The Role of the Indonesian Throughflow in Equatorial Pacific Thermocline Ventilation

Home , Boundary current, Indonesian Throughflow

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 104, NO. C9, PAGES 20,551-20,570, SEPTEMBER 15, 1999

The role of the Indonesian Throughflow in equatorial Pacific thermocline ventilation

Keith B. Rodgers,• Mark A. Cane,and NaomiH. Naik Lamont-Doherty Earth Observatory, Columbia University, Palisades, New York

Daniel P. Schrag Department of Earth and Planetary Sciences,Harvard University, Cambridge, Massachusetts

Abstract. The role of the IndonesianThroughflow (ITF) in the thermocline circulation of the low-latitude Pacific Ocean is explored using a high-resolution primitive equation ocean circulation model. Seasonally forced runs for a domain with an open Indonesian passageare compared with seasonally forced runs for a closedPacific domain. Three casesare considered: one with no throughflow, one with 10 Sv of imposed ITF transport, and one with 20 Sv of ITF transport. Two idealized tracers, one that tags northern component subtropical water and another that tags southern component subtropical water, are used to diagnosethe mixing ratio of northern and southern component waters in the equatorial thermocline. It is found that the mixing ratio of north/south componentwaters in the equatorial thermocline is highly sensitiveto whether the model accountsfor an ITF. Without an ITF, the source of equatorial undercurrent water is primarily of North Pacific origin, with the ratio of northern to southern component water being approximately 2.75 to 1. The ratio of northern to southern component water in the Equatorial Undercurrent with 10 Sv of ITF is approximately 1.4 to 1, and the ratio with 20 Sv of imposed ITF is I to 1.25. Estimates from data suggesta mean mixing ratio of northern to southern component water of less than I to 1. Assuming that the mixing ratio changesapproximately linearly as the ITF transport varies between 10 and 20 Sv, an approximate balance between northern and southern component water is reached when the ITF transport is approximately 16 Sv. It is also shown that for the isopycnal surfaces within the core of the equatorial undercurrent, a 2øC temperature front exists acrossthe equator in the western equatorial Pacific, beneath the warm pool. The implications of the model results and the temperature data for the heat budget of the equatorial Pacific are considered.

1. Introduction 1997]. Our investigationdiffers from previousstudies in A series of model runs have been conducted to look that we focus specifically on how the ITF affects circulation fields and thermal budgets within the pycnocline at the role of the IndonesianThroughflow (ITF) in the of the equatorial Pacific Ocean. thermocline circulation of the equatorial Pacific Ocean. Our goal is to develop a better quantitative under- The ITF refers to the mean export of waters from standing of the sources of the water that upwells in the western equatorial Pacific to the Indian Ocean via the equatorial Pacific. We seek to identify not only the Indonesian Straits separating the Australian and the range of density surfaces that contributes to the Asian continents. The role of the ITF in the global equatorial upwelling but also the relative contributions thermohaline circulation has been discussedby Gordon of Northern and Southern Hemispheric component wa- [1986],and its impacton globalcirculation and thermal ter that upwells along the equator. In ocean circula- budgets has since been addressedin several modeling tion models that treat the Pacific Ocean as a closed studies[Hirst and Godfrey,1993; Shriver and Hurlbutt, basin(thereby ignoring the transportof the ITF), equatorial upwelling is primarily supplied by water from 1Now at Max-Planck-Institutfiir Meteorologie,Hamburg, the Northern Hemisphere subtropics. This is incon- Germany sistent with salinity measurements in the equatorial thermocline[Tsuchiya et al., 1989],which suggest that Copyright 1999 by the American GeophysicalUnion. the Southern Hemispheric componentof the equatorial Paper number 1998JC900094. undercurrent is larger than the Northern Hemispheric 0148-0227/ 99/ 1998JC 900094509.00 component. One of our primary objectives is to quan-

20,551 20,552 RODGERS ET AL.' INDONESIAN THROUGHFLOW IN EQUATORIAL PACIFIC

tify the model sensitivity of the mixing ratio of north- acting on the mean background temperature gradients ern to southern component thermocline water in the is not important. It also assumeslow diffusivity and equatorial upwellingto the opening and closingof the velocity shear within the thermocline, so that the ex- Indonesian Straits. This has implications for climate tratropical anomaliesare relatively intact upon reaching variability becauseisopycnal surfaces from northern and the equator. This amounts to assumingthat the STC southern sourceshave different temperatures. is a closed system, where the changesin temperature This exerciseis motivated by two broader scientific on isopycnal surfacesare described by questions: What dynamical processesdetermine the OT • mean thermal structure of the pycnoclinein the equato- Ot rial Pacific? What dynamical processescontrol the time variability of the thermal structure of the pycnocline? However,the low prebombA•4C valuesrecorded by Modeling studieshave shownthat changesin the ther- Galapagos(1øS, 90øW) co,'als [Druffe, 1981,1987] sug- mal structure of the equatorial thermoclinecan change gest that the STC is not a closed system, in that sig- the decadal variability of the E1 Nifio-Southern Oscil- nificant amounts of water from below the equatorial lation (ENSO). In fact, of all the parametersensitivity pycnocline are entrained into the equatorial upwelling studiesthat have been conductedwith the coupled [Toggweileret al., 1991]. The term u', whichincludes biakand Cane[1991] model, changing the temperature changesin upwelling velocities, is known to be signif- structure of the thermoclinehas the strongestinfluence icant in the equatorial Pacific. In addition, the equa- on model behavior. torial VT, which includes the time mean temperature To first order, in a zonally averagedsense, a subtrop- gradient along the isopycnal surface correspondingto ical cell (STC)[McCreary and Lu, 1994],exists in the the core of the equatorial undercurrent, is significantly upper thermocline of the Pacific Ocean, whereby much larger than the VT • term emphasizedby Deser et al. of the water that upwells along the equator has sub- [19961ß ducted in the subtropicalgyres of the South and North In this paper we investigate the sensitivity of the cir- Pacific. The upwelled water returns to the subduct- culation within the equatorial thermocline to the trans- ing regionsof the subtropicsthrough the surfaceEk- port of the ITF. First, we present an overview of esti- man flow. Temporal variability in the thermal structure mates of throughflow transport. This is followed by a of the equatorialpycnocline (below the euphoticzone) presentationof our modeling results. We finish with a on isopycnal surfaces is described by the advection- discussionof thermal budgets within the equatorial Pa- diffusionequation for temperature anomalies cific thermocline, where we refer to temperature and OT • salinity measurements within the equatorial thermo- + •.VT' + u'.V'7- K + K' (1) cline.

whereK representsisopycnal mixing and K • represents 2. Observations diapycnalmixing. The terms • and T representthe tinhemean three-dimensional velocity and temperature 2.1. Thermocline Circulation for the Western Pacific fields,respectively, and u • and T • representperturba- tions to the time mean fields. We have assumed that The mean flow configurationwithin the thermocline temperature anomaliescan be treated as passivetrac- of the low-latitude western Pacific is shown schemati- ers,and we haveignored the perturbationproduct term. Callyin Figure 1. The focusis on flow within the pycno- The secondterm on the left-handside of (1) represents cline, which correspondsto the potential densityrange the advectionof temperatureanomalies by the mean as-25.0 to 25.5 and which lies at a mean depth of be- flow, and the third term on the left-handside represents tween 150 and 200 m in this region. This is the density changesassociated with anomalouscurrents acting on range feeding the core of the Equatorial Undercurrent the tinhemean temperature gradients along isopycnals (EUC) in the westernPacific. in the equatorial pycnocline. The SouthEquatorial Current (SEC) and the North Recently a mechanismwas proposedby Deser et EquatorialCurrent (NEC) representthe westwardflow- al. [1996],Gu and Philander[1997], and Zhanget ing, equatorward branchesof the southern and northern al. [1998],whereby temperature anomalies of order subtropicalgyres, respectively.The Mindanao Current 0.5øC originatingat the sea surfacein the subtropics and the New GuineaCoastal Undercurrent (NGCUC) can advect subsurfaceinto the equatorial thermocline are both low-latitude westernboundary currents,which via the STC and thus changethe thermal structureof transport high-salinity subtropical thermocline water the equatorialpycnocline. It was suggestedthat it is into the equatorial region. Much of the high-salinity the advectivetimescale of intergyreexchange between southern component water has reached 143øE via the the subtropicalsubducting regions and the equatorial Vitiaz Strait, stretching between Papua New Guinea upwellingthat determinesthe period for decadal-scale and New Britain at approximately6øS, 148øE (see Fig- climate variations. The proposedmechanism assumes ure 1). Murray et al. [1995]measured a transport that advectionassociated with perturbation velocities through the Vitiaz Strait of approximately 15 Sv in RODGERS ET AL.: INDONESIAN THROUGHFLOW IN EQUATORIAL PACIFIC 20,553

North EquatorialCurrent (NEC)

Mindanao Current

EquatorialUndercurrent (EUC)

New Guinea Indonesian CoastalUnderct•ixent _ (NGCUC) ;- (ITF) .,

110øE 120øE 130•E 140øE 150øE 160øE 170øE Longitude

Figure 1. Schematicrepresentation of thermoclineflow in the westernequatorial Pacific Ocean.

1992. The magnitudeof this transport is quite large, CirculationStudy (WEPOCS) II surveyalong 143øE to especiallywhen consideringthat the channelthrough arguethat 1/2 to 2/3 of the water feedingthe equato- which it flowsis only 50 km wide and i km deep. rial undercurrentin the westernPacific is suppliedfrom the South Pacific. The WEPOCS II section at 143øE is 2.2. Hydrography of the Pacific Thermocline near the confluence of the Mindanao Current and the There is a meridionalsalinity gradient along isopy- NGCUC, and thus salinity measurementscollected on cnals across the equatorial thermocline of the Pacific the equator at this longitude can be used to infer the Ocean that results fi'om differences in freshwater fluxes relative mixing ratio of northern subtropical component at the sea surface between the North Pacific and the water to southern subtropical component water within South Pacific. As a consequence,isopycnals in the Pa- the Equatorial Undercurrent. Since, as discussed be- cific thermocline are warmer and saltier in the South low, interannual variability in temperature and salinity Pacific than they are in the North Pacific. This can along isopycnal surfacesin this region is substantial, be seenin Figure 2, which showsthe salinity distribu- one should exercisecaution in interpreting mixing ra- tion on the rr0=25.0 isopycnalsurface, where Levitus et tios from one survey. al. [1994]annual mean salinity data have been linearly interpolatedonto the rr0=25.0surface calculated using 2.3. Indonesian Throughflow and Its Connection to Pacific Circulation both the annualmean temperature [Levitus and Boyer, 1994]and salinityclimatologies. This densitysurface The thermocline circulation in the western equato- was chosenbecause it correspondsto the densityof the rial Pacific is complicated by the presence of the In- EUC core. The freshest waters can be seen within the donesianThroughflow. Ffield and Gordon[1992] have CaliforniaCurrent in the northeasternpart of the plot, demonstrated using hydrographic measurements that advectingsouthward within the thermocline. The salti- the source of the ITF is North Pacific thermocline wa- est watersare foundin the subtropicalgyre of the South ter. This North Pacific water mass is supplied to the Pacific. ITF by the Mindanao Current. Given that the mean Wyrtki [1981]used thermal budgets to arguethat ITF transport (0(10 Sv))is an order of magnitude the mean upwellingflux in the equatorial Pacific is 50 larger than flow throughthe Bering Strait (0(1 Sv)), Sv. Bothtritium (3H) to helium3 (3He)ratios [Jenk- mass conservation requires that the ITF transport be ins, 1996 ] andchlorofiuorocarbons [ Warner et al., compensated by cross-equatorial flow Dom the South- 1996]reveal a 10-20year ventilationage for the core ern Hemisphere. Thus although the NGCUC supplies of the equatorialundercurrent. Tsuchiya et al. [1989] perhaps 15 Sv to the equatorial pycnocline from the usedmeasurements of the meridionalsalinity gradient Southern Hemisphere, it does not itself close the flow collectedduring the WesternEquatorial Pacific Ocean around Australia by directly feedingthe ITF. Rather, it 20,554 RODGERS ET AL.: INDONESIAN THROUGHFLOW IN EQUATORIAL PACIFIC

120øE 150øE 180 ø 150øW 120øW g0øw x

Figure 2. The salinityon the a0=25.0isopycnal surface, calculated from the Levituset al. [1994] and Levitusand Boyer [1994]annual mean climatologies.

retrofiectsnear 140øE in the Halmahera eddy and feeds drographicmeasurements, which cluster around 12 Sv. the EUC. Upon upwellingin the easternequatorial Pa- Part of this discrepancyis undoubtedly due to the fact cific, some of this water is advected within the surface that the zonal componentof the Hellerman and Rosen- Ekman flow into the Northern Henrisphere.Through stein[1983] wind stressclimatology is knownto be too its exposureto surfacefluxes, it undergoesa water mass strong in the tropics, which would lead to an overesti- transformation,before eventually subducting as north- mate of the ITF transport. ern component thermocline water. Many attempts have been made to estimate the trans- 2.4. Previous Modeling Studies port of the ITF using hydrographicdata. An overview of recentestimates has been presented by Godfrey[1996], Liu et al. [1994],Lu and McCreary[1995], Gu and and they rangefrom-2.6 Sv [Fieuxet al., 1996]to 18.6 Philander[1997], and Rothsteinet al. [1998]have stud- Sv [Fieuxet al., 1994. However,there are manycom- ied Pacific equatorial thermocline ventilation due to in- plicationsinvolved with this method. The strongtidal tergyre exchangeusing ocean models with varying de- mixing within the straits makesgeostrophic calculations greesof vertical and horizontal resolution. However, all of doubtful validity. of thesestudies neglect the effect of the ITF by using a Another approachto estimatingthe transport of the closedPacific domain. There are also two global mod- throughflow,Godfrey's [1989] Island Rule calculatesthe eling studiesthat are relevant to our discussion,as they mean transport of the ITF from the annual mean wind address the influence of the ITF on thermocline circula- stressr over the surfaceocean. It makesa few approx- tion within the PacificOcean. Hirst and Godfrey[1993] imationsregarding the dynamicalequations governing used a relatively coarseresolution global version of the the mean flow in the oceaninterior, includingthe as- Modular Ocean Model (Ax-2.8 ø longitude, Ay-l.6 ø sumption that the throughflow in the Indonesian Seas latitude,with six layersin the upper500 m) in orderto is semigeostrophic,thus bypassingmany of the prob- compare a model run that allowed for ITF with a model lemsinherent in estimatingthe throughflow.Using the run that did not allow for ITF. For their run with an annual mean wind stressclimatology of Hellerman and open ITF, a sill depth of 800 m was chosenfor the In- Rosenstein[1983] for his analysis,Godfrey obtained an donesianStraits, and the passagewas given a minimum annual mean transport estimate of 164-4 Sv. This es- width of four model grid points. The annual mean ITF timate for the ITF transport is somewhatlarger than transport was determined by the model to be 17.1 Sv the estimatesof annualmean transportmade from by- with the run for which a sill depth of 800 m was used. RODGERSET AL.: INDONESIAN THROUGHFLOW IN EQUATORIAL PACIFIC 20,555

The Hirst and Godfrey[1993] model runs were forced 3. Model Description with annual mean wind stresses, and the formulation of Hah [1984]was used to calculatesurface heat fluxes. The model runs discussedherein are designedto test They found that for the casewhere the ITF was open, the sensitivityof the mixing ratio of northern and south- the sea surfacetemperatures (SSTs) in the central and ern component subtropical thermocline water in the eastern equatorial Pacific were in excessof 0.4øC colder equatorial undercurrent to the opening and closing of than they were for the case where there is no ITF. This the ITF. For the casewhere the ITF is open, the geom- is a consequenceof the oceanicheat transport within etry of the throughflow region is highly simplified, and their model when the ITF is open; the water exported the barotropic transport around Australia is imposed to the Indian Oceanthrough the Indonesianpassages is rather than part of the solution. Thus configured,this warmer than the compensatinginflow from the extrat- processstudy is meant to convey the importance of in- ropical South Pacific, with the result that the thermo- cludingthe ITF in oceanmodels that focuson intergyre cline is cooler than for the case with no ITF. exchangefor the Pacific Ocean.

The correspondingsurface heat flux in the central and 3.1. Circulation Model easternPacific was also 15 to 20% larger for the ITF opencase (note that $eageret al. [1995]have shown The primitive equation Lamont Ocean-AML (Atmo- that the Han [1984]heat flux formulationis likely to sphericMixed Layer) Model (LOAM) is usedfor this underestimatefluxes in this region). In addition, the study. This is a level version of the Gent and Cane temperature in the warm pool was slightly cooler for [1989]model, which includes a barotropicsolver [Naik the caseof an openITF, sothat the overalltemperature et al., 1995]with realisticbathymetry. A Lorenzfour gradientacross the equatorialPacific is iøC higherfor cycle is used for time differencing,and model calcula- the case with an open ITF. tions are performed on an A grid. A fundamentallimitation of the Hirst and Godfrey For surface momentum forcing the seasonal wind [1993]study, given our interestin the Pacificpycno- stressclimatology of daSilva[1994] is used. The heat cline, is one of resolution. With meridional resolution flux componentof surfacebuoyancy forcing is calculated of 1.6ø, the model does not capture the critical scales usingthe AML model of Seagetet al. [1995],which of the equatorial undercurrent, and thus the ventilation now includes a parameterization for wintertime storm pathwaysfor intergyre exchangeand equatorial thermo- tracksin the extratropics(described by Hazelegeet al., cline ventilation are representedtoo crudely. manuscriptin preparation, 1999). Sea surfacesalinities are relaxedto the Levituset al. [1994]monthly clima- $hriver and Hurlburt [1997]used a versionof the tology with a restoring time constant of 1 month. NavalResearch Laboratory (NRL) modelto investigate Two different horizontal domains were used for this surface and thermocline flow in the low latitude western Pacific basin. The horizontal resolution used for study and are shownin Figure 3. The first (Figure 3a) does not allow for any flow around Australia and thus their study is 0.5 ø, which is sufficient to resolve the critical scales of both the EUC and low-latitude west- doesnot permit any ITF transport. The seconddomain (Figure 3b) doesaccount for an ITF. A channelrepre- ern boundarycurrents (LLWBCs), and their study il- lustrates nicely the deviation of the flow fields for the senting the Indonesian Straits connectsthe equatorial Pacific and Indian Oceans. The channel has simplified westernPacific thermoclineand surfaceocean gener- geometry and has a constant depth of 800 m, which is ated usinga nonlinearprimitive equationmodel as op- posed to linear Sverdrup dynamics. approximately the same depth as that usedin the study of Hirst and Godfrey[1993]. Given our interestin the thermal budgetsassociated Although our representationof the throughflow re- with intergyreand interbasinexchange, the most sig- gion is simplified, it is justified a postJori, in that the nificantshortcoming of the $hriverand Hurlburt [1993] vertical structure of the mean transport, with maximum study is that their model doesnot include thermody- transport near 100 m depth and most of the transport namics.Their focuswas on circulationpathways in the occurring in the upper 400 m, is in accord with observa- upper Pacific Ocean connectingthe ACC to the ITF, tions. Within the sliver of the Indian Ocean included in and thus they ignoredthermal budgetsassociated with the modeldomain, the temperature[Levitus and Boyer, heat transportswithin the thermoclineand buoyancy 1994]and salinity [Levitus et al., 1994]fields are relaxed forcing at the sea surface. Also, the model used for to seasonalvalues on a timescale of 30 days. This is nec- the Shriver and Hurlburr study has only six layersin essary to maintain the baroclinic structure of the flow the vertical, and much of the thermocline flow is con- in the channel. fined to the singlelayer just below the surfacelayer. As the focusof this study is intergyre exchange,hor- We shall seethat for modelswith high resolution,in- izontal grid stretching has been used to achieve high tergyreexchange between the subtropicsand equatorial resolutionin dynamically sensitiveregions. The merid- upwellingregions favors the westernboundary route for ional and zonal resolutionis showngraphically in Fig- deeperisopycnals and the interiorventilation pathway ure 4. The meridional resolution (Figure 4a) is ap- for shallow isopycnals. proximately1/3 ø alongthe equatorand falls off to ap- 20,556 RODGERS ET AL.' INDONESIAN THROUGHFLOW IN EQUATORIAL PACIFIC

120 øE 150 øE 180 ø 150øW 120øW 90øW 60øW x

(a)

•44000•35/J

120øE 150øE 180' 150øW 120øW 90øW 60øW x

(b)

Figure 3. Model domain and bathymetry (a) for casewith no IndonesianThroughflow (ITF) and (b) for caseswhere ITF transportis imposedthrough barotropic solver.

proximately 2ø in the extratropics. The zonal resolu- of the GriJfies[1998] implementation of the Gent and tion (Figure 4b) is about 1ø in the longitudebands of Mc Williams[1990] scheme. As a stretchedgrid is used the Mindanao Current, the NGCUC, and the Peru up- for the model runs discussedhere, the lateral mixing co- welling, with resolution falling off to slightly less than efficientshave been chosento vary linearly with horizon- 2ø for most of the ocean interior. There are 30 layers in tal resolution, so that the larger coefficientsneeded in the vertical, 15 of which are in the upper 300 m. High the extratropics where the resolution is slightly greater vertical resolution within the thermocline is necessary than 2ø do not compromisethe solution in regionswhere for the intergyre exchangeproblem, since water parcels resolution is less than 1ø . The coefficient correspond- that subduct in the subtropics and upwell along the ing to both the isopycnalmixing and the eddy-induced equator can attain a depth of several hundred meters transportvelocity is approximately200 m2 s-x where en route. resolutionis 1/3ø and approximately 1600 m 2 s-x where Lateral mixing of temperature and salinity is per- resolution is larger than 2ø . The model sensitivity to formed using a modified version of the GriJfies et al. this coefficient will be addressedin a further study. [1998]implementation of the Redi[1982] isopycnal mix- Shapiro filtering is used on the momentum equations, ing tensor, and adiabatic eddy-induced transport of asdescribed by Gentand Cane[1989]. In addition,mo- these tracers is accomplishedusing a modified version mentum is diffused laterally using a coefficientof 1500 RODGERS ET AL.. INDONESIAN THROUGHFLOW IN EQUATORIAL PACIFIC 20,557

(a)

:::) 2.5

z 1.5

N • 1- • 0.5

0 -60 4'O 6O LATITUDE (b) 3 , i i i i i i 2.5

1.5

0.5

I I I I I I 120 140 160 180 200 220 240 260 280 300 LONGITUDE

Figure 4. Model (a) meridionaland (b) zonalresolution.

m2s-•. Thisviscosity decays in the verticalfrom its full 3.2. Initialization and Forcing of Idealized value at the sea surfacewith an e-foldingdepth of 200 Tracers m. This value was chosen to tune the structure of the Two idealized tracers are advected on line for the model's equatorial undercurrent,as with only Shapiro model runs, which impose 0, 10, and 20 Sv of ITF filtering, the undercurrentis too strong and outcrops transport, and are used to illustrate mixing ratio of at the surface east of the dateline in the annual mean. northern to southern component water in the equatorial The physical justification for this parameterization is thermocline. One of the tracers represents a Northern that high-frequencymeridional winds in the equatorial Hemispherictracer, and the other representsa Southern Pacific in the real world lead to a meridionalmixing Hemispherictracer. Initially, the tracer concentration is of momentum in the upper ocean acrossthe equator set equal to zero everywhere, and the tracer concentra- and that this effect will not be captured for oceanmod- tion is clamped to a value of 1 poleward of 18ø in the re- els forcedwith monthly mean wind climatologies.This spectivehemispheres as the model integrates. This for- value of momentum diffusionis rather small, and its ef- mulationis similarto that of Lu andMcCreary [1995], fect is only felt in the EUC, wherethe resolutionis high except that we treat the northern and southern ventila- and shearsare high. For vertical mixing, in addition to tion tracers separately. The advection schemeused for convectiveadjustment, the mixing parameterization of passivetracers is identical to that used for temperature Pacanowskiand Philander[1981] is used,with a uni- and salinity in the circulation model, and the samemix- form backgrounddiffusivity of 0.1 cm2 s-•. ing parameterizationsthat operate on temperature and Three model runs were made using the the geome- salinity (namely lateral diffusion and the eddy-induced try shownin Figure 3 with seasonalforcing' (1) No transport) are alsoused for the passivetracers. ITF, with seasonalwinds; (2) 10 Sv of imposedITF, with seasonalwinds; and (3) 20 Sv of imposedITF, 4. Results with seasonalwinds Case (1) usesthe domainshown 4.1. Model Flow Fields in Figure 3a, and cases(2) and (3) use the geometry shownin Figure 3b. In cases(2) and (3) the time- For both the 10 and 20 Sv cases the transport in invariantbarotropic ITF transportis imposedthrough the channel separating the Pacific and Indian Oceans the model's barotropic solver. is maximum at 90 m depth. This is consistent with For each run discussedhere, the model is initialized the observed vertical structure of flow in the Indone- with temperature and salinity fields from Levitus and sian Straits, where flow is observed to be maximum Boyer[1994] and Levitus et al. [1994],respectively. The at approximately100 m depth (A.L. Gordon,personal model is spun out for 10 years in a robust diagnostic communication,1998). This is significant,because the mode with a time relaxation constant of 1 year. Af- model is siphoning off water at the correct density hori- ter this spin-upperiod, tracer initialization and forcing zon for the Pacific Ocean, despite its simplified repre- occur, and the model is integrated for 40 more years. sentation of the geometry of the Indonesian Straits. 20,558 RODGERS ET AL' INDONESIAN THROUGHFLOW IN EQUATORIAL PACIFIC

The flow at a depth of 90 m is shown in Figure 5a north of the equator. The flow acrossthis sectionfor the for the casewith no ITF and in Figure 5b for the case casewith 10 Sv of ITF (Figure6b) revealsthe vertical with 10 Sv of imposedITF transport. The 10 Sv of ITF structure of the retrofiection of NGCUC water shown in siphonsoff a significantfraction of the southwardflow- Figure 5. The strongwestward flow within the thermo- ing Mindanao Current water toward the Indian Ocean. cline along the northern coastof New Guinea is nearly An important difference between the two caseslies in as large as that of the eastwardflowing EUC and is the flow along the northern coast of New Guinea. The consistentwith the findingsof the WEPOCS program. strong retrofiectionjust north of New Guinea for the The meridional componentof flow across 5.7øS is case with the open throughflow is consistentwith the shownin Figure 6c for the case with no ITF and in flow fields measured during the WEPOCS II survey Figure 6d for the casewith 10 Sv of ITF. As the ITF at 143øE[Tsuchiya et al., 1989]. At 195 m depthin transport increases,so does the influx of water from the thermocline,for the casewith no ITF (Figure 5c), the South Pacific across7øS, as required by continuity. nearly all of the equatorwardflowing Mindanao Current Significantly,the compensationoccurs primarily within feedsthe EUC, whilewith 10 Sv of ITF (Figure5d) the the NGCUC. Mindanao Current splits at approximately 4øN and its 4.2. Model Results for Hemispheric Tracers contribution to the EUC is significantlydiminished. Next we consider the zonal flow across two transects The Northern and Southern Hemispheric tracer dis- in the thermocline of the western Pacific. The first tributions for the casewith no ITF are shownprojected of these is along 142øE between 3øS and 7øN, which ontothe •rs=25.0isopycnal surface in Figures7a and 7b. correspondsto the WEPOCS II section described by These plots representsnapshots 10 years after tracer Tsuchiyaet al. [1989]. The flow acrossthis section initialization. The Northern Hemisphere tracer can be for the casewith no ITF (Figure 6a) showsmaximum seento accessthe equator primarily through advection zonalflow within the EUC of 55 cm s-•, centeredjust within the Mindanao Current. Along the equator the

Figure 5. Vector plots of flow fields at 90 and 195 m depth in the westernPacific, showingflow at (a) 90 m depthfor casewith no ITF, (b) 90 m depthfor casewith 10 Sv of imposedITF, (c) 195m depthfor casewith no ITF, and (d) 195 m depthfor casewith 10 Sv of imposedITF. RODGERSET AL.. INDONESIAN THROUGHFLOWIN EQUATORIAL PACIFIC 20,559

',, ",,.../// o i::i::i::ii:•iiiiii:: ' • \ / /

3'S 2'S 1'S 0' I'N 2'N 3øN 4'N 5'N 6'N 7'N 3'S 2øS 1'S 0' I'N 2'N 3'N 4'N 5'N 6'N 7'N Y Y (a) (b)

144'E 146'E 148'E 150'E 152'E 154'E 156'1 146'E 148'E 150'E 152øE 154'E 156'E x x

Figure 6. Annual meanzonal velocityas a functionof latitude and depth at 142øEfor the case (a) with no ITF and (b) with 10 Sv of ITF. The contouringinterval for Figures6a and6b is 0.1 m s-•. Annualmean meridional velocity across 5.7øS for the case (c) withno ITF and(d) with 10 Sv of ITF. The contouringinterval for Figures6c and 6d is 0.05 m s-1.

concentrationof the NorthernHemisphere tracer is sig- rent hasnow beendiverted into the throughflowand can nificantlylarger than that of the SouthernHemisphere be seenexiting the Pacific in the throughflowregion. tracer. The Southern Hemisphere tracer in the 10 Sv case The unventilatedregion along 10øN in the eastern (Figure7d) showsconcentrations in the equatorialun- Pacificin Figure 7a is the shadowzone of the North Pa- dercurrentthat are significantlylarger than they were cific subtropicalgyre and is an oxygen-minimumzone for the case with no ITF. The maximum in tracer con- in the real ocean. The minimumzone along4øN seen centration along the equator can be traced back to the stretchingfrom the easternedge of the Mindanao Cur- NGCUC at the western boundary, which is the domi- rent acrossthe basin has been attributed by Lu and nant equatorial ventilation pathway from the Southern McCreary[1995] to the potentialvorticity barrier posed Hemisphereon this isopycnalhorizon. The casewith 20 by the Inter TropicalConvergence Zone (ITCZ). Sv of imposedITF is shownin Figure 7e and 7L The For the SouthernHemisphere tracer shownin Fig- overall pattern for both hemispheresremains the same, ure 7b, both the interior and westernboundary venti- although the tracer concentrationalong the equator is lation routescontribute to the equatorial undercurrent. once again diminished for the northern tracer and en- The interior ventilation route penetratesclosest to the hanced for the southern tracer. equator at 140øW. In Figure 8 we considermeridional sectionsof tracer The tracer distributions for the case with 10 Sv of im- concentration at 150øW for the three model runs. For posedITF after 10 yearsare shownin Figures7c and the casewith no ITF the Northern Hemispheretracer 7d. With the throughflowregion open, Northern Hemi- (Figure 8a) is dominant at all depths within 2ø of spheretracer concentrations are lower along the equator the equator. The maximum concentration of North- than they werefor the casewith no throughflow.Much ern Hemisphere tracer within the undercurrent is three of the tracer that flows south within the Mindanao Cur- times larger than the maximum concentrationof South- 20,560 RODGERSET AL' INDONESIAN THROUGHFLOW IN EQUATORIAL PACIFIC

•:ii•'::•ii!"'::•'!J.i.:•i•.....,• co5 ::iii!:•iiiiiiiii:•111i

120'E 150øE 180' 150'W 120'W 90'W 60'W 20'E 150'E 180' 150'W 120'W 90'W 60øW x x (a) (b)

120'E 150øE 180' 150'W 120øW 90'W 60'W 120'E 150'E 180' 150'W 120øW 90•W 60øW x x (c) (d)

120'E 150'E 180' 150øW 120•W 90'W 60'W 120'E 150'E 180' 150'W 120•W 90'W 60'W x x (e) (f)

Figure 7. Tracerdistribution on the •r0=25.0surface after 10 years,showing (a) the Northern Hemispheretracer for the casewith no ITF, (b) the SouthernHemisphere tracer for the case with no ITF, (c) the NorthernHemisphere tracer for the casewith 10 Sv ITF, (d) the Southern Hemispheretracer for the casewith 10 Sv ITF, (e) the NorthernHemisphere tracer for the case with 20 Sv ITF, and (f) the SouthernHenrisphere tracer for the casewith 20 Sv ITF.

ern Hemisphere tracer. The Northern Hemispheretracer Northern Hemispheric tracer along the equator is less concentration within the EUC is a pronounced local than 0.7. and there is a significant reduction in the maximum (with tracer concentrationin excessof 0.9) off-equatorialconcentration as well when comparedto associated with the eastward transport of the EUC, the casewith no ITF. The Southern Hemispheric tracer whereas for the Southern Hemisphere there is no lo- shows elevated concentrations between 2øS and 2øN rel- cal maximum of tracer concentration within the EUC ative to the case with no ITF, and the Southern Hemi- (Figure8b). The SouthernHemisphere tracer is largely spherictracer now accessesthe EUC. With 20 Sv of confined to the region south of 2øS. imposedITF the maximum concentrationof Northern For the run that imposed 10 Sv of ITF (Figures 8c Hemisphericwater in the EUC is now lessthan 0.6, and and 8d). the maximum in the concentration of the there is now an isolated local maximum of Southern RODGERS ET AL.' INDONESIAN THROUGHFLOW IN EQUATORIAL PACIFIC 20,561

., , i i i i I

10'S 5'N 10'N

10øS 5'S 0' 5'N 10'N Y

_ I I I I ' '

10'S 5øS 0•/ 5'N 10øN 10'S 5'S 0; 5'N 10'N

(e) (f)

Figure 8. Tracer distribution at 150øW after 20 years of model integration, showing(a) the Northern Hemispheretracer for the case with no ITF, (b) the Southern Hemisphere tracer for the casewith no ITF, (c) the NorthernHemisphere tracer for the casewith 10 Sv ITF, (d) the SouthernHemisphere tracer for the casewith 10 Sv ITF, (e) the NorthernHemisphere tracer for the casewith 20 Sv ITF, and (f) the SouthernHemisphere tracer for the casewith 20 Sv ITF.

Hemisphericwater within the EUC. In addition, there the Northei'n Hemisphere ventilates more rapidly than is some cross-equatorialtransport of Southern Henri- the signalfro TM the SouthernHemisphere. The mixing sphere tracer into the NECC, which is consistent with ratio changesvery little afteryear 20 for all threecases. observations[Gordon, 1995]. The results for the hemispheric tracer are summa- The evolutionwith time of the mixing ratio of south- rized in Table 1. A rectangular box ranging in depth ern component to northern component waters in the from 50 to 150 m and in latitude from 2øN to 2øS has equatorial thermocline for the case with no ITF, the been chosen to correspond to the equatorial thermo- casewith 10 Sv of ITF, and the casewith 20 Sv of ITF, cline. Within this box the averageconcentration of both is shownin Figure 9. For all three casesthe mixing the northern and southern tracer has been tabulated for ratio is smallestafter 5 years,as the initial signalfrom the end of the 20th year of model integration. 20,562 RODGERS ET AL.' INDONESIAN THROUGHFLOW IN EQUATORIAL PACIFIC

1.5

1.0 Z

• OSv ,10Sv • 20 Sv

0.5

0.0 5.0 10.0 15.0 20.0 Time (years)

Figure 9. Evolutionof mixing ratio of southernto northerncomponent water in Equatorial Undercurrentfor three modelruns over20 years(Lainout OceanModel, levelversion of Gent and Cane[1989]). Although the systemhas not yet reachedsteady state, this timescaleis longerthan the 10-15 year ventilation 'age' of equatorialpycnocline water inferred from chlorofiuorocarbon measurementsin the Pacific[Warner et al., 1996].

In Figure 10 the mixing ratio of the Northern Hemi- across30øS is roughly 30 Sv for each case. However, spheric tracer to the Southern Henrisphere tracer is the casesdiffer in their calculated transport of the East shown as a function of depth and longitude after 20 Australia Current, the western boundary current of the yearsfor each of the three cases(0, 10, and 20 Sv). For South Pacificsubtropical gyre. For the 0 Sv of imposed each case the mixing ratio has been averagedbetween ITF transport, the recirculationwithin the East Aus- 2øN and 2øS.With 0 Sv of ITF (Figure 10a), we seethat tralia Current is roughly 30 Sv, whereaswith 10 or 20 the mixing ratio is consistentlygreater than 2 over the Sv of imposedITF transport, the east Australia Cur- equatorial pycnocline. However, this plot also reveals rent is diminishedby an amount equal to the imposed significant zonal and vertical structure, with the ratio ITF transport. This is a consequenceof continuityand tending to increase below 100 m depth. The largest gradients near 140øE, between 100 and 300 m depth, are associatedwith the retrofiection shown in Figure 5. Table 1. Percentageof Different ComponentWaters The casefor 10 Sv of ITF (Figure 10b) also reveals After 20 Years as a Function of Imposed Indonesian significant vertical and zonal structure. Here the mix- ThroughflowTransport ing ratio in the upper 200 m is between1 and 2 for most of the region between 150øE and the eastern boundary, Component 0 Sv 10 S½ 20 Sv with the relative contribution of the Northern Hemi- spheric tracer increasing to the east and with depth. Northern Hemisphere 66 43 33 Southern Hemisphere 24 31 41 With 20 Sv of ITF (Figure 10c) the ratio is lessthan 1 Other 10 26 26 over most of the region. As with the casewith 10 Sv, the relative contribution of the northern component water increasesto the east and with depth. The percentagesare calculatedby taking the averagecon- centrationof the hemispherictracers within a box bounded The barotropic stream function calculated by the in the vertical by the 50 and 150 m depth horizonsand model is nearly identical over the interior regions of by 2øN and 2øSxneridionally over the entire equatorialPa- the Pacific for the three cases(0, 10, and 20 Sv of im- cific. "Other" is calculated by subractingthe sum of the two posed ITF) discussedhere. The Sverdrup circulation hexnispherictracer concentrationsfrom unity for eachrun. RODGERS ET AL.. INDONESIAN THROUGHFLOW IN EQUATORIAL PACIFIC 20,563

focuses on temperature anomalies of order 0.5øC ad- veering into the equatorial region from poleward of 30ø. In essence, this involves a temperature anomaly ad- veered by a mean flow, eventually impacting the temperature of the thermocline downstream. It is known, however, that the equatorial circulation is quite variable. Thus it is worth considering the basin-scale time mean temperature gradients within the pycnocline (VT of (1)) correspondingto the salinitydistribution shown in Figure 2. 150øE 180' X150øW 120øW 90øW The temperature distribution on the •s-25.0 isopycnal surface between 30øN and 30øS calculated fi'om the Levituset al. [1994]and Levitusand Boyer [1994] annual mean temperature and salinity fields is shown in Figure 11a. As before, the •=25.0 surface is chosen becauseit correspondsto the approximate density horizon of the core of the equatorial undercurrent. A meridional temperature gradient exists acrossthe equator. Minimum temperatures of lessthan 15øC are found within the California Current off Baja California near 120øW. Maximum temperatures in excessof 22øC can be found between 10 ø and 20øS across the South Pa- cific subtropical gyrc. As with the hemispheric salinity gradients in Figure 2, the mcridional temperature gra- 150øE 180' 150'W 120øW 90'W x dients on this isopycnal surface are due to differences in (b) surface freshwater fluxes in the Northern and Southern Hemispheres. Figure 1lb showsa zonally averagedplot of tempera- turc as a function of latitude and density from the Lev- itus and Boyer [1994]climatology for the Pacific. The 2øC change in zonally averaged temperature between 8øN and 8øS on the •=25.0 surface can be seen to exist 2 12 over the density range •-24.0 to 26.5. As the horizontal smoothingradius for the Levitusand Boyer [1994] and Leviilt8ei ai. [1994]data setsis of order500 kin, we cannot expect to resolve fronts or features on finer scales than this. This temperature distribution representsthe 150'E 180' 150'W 120øW 90'W x T of (1), and for the real ocean,OT/Oy within the equa- (c) torial thermoclineis of order 2øC/1600km alongisopy- cnal surfaces. Figure 10. The mixing ratio of northern to southern Having consideredthe large-scaletime mean temper- componentwater plotted as a function of longitude and depth along the equator. The plotted quantity repre- ature distribution on an isopycnal surface within the sents an average over the latitude range 2øN to 2øS. Pacific pycnocline, we turn our attention to a transect Shownare caseswith (a) 0 Sv ITF, (b) 10 Sv ITF, and at 165øE that was sampled at least twice per year be- (c) 20 Sv ITF. The contouringlevels for the mixingra- tween 1984 and 1990. The data set discussed here con- tio are 0.1, 0.2, 0.4. 0.6, 0.8, 1.0, 1.2, 1.4, 1.6, 1.8, 2, sists of conductivity-temperature-depth(CTD) casts 2.5, 3, 4.5, 7, 10, 50, 100. fi'om 10 cruises by the Surveillance Trans-Oc•anique du Pacifique (SURTROPAC) program conductedby the Institut Fran•ais de Recherche Scientifique pour le D•veloppement en Cooperation (ORSTOM) group the fact that the interior circulation is determined by from New Caledonia, three cruises from the Production Sverdrup dynamics. P•lagique dansle Pacifique(PROPPAC) program,and sevencruises from the United States/People'sRepublic 5. Variability in Hydrographic Data of China (US/PRC) program. The time averagedfields along this transect have been describedby Gouriou and The mechanismproposed by Deser et al. [1996],Gu Toole[1993]. andPhilander [1997], and Zhang et al. [1998]for chang- Figure 12 shows a scatterplot of temperature and ing the thermal structure of the equatorial pycnocline salinity measurements(indicated by open circles) for 20,564 RODGERS ET AL.' INDONESIAN THROUGHFLOW IN EQUATORIAL PACIFIC

120øE 150øE 180 ø 150øW 120øW 90øW x

(a)

10øS 5øS 0 ø 5øN 10øN Y

(b)

Figure 11. Temperaturesfrom the Levitusand Boyer[1994] annual mean climatology (a) of the cr0-25.0 isopycnalsurface and (b) as a fimction of latitude and density(zonally averaged).

the CTD measurementsfor equatorialstations (165øE, ture as a function of latitude and time for this data set 0øN)for all of the cruisescombined. The correspondingalong the isopycnalsurface cr0=24.5. Along this surface, or0surfaces are superposedon the figure, and the Min- which lies in the upper part of the undercurrent, the danaoCurrent (cold and fresh) and New GuineaCoastal temperature gradient of approximately 2øC can be seen U•dercurrent(warm and salty) end-members areshown acrossthe equator. Between 2øS and 2øN, which corre- as stars. Figure 12 showsthat the temperature vari- spondsto the latitude range of the Equatorial Under- ability existson the equator when viewedisopycnally, current, there is low-frequency variability, with cooler although this plot tells us nothing about the phaseof temperatures measured in late 1987 followed by warmer the variability. For the density range betweencr•=25.0 temperatures. There is also low-frequency temperature and 26.0, which roughly correspondsto the equatorial variability within the South Equatorial Current at ap- undercurrent, the variance is of order 1øC. Also worth proximately 4øS, where the temperatuTM in early 1987 notingis the fact that the meanmixing ratio of southern is approximately 1øC cooler than in early 1988. to northerncomponent water is slightlylarger than 0.5 betweencr0=25.0 and 26.0 and increaseswith depth, so that by cr0=27.0 the source of the water at this location 6. Discussion is dominatedby the southerncomponent. We have seen from our modeling results that the Figure 13 showsa Hovmoellerdiagrams for tempera- equatorial thermocline circulation is indeed quite sen- RODGERS ET AL' INDONESIAN THROUGHFLOW IN EQUATORIAL PACIFIC 20,565

density 25

o cy;••o 20 c•'-' 0 •.0 O0 o

0 C

10 34.6 34.8 315 35.2I 35.4 35.6 salt

Figure 12. T/$ scatterplotfor conductivity-temperature-depth(CTD) data collectedat 165øE, 0øN (open circles)between 1984 and 1990 (Data courtesyof John Toole, WoodsHole Oceano- graphicInstitution, 1998). The starsrepresent OTD measurementsfrom the low-latitudewestern boundarycurrent (LLWBC) regions,namely, the freshMindanao Current (left) for the northern hemisphere,and the coastof New Guineafor the southernhemisphere (right). Contoursof potential density anomaly are also inchtded.

--21.5' • .?.5 21.5j 21.7522

-2

-4

-6

1984 1985 1986 1987 1988 1989 1990 Year

Figure 13. Temperature(tinhe, latitude) at 165øEon rr0:24.5 between1984 and 1990. These CTD data are courtesyof John Toole, Woods Hole OceanographicInstitution, and have been interpolated to isopycnal surfaces. 20,566 RODGERS ET AL.: INDONESIAN THROUGHFLOW IN EQUATORIAL PACIFIC sitive to the transport of thermocline water from the existence of the ITF adds an additional pathway not Pacific to the Indian Ocean through the Indonesian! availablefor a closeddomain. Fi•,e [1985]has pointed Straits. Only when the imposed transport is between out that the tritium measured in the thermocline of the 10 and 20 Sv does the mixing ratio of the Southern to equatorialIndian Ocean during the GeochemicalOcean the Northern Hemispheric tracer become larger than 1 SectionsSurvey (GEOSECS) of the 1970soriginated in for the equatorial pycnocline, in agreement with what the North Pacific and thus subducted in the "ITF win- is seenin the T/$ scatterplotof Figure 12. Assuming dow" of the North Pacific subtropical gyre. that the mixing ratio increasesapproximately linearly The three model runs considered in this study did with the imposedITF transport and given the salinity- not allow for any tinhe variability in the ITF trans- derivedestimate of 1/2 to 2/3 of the thermoclinederive port. Given that there exists a 2ø temperature gra- from the SouthernHemisphere [Tsuch,'iya et al., 1989], dient along isopycnalsbetween the northern and south- our results are consistentwith Godfrey's estimate of 16 ern component water in the western equatorial Pacific Sv calculatedusing the Island Rule. However, this value thermocline, as well as strong sensitivity of the mixing of 16 Sv is approximately 50 per cent larger than the ratio of northern to southern component water in the meanvalues inferred from observations[Godfrey, 1996]. EUC to the transport of the ITF, a question naturally For the hemispheric tracers the results for the three arises: to what extent could real-world variability in runs considered here are consistent with the results of ITF transport lead to variability in the thermal struc- L'ua•,d McCreary [1995], insofar as the ITCZ presentsa ture of the equatorial pycnocline?More specifically,we potential vorticity barrier that inhibits interior ventila- are interested in investigating the relative magnitudes tion from the Northern Hemisphere. The barrier formed of the termsin (1) and workingquantitatively through by the ITCZ is not impermeable, as interior ventilation the implications for climate variability. This will be does occur for the Northern Hemisphere along isopyc- investigated in a future study. hals above the core of the undercurrent. Nevertheless, As mentionedearlier, H'irst and Godfrey[1993] found the ITCZ barrier is responsiblefor a hemispheric asym- that easternequatorial Pacific SSTs are coolerwhen the merry, in that interior ventilation occursmore readily IndonesianThroughflow is open in their model. This is from the Southern Hemisphere. despite the fact that for this case, more water is being The choice of 18 ø latitude as the cutoff for the tracer- drawn from the Southern Hemisphere, where we expect forcing function was motivated by our interest to be temperatures to be warnherfor the sanheisopycnal sur- consistent with the model experiment of Lu and Mc- face. Hirst and Godfrey attributed this to the water Creary[1995]. However, their choiceof 18ø latitudefor that exits the Pacific thermocline through the ITF com- their tracer forcing function was motivated by the fact ing from a shallowerlevel then the compensatinginflow that this is the cutoff latitude for their subduction pa- from the South Pacific, with the net effect that the Pa- rameterization. Since this choice of a cutoff latitude cific thermocline shallowsfor an open ITF. doesnot have specialsignificance for the physicalprob- In this study the SST is also quite sensitiveto whether lem of intergyre exchange,the sensitivity of the results the ITF is open or closed. Figure 14 shows that in the presentedhere to the cutoff latitude of the tracer forcing casewith 10 Sv of imposed ITF, the annual mean SST function will be pursued in our further study. is as much as iøC colder than in the case with no ITF. Continuity requiresthe ITF transport to be balanced The difference in SST patterns between the case with by a cross-equatorial flow of water from the Southern 10 Sv of imposedtransport and the casewith 20 Sv of Hemisphere. Southern component thermocline waters imposedtransport is quite small in comparison. Our re- do not crossthe equator easily in the western or central suits in the eastern equatorial Pacific are consistentwith Pacific; note the strong retroflection along the north the resultsof H'irstand Godfrey[1993], except that the coast of New Guinea in Figures 5b and 5d, as well as SST differencesare less confined to the equator for our the fact that the strong equatorial fronts in tempera- model. This can be attributed in large part to the dif- ture and salinity are maintained in steady state for the ferences in the surface heat flux parameterization used real ocean. Upon reachingthe northern coast of Papua in the two studies. Hirst and Godfrey[1994] used the New Guinea near 140øE, nearly all of the northwestward parameterizationof Hah [1984],which includes a strong flowingNGCUC water retroflectsto feedthe EUC. Only damping term on SST. By using the AML of $eager et after traversing the Pacific from west to east within the al. [1995],we haveallowed changes in the temperature EUC, upwellingin the easternPacific, and undergoing of the Peru upwelling to propagate into the ocean inte- a water mass transformation at the sea surface, is the rior through the influence of the trade winds. majority of this southern component water able to enter The sensitivity of SST in the tropics to the imposed the Northern Hemisphere. ITF transport should be understood as the surface ex- In a seriesof oceanmodel experimentsusing a closed pression of a response in the depth of the pycnocline rectangular domain with idealized winds, Liu and Ph'i- to the ITF. When the ITF is open, the entire equato- lander[1995] distinguished between "recirculation win- rial pycnocline is shallower than for the case where the dows" and "exchangewindows" for the subductingre- ITF is closed. The changesin SST associatedwith the gionsof the subtropicalgyres. For the North Pacific the changesin pycnocline depth are large enough to more RODGERS ET AL.: INDONESIAN THROUGHFLOW IN EQUATORIAL PACIFIC 20,567

120øE 150øE 180' 150øW 120øW 90øW 60øW X

Figure 14. Differencein annual mean SST bd;ween case with 10 Sv of imposed ITF and case with no ITF. Units are øC, and negative contoursin the Eastern Equatorial Pacific indicate that temperatures are cooler when the ITF is open.

than compensatefor the meridional temperature gra- diapycnal mixing within the model due to the mixing dients seenon isopycnalsurfaces within the pycnocline parameterizationof Pacanowskiand Philander[1981], of Figure 11a. The SST sensitivity poleward of 20ø in in our model. For the henrispheric tracers used here, either hemisphereis not insignificantand again reflects one cannot readily distinguish between cool and fresh changesin the depth of the pycnocline on a basin-wide water that has been locally entrained from below and scale. The dynamical mechanismsresponsible for this subtropical water that is merely slow in accessingthe strong non local responseto changesin ITF transport equatorial pycnocline along isopycnal surfaces. will be the subject of future investigation. The studyof Lu et al. [1998],which used a 3-1/2 layer Another subject for further investigationwill be the model with an account of the Indonesian Throughflow, variation in pycnocline depth with variations in Indone- suggeststhat the mixing ratio of southern to northern sian Throughflow transport on seasonalto interannual component water in the equatorial pycnocline should be timescales.All the studiesdiscussed ignore seasonal and dependent on the extent of diapycnal fluxes within the interannual variability in ITF transport. If the cooler equatorial pycnocline. Increased entrainment of water SSTs found in our study (as well as in the Hirst and from below the directly ventilated thermocline tends to Godfrei/[1994]s•udy) for the casewhere the ITF is open increase the relative contribution of southern compo- were to correspondto the responseof the real ocean to nent water. If this is true, then the fact that the ITF increased throughflow transport, then this would offer requiredby our model is roughly 50% larger than ob- a feedback mechanism with potential implications for servationscould be a manifestation of diapycnal mixing the ocean thermostat mechanismproposed by •lement with the equatorial pycnocline being too small for our et al. [1996]. Assumingthat the zonal SST gradient model runs. In a future study the hemispheric tracers increasesin responseto increasedgreenhouse forcing, used here will be generalized such that they tag partic- the trade winds respondby increasingin strength, thus ular isopycnal layers from each hemisphere. In this way increasingthe differencein sea level height between the we can quantitatively addressthe extent of entrainment Pacific and Indian Oceans. Owing to the increased associatedwith intergyre exchange. pressure head difference between the Pacific and In- Two caveats are in order with regard to the compari- dian Oceans,the transport of the ITF would increase, son of the model and the data. First, our comparison is thereby cooling the sea surfacein the eastern equatorial complicated by the fact that for the model, there is sig- Pacific, thus constituting a positive feedback. nificant zonal structure to the mixing ratio of northern Accordingto Table 1, approximately25% of the wa- to southern component water in the equatorial thermo- ter within the equatorial pycnocline after 20 years falls cline (Figure 10). This zonal structurewill also exist into the 'other' category for the two runs that include when the mixing ratio is evaluated along isopycnal sur- ITF transport. This raises the question of the extent of faces in the equatorial pycnocline. For the case with 10 20,568 RODGERS ET AL.: INDONESIAN THROUGHFLOW IN EQUATORIAL PACIFIC

Sv of imposedITF (Figure 10b), there are roughlyequal with 0 Sv of ITF, nearly all of this northern component proportions of northern to southern componentwater in water directlyfeeds the equatorialundercurrent (EUC) the region to the west of the dateline, which corresonds upon reaching the equator. whereaswith 10 Sv of ITF, to the region of sampling used for the salinity-based much of the Mindanao Current water is diverted to the estimates[Tsuchiya et al., 1989, Gouriou and Toole, Indian Ocean. As for the inflow from the Southern 1993]. Hemispheric LLWBC, the NGCUC is relatively weak This underscoresthe need for a more complete anal- with 0 Sv of ITF but is significantly strengthenedfor ysis of the existing databaseof hydrographicmeasure- the casewith 10 Sv of ITF, as seenin Figures 6c and 6d. ments in the equatorial Pacific thermocline, in order to It is principally the sensitivity of these LLWBCs to the further our understandingof the sourcesof upwelling strength of the imposed ITF transport that is reflected equatorial thermocline water. This should include hy- in the mixing ratio of our idealized tracers. drographic measurementsfrom a range of longitudes Given that this strong sensitivity exists, one would acrossthe Pacific in order to better establishthe spatial expectthat the mixing ratio would be roughly50/50 stucture of the mixing ratio in the real oceanas revealed for the case with 0 Sv of ITF and dominated by the in the salinity field. In our future work we will also in- southern source for the case with 10 Sv of ITF. The corporate measurements of the transient tracers bomb point we wish to emphasize here is that the sensitiv- radiocarbonand bomb tritium in the equatorialthermo- ity exists but that it is not yet clear why the north- cline to adressthis question (as suggestedby Broecker ern component is dominant for a closed Pacific domain. et •. [1995]). Our results are not inconsistent with those of Lu and The second caveat is that caution must be exercized McCreary[1995], since for all three cases(0, 10, and when using information derived from passivetracers to 20 Sv) an ITCZ "barrier" is maintained in the North- infer the behavior of temperature and salinity anoma- ern Hemisphere. Lu and McCreary did not quantify lies in the ocean. Some important differencesbetween a mixing ratio of northern to southern component wa- passiveand active tracers have been discussedby ter in their study, as their single passivetracer did not and Huang[1988]. Nevertheless,our hemispherictrac- distinguish between northern and southern sources. It ers provide a usetiff diagnostic for the sourcesof the would be interesting to compare the results presented model's equatorial thermocline water for the mean state here with results obtained with other models such as with climatologicalforcing, and the resultsprovide mo- theirs. tivation for a more carefulreassessment of the existing A meridional temperature gradient of amplitude 2øC database of hydrographicand transient tracer data. exists on isopycnalsurfaces in the western Pacific pycnocline. and thus changesin the mixing ratio of north- 7. Conclusions ern to southern component water have the potential to change the thermal structure of the pycnocline. De- The transport of the ITF strongly influencesthe ratio pending on the amplitude of these changes,this has the of northern to southerncomponent water in the equato- potential to change the thermal structure of the equa- rim pycnocline. High-resolution ocean models of Pacific torial pycnocline and thus to influence climate. As has circulation that do not accotint for an ITF are char- been shownwith the CTD data collectedalong the re- acterized by an EUC that is dominated by northern peat section at 165øE between 1984 and 1990, there component water. This is inconsistent with the ratio is significant low-frequency variability in temperature of northern to southern component water inferred from and salinity within the pycnoclineof the western equa- salinity measurementsin the equatorial pycnocline. In torial Pacific. This is consistentwith the recentfindings our model runs. only when an ITF transport of be- of W.S. Kessler(personal communication, 1998), who tween 10 and 20 Sv is imposeddoes the mixing ratio has shown that low-fi'equencyvariability in the salin- of northern to southern component waters come into ity structure of the western Pacific pycnocline accotints accord with observations. for as much as one third of the variance observed in The strong sensitivityto the strength of the ITF was dynmnic height. certainly an unexpected restilt. We have seen that the It remains to be determined whether the mechanism dynamical reasonfor this sensitivity can be tinderstood responsible for the observed subsurface temperature by consideringmodel flow fields within the pycnocline variability on isopycnals at 165øE is due to a mecha- (shownin Figtires 5c and 5d) for the caseswith 0 and nismof the type proposedby Gu and Philander[1997], 10 Sv of ITF, respectively.The most significantimpact by variability of the ITF transport as proposed in this of the changingthe ITF transport occursin the equa- study, by changes in diapycnal exchange rates, or by torwardflowing LLWBCs of eitherhemisphere, namely. some combination thereof. The studies of Gu and Phi- the MindanaoCurrent in the NorthernHenrisphere, and lander[1997] and Blankeand Raynaud [1997] have used the New GuineaCoastal Undercurrent (NGCUC) in the point tracers to illustrate ventilation pathways for the SouthernHemisphere. The equatorwardtransport of Pacific thermocline circulation. Sincethese tracers only the Mindanao Current across 5øN at the western bound- follow ventilation pathways determined by the model's ary is nearly the same for both model runs. However, advection equation, they are not influenced by the mix- RODGERS ET AL.: INDONESIAN THROUGHFLOW IN EQUATORIAL PACIFIC 20,569 ing processesthat occur within the model. However, Fine, R.A., Direct evidenceusing Tritium data for through- we knowfrom transienttracer measurements[Quayet flow from the Pacific into the Indian Ocean, Nature, 315, al., 1983]that suchprocesses are not insignificantin the 478-480, 1985. equatorial Pacific pycnocline. Gent, P. R., and M. A. Cane, A reduced gravity, primitive equation model of the upper equatorial ocean, J. Cornput. As we have seenin Table 1, for our model runs, en- Phys., 81,444-480, 1989. trainment of cool water from below within the equa- Gent, P. R., and J. C. McWilliams. Isopycnal mixing in torial Pacific provides up to 25% of the water within ocean circulation models, J. Phys. Oceanogr., 20, 150- the equatorial pycnocline for the cases where we ac- 155, 1990. count for the IndonesianThroughflow. Viewed in light Gouriou, Y., and J. M. Toole, Mean circulation of the upper of the modelresults of Lu et al. [1998],this couldbe less layers of the western equatorial Pacific Ocean, J. Geophys. Res., 98, 22,495-22,520, 1993. than the entrainment fluxes that occur in the real ocean, Godfrey, J.S.. A Sverdrup model of the depth-integrated which nfight explain why the ITF transport required by flow for the World Ocean allowing for Island Circulations, our model is somewhat larger than the estimates from Geophys. Astrophys. Fluid Dyn., •5, 89-112, 1989. observations. Godfrey, J.S., The effect of the Indonesian throughflow on Future modelingwork will involveforcing the model ocean circulation and heat exchangewith the atmosphere: A review, J. Geophys. Res., 101, 12,217-12,237, 1996. with seasonaland interannual winds and letting the Gordon, A.L., Interocean exchangeof thermocline water, J. model internally calculate its own ITF transport. This Geophys. Res., 91, 5037-5046, 1986. will allow us to addressthe magnitude of the effect of Gordon, A.L., When is appearance reality? A comment on realistic ITF variability on the thermal budget of the Why Does the Indonesian Throughflow Appear to Origi- equatorial pycnocline. Future model runs will also in- nate from the North Pacific, J. Phys. Oceanogr., 25, 1560- 1570, 1995. clude tracers that facilitate quantification of diapycnal Griffies, S.M.. The Gent-McWilliams skew-flux, J. Phys. exchangeswithin the model. Oceanogr., 28, 831-841, 1998. Griffies, S.M., A. Gnanadesikan, R.C. Pacanowski, V.D. Acknowledgments. Larichev. J.K. Dukowicz, and R.D. Smith, Isoneutral dif- We would like to thank Arnold Gordon, Bill Jenkins. and fusion in a z-coordinate ocean model, J. Phys. Oceanogr., Martin Visbeck for constructive comments and feedback. 28, 805-830, 1998. We wouldalso like to thank John Tooleat WHOI for making Gu. D., and S.G.H. Philander. Interdecadal climate fluctu- the CTD data at 165øEavailable to us. This work was sup- ported by National Science Foundation contract numbers ations that depend on exchangesbetween the tropics and the extratropics, Science, 275, 805-807. 1997. OCE-9633375 and OCE-9796253. This is Lamont-Doherty Hah. Y., A numerical world ocean general circulation model, Earth Observatory contribution 5955. II, A Baroclinic Experiment, Dyn. Atrnos. Oceans. 8, 141-172, 1984. References Hellerman, S., and M. Rosenstein, Normal monthly wind Blanke.B., and S. Raynaud.Kinematics of the PacificEqua- stressover the World Ocean with error estimates, J. Phys. torial Undercurrent: An Eulerian and Lagrangian ap- Oceanogr., 13, 1093-1104, 1983. proachfrom GCM results. J. Phys. Oceanogr..27, 1038- Hirst, A. C., and J. S. Godfrey, The role of Indonesian 1053, 1997. throughflow in a global Ocean GCM, J. Phys. Oceanogr., Broecker,W.S., S. Sutherland,W. Smethie, T.-H. Peng, 23, 1057-1086, 1993. and G. Ostlund, Oceanicradiocarbon: separationof the Jenkins, W.J., The exchange and upwelling of subtropi- natural and bomb components,Global Biogeoch. Cycles. cal waters in the Tropical Pacific (abstract), EOS Trans 9, 263-288, 1995. AGU, 77(46), Fall Meet. Suppl., 359, 1996. Clement,A.C.. R. Seaget,M.A. Cane, and S.E. Zebiak, An Levitus, S., R. Burgeft. and T. P. Boyer, World Ocean Atlas oceandynamical thermostat, J. Clim., 9, 2190-2196, 1996. 199•, vol. 3, Salinity. 99 pp., U.S. Dep. of Cornruer., daSilva, A., C. Young, and S. Levitus, Atlas of surfacema- Washington, D. C., 1994. rine data 1994, vol. 1, Algorithms and procedures,NOAA Levitus, S., and T. P. Boyer, World Ocean Atlas 199•, vol. Atlas NESDIS 6, U.S. Dep. of Cornmet..Washington, 4, Temperature, 117 pp., U.S. Dep. of Commer., Wash- D.C., 1994. ington, D. C.. 1994. Deser, C., M.A. Alexander,and M.S. Timlin, Upper-ocean Liu, Z., and B. Huang, Why is there a tritium maximum thermal variationsin the North Pacificduring 1970-1991, in the Central Equatorial Pacific thermocline?, J. Phys. J. Clim., 9, 1840-1855, 1996. Oceanogr., œ8,1527-1533, 1998. Druffel,E., Radiocarbonin annualcoral rings from the east- Liu, Z.. and S. G. H. Philander, How different wind stress ern tropical Pacific Ocean, Geophys.Res. Left., 8, 59-62. patterns affect the tropical-subtropical circulations in the 1981. upper ocean, J. Phys. Oceanogr., 25, 449-462. 1995. Druffel, E., Bomb radiocarbon in the Pacific, J. Mar. Res., Liu, Z., S. G. H. Philander, and R. C. Pacanowski, A GCM 19, 35-46. 1987. study of tropical-subtropical upper-oceanwater exchange, Ffield, A., and A.L. Gordon, Vertical mixing in the Indone- J. Phys. Oceanogr., 2•, 2606-2623. 1994. sian thermocline, J. Phys. Oceanogr.,22, 184-195, 1992. Lu, P., and J.P. McCreary, Influence of the ITCZ on the flow Fieux, M., C. Andtie, P. Delacluse, A.G. Ilahude, A. Kar- of thermocline water from the subtropical to the Equato- tavseff, F. Mantisi, R. Molcard, and J.C. Swallow, Mea- rial Pacific Ocean, J. Phys. Oceanogr., 25, 3076-3088, surementswithin the Pacific-Indian OceansThroughflow 1995. region, Deep Sea Res., Part 1, •1, 1091-1130, 1994. Lu, P., J.P. McCreary, and B.A. Klinger, Meridional circu- Fieux, M., R. Molcard, and A.G. Ilahude, Geostrophic lation cellsand the sourcewaters of the Pacific Equatorial transport of the Pacific-Indian Oceans throughflow. J. Undercurrent, J. Phys. Oceanogr.,28, 62-84, 1998. Geophys. Res., 101, 12,421-12,432. 1996. McCreary, J.P., and P. Lu. Interaction betweenthe subtrop- 20,570 RODGERSET AL.: INDONESIAN THROUGHFLOW IN EQUATORIAL PACIFIC

ical and the equatorial ocean circulations: The subtropical Tsuchiya,M., R. Lukas,R. A. Fine, E. Firing, and E. Lind- cell, J. Phys. Oceanogr., 2•, 466-497, 1994. strom, Sourcewaters of the PacificEquatorial Undercur- Murray, S.P., S.P., E. Lindstrom, J. Kindle, and E. Weeks, rent, Prog. Oceanogr.,23, 101-147, 1989. Transport through Vitiaz Strait, WOCE Notes, 7(1), 21- Warner, M.. J. Bullister, D. Wisegarves,R. Gammon, and 23, 1995. R. Weiss, Basin-wide distributions of chlorofiuorocarbons Naik, N., M. A. Cane, S. Basin, and M. Israeli, A solver CFC-11 and CFC-12 in the North Pacific: 1985-1989, J. for the barotropic mode in the presence of variable to- Geophys. Res., 101, 22,525-22,542, 1996. pography and islands, Mon. Weather Rev., 123, 817-832, Wyrtki, K., An estimateof equatorialupwelling in the Pa- 1995. cific, J. Phys. Oceanogr.,11, 1205-1214, 1981. Pacanowski, R.C., and S.G. Philander, Parameterization of Zebiak, S.E., and M.A. Cane, Natural climate variability vertical mixing in numerical modelsof the tropical oceans, in a coupledmodel. in Workshopon Greenhouse-Gas- J. Phys. Oceanogr., 11, 1443-1531, 1981. Induced Climate Change: A Critical Appraisal of Simu- Quay, P.D., M. Stuiver, and W.S. Broecker, Upwelling rates lations and Observations,edited by M.E. Schlesinger,pp. for the Equatorial Pacific Ocean derived from the bomb 457-470, Elsevier, New York, 1991. x4C distribution,J. Mar. Res., J1, 769-793,1983. Zhang,R.-H., L.M. Rothstein,and A.J. Busalacchi,Origin Redi. M.H., Oceanic isopycnal mixing by coordinate rota- of upper-oceanwarming and E1 Nino changesin the trop- tion. J. Phys. Oceanogr., 12, 1154-1158, 1982. ical Pacific Ocean, Nature, 391, 879-883, 1998. Rothstein, L.M., R.H. Zhang, A.J. Busalacchi, and D. Chen, A numerical simulation of the mean water pathways in the subtropical and tropical Pacific Ocean, J. Phys. Oceanogr., K.B. Rodgers, Max-Planck-Institut fiir Meteorologie, 28, 322-343, 1998. Bundesstrasse55, D-20146 Hamburg, Germany. (e-mail: Seager, R., B. Blumenthal, and Y. Kushnir, An advective [email protected]) atmospheric mixed layer model for ocean modeling pur- M.A. Cane and N. Naik, Lamont-Doherty Earth Obser- poses: Global simulation of surface heat fluxes, J. Clim., vatory, Columbia University, Palisades,NY, 10964. (e-mail: 8, 1951-1964, 1995. mcane(@ldeo.columbia.eduand [email protected]) Shriver, J.F.. and H.E. Hurlhurt, The contribution of the D.P. Schrag, Department of Earth and Planetary Sci- global thermohaline circulation to the Pacific to Indian ences,Harvard University, Cambridge. MA. 02138. (e-mail: Ocean throughflow via Indonesia, J. Geophys.Res., 102. [email protected]) 5491-5511, 1997. Toggweiler,J. R., K. Dixon, and W. S. Broecker.The Peru upwelling and the ventilation of the South Pacific ther- (ReceivedApril 6, 1998: revised November 4. 1998: mocline, J. Geophys. Res., 96, 20,467-20,497, 1991. acceptedNovember 17. 1998.)