariations.The
pycnoclinedepthchangesassociatedwiththe1976–77climateregimeshiftdonotseemtobeconsistentwith
Sverdrup dynamics, raising questions about the nature of the adjustment of the Alaska Gyre to low-
frequencywindstressvariability.
1. Introduction be indicative of changes in upwelling, a process that
influencestheexchangeofpropertiesbetweenthedeep
Theoceanicpycnocline,asubsurfacelayercharacter- and upper ocean. Large areas of the northeast Pacific
ized by large vertical density gradients, can be viewed Oceanarecharacterizedbyafreshandwell-mixedsur-
astheinterfacebetweenthesurfaceoceanmixedlayer face layer, separated from the deeper ocean by large
andthedeepocean.Changesinpycnoclinedepthmay salinitygradients,orhalocline.Athighlatitudes,where
temperaturesarelow,salinityhasadominantinfluence
on density, particularly in winter, so that pycnocline
depth is very similar to winter mixed layer depth
*TheNationalCenterforAtmosphericResearchissponsored
bytheNationalScienceFoundation. (Freeland et al. 1997). Thus, understanding the pro-
cesses governing pycnocline variability can also help
understand the changes in mixed layer depth, a quan-
Correspondingauthoraddress:AntoniettaCapotondi,NOAA/
tity that has a large influence upon biological produc-
CIRES Climate Diagnostics Center, R/CDC1, 325 Broadway,
tivity (Polovina et al. 1995; Gargett 1997).
Boulder,CO80305-3328.
E-mail:[email protected] What processes control pycnocline depth variability
©2005AmericanMeteorologicalSociety
JPO2757
1404 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME35
inthenortheastPacific?Usingsubsurfaceobservations steady-state response of the large-scale ocean circula-
over the period 1968–90, Lagerloef (1995) examined tiontowindforcingistheSverdruprelation(Sverdrup
interannual and decadal variations of the dynamic 1947), which expresses the linear vorticity balance be-
height field, which is closely related to pycnocline to- tween wind stress curl and meridional advection of
pography,overtheGulfofAlaska.Thedominantmode planetaryvorticity.AlthoughtheSverdruprelationwas
of variability of the dynamic height field, identified originallyderivedasatheoryforthesteady-stateocean
throughempiricalorthogonalfunction(EOF)analysis, circulation, it has proven very useful to explain circu-
captured the 1976–77 climate shift and was well corre- lation changes associated with large-scale variations of
latedwithotherclimateindices.Asimplemodelforthe the wind forcing. Deser et al. (1999) have related the
evolution of the dynamic height field, in which the intensificationoftheKuroshioExtensionafterthemid-
forcingwassuppliedbythelocalEkmanpumping,and 1970stothestrengtheningofthewesterliesinthecen-
dissipation was modeled with a linear damping term, tral Pacific using Sverdrup dynamics. In the Gulf of
appeared to reproduce a large fraction of the low-fre- Alaska,theAlaskanStreamcanbeviewedasthewest-
quency dynamic topography variations. However, the ernboundarycurrentoftheAlaskagyre,anditiscon-
hindcast of dynamic heights underestimated the ob- ceivable that some of the changes in the upper-ocean
served variability in the western half of the gyre, a re- density structure and circulation associated with the
sultthatLagerloef(1995)attributedtothepresenceof 1976–77climateshiftcanbeexplainedintermsofSver-
westward propagating baroclinic Rossby waves. drup dynamics. Qiu (2002) has interpreted the annual
Cummins and Lagerloef (2002, CL02 hereinafter) modulation of the Alaska gyre intensity during 1993–
used the same Ekman pumping model described by 2000 in terms of time-dependent Sverdrup balance.
Lagerloef(1995)toexaminethestructureofpycnocline However, the dynamical balances governing the varia-
depth variability forced by the dominant patterns of tionsoftheAlaskagyreatdecadaltimescalesarestill
anomalous Ekman pumping over the northeast Pacific unclear.
(30°–60°N,180°–120°W)during1948–2000.Theevolu- In this study we examine the physical mechanisms
tion of pycnocline depth predicted by the simple Ek- responsibleforlow-frequency(interannualtointerdec-
man pumping model reproduced well the observed adal)pycnoclinevariabilityinthenortheastPacificdur-
variations at Ocean Weather Station Papa (OWS P: ing 1958–97. Both interannual and decadal time scales
50°N,145°W).However,CL02wereunabletotestthe will be examined, but our major focus will be on the
performance of the simple model at other locations in decadal–interdecadal variations and, in particular,
the Gulf of Alaska and to verify the accuracy of the those associated with the 1976–77 climate shift. The
spatial patterns of pycnocline depth changes produced specific questions we ask are: What fraction of pycno-
by the model over the entire northeast Pacific. Cum- cline variability can be explained by local Ekman
mins and Lagerloef (2004, CL04 hereinafter) further pumpingindifferentareasoftheGulfofAlaska?Ifthe
investigated the performance of the local Ekman simple model of CL02 and CL04 fails in some areas,
pumping model of CL02, by estimating its ability to what other processes control pycnocline changes in
reproduce observed SSH variability in the northeast those areas? What is the role of baroclinic Rossby
Pacific during 1993–2003, and concluded that interan- wavesinpycnoclinevariability?DoesSverdrupbalance
nual variability in the Gulf of Alaska is dominated by holdatdecadaltimescales?Toanswerthesequestions
the local response to wind forcing. we will use the output from an ocean general circula-
ThesuccessofthesimpleEkmanpumpingmodelin tion model (OGCM) driven by observed surface forc-
explainingpycnoclinevariabilityissomewhatsurprising ing.Aftertestingthemodel’sperformanceatlocations
giventheroleofRossbywavesintheadjustmentofthe where long-term observations are available, the
large-scaleoceancirculationtochangesinsurfaceforc- OGCM output is used to test the different dynamical
ing. Indeed, satellite altimetry shows westward- hypotheses.OurstudywillextendtheworkofCL02by
propagating sea surface height (SSH) anomalies in the testingthelocalEkmanpumpingmodelovertheentire
northeast Pacific (Kelly et al. 1993; Fu and Qiu 2002). northeast Pacific. It will also extend and complement
Further, Qiu (2002) has shown that SSH variability the work of CL04 by examining the changes that took
from the Ocean Topography Experiment (TOPEX)/ place across the mid-1970s.
Poseidon (T/P) altimeter mission (1993–2001), areally TheOGCMusedforthisstudyisdescribedinsection
averaged over the offshore region of the Gulf of 2, and in section 3 the performance of the OGCM is
Alaska, can be largely explained by first-mode baro- examined by comparing the pycnocline depth changes
clinic Rossby wave propagation. in the OGCM with those observed at two oceano-
The simplest dynamical paradigm to describe the graphicstationsindifferentareasoftheGulfofAlaska.
AUGUST2005 CAPOTONDI ET AL. 1405
The influence of local Ekman pumping upon pycno-
clinevariabilityisexaminedinsection4,andinsection
5theroleofbaroclinicRossbywavesisconsidered.The
validity of the Sverdrup balance at decadal time scales
is examined in section 6, and the processes governing
the evolution of pycnocline depth anomalies along the
coastisanalyzedinsection7.Weconcludeinsection8.
2. The OGCM
TheOGCMusedforthisstudyistheNationalCen-
ter for Atmospheric Research ocean model (NCOM)
FIG. 1. Pycnocline depth changes associated with the 1976–77
thathasbeendescribedindetailbyLargeetal.(1997),
climateregimeshiftfromtheoutputoftheNCAROGCM.The
Gentetal.(1998),andLargeetal.(2001).Thespecific depth of the 26.4 (cid:2)(cid:3)isopycnal is used as proxy for pycnocline
numerical simulation analyzed here is described in depth.ThepositionsofOWSP(P:50°N,145°W)andGAK1(G:
Doney et al. (2003). In this section we only provide a 59°N,149°W)andthelocationofKodiakIsland(K),Seward(S),
QueenCharlotteIsland(Q),theAlaskaPeninsula,andtheAleu-
brief summary of the basic model characteristics and
tianIslandsareindicatedforreference.
information about the surface forcing used for this
simulation. from the mean state characteristic of the 40-yr experi-
NCOM is derived from the Geophysical Fluid Dy- ment.Thenthemodelwasrunfortwo40-yrcycles,the
namics Laboratory (GFDL) Modular Ocean Model second cycle starting from the conditions achieved at
with the addition of a mesoscale eddy flux parameter- the end of the first 40-yr segment. The mismatch be-
izationalongisopycnalsurfaces(GentandMcWilliams tweenthemodelstateandtheforcingatthebeginning
1990) and a nonlocal planetary boundary layer param- of the second cycle did not seem to produce any long-
eterization (Large et al. 1994). The model is global, termtransientbehavior,butsomeresidualdriftintem-
with a horizontal resolution of 2.4° in longitude and perature and salinity can be detected at depths larger
varyingresolutioninlatituderangingfrom0.6°nearthe thanapproximately500m(Doneyetal.2003).Herewe
equator to 1.2° at high latitudes. The model version analyze the output for the second 40-yr period using
usedforthisstudyincludesananisotropicviscositypa- monthly mean values.
rameterization (Large et al. 2001) with enhanced vis-
cosityclosetooceanboundariesandmuchweakervis- 3. How realistic is pycnocline variability in the
cosity in the ocean interior. OGCM?
The surface forcing includes momentum, heat, and Lagerloef(1995)hasshownthatasignificantfraction
freshwater fluxes for the period 1958–97. The wind (28%)ofthesubsurfacevarianceovertheperiod1960–
stressiscomputedfromthereanalysesfieldsproduced 90 is associated with the 1976–77 climate regime shift.
at the National Centers for Environmental Prediction Thus,westartbyexaminingthepycnoclinedepthvaria-
(NCEP) (Kalnay et al. 1996) using bulk formulas. The tions exhibited by the OGCM in association with the
sensible and latent heat fluxes are computed from the 1976–77 climate shift. The depth of the 26.4 (cid:2) isopyc-
(cid:3)
NCEP winds and relative humidity and the model’s nal, which lies in the core of the main pycnocline, is
SSTs using standard air–sea transfer equations (Large usedasproxyforpycnoclinedepth.Themeandepthof
and Pond 1982; Large et al. 1997). Sensible and latent the26.4(cid:2) isopycnalrangesfrom(cid:1)100minthecenter
(cid:3)
heatfluxesdependonthedifferencebetweenSSTand oftheAlaskagyreto150–180maroundtherimofthe
surfaceairtemperature.SinceSSTandairtemperature gyre.Thechangesinpycnoclinedeptharecomputedas
closely track each other, when observed air tempera- thedifferencebetweenthepycnoclinedepthinthepe-
tures are used in the bulk formulas, as in the present riod1977–97(period2)andthepycnoclinedepthinthe
model simulation, the model’s SST is relaxed toward period1960–75(period1).Afterthemid-1970s,thepy-
observations(Haney1971).Therelaxationtimescaleis cnoclinewasshallowerinthecentralpartoftheGulfof
relatively short (30–60 days for typical mixed layer Alaskaanddeeperinabroadbandfollowingthecoast
depths),sotheSSTinthemodelcanbeexpectedtobe (Fig. 1). The deepening was more pronounced in the
strongly constrained by the surface forcing rather than westernpartoftheGulfofAlaska,tothesouthwestof
by the interior ocean dynamics. Kodiak Island (K in Fig. 1), following approximately
The numerical simulation is started from an initial the Alaska Peninsula. A similar pattern of pycnocline
condition obtained from a preliminary climatological changes was found by Miller et al. (1994) in a coarser-
integration,sotheinitialmodelstateisnottoodifferent resolution ((cid:1)4°) ocean model hindcast, using surface
1406 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME35
FIG.2.Zonalsectionofmeanpotentialdensityalong53.8°Nfor
the periods 1958–75 (solid contours) and 1977–97 (dot–dashed
contours).Contourintervalis0.4(cid:2)(cid:3)units.The26.4(cid:2)(cid:3)isopycnal,
used as proxy for pycnocline depth, is highlighted with a thick
contourforbothperiods.
forcing fields from the COADS dataset instead of
NCEP.Theleadingmodeofasimplestochasticmodel
FIG. 3. Pycnocline depth variations (m) at (top) OWS P and
forced by Ekman pumping also displayed the spatial
(bottom)StationGAK1.Positiveanomaliesindicatedeeperpyc-
structure and temporal evolution characteristics of the nocline, while negative anomalies correspond to a shallower
1976–77 regime shift (Fig. 4 in CL02). Changes in the pycnocline.Thedotsconnectedbythedashedlinearethewinter
density structure are detectable to depths greater than averages of observed values at the two stations (usually a few
(cid:1)800 m, but appear to be more pronounced in the values for each winter), while the solid line shows the winter
pycnocline depth anomalies from the NCAR OGCM, based on
upperoceanasshownbyazonaldensitysectionalong
monthlyaverages.Correlationcoefficientsbetweentheobserved
53.8°N (Fig. 2). Above 400 m, the isopycnals deepen andmodeledwintervaluesare0.7and0.77atPapaandGAK1,
towardtheeasternandwesternedgesofthegyreafter respectively. The observed time series at Papa was kindly pro-
themid-1970s,resultinginamorepronounceddoming videdbyP.Cummins.
of the density structure.
Is the pattern of pycnocline variations in Fig. 1 real- byaweathership.Freelandetal.(1997)havecomputed
istic?Subsurfaceobservationsaregenerallysparseand pycnoclinedepthsbyassumingatwo-layerrepresenta-
salinitydatainparticularareverylimited.Sincesalinity tion of the density structure and using a least squares
is an important factor controlling the potential density approachtodeterminethethicknessoftheupperlayer
distribution in the subarctic Pacific, the paucity of sa- from the observed density profiles. The two-layer rep-
linity data limits the possibility of testing the perfor- resentation is appropriate for wintertime (December–
manceoftheOGCMoverthewholenortheastPacific. April) conditions, and it tracks the core of the main
However,thereareafewoceanographicstationsinthe pycnocline in winter. CL02 have used the additional
Gulf of Alaska where measurements of both tempera- observations after 1994 to extend the time series of
tureandsalinityareavailableoveralongperiodoftime meanwinterpycnoclinedepthfrom1994to1999.Thus,
including the 1976–77 climate shift. We will use mea- the whole time series covers the period 1957–99, with
surements from two stations, Ocean Weather Station gaps in 1985, 1988, 1990, and 1994. The resulting time
Papa(OWSP,PinFig.1),whichislocatedinthearea seriesiscomparedinFig.3awiththewinterpycnocline
where the pycnocline becomes shallower, and station depth from the OGCM. The correlation between the
GAK1 (G in Fig. 1), which is in the band where the twotimeseriesis0.7,andtheamplitudeoftheOGCM
pycnocline deepens. pycnocline depth anomalies is also approximately cor-
OWS P (50°N, 145°W) provides a long record of rect(thestandarddeviationoftheOGCMtimeseriesis
bottlecastandCTDdataovertheperiod1957–94.The 9.7mandtheobservedtimeseriesstandarddeviationis
time series has been recently augmented by more re- 11.3 m). Notice that the observed pycnocline depth
cent observations from 1995 to 1999. Generally, the anomalyvaluesareoftenaveragesoffewdataoverthe
density of observations is higher in the earlier period, winter months, while the winter pycnocline depths es-
1957–81,whenOWSPwasoccupiedonaregularbasis timated from the OGCM are based on monthly aver-
AUGUST2005 CAPOTONDI ET AL. 1407
ages of potential density, which is available at each
modeltimestep(oftheorderofoneday).Thus,higher-
frequency processes (e.g., mesoscale eddies and baro-
clinictides),whicharefilteredoutintheOGCM,may
contributetosomeoftheobservedpycnoclinevariabil-
ity. Both model and observations indicate a shoaling
trendofthemainpycnocline,asdiscussedbyFreeland
et al. (1997) and CL02.
StationGAK1(59°N,149°W)islocatedatthemouth
of Resurrection Bay near Seward, Alaska, where the
OGCM indicates a deepening of the pycnocline after
FIG. 4. Winter Ekman pumping changes associated with the
the mid-1970s (Fig. 1). Profiles of temperature and sa- 1976–77 climate regime shift. Contour interval is 0.1 (cid:7) 10(cid:10)4
linitytoadepthof250mhavebeenmeasuredstarting cms(cid:10)1.Solidcontoursindicatepositivevalues,whiledashedcon-
inDecember1970,andarestillongoing.Thenumberof toursareusedfornegativevalues.Anomalieslargerthan0.3(cid:7)
observations over the winter period varies from less 10(cid:10)4 cms(cid:10)1 are dark shaded, while values smaller than (cid:10)0.3 (cid:7)
10(cid:10)4cms(cid:10)1arelightshaded.ThemeanEkmanpumpingisposi-
than5measurementsin1972,1980,1981,and1985toa
tiveovertheGulfofAlaska.NegativeEkmanpumpinganomalies
maximum of 15 values in 1999. Starting from the indi-
implyreducedupwellingandviceversa.
vidualdensityprofiles,wehaveestimatedthedepthof
the25.4(cid:2) isopycnal,whichislocatedclosetothebase
(cid:3)
of the observed winter pycnocline. Winter values have The changes in the Ekman pumping field associated
then been computed by averaging all the values be- withthe1976–77climateshiftareshowninFig.4asthe
tweenDecemberandApril.Standarddeviationsabout difference between Ekman pumping during winter
eachwintervalue(notshown)canbeaslargeas50m. (December–April) in 1977–97 (period 2) and 1960–75
StationGAK1issituatedveryclosetoshore,andthere (period 1). The resulting pattern shows negative W
E
isnomodelgridpointavailableattheexactlocationof changes(anomalousdownwelling)inabroadbandthat
the station. So we have estimated winter pycnocline followsthecoasttotheeastofKodiakIsland(indicated
depthsfromtheOGCMattheclosestmodelgridpoint with a K in Fig. 1), and positive changes (anomalous
(58.6°N, 147.6°W). The comparison between the ob- upwelling)elsewhereintheGulfofAlaska,withmaxi-
servedandsimulateddepthanomaliesisshowninFig. mumvaluesextendingtothesoutheastoftheAlaskan
3b.Althoughtheobservedtimeseriesexhibitsinteran- Peninsula. The largest negative difference is achieved
nual fluctuations of larger amplitude than the time se- in the very northern part of the Gulf of Alaska, be-
ries from the OGCM, which are based on monthly av- tween approximately Seward (S in Fig. 1) and Queen
eragesofvaluesateachmodeltimestep,thetwotime CharlotteIsland(QinFig.1).Sincethetime-averaged
series agree quite well (the correlation coefficient is Ekman pumping is positive (e.g., upwelling) over the
0.77)onbothinterannualandlongertimescales(apart whole Gulf of Alaska, a positive (negative) difference
from a discrepancy in 1979–81). In particular, the low- implies that upwelling is enhanced (diminished) after
frequency behavior is very similar in both time series the1976–77climateshift.ThepatternofW changesin
E
and indicates a deepening of the pycnocline after the Fig. 4 is very similar to that computed by Lagerloef
mid-1970s. Thus, the comparison with observations in- (1995) using an empirical orthogonal function ap-
dicates that the pattern of pycnocline depth changes proach.
producedbytheOGCMisbelievable.Nextweusethe The pattern of W bears some similarity to the dis-
E
OGCMtoassesswhichfractionofthepycnoclinevari- tributionofpycnoclinechangesinFig.1:areasofnega-
ability can be explained by local Ekman pumping. tive (positive) W differences generally correspond to
E
areasofdeeper(shallower)pycnoclineafter1977.The
4. Where is Ekman pumping important? majordiscrepancybetweenthetwofieldsoccursinthe
coastalbandextendingsouthwestwardfromKodiakIs-
The vertical velocity at the base of the Ekman layer
landtotheAleutianIslands,whereapositivechangein
due to the divergence of the Ekman currents (Ekman
W from period 1 to period 2 is associated with a
pumping) is defined as the vertical component of the E
curlofthewindstress(cid:1)dividedbytheCoriolisparam- deeper pycnocline. The local correlation between Ek-
eter f and the mean density of seawater (cid:4): man pumping and pycnocline depth from the OGCM
(cid:1) (cid:2) (cid:3)(cid:4) (cid:5) (h )isshowninFig.5ausinglow-passfilteredEk-
(cid:1) OGCM
W (cid:6) (cid:2)(cid:7) . (cid:8)1(cid:9) manpumpingdatatoremoveperiodsshorterthan2yr
E (cid:1)0f z and finding the lag at which the correlation is maxi-
1408 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME35
TofurtherexaminetheinfluenceofthelocalEkman
pumping upon pycnocline variability and understand
thephaserelationshipsbetweenthetwofields,wecon-
sider the simple model used by Lagerloef (1995) and
CL02:
dh
(cid:6)(cid:10)W (cid:10)(cid:2)h, (cid:8)2(cid:9)
dt E
where h is the pycnocline depth, W is the Ekman
E
pumping,and(cid:11)isalineardampingcoefficientwiththe
unitsoftheinverseoftime.Equation(2)isforcedwith
monthlyanomaliesofEkmanpumpingcomputedfrom
the NCEP–NCAR reanalyses wind stress, the same
forcingwhichdrivestheOGCM.Equation(2)issolved
using a second-order accurate trapezoidal scheme as
hn(cid:12)1(cid:6)(cid:3)hn(cid:12)(cid:3)Wn(cid:12)1(cid:4)2, (cid:8)3(cid:9)
1 2 E
where (cid:13) (cid:6) (2 (cid:10) (cid:11)(cid:14)t)/(2 (cid:12) (cid:11)(cid:14)t) and (cid:13) (cid:6) (2(cid:14)t)/(2 (cid:12)
1 2
(cid:11)(cid:14)t). We have used the OGCM pycnocline depth
anomalies at the initial time (January 1958) as initial
conditions for the simple model.
A question remains concerning the value of the
FIG.5.(a)Maximumlagcorrelationsbetweenlow-passfiltered dampingtimescale(cid:11)(cid:10)1.CL02foundthatthetimeevo-
Ekman pumping (periods shorter than 2 yr are removed) and lution of the pycnocline depth produced by the simple
pycnocline depth anomalies. Pycnocline depth anomalies have model (h) is largely controlled by the damping time
beenmultipliedby(cid:10)1toachievepositivecorrelations(apositive scale (cid:11)(cid:10)1. Longer damping time scales tend to empha-
depthanomalyindicatesdeeperpycnocline,andshouldbeasso-
size the oceanic response at lower frequencies, and to
ciatedwithnegativeW anomalies,andviceversa).Valueslarger
E
than 0.5 are light shaded, while values larger than 0.7 are dark increase the lag between Ekman pumping and pycno-
shaded.(b)Lag(months)betweenW andh atwhichthe cline depth variations. As shown by (3), (cid:13) can be in-
E OGCM 1
maximumcorrelationsin(a)areachieved.Positivelagsindicate terpreted as the lag-1 autocorrelation of pycnocline
thatW leadsthechangesinpycnoclinedepth.
E depth.Usingthelag-1autocorrelationofyearlypycno-
clinedepthanomaliesatOWSP,CL02foundavalueof
mized.Thecorrelationishighestinabandthatextends 1.4 yr for (cid:11)(cid:10)1, while the lag-1 autocorrelation of the
westwardfromthecenteroftheAlaskangyre((cid:1)50°N, Pacific decadal oscillation (PDO) index [the leading
145°W) to the Aleutian Islands, with values of (cid:1)0.7 principal component of monthly SST anomalies over
(Fig.5a).ThecorrelationsdecreasetowardtheAlaska theNorthPacific,ascomputedbyMantuaetal.(1997)]
Peninsulawheretheydroptovaluesveryclosetozero. yielded (cid:11)(cid:10)1 (cid:6) 1.6 yr
Along the coastal band to the east of Kodiak Island, CL02usedasinglereferencevalueof(cid:11)(cid:10)1(cid:6)1.5yrfor
correlations tend to be higher (0.3–0.5) than on the their entire North Pacific domain. However, if the
western side, with maximum values of (cid:1)0.5 between dampingterm(cid:10)(cid:11)hin(2)isaparameterizationofoce-
Seward and Queen Charlotte Island. Figure 5b shows anicprocesses,thedampingtimescalemaybeexpected
the lags between Ekman pumping and pycnocline to vary spatially depending on the strength of those
depth that yield the maximum correlations. Ekman processes. Using the OGCM as a term of comparison,
pumpingvariationsalwaysleadpycnoclinedepthvaria- we have determined the value of (cid:11)(cid:10)1 that maximizes
tions with lags ranging from 6–8 months over a large theinstantaneouscorrelationwiththeOGCMtimese-
areainthecentralpartoftheGulfofAlaskatovalues ries at each grid point. The resulting distribution (Fig.
aslargeas10–15monthsinthenorthernportionofthe 6a) shows indeed large spatial variability of the damp-
GulfofAlaskaandclosetothesoutheasterncornerof ingtimescale,rangingfrom14–16monthsinthecentral
the domain. Lags are much smaller from the southern part of the Gulf of Alaska to values as large as 90
part of the Alaska Peninsula to the Aleutian Islands months along the coastal band to the east of Kodiak
where Ekman pumping and pycnocline depth are only Island. In the area along the Alaska Peninsula from
weakly correlated. KodiakIslandtotheAleutianIslands(cid:11)(cid:10)1iscloseto12
AUGUST2005 CAPOTONDI ET AL. 1409
ally larger in areas where (cid:11)(cid:10)1 is larger. In the central
part of the Gulf of Alaska (cid:11)(cid:10)1 (cid:1)15 months. Consider-
ing a period of 3 yr for the forcing (recall that the
Ekmanpumpingusedforthecorrelationsonlycontains
periods longer than 2 yr) the time lag is (cid:1)7 months,
very close to the lag between W and OGCM pycno-
E
cline depth in the same area (Fig. 5b).
The pattern of correlations between h and h
OGCM
obtained with the values of (cid:11)(cid:10)1 in Fig. 6a is shown in
Fig. 6b. The overall structure is similar to the correla-
tion pattern between W and h , but the correla-
E OGCM
tions between h and h are higher than the corre-
OGCM
lations between W and h because of a larger
E OGCM
level of high-frequency variability in the W field with
E
respect to the h field. The correlations in Fig. 6b indi-
cate that the simple model gives an excellent fit to the
OGCM in the central part of the Gulf of Alaska, and
west of (cid:1)145°W, with maximum correlations (0.9) at
approximately50°–52°N,155°–150°W.Correlationsde-
creaseonshore,especiallytowardtheAlaskaPeninsula
andtothesouthof(cid:1)45°N.TotheeastofKodiakIsland
valuesrangefrom0.5southofQueenCharlotteIsland
to0.7betweenQueenCharlotteIslandandKodiakIs-
FIG.6.(a)Spatialdistributionofthedampingtimescales(cid:11)(cid:10)1 land, while southwestward of Kodiak Island correla-
(months), which maximizes the correlation between pycnocline tionsdroptonearzero.Thus,theareatothesouthwest
depthanomaliesproducedbytheEkmanpumpingmodel(2)and ofKodiakIslandistheregionwherethesimplemodel
thepycnoclinedepthanomaliesintheOGCM.(b)Spatialdistri-
completely fails to reproduce the OGCM pycnocline
butionofthecorrelationbetweenhandh obtainedusingthe
values of (cid:11)in (a). Light shading is forOvGaClMues larger than 0.7, depth changes. This is the region of the Alaskan
intermediateshadingisforvalueslargerthan0.8,anddarkshad- Stream,astrongwesternboundarycurrent.Advection
ingisforvalueslargerthan0.9. processes can be expected to be important in this area
andberesponsiblefortheinabilityofthesimplemodel
to reproduce the OGCM pycnocline variability. Satel-
months,butcorrelationsareverylowinthisarea(Fig. lite observations have also shown the presence of pro-
6b). The values we find for the damping time scale in nounced anticyclonic meanders that develop along the
the central part of the Gulf of Alaska (14–18 months) streamasaresultofthestreaminstabilities(Crawford
areinagreementwithLagerloef(1995)andCL02’ses- etal.2000).Thesemeanders,whicharenotcapturedby
timates. theOGCMbecauseofitscoarsehorizontalresolution,
The pattern of (cid:11)(cid:10)1 in Fig. 6a is generally consistent will certainly produce large, low-frequency variability
withthespatialdistributionofthelagsbetweenW and in the local pycnocline that cannot be reproduced by
E
pycnocline depth in Fig. 5a. However, it may be seen the simple Ekman pumping model.
from (2) that the phase lag between h and W is fre- To further examine the simple model fit to the
E
quency dependent and given by (cid:15) (cid:6) atan ((cid:16)(cid:11)(cid:10)1), OGCM, we compute the difference between pycno-
where (cid:16) is the angular frequency. Thus, (cid:15) depends clinevariabilityintheOGCMandpycnoclinevariabil-
upontheratiobetweenthefrictionaltimescale(cid:11)(cid:10)1and ity produced by Eq. (2) using the spatially varying (cid:11)(cid:10)1
the period of the forcing 2(cid:17)(cid:16)(cid:10)1, varying from values thatmaximizesthecorrelation.Thestandarddeviation
verycloseto0°whentheperiodoftheforcingismuch ofthisresidualfield,whichisameasureofthemisfitof
longer than the frictional time scale to (cid:17)/2 for forcing thesimplemodeltotheOGCM,maximizesinthewest-
periods much shorter than the frictional time scale. ernpartoftheGulfofAlaska(Fig.7b)wherethevari-
Thus, at low frequencies Ekman pumping and pycno- anceofh isalsoverylarge(Fig.7a).Inparticular,
OGCM
clinedepthvariationstendtobeinphase,whileathigh to the southwest of Kodiak Island the residual field
frequencies they are in quadrature. Comparison be- accounts for 80%–90% of the variance of h (Fig.
OGCM
tween Figs. 5b and 6a shows that the lag between Ek- 7c).Theminimumoftheresidualstandarddeviationis
man pumping and pycnocline depth response is gener- found in the central part of the Gulf of Alaska (cen-
1410 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME35
FIG.7.(a)StandarddeviationofpycnoclinedepthintheOGCM.Contourintervalis5m.
Valueslargerthan10mareshaded.(b)Standarddeviationoftheresidualfieldcomputedas
the difference between pycnocline depth anomalies in the OGCM and pycnocline depth
anomaliesfromtheEkmanpumpingmodel(2).Contourintervalis5m.Valueslargerthan10
mareshaded.(c)PercentageoftheOGCMpycnoclinedepthvarianceaccountedforbythe
residualfield.Valueslargerthan60%arelightshaded,andvalueslargerthan70%aredark
shaded.
tered around 52°N, 150°W), where (2) produces the waves (Kelly et al. 1993; Qiu 2002; Fu and Qiu 2002).
best fit to the OGCM. In this area the variance of the WhatfractionofpycnoclinevariabilityintheOGCMis
residual field accounts for 10%–30% of the OGCM duetoRossbywavepropagation?Toanswerthisques-
pycnocline depth variance (Fig. 7c). tion we compare the performance of the local Ekman
pumpingmodel(2)withthatofaRossbywavemodel.
Adirectcomparisoniscarriedoutalongtwolatitudes,
5. Rossby wave dynamics
40.5° and 49°N. As seen in Fig. 6b, the local Ekman
Equation(2)emphasizestheimportanceoflocalEk- pumping model gives a good fit to the OGCM along
man pumping in driving thermocline variability. Satel- 49°N with correlations of (cid:1)0.8 west of 140°W, but its
litealtimetry,ontheotherhand,hasprovidedevidence performance is worse along 40.5°N where the maxi-
ofwestwardpropagatingSSHanomalies,whichappear mum correlations are less than 0.7.
to be associated with first-mode baroclinic Rossby Following CL04, we consider a linear, reduced-
AUGUST2005 CAPOTONDI ET AL. 1411
gravity, quasigeostrophic model to simulate upper- 1. Westward propagation is evident at times: for ex-
ocean variability that includes Rossby wave dynamics. ample the positive anomaly centered at (cid:1)155°W in
The wavelength of Rossby waves at interannual time 1965 and the negative anomaly centered around the
scales is typically longer than (cid:1)2000 km (Chelton and samelongitudein1987.However,inmostcasespycno-
Schlax 1996; Fu and Chelton 2001), while the Rossby clinedepthanomaliesexhibitastandingcharacter.The
radiusofdeformationis(cid:1)30kmaround40°N(Chelton evolutionproducedbytheEkmanpumpingmodel(Fig.
et al. 1998) and decreases poleward. Thus, the long- 8b)doesnotcapturethewestwardpropagationseenin
wave approximation is valid poleward of 40°N. In the theOGCMbutreproducesrelativelywellotheraspects
long-wave limit the equation describing the evolution ofthevariability.TheRossbywavemodel,ontheother
of pycnocline depth is hand, gives a good representation of the westward-
propagating anomalies but does not capture other fea-
(cid:5)h (cid:5)h tures,forexampletheanomalieswestof170°Wduring
R(cid:12)c R(cid:6)(cid:10)W (cid:10)(cid:2)h , (cid:8)4(cid:9)
(cid:5)t R (cid:5)x E 1 R 1958–66 and 1975–87 (Fig. 8c). It is worth noting that
the period examined by Fu and Qiu in the altimeter
where h is the pycnocline depth anomaly in the pres- data (1993–2000) contains a prominent westward-
R
ence of Rossby wave propagation, c (cid:6) (cid:10)(cid:18)R2 is the propagating signal. To quantify the goodness of the fit
R
phasespeedoflongbaroclinicRossbywaves,whereR oftheEkmanpumpingandRossbywavemodelstothe
istheRossbyradiusofdeformationand(cid:18)isthelatitu- OGCMweshowinFig.10atheinstantaneouscorrela-
dinal gradient of the Coriolis parameter, and (cid:11) is a tionsofeachmodelwiththeOGCMalong49°N.Over-
1
Rayleighfrictioncoefficient.FollowingKillworthetal. all,theEkmanpumpingmodelproducesabetterfitto
(1997), who took into account the influence of a back- the OGCM than the Rossby wave model, indicating
ground mean flow, we have chosen c (cid:6) 0.8 cms(cid:10)1 thatoverthe1958–97periodRossbywavepropagation
R
along 49°N and 1.5 cms(cid:10)1 along 40.5°N. To optimize is not the controlling dynamics along 49°N.
thecomparisonwiththeOGCM(cid:11) hasbeenchosento Along40.5°N,theevolutionofthepycnoclinedepth
1
be (4 yr)(cid:10)1. Equation (4) can be solved by integrating anomaliesintheOGCMdoesnotshowanyclearindi-
along Rossby wave characteristics in the x–t plane: cationofthe1976–77climateshift.Westwardpropaga-
tion,asindicatedbytheslopeofthephaselines,canbe
h (cid:8)x,t(cid:9)(cid:6)h (cid:8)x ,t(cid:10)t (cid:9)e(cid:10)(cid:2)1tE noticedforsomeeventseastof(cid:1)165°W.However,the
R R(cid:5)E E
large-scale anomalies appear to be primarily of stand-
x W (cid:8)(cid:6),t(cid:10)t (cid:9)
(cid:12) E (cid:6) e(cid:10)(cid:2)1t(cid:6)d(cid:6), (cid:8)5(cid:9) ing nature, an aspect captured relatively well by the
x cR(cid:8)(cid:6)(cid:9) Ekmanpumpingmodel.Alongthislatitudetheperfor-
E
mance of the two simple models is comparable, as in-
wherethesolutionateachpointxandtimetisobtained dicated by the instantaneous correlation with the
asthesuperpositionofthedisturbancesgeneratedeast OGCM (Fig. 10b).
ofpointxatprevioustimes.Thefirsttermontherhsof
(5) is the contribution of the signals originating at the
6. Sverdrup dynamics
easternboundaryx ,andreachingpointxattimetwith
E
a transit time t (cid:6) (cid:19)x [d(cid:20)/c ((cid:20))]. The second term on AclassicalmodelofthemeancirculationoftheGulf
E xE R
the rhs of (5) is the contribution of the disturbances of Alaska would include a Sverdrup interior with the
generated by the Ekman pumping east of the target Alaskan Stream as the western boundary current. Are
pointxwitht (cid:6)(cid:19)x[ds/c (s)].Bothboundaryandwind- thechangesintheAlaskanStreamthattookplaceafter
(cid:20) (cid:20) R
forcedtermsdecaywhilepropagatingcounterclockwise the mid-1970s consistent with Sverdrup dynamics? If
with the frictional time scale 1/(cid:11). theAlaskanStreamdoesrespondtotheadjustmentof
1
Figures 8 and 9 show the comparisons between the the Alaska gyre according to Sverdrup dynamics, this
evolution of pycnocline depth anomalies from the couldpartiallyexplainwhyitsvariabilityisonlyweakly
OGCM and those obtained from the Ekman pumping affected by the local wind stress curl.
model (2) and Rossby wave model (4) along 49° and Sverdrup balance (Sverdrup 1947) is derived from
40.5°N,respectively.Thepycnoclinedepthvariationsin the vertical integral of the vorticity equation
theOGCMalong49°N(Fig.8a)exhibitaclearchange (cid:5)w
in character after 1976 with positive anomalies west of (cid:7)(cid:8)(cid:6)f , (cid:8)6(cid:9)
(cid:5)z
(cid:1)145°W before the shift and negative afterward. Pyc-
nocline depth anomalies east of 145°W tend to be of where (cid:21) is the meridional velocity, w is the vertical
oppositesigntothosefartherwest,consistentwithFig. velocity, and z is the vertical coordinate. Equation (6)
1412 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME35
FIG.8.Comparisonbetweentheevolutionofpycnoclinedepthalong49°Nin(a)theOGCM,(b)theonecomputedfromtheEkman
pumpingmodel,and(c)theoneobtainedfromtheRossbywavemodel.ThesolutionfortheEkmanpumpingmodelwasobtainedby
usingtheoptimalvaluesof(cid:11)inFig.6aalong49°N.Dashedcontoursandblueshadingarefornegativepycnoclinedepthanomalies
(shallowerpycnocline),whilesolidcontoursandorangeshadingareforpositiveanomalies(deeperpycnocline).Contourintervalis10
minallthreepanels.
states that the potential vorticity changes associated becomputedbyintegratingwestwardfromtheeastern
withtheverticalstretchingofthewatercolumnmustbe boundary:
balanced by the changes in planetary vorticity due to (cid:5)
f x
meridionalmotions.Integrating(6)fromalevelofno- (cid:9)(cid:8)x(cid:9)(cid:6)(cid:10) W dx(cid:10). (cid:8)9(cid:9)
motion to the base of the Ekman layer, we obtain the g (cid:7) E
x
E
classical form of the Sverdrup balance:
The time averaged barotropic streamfunction in the
(cid:7)V (cid:6)fW , (cid:8)7(cid:9) model describes the cyclonic circulation of the Alaska
g E
gyre(Fig.11a).ThemeantransportofthemodelAlas-
where V is the vertically integrated geostrophic me-
g kan Stream is (cid:1)7–8 Sv (Sv (cid:23) 106 m3s(cid:10)1), a value
ridional velocity. The condition of nondivergence for
smaller than some observational estimates. Geo-
the vertically integrated geostrophic flow allows to in-
strophic transports relative to 1500 dbar are typically
troduce a streamfunction (cid:22), such that
g 12–18Sv(Musgraveetal.1992).However,intermittent
(cid:5)(cid:9) surveysoftheAlaskanStreamyieldedabroadrangeof
Vg(cid:6) (cid:5)xg. (cid:8)8(cid:9) transport estimates with values as low as 2 Sv and as
large as 20 Sv (Reed et al. 1980; Royer 1981). The
Using the boundary condition that (cid:22)(x ) (cid:6) 0 at the barotropic streamfunction difference between 1977–97
g E
easternboundaryx oftheoceanbasin,thegeostrophic versus 1960–75 (Fig. 11b) shows a slight weakening of
E
streamfunctionatlongitudexalongagivenlatitudecan the circulation in the eastern part of the basin and a
Fig 8live4/C