WHY DO INTRATHERMOCLINE EDDIES FORM IN THE JAPAN / EAST SEA A Modeling Perspective

Abstract : Intrathermocline eddies (ITEs) are characterized by a subsurface lens of relatively homogeneous water. By definition, they are situated within the thermocline and therefore split the stratified water column, taking the form of a dome in the upper part of the thermocline and a bowl in the lower part. Observations of ITEs in diverse regions of the world ocean indicate typical spatial scales of 10-100 km horizontally and 100 m vertically. In the Japan/East Sea (JES) there are at least three mechanisms for the formation of ITEs from pre-existing non-ITE eddies based on results from the HYbrid Coordinate Ocean Model (HYCOM). Those mechanisms include advection of the stratified seasonal variations of temperature and salinity through the Tsushima Strait, restratification of the upper water column due to seasonal heating and cooling of the upper ocean, and subduction of ITE water originating from the Tsushima Strait beneath the wintertime Subpolar Front. The formation mechanisms are not mutually exclusive. Indeed, all three are shown to be interactively affecting the formation of an ITE in at least one case. Gordon et al. (2002) reported the existence of ITEs in the JES based on observations from SeaSoar instrumentation, conductivity-temperature depth (CTD) sensors, and airborne expendable bathythermographs (AXBTs). Their paper contains extensive analysis of ITEs in the JES and observational evidence of formation mechanisms based on cruise data collected during 1999-2000 as part of the Office of Naval Research (ONR) JES Department Research Initiative as well as results from earlier studies. The Gordon et al. work inspired a numerical modeling study to examine whether or not similar features could be simulated. If they could be simulated, could the ocean model be used as a tool to elucidate the formation mechanisms of the ITEs? This study uses HYCOM to simulate JES ITEs that have domed stratification at the top, forming a lens-shaped interior of nearly unstratified water.

Intrathermocline eddies (ITEs) are characterized by a subsurface lens of relatively homogeneous water.By defi nition, they are situated within the thermocline and therefore split the stratifi ed water column, taking the form of a dome in the upper part of the thermocline and a bowl in the lower part.Observations of ITEs in diverse regions of the world ocean (Kostianoy and Belkin, 1989) indicate typical spatial scales of 10-100 km horizontally and 100 m vertically.In the Japan/East Sea (JES) (Figure 1) there are at least three mechanisms for the formation of ITEs from pre-existing non-ITE eddies based on results from the HYbrid Coordinate Ocean Model (HYCOM).Those mechanisms include advection of the stratifi ed seasonal variations of temperature and salinity through the Tsushima Strait, restratifi cation of the upper water column due to seasonal heating and cooling of the upper ocean, and subduction of ITE water originating from the Tsushima Strait beneath the wintertime Subpolar Front.The formation mechanisms are not mutually exclusive.Indeed, all three are shown to be interactively affecting the formation of an ITE in at least one case.Gordon et al. (2002) reported the existence of ITEs in the JES based on observations from SeaSoar instrumentation, conductivity-temperature-depth (CTD) sensors, and airborne expendable bathythermographs (AXBTs).Their paper contains extensive analysis and discussion of ITEs in the JES and obser-vational evidence of formation mechanisms based on cruise data collected during 1999-2000 as part of the Offi ce of Naval Research (ONR) JES Departmental Research Initiative (DRI), as well as results from earlier studies.The observed ITEs have diameters of ~100 km, thicknesses of ~100 m, and are characterized by a nearly homogeneous core rotating anticyclonically (clockwise) with temperature ~10°C and salinity ~34.12 psu (Figure 2).
The Gordon et al. (2002) work inspired a numerical-modeling study to examine whether or not similar features could be simulated.If they could be simulated, could the ocean model be used as a tool to elucidate the formation mechanisms of the ITEs?Here, we use HYCOM to simulate JES ITEs.Eddies are prolifi c in the JES, and are often associated with quasi-stationary meanders of the East Korea Warm Current (EKWC), Subpolar Front, and the Tsushima Warm Current (TWC) located near the northwestern coast of Japan.The anticyclonic eddies in the JES: (1) can be unstratifi ed from the sea surface down to the thermocline within the eddy (i.e., have a mixed layer throughout the eddy core down to the thermocline), ( 2

THE NUMERICAL MODEL
HYCOM was developed from the Miami Isopycnal Coordinate Ocean Model (MICOM) using the theoretical foundation set forth in Bleck and Boudra (1981), Bleck andBenjamin (1993), andBleck (2002).HYCOM is a hydrostatic      (Fox et al., 2002), which was also used to initialize the model state.

JES INTR ATHERMOCLINE EDDY CHAR ACTERISTICS AND FORMATION MECHANISMS
In earlier work, Hogan andHurlburt (2000, 2005)    of Gordon et al., 2002).This result is not surprising based on the nondeterministic nature of the JES eddies simulated by Hogan and Hurlburt (2005).In the simulation used here, ITEs are most easily identifi ed during the summer-fall by maps of layer-7 thickness, as this is the layer with the target density (26.0 σ θ isopycnal layer) that carries the bulk of the ITE characteristic temperature (~10°C) and salinity (~34.1 psu) (Figure 4d-f and Table 1) in close agreement with the observations reported by Gordon et al. (2002).The ITEs are much less discernible in the winter and early spring when the mixed-layer depth typically extends as deep as the ITE water (Figure 5a-c (Figure 4d).This ITE phenomenon is also seen in observations (e.g., Figure 2) (Gordon et al., 2002).The water in this layer has higher temperature and salinity than the ambient JES water and is advected to the ITEs via the current systems of the JES.During the fall, the ITEs have a clear signature on a map of the thickness of layer 7 (Figure 4).At the locations of the ITEs, layer 7 typically has a central depth of 150-200 m.This layer is always shallower than that north of the Subpolar Front where isopycnal outcropping occurs (Hogan and Hurlburt, 2000).
Interestingly, even north of the Subpolar Front, HYCOM shows evidence of an ITE centered just north of 41°N in Figure 5.It is best seen in the November cross section (Figure 5f) in layers 8     (3) depicts an ITE where it is deepest (Figure 5).

SUMMARY AND CONCLUSIONS
In this paper a numerical model with an advanced vertical coordinate design was used to simulate ITEs in the JES.ITEs are unique, poorly understood features, but were observed in the JES as reported by Gordon et al. (2002) and Talley et al.

Figure 1 .
Figure1.Th e HYbrid Coordinate Ocean Model (HYCOM) simulates the major circulation features in the Japan/East Sea as depicted here by mean sea surface height from model years 7-10 of the simulation.Th e names of the circulation features are overlaid as are the names of key geographical features mentioned in the text.

Figure 2
Figure 2. (a)In May 1999 two intrathermocline eddies (ITEs) were seen in observed cross sections of potential temperature, salinity, and density along 37°45' (top to bottom).Th e ITEs are characterized by a relatively homogeneous interior with a bowl-shaped bottom and a domed top in all three cross sections.(b) When the cross section was repeated during January 2000, the two eddies had lost their domed-shaped tops and could no longer be classifi ed as ITEs.Both (a) and (b) are fromGordon et al. (2002).
retains particular advantages associated with the different vertical coordinate types: (1) the retention of water-mass characteristics for centuries (a characteristic of isopycnal coordinates), (2) high vertical resolution in the surface mixed layer and unstratifi ed or weakly stratifi ed regions of the ocean (a characteristic of z-level coordinates), and (3) high vertical resolution in coastal regions (a characteristic of terrain-following coordinates).The simulation used in this study covers the JES (127.5°E-143°E,34.5°N-52°N) (See Figure 1).The horizontal grid resolution is .04°x .04°,or about 3.5 km.Fifteen layers are used in the vertical, with a target density assigned to each layer (based on climatology) ranging from 23.0 σ θ at the surface to 27.4 σ θ in the abyssal layer, as noted in Table 1.Target densities are chosen so that the top layer is always a fi xed-pressure coordinate, and usually it is the only one present during the stratifi ed summer-fall months.In the winter-spring months, the top six layers are typically at fi xed depths, thereby ensuring adequate resolution of the mixed layer when it is deepest.Sigma (terrain-following) coordinates are allowed in this simulation, but are rarely present because there is little shelf area in the JES.For the simulation described here, the K-Profi le Parameterization (KPP) turbulence closure model of Large et al. (1997) is used to invoke vertical mixing and determine the surface boundary-layer depth.There is no assimilation of ocean data in the simulation used here (except weak relaxation to a monthly surface salinity climatology), and the only forcings are surface wind and thermal fl ux at the ocean surface and fl ow in through the Tsushima Strait and out through the Tsugaru and Soya Straits.For the surface forcing, a monthly climatology from the European Centre for Medium-Range Weather Forecasts (ECMWF) 10-m reanalysis covering 1979-1993 was used, but six hourly sub-monthly wind fl uctuations, derived from September 1994 to August 1995, were superimposed onto the temporally interpolated monthly means.The time period of the high-frequency fl uctuations was chosen for reasons other than those affecting the JES, but served the purpose of adequately energizing the mixed layer (which monthly climatological wind forcing does not).For the lateral boundary conditions, fl ow through the straits was decomposed into an annually constant barotropic (depthaveraged) component and a monthly varying baroclinic component that carried the deviations from the mean in each model layer.For the barotropic part, the infl ows were 1.5 Sv and 0.5 Sv in the western and eastern channels of the Tsushima Strait, respectively.The outfl ow through Tsugaru was 1.3 Sv and through Soya, 0.7 Sv.Thus, the total throughfl ow was 2 Sv.For the baroclinic parts, monthly variations were obtained by relaxing to the Navy's Modular Ocean Data Assimilation System (MODAS) climatology performed extensive investigations of JES circulation dynamics, including the roles of wind and straits forcing, fl ow instabilities, isopycnal outcropping, and the impact of upper ocean-topographical coupling via the fl ow instabilities on the mean circulation and preferred regions of eddy generation.Of particular interest were the dynamics of EKWC separation from the Korean coast, a topic of relevance to the formation of the semi-permanent Patrick J. Hogan (hogan@nrlssc.navy.mil) is Oceanographer, Naval Research Laboratory, Ocean Dynamics and Prediction Branch, Stennis Space Center, MS, USA.Harley E. Hurlburt is Senior Scientist for Ocean Modeling and Prediction, Naval Research Laboratory, Oceanography Division, Stennis Space Center, MS, USA.

Figure 3 .
Figure 3. Th ree ITEs are clearly depicted by mapping the layer thickness (m) between the 8°C and 11°C isotherms (the interior of the ITEs) as observed during the R/V Revelle cruises of May-July 1999 and January-February 2000, and from the Hakuho-Maru in October 1999.Th e color coding shows the mean salinity within this temperature interval.July R/V Revelle stations are shown as red dots; Hakuho-Maru are shown as black dots; SeaSoar tracks of the May and January R/V Revelle cruises are shown as dotted lines.From Gordon et al. (2002).
).During February through April, the JES is subject to signifi cant cold-air outbreaks and the mixed layer south of the Subpolar Front typically extends to ~300-m depth.This deep mixing results in the destratifi cation of many eddies, giving them homogeneous water properties from the surface to the thermocline within the eddy.February through April is also when the coldest and most saline water enters the JES through the Tsushima Strait.Model layer 7 lies above the sill depth of the Tsushima Strait (~204 m; see Preller and Hogan, 1998) from ap-proximately February through August, and is the primary source of highersalinity water for the subsurface water masses in the JES.During this time, the salinity of the eddies within layer 7 increases via mixing along the isopycnal layer (Figure 5).It decreases when the isopycnal cap is removed by deepening of the mixed layer during winter.Starting around May, the surface heat fl ux begins to increase (net heat into the ocean) and the eddies develop a stratifi ed cap (Figure 5c).However, an additional process is required to create an ITE, which has a cap of doming isopycnals.Above model layer 7, doming of the near-surface isopycnals capping an ITE (Figure 5d-f) occurs when advecting plumes of warm saline surface water originating from the Tsushima Strait wrap around the cap of the ITE as illustrated by the layer-6 salinity maps in Figure 4, especially the map for July

Figure 4 .
Figure 4. Bi-monthly, onemonth means of salinity (color) in model layer 6 (the 25.5 σ θ isopycnal layer) and the thickness of ITE interiors (line contours) as represented by model layer 7 (the 26.0 σ θ isopycnal layer).Plumes of high salinity originating from the Tsushima Strait can be seen wrapping around the caps of the ITEs.

Figure 5 .
Figure5.An annual cycle of the Ulleung ITE can be seen in bi-monthly snapshots of cross sections of temperature and salinity along 130.8°E.Th e model layer interfaces are superimposed as line contours and the layer numbers are linked to their target densities in Table1.Th e erosion of the Ulleung ITE cap is most evident in March and the restratifi cation with a domed cap is most evident in November.
Figure 6.Subduction along the Subpolar Front is evident in salinity and temperature from the model and in a January 2000 cross section of salinity from the observations.Subduction is illustrated by January 4 cross sections of (a) temperature and (b) salinity from the model and in (c) by an observed cross section of salinity (color) and temperature (line contours) from January 2000 (the latter fromGordon et al., 2002).In the model, the subduction feeds into an ITE in layer 7 (σ θ =26.00).

Table 1 .
Layer Numbers and Corresponding Target Densities Used in HYCOM to Simulate the ITEs

Table 1 .
Th e erosion of the Ulleung ITE cap is most evident in March and the restratifi cation with a domed cap is most evident in November.