In this paper, results of a high-resolution regional model of the Kuroshio–Oyashio confluence, where the mixed water region (MWR) forms off the northeastern coast of Japan, are discussed. The model simulates major characteristics of the Kuroshio and the Oyashio system well, such as the separation of the Kuroshio Extension from the Japanese coast and southward intrusion of the Oyashio. Further, potential temperature and salinity structures in the intermediate layer σθ = 27.0 resemble those obtained from historical data. Upon the success of this simulation, the authors focus on the diagnosis of the Oyashio water pathways intruding into the subtropics. It is found that the pathways of the Oyashio water form in the vicinity of the Japanese coast, where warm core rings and the Oyashio intrusion are active. These pathways are shown to be primarily eddy driven. Of particular interest is the water that originates in the Sea of Okhotsk, characterized by low potential vorticity (PV). Impacts of the Okhotsk water are identified by conducting an experiment in which the exchange of waters between the Pacific Ocean and the Sea of Okhotsk is blocked. The impacts are striking. If the exchange were blocked, the pathways would not form in the MWR. Instead, a strong cyclonic recirculation, caused by separation of the Kuroshio from the Japanese coast, dominates the MWR and advects warm and salty Kuroshio water northwestward, letting it occupy the entire MWR. It is found that the low-PV flux from the subpolar region tends to reduce the cyclonic circulation in the MWR. As a result, a southward intrusion of the Oyashio is induced. Concurrently, this intrusion blocks the northward advection of the Kuroshio water, maintaining the Oyashio water pathways in the MWR.
The western North Pacific off the east coast of Japan is a crossroads of water masses that are carried by the Kuroshio, Oyashio, and the outflows from marginal seas. The Kuroshio flows poleward and transports a large amount of heat and salt into the midlatitude ocean. The Oyashio is the western boundary current of the subpolar gyre, transporting cold and freshwaters that originated in the Sea of Okhotsk as well as from the east Kamchatka Current. The Kuroshio and Oyashio meet off the east coast of Japan and outflow eastward, forming the Kuroshio Extension (KE) and the Oyashio Front (OF), respectively, exhibiting very complicated current structures. Strong decadal signals have been found in the KE and the OF, and this has generated interest in recent years from the viewpoint of North Pacific climate variations (Qiu 2000; Miller et al. 1998; Nakamura et al. 1997) and their impacts on fisheries (Sugimoto et al. 2001; Noto and Yasuda 1999).
The KE consists of an eastward inertial jet around 35°N and is accompanied by large-amplitude meanders and vigorous pinched-off eddies. In this region, the Sverdrup constraint on the integrated mass transport no longer holds, because of deep recirculation gyres located south (and possibly north) of the KE. The Oyashio flows southward along the northeast coast of Japan as far as 38°N in the climatological mean (e.g., Qu et al. 2001). The Oyashio then forms the OF northeastward from 38°N nearshore to 40°–45°N offshore. The offshore location of the OF is close to the line of the zero wind curl, consistent with Sverdrup theory (e.g., Talley 1993; Hurlburt et al. 1996; Qu et al. 2001).
It should be noted that the KE and the OF are separate fronts. The mixed water region (MWR) forms between these fronts where the Kuroshio water and the Oyashio water meet and are vigorously mixed by active eddies (e.g., Kawai 1972). Mixing of these waters is likely to occur off the northeastern coast of Japan (e.g., Talley et al. 1995; Yasuda et al. 1996), which may involve cabbeling and double-diffusive processes (e.g., Maximenko and Shcherbina 1996; Talley and Yun 2001). The mixed water is then advected eastward in the MWR by the KE and then redistributed in the subtropical and subpolar gyres. Therefore, the MWR is recognized as an origin of the North Pacific Intermediate Water (NPIW) that has a well-defined salinity minimum in the subtropical gyre of the Pacific Ocean (Hasunuma 1987; Talley 1993; Yasuda 1997). A part of the MWR water may also flow poleward across the OF to form a subsurface temperature and salinity maximum (mesothermal layer) in the subpolar gyre (Ueno and Yasuda 2000, 2001).
The Oyashio water properties are strongly influenced by the water originating in the Sea of Okhotsk. The Okhotsk water is characterized by its low potential vorticity (PV) around the 26.8 σθ surface as compared with the surrounding water masses of subpolar origin (Yasuda 1997; Qu et al. 2001; McDonald et al. 2001). The Okhotsk water occupies the upper Oyashio. Yasuda (1997) hypothesized that the low PV is an important property for the upper Oyashio so that it may flow as a coastally trapped density-driven current. Farther to the south, high-density CTD surveys have revealed distinct freshwater pathways that have low PV in the MWR (Fig. 1, after Yasuda et al. 1996). Therefore, the Okhotsk water crosses the OF and modifies the water mass properties of the MWR. The transport of the upper Oyashio across the OF has been estimated to be about 3 Sv (Sv ≡ 106 m3 s−1) (e.g., Talley 1997).
Modeling of the Kuroshio and Oyashio system and their pathways is challenging. It is well known that in many models the Kuroshio overshoots northward and joins the OF around 45°N immediately after it separates from the coast. This is a typical problem for coarse-resolution models in which linear Sverdrup constraints are dominant. In a series of numerical experiments, Hurlburt et al. (1996) have demonstrated that higher resolution greatly improves upper-ocean pathways of the KE, the Oyashio, and the OF. They showed that, once the model resolution becomes high enough so that the baroclinic Rossby radius (∼30 km) is resolved properly, the amplitude of mesoscale fluctuations due to baroclinic instabilities in the KE and the OF becomes comparable to observations. The baroclinic instabilities then produce bottom currents; they interact with bottom topographic features and in turn affect the surface currents. Many aspects of the Kuroshio–Oyashio system such as the KE's quasi-stationary meanders and the Oyashio's southward penetration are likely to be controlled by bottom topographic features.
Processes associated with the formation of the Oyashio water pathways in the intermediate layer have not been modeled properly. Coarse-resolution models (e.g., Yamanaka et al. 1998) showed that the outflow from the Sea of Okhotsk is essential for the NPIW to form. However, their KE and OF coincide so that subtropical subpolar-gyre exchange occurs through diffusive processes. In reality, the KE and the OF are separate fronts having the MWR in between. The exchange occurs via distinct pathways in the MWR as in Fig. 1. Many questions may be raised associated with the pathway formation. For example, why and how can the Oyashio water in Fig. 1 cross the OF? What roles do eddies play? Are the pathways associated with a geostrophic mean flow field or are they associated with eddy transport processes? What are the impacts of the low-PV Okhotsk water on the dynamics of the MWR and the KE?
Here we present results of a high-resolution regional model of the Kuroshio–Oyashio system based on the Princeton Ocean Model (POM). It is shown that the model is capable of representing major features of the Kuroshio and Oyashio system including both surface and intermediate layer circulations. As for the surface circulation, the mean KE and OF positions are regulated by the bottom topography in the present model, consistent with Hurlburt et al. (1996). Based upon the success in simulating the major circulations, we focus our attention on processes in the Oyashio water pathways in the intermediate layer where the Kuroshio water and the Oyashio meet and mix.
The present paper is organized as follows. Section 2 describes the model configuration. Section 3 presents an overview of the model results and climatology. After describing the intermediate-layer circulation in section 4, we discuss processes associated with the Oyashio water pathways in section 5. Section 6 describes impacts of the Okhotsk outflow by conducting an experiment with the exchange between the Sea of Okhotsk and the Pacific Ocean along the Kuril Islands blocked. Section 7 summarizes the results.
2. Numerical model
The model is a version of the sigma-coordinate primitive equation solver (POM) with the domain spanning the Kuroshio along the southern Japanese coast, the Oyashio in the north, and the KE region (Mitsudera et al. 2001; Waseda and Mitsudera 2002). The approximate domain is from 20° to 52°N, 125° to 170°E, configured to have a curvilinear coordinate system in which the horizontal axes follow the mean geometry of the Kuroshio. The set of dynamical and thermodynamic primitive equations is described in Blumberg and Mellor (1983). The horizontal resolution of the model varies between 1/6° and 1/12° within the model domain, and 32 sigma levels are configured in depth.
Bottom topography within the model domain is shown in Fig. 2. This is based on topographic data of 1/12° by 1/12° resolution (NOAA 1988). The data are first interpolated to each model grid point using a Gaussian filter with e-folding scale of 100 km and then smoothed so that the ratio between the difference and the average of two adjacent-grid depths does not exceed 0.35. With this configuration, spurious flows that could be caused by errors in the pressure gradient terms are negligible, as confirmed with a model run with zero forcing (figures not shown here).
Monthly mean climatological runs were made using Hellerman and Rosenstein (1983) wind stress and the Comprehensive Ocean–Atmosphere Data Set (COADS) heat flux (Slutz et al. 1985) as surface boundary conditions. The Levitus (1982) climatology is used for the initial temperature and salinity fields as well as for the lateral boundary restoration at the open boundaries. It is important to note that the Okhotsk water is represented by a sponge layer west of the Kuril Islands. The exchange of water masses between the Pacific Ocean and the Sea of Okhotsk is allowed through the straits along the Kuril Islands. In the present model, this exchange is 4 Sv for both inflow and outflow, with the net transport being zero. We also conducted an experiment in which the exchange is blocked in order to isolate impacts of the Okhotsk water outflow (see section 6).
For the present numerical experiments, we have used a model spun up with a fixed 25-Sv Kuroshio inflow (e.g., Kawabe 1995). The lateral inflow transport at the eastern boundary of the domain is determined by referencing to those obtained from a global high-resolution model by Ishida et al. (1998), where the east Kamchatka Current inflow is 20.5 Sv on average with 28 Sv in winter and 18 Sv in the other seasons (cf. Rogachev 2000). The salient parameter values are summarized in Table 1.
3. Overview of model phenomenology
In this section we provide a brief overview of the surface circulation of the model climatology. Figure 3 represents the SSH fields averaged over seven model years (from the 4th to the 10th model year), the velocity field at 100-m depth, its standard deviation, and depth-averaged streamfunctions.
a. Kuroshio Extension and recirculation gyres
The Kuroshio off the south coast of Japan is very narrow and hence strong, with maximum speeds exceeding 1.5 m s−1. In Fig. 3, the Kuroshio shows a typical nonlarge meander path (Kawabe 1995). The Kuroshio finally separates from the Japanese coast at about 35°N, which is one of the successful features simulated by the model (e.g., Figs. 3a and 3b). As in reality, the model's Kuroshio is followed by the narrow and undulating KE even in the climatology. The stationary undulations of the model KE axis are consistent with those from various recent observations and analyses (e.g., Mizuno and White 1983: Qiu 1995; Qu et al. 2001; Niiler et al. 2003).
Associated with the undulations are well-defined anticyclonic recirculation gyres to the south of the KE. The recirculations are particularly pronounced in the depth-integrated streamfunctions (Fig. 3d), indicating that they have significant deep velocity. The recirculation centering around 34°N, 143°E, at 130 Sv, is strongest (Fig. 3d). This transport is consistent with results from World Ocean Circulation Experiment (WOCE) P10 (Wijffles et al. 1998), measured by a lowered acoustic Doppler current profiler (LADCP), giving more than 150 Sv in this region as an instantaneous value. The recirculation gyres are strongly influenced by bottom topography, that is, seamounts around the recirculation locations, as pointed out previously (e.g., Hurlburt et al. 1996). Figure 3d also shows that cyclonic recirculation gyres exist northward of the KE.
The standard deviation of the model SSH (Fig. 3c) shows large variability at troughs of the KE undulation. In particular, variability of the first and second troughs upstream of 150°E is large, exceeding 30 cm in standard deviation, which is comparable to satellite altimeter observations (e.g., Kuragano and Kamachi 2000; Qiu 1995). Farther to the east, troughs around 155° and 160°E have local maxima in variability, although they are weaker than the observations.
b. Oyashio and Oyashio Front
Another successful feature in this model is the southward intrusion of the Oyashio to the east coast of Japan. As seen in Fig. 3b the Oyashio flows southwestward with considerable strength, exceeding 20 cm s−1, along the Kuril Islands. Figure 4 displays the model vertical structure of the Oyashio off Cape Erimo (42.5°N, 143°E), Hokkaido at the northeast coast of Japan. The current is surface intensified, and the velocity is greater than 20 cm s−1 in the upper 350 m (Fig. 4a). Therefore, the upper Oyashio has a baroclinic character. At the same time it has a deep root over the continental slope, indicating that barotropicity is also important. In the model, the surface current is centered around the shelf break at about 41.7°N. This indicates that the Oyashio could be steered strongly by the bottom slope.
The upper Oyashio contains the Okhotsk water. A salinity section from the model (Fig. 4c) shows coastally trapped low-salinity water (<33.8 psu) down to 400 m and extending 100 km offshore. The model temperature along the same section (Fig. 4b) is almost homogeneous vertically at 3°C. Therefore, the water inshore is lighter than offshore, implying that the upper Oyashio flow is partly a density-driven current. These basic features of the model Oyashio are consistent with observations (e.g., Yasuda 1997; Uehara and Miyake 1999).
When it reaches 38°N off the east coast of Japan, the Oyashio encounters an anticyclonic circulation associated with quasi-stationary warm core rings (WCRs). The Oyashio turns northeastward at that point and flows along the OF afterward (Fig. 3b). By comparing with the topography in Fig. 2 (see, e.g., the 5500-m isobath), we can see that the OF apparently flows along a gentle seamount range in the subpolar region of the western North Pacific. This character of the OF separation is consistent with that found in the high-resolution models by Hurlburt et al. (1996).
In the western subpolar gyre between the Oyashio and the OF, there is a paired northeastward and southwestward flow. This is caused by a chain of anticyclonic eddies propagating from the MWR. This warm-eddy chain is also an observed feature (Yasuda et al. 2000; Rogachev 2000), although the model eddies appear to be stronger than those observed.
c. Mixed water region
The MWR lies between the KE and the OF where eddy activity is intense. WCRs are frequently pinched off from the KE and are observed to propagate westward in the MWR. The WCRs then tend to stay off the northeastern coast of Japan. There are many observations of WCRs in this region (e.g., Kawai 1972; Yasuda et al. 1992; Talley et al. 1995; Okuda et al. 2001). The residence time of observed WCRs is usually long, reviving occasionally through merger among eddies as well as through entrainment of warm water as they interact with the Kuroshio (e.g., Yasuda et al. 1992). In this model, the main WCR stays near 37.5°N and is manifested in the mean field as an anticyclonic circulation (Fig. 3a). The WCRs are of particular interest in this study because they are likely to play a major role in transporting Oyashio water into the MWR and the KE.
The model surface flow is generally eastward in the MWR. Further, Figs. 3a and 3b show the northern branch of the KE approximately along 37°N. There are minima in the zonal velocity along 36°N between the KE main axis and the northern branch. Comparing with the model streamfunctions (Fig. 3d), we can see that those minima are associated with cyclonic recirculation north of the KE. The banded structure is consistent with drifter (Maximenko et al. 1997; Niiler et al. 2002) and XBT (e.g., Mizuno and White 1983) observations. Northward bifurcation of the KE is seen around 34°N, 155°E, which indicates significant influence of the Shatsky Rise on the surface flow (e.g., Hurlburt and Metzger 1998).
Figures 5a and 5b show the vertical section of potential temperature and salinity of the model along 155°E. Temperature and salinity structures are simulated well in comparison with historical data (e.g., Fig. 9 of Qu et al. 2001). In both model and observations, there is a large tilting of the thermocline around 35°N associated with the KE. The thermocline becomes relatively flat immediately north of the KE and tilts again farther to the north. This corresponds to the banded double-jet structure of the KE. The MWR is bounded by the temperature and salinity front corresponding to the OF, located around 43°N. In the salinity section, a salinity minimum is the most prominent feature, which extends from the subpolar region to the MWR. This occurs around the 27.0-σθ surface in this model instead of 26.65–26.8 σθ as in reality.
4. Intermediate-layer circulation
We now focus our attention on the intermediate circulations. As shown in the previous section, the salinity minimum forms around the 27.0-σθ surface in this model. We will therefore examine the water properties on the 27.0-σθ surface, in order to investigate the circulation of the intermediate layer and subpolar–subtropical exchanges.
Figures 6a–d show the climatology of potential temperature, salinity, flow field, and PV, respectively, on the 27.0-σθ surface. In order to describe the flow field, and approximate form of the Bernoulli function, Γ(σθ), is defined to be (e.g., Rothstein et al. 1998)
where η is the surface elevation and ρ is the density. This depicts the geostrophic flow field referenced to the variable surface pressure. Further, we define the potential vorticity Q such that
where ζ denotes the relative vorticity.
Figures 6a and 6b show that the Kuroshio and southern recirculation gyre contain a warm and saline water mass, while the subpolar region is occupied by cold and fresh water. The potential temperature and salinity have large gradients in the MWR, which is also a marked character of the observed MWR (see Fig. 6 of Qu et al. 2001). This implies that the cold freshwater from the subpolar gyre crosses the OF and mixes with the Kuroshio water in the MWR. Further, the potential temperature and salinity decrease eastward in the KE and the southern recirculation gyre, indicating intrusion of Oyashio water into the subtropics.
The circulation field on this isopycnal surface (Fig. 6c) is similar to that of the SSH (Fig. 3a). The KE can be identified as a relatively narrow front around 35°N. The Oyashio and the OF are also clearly seen in the Γ field. In particular, the OF that runs northeastward from 38°N, 147°E, is evident.
The MWR is relatively quiescent in the climatological velocity field except near the western boundary region. As in the surface flow field, a banded zonal flow structure consisting of the KE main axis and northern branch is represented with a relatively stagnant region in between (Fig. 6c). Near the western boundary, however, relatively strong flow associated with a pair of cyclonic and anticyclonic circulations exists. In comparison with the SSH, the cyclonic recirculation is very evident in the intermediate layer, because it shows closed contours in Γ.
Consistent with observations, the cross-frontal pathways from the subpolar region to the subtropics, represented by the low-PV water of about 1.0 × 10−10 m−1 s−1 (Fig. 6d), are well exhibited in this model. The pathway along the coast is prominent. We can also see low-PV water in the northern MWR extending eastward along the OF. Further discussion on this cross-frontal pathway will follow in the next section.
Much of the low-PV water is finally entrained into the KE near Japan and advected eastward, resembling the observation shown in Fig. 1. It then spreads both southward and northward out of KE by eddies and meanders. This process is an efficient mechanism for the Oyashio water to be redistributed as discussed in Yasuda et al. (1996). In this model, the redistribution occurs around promarily 150°–160°E, west of the Shatsky Rise. The southward branch is a source of the NPIW in the subtropical gyre, which is well represented in this model. To the north, there is a well-defined northeastward pathway of the low-PV water around 155°E, bifurcating from the KE main axis into the MWR. This implies a large influence of the Shatsky Rise, consistent with recent observations (e.g., Yoshinari et al. 2001) and modeling (e.g., Hurlburt and Metzger 1998). Similar cross-frontal exchange processes were observed in the Gulf Stream (e.g., Bower and Rossby 1989; Watts et al. 1995).
In Fig. 7 we summarize the water properties on the 27.0-σθ surface in the MWR, averaged over the box 36°–42°N, 145°–160°E. Figure 7 also displays the observed properties presented in Qu et al. (2001). The model simulates the overall properties of the MWR water well. The potential temperature, salinity, PV, and depth in the model are 4.67°C, 34.08 psu, 1.17 × 10−10 s−1 m−1, and 510 m. The analysis by Qu et al. (2001) indicates that the potential temperature and salinity are 4.22°C and 34.04 psu. Thus the model water is slightly warmer and more saline. Depths of the 27.0-σθ surface are comparable between the model and observations. Observed PV is 1.1 × 10−10 s−1 m−1, and the model value is slightly larger.
5. Cross-frontal pathways of the Oyashio water
a. Mean pathways
A striking feature simulated in this model is the well-defined pathways from the subpolar region to the subtropics, as displayed in the PV field (Fig. 6d). In this section, we investigate the pathways in more detail.
Figure 8 displays PV on σθ = 27.0 near Japan with the mean geostrophic flow Γ superimposed. The labels A–F in the figure are shown for convenience and are used later in the discussion of the dynamical system; see section 5b(2). The climatology of the Oyashio has two branches off the coast of Japan, a well-known feature from observations (e.g., Kawai 1972). The southward flow along the northern coast represents the Oyashio First Branch, which carries the low-PV water southward from the subpolar gyre. The mean geostrophic flow then separates eastward at about 39°N near A, forming a front that divides the subpolar circulation and the main WCR. It is clearly seen, however, that the low-PV water (<1.1 × 10−10 s−1 m−1) crosses the front and enters the WCR (A to B). Then it flows in the vicinity of the Japanese coast toward C and finally is entrained in the KE. This pathway may be considered as the coastal intrusion of the Oyashio First Branch (e.g., Shimizu et al. 2001). As shown in Fig. 8, this pathway does not follow the mean geostrophic flow, implying that the transport of low-PV water is dominated by eddy processes.
Farther eastward around D (40°N, 146°E), Γ representing the front, deflects southward again. This is the Oyashio Second Branch in this model. The contours then bifurcate around E. One of the bifurcation branches represents entrainment into the WCR. The other bifurcates northeastward, and the low-PV water is advected by the mean geostrophic current along the OF. The Oyashio water along this pathway cools and freshens the water in the northern part of the MWR as in Figs. 3a and 3b. Note that much of this low-PV water exists south of the OF so that it has already crossed the front upstream.
b. Exchange processes
1) Instantaneous fields
Consistent with observations (e.g., Yasuda et al. 1996; Okuda et al. 2001), the mean field implies two types of Oyashio water intrusion: one is a coastal intrusion associated with the Oyashio First Branch, and the other is an offshore intrusion associated with the second branch. These two branches are also seen in instantaneous fields.
We first display an example of the offshore intrusion at the second branch. Figure 9a shows the PV field with the SSH superimposed. The pathway of the Oyashio water from the subpolar region to the KE is evident. One of the most pronounced features is a large, low-PV reservoir at the center of the main WCR around 37.5°N, 145°E.
The potential temperature field (Fig. 9b) demonstrates the formation process of the reservoir. It shows that the southward flow of the cold Oyashio water is first blocked by the WCR. The flow deflects eastward along the front, then intrudes southward around 39.5°N, 146.5°E, marked as a “+,” forming the Oyashio Second Branch. After that, the Oyashio water is entrained into the WCR, coiling around its periphery forming the reservoir. Part of this water flows out from the reservoir and is advected around the coastal side of the cyclonic recirculation. Finally the low-PV water is entrained into the KE.
The Oyashio water intrusion below the WCR is clearly seen in a vertical section (Fig. 9c). The WCR is expressed by the deep isopycnal surface around 37.5°N. The low-salinity water subducts around the 27.0-σθ surface from the salinity front associated with the OF (at about 39.0°N). This water, partly mixed with the Kuroshio water, is found near the center of the WCR, making a freshwater reservoir at about 500 m. This type of entrainment has been observed in many field surveys (e.g., Talley et al. 1995).
Figure 10 displays an example for the coastal intrusion (Oyashio First Branch). Although there is a front at 37.5°N, the Oyashio water crosses the OF in the vicinity of the coast. This pathway is located between a small-scale WCR and the cyclonic recirculation gyre. The circulation around the pathway resembles some observations of the coastal intrusion, for example, by Shimizu et al. (2001), although the origins of the eddies may be different.
In the mean PV field (Fig. 8) we can further recognize a low-PV band around F that connects the main WCR (or the Oyashio Second Branch) and the second crest of the KE meander. Direct intrusion of the Oyashio Second Branch into the KE sometimes occurs, forming eastern low-PV pathways. However, we also note that another type of exchange occurs frequently between the KE and the main WCR in conjunction with the northwestward propagation of pinched-off eddies. Figure 11 shows low-PV water evolution associated with an eddy. In this particular case, the eddy around 35°N, 146°E is first pinched off (Fig. 11a) and propagates along the periphery of the cyclonic recirculation (Fig. 11b). It then arrives at about 37°N (Fig. 11c), merging with the main WCR. We can further see another eddy propagating westward, which also merges with and amplifies the main WCR eventually. The pinched-off eddies also provide the main WCR with the low-PV water from the KE to the main WCR. This is a feedback of the low-PV water. This reduces the PV contrast across the front, making the cross-frontal exchange easier.
2) Considerations from the viewpoint of dynamical systems
A theoretical framework that originates from dynamical systems theory may provide useful insights into understanding the formation of these pathways. Typically, the flow field is best characterized by hyperbolic stagnation points. Each hyperbolic point is an intersection of a pair of bounding streamlines: stable manifolds that attract fluid particles to and unstable manifolds that repels fluid particles from the hyperbolic point (Fig. 12a).
If a flow is steady, the motion of the fluid particles should follow streamlines and no exchange would occur across the bounding streamlines. However, once perturbations are applied to the flow field, the manifolds originating from the hyperbolic points will inter twine to an infinite complexity, and the particle trajectories become chaotic (e.g., Wiggins 1992). Fluid particles then cross the mean bounding streamlines at the hyperbolic points, resulting in vigorous mixing and formation of cross-frontal transport. Application in oceanographic context may be found in Lozier et al. (1997), analyzing RAFOS float trajectories in the Gulf Stream, and in Samelson (1992), Rogerson et al. (1999), and Couliette and Wiggins (2000), exploring mixing in meandering jets and intergyre exchange. An example is shown in Fig. 12a. A periodic perturbation is applied to a flow consisting of a vortex filament and a zonal flow; without perturbation, a hyperbolic point and a homoclinic orbit form (black curve in Fig. 12a). When the vortex strength is periodically oscillated, the manifolds originating from the hyperbolic point tangle with increasing complexity per cycle. In Fig. 12a, the red curve indicates an approximate shape of the manifold after four cycles of oscillation. Note that at this point, the shape of the mean bounding streamline and the location of the hyperbolic point remain as the original (black) curve. Because the manifolds are good indicators of the trajectories of the fluid particles, the crossing of the manifolds and the mean bounding streamline indicates that there was an exchange of particles across the bounding streamline. This is further illustrated in the Poincaré map of particles that were initially released in a square region inside the mean bounding streamline (Fig. 12b). After four cycles of oscillation, the location of the particles shown in red indicates that part of the particles have been discharged outside of the mean bounding streamline through a relatively narrow channel near the hyperbolic point.
Let us restate the discussion made here. Even when a flow has a well-defined mean field and a closed streamline, particles inside the mean bounding streamline can be discharged out of an enclosed region as a result of chaotic transport. The key to understand this is identifying a hyperbolic point in the mean field. Then, it is likely that any perturbation, whether periodic or quasiperiodic, would result in an exchange of particles across the mean bounding streamline in the neighborhood of the hyperbolic point.
Now we discuss the Oyashio intrusion using this concept as a guideline. Figure 12c displays the depth-integrated streamfunctions near the Japanese coast together with the low-PV pathways. The depth-integrated streamfunction is found to characterize the Oyashio water pathways well, possibly due to the fact that PV water is not a passive tracer but a dynamically active tracer, so that the coupling with the deeper current is important in determining its pathways.
We can identify several hyperbolic points in the streamfunctions near the western boundary, labeled as A–F. Here, only hyperbolic points that are related to the intrusion of the Oyashio water are indicated. It should be noted that the labels are the same as those in Fig. 8.
Consistent with dynamical systems theory, the low-PV water pathways tend to form around these hyperbolic points. This is most evident in the coastal-intrusion process across the OF. The coastal intrusion is related to A and B; the Oyashio separates southward at A, and then bifurcates around B toward C as well as toward D. The cross-frontal transport of the low-PV water across the OF occurs around B along the streamfunction where the flow diverges. The feature resembles that around the hyperbolic point, in Fig. 12b, where the chaotic transport occurs along the unstable manifolds.
The hyperbolic point D is likely to characterize the offshore intrusion associated with the Oyashio Second Branch. The process was essentially described in the previous section (and Fig. 9), where hyperbolic points that spontaneously appear associated with eddies, such as that denoted by a “+” in Fig. 9, play a vital role (cf. Poje and Haller 1999; Waseda and Mitsudera 2002).
The point E is located along the second branch, describing the bifurcation to the OF branch and to the WCR branch. The OF branch transports the low PV water to the northern MWR. The WCR branch is further characterized by points F and B, which determine whether the Oyashio water recirculates inside the WCR or flows out from the WCR into the KE. An example is shown in Fig. 9, where we can see the Oyashio water is discharged from the reservoir around B.
In this section, we have suggested that a possible mechanism for the formation of the Oyashio water pathways is chaotic transport. Originating from geometric methods of dynamical systems theory, the key for such analysis is to identify hyperbolic points in a two-dimensional flow field. Unlike the simple analytical flow example in this section (Fig. 12a), finding a well-defined hyperbolic point in a mean velocity field is a major challenge when analyzing eddy-resolving model output. In fact, finite-time approximations of the manifolds were made by Poje and Haller (1999) to overcome this difficulty when analyzing the velocity field of a ring formation from a jet. It is, therefore, rather amazing that we were able to identify hyperbolic points that may play an important role in determining the Oyashio water pathways in the mean streamfunction. This implies that the WCRs are relatively stationary off the coast of Japan in this model. Quasi-permanent eddies are well known from observations (e.g., Kawai 1972). The eddies tend to propagate northward slowly, continually being reshaped by merger and cooling (e.g., Yasuda et al. 1992). We will further discuss the WCR in the mean field, comparing it with observations in section 7.
3) Eddy fluxes and the pathways
The role of turbulent flows and eddies in forming the Oyashio water pathways may be illustrated by eddy fluxes across the mean geostrophic flow. A question here is whether the eddy fluxes across streamlines have any relationship with hyperbolic points. According to dynamical systems theory, the cross-frontal transports preferentially occur around these points, implying that large eddy fluxes may occur there.
In Fig. 13a, we display time series of the eddy heat flux, υ′T′, at the OF near the coast of Japan. Velocity and temperature anomalies, denoted as υ′ and T′, respectively, are also displayed in the figure. The location where the time series is sampled is close to the hyperbolic point A in Fig. 12b, where vigorous cross-frontal transport occurs. Figure 13a indicates that the eddy heat flux is positive most of the time. Further, the heat flux often shows large values in a few hundred days, that is, a mesoscale eddy time scale. This large positive flux is associated with intermittent southward intrusion of the low temperature water (υ′ < 0, T′ < 0) from the subpolar region to the WCR. Since variations of potential temperature and salinity are compensated. on isopycnal surfaces, the low-salinity water is also intermittently transported southward by eddies.
The eddy heat flux is directed primarily northward. Figure 13b shows the mean eddy-heat-flux field, F = (u′T′, υ′T′), on the 27.0-σθ surface. The eddy flux vectors F depict grossly the pathways of the Oyashio water, although the direction is opposite to the Oyashio intrusion. The flux is largest off the Japanese coast around A. Both the coastal and the offshore intrusions across the OF can also be seen.
Cross-frontal heat transport is expressed by the flux normal to the mean geostrophic flow, Fn, indicated by color contours in Fig. 13b. There are several maxima in Fn. They appear to be located adjacent to some hyperbolic points. First, Fn is particularly large around the hyperbolic point A on the OF. We can also see Fn maxima to the north and south of B along the coastal intrusion pathway. These support the idea that the coastal intrusion is mainly driven by WCRs and fluctuations in the Oyashio First Branch. There is relatively large flux across the mean current around the hyperbolic point F as well. This flux is often related to the intrusion of the Oyashio Second Branch. Cross-boundary flux is also seen between the KE and the cyclonic recirculation gyre, although this is mainly associated with the northward eddy flux of the Kuroshio water.
In summary, there are similarities between the features in the eddy heat flux and those in the Oyashio water pathways. Therefore, the intrusions and the pathways are likely to be predominantly eddy driven. The eddy freshwater flux should show a similar behavior because changes of potential temperature and salinity compensate each other on an isopycnal surface.
c. Cross-frontal transport of the Oyashio water
The properties of the model intermediate water on the 27.0-σθ surface have realistic values, as shown in Fig. 7. Transport of the subpolar water across the OF may thus be estimated using a method that assumes an isopycnal mixture of the two water masses originated in the subtropical and subpolar regions (e.g., Talley et al. 1995; Shimizu et al. 2001). Here, the subtropical water mass is taken off the south coast of Japan, and the subpolar water mass is taken inside the Sea of Okhotsk from the Levitus (1982) climatology. Only the eastward flow is considered here. Southward from the OF (43°N, 155°E) in the layer between 26.6 and 27.4 σθ, we obtain 3.1 Sv (Sv ≡ 106 m3 s−1) as the cross-frontal transport of the Oyashio water out of the total transport of 13.4 Sv across the 155°E section. This transport is comparable with 3 Sv that Talley (1997) estimated from hydrographic sections along 152°E. The total transport of the model is also similar to geostrophic transport estimates of Talley (1997) using several reference levels. Therefore, the model represents the cross-frontal transport of the subpolar water quite well.
6. Impacts of the Okhotsk water on the intermediate-layer circulation
One of the essential components of the Kuroshio and Oyashio system is the outflow from the Sea of Okhotsk. This outflow is thought to transform into the new NPIW in the MWR by mixing with old NPIW from the subtropics (e.g., Talley et al. 1995; Yasuda et al. 1996). Previous numerical models also showed that the NPIW would not form properly unless the outflow from the Sea of Okhotsk is included (Yamanaka et al. 1998). It is not known, however, what impact the low-PV water of the Okhotsk Sea has on the dynamics in the Kuroshio and Oyashio confluence zone. We will try to address this question in this section. Here, the low-PV water originating in the Sea of Okhotsk is specifically referred to as the Okhotsk water, noting that it occupies the upper part of the Oyashio water in the subpolar region.
In order to isolate the effects of the Okhotsk water on dynamics, we conducted an experiment in which the exchange between the Pacific and the Sea of Okhotsk (Okhotsk Sea exchange) is blocked, and we compared the results with those of the case discussed in the previous sections. We label the case with the exchange “EX” and the one without the Okhotsk Sea exchange as “NEX.” Note that the net exchange is zero in both cases because, even in the EX case, inflow and outflow of transport are equal because of mass conservation in the modeled Okhotsk Sea. In addition, the subpolar gyre in both cases is forced by the same Oyashio inflow– outflow transport at the eastern boundary of the model (see Section 2). For these reasons, the NEX streamfunction field (not shown) closely resembles the EX streamfunction in Fig. 3d. Therefore, this experiment examines the impacts of the Okhotsk water without changing the total transport in the subpolar region. Here we focus on impacts of the Okhotsk water exchange on the intermediate-layer (27.0 σθ) circulation.
a. Oyashio structure in the absence of Okhotsk Sea exchange
The potential temperature and salinity of the Oyashio in the NEX case are displayed in Fig. 14. Recall that the EX counterpart was shown in Fig. 4. It is found that the water properties near the coast are greatly different between the EX and NEX cases. In the NEX case, temperature is strongly stratified (Fig. 14a); the surface water is warmer than 10°C, and it is about 4°C around 300-m depth near the coast. In contrast, the EX temperature is rather uniform near the coast where the Oyashio water colder than 3°C occupies the layer from the surface to 400 m (Fig. 4b). Further, water saltier than 34.0 psu occupies the coastal region in the NEX case, indicating the influence of the Kuroshio. This is contrary to the EX case in which salinity is less than 34.0 psu in the upper 500 m near the coast.
The differences in the water properties are attributed to the difference in the Oyashio structure. The NEX velocity section (Fig. 14c) indicates that the Oyashio is predominantly bottom trapped in this case. Further, the thermocline tilts upward toward the coast (Fig. 14a). Therefore, the surface current tends to flow northward, advecting warm water from the KE. The Kuroshio water thus dominates in this NEX section, which is far from reality. This clearly indicates that the outflow from the Sea of Okhotsk has strong impacts on the structure of the surface Oyashio in the subpolar region in this model.
b. Intermediate layer circulation in the absence of Okhotsk Sea exchange
Shown in Fig. 15 is the climatology of water properties on the 27.0-σθ surface from the NEX case. The averaged NEX properties in the MWR are also shown in Fig. 7, together with those of the EX case and from observations (Qu et al. 2002).
The impact of the absence of the Okhotsk water is striking. The MWR is entirely occupied by the warm and salty water from the KE (Figs. 15a,b). Figure 7 shows that the averaged potential temperature and salinity in the MWR are 5.8°C and 34.26 psu. These are close to typical values of the Kuroshio water on 27.0 σθ. Therefore, there is almost no flux of the Oyashio water entering into the MWR. Low-temperature and low-salinity water is present along the coastal region, but it does not spread offshore. This is probably because the Oyashio has a bottom-intensified structure (Fig. 14c).
The geostrophic flow (Fig. 15c) indicates that the KE accompanies zonally elongated, strong recirculation both to the north and to the south of the KE. In particular, the northern recirculation in this case is much stronger than that in the EX case (see Fig. 6c). Therefore, although the KE separates at the right location, the strong northwestward flow associated with this recirculation makes the KE water retroflect toward the coast of Japan. The flow then deflects northward and even overrides the Oyashio, preventing the surface Oyashio water from flowing southward. Instead, this flow advects warm and salty water from the KE, letting it occupy the entire MWR.
The PV value is much higher in the NEX case, which is about 1.5 × 10−10 m−1 s−1 on average in the MWR (Fig. 7). Most prominently, the PV inside the cyclonic recirculation north of the KE is particularly high (≥2 × 10−10 m−1 s−1), where it is nearly homogenized. This high PV is most likely produced along the south coast of Japan due to lateral friction, in a similar manner as discussed in Dengg (1993). The PV outside the recirculation is also larger than that of the EX case. This is consistent with the advection of the relatively high PV water from the KE.
The strong cyclonic recirculation after separating from the Japanese coast is a common feature for many high-resolution models, in particular reduced-gravity models. Hurlburt et al. (1996) have shown that bottom topography is important in regulating how the KE pathways connect to the subpolar front. Bottom topography is important in the present model as well (see section 3). In addition to this effect, we will show that the low-PV input from the Oyashio is likely to be important for the intermediate layer pathways to form in the MWR. This will be discussed further in the next section.
c. Impacts of the outflow from the Sea of Okhotsk
1) Potential temperature field
The difference between the EX and the NEX cases in potential temperature, ΔT, on the 27.0-σθ surface is displayed in Fig. 16. Here, ΔT is defined to be the EX case relative to the NEX case, ΔT = TEX − TNEX. The difference is prominent in the northern part of the MWR, where the EX case is colder than the NEX case. Maximum impact is seen around 39°N near the Japanese coast where the cross-frontal exchange occurs in the EX case (see also Fig. 8). This difference extends farther to the east and to the south, showing the extensive influence of the Okhotsk water. The difference in the salinity field is similar because variations of potential temperature and salinity compensate each other on an isopycnic surface.
The difference originates from the flow fields of the two cases (Fig. 6c and Fig. 15c). As we note, the major difference of the flow field is in the strength of the northern cyclonic recirculation. In the EX case, the cyclonic recirculation is relatively weak, while the anticyclonic circulation in the main WCR is strong. Therefore, the Oyashio Second Branch can intrude southward and effectively block the northward advection of the Kuroshio water.
In the NEX case, however, there are no hyperbolic points associated with the Oyashio Second Branch (like F) because of the strong cyclonic circulation centered at 37°N, 144°E in the MWR. The Kuroshio water is directly advected to the north, and then to the east, by the mean geostrophic current in the MWR. Further, the Oyashio tends to retreat northward. Consequently, the Kuroshio water occupies the whole MWR. Therefore, the difference in potential temperature in Fig. 16 reflects the difference in the Oyashio intrusion pathways, even though the pathway changes are confined to the vicinity of the Japanese coast.
2) Potential vorticity and circulation in MWR
What causes the circulation differences in the MWR in the absence of Okhotsk Sea exchange? To address this question, we display the difference in the PV field, ΔQ = QEX − QNEX, as well as the difference in the mean geostrophic flow, ΔΓ = ΓEX − ΓNEX, in Fig. 17. The PV difference, ΔQ, is negative in most of the region, that is, the EX case has smaller PV. The difference is particularly large, |ΔQ| > 0.7 × 10−10 m−1 s−1, near the western boundary centered around 37°N. A relative anticyclonic circulation (ΔΓ > 0) coincides with this large |ΔQ|. The location of the peak difference in PV is farther south of that in the potential temperature field.
The PV difference (ΔQ) is mostly attributed to the difference in the density stratification, ∂ρ/∂z; the density stratification is reasonable in the EX case, while it is too large in the NEX case. On the other hand, the difference in relative vorticity, ζ, hardly affects ΔQ because it is at most 10% of the planetary vorticity, that is, |ζ| ≤ 0.1f (except at the edge of the KE where the relative vorticity can be as large as 20%).
We thus conclude that the influx of the low-PV Okhotsk water weakens the density stratification in the MWR, resulting in a reduction of PV in the cyclonic recirculation. As discussed previously, the influx of Oyashio water occurs through eddy processes. Since the PV becomes nearly homogenized inside closed streamlines of the cyclonic recirculation, the dynamics of the recirculation may be readily influenced by the eddy PV flux (e.g., Jayne et al. 1996). This suggests that the Okhotsk water is not merely a dye useful for visualizing the Oyashio water pathways, but is a dynamically active tracer that can change the MWR circulation.
The discussion above suggests that the MWR circulation may be considered as a sum of the anticyclonic circulation caused by the low-PV Okhotsk water and the cyclonic recirculation produced by the high-PV water from the southern coast of Japan. The hyperbolic stagnation point at the tip of the Oyashio Second Branch, as seen in the EX case (F in Fig. 12c), tends to emerge because of this superposition. Further, the impact of this change on the recirculation does not remain local; the water properties to the east are dramatically modified because of this pathway change, as discussed in the previous section.
In this paper we have presented a simulation of the Kuroshio–Oyashio system, focusing on the processes in the intermediate layer where water masses of subtropical and subarctic origin meet and mix. The model exhibits many realistic features of the confluence, including the Oyashio water pathways into the subtropics. Results are summarized as follows:
The model represents the major features of the Kuroshio and the Oyashio systems successfully. The mean Kuroshio Extension and the Oyashio Front are situated at observed locations, determined by bottom topographic features.
The salinity minimum forms around 27.0 σθ in this model. The water properties on that density surface are well simulated, although the salinity minimum is slightly deeper than in reality.
Pathways of the Oyashio water to the subtropics have been successfully simulated in this model, and they closely resemble those estimated from CTD surveys (e.g., Yasuda et al. 1996).
There are two major intrusion pathways from the subpolar region across the OF: one is associated with the Oyashio First Branch and the other with the Oyashio Second Branch. These are strongly eddy driven as they tend to cross the mean geostrophic streamlines. Typically, the Oyashio water first subducts below warm core rings across the OF, forming a low PV reservoir there. It then flows out and joins the KE near the Japanese coast. This water is advected eastward in the KE and redistributed to the MWR and the southern recirculation gyre. Further, there is another pathway in the northern MWR extending from the WCR reservoir. Consistent with dynamical systems theory, these pathways tend to form around hyperbolic stagnation points. Cross-frontal transport of the Oyashio water is estimated to be about 3.1 Sv, comparable to estimates from observations.
The surface Oyashio water is strongly influenced by the water originating in the Sea of Okhotsk, characterized by low potential vorticity. Impacts of the Okhotsk water are isolated by conducting an experiment in which the exchange of water mass between the Pacific Ocean and the Sea of Okhotsk is blocked. The impacts are striking. In the absence of the exchange, the warm and salty water originating from the Kuroshio occupies the whole region of the MWR. This is a consequence of the bottom-intensified structure of the Oyashio in the absence of exchange; the Oyashio water hardly spreads offshore even though it reaches beyond the northeastern coast of Japan.
The intrusion of the low-PV Okhotsk water into the MWR causes dynamical impacts on the circulation there. It tends to induce an anticyclonic circulation (or to reduce the cyclonic circulation) in the MWR, which consequently blocks the northward advection of the Kuroshio water, and enhances the southward intrusion of the Oyashio water at the first and second branches.
In this paper, we have discussed the importance of WCRs off the northeastern coast of Japan in the formation of the Oyashio water pathways. The model exhibits a WCR in the mean flow field, and hence the Oyashio water pathways form close to the hyperbolic points associated with the WCR. One may ask whether we can find the mean WCR and hyperbolic points in reality, although quasi-permanent WCRs off the northeast coast of Japan are well known from many observations (e.g., Kawai 1972; Yasuda et al. 1992; Talley 1995). As an example we display the absolute sea level estimated from drifters and satellite altimeters (see Niiler et al. 2003) integrated over one year in 1994 (Fig. 18). The closed contours around 38°N, 144°E indicate the existence of a mean WCR and the hyperbolic points around it. This reveals that the WCR was maintained for about a year over the region. The SSH field in Fig. 18 resembles that in the model (Fig. 3a); after separating from the coast, the Oyashio interacts with the WCR and forms the second branch to the east of the WCR. Hyperbolic points are likely to form between the WCR and the KE, potentially characterizing pathways that connect the water below the WCR with the KE. Therefore, the observations also suggest a cross-frontal transport mechanism similar to the one discussed in this paper.
This is the best observational evidence of a near-stationary WCR that we can provide at this time. A depth-integrated streamfunction could exhibit such a WCR better but it is difficult to estimate from observations. Another difficulty arises because, in nature, there are longer time variations that could obscure the WCR feature and the hyperbolic points in the mean SSH field when taken over time much longer than a year. For example, individual WCRs tend to propagate northward along the northeastern coast of Japan (e.g., Yasuda et al. 1992), so that the WCR feature becomes vague in the long-term mean. Further, the SSH decreases in winter due to strong cooling, which makes it difficult for the WCR to be manifested in the SSH field after the winter. Nevertheless, the suggested mechanism for the transport of the Oyashio water to the KE would work efficiently as far as the WCR exists (e.g., Talley et al. 1995) and the cumulated transport due to several of those events may be significant. We can foresee the possibility of extending the finite-time approximation of the Lagrangian manifold by Poje and Haller (1999) to allow for the evolution of lobes in an unsteady, nonperiodic flow, although this is well beyond the scope of this work.
Why does the salinity minimum layer not form on the 26.8-σθ surface in this model, but on the 27.0-σθ surface instead? Figure 19 displays the PV on the 26.8-σθ surface of this model. The PV is considerably higher in the MWR when compared with those in both subpolar and subtropical regions. That is, the PV is maximum in the MWR, and hence the MWR acts as a barrier for interfrontal exchanges rather than a blender (Bower and Rossby 1989). This barrier works for the main WCR off the east coast of Japan when it entrains the Oyashio water below (see Fig. 10c); the water around 27.0 σθ and below is entrained selectively, while the water on 26.8 σθ is blocked.
We note here that the source of the high PV is located along the south coast of Japan, produced frictionally in the lateral boundary layer. It does not originate from the sea surface. This suggests that careful modeling of the Kuroshio and the frictional boundary layer may be needed for the simulation of the subpolar–subtropical pathways. The Kuroshio in the present model may be too deep along the south coast of Japan, exceeding 20 cm s−1 at 1000 m.
Improvements in mixing parameterizations may lead to better simulation of the Oyashio water pathways. The Oyashio water in reality often keeps its pure characteristics along the pathways in the MWR (Talley et al. 1995; Yasuda et al. 1996; Shimizu et al. 2001). As compared with this, potential temperature and salinity in the model appear to be diffusive. Further, cabbeling and double diffusion should be considered for the transformation of the Oyashio water and NPIW formation (Talley and Yun 2001).
This work suggests significant influences of the low-PV water in the intermediate layer on the dynamics of the recirculation gyres. In the future, we will continue our investigation on the sensitivity of the recirculation gyres to the flux of the low-PV water from the Sea of Okhotsk. Another factor to be considered is inclusion of the Tsugaru Warm Current, which outflows from the strait between Hokkaido and Honshu around 41°N. It shows a characteristic water mass around 26.4–26.7 σθ (e.g., Talley et al. 1995). Transport of this current is comparable to the cross-frontal transport of the Oyashio water. Therefore, its impact could be significant, although it scarcely spreads offshore. These issues await further studies.
Our research benefited from discussions with Profs. I. Yasuda, K. Ohshima, B. Qiu, M. Endoh, and T. Yamagata. We also thank Dr. Potemra for a critical reading of the manuscript. The new absolute SSH is kindly provided by Dr. Maximenko, which is greatly appreciated. The manuscript was improved significantly by the editor and reviewer comments. This work was initiated as a part of the Midlatitude Ocean Project of the Ocean Research Department, JAMSTEC and was also supported by the Frontier Research System for Global Change through its sponsorship of the International Pacific Research Center (IPRC). Author HM completed this work at ILTS of Hokkaido University.
Corresponding author address: Humio Mitsudera, Institute of Low Temperature Science, Hokkaido University, Nishi-8, Kita-19, Sapporo 060-0819, Japan. Email: firstname.lastname@example.org
School of Ocean and Earth Science and Technology Contribution Number 6309 and International Pacific Research Center Contribution Number 252.