a model for quantifying oceanic transport and mesoscale variability in the coral triangle of the...

22
A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago Frederic S. Castruccio, 1 Enrique N. Curchitser, 2,3 and Joan A. Kleypas 1 Received 12 June 2013 ; revised 11 September 2013 ; accepted 7 October 2013 ; published 19 November 2013. [1] The Indonesian Throughflow region (ITF) continues to pose significant research challenges with respect to its role in the global ocean circulation, the climate system, and the ecosystem sustainability in this region of maximum marine biodiversity. Complex geography and circulation features imply difficulties in both observational and numerical studies. In this work, results are presented from a newly developed high-resolution model for the Coral Triangle (CT) of the Indonesian/Philippines Archipelago specifically designed to address regional physical and ecological questions. Here, the model is used to quantify the transport through the various passages, sea surface temperature and mesoscale variability in the CT. Beyond extensive skill assessment exhibiting the model ability to represent many conspicuous features of the ITF, the high-resolution simulation is used to describe the mesoscale and submesoscale circulation through the application of Finite Size Lyapunov Exponents (FSLEs). The distribution of FSLEs is used to quantify the spatiotemporal variability in the regional mixing characteristics. The modeled seasonal and interannual variability of mixing suggests a link to large-scale climate signals such as ENSO and the Asian-Australian monsoon system. Citation : Castruccio, F. S., E. N. Curchitser, and J. A. Kleypas (2013), A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago, J. Geophys. Res. Oceans, 118, 6123–6144, doi:10.1002/2013JC009196. 1. Introduction [2] The Coral Triangle (CT) is a marine region that spans parts of Indonesia, Malaysia, Papua New Guinea, the Philippines, the Solomon Islands, and Timor-Leste (Figure 1). This region covers nearly 6 million km 2 , which is roughly three-quarters the land area of Australia and encompasses portions of two biogeographic regions : the Indonesian-Philippines Region and the Far Southwestern Pacific Region. Often referred to as the maritime continent, this region is located at the confluence of tropical waters from the North and South Pacific and within the pathways of the inter-ocean exchange between the Pacific and Indian oceans. The maritime continent is recognized both as a key driver of atmospheric circulation due to its enormous abil- ity to transfer heat from the ocean to the atmosphere [Neale and Slingo, 2003] and as a key checkpoint for the global thermohaline circulation [Gordon, 2005]. [3] The oceanographic complexity of this region (Figure 1) presents major challenges to both field oceanog- raphers and numerical modelers [Gordon and Kamenko- vich, 2010]. Within the CT, the Indonesian Archipelago (IA) represents a complex array of passages linking inter- connected shelves, deep basins, shallow and deep sills, and submerged ridges, that collectively provide a sea link between two oceans [Gordon et al., 2003]. Known as the Indonesian Throughflow (ITF), it is recognized as a key component of the global thermohaline circulation [Gordon and Fine, 1996; Hirst and Godfrey, 1993; Wajsowicz and Schneider, 2001]. It serves as the main return flow of upper ocean warm waters from the tropical Pacific Ocean to the tropical Indian Ocean that balances the spreading of deep waters that form at high latitudes. Since water in the west- ern tropical Pacific is warmer and fresher than in the Indian Ocean, the ITF transport impacts the temperature and salin- ity in the Pacific Ocean, Indian Ocean, and Indonesian seas and also affects the air-sea heat exchange patterns strongly influencing the Indo-Pacific climate [Song et al., 2007]. Observation-based estimates of the ITF transport are 15 Sv (1 Sv 5 10 6 m 3 s 21 ). As the water is transported, its hydro- logical characteristics are altered by heat and freshwater inputs from the Indonesian seas and by strong vertical mix- ing. On a local scale, tides and winds, which are primarily monsoonal, are the dominant forcings but the large-scale pressure gradient between the Pacific and Indian oceans is the main force driving the flow of Pacific water through the Indonesian Archipelago into the Indian Ocean. As a result, the structure and magnitude of the ITF varies on timescales 1 National Center for Atmospheric Research, Climate and Global Dynamics Division, Boulder, Colorado, USA. 2 IMCS, Rutgers University, New Brunswick, New Jersey, USA. 3 DES, Rutgers University, New Brunswick, New Jersey, USA. Corresponding author: F. S. Castruccio, National Center for Atmos- pheric Research, Climate and Global Dynamics Division, P.O. Box 3000, Boulder, CO 80307, USA. ([email protected]) V C 2013. American Geophysical Union. All Rights Reserved. 2169-9275/13/10.1002/2013JC009196 6123 JOURNAL OF GEOPHYSICAL RESEARCH : OCEANS, VOL. 118, 6123–6144, doi :10.1002/2013JC009196, 2013

Upload: joan-a

Post on 03-Feb-2017

214 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago

A model for quantifying oceanic transport and mesoscale variabilityin the Coral Triangle of the Indonesian/Philippines Archipelago

Frederic S. Castruccio,1 Enrique N. Curchitser,2,3 and Joan A. Kleypas1

Received 12 June 2013; revised 11 September 2013; accepted 7 October 2013; published 19 November 2013.

[1] The Indonesian Throughflow region (ITF) continues to pose significant researchchallenges with respect to its role in the global ocean circulation, the climate system, andthe ecosystem sustainability in this region of maximum marine biodiversity. Complexgeography and circulation features imply difficulties in both observational and numericalstudies. In this work, results are presented from a newly developed high-resolution modelfor the Coral Triangle (CT) of the Indonesian/Philippines Archipelago specifically designedto address regional physical and ecological questions. Here, the model is used to quantifythe transport through the various passages, sea surface temperature and mesoscalevariability in the CT. Beyond extensive skill assessment exhibiting the model ability torepresent many conspicuous features of the ITF, the high-resolution simulation is used todescribe the mesoscale and submesoscale circulation through the application of Finite SizeLyapunov Exponents (FSLEs). The distribution of FSLEs is used to quantify thespatiotemporal variability in the regional mixing characteristics. The modeled seasonal andinterannual variability of mixing suggests a link to large-scale climate signals such asENSO and the Asian-Australian monsoon system.

Citation: Castruccio, F. S., E. N. Curchitser, and J. A. Kleypas (2013), A model for quantifying oceanic transport and mesoscale variabilityin the Coral Triangle of the Indonesian/Philippines Archipelago, J. Geophys. Res. Oceans, 118, 6123–6144, doi:10.1002/2013JC009196.

1. Introduction

[2] The Coral Triangle (CT) is a marine region thatspans parts of Indonesia, Malaysia, Papua New Guinea,the Philippines, the Solomon Islands, and Timor-Leste(Figure 1). This region covers nearly 6 million km2, whichis roughly three-quarters the land area of Australia andencompasses portions of two biogeographic regions: theIndonesian-Philippines Region and the Far SouthwesternPacific Region. Often referred to as the maritime continent,this region is located at the confluence of tropical watersfrom the North and South Pacific and within the pathwaysof the inter-ocean exchange between the Pacific and Indianoceans. The maritime continent is recognized both as a keydriver of atmospheric circulation due to its enormous abil-ity to transfer heat from the ocean to the atmosphere [Nealeand Slingo, 2003] and as a key checkpoint for the globalthermohaline circulation [Gordon, 2005].

[3] The oceanographic complexity of this region(Figure 1) presents major challenges to both field oceanog-raphers and numerical modelers [Gordon and Kamenko-vich, 2010]. Within the CT, the Indonesian Archipelago(IA) represents a complex array of passages linking inter-connected shelves, deep basins, shallow and deep sills, andsubmerged ridges, that collectively provide a sea linkbetween two oceans [Gordon et al., 2003]. Known as theIndonesian Throughflow (ITF), it is recognized as a keycomponent of the global thermohaline circulation [Gordonand Fine, 1996; Hirst and Godfrey, 1993; Wajsowicz andSchneider, 2001]. It serves as the main return flow of upperocean warm waters from the tropical Pacific Ocean to thetropical Indian Ocean that balances the spreading of deepwaters that form at high latitudes. Since water in the west-ern tropical Pacific is warmer and fresher than in the IndianOcean, the ITF transport impacts the temperature and salin-ity in the Pacific Ocean, Indian Ocean, and Indonesian seasand also affects the air-sea heat exchange patterns stronglyinfluencing the Indo-Pacific climate [Song et al., 2007].Observation-based estimates of the ITF transport are �15 Sv(1 Sv 5 106 m3 s21). As the water is transported, its hydro-logical characteristics are altered by heat and freshwaterinputs from the Indonesian seas and by strong vertical mix-ing. On a local scale, tides and winds, which are primarilymonsoonal, are the dominant forcings but the large-scalepressure gradient between the Pacific and Indian oceans isthe main force driving the flow of Pacific water through theIndonesian Archipelago into the Indian Ocean. As a result,the structure and magnitude of the ITF varies on timescales

1National Center for Atmospheric Research, Climate and GlobalDynamics Division, Boulder, Colorado, USA.

2IMCS, Rutgers University, New Brunswick, New Jersey, USA.3DES, Rutgers University, New Brunswick, New Jersey, USA.

Corresponding author: F. S. Castruccio, National Center for Atmos-pheric Research, Climate and Global Dynamics Division, P.O. Box 3000,Boulder, CO 80307, USA. ([email protected])

VC 2013. American Geophysical Union. All Rights Reserved.2169-9275/13/10.1002/2013JC009196

6123

JOURNAL OF GEOPHYSICAL RESEARCH: OCEANS, VOL. 118, 6123–6144, doi:10.1002/2013JC009196, 2013

Page 2: A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago

from the interannual El Ni~no-Southern Oscillation (ENSO)signal to the semidiurnal tidal signals.

[4] The CT region is also strongly influenced by theSouth China Sea Throughflow (SCSTF). A recent study byQu et al. [2009] utilizing existing observations and resultsfrom ocean GCMs showed that the SCSTF is a heat andfreshwater conveyor, which may have an important influ-ence on the South China Sea (SCS) heat content, the path-way and vertical structure of the ITF, and the heat andfreshwater transport from the Pacific into the Indian Ocean.The interplay of the monsoon and the SCSTF and theresulting effect on the strength of the ITF are key to under-standing the regional climate variability and its implica-tions on a global scale.

[5] In addition to its importance in the global ocean andclimate variability, the CT region is also widely consideredthe apex of marine biodiversity for several major taxo-nomic groups [Tittensor et al., 2010], and particularly forzooxanthellate corals [Veron et al., 2009]. Over 120 millionpeople live in the CT and rely on its fisheries and coralreefs for food, income, and protection from storms. Conser-vation in the CT has thus become a top priority of stategovernments and international conservation efforts, withthe six Coral Triangle countries establishing the Coral Tri-angle Initiative (CTI) [Coral Triangle Secretariat, 2009] in2007. Conservation efforts recognize that because of theCT’s oceanographic complexity, changes in SST and other

oceanographic conditions are likely to vary spatially inresponse to climate change. Based on AVHRR PathfinderSea Surface Temperature (SST) for 1985–2006, Pe~nafloret al. [2009] found that SST in the CT has increased anaverage of 0.2�C per decade but with considerable variabil-ity across the region.

[6] The oceanographic complexity and large areal extentof the CT, however, present challenges for understandingthe roots of this spatial variability. Oceanographic modelsmust consider the complex interactions between topogra-phy, large-scale oceanic currents, surface heat fluxes, tidalmixing, and wind-forced variations in thermocline depth ofboth the Indian and Pacific oceans (as reviewed by Quet al. [2005]). In addition to the need to resolve the narrowpassages between the numerous islands of the CT, themajor factors that should be addressed to accurately simu-late ocean conditions in the CT are the wind field [Godfrey,1996], the tides [Ffield and Gordon, 1996; Koch-Larrouyet al., 2007], and a proper treatment of boundary conditionsthat respects the mean flow currents from the Pacific to theIndian Ocean [Sprintall et al., 2009].

[7] Several high-resolution modeling studies have beenconducted in this region, but most have targeted particularsubregions and/or specific processes. Robertson and Ffield[2008] used a regional high-resolution ocean model to sim-ulate the barotropic and baroclinic tides in the Indonesiaseas and examine tide-induced mixing processes at the

Figure 1. Schematic of ocean circulation in the Coral Triangle region. The dashed orange line delineatesthe Coral Triangle following Veron et al. [2009]. Numbered passages are: (1) Makassar Strait, (2) Lifama-tola Strait, (3) Lombok Strait, (4) Ombai Strait, (5) Timor Passage, (6) Luzon Strait, (7) Karimata Strait,(8) Mindoro Strait, (9) Sibutu Strait, and (10) Torres Strait. Abbreviations are: NEC, North Equatorial Cur-rent; NECC, North Equatorial Countercurrent; SEC, South Equatorial Current; SECC, South EquatorialCountercurrent; ME, Mindinao Eddy; HE, Halmahera Eddy; and NGCC, New Guinea Coastal Current.

CASTRUCCIO ET AL.: MESOSCALE MODELING IN THE CORAL TRIANGLE

6124

Page 3: A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago

INSTANT mooring locations [Robertson, 2010] and theinteraction and transfer of energy among tidal constituents[Robertson, 2011]. Metzger et al. [2010] analyzed the path-way of ITF by using a global high-resolution model drivenby atmospheric forcing. Both of these high-resolution stud-ies described the Indonesian seas using a single forcing,either tidal or atmospheric. Kartadikaria et al. [2011]implemented a regional high-resolution ocean model forthe Indonesian seas that combined the tidal and atmos-pheric forcings. Han et al. [2009] used a regional high-resolution model to characterize seasonal surface oceancirculation and dynamics in the Philippine Archipelagoregion during the Philippine Archipelago Experiment(PhilEx). Several studies have used high-resolution model-ing in studies of the South China Sea [e.g., Metzger andHurlburt, 1996; Qu et al., 2005; Wang et al., 2009; Quet al., 2009; Xie et al., 2011], and another [Melet et al.,2010] used high-resolution modeling to describe thermo-cline circulation pathways in a the Solomon Sea.

[8] Here we present results from a Regional Ocean Mod-eling System (ROMS) configuration for the entire CT (CT-ROMS), including both the ITF and the SCSTF, with a5 km horizontal resolution. This resolution is appropriatefor capturing the complex ocean dynamic of the regionwhile still permitting the long integrations needed forstudying the dynamical response to rising temperature andits regional variability. The large-scale forcing is setthrough accurate ocean boundary conditions from an assim-ilation product; CT-ROMS also incorporates the two pri-mary forcings for the region at the local scale: tidal andatmospheric. The model explicitly solves the tides so thatthe mixing processes are not artificially parameterized (asin the study by Koch-Larrouy et al. [2007]).

[9] In this paper, we assess CT-ROMS model dynamicsagainst the recently observed ocean dynamics in the areaand use the resulting fields to quantify mesoscale variability.A description of the model is provided (section 2), followedby a description of important modeled circulation pathwaysand a comparison with the observed transports (section 3).We pay particular attention to the validation of the simulatedtides (section 4), and to the comparison of simulated SSTswith satellite derived and in situ data (section 5). Finally, weexamine the surface mixing and the mesoscale turbulencesimulated by the model, through application of the finite-size Lyapunov exponent (FSLE) method (section 6).

2. Model Description

2.1. Regional Ocean Modeling System (ROMS)

[10] The numerical simulations were performed with theRegional Ocean Modeling System (ROMS; http://www.myroms.org). ROMS is widely used for applicationsfrom the basin to coastal and estuarine scales [e.g., Curch-itser et al., 2005; Danielson et al., 2011; Haidvogel et al.,2000; Lemarie et al., 2012; Marchesiello et al., 2003, 2009;Warner et al., 2005a]. ROMS solves the incompressible,hydrostatic Boussinesq primitive equations in finite differ-ence form with a free-surface and within an Arakawa C-gridcurvilinear horizontal coordinate system and a generalizedstretched terrain-following vertical coordinate system [Haid-vogel et al., 2008]. Shchepetkin and McWilliams [2003,2005, 2009] describe in detail the algorithms that comprise

the ROMS computational kernel. ROMS makes use of veryaccurate and efficient physical and numerical algorithms. Inparticular, it utilizes consistent temporal averaging of thebarotropic mode to guarantee both exact conservation andconstancy preservation properties for tracers and yieldsmore accurate resolved barotropic processes while prevent-ing aliasing of unresolved barotropic signals into the slowbaroclinic motions. Accuracy of the mode splitting is furtherenhanced due to redefined barotropic pressure-gradientterms to account for the local variations in the density field(i.e., the pressure-gradient truncation error that has previ-ously plagued terrain-following coordinate models is greatlyreduced) while maintaining the computational efficiency ofa split model. ROMS has various options for advectionschemes: second-order and forth-order centered differences;and third-order upstream biased. The vertical mixingschemes include several subgrid-scale parameterizations.The horizontal mixing of momentum and tracers can bealong vertical levels, geopotential surfaces, or isopycnalsurfaces. The vertical mixing parameterization in ROMS canbe either by the local Generic Length Scale (GLS) closurescheme by Umlauf and Burchard [2003], or the nonlocal, K-profile boundary layer formulation by Large et al. [1994].

2.2. CT-ROMS Configuration

[11] The CT-ROMS model domain spans the regionfrom about 95�E to 170�E, and 25�S to 25�N (Figure 2).This is about 8350 km 3 5500 km, encompassing the entireCT as defined by Veron et al. [2009]. The horizontal gridresolution is 5 km on average, resulting in a 1280 3 640points grid. CT-ROMS uses 50 vertical levels in terrain-following sigma-coordinates, weighted toward the surfacein order to better resolve the mixed layer. The vertical coor-dinate transformations and stretching function of Shchepet-kin and McWilliams [2009] are used, so that the upperlayers are closer to geopotential surfaces, which reducesspurious advection in the ocean surface mixed layer as wellas the errors in the pressure gradient. The model grid isslightly rotated relative to constant longitude/latitude linesin order to encompass the CT region while maximizing thewet points ratio (i.e., the number of sea points over the totalnumber of points in the grid) and minimizing computationover land points. The model bathymetry was interpolatedfrom the global SRTM30_PLUS product, which has a rawbathymetric resolution of 30 s or roughly 1 km (Figure 2).In order to reduce the intrinsic error in the horizontal pres-sure gradient associated with sigma-coordinates, the modelbathymetry was smoothed using a two-step method com-bining a Shapiro filter [Shapiro, 1975] and the direct itera-tive technique proposed by Martinho and Batteen [2006].

[12] CT-ROMS was integrated from June 2003 to the endof 2006, to coincide with the observational period of theInternational Nusantara Stratification and Transport Program(INSTANT) [Gordon et al., 2010]. The simulation was ini-tialized using an initial condition interpolated on the CT-ROMS grid from the Simple Ocean Data Assimilation(SODA) [Carton et al., 2000a, 2000b] retrospective analysis.A short 6 month period was used to spin up the model whichis long enough for the upper ocean, our region of interest, toreach a dynamical balance. All analyses were performed overthe 2004–2006 period, coincident with the INSTANT obser-vations. INSTANT’s primary objective was to measure the

CASTRUCCIO ET AL.: MESOSCALE MODELING IN THE CORAL TRIANGLE

6125

Page 4: A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago

Fig

ure

2.B

athy

met

ry(i

nm

eter

s)an

dla

nd-s

eam

ask

used

byC

T-R

OM

S.

The

four

inse

tssh

owbl

owup

sov

erth

eke

yIT

Fpa

ssag

esm

onit

ored

dur-

ing

the

INS

TA

NT

prog

ram

.The

sam

eco

lorb

aris

used

for

all

plot

s.

CASTRUCCIO ET AL.: MESOSCALE MODELING IN THE CORAL TRIANGLE

6126

Page 5: A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago

Indonesian Throughflow (ITF) simultaneously across multi-ple passages from the Pacific inflow at Makassar Strait andLifamatola Passage to the Indian Ocean export channels ofTimor, Ombai, and Lombok (Figure 1) and thus capture theITF seasonal and annual cycle over a range of ENSO phases.

[13] SODA was also used to provide the model openboundary temperature, salinity, and velocity using a hybridof nudging and radiation approaches [Marchesiello et al.,2001]. The tidal forcing is naturally implemented at theboundary by providing the tidal elevation and barotropicflows from the global model of ocean tides TPXO 7.2,which best fits in a least-squares sense the Laplace TidalEquations and along track averaged data from TOPEX/Pos-eidon and Jason [Egbert and Erofeeva, 2002]. Because themodel domain is large, an astronomical tide-generatingpotential is also added as a body force in the momentumequation to ensure correct tidal phasing.

[14] The surface forcing for the CT-ROMS model wasderived from the Modern Era-Retrospective Analysis forResearch and Applications (MERRA) reanalysis [Rieneckeret al., 2011]. MERRA is a NASA reanalysis for the satelliteera using the Goddard Earth Observing System Data Assimi-lation System Version 5 (GEOS-5). MERRA provides anextensive suite of global atmospheric fields with high tempo-ral (hourly) and spatial (1=2� latitude 3 2=3� longitude) resolu-tions. Air temperatures, sea level pressure and specifichumidity, daily short-wave and downwelling long-waveradiation, and precipitation were used to compute air-seaheat and momentum fluxes using bulk formulae [Large andYeager, 2009]. River discharge was implemented as a freshwater flux using the global river flow and continental dis-charges estimated by Dai and Trenberth [2002].

[15] The vertical mixing in the interior layers was calcu-lated with the local generic two-equation turbulence clo-sure scheme (GLS) [Warner et al., 2005b]. The bottomstress was empirically parameterized with a spatially vari-able linear coefficient of friction based on total water col-umn depth. Sea surface salinity (SSS) was weakly restoredtoward monthly observed SSS in order to prevent modeldrift while allowing the model’s own variability in the sur-face salinity and deep circulation to develop.

[16] Satellite-derived ocean color data have revealed thespatially complex structure of near-surface biooptical prop-erties in open-ocean frontal areas and in coastal waters[Ackleson, 2001]. In order to simulate the spatial variationin light attenuation, we implemented a spatially varyingwater type in ROMS, following the five ocean water typesintroduced by Jerlov [1976]. Water types were approxi-mated using a simple depth relationship, with deep openocean cells having the clearest water, i.e., water Type I ofJerlov [1976], and shallow/coastal grid cells having theleast transparent, i.e., water Type III. This is important, par-ticularly in the tropics where the shortwave heat flux islarge, as a decrease (increase) of the attenuation depth willincrease (decrease) the static stability in the surface watercolumn and thus affect the surface layer temperature.

3. Circulation Pathways in CT-ROMS

[17] The complex circulation and transport pathways ofthis region have been described in previous observations[e.g., Gordon et al., 2010; Qu et al., 2009] and modeling

studies [e.g., Metzger et al., 2010; Hurlburt et al., 2011; Duand Qu, 2010]. The circulation patterns simulated in CT-ROMS are in good agreement with these previous studies,and the ability of the model to reproduce the most importantfeatures of the circulation is briefly described here.

3.1. Major Circulation Patterns

[18] Both the ITF and SCSTF are clearly identified fea-tures of the near-surface current, which we define as thedepth-averaged circulation between the surface and 250 mdepth (Figure 3). In the North Equatorial Pacific, the NorthEquatorial Current (NEC) flows westward until it reachesthe Philippines Archipelago where it bifurcates into thenorthward-flowing Kuroshio Current and the southward-flowing Mindanao Current (MC) at around 13�N, consistentwith previous findings [Lukas et al., 1996; Qu and Lukas,2003].

[19] The MC flows southward along Mindanao Islanduntil around 5�N, where most of the flow turns eastward toform the North Equatorial Counter Current (NECC). Thestrong recirculation east of Mindanao Island is the Minda-nao Eddy (ME). A significant portion of the MC water alsoleaks into the Celebes Sea over the Sangihe Ridge. Whilesome of this water recirculates and reexits the Celebes Sea,a large portion continues southward through the MakassarStrait to form the main branch of the ITF. At the DewakangSill in the southern Makassar Strait, it splits into a westbranch directly exiting Flores Sea via Lombok Strait andan east branch flowing along the north side of Lesser SundaIsland into the Banda Sea. Also obvious in Figure 3 is theinflow through the Halmahera Sea into the Banda Seabranching from the New Guinea Coastal Current (NGCC),a northern branch of the South Equatorial Current (SEC).Water passes from the Banda Sea into the Indian Ocean viatwo main passages, Ombai Strait and Timor Passage.

[20] Most of the northward flowing Kuroshio Currentbypasses Luzon Strait and continues along the continentalslope, east of China. A small fraction intrudes the SCS,most of which flows southward along the continental slopeand crosses the entire basin, ultimately flowing throughKarimata Strait into the Java Sea or through Mindoro Straitinto the Sulu Sea and eventually reaching the Celebes Seato the south mostly through the shallow Sibutu Passage.

3.2. Mean Transports

[21] The simulated mean volume transport in Sverdrup(1 Sv 5 106 m3 s21) over the 2004–2006 period (Figure 4)is calculated for the full water column from sidewall tosidewall with negative transport defined as from the PacificOcean to the Indian Ocean. INSTANT observational esti-mates of the volume transport, where available, are pro-vided for comparison.

[22] Following Sprintall et al. [2009], we define the totalITF transport as the sum of the three main outflow passagesof Lombok Strait, Ombai Strait, and Timor Passage. Thesimulated total ITF transport in CT-ROMS is 217.5 Svover the 2004–2006 INSTANT time frame, which is ingood agreement with the best estimate based on theINSTANT moorings data: 215 Sv with values rangingfrom 210.7 to 218.7 Sv depending upon how the observa-tions were extrapolated to the surface and sidewalls [Sprin-tall et al., 2009]. The simulated Lombok Strait transport of

CASTRUCCIO ET AL.: MESOSCALE MODELING IN THE CORAL TRIANGLE

6127

Page 6: A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago

22.6 Sv (15% of the total ITF transport) is identical to theobservational estimate. However, the simulated transportsthrough Ombai Strait (29.8 Sv; 56% of the total ITF) andTimor Passage (25.1 Sv; 29% of the total ITF) differ fromINSTANT observations of 24.9 Sv (33%) and of 27.5 Sv

(50%), respectively, although the combined flow throughthese two passages agree well with observations. A detailedverification of the model bathymetry in the area, with par-ticular care regarding the sill depths upstream and down-stream of the two passages, did not reveal a bathymetric

Figure 3. Mean near-surface currents (depth averaged between the surface and 250 m depth, in ms21)simulated by CT ROMS over the 2004–2006 INSTANT period.

Figure 4. Total mean volume transport (in Sverdrup, 1 Sv 5 106 m3 s21) simulated by CT-ROMS(value on the left) and observed by INSTANT (value on the right) over 2004–2006. Negative transport istoward the Indian Ocean. Simulated transport is calculated for the full water column from sidewall tosidewall. The red lines indicate the transects used to diagnose the transport in the model.

CASTRUCCIO ET AL.: MESOSCALE MODELING IN THE CORAL TRIANGLE

6128

Page 7: A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago

cause for this discrepancy. Interestingly, Kartadikaria et al.[2011] report a similar issue in their atmospheric-tidalforced model with positive meridional velocity at depthdecreasing the transport toward the Indian Ocean for theTimor Passage, and blamed a shear stress issue at the bot-tom of the water column to explain this discrepancy. Arbicet al. [2010] stressed the importance of an additional bot-tom dissipative term for tidal models to parameterize topo-graphic wave drag. Such parameterization is currently notimplemented in ROMS. Instead, the bottom stress wasempirically parameterized with a spatially variable linearcoefficient of friction based on total water column depth. Itwill be interesting to incorporate a parameterized topo-graphic wave drag similar to the one described by Arbicet al. [2010] into ROMS to see if the transport mismatchbetween CT-ROMS and the INSTANT estimate for TimorPassage is indeed imputable to a shear stress issue at thebottom of the water column.

[23] Transports through two other major passages weremonitored during the INSTANT program: Makassar Straitand Lifamatola Passage. In Makassar Strait, Gordon et al.[2008] reported a transport of 211.6 Sv, which accountsfor 77% of the total ITF. A more recent estimate, using thesame velocity data from the INSTANT moorings but amore accurate bathymetry suggests a net transport of212.7 Sv at Makassar for the 2004–2006 time frame [Sus-anto et al., 2012]. CT-ROMS simulated the Makassar Straittransport at 213.1 Sv (75% of total ITF), which agreeswell with these observations.

[24] Transport through Lifamatola Passage was esti-mated at 21.1 Sv [van Aken et al., 2009], while the CT-ROMS simulated transport was 21.5 Sv. The strongbottom-intensified overflow in the Seram Sea is well repre-sented in the model with velocity close to the bottom of the�2000 m deep sill regularly surpassing 1 m s21. Note thatthe flow in the upper layer is northward (shown in Figure3).

[25] Figure 4 also shows the transport simulated by CT-ROMS for some important straits not monitored during theINSTANT program. For example, the model shows a sig-nificant transport of 23.2 Sv through the Halmahera Sea.This flow is surface intensified and carries water fromSouth Pacific origin into the Indonesian seas (Figure 3).

[26] An alternate pathway to the ITF for exchange oftropical waters between the Pacific and Indian oceans isTorres Strait connecting the Gulf of Carpentaria to the con-tinental shelf of the Great Barrier Reef. Torres Strait isvery shallow (less than 10 m deep) with the presence ofnumerous islands and reefs. Based on in situ observations,Wolanski et al. [1988] found strong tidal flow but veryweak mean flow through Torres Strait. The transport simu-lated by CT-ROMS is 20.3 Sv, less than 2% of the ITFtransport.

[27] The South China Sea throughflow (SCSTF) isanother important pathway that involves inflow of cold,salty water through the deep Luzon Strait and outflow ofwarm, fresh water through the shallow Karimata and Mind-oro Straits [e.g., Qu et al., 2005; Fang et al., 2005; Yuet al., 2007]. Many descriptive and quantitative studieshave focused on the westward intrusion of the WesternPacific water into the SCS through the Luzon Strait [e.g.,see Fang et al., 2005; Qu et al., 2009 for a comprehensive

overview of the previous published works for the area].Applying Godfrey’s [1989] Island Rule to the Luzon Straityielded an annual mean transport estimate (from the Pacificinto SCS) of approximately 4.2 Sv [Qu et al., 2000]. Otherobservational and modeling studies arrived at a range ofestimates from 0.5 to 10 Sv [e.g., Metzger and Hurlburt,1996; Qu et al., 2000; Fang et al., 2003]. The mean valueof all estimates is 4.4 Sv. The CT-ROMS simulated trans-port is close to this average with a transport estimated at5.3 Sv. About 1.1 Sv of this Pacific water exits toward thenorth through Taiwan Strait and the remainder flowstoward the south through Karimata Strait (0.7 Sv) andMindoro Passage (3.5 Sv) and eventually reaches theIndian Ocean. The seasonal variability for all three straits isvery large. Karimata Strait transport, for example, rangedfrom approximately 2 Sv southward during the Northeastmonsoon (November to March) to 0.5 Sv northward duringthe Southwest monsoon (May to September) with a veryabrupt transition from one regime to another.

3.3. Makassar Strait Variability

[28] The model confirms observations that MakassarStrait is the primary inflow passage for Pacific water, carry-ing about three quarters of the total ITF transport. TwoINSTANT moorings were deployed for nearly 3 years oneach side of the Labani Channel, a constriction in MakassarStrait (MAK-west: 2�51.900 S, 118�27.300 E; MAK-east :2�51.500 S, 118�37.700 E). For comparison, two virtualmoorings were deployed in CT-ROMS to record hourlymodel output at the same MAK-west and MAK-east loca-tions (Figure 5). The results are similar to plots based onthe INSTANT observations of Gordon et al. [2008] andSusanto et al. [2012] (not shown). The time-series sectionof the along-axis current (Figure 5, top) exhibits thermo-cline intensification in agreement with the observations,although the depth of the modeled maximum southwardcurrent is somewhat shallower than the 120 m depth in theobservations [Susanto et al., 2012]. The model accuratelycaptures the seasonal cycle with a deepening of the maxi-mum southward current during the northwest monsoon(February to April) and a shallowing and intensification ofthe maximum southward current during the southeast mon-soon (July to September). The signatures of semiannualKelvin waves below 200 m depth propagating from theIndian Ocean [Sprintall et al., 2000; Susanto et al., 2012]and weakening the southward flow are clearly visible inMay and October. This feature is absent in May 2006, inagreement with the observations [Gordon et al., 2008].During January to April 2006, weak La Ni~na condition pre-vailed and stands as a period of sustained throughflow, bothin the model and the observed time sections. The modelsimulated a short period of reversal flow during the north-west monsoon in the surface layer, also consistent with theobservations [Gordon et al., 2008]. During 2004, a seriesof surface flow reversals occurred throughout the year.

[29] The temperature time section (Figure 5, middle)shows a relatively weak variability over the 2004–2006period. This period is free of major ENSO events and theweak variability is in agreement with the INSTANT obser-vations [Susanto et al., 2012]. As seen in the observed timesection [Susanto et al., 2012], the model simulates a deep-ening of the upper thermocline isotherms associated with

CASTRUCCIO ET AL.: MESOSCALE MODELING IN THE CORAL TRIANGLE

6129

Page 8: A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago

the weak La Ni~na condition in January to April 2006.Unlike the temperature, the salinity time section shows sig-nificant seasonal variability. The low salinity values in thesurface layer during the northwest monsoon associatedwith reversal flow (Figure 5, top) are consistent with theidea that the intrusion of fresh buoyant SCS water intoMakassar Strait weakens the ITF surface flow and forcesthe ITF to a deeper level (near the thermocline) [Qu et al.,2009; Gordon et al., 2012].

[30] Figure 6 shows the results of a power-spectra analy-sis of the observed and simulated along-channel velocitytime series at MAK-west for a point within the thermoclineat 150 m depth and for a deeper point at 750 m depth. Atlow frequencies, the model is in good agreement with theobservations both within the thermocline and at depth. Thespectral analysis shows that velocity fluctuations peak atthe dominant tidal frequencies, the diurnal (O1 and K1) andsemidiurnal (M2 and S2), with the same intensity in themodel and the observations (Figure 6). Additional peaksare found at the higher harmonics of the diurnal and semi-diurnal frequencies. The high frequencies contain lessenergy in the model than the observations. This is particu-larly true within the thermocline. The shape of the spec-trum based on the observed thermocline currents, however,suspiciously deviates from the characteristic 25/3 powerlaw and the expected cascade of energy to smaller scales in

the inertial subrange of three-dimensional isotropic turbu-lence. This high energy at high frequencies is likely theresult of the bobbing up and down of the mooring, as Rob-ertson and Ffield [2008] and Susanto et al. [2012] reportedsignificant mooring blowdown at MAK-west by ocean cur-rent and tides. Figure 6 confirms that tides are a dominantforcing in the area and that the tidal current is dominatedby the diurnal (O1 and K1) and semidiurnal (M2 and S2)frequencies.

4. Tides

[31] The ITF is not only critical in transferring mass,heat, and salt between the Pacific and Indian oceans, it isalso a region of strong water mass transformation. Severalstudies suggest that internal tides cause the intense mixingrequired for this transformation in the ITF region [Schiller,2004; Hatayama, 2004; Robertson and Ffield, 2005;Koch-Larrouy et al., 2007]. Tides are also a critical compo-nent of the local ocean dynamic, as they affect the flowthrough the straits by the generation of residual currents

Figure 6. Power spectral density of the along channel(170�) velocity for the model (blue solid line) and forobservations (red solid line) at (top) 150 m depth and (bot-tom) 750 m depth at the MAK_west location. Peaks at thedominant semidiurnal and diurnal component are wellidentified. The spectra are based on successive 180 daysubperiods during the 2004–2006 period with an overlap of90 days. The 95% confidence intervals are indicated by theshading in the corresponding color.

Figure 5. Makassar Strait (top) along-channel velocity(in ms21), (middle) temperature (in �C), and (bottom)salinity time sections from January 2004 to December 2006simulated by CT-ROMS. The quantities represent an aver-age of virtual moorings at MAK-west and MAK-east loca-tions deployed in CT-ROMS. A monthly low-pass filter hasbeen applied before contouring. In the top plot, negativevelocities denote flow toward 170� along the Labani Chan-nel axis.

CASTRUCCIO ET AL.: MESOSCALE MODELING IN THE CORAL TRIANGLE

6130

Page 9: A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago

and by the density-driven flows resulting from tidal mixing.Simulation of the tides in CT-ROMS is particularly chal-lenging because of the region’s complex bathymetry andbecause the interactions between the Pacific and Indianoceans tides within the Indonesian seas result in complexbarotropic and baroclinic tidal fields.

[32] The simulated tides in CT-ROMS were comparedwith the time-series observations of sea level and derivedconstant harmonic tidal components from University ofHawaii Sea Level Center (http://ilikai.soest.hawaii.edu/uhslc/datai.html) (Figure 7). For clarity, only a 1 month period ispresented for each station, but the results are consistentthroughout the 3 year simulation. The correlations betweenthe observed and simulated time series are at least 80% forall the stations and more than 90% for Bintulu, Cebu, Mala-kal, Manila, Pohnpei, Puerto Princesa, and Sandakan.

[33] Harmonic analysis was used to compare the ampli-tude and phase of four dominant tidal constituents, i.e., M2,K1, S2, and O1, from the observed data and CT-ROMS sim-ulations (Table 1 and Figure 8). The error of the simulatedM2 constituent was greater than that of the other compo-nents (RMS difference 5 17 cm; Table 1). The reason forthis is explained by several factors. M2 is the dominantcomponent in this region, small-scale coastal topography(seafloor slope, mouths of rivers and bays) can intensify thetide but are not resolved at the 5 km horizontal resolutionof the model and semidiurnal tides are amplified whereshelf resonance occurs, such as over the broad and shallowAustralian North West Shelf, a region of strong barotropicand baroclinic tides. Port Darwin, for example, experiencesextremely large tidal amplitudes with a maximum tidalrange of 7.8 m. This is attributed to the semidiurnal tideamplification when entering Darwin Harbor. Because the5 km resolution of CT-ROMS does not adequately resolveDarwin Harbor, the modeled M2 amplitude fails to capturethe observed tide for this location. If Port Darwin isremoved from the computation, the RMS differencereduces to 8 cm for M2. The RMS error in amplitude forcomponents K1, S2, and O1 are 5 cm, 3 cm, and 3 cm(excluding Port Darwin), respectively. The phasing of themodel tides is also in agreement with the observations withan RMS error of less than 20 degrees for three of the fourdominant components (the M2 RMS error is 25� ; Table 1).

[34] The horizontal distributions of semidiurnal (M2) anddiurnal (K1) components computed from CT-ROMS areshown in Figure 9. The primary semidiurnal response isdominated by the large M2 tide from the Indian Ocean,with amplitudes larger than 1 m off northwest Australia.This wave is delayed slightly as it passes into the Bandaand Flores seas, which are deep enough that high tideoccurs almost simultaneously throughout both basins. TheM2 tide from the Indian Ocean also leaks into the FloresSea through Lombok Strait. Another component of thesemidiurnal tides enters the Indonesian seas from thePacific Ocean. The Indian and Pacific M2 tides meet in thesouth of Makassar Strait and the Maluku and Halmaheraseas. The Indian Ocean semidiurnal tide also propagatesslowly westward from the Flores Sea across the Java Sea.

[35] In the SCS, the M2 tide propagates mainly from thePacific into the SCS. M2 tide entering through the LuzonStrait behaves as a decaying wave while propagating south-westward in the SCS. Its amplitude drops rapidly from 40

to 20 cm while passing through the Luzon Strait, which isassociated with strong tidal energy dissipation by the localtopography. However, the M2 tide is amplified in thecoastal regions with strong shoaling and narrowing effects,i.e., around the western and southern parts of the MalayPeninsula, south of the Indo-China Peninsula, west of Bor-neo, around Leizhou Peninsula and in Taiwan Strait.

[36] The K1 diurnal tide, in contrast to the semidiurnal,easily propagates from the Pacific to the Indian Oceanthrough the Celebes Sea and the Sulawesi and Halmaheraseas. From the Celebes Sea, the wave passes throughMakassar Strait and either propagates into the Java Seawhere it encounters another K1 component coming fromthe SCS, or propagates into the Banda Sea where it encoun-ters the Pacific K1 tide coming through the Sulawesi andHalmahera seas. From the Banda Sea, this diurnal compo-nent flows to the Indian Ocean.

[37] Similar to the conditions of the M2 tide, relativelyhigh amplitude of the K1 tide appears on the continentalshelf. But, unlike the M2 tide, the amplitude of K1 is mark-edly increased in the SCS basin (about 0.4 m) after propa-gating from the Pacific (about 0.2m) through the LuzonStrait. Since the SCS is separated from Pacific Ocean forc-ing by Luzon Strait, and given that the phase and amplitudeof the K1 tide in the SCS basin are nearly constant, theamplified K1 is likely caused by the Helmholtz resonanceinside the SCS [Zu et al., 2008]. The K1 tide continuessouthward into the Java Sea where it encounters another K1

component originating from the Pacific Ocean throughMakassar Strait. The two components intersect to form acomplicated system of large-amplitude, nearly amphi-dromic systems west of Borneo.

[38] The distribution pattern, magnitude and phase of thedominant semidiurnal and diurnal tide described above(Figures 8 and 9) are generally similar to those found byRay et al. [2005], Robertson and Ffield [2008], and Zuet al. [2008] and comprise a considerable range in tidaltypes. Diurnal, semidiurnal, and mixed tides are foundthroughout the Coral CT. In general, semidiurnal prevailingtides are found in the Indonesian seas, as well as in theadjoining Pacific and Indian oceans. Diurnal tides prevailin the SCS due to the Helmholtz resonance. Some regionssuch as Manila, Philippines, experience a purely diurnaltide. Purely semidiurnal tides can be found in MalaccaStrait or along the Northwest Australia shore, but mixedtide are more widely encountered ranging from mixed diur-nal dominant (e.g., Bintulu, Malaysia) to mixed semidiur-nal dominant (e.g., Guam).

5. Tidal Mixing and Upper Ocean Temperature

[39] Because of its deep cloud convection the CT is rec-ognized as a primary energy source for the entire climatesystem. This energy is mainly supplied as latent heating,released from the condensation of water vapor when cloudsand precipitation form due to cumulus convection. Therelationship between Sea Surface Temperature (SST) andconvective activity is highly sensitive and the local SST isof major importance to atmospheric state not only over theCT itself, but globally [Neale and Slingo, 2003].

[40] The mechanisms that generate and maintain SSTwithin the Indonesian seas are a consequence of the complex

CASTRUCCIO ET AL.: MESOSCALE MODELING IN THE CORAL TRIANGLE

6131

Page 10: A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago

Figure 7. Water level (in meters) simulated by CT-ROMS (blue solid line) and tide gauge data (redsolid line) for a 1 month period in 2004. The mean has been removed from all time series. The timeperiod is not the same for all location. r is the correlation between the observed and simulated timeseries. Both the model and the data are in GMT time zone format.

Table 1. Amplitude and Phase RMS Error for the Dominant Tidal Constituents M2, K1, O1, and S2 Between CT-ROMS and theObserved Water Levela

M2 K1 O1 S2

Amplitude Phase Amplitude Phase Amplitude Phase Amplitude Phase

RMS Error 17 cm (8 cm) 25� 5 cm 15� 3 cm 18� 8 cm (3 cm) 18�

aTwelve tide gauges are used. The value inside parentheses for M2 and S2 is the value obtained when the Darwin tide gauge is excluded.

CASTRUCCIO ET AL.: MESOSCALE MODELING IN THE CORAL TRIANGLE

6132

Page 11: A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago

topography and connectivity between the Pacific and Indianoceans. In addition to surface heat fluxes, intense tidal mix-ing of surface and thermocline waters, and variability in

thermocline depth driven remotely by winds over the Pacificand Indian oceans play a role in generating and maintainingSST [Qu et al., 2005]. Koch-Larrouy et al. [2007], among

Figure 9. (left) Amplitude (in meters) and (right) Greenwich phase (in degrees) for (top) the M2 tidalconstituent and(bottom) the K1 tidal constituent computed using CT-ROMS outputs. M2 and K1 are,respectively, the dominant semidiurnal and diurnal tidal constituents over the area.

Figure 8. Tidal chart for the four major tidal constituents (M2, K1, S2, and O1) driving the tidal forc-ing in the Indonesian seas. The blue arrows are for CT-ROMS and the red arrows are for the observa-tions. The length of the arrow is proportional to the tidal amplitude and the angle represents the tidalphase. The amplitude and phase of each tidal constituent were computed using a 6 month time series ofthe water level.

CASTRUCCIO ET AL.: MESOSCALE MODELING IN THE CORAL TRIANGLE

6133

Page 12: A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago

others, have illustrated the importance of an appropriate tidalmixing parameterization in order for an OGCM to accu-rately represent the water masses and the SST in the area. Inmost OGCMs without tidal mixing [Gordon and Susanto,1998; Schiller, 2004; Koch-Larrouy et al., 2007] the watermasses remain almost unchanged during their journeythrough the Indonesian seas. This feature does not agreewith the observations and various tidal mixing parameteriza-tions have been introduced [e.g., Koch-Larrouy et al., 2007]in order to artificially enhance the vertical mixing. In CT-ROMS, the barotropic tides are explicitly and accuratelyresolved as described in section 4. As a result, baroclinicityforms three dimensionally and internal tides are generated.The mixing associated with the tides in considerablyimproved and more naturally generated without having totune the model diffusivity to artificially enhance the mixinglike in the work by Koch-Larrouy et al. [2007].

[41] Strong vertical mixing affects the modeled watermasses as they flow along the ITF pathways (Figure 10). Inthe two entrance seas (North and South Pacific, red andgreen points), the TS diagrams display the typical structurefor Pacific water, with a salinity maximum associated withthe Pacific subtropical water and a salinity minimum asso-ciated with the Pacific intermediate water. These patternsare similar to TS diagrams based on the Levitus climatol-ogy [Levitus et al., 1998] (Figure 10). By the time the waterreaches the Banda Sea (magenta point in Figure 10), thetwo salinity maxima have been eroded, and the signature of

the Pacific subtropical water has been lost. Instead, thePacific incoming water has been transformed into Indone-sian Throughflow water, a unique water mass with analmost uniform salinity below the 20�C isotherm. Charac-terizing and quantifying those transformations using CT-ROMS is out of the scope of this paper but is clearly aninteresting topic for subsequent work.

[42] Figure 11 shows the mean sea surface temperature(MSST) and the standard deviation from both CT-ROMSand the Coral Reef Temperature Anomaly Database (CoR-TAD) [Selig et al., 2010]. CoRTAD was developed usingdata from the AVHRR Pathfinder Version 5.0 [Selig et al.,2010]. CoRTAD contains global, approximately 4 km reso-lution SST data on a weekly time scale from 1982 through2010 and related thermal stress metrics, developed specifi-cally for coral reef ecosystem applications. This is a widelyused data set for coral reef temperature analyses [e.g.,Pe~naflor et al., 2009; McLeod et al., 2010] and becauseCT-ROMS will be used in subsequent work to investigatecoral bleaching we evaluate CT-ROMS SST against theCoRTAD data.

[43] Both the MSST and its variability from theCT-ROMS simulations agree well with the satellite observa-tions (Figure 11). Overall, the RMS errors between CoR-TAD and the model SSTs are very small with typical valuesless than 1�C, although some areas close to shore havehigher RMS errors between 1 and 2�C (Figure 12). The biasbetween the model SST and the satellite SST (not shown) isas low as 0.4�C with a mean RMS error for the entiredomain of 0.7�C.

[44] As with the surrounding western Pacific and easternIndian oceans warm pools, the MSST is high and the SSTvariability is generally small. The relatively cold MSSTover the Flores and the Banda seas observed in CoRTAD isclearly identified in the model MSST illustrating the role oftidally enhanced vertical mixing of surface warm waterswith colder waters from below, which results in a meancooling of the SST over the Indonesian seas consistent withobservations. In the SCS, similar to the satellite observa-tions, the simulated MSST is colder on the north and westsides of the basin, reflecting the influence of the monsoonalwinds. During the winter, a cyclonic gyre circulation in thecentral basin with a strong southward western boundarycurrent is generated by northeasterly monsoon winds. Theresulting cold advection along the western boundary leadsto a cold tongue that is strongest from November to Febru-ary [Liu et al., 2011; Varikoden et al., 2010]. In summer,as southwesterly monsoon winds prevail, an anticyclonicocean eddy develops off the coast of South Vietnam, lead-ing to a cold filament located north of the wind speed maxi-mum and ocean upwelling off the coast, which helps tocool the ocean surface [Xie et al., 2003].

[45] Areas with high temporal variability in SST high-light in particular those areas with strong seasonal upwell-ing, such as the southern coasts of Java and Flores Island,the southern coast of West Papua, the eastern Banda Sea,and the south coast of Vietnam. The SST variability overmost of the Pacific Ocean (standard deviation of approxi-mately 0.4�C) is less in the simulations than in the satellite-based observations (standard deviation of approximately0.6–0.8�C). The reasons for this discrepancy are addressedbelow.

Figure 10. TS diagrams for (left) CT-ROMS and (right)Levitus climatology, in the Indonesian seas entrances (red,North Pacific, and green, South Pacific) and in the BandaSea (magenta). The color scale on each diagram representsthe depth (in meters). Isopycnic lines are overlaid.

CASTRUCCIO ET AL.: MESOSCALE MODELING IN THE CORAL TRIANGLE

6134

Page 13: A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago

[46] Figure 13 shows the 3 year time series of SST forTAO/TRITON moorings and for their model and CoRTADequivalents. The TAO/TRITON Array, designed for thestudy of year-to-year climate variations related to ENSO[McPhaden et al., 1998], includes 13 mooring within theCT-ROMS domain. We only show the time series for threeTAO/TRITON mooring locations but the results wereequivalent for all locations within the CT-ROMS domain.The modeled SSTs are in good agreement with the observa-tions, with no significant biases in the means, correlationsabove 0.8 and RMS differences smaller than 0.3�C (Figure13). CoRTAD SSTs appear to have a cold bias and greatervariability than the in situ TAO/TRITON observations.One possible explanation for CT-ROMS warmer SST withsmaller variability when compared to the CoRTAD dataacross the western equatorial Pacific area (Figure 11) couldbe cloud contamination in the CoRTAD data. The CT ischaracterized by high cloud cover and rainfall in response tothe ascending air over the western Pacific Warm Pool andthe Maritime Continent associated with the Walker Circula-tion [Bjerknes, 1969].

[47] The distribution of observations included in theCoRTAD data set (Figure 14) illustrates the high cloudcover over the area. The presence of clouds reduces thenumber of the valid observations that can be made, result-ing in large areas where observations are available for only50% of the time. The mean and standard deviation of SSTvalues for the 13 TAO/TRITON locations, compared tothose from CT-ROMS and CoRTAD, show a small but per-sistent cold bias in the CoRTAD SSTs, as well as greater

SST variability (Table 2). The mean SST for 12 locations(excluding the mooring at 0�N, 137�E where TAO/TRI-TON data are only available for a small portion of the2004–2006 period) is 29.56�C for the TAO/TRITON and29.32�C for CoRTAD. The variability is also systemati-cally larger for CoRTAD (0.66�C over the 12 locations)than for the in situ TAO/TRITON data (0.42�C). The coldbias (0.24�C) and larger variability for CoRTAD suggestthat cloud contamination of the satellite data is the sourcefor an overall negative temperature bias in AVHRR-derived SST [Reynolds et al., 2002]. CT-ROMS performswell when compared to TAO/TRITON with a mean SST of29.56�C and a standard deviation of 0.44�C over the 12locations (Table 2).

6. Mixing and Mesoscale Features

[48] The CT is characterized by intense mesoscale andsubmesoscale activity, such as eddies, fronts and shear cur-rents, which are clearly visible in both high-resolutionobservations of the ocean surface and in CT-ROMS SSTand SSH fields. The stirring of water masses with contrast-ing properties affects a large number of biophysical proc-esses, including mixing, lateral and vertical transport, thestructure of phytoplankton communities and the dispersalof larvae and pollutants. It is also well recognized that mes-oscale to submesoscale mixing plays a critical role in mod-ulating large-scale circulation [van Haren et al., 2004].

[49] We describe here the mesoscale and submesoscalecirculation of CT-ROMS. Eddy kinetic Energy (EKE) is

Figure 11. (left) Mean SST (in �C) and (right) standard deviation (in �C) over the 2004–2006 periodfor (top) CT-ROMS and the CoRTAD satellite data. Note that the standard deviation is saturated at 2�C.Values larger than 3�C, both in the model and in the observations, are found in the SCS along the coastof China with the larger value (>4�C) being found in the extreme north of the Gulf of Tonkin. Largevalue around and above 3�C are also found near the coast in the Carpentaria Bay.

CASTRUCCIO ET AL.: MESOSCALE MODELING IN THE CORAL TRIANGLE

6135

Page 14: A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago

often computed to characterize the mesoscale and eddy var-iability. While satellite altimetry has provided an uniquecontribution to the observation of eddy variability,satellite-derived EKE is not reliable within the 65� bandstraddling the equator as the geostrophic approximationbreaks down near the equator where the Coriolis force van-ishes. For that reason we use SODA, which assimilates

altimetry to evaluate the model ability to simulate meso-scale features (Figure 15). The distribution and intensity ofEKE simulated by CT-ROMS compare well with theassimilated products. Two active areas, respectively, asso-ciated with the unstable jet leaving the Vietnam coast andthe Kuroshio intrusion are found in the SCS. Large EKEvalues are also found in the Indonesian seas, with MakassarStrait, the Flores Sea, and the southern part of the BandaSea being the more active regions. The Halmahera Eddy aswell as the NGCC also show a strong signature in the EKEfield.

[50] In an attempt to further characterize the turbulentcirculation in the model, we have also computed theLagrangian coherent structures (LCSs) to quantify theturbulent circulation in the model. In a recent study,Harrison et al. [2013] show that the pathways of simu-lated Lagrangian particles are often organized into fila-ments between mesoscale eddies that correspond toattracting LCSs. These flow features are material curvesthat map filamentation and transport boundaries. Fluidparticles straddling an LCS will either diverge (‘‘repel-ling’’) or converge (‘‘attracting’’) in forward time [Hal-ler and Yuan, 2000]. LCSs thus delineate the boundarybetween dynamically distinct regions of the flow field,effectively allowing us to visualize the skeleton of tur-bulent transport [e.g., Haller, 2002; d’Ovidio et al.,2004; Shadden et al., 2005]. The spatial organization ofthese structures has a large impact on the coastal envi-ronment not only because they influence the dispersionof tracers in the water but also because by separatingdynamically distinct regions of the flow they can definefluid dynamical niches, which contribute to the structur-ing of marine ecosystems [d’Ovidio et al., 2010].

[51] Near LCSs, neighboring fluid parcels are straineddiffering amounts by the flow field. These differences instretching rates allow the detection of LCSs by the Lyapu-nov exponent k, which quantifies the exponential rates ofdivergence or convergence of initially infinitesimally closetrajectories averaged over infinite time (the relative

Figure 12. SST root mean square (RMS) errors (in �C) against CoRTAD satellite SST averaged overthe 2004–2006 period.

Figure 13. SST time series (in �C) at three TAO/TRI-TON locations (red dots on map) over the 2004–2006period for CT-ROMS (blue solid line), CoRTAD (green),and TAO/TRITON moorings (red solid line).

CASTRUCCIO ET AL.: MESOSCALE MODELING IN THE CORAL TRIANGLE

6136

Page 15: A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago

dispersion), such that the distance d0 between two nearbyparticles changes in time as

dt5d0ekt (1)

[52] In realistic situations the infinite time limit in thedefinition makes the Lyapunov exponent a quantity of lim-ited practical use. Instead, two of the most commonly usedapproximations are the finite-time Lyapunov exponents(FTLEs) [Haller and Yuan, 2000] and the finite-size Lya-punov exponents (FSLEs) [Aurell et al., 1997; Boffettaet al., 2001]. FSLEs appear to be better suited in a context

of regional modeling like CT-ROMS where boundaries atfinite distances strongly constrain the circulation. To mea-sure the FSLE at a point x, a reference particle is startedfrom x at time t, simultaneously with another particle at adistance d0 from x. The time t required to reach a separationdf is measured, so that the FSLE is defined as

k x; t; d0; df

� �5

1

tlog

df

d0(2)

[53] The FSLE is inversely proportional to the time atwhich two tracers reach a prescribed separation; large

Figure 14. CoRTAD observational ratio for each point on the gridded product for the 2004–2006period. The observational ratio is defined as the ratio between the number of weeks for which an SSTobservation is available over the total number of weeks for the period. Areas with small observationalratios identify regions with persistent and/or recurrent gaps in the CoRTAD satellite SST product.

Table 2. Mean and Standard Deviation (in �C) for the 2004–2006 SST Time Series for TAO/TRITON In Situ Observations,CT-ROMS, and CoRTAD Satellite Observationsa

137�E 147�E 156�E

Mean Standard Mean Standard Mean Standard

8�N TAO 29.03 0.58 29.19 0.51ROMS 28.95 0.55 29.19 0.37

CoRTAD 28.55 0.72 28.80 0.765�N TAO 29.29 0.50 29.40 0.36 29.47 0.31

ROMS 29.02 0.57 29.41 0.32 29.59 0.28CoRTAD 29.04 0.73 29.21 0.60 29.26 0.57

2�N TAO 29.55 0.38 29.73 0.35 29.76 0.35ROMS 29.20 0.58 29.74 0.39 29.94 0.36

CoRTAD 29.46 0.65 29.67 0.57 29.57 0.600�N TAO 29.61 0.28 29.80 0.37 29.89 0.33

ROMS 29.32 0.55 29.72 0.47 30.03 0.34CoRTAD 29.36 0.59 29.59 0.55 29.68 0.60

2�S TAO 29.86 0.34ROMS 30.04 0.40

CoRTAD 29.70 0.725�S TAO 29.78 0.50

ROMS 29.83 0.50CoRTAD 29.25 0.77

aTable shows all the TAO/TRITON mooring locations within the CT-ROMS domain.

CASTRUCCIO ET AL.: MESOSCALE MODELING IN THE CORAL TRIANGLE

6137

Page 16: A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago

values identify regions where the stretching induced bymesoscale and submesoscale activity is strong. Large FSLEvalues are typically organized in convoluted lines encir-cling submesoscale filaments. A line of local maxima ofFSLEs (more precisely, a ridge) can be used to predict pas-sive tracer fronts induced by horizontal advection and stir-ring. In practice, FSLEs provide a direct method forcharacterizing the mixing activity and the coherent struc-tures that control transport at a given scale and can be usedto analyze dispersion processes in the ocean [e.g., d’Ovidioet al., 2009; Hernandez-Carrasco et al., 2011].

[54] We estimated the Lyapunov exponents using thesurface velocity field from CT-ROMS. The FSLE dependson the choice of two length scales: the initial separation d0

and the final separation df. Here, we are interested in thespatial distribution of FSLEs and we naturally use d0 equalto the grid spacing to calculate the FSLE, i.e., 0.02�. Sincewe are interested in mesoscale and submesoscale struc-

tures, we set df 5 0.6�, i.e., a separation of about 66 km.The FSLE thus represents the inverse time scale for mixingfluid parcels over length scale df characteristic of the meso-scale and submesoscale structures. The spatial distributionof FSLE for a particular day of the simulation is shown inFigure 15. Typical values are in the order of 0.120.6days21, corresponding to mixing times for mesoscale dis-tances of 1.7210 days. Several prominent features emergefrom the FSLEs analysis of the CT-ROMS simulation (Fig-ure 15). First, the Indonesian seas and SCS emerge asregions of intense mesoscale and submesoscale activityassociated with strong horizontal mixing. Second, theFSLEs sharply identify well-known structures such as theHalmahera Eddy, which effectively isolates a large parcelof surface water from the surrounding ocean. Third, theregion includes large areas with weak mesoscale activity,such as the North West Australian Shelf and the Gulf ofCarpentaria.

Figure 15. Map of Eddy Kinetic Energy (EKE, in cm2 s22) for (top) SODA and (bottom) CT-ROMS.SODA, which assimilates altimetry is used for validation because the EKE derived from altimetry is notreliable within the 65� band straddling the equator, as the geostrophic approximation breaks down nearthe equator where the Coriolis force vanishes.

CASTRUCCIO ET AL.: MESOSCALE MODELING IN THE CORAL TRIANGLE

6138

Page 17: A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago

[55] As observed in previous works, maximum values ofthe distribution organize in lines or ridges that coincidewith repelling LCSs. Computing the FSLEs from trajectoryintegrations backward in time can also be used to approxi-mate attracting LCSs. Figure 16 shows both the attractingand repelling LCSs for a zoom over the Indonesian seas.Since LCSs are material lines that cannot be crossed by

particle trajectories, such lines strongly constrain andorganize the fluid motion in the surface layer and act astransport barriers for passive particles and tracers advectedby the large-scale flow [Shadden et al., 2005].

[56] We also examined the seasonal variability of theforward FSLEs for the 2004–2006 period (Figure 17). Thelarge FSLE values (short turbulence time scales) are the

Figure 16. FSLE spatial distribution computed forward in time for the entire CT-ROMS domain for14 September 2004, dx 5 d0 5 0.02, df5 0.6. Units for FSLEs are day21. Mesoscale structures and vorti-ces can be clearly seen. Small values of FSLEs (low-dispersion rates) are found in the core of eddies. Onthe contrary, large values of the FSLEs can be found in the outer part of eddies, where the stretching ofthe fluid parcels is particularly important, and in lines indicating robust transport barriers.

Figure 17. Lagrangian coherent structures (LCSs) for 14 September 2004. Blue and red lines representattracting and repelling LCSs, respectively. Repelling (attracting) LCSs coincide with ridges in theFSLE spatial distribution computed forward (backward) in time.

CASTRUCCIO ET AL.: MESOSCALE MODELING IN THE CORAL TRIANGLE

6139

Page 18: A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago

signature of the strong divergence induced by the flow pat-terns. Several regions stand out as having large FSLE val-ues. These include the Kuroshio Current (year round), andthe Flores and the western Banda seas. The mixing in theFlores Sea is particularly intense during the winter whenwinds of the Northeast Monsoon advect SCS surface watereastward into the Flores Sea through Karimata Strait, com-bining with the ITF flow and resulting in strong surfacecurrents. In Makassar Strait, winter is a period of weakmixing, corresponding to its seasonal minimum transport.During the Southwest Monsoon (summer), however, astrong transport barrier develops in Makassar Strait andpersists during the fall season. This transport barrierappears to be associated with Southwest Monsoon windforcing. West of the barrier, the main surface current flowswestward from the Java Sea through the Karimata Straitand eventually reaches the SCS. East of the barrier the ITFflows from Makassar Strait and turns eastward into theFlores and Banda seas.

[57] Other interesting patterns also emerge from Figure17. One example is found in the SCS. Monsoonal windsforce the upper ocean circulation in the SCS, alternatingfrom a strong basin-wide cyclonic gyre during the winterNortheast Monsoon to a double gyre during summer South-west Monsoon. The Ceram Sea, the eastern passage betweenthe Halmahera and Banda seas, appears to be relatively quietduring the winter months but shows large FSLE values andintense mesoscale activity during the summer months.

[58] Finally, persistently large FSLE values seem tocoincide with biogeographic provinces, illustrating theirpotential for identifying ‘‘invisible barriers’’ to larval dis-persal. The high FSLE region that extends across the mouth

of the Cenderawasih Bay (the large bay along the northshore of western Papua New Guinea) almost year aroundindicates that while particles to the north of the line will becarried by the NGCC and continue westward, particles onthe south side of this line will remain in the bay and recir-culate. Another example is the Halmahera Eddy, which iswell developed during the northern summer monsoon whenthe South Pacific water from the NGCC curves back intothe NECC. Interestingly, coral biogeographic studies alsosuggest that Cenderawasih Bay and the Halmahera regionare rather isolated systems [Carpenter et al., 2011].

[59] The time series of the forward FSLEs averaged overthe whole CT-ROMS domain for the 2004–2006 period(Figure 19) confirms the significant seasonal variabilityseen in Figure 18. In 2004, the two peaks of enhanced mix-ing coincide with the Australian monsoon (March) and theWestern North Pacific-East Asian Monsoon (August andSeptember). The absence of a well-defined peak in early2005 may be due to the weak Australian monsoon in thatyear, although the second peak in 2005 indicates the returnof vigorous mixing in phase with the Western NorthPacific-East Asian Monsoon. The second half of the time-series appears to reflect the effects of a weak and short-lived La Ni~na which lasted from autumn 2005 to spring2006, followed by an El Ni~no phase, which lasted from fall2006 until early 2007. Although our time series is too shortto draw any definitive conclusions, Figure 18 does suggestthat mixing strength in the CT has a seasonal and interan-nual variability that is linked to large-scale climate signalssuch as ENSO and/or the Asian-Australian monsoon sys-tem. We will pursue the investigation of those possiblelinkages in subsequent work using longer simulations.

Figure 18. Seasonal averages (for the 2004–2006 period) of the FSLEs for the entire CT-ROMSdomain; spring 5 MAM; summer 5 JJA, fall 5 SON, and winter 5 DJF. FSLE units are day21, largevalues are equivalent to the shortest turbulence time scales.

CASTRUCCIO ET AL.: MESOSCALE MODELING IN THE CORAL TRIANGLE

6140

Page 19: A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago

7. Conclusions

[60] We present a model specifically designed to tackleimportant questions regarding oceanographic circulationwithin the Coral Triangle and its relationship to climateand marine ecosystems. Coral Triangle ROMS (CT-ROMS) includes the South China Sea, Indonesia, and Phil-ippines, at a 5 km resolution. The primary forcings for theregion include the large-scale oceanic pressure gradientsthat set the background inter-ocean exchange, atmosphericdrivers, and importantly, explicit tides. The high-resolutionatmospheric forcing combined with the region’s intricatetopography, as well as interactions between the Pacific andIndian oceans tides within the Indonesian seas, result in acomplex barotropic and baroclinic ocean circulation. Themodel was integrated for the years 2004–2006, coincidentwith the INSTANT observational period, and results wereevaluated against INSTANT and other observations. Thesimulated total ITF transport of 217.5 Sv over the 2004–2006 period is in good agreement with the 215 6 3 Sv esti-mated ITF transport based on the INSTANT observations.The model is able to accurately reproduce the observed var-iability in Makassar Strait, the main ITF branch, with asimulated mean transport of 13.1 Sv compared to the mostrecent observational estimate of 12.7 Sv. The simulatedtides, with a correlation around 0.9, an RMS difference forthe four dominant tidal constituents (i.e., M2, K1, S2, andO1) of 5 cm for the amplitude and 19� for the phase, arealso in agreement with the observations over the modeldomain. The simulated SSTs are in agreement with boththe CoRTAD satellite and the in situ TAO-TRITON moor-ing SSTs. The mean bias is as low as 0.4�C with a meanRMS error for the entire domain of 0.7�C. When comparedto the in situ TAO-TRITON mooring SSTs, the modelshows a correlation of 0.8 or better, with no significantbiases in the means (bias of 0.03�C) and RMS differencessmaller than 0.3�C.

[61] By explicitly solving the tides, CT-ROMS is able togenerate the mixing due to vigorous internal tides, whichwere observed in both the simulated velocity and the seasurface height fields (not shown). Internal tides are gener-ated through interactions of the barotropic tides with thelocal topography consisting of a complex array of passageslinking interconnected shelves, deep basins, shallow anddeep controlling sills and submerged ridges, which providea sea link between the two oceans. Vertical tidallyenhanced mixing erodes the Pacific subtropical water salin-

ity maximum as seen in the TS diagram and acts to transfercold water from below into the thermocline and the surface.The relatively cold MSST over the Flores and the Bandaseas observed in both the CoRTAD and the modeled MSSTis the signature of this upwelled water. These improve-ments in simulating ITF circulation encourage using themodel to further characterize and quantify tidal mixing inthe region, although such analyses are beyond the scope ofthis paper.

[62] The comparison of simulated SSTs with those fromthe TAO-TRITON mooring and satellite-based SSTs illus-trates the skills of the model as equal to or better than thesatellite product in the region, which is commonly ham-pered by dense cloud cover.

[63] To our knowledge, this is the first modeling study tocharacterize the rich mesoscale eddy activity in the regionusing Finite-size Lyapunov exponents (FLSEs). FSLEswere computed using the modeled velocity fields and usinga final separation of df 5 0.6� in order to identify the coher-ent structures that control transport and stirring at scalescharacteristic of the mesoscale and submesoscale struc-tures. Spatial structures ranging from the small scales to theones typical of mesoscale vortices were sharply identifiedin the modeled flow. Our analysis shows that the CT is aregion of intense mixing that displays strong seasonal andinterannual variability. The seasonal average of the spatialdistribution of FSLEs highlighted the regional differencesin mixing activity as well as robust transport barriers in thesimulated flow. The seasonal transport barrier in MakassarStrait is one such example of a transport barrier visiblethrough FSLE analysis that is not easily identified in otherfields such as SST or SSH. A time series of the spatiallyaverage FSLEs allows quantifying the mesoscale activityand stirring in the CT at a given temporal window.Although the 3 year time series is too short to draw anydefinitive conclusions, it clearly suggests possible linkagesto large-scale climate signals such as ENSO and/or theAsian-Australian monsoon.

[64] The repelling and attracting LCSs, which were iden-tified using ridges in the FSLEs integrated forward andbackward in time, respectively, allow us to visualize theskeleton of the turbulent transport. Oceanic mesoscale andsubmesoscale structures play a critical role in modulatinglarge-scale circulation [van Haren et al., 2004]. Submeso-scale filaments have also been shown to have a structuringrole on marine ecosystems [e.g., Bakun, 1996; Rossi et al.,2008]. For example, d’Ovidio et al. [2009] found that thephytoplanktonic landscape is organized in submesoscalepatches (10–100 km) of dominant types separated by physi-cal fronts induced by horizontal stirring. A recent study byHarrison et al. [2013] also suggested that LCSs play animportant role in pelagic transport of marine larvae andshow that the pathways of the simulated larvae are oftenorganized into filaments found between mesoscale eddies.While particle-tracking experiments are beyond the scopeof the present paper, the FSLE analysis illustrates its suit-ability for research on coral larval transport and connectiv-ity in the CT.

[65] The CT supports the highest marine biodiversityknown, which is undoubtedly maintained by its complexoceanographic, topographic, and climatic complexity [Car-penter et al., 2011]. With such complexity, it is unlikely

Figure 19. Time series of the FSLEs spatial average forthe entire CT-ROMS domain over the 2004–2006 period.Unit is day21.

CASTRUCCIO ET AL.: MESOSCALE MODELING IN THE CORAL TRIANGLE

6141

Page 20: A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago

that the vulnerability of coral reefs and other marine eco-systems to climate change will be spatially uniform acrossthe CT. This region is now the focus of multiple conserva-tion efforts, in particular the Coral Triangle Initiative onCoral Reefs, Fisheries, and Food Security [Coral TriangleSecretariat, 2009], which deals with the multiple threats tocoastal and marine ecosystems and the people that dependon them.

[66] The ability of CT-ROMS to simulate oceanographiccirculation and SSTs will improve our ability to understandhow patterns of coral bleaching are linked to dominantmodes of variability, such as ENSO and the IOD, the inter-play of the ITF and SCSTF, as well as climate change. Animportant component of the Coral Triangle Initiative is theestablishment of Marine Protected Area (MPA) networksthat plan for resilience to climate change through applica-tion of biophysical principles [Fernandes et al., 2012;McLeod et al., 2009, 2012]. Several features of CT-ROMSrender it particularly useful to such designs such as theaccurate simulation of SSTs and the ability to couple themodel with ecosystem models. The LCSs and FSLEs anal-yses also illustrate the ability of CT-ROMS to resolve mes-oscale and submesoscale turbulence patterns that areknown to have large impacts on marine larval dispersal[Harrison et al., 2013] and show that these patterns varyboth temporally and spatially across the CT.

[67] Acknowledgments. This work was funded by a National ScienceFoundation grants OCE-1233430 and OCE-1234674. Additional supportwas provided by Rutgers University and by the National Center for Atmos-pheric Research, which is sponsored by NSF. Numerical integrations wereperformed with computational resources granted through the Janus super-computer (NSF-MRI grant CNS-0821794 and the University of ColoradoBoulder, as a joint effort of the University of Colorado Boulder, the Uni-versity of Colorado Denver and the National Center for AtmosphericResearch); Rutgers University; and the Texas Advanced ComputingCenter.

ReferencesAckleson, S. (2001), Ocean optics research at the start of the 21st Century,

Oceanography, 14(3), 5–8.Arbic, B. K., J. A. Wallcraft, and E. J. Metzger (2010), Concurrent simula-

tion of the eddying general circulation and tides in a global ocean model,Ocean Modell., 32(3–4), 175–187, doi:10.1016/j.ocemod.2010.01.007.

Aurell, E., G. Boffetta, A. Crisanti, G. Paladin, and A. Vulpiani (1997),Predictability in the large: An extension of the concept of Lyapunovexponent, J. Phys. A Math. Gen., 30(1), 1–26, doi:10.1088/0305–4470/30/1/003.

Bakun, A. (1996), Patterns in the Ocean: Oceanic Processes and MarinePopulation Dynamics, Univ. of Calif. Sea Grant, San Diego, Calif., incooperation with Centro de Investigaciones Biologicas de Noroeste, LaPaz, Baja California Sur, Mexico.

Bjerknes, J. (1969), Atmospheric teleconnections from the equatorialPacific, Mon. Weather Rev., 97, 163–172.

Boffetta, G., G. Lacorata, G. Radaelli, and A. Vulpiani (2001), Detectingbarriers to transport: A review of different techniques, Physica D,159(1–2), 58–70, doi:10.1016/s0167–2789(01)00330-x.

Carpenter, K. E., et al. (2011), Comparative phylogeography of the CoralTriangle and implications for marine management, J. Mar. Biol., 2011,1–14, doi:10.1155/2011/396982.

Carton, J. A., G. Chepurin, and X. H. Cao (2000a), A Simple Ocean DataAssimilation analysis of the global upper ocean 1950–95. Part II :Results, J. Phys. Oceanogr., 30(2), 311–326, doi:10.1175/1520-0485(2000)030<0311:asodaa>2.0.co;2.

Carton, J. A., G. Chepurin, X. H. Cao, and B. Giese (2000b), A SimpleOcean Data Assimilation analysis of the global upper ocean 1950–95.Part I: Methodology, J. Phys. Oceanogr., 30(2), 294–309.

Coral Triangle Secretariat (2009), Coral Triangle Initiative on Coral Reefs,Fisheries and Food Security (CTI-CFF) Regional Plan of Action, CoralTriangle Secr., Jakarta.

Curchitser, E. N., D. B. Haidvogel, A. J. Hermann, E. L. Dobbins, T. M.Powell, and A. Kaplan (2005), Multi-scale modeling of the North PacificOcean: Assessment and analysis of simulated basin-scale variability(1996–2003), J. Geophys. Res., 110, C11021, doi:10.1029/2005JC002902.

d’Ovidio, F., V. Fernandez, E. Hernandez-Garcia, and C. Lopez (2004),Mixing structures in the Mediterranean Sea from finite-size Lyapunovexponents, Geophys. Res. Lett., 31(17), L17203, doi:10.1029/2004GL020328.

d’Ovidio, F., J. Isern-Fontanet, C. Lopez, E. Hernandez-Garcia, and E.Garcia-Ladon (2009), Comparison between Eulerian diagnostics andfinite-size Lyapunov exponents computed from altimetry in the Algerianbasin, Deep Sea Res., Part I, 56(1), 15–31, doi:10.1016/j.dsr.2008.07.014.

d’Ovidio, F., S. De Monte, S. Alvain, Y. Dandonneau, and M. Levy (2010),Fluid dynamical niches of phytoplankton types, Proc. Natl. Acad. Sci. U.S. A., 107(43), 18,366–18,370, doi:10.1073/pnas.1004620107.

Dai, A., and K. E. Trenberth (2002), Estimates of freshwater discharge fromcontinents: Latitudinal and seasonal variations, J. Hydrometeorol., 3(6),660–687, doi:10.1175/1525–7541(2002)003<0660:eofdfc>2.0.co;2.

Danielson, S., E. Curchitser, K. Hedstrom, T. Weingartner, and P. Stabeno(2011), On ocean and sea ice modes of variability in the Bering Sea,J. Geophys. Res., 116, C12034, doi:10.1029/2011JC007389.

Du, Y., and T. D. Qu (2010), Three inflow pathways of the Indonesianthroughflow as seen from the simple ocean data assimilation, Dyn.Atmos. Oceans, 50(2), 233–256, doi:10.1016/j.dynatmoce.2010.04.001.

Egbert, G. D., and S. Y. Erofeeva (2002), Efficient inverse modeling of bar-otropic ocean tides, J. Atmos. Oceanic Technol., 19(2), 183–204.

Fang, G. H., Z. X. Wei, B. H. Choi, H. Wang, Y. Fang, and W. Li (2003), Inter-basin freshwater, heat and salt transport through the boundaries of the Eastand South China Seas from a variable-grid global ocean circulation model,Sci. China Ser. D Earth Sci., 46(2), 149–161, doi:10.1360/03yd9014.

Fang, G. H., D. Susanto, I. Soesilo, Q. A. Zheng, F. L. Qiao, and Z. X. Wei(2005), A note on the South China Sea shallow interocean circulation,Adv. Atmos. Sci., 22(6), 946–954.

Fernandes, L., et al. (2012), Biophysical principles for designing resilientnetworks of marine protected areas to integrate fisheries, biodiversity andclimate change objectives in the Coral Triangle, report prepared by TheNature Conservancy for the Coral Triangle Support Partnership, 152 pp.

Ffield, A., and A. L. Gordon (1996), Tidal mixing signatures in the Indone-sian seas, J. Phys. Oceanogr., 26(9), 1924–1937, doi:10.1175/1520-0485(1996)026<1924:tmsiti>2.0.co;2.

Godfrey, J. S. (1989), A Sverdrup model of the depth-integrated flow forthe world ocean allowing for island circulations, Geophys. Astrophys.Fluid Dyn., 45(1–2), 89–112, doi:10.1080/03091928908208894.

Godfrey, J. S. (1996), The effect of the Indonesian throughflow on oceancirculation and heat exchange with the atmosphere: A review, J. Geo-phys. Res., 101(C5), 12,217–12,237, doi:10.1029/95JC03860.

Gordon, A. L. (2005), Oceanography of the Indonesian Seas and theirthroughflow, Oceanography, 18(4), 14–27.

Gordon, A. L., and R. A. Fine (1996), Pathways of water between thePacific and Indian oceans in the Indonesian seas, Nature, 379(6561),146–149, doi:10.1038/379146a0.

Gordon, A. L., and R. D. Susanto (1998), Makassar Strait transport : Initialestimate based on Arlindo results, Mar. Technol. Soc. J., 32(4), 34–45.

Gordon, A. L., and V. M. Kamenkovich (2010), ‘‘Modeling and Observingthe Indonesian Throughflow’’ a special issue of dynamics of atmosphereand ocean, Dyn. Atmos. Oceans, 50(2), 113–114, doi:10.1016/j.dynatmoce.2010.04.003.

Gordon, A. L., R. D. Susanto, and K. Vranes (2003), Cool Indonesianthroughflow as a consequence of restricted surface layer flow, Nature,425(6960), 824–828, doi:10.1038/nature02038.

Gordon, A. L., R. D. Susanto, A. Ffield, B. A. Huber, W. Pranowo, and S.Wirasantosa (2008), Makassar Strait throughflow, 2004 to 2006, Geo-phys. Res. Lett., 35, L24605, doi:10.1029/2008GL036372.

Gordon, A. L., J. Sprintall, H. M. Van Aken, D. Susanto, S. Wijffels, R.Molcard, A. Ffield, W. Pranowo, and S. Wirasantosa (2010), The Indone-sian throughflow during 2004–2006 as observed by the INSTANT program,Dyn. Atmos. Oceans, 50(2), 115–128, doi:10.1016/j.dynatmoce.2009.12.002.

Gordon, A. L., B. A. Huber, E. J. Metzger, R. D. Susanto, H. E. Hurlburt,and T. R. Adi (2012), South China Sea throughflow impact on the Indo-nesian throughflow, Geophys. Res. Lett., 39, L11602, doi:10.1029/2012GL052021.

CASTRUCCIO ET AL.: MESOSCALE MODELING IN THE CORAL TRIANGLE

6142

Page 21: A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago

Haidvogel, D. B., H. G. Arango, K. Hedstrom, A. Beckmann, P. Malanotte-Rizzoli, and A. F. Shchepetkin (2000), Model evaluation experiments inthe North Atlantic Basin: Simulations in nonlinear terrain-followingcoordinates, Dyn. Atmos. Oceans, 32, 239–281.

Haidvogel, D. B., et al. (2008), Ocean forecasting in terrain-following coor-dinates: Formulation and skill assessment of the Regional Ocean Model-ing System, J. Comput. Phys., 227(7), 3595–3624, doi:10.1016/j.jcp.2007.06.016.

Haller, G. (2002), Lagrangian coherent structures from approximate veloc-ity data, Phys. Fluids, 14(6), 1851–1861, doi:10.1063/1.1477449.

Haller, G., and G. Yuan (2000), Lagrangian coherent structures and mixingin two-dimensional turbulence, Physica D, 147(3–4), 352–370, doi:10.1016/s0167–2789(00)00142-1.

Han, W. Q., A. M. Moore, J. Levin, B. Zhang, H. G. Arango, E. Curchitser,E. Di Lorenzo, A. L. Gordon, and J. L. Lin (2009), Seasonal surfaceocean circulation and dynamics in the Philippine Archipelago regionduring 2004–2008, Dyn. Atmos. Oceans, 47(1–3), 114–137, doi:10.1016/j.dynatmoce.2008.10.007.

Harrison, C. S., D. A. Siegel, and S. Mitarai (2013), Filamentation andeddy-eddy interactions in marine larval accumulation and transport,Mar. Ecol. Prog. Ser., 472, 27–44, doi:10.3354/meps10061.

Hatayama, T. (2004), Transformation of the Indonesian throughflow waterby vertical mixing and its relation to tidally generated internal waves,J. Oceanogr., 60(3), 569–585, doi:10.1023/B:JOCE.0000038350.32155.cb.

Hernandez-Carrasco, I., C. Lopez, E. Hernandez-Garcia, and A. Turiel(2011), How reliable are finite-size Lyapunov exponents for the assess-ment of ocean dynamics?, Ocean Modell., 36(3–4), 208–218, doi:10.1016/j.ocemod.2010.12.006.

Hirst, A. C., and J. S. Godfrey (1993), The role of the Indonesian through-flow in a global ocean GCM, J. Phys. Oceanogr., 23(6), 1057–1086, doi:10.1175/1520-0485(1993)023<1057:troiti>2.0.co;2.

Hurlburt, H. E., E. J. Metzger, J. Sprintall, S. N. Riedlinger, R. A. Arnone,T. Shinoda, and X. B. Xu (2011), Circulation in the Philippine Archipel-ago simulated by 1/12� and 1/25� global HYCOM and EAS NCOM,Oceanography, 24(1), 28–47.

Jerlov, N. H. (1976), Marine Optics, 231 pp., Elsevier, Amsterdam.Kartadikaria, A. R., Y. Miyazawa, S. M. Varlamov, and K. Nadaoka

(2011), Ocean circulation for the Indonesian seas driven by tides andatmospheric forcings: Comparison to observational data, J. Geophys.Res., 116, C09009, doi:10.1029/2011JC007196.

Koch-Larrouy, A., G. Madec, P. Bouruet-Aubertot, T. Gerkema, L.Bessieres, and R. Molcard (2007), On the transformation of PacificWater into Indonesian Throughflow Water by internal tidal mixing, Geo-phys. Res. Lett., 34(4), L04604, doi:10.1029/2006GL028405.

Large, W. G., J. C. McWilliams, and S. C. Doney (1994), Oceanic ver-tical mixing—A review and a model with a nonlocal boundary-layerparameterization, Rev. Geophys., 32(4), 363–403, doi:10.1029/94RG01872.

Large, W. G., and S. G. Yeager (2009), The global climatology of an inter-annually varying air-sea flux data set, Clim. Dyn., 33(2–3), 341–364,doi:10.1007/s00382-008-0441-3.

Lemarie, F., J. Kurian, A. F. Shchepetkin, M. J. Molemaker, F. Colas, andJ. C. McWilliams (2012), Are there inescapable issues prohibiting theuse of terrain-following coordinates in climate models?, Ocean Modell.,42, 57–79, doi:10.1016/j.ocemod.2011.11.007.

Levitus, S., T. P. Boyer, M. E. Conkright, T. O’Brien, J. Antonov, C.Stephens, L. Stathoplos, D. Johnson, and R. Gelfeld (1998), NOAA AtlasNESDIS 18, World Ocean Database 1998: Volume 1: Introduction, p.346, U.S. Gov. Printing Office, Washington, D. C.

Liu, Q. Y., M. Feng, and D. X. Wang (2011), ENSO-induced interannualvariability in the southeastern South China Sea, J. Oceanogr., 67(1),127–133, doi:10.1007/s10872-011-0002-y.

Lukas, R., T. Yamagata, and J. P. McCreary (1996), Pacific low-latitudewestern boundary currents and the Indonesian throughflow, J. Geophys.Res., 101(C5), 12,209–12,216, doi:10.1029/96JC01204.

Marchesiello, P., J. C. McWilliams, and A. Shchepetkin (2001), Openboundary conditions for long-term integration of regional oceanic mod-els, Ocean Modell., 3(1–2), 1–20, doi:10.1016/s1463–5003(00)00013-5.

Marchesiello, P., J. C. McWilliams, and A. Shchepetkin (2003), Equilib-rium structure and dynamics of the California Current System, J. Phys.Oceanogr., 33, 753–783.

Marchesiello, P., L. Debreu, and X. Couvelard (2009), Spurious diapycnalmixing in terrain-following coordinate models: The problem and a solution,Ocean Modell., 26(3–4), 156–169, doi:10.1016/j.ocemod.2008.09.004.

Martinho, A. S., and M. L. Batteen (2006), On reducing the slope parameterin terrain-following numerical ocean models, Ocean Modell., 13(2),166–175, doi:10.1016/j.ocemod.2006.01.003.

McLeod, E., R. Salm, A. Green, and J. Almany (2009), Designing marineprotected area networks to address the impacts of climate change, Front.Ecol. Environ., 7(7), 362–370, doi:10.1890/070211.

McLeod, E., R. Moffitt, A. Timmermann, R. Salm, L. Menviel, M. J. Palmer,E. R. Selig, K. S. Casey, and J. F. Bruno (2010), Warming seas in theCoral Triangle: Coral reef vulnerability and management implications,Coastal Manage., 38(5), 518–539, doi:10.1080/08920753.2010.509466.

McLeod, E., et al. (2012), Integrating climate and ocean change vulnerabil-ity into conservation planning, Coastal Manage., 40(6), 651–672.

McPhaden, M. J., et al. (1998), The tropical ocean global atmosphereobserving system: A decade of progress, J. Geophys. Res., 103(C7),14,169–14,240, doi:10.1029/97JC02906.

Melet, A., L. Gourdeau, W. S. Kessler, J. Verron, and J. M. Molines(2010), Thermocline circulation in the Solomon Sea: A modeling study,J. Phys. Oceanogr., 40(6), 1302–1319, doi:10.1175/2009JPO4264.1.

Metzger, E. J., and H. E. Hurlburt (1996), Coupled dynamics of the SouthChina Sea, the Sulu Sea, and the Pacific Ocean, J. Geophys. Res.,101(C5), 12,331–12,352, doi:10.1029/95JC03861.

Metzger, E. J., H. E. Hurlburt, X. Xu, J. F. Shriver, A. L. Gordon, J.Sprintall, R. D. Susanto, and H. M. van Aken (2010), Simulated andobserved circulation in the Indonesian Seas: 1/12� global HYCOM andthe INSTANT observations, Dyn. Atmos. Oceans, 50(2), 275–300, doi:10.1016/j.dynatmoce.2010.04.002.

Neale, R., and J. Slingo (2003), The maritime continent and its role in theglobal climate: A GCM study, J. Clim., 16(5), 834–848, doi:10.1175/1520-0442(2003)016<0834:tmcair>2.0.co;2.

Pe~naflor, E. L., W. J. Skirving, A. E. Strong, S. F. Heron, and L. T. David(2009), Sea-surface temperature and thermal stress in the Coral Triangleover the past two decades, Coral Reefs, 28(4), 841–850, doi:10.1007/s00338-009-0522-8.

Qu, T. D., and R. Lukas (2003), The bifurcation of the North EquatorialCurrent in the Pacific, J. Phys. Oceanogr., 33(1), 5–18, doi:10.1175/1520-0485(2003)033<0005:tbotne>2.0.co;2.

Qu, T., Y. Du, J. Strachan, G. Meyers, and J. Slingo (2005), Sea surfacetemperature and its variability in the Indonesian region, Oceanography,18(4), 50–61.

Qu, T. D., H. Mitsudera, and T. Yamagata (2000), Intrusion of the NorthPacific waters into the South China Sea, J. Geophys. Res., 105(C3),6415–6424, doi:10.1029/1999JC900323.

Qu, T. D., Y. T. Song, and T. Yamagata (2009), An introduction to theSouth China Sea throughflow: Its dynamics, variability, and applicationfor climate, Dyn. Atmos. Oceans, 47(1–3), 3–14, doi:10.1016/j.dynatmoce.2008.05.001.

Ray, R. D., G. D. Egbert, and S. Y. Erofeeva (2005), Tides in the Indone-sian seas, Oceanography, 18, 74–79.

Reynolds, R. W., N. A. Rayner, T. M. Smith, D. C. Stokes, and W. Q.Wang (2002), An improved in situ and satellite SST analysis for climate,J. Clim., 15(13), 1609–1625, doi:10.1175/1520-0442(2002)015<1609:aiisas>2.0.co;2.

Rienecker, M. M., et al. (2011), MERRA: NASA’s modern-era retrospec-tive analysis for research and applications, J. Clim., 24(14), 3624–3648,doi:10.1175/jcli-d-11–00015.1.

Robertson, R. (2010), Tidal currents and mixing at the INSTANT mooringlocations, Dyn. Atmos. Oceans, 50(2), 331–373, doi:10.1016/j.dynatmoce.2010.02.004.

Robertson, R. (2011), Interactions between tides and other frequencies inthe Indonesian seas, Ocean Dyn., 61(1), 69–88, doi:10.1007/s10236-010-0343-x.

Robertson, R., and A. Ffield (2005), M2 Baroclinic tides in the Indonesianseas, Oceanography, 18, 62–73.

Robertson, R., and A. Ffield (2008), Baroclinic tides in the Indonesianseas: Tidal fields and comparisons to observations, J. Geophys. Res.,113, C07031, doi:10.1029/2007JC004677.

Rossi, V., C. L�opez, J. Sudre, E. Hern�andez-Garc�ıa, and V. Garcon (2008),Comparative study of mixing and biological activity of the Benguela andCanary upwelling systems, Geophys. Res. Lett., 35, L11602, doi:10.1029/2008GL033610.

Schiller, A. (2004), Effects of explicit tidal forcing in an OGCM on thewater-mass structure and circulation in the Indonesian throughflow region,Ocean Modell., 6(1), 31–49, doi:10.1016/s1463–5003(02)00057-4.

Selig, E. R., K. S. Casey, and J. F. Bruno (2010), New insights into globalpatterns of ocean temperature anomalies: Implications for coral reef

CASTRUCCIO ET AL.: MESOSCALE MODELING IN THE CORAL TRIANGLE

6143

Page 22: A model for quantifying oceanic transport and mesoscale variability in the Coral Triangle of the Indonesian/Philippines Archipelago

health and management, Global Ecol. Biogeogr., 19(3), 397–411, doi:10.1111/j.1466–8238.2009.00522.x.

Shadden, S. C., F. Lekien, and J. E. Marsden (2005), Definition and proper-ties of Lagrangian coherent structures from finite-time Lyapunov expo-nents in two-dimensional aperiodic flows, Physica D, 212(3–4), 271–304, doi:10.1016/j.physd.2005.10.007.

Shapiro, R. (1975), Linear filtering, Math. Comput., 29(132), 1094–1097,doi:10.2307/2005747.

Shchepetkin, A. F., and J. C. McWilliams (2003), A method for computinghorizontal pressure-gradient force in an oceanic model with a nonalignedvertical coordinate, J. Geophys. Res., 108(C3), 3090, doi:3010.1029/2001JC001047.

Shchepetkin, A. F., and J. C. McWilliams (2005), The regional oceanicmodeling system (ROMS): A split-explicit, free-surface, topography-following-coordinate oceanic model, Ocean Modell., 9(4), 347–404,doi:10.1016/j.ocemod.2004.08.002.

Shchepetkin, A. F., and J. C. McWilliams (2009), Ocean forecasting interrain-following coordinates: Formulation and skill assessment of theregional ocean modeling system (vol 227, pg 3595, 2008), J. Comput.Phys., 228(24), 8985–9000, doi:10.1016/j.jcp.2009.09.002.

Song, Q., G. A. Vecchi, and A. J. Rosati (2007), The role of the IndonesianThroughflow in the Indo-Pacific climate variability in the GFDL CoupledClimate Model, J. Clim., 20(11), 2434–2451, doi:10.1175/jcli4133.1.

Sprintall, J., A. L. Gordon, R. Murtugudde, and R. D. Susanto (2000), Asemiannual Indian Ocean forced Kelvin wave observed in the Indonesianseas in May 1997, J. Geophys. Res., 105(C7), 17,217–17,230, doi:10.1029/2000JC900065.

Sprintall, J., S. E. Wijffels, R. Molcard, and I. Jaya (2009), Direct estimatesof the Indonesian Throughflow entering the Indian Ocean: 2004–2006,J. Geophys. Res., 114, C07001, doi:10.1029/2008JC005257.

Susanto, R. D., A. Ffield, A. L. Gordon, and T. R. Adi (2012), Variabilityof Indonesian throughflow within Makassar Strait, 2004–2009, J. Geo-phys. Res., 117, C09013, doi:10.1029/2012JC008096.

Tittensor, D. P., C. Mora, W. Jetz, H. K. Lotze, D. Ricard, E. Vanden Berghe,and B. Worm (2010), Global patterns and predictors of marine biodiversityacross taxa, Nature, 466(7310), 1098–1107, doi:10.1038/nature09329.

Umlauf, L., and H. Burchard (2003), A generic length-scale equation forgeophysical turbulence models, J. Mar. Res., 61(2), 235–265, doi:10.1357/002224003322005087.

van Aken, H. M., I. S. Brodjonegoro, and I. Jaya (2009), The deep-watermotion through the Lifamatola Passage and its contribution to the Indo-

nesian throughflow, Deep Sea Res., Part I, 56(8), 1203–1216, doi:10.1016/j.dsr.2009.02.001.

van Haren, H., L. S. Laurent, and D. Marshall (2004), Small and mesoscaleprocesses and their impact on the large scale: An introduction, Deep SeaRes., Part I, 51(25–26), 2883–2887, doi:10.1016/j.dsr2.2004.09.010.

Varikoden, H., A. A. Samah, and C. A. Babu (2010), The cold tongue inthe South China Sea during boreal winter and its interaction with theatmosphere, Adv. Atmos. Sci., 27(2), 265–273, doi:10.1007/s00376-009-8141-4.

Veron, J. E. N., L. M. DeVantier, E. Turak, A. L. Green, and S.Kininmonth (2009), Delineating the coral triangle, Galaxea, 11(2), 91–100.

Wajsowicz, R. C., and E. K. Schneider (2001), The Indonesian through-flow’s effect on global climate determined from the COLA coupled cli-mate system, J. Clim., 14(13), 3029–3042.

Wang, Q. Y., H. Cui, S. W. Zhang, and D. X. Hu (2009), Water transportsthrough the four main straits around the South China Sea, Chin. J. Oce-anol. Limnol., 27(2), 229–236, doi:10.1007/s00343-009-9142-y.

Warner, J. C., W. R. Geyer, and J. A. Lerczak (2005a), Numerical modelingof an estuary: A comprehensive skill assessment, J. Geophys. Res., 110,C05001, doi:05010.01029/02004JC002691.

Warner, J. C., C. R. Sherwood, H. G. Arango, and R. P. Signell (2005b),Performance of four turbulence closure methods implemented using aGeneric Length Scale Method, Ocean Modell., 8, 81–113.

Wolanski, E., P. Ridd, and M. Inoue (1988), Currents through Torres Strait,J. Phys. Oceanogr., 18(11), 1535–1545, doi:10.1175/1520-0485(1988)018<1535:ctts>2.0.co;2.

Xie, S. P., Q. Xie, D. X. Wang, and W. T. Liu (2003), Summer upwellingin the South China Sea and its role in regional climate variations, J. Geo-phys. Res., 108(C8), 3261, doi:10.1029/2003CJ001867.

Xie, J., F. Counillon, J. Zhu, and L. Bertino (2011), An eddy resolvingtidal-driven model of the South China Sea assimilating along-track SLAdata using the EnOI, Ocean Sci. Discuss., 8, 873–916, doi:10.5194/osd-8–873-2011.

Yu, Z., S. Shen, J. P. McCreary, M. Yaremchuk, and R. Furue (2007),South China Sea throughflow as evidenced by satellite images andnumerical experiments, Geophys. Res. Lett., 34, L01601, doi:10.1029/2006GLl028103.

Zu, T. T., H. P. Gan, and S. Y. Erofeeva (2008), Numerical study of the tideand tidal dynamics in the South China Sea, Deep Sea Res., Part I, 55(2),137–154, doi:10.1016/j.dsr.2007.10.007.

CASTRUCCIO ET AL.: MESOSCALE MODELING IN THE CORAL TRIANGLE

6144