Journal of Marine Systems 78 (2009) 351–362

Contents lists available at ScienceDirect

Journal of Marine Systems

j ourna l homepage: www.e lsev ie r.com/ locate / jmarsys

Interannual variation of the Polar Front in the Japan/East Sea from summertimehydrography and sea level data

Byoung-Ju Choi a,⁎, Dale B. Haidvogel a, Yang-Ki Cho b

a Institute of Marine and Coastal Sciences, Rutgers—the State University of New Jersey, New Brunswick, New Jersey 08901-8521, USAb Department of Oceanography, Chonnam National University, Kwangju, 500-757, Republic of Korea

⁎ Corresponding author. Present address: DepartmeNational University, San 58, Miryong-dong, Gunsan, JeoKorea. Tel.: +82 63 469 4607; fax: +82 63 469 4990.

E-mail address: bjchoi@kunsan.ac.kr (B.-J. Choi).

0924-7963/$ – see front matter © 2009 Elsevier B.V. Aldoi:10.1016/j.jmarsys.2008.11.021

a b s t r a c t

a r t i c l e i n f o

Article history:

The Polar Front in the Japan Received 17 October 2006Received in revised form 23 May 2008Accepted 7 November 2008Available online 20 February 2009

Regional index terms:Northwestern Pacific OceanJapan/East Sea126°E–142°E34°N–52°N

Keywords:Polar FrontHydrographic dataAltimeter dataNumerical modelInterannual variationJapan/East Sea

/East Sea separates the southernwarmwater region from the northern cold waterregion. A merged TOPEX/POSEIDON and ERS-1/2 altimeter dataset and upper water temperature data wereused to determine the frontal location and to examine the structure of its interannual variability from 1993 to2001. The identified frontal location, where sea surface height gradient has a maximum about 10–20 cm overthe horizontal distance of 100 km, corresponds well to the maximum subsurface horizontal temperaturegradient. The front migrates more widely (36°N–41°N) in the western part of the sea than in the eastern part.The interannual migration induces large variability in upper water temperatures and sea surface height in thewestern region. Responsible physical mechanisms were studied using a reduced-gravity model. Differencesbetween inflow and outflow change the total volume of warmwater, and total warmwater volume change inthe warmwater region uniformly pushes the front in the meridional direction across its mean position in themodel simulation. Interannual variation of wind stress causes relatively wide migration of the modeled frontin the western part.

© 2009 Elsevier B.V. All rights reserved.

1. Introduction

The Polar Front marks an important climate boundary in terms ofboth water mass formation and air–sea fluxes. Accordingly, variationof the Polar Front in the Southern Ocean and the North Pacific Oceanhas been studied extensively (Roden et al., 1982; Moore et al., 1999).The Polar Front of the North Pacific extends from 57°N in the Gulf ofAlaska to 40°N off Japan. The front is relatively stable, except along170°E, where it shifts north–south by 400 km every 6 years (Belkinet al., 2002). The western end of the Polar Front in the North Pacificextends into the Japan/East Sea, and is located between 38°N and 40°N(Tomczak and Godfrey, 1994; Belkin and Cornillon, 2003). The PolarFront in the Japan/East Sea is often referred to as the subpolar front inother recent studies (Park et al., 2004; Talley et al.; 2006).

The Japan/East Sea is a marginal sea of the North Pacific Ocean(Fig. 1a). Its dimensions are about 1600×900 km and the mean depthis about 1350 m. It communicates with the East China and Yellow Seasto the south, with the Pacific Ocean to the east, and with the Sea of

nt of Oceanography, Kunsanllabuk-do 573-701, Republic of

l rights reserved.

Okhotsk to the north. Water is exchanged through narrow channelswith sill depths not exceeding 135 m. Warm and saline water entersfrom the Kuroshio, flows through the southern part of the sea, andexits to the subpolar gyre with a throughflow transport of about2.5×106m3/s (Sv). The warm water occupies the depth range of 0–200 m and is characterized by a shallow salinity maximum at about50 m depth in the warm southern Japan/East Sea (Talley et al., 2006).Below 200m, thewater is remarkably uniformwith temperature of 0–1 °C and salinity of 34.1 (Tomczak and Godfrey, 1994; Preller andHogan, 1998).

The Polar Front separates the southern warm water from thenorthern cold water in the Japan/East Sea. In Fig. 1a, solid lines witharrows indicate surface currents in the southern warm water regionand dashed lines denote those in the northern cold water region. Thedotted line is an approximate position of the Polar Front. A line of opencircles represents hydrographic stations where upper layer tempera-ture data were obtained to identify vertical structure of the upperwater across the Polar Front in July 1993. The vertical section of upperocean temperature along the north–south transect shows thestructure of the Polar Front (Fig. 1b). Due to summer heating, theupper 50-m layer is highly stratified and the front submerges belowthe surface. There is a strong subsurface horizontal temperaturegradient at depths of 50–150 m across 39.2°N. Horizontal maps of the

Fig. 1. (a) Schematic surface currents in summer and location of the Polar Front. UIstands for Ulleung Island. (b) Comparison of upper layer temperature and sea surfaceheight. PF stands for the Polar Front in (b).

352 B.-J. Choi et al. / Journal of Marine Systems 78 (2009) 351–362

subsurface temperature give relatively better identification of thefront position (Uda, 1938; Isoda et al., 1991; Talley et al., 2006). Oncesea surface height measured from satellite altimeters is plotted on topof the vertical temperature section to compare with the upper layerheat content as in Fig.1b, the sea surface height is closely related to theupper water heat content.

The front is identified by the maximum horizontal temperaturegradient and, in general, it coincides with closely spaced zonallydirected isotherms. It is about 100–150 km wide and is a boundary ofphysical and chemical properties such as temperature, salinity anddissolved oxygen (Kim and Kim, 1999) and nutrients (Talley et al.,2006). Sinking of water immediately north of the Polar Front formsthe Japan/East Sea intermediate water (Senjyu and Sudo, 1994;Yoshikawa et al., 1999). Intrathermocline eddies may form throughfrontal convergence and subduction at the Polar Front (Ou andGordon, 2002; Gordon et al., 2002). Migration of the front affectsmeso-scale circulations, water mass formation, eddy generation andnutrient distributions in the Japan/East Sea.

Sea level measured at Ulleung Island (130.9E°E, 37.4°N) has largeinterannual variations comparable to themean seasonal variation, andthe interannual variation in the southwestern part of the Japan/EastSea was found to be related to migration of the Polar Front from upper

water temperature data (Kim et al., 2002). Statistical analysis of seasurface and upper water temperature in the Japan/East Sea revealedthat interannual variation of the upper water temperature is larger inthe western part than in the eastern part (Chu et al., 1998; Minobeet al., 2004; Choi et al., 2004a). A physical mechanism for the strongvariability in the western part was thought to be the migration of thePolar Front. Seasonal variation of the Polar Front has been investigatedusing satellite sea surface temperature (SST) data (Isoda et al., 1991),climatological temperature and salinity data (Chu et al., 2001), and innumerical models (Kim and Yoon, 1996). Belkin and Cornillon (2003)examined thermals fronts of the Pacific coastal and marginal seasincluding the entire Japan/East Sea using the Pathfinder AVHRR SSTdata from 1985 to 1996. Park et al. (2004) identified the locations ofthe sea surface temperature front from an analysis of satellite SSTimages for the entire Japan/East Sea. Because of sparseness andirregularity of measured upper water temperature data, the inter-annual variability of the Polar Front over the entire Japan/East Sea hasnot been well investigated in previous studies.

Altimeter data have been used to examine seasonal variation of seasurface circulation and to determinemeso-scale eddy variability in theJapan/East Sea. Twelve ground tracks of TOPEX/POSEIDON (T/P) passover the Japan/East Seawith an orbit repeat period of 9.9 days. ERS-1/2have forty-one ground tracks and their repeat cycle is 35days. FromT/Pand ERS-2 altimeter data, Morimoto et al. (2000) found high root-mean-square (RMS) variabilities of sea surface height (SSH) in theYamato Basin, the Ulleung Basin, east of North Korea, and the easternpart of the YamatoRise.Morimoto andYanagi (2001) showed variationof sea surface circulation by EOF analysis of the 3.5 years of ERS-2altimeter data.

In this paper, we identify locations of the Polar Front from 1993 to2001 and examine the structure of its interannual variability based onthe merged altimeter data and hydrographic data in section 2. To seekpossible responsible factors for the interannual variation, numericalsimulations of the surface circulation are performed and the effects ofthe interannual variations in the wind stress and total warm watervolume are investigated in section 3.

2. Interannual variation of the Polar Front

Before the satellite remote sensing era, locations of the Polar Frontwere observed by on-board temperature measurements, and the in-situ measurements were limited in time and space. Since satelliteradiometers have sensed radiation emitted from the sea surface, thefront has been easily identified in maps of SST inferred fromcomposite infrared (IR) images (Isoda et al., 1991; Park et al., 2004).However, the measurements of SST from the satellite are frequentlyblocked by clouds in the mid-latitude. The data need to beinterpolated in time and space to fill in the cloud-contaminateddata. Additionally, the maps of summer SST do not capture the properlocations of the front because the seasonal surface heating in theupper 50 m of the water column reduces horizontal SST gradients. Onthe other hand, SSH measurements from satellite altimeters are notaffected by cloud clover and SSH is closely related to the integratedheat content in the upper layer in the Japan/East Sea (Kim et al., 2002)so that maps of SSH are better to identify the Polar Front in the Japan/East Sea.

Satellite altimeters have provided sea level anomalies (SLA), butlack of accurate geoid information in the Japan/East Sea prevents thealtimeters from providing the absolute SSH. Composite dynamic sealevel (DSL) can be calculated by combining themean sea surface stericheight calculated from the long-term mean hydrographic data andSLA derived from the altimeters (Morimoto and Yanagi, 2001;Korotaev et al., 2003; Choi et al., 2004a). The mean sea surface stericheight was calculated relative to 500 dbar using temperature andsalinity climatological data from the Japan Oceanographic Data Center(JODC) and the Korea Oceanographic Data Center (KODC). SLA data

Fig. 2. Composite dynamic sea level (DSL) in July from 1993 to 2001. Contour interval is 2 cm. In panels (d) and (f), two lines of circles are longitudinal cross section lines, V and PM. Solid (open) circles indicate hydrographic stations whereupper water temperature data are (are not) available.

353B.-J.Choi

etal./

Journalof

Marine

Systems78

(2009)351

–362

354 B.-J. Choi et al. / Journal of Marine Systems 78 (2009) 351–362

are the merged T/P and ERS-1/2 altimeter data at 1/3° spatial intervalsand with a repeat cycle of 7 days, which were obtained from the SpaceOceanography Division of Collecte Localisation Satellite (CLS) inToulouse, France (Le Traon et al., 1998; Ducet et al., 2000). Because thereference level for themean steric height is 500 dbar, composite DSL isthe steric height relative to 500 dbar in this paper. It has the samephysical meaning as the dynamic height in the analysis of surfacecirculation. Horizontal maps of composite DSL are also closely relatedto the integrated heat content in the upper layer.

Maps of composite DSL are plotted to see surface circulation in theJapan/East Sea and to find locations of the Polar Front in July from1993 to 2001 (Fig. 2). SSH data are available three or four times amonth because sea level anomalies are obtained and gridded every7 days based on objective analysis by interpolating observed data intime and in space. A map of sea level anomalies in Fig. 2 representscomposite DSL on a given day in July. Note that SSH data, which areused to produce composite DSL, were interpolated with an interpola-tion-time window of 40 days. Assuming geostrophic balance, the seasurface current flows along the contour lines and strong surfacecurrents appear in closely spaced contour lines. The Polar Front isapproximately located at the northern edge of maximum horizontalgradients in maps of the composite DSL (Fig. 1b). In July 1994, 1996and 2000, the front retreated to the south and cold water occupied thewestern part of the basin (Fig. 2). In contrast, the front moved back tothe north and aligned along 40°N in July 1993, 1998 and 1999. In thesewarm periods, warm water advanced in the western part of the basinand many eddies developed to the south of the front. The meridionalmigration range of the Polar Front is between 36.5°N and 41°N in thewestern part (along 131°E), and between 39°N and 40°N in the easternpart (along 137°E).

Fig. 3. Temperature cross-sections and composite dynamic sea level, along the V and PM linhydrographic stations where upper water temperature data are (are not) available. PF stan

To examine the vertical upper water temperature structure acrossthe front, we examined temperature sections along the two long-itudinal lines in the western and central parts. The vertical tempera-ture sections are also compared with the composite DSL in Fig. 3. Dueto summer surface heating, the upper 50m of water is highly stratifiedin the vertical, and sea surface temperatures have weak horizontalgradients relative to subsurface horizontal temperature gradients. InJuly 1996, the Polar Front was located at 36.5°N along the V line and at38.5°N along the PM line. Thewarmwater was confined to the south ofthe front. In July 1998, thewarmwater advancedup to 40°Nalong theVline and replaced the cold water in the water column above thepermanent thermocline, which is the strongly stratified layersbetween 5 and 10 °C. Warm eddies developed to the south of thefront. Generally, the compositeDSL representswell the subsurface heatcontent and it resolves the warm eddies at 38.2°N and 39.5°N in July1998. The slope of sea surface across the subsurfacemaximumgradientof temperature is approximately 10–20 cm over the horizontaldistance of 100 km.

To confirm the interannual variations in warm water distributionin July 1996 and 1998, horizontal maps of temperature at 100 m depthare plotted in Fig. 4. Temperature data were obtained from JODC andKODC. Warm water refers to water whose temperature is higher than10 °C at the depth of 100 m, a conservative measure because watertemperature is 2–3 °C at 100 m depth in the cold water region (Fig. 3).In July 1996, the surface current from the Korea Strait flows near thesoutheastern coastal boundary instead of flowing northward along thewestern boundary (Fig. 2d), and warm water is confined offshore ofJapan (Fig. 4a). Cold water (<5 °C) was found offshore of Korea andreaches down to the Korea Strait. In July 1998, the surface current fromthe Korea Strait flows along the east coast of Korea (Fig. 2f) and

es in Fig. 2, in July 1996 and 1998. Contour interval is 1 °C. Solid (open) circles indicateds for the Polar Front.

Fig. 4. Maps of 100-m depth temperature in July 1996 and 1998. The contour interval is 2.5 °C and small dots represent temperature observation stations.

355B.-J. Choi et al. / Journal of Marine Systems 78 (2009) 351–362

provides warm water in the Ulleung Basin (Fig. 4b). Then, the warmwater replenishes the Ulleung Basin and the southern Yamato Basin.In most of the warm water region, the distance between the 5 °C andthe 10 °C contours is more close in July 1998 than in July 1996.

In order to quantify meridional migration of the Polar Front in thewestern, central and eastern parts of the Japan/East Sea, latitudinallocations of the Polar Front along 130.5°E, 134.0°E and 137.5°E areplotted from 1993 to 2001 (Fig. 5). The location of the Polar Front wasdetermined from the north–south gradient of composite DSL. Frontalstructures in Fig. 2 show a double-frontal structure in July 1997 and2001, and the frontal location is also largely affected by local meso-scale eddies in July 1993 and 1994. Thus, an automatic algorithm usedin the frontal location determination program does not always work.To find appropriate location of the front, we used an assumption thatthe front continuously extends from the west to the east. Thesubsurface front is located where SSH gradient has the maximum oneach meridional section. Latitudinal locations of the front (three linesin Fig. 5) closely go together in general except for the periods from1996 to early 1997 and from 2000 to early 2001. Interannual variationis highest in the western part (along 130.5°E) and lowest in theeastern part (along 137.5°E). The front migrates about 400 km from36.4°N to 40.1°N in the western part while it migrates about 250 kmfrom 38.5°N to 40.8°N in the eastern part. In the central part (134°E)the Polar Front migrates north–south between 37.2°N and 40.7°N.

3. Numerical simulations

To find the physical mechanisms responsible for the interannualvariation of the Polar Front in the western Japan/East Sea, a set of

Fig. 5. Latitudinal location of the Polar Front along 130.5°E, 134

numerical simulations was performed using a reduced-gravity model.The ocean is thus assumed to consist of a single active layer of constantdensity and variable layer thickness, overlying a denser, deep andmotionless layer; the motion of the upper layer represents the firstbaroclinic mode. All thermodynamic effects were neglected. Theinterface between the two fluid layers is a material surface whichrepresents the permanent thermocline.

A reduced-gravity model is particularly suitable for the applicationto the southern Japan/East Sea, which has a distinct two-layerstructure. Kim and Yoon (1996) developed a reduced-gravity modelbased on finite differences to simulate the Japan/East Sea circulation,and it has been used in various studies (Hirose et al., 1999; Hirose andOstrovskii, 2000). Since the Polar Front is a discontinuity where thewarm water layer vanishes (layer thickness of the upper warm waterbecomes zero), a model which allows layer vanishing is a mostfavorable choice for the proper simulation of the surface circulation.The reduced-gravity models of Kim and Yoon (1996) and Hogan andHurlburt (2000) do not allow vanishing of the upper layer. Hogan andHurlburt (2005) used a multi-layered ocean model to simulate oceancirculation in the Japan/East Sea and investigated the circulationsensitivity to the choice of wind-forcing product. Here, a reduced-gravity model (Choi et al., 2004b) based on the spectral finite volume(SFV) method is used. The SFV model can allow the upper layer tovanish and can represent irregular coastal boundaries.

In the model, the upper layer thickness at rest is 100 m, thereduced gravity is 0.02 m/s2, and the average grid size is about 10 km(Fig. 6). The model grid barely resolves the first baroclinic Rossbyradius of about 15 km. Free-slip boundary conditions are applied alongthe coastal boundaries, and a no-normal flux of mass and momentum

.0°E and 137.5°E in the Japan/East Sea from 1993 to 2001.

Fig. 7. Snapshot of layer thickness (m) at year 10 with the steady transport and steadywind stress. The upper layer thickness at rest is 100 m.

Fig. 6. Each quadrilateral represents an element and each element contains 9 quadrilateral cells. Average grid size is about 10 km.

356 B.-J. Choi et al. / Journal of Marine Systems 78 (2009) 351–362

condition is imposed along the boundaries. In order to spin-up thecirculation in the model, constant in- and outflow transports of 2.5 Svare given through the straits, and the steady long-term mean windstresses are forced for 10 years. After 4 years of numerical integration,the total kinetic and potential energies reach an almost statisticalsteady state. The final snapshot of year 10 is used as the initialcondition of further experiments in section 3 (Fig. 7). The snapshot is atypical mean state of model upper ocean circulation.

Assuming geostrophic balance, the inflow enters through theKorea Strait and flows along the western boundary until it meets thenorthern cyclonic gyre (Fig. 7). The upper layer vanishes in the interiorof the northern cyclonic gyre, and a strong boundary current flowsalong the northwestern boundary. The confluence of the inflow andthe northern cyclonic recirculation gyre makes a distinct front in themid-latitude of the basin (Naganuma, 1977; Preller and Hogan, 1998).We performed four different numerical experiments (Table 1). Allexperiments are performed from 1993 to 2001 using the specifiedforcing fields: reference case (R) with steady forcing, variable inflowand outflow (I), variable wind stress (W). Themeridional location andshape of the front changes from 1993 to 2001 in each numericalexperiment (R, I, W and IW).

Unsteady monthly mean transports of the inflow and outflow aregiven as open boundary conditions for experiments I and IW. Themonthly mean inflow and outflow transports are estimated from thesea level difference across the Korea Strait (Lyu and Kim, 2003) andtotal warm water volume change within the Japan/East Sea.Differences in transports between the inflow and the outflow throughthe straits produce convergence or divergence of warm water withinthe basin; i.e., the total water volume of the JES changes in time. Therelative ratio of outflow through the Tsugaru Strait and the Soya Straitis 7 to 3. Unsteady monthly mean wind stress fields are applied at the

surface for experiments W and IW, and the wind stress data wereobtained from the Navy Operational Global Atmospheric PredictionSystem (NOGAPS) (Rosmond, 1992).

Table 1The numerical experiments: R, I, W and IW.

Experiment Inflow and outflow Wind stress

R Constant SteadyI Monthly SteadyW Constant MonthlyIW Monthly Monthly

The constant transport is 2.5×106 m3/s (2.5 Sv) and the steady wind stress is a long-termmeanwind stress from 1993 to 2001 which varies in space and is constant in time.The monthly transport (wind stress) is unsteady monthly mean transport (wind stress)from 1993 to 2001.

357B.-J. Choi et al. / Journal of Marine Systems 78 (2009) 351–362

3.1. Volume changes of the warm water

Total volume changes of water in the Japan/East Sea are estimatedfrom the merged T/P and ERS-1/2 altimeter dataset. The resolution ofthe merged altimeter data is 1/3° in space and 7 days in time. Thespatial mean of the sea level anomalies represents the total volumechange in the basin (Fig. 8a). The total volume is altered by thermalexpansion through sea surface heat exchanges and by horizontaladvection through the straits.

The dominant seasonal variation in the total volume is related tothe thermal steric height change of the water column due to surfaceheat fluxes, i.e., to surface heating and cooling (Stammer, 1997; Vivieret al., 1999). The thermal steric height change is estimated by

ηheat tð Þ ≈ ηheat t0ð Þ + 1ρ cp

Z t

t0

αT Q net − < Q net >ð Þdt ð1Þ

where ηheat is the thermal steric height; ρ the water density; cp thespecific heat of seawater; αT the coefficient of thermal expansion;Q net the net sea surface heat flux; < Q net > the long-termmean of thenet heat flux through the sea surface; and t the time.

The thermal expansion coefficient αT varies temporally andspatially. Monthly temperature (T) and salinity (S) fields of theWorld Ocean Atlas 2001 are used to estimate the mixed layer (ML)

Fig. 8. Changes in the total water volume in the Japan/East Sea. (a) Sea surface height (SSHrelated to the advection of warm water through the straits. The residuals are decomposed involume convergence (thick line). Note that vertical scale varies from panel (a) to panel (b)

depth, temperature, and salinity as a function of time and space. TheML depth is computed using a potential density criterion of 0.125from the surface value. These time-varying T and S values averagedover the ML depth are used to compute αT. The Q net is obtained fromdaily means of the National Centers for Environmental Prediction(NCEP) reanalysis data. Density (ρ) and heat capacity (cp) of seawaterare constant within a few percent throughout the basin. The timeseries of the estimated thermal steric height (ηheat) shows clearseasonal variation with a range of 13 cm (Fig. 8b). Interannualvariation of the thermal steric height is smaller than the seasonalvariation.

The spatial mean thermal steric change is removed from the spatialmean sea level (Fig. 8a and b). Then, the residuals are divided into twocomponents (Choi et al., 2004a): the basin-wide near-uniformoscillations (thin line) and the volume convergence (thick line) inFig. 8c. As the sea level changes outside of the Korea Strait withintraseasonal timescale, the signal enters the Japan/East Sea asbarotropic waves. Since the average depth of the Japan/East Sea isdeep enough (about 1350 m) and its horizontal length scale is about1000 km, the signals cross the basin within several hours and the sealevels within the basin appear to near-uniformly adjust to the sea leveloutside of the Korea Strait (Lyu et al., 2002). The basin-wide near-uniform oscillations change sea levels in both cold and warm waterregions with periods of 155, 126 and 100 days (Choi et al., 2004a). Theintraseasonal oscillations are not related to the warm water volumeconvergence in the southern Japan/East Sea. The near-uniformoscillation component is estimated by averaging the sea level northof 42°N since the northern region is not directly affected by the warmwater volume convergence. The transport difference between theinflow and the outflow canmake thewarmwater convergence and sealevel changes in the southern warm water region. The warm waterconvergence component is obtained by subtracting the near-uniformoscillation component from the residuals. The time series of the warmwater volume has seasonal and interannual variation. The watervolume change will be used for the estimation of the outflowtransport in the next section.

); (b) thermal steric height by sea surface heat exchange; and (c) residuals which areto two components: basin-wide near-uniform oscillations (thin line) and warm waterto panel (c).

358 B.-J. Choi et al. / Journal of Marine Systems 78 (2009) 351–362

3.2. Estimation of the inflow and the outflow transports

Lyu and Kim (2003) found a linear relationship between thetransport (from cable voltage and ADCP measurements) and sea leveldifference across the Korea Strait. A correction was applied to the sealevel difference using cross-strait hydrographic sections to removebaroclinic effects, which lead to an overestimation of the transport byup to 40% in summer. Transport of the inflow through the Korea Straitis computed from the sea level difference. Inflow transport hasinterannual variations as well as seasonal variation. The meantransport is about 2.5×106m3/s (Sv). The outflow transport isestimated by integrating the following equation in time

dV tð Þdt

= Tin tð Þ− Tout tð Þ ð2Þ

where Tin is the inflow transport from the sea level difference, dV/dt isthe total volume change within the basin estimated from the mergedaltimeter data, and Tout is the outflow transport which is unknown.We will use the inflow and outflow transports as open boundaryconditions of our numerical model. Since the transports of the inflowand the outflow are different in time, the total volume of waterchanges in the model domain.

Fig. 9. The root-mean-square (RMS) values of layer thickness anomaly from the

3.3. Results of simulations

Layer thickness h(x, y, t) is sampled every 6 days at every gridpoint for further analysis. Root-mean-square (RMS) values of the layerthickness anomalies are plotted in Fig. 9. Most of the layer thicknessvariability appears near the mean position of the Polar Front, and thelayer thickness variability is high in the southwestern part of themodel basin. Without any interannual variability in forcing (experi-ment R), nonlinear behavior of the surface flowmakes the Polar Frontmigrate across 38°N in the east of Ulleung Island (Fig. 9a). Previousnumerical model studies of the Japan/East Sea circulation showed thatan intrinsic variability can arise on interannual timescale even with afixed wind forcing due to nonlinearity of the equations of motion(Holloway et al., 1995; Hirose and Ostroviskii, 2000; Hogan andHurlburt, 2005). As the total volume of warm water changes(experiment I), the front migrates north–southward across its meanposition. Warm water volume change enhances the frontal migration(Fig. 9b). As unsteadywind stress is forced on the surface (experimentW), there is relatively high variability in the southwestern partbetween 36°N and 38°N (Fig. 9c). When the inflow, outflow and windstress are all unsteady (experiment IW), the north–south migration ofthe Polar Front is relatively enhanced. The unsteady wind stressinduces larger variability of the front than the unsteady transports andintrinsic nonlinearity of the flow. Note that the variability in Fig. 9

results of numerical simulations (R, I, W and IW). Contour interval is 10 m.

Fig. 10. Distribution of layer thickness from experiment W from 1993 to 2001. Contour interval is 10 m.

359B.-J.Choi

etal./

Journalof

Marine

Systems78

(2009)351

–362

Fig. 11. Nonseasonal component of zonal wind stress over Ulleung Basin (129°–133°E, 35°–39°N). Thin line is for monthly mean value; thick line is for 3-month average.

Fig. 12. The root-mean-square (RMS) values of sea surface height (SSH) after thermalsteric height and basin-wide near-uniform oscillations were removed from theobserved SSH at each grid point.

360 B.-J. Choi et al. / Journal of Marine Systems 78 (2009) 351–362

includes seasonal and interannual time-scale variability which wasinduced by the unsteady transports and unsteady wind.

Because variation of the wind forcing causes more energeticmigration of the Polar Front than those of the inflow and outflow inthe model simulation, the layer thickness maps from experimentWareplotted in Fig. 10 in order to examine spatial pattern of the variability.Distribution of layer thickness in the model is related to that of seasurface height on a day in July from 1993 to 2001, and the sea surfacecurrent flows along the contour lines assuming geostrophic balance ofthe surface flow. Southward flow from the northern cyclonic gyre andnorthward flow from the southern gyre combine around 37°N in thewestern boundary and make a strong flow to the northeast toward theTsugaru Strait. The northeastward flow corresponds to the Polar Frontand the flow meanders in time. Because layer thickness in Fig. 10 is arepresentation of upper layer circulation in July for 9 years, comparisonof the upper layer circulation of each year can provide ideas oninterannual variation of the Polar Front in summertime. In July 1996, thefront moves south to about 37.5°N and the warm water accumulatesalong the coast of Japan. Meanwhile, the front moves northward in thewestern and central parts of the Japan/East Sea and the warm waterregion extends to the north in July 1998. In most years, the front islocated near 39°N in the western part. The meridional migration of thefront is wider in the western part of the basin than in the eastern part.The model did not reproduce southward migration of the Polar Front inJuly 2000.

4. Discussion

The unsteady wind stress induces relatively larger interannualmigration of the front than the unsteady transports through the openboundaries in the southwesternpart of themodel domain from1993 to2001. To find the responsible wind factors, wind stress andwind stresscurl were analyzed. Winds over the Japan/East Sea are dominated bythe seasonal monsoon so that the mean seasonal variation wasremoved from the data. The meridional component of the wind stress,and the wind stress curls did not have notable interannual variations.The zonal component of the nonseasonal wind stress has interannualvariations over the southwestern and eastern parts of the basin.Nonseasonal zonal wind stress over the southwestern part of the basin(129°–133°E, 35°–39°N) is plotted in Fig. 11. The zonal wind stressanomaly is positive from July 1995 to April 1996, and negative fromNovember 1997 to December 1998. Those strong eastward (westward)anomalies of wind stress may cause southward (northward) anoma-lies of the Ekman transport in the reduced-gravity model.

The reduced gravity model did not exactly reproduce interannualvariation of the front in the Japan/East Sea because of limited capabilityof the model: coarse horizontal resolution and one active layer. Thefront and the boundary currents from the model simulations arebroader than reality owing to the coarse resolution of the model gridsize. In the coarse resolution model, inherent numerical diffusionprevents the front and the surface currents frommeandering. Nonlinear

nature of the equations of motion in a numerical model may generatemeandering of currents and a front (Hirose and Ostrovskii, 2000; Changet al., 2001). In order to include nonlinearity of oceanic currents and toresolve meso-scale eddies, a numerical model needs high (<3.5 km)horizontal resolution (Hogan and Hurlburt, 2005; Mooers et al., 2006).Since themodel has one active layer and the active layer vanishes in theinterior of the northern cyclonic gyre, the wind stress variability overthe northern cyclonic gyre does not fully affect the model circulation.Multi-layer or multi-level models will be valuable to understand thedynamical process of how surface wind stress variability affectssubsurface front migration.

Nearshore Branch, a branch of Tsushima Current flowing along theJapanese coast, is absent in the model results (Figs. 7 and 10) becausethe bottom topography was not represented in this reduced-gravitymodel. Hogan and Hurlburt (2000) used a multi-layered model and athin top-most layer was essential in the generation of NearshoreBranch in their model experiments. The reason why the reduced-gravity model could not reproduce southward migration of the PolarFront in July 2000 (Fig. 10) may be (1) lack of enough nonlinearity inthe coarse resolution model, (2) absence of interactions betweenupper layer and bottom layer circulation, or (3) deficiency of flow–

topography interactions (Hogan and Hurlburt, 2005).

361B.-J. Choi et al. / Journal of Marine Systems 78 (2009) 351–362

After the thermal steric height change and the basin-wide near-uniform oscillation componentwere removed at each grid point of themerged altimeter data, RMS values of the remaining signals werecalculated (Fig. 12). They are small (about 3 cm) in the northern coldwater region and relatively large (4–10 cm) in the southern warmwater region. The sea level variation and the upper layer thickness areapproximately related as

Δη ≈ Δρρ

Δh ≈ 2 × 10−3 Δh ð3Þ

where Δη is the sea level anomaly, Δρ the density difference betweenthe two layers and Δh the upper layer thickness anomaly. 6 cm changeof sea level in the southwestern part (Fig. 12) is approximatelyexplained by a 30 m change of upper layer thickness in the numericalmodel (Fig. 9c and d) and in the real ocean (Fig. 3).

The large variability of sea level in the Yamato Basin (135°E, 38°N and137°E, 39°N) was not reproduced in our numerical experiments (Figs. 9and 12). The high RMS values of sea level anomaly in the Yamato Basinmay be a result of the meandering of the Polar Front and movement ofeddies (Isoda, 1994; Hirose and Ostrovskii, 2000). Holloway et al. (1995)found that inclusion of eddy and topography interactions in a three-dimensional numericalmodel of the Japan/East Sea generates strongdeepcirculation and a distinct cold eddy in the southern Yamato Basin (135°E,38°N). Hogan and Hurlburt (2000, 2005) suggested that the horizontalgrid size should be smaller than 3.5 km to capture the bottom topographyeffects on surfacemeso-scale surface circulations in the Japan/East Sea. Toinvestigate the effects of meandering currents andmeso-scale eddies, wemayneed to use amulti-layered or three-dimensionalmodelwith bottomtopography and sloping lateral boundaries.

5. Summary and concluding remarks

The Polar Font separates the southern warm water region from thenorthern cold water region in the Japan/East Sea. The merged T/P andERS-1/2 altimeter dataset from CLS Space Oceanography Division andthe temperature data from JODC and KODC were used to identifylocations of the Polar Front and to examine its interannual variability.Periodic sea level measurements from the satellite altimeters allowscontinuousmonitoring of the front. Locations of the front are effectivelyidentified frommaps of sea levelwhile surface thermal heatingmakes itsubmerge to subsurface and horizontal gradients of SST diminish insummer. Locations of the frontwere identified based on themaps of sealevel from 1993 to 2001 and they were confirmed by comparison withthe upper water temperature distribution. It was found that meridionalmigrationof the front ismore active in thewesternpart of the Japan/EastSea than in the eastern part. The interannual variation in the location ofthe front causes large variationsof sea level andupper layerheat contentin the western part.

The effects of total warm water volume change and wind stressvariation on front migration were examined using a reduced-gravitymodel. The total warm water volume change in the basin inducesrelatively small migration of the front around its mean position in themodel. Unsteady wind stress from NOGAPS induces migration of themodeled front between 37°N and 39°N in the western part of thebasin. The combination of these two factors enhances the migration ofthe front across its mean position.

To find wind factors responsible for inducing the interannualmigration of the front, zonal andmeridional components of wind stressandwind stress curl were analyzed. Themeridional wind stress, and thewind stress curl have relatively small variations in time. Interannualvariation of zonal wind stress over the southwestern part may beresponsible for thewide migration of the front. Positive anomalies fromJuly 1995 to April 1996may increase the southward Ekman transport inthe reduced-gravitymodel; negative anomalies fromNovember 1997 toDecember 1999 may enhance the northward Ekman transport.

Our numerical model simulations did not include eddy generation,deep circulation and topographic effects. High variability of sea levelin the Yamato Basin, where generation and movement of eddies areactive, needs to be investigated using a high resolution multi-layeredor three-dimensional model.

Acknowledgements

We appreciate support from the Institute of Marine and CoastalSciences, Rutgers—the State University of New Jersey. This work wassupported by the Korean Meteorological Agency (ARGO program) andthe Korea Research Foundation Grant (KRF-2005-070-C00142). The CLSSpaceOceanographyDivision supplied themergedT/PandERS-1/2data.JODC and KODC provided the upper water temperature data. Dr. S. J. Lyugranted access to the inflow transport through the Korea Strait. NOGAPSwind stress datawere obtained from the Center for Ocean-AtmosphericPrediction Studies, Florida State University.

References

Belkin, I., Krishfield, R., Honjo, S., 2002. Decadal variability of the North Pacific PolarFront: subsurface warming versus surface cooling. Geophys. Res. Lett. 29 (9), 1351.doi:10.1029/2001GL013806.

Belkin, I., Cornillon, P., 2003. SST fronts of the Pacific coastal and marginal seas. PacificOceanogr. 1, 90–113.

Chang, K.-I., Ghil, M., Ide, K., Lai, C.-C.A., 2001. Transition to aperiodic variability in awind-driven double-gyre circulation model. J. Phys. Oceanogr. 31, 1260–1286.

Choi, B.-J., Haidvogel, D.B., Cho, Y.-K., 2004a. Nonseasonal sea level variations in theJapan/East Sea from satellite altimeter data. J. Geophys. Res. 109, C12028.doi:10.1029/2004JC002387.

Choi, B.-J., Iskandarani, M., Levin, J., Haidvogel, D.B., 2004b. A spectral finite volumemethod for the shallow water equations. Mon. Wea. Rev. 132, 1777–1791.

Chu, P.C., Chen, Y.C., Lu, S.H., 1998. Temporal and spatial variabilities of Japan Sea surfacetemperature and atmospheric forcings. J. Oceanogr. 54, 273–284.

Chu, P.C., Lan, J., Fan, C., 2001. Japan Sea thermohaline structure and circulation. Part I:climatology. J. Phys. Oceanogr. 31, 244–271.

Ducet, N., Le Traon, P.Y., Reverdin, G., 2000. Global high-resolution mapping of oceancirculation from the combination of T/P and ERS-1/2. J. Geophys. Res. 105,19,477–19,498.

Gordon, A., Giulivi, C., Lee, C.M., Furey, H.H., Bower, A., Talley, L., 2002. Japan/East Seaintrathermocline eddies. J. Phys. Oceanogr. 32, 1960–1974.

Hirose, N., Fukumori, I., Yoon, J.-H., 1999. Assimilation of TOPEX/POSEIDON altimeterdata with a reduced gravity model of the Japan Sea. J. Oceanogr. 55, 53–64.

Hirose, N., Ostrovskii, A.G., 2000. Quasi-biennial variability in the Japan Sea. J. Geophys.Res. 105, 14,011–14,027.

Hogan, P.J., Hurlburt, H.E., 2000. Impact of upper ocean–topographical coupling andisopycnal outcropping in Japan/East sea models using 1/8° to 1/64° resolution.J. Phys. Oceanogr. 30, 2535–2561.

Hogan, P.J., Hurlburt, H.E., 2005. Sensitivity of simulated circulation dynamics to thechoice of surface wind forcing in the Japan/East Sea. Deep Sea Res. 52, 1464–1489.

Holloway, G., Sou, T., Eby, M., 1995. Dynamics of circulation in the Japan seas. J. MarineRes. 53, 539–569.

Isoda, Y., 1994. Warm eddy movements in the Eastern Japan Sea. J. Oceanogr. 50, 1–15.Isoda, Y.S., Saitoh, S., Mihara, M., 1991. SST structure of the polar front in the Japan Sea.

In: Takano, K.P. (Ed.), Oceanography of Asian Marginal Seas. Elsevier OceanographySeries, vol. 54. Elsevier, Amsterdam, pp. 103–112.

Kim, C.-H., Yoon, J.-H., 1996. Modeling of the wind-driven circulation in the Japan Seausing a reduced gravity model. J. Oceanogr. 52, 359–373.

Kim, Y.-G., Kim, K., 1999. Intermediate waters in the Japan/East Sea. J. Oceanogr. 55,123–132.

Kim, K., Cho, Y.-K., Choi, B.-J., Kim, Y.-G., Beardsley, R.C., 2002. Sea level variability atUlleung Island in the East (Japan) Sea. J. Geophys. Res. 107 (C3), 3015. doi:10.1029/2001JC0008.

Korotaev, G., Oguz, T., Nikiforov, A., Koblinsky, C., 2003. Seasonal, interannual, andmesoscale variability of the Black Sea upper layer circulation derived from altimeterdata. J. Geophys. Res. 108 (C4), 3122. doi:10.1029/2002JC001508.

Le Traon, P.Y., Nadal, F., Ducet, N., 1998. An improved mapping method of multisatellitealtimeter data. J. Atmos. Oceanic Technol. 15, 522–533.

Lyu, S.J., Kim, K., Perkins, H.T., 2002. Atmospheric pressure-forced subinertial variationsin the transport through the Korea Strait. Geophys. Res. Lett. 29, 1294. doi:10.1029/2001GL014366.

Lyu, S.J., Kim, K., 2003. Absolute transport from the sea level difference across the KoreaStrait. Geophys. Res. Lett. 30, 1285. doi:10.1029/2002GL016233.

Minobe, S., Sako, A., Nakamura, M., 2004. Interannual to interdecadal variability in theJapan Sea based on a new gridded upper water temperature dataset. J. Phys.Oceanogr. 34, 2382–2397.

Mooers, C.N.K., Kang, H.-S., Bang, I., Snowden, D.P., 2006. Some lessons learned fromcomparisons of numerical simulations and observations of the JES circulation.Oceanography 19, 86–95.

362 B.-J. Choi et al. / Journal of Marine Systems 78 (2009) 351–362

Moore, J.K., Abbott, M.R., Richman, J.G., 1999. Location and dynamics of the AntarcticPolar Front from satellite sea surface temperature data. J. Geophys. Res. 104,3059–3073.

Morimoto, A., Yanagi, T., 2001. Variability of sea surface circulation in the Japan Sea.J. Oceanogr. 57, 1–13.

Morimoto, A., Yanagi, T., Kaneko, A., 2000. Tidal correction of altimeter data in the JapanSea. J. Oceanogr. 56, 31–41.

Naganuma, K., 1977. The oceanographic fluctuations in the Japan Sea. Mar. Sci. (KaiyoKagaku) 9, 137–141 (in Japanese with English abstract).

Ou, H.W., Gordon, A., 2002. Subduction along a midocean front and the generation ofintrathermocline eddies: a theoretical study. J. Phys. Oceanogr. 32, 1975–1986.

Park, K.-A., Chung, J.Y., Kim, K., 2004. Sea surface temperature fronts in the East (Japan)Sea and temporal variations. Geophys. Res. Lett. 31, L07304. doi:10.1029/2004GL019424.

Preller, R.H., Hogan, P.J., 1998. Oceanography of the Sea of Okhotsk and Japan/East Sea,Coastal segment (11,S). In: Robinson, A.L., Brink, K.H. (Eds.), The Sea, vol. 11. JohnWiley & Sons, New York, pp. 429–481.

Roden, G.I., Taft, B.A., Ebbesmeyer, C.C., 1982. Oceanographic aspects of the Emperorseamounts region. J. Geophys. Res. 87 (C12), 9537–9552.

Rosmond, T.E., 1992. The design and testing of the Navy Operational Global AtmosphericPrediction System. Weather Forecast. 7, 262–272.

Senjyu, T., Sudo, H., 1994. The upper portion of Japan Sea proper water, its source andcirculation as deduced from isopycnal analysis. J. Oceanogr. Soc. Japan 50, 663–690.

Stammer, D., 1997. Steric and wind-induced changes in TOPEX/POSEIDON large-scalesea surface topography observations. J. Geophys. Res. 102, 20,987–21,010.

Talley, L.D., Min, D.-H., Lobanov, V.B., Luchin, V.A., Pnomarev, V.I., Salyuk, A.N.,Shcherbina, A., Tishchenko, P.Y., Zhabin, I., 2006. Japan/East Sea water massesand their relation to the sea's circulation. Oceanography 19, 32–49.

Tomczak, M., Godfrey, J.S., 1994. Regional Oceanography: An Introduction. Pergamon,Oxford.

Uda, M., 1938. Researches on “siome” or current rip in the seas and oceans. Geophys.Mag. 11, 307–372 (in Japanese).

Vivier, F., Kelly, K.A., Thompson, L., 1999. The contributions of wind forcing, waves, andsurface heating to sea surface height observations in the Pacific Ocean. J. Geophys.Res. 104, 20,767–20,788.

Yoshikawa, Y., Awaji, T., Aktomo, K., 1999. Formation and circulation processes ofintermediate water in the Japan Sea. J. phys. Oceanogr. 29, 1701–1722.