On the eddy-kuroshio interaction: Meander formation process

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 108, NO. C7, 3220, doi: /2002jc001583, 2003 On the eddy-kuroshio interaction: Meander formation process Takuji Waseda and Humio Mitsudera Frontier Research System for Global Change, and International Pacific Research Center, SOEST, University of Hawaii, Honolulu, Hawaii, USA Bunmei Taguchi Department of Meteorology, SOEST, University of Hawaii, Honolulu, Hawaii, USA Yasushi Yoshikawa Japan Marine Science and Technology Center, Yokosuka, Japan Received 2 August 2002; revised 27 February 2003; accepted 7 April 2003; published 9 July [1] The Kuroshio flows along the southern coast of Japan during its non-largemeandering state, then separates from the coast near the Kii Peninsula and attaches at the north of the Izu Ridge. The amplitude of the offshore displacement of the Kuroshio changes as a result of mesoscale perturbations. Satellite SSH and SST observations (TOPEX/Poseidon: T/P and NOAA AVHRR) have suggested that the short-term Kuroshio meander formation is triggered by anticyclonic eddies originating in the Kuroshio Extension. To simulate such an event, a numerical experiment using a highresolution regional GCM was conducted by initializing the eddy with the observed T/P sea level anomaly. An anticyclonic eddy was injected to the south of the Kuroshio (140 E, 30 N) by means of sequential data assimilation of the T/P data of October 1992 for 30 days. The volume transport of the model Kuroshio was kept constant at 25 Sv, a condition which would not cause large meander of the Kuroshio. The simulation successfully reproduced the four phases of the interaction: (1) westward propagation of the eddy; (2) advection of the eddy by the Kuroshio; (3) meander formation; and (4) detachment of the eddy from the Kuroshio and their repetition. The analyses revealed that the inshore high potential vorticity (PV) water is generated at the sharp coastal topography (the Kii peninsula). The cyclonic eddies shed will eventually coalesce with the existing inshore cyclonic circulation and the meander grows. During the growth of the meander, the necessary barotropic kinetic energy is produced through the shallowing of the thermocline of the anticyclonic eddy as it elongates and splits. The growth of the meander ceases when the split anticyclonic eddies merge, the thermocline deepens, and the eddy detaches itself from the Kuroshio as a result of its own westward thrust. Simultaneously, the accumulated high PV inshore is released to the Kuroshio Extension region as cyclonic eddies. This study provides evidence for the active role of the anticyclonic eddy in causing the variability of the Kuroshio path and suggests a mechanism of rapid discharge/recharge of the available potential energy of the eddy and production/release of high PV inshore Kuroshio to cause the short-term Kuroshio meander. INDEX TERMS: 4255 Oceanography: General: Numerical modeling; 4520 Oceanography: Physical: Eddies and mesoscale processes; 4528 Oceanography: Physical: Fronts and jets; 4576 Oceanography: Physical: Western boundary currents; KEYWORDS: Kuroshio, meander, anticyclonic eddy, TOPEX/Poseidon, Princeton Ocean Model, AVHRR Citation: Waseda, T., H. Mitsudera, B. Taguchi, and Y. Yoshikawa, On the eddy-kuroshio interaction: Meander formation process, J. Geophys. Res., 108(C7), 3220, doi: /2002jc001583, Introduction [2] The Kuroshio current begins at the east of Luzon Island where the North Equatorial Current branches northward and southward. The Kuroshio then flows northward Copyright 2003 by the American Geophysical Union /03/2002JC east of Taiwan, bends eastward through the Tokara strait, and flows along the Japan coast until it separates from the coast and enters the Pacific basin as a free jet called the Kuroshio Extension. Because of its massive volume and heat, which are transported, its impact on climate is not insignificant, not only to Japan but to the world. Thus the understanding of local Kuroshio dynamics that will lead to 13-1

2 13-2 WASEDA ET AL.: EDDY-KUROSHIO INTERACTION: MEANDER FORMATION Figure 1. Schematic diagram of the Kuroshio path south of Japan. The solid lines, Phase 1 (red), Phase2 (blue), and Phase 3 (yellow), indicate the transition of the Kuroshio path during the short-term meander caused by an anticyclonic eddy A. YT, Yaku-shima and Tanega-shima Islands; CA, Cape Ashizuri; CM, Cape Muroto; KP, Kii Peninsula; BP, Boso Peninsula; SRG, Shikoku Recirculation Gyre; IR, Izu Ridge. The Kuroshio separates from the coast at points indicated by S and reattaches at points indicated by A. an improved predictability of the Kuroshio paths is important. In this study, the focus is on the dynamics of the interaction of the Kuroshio south of Japan and the Kuroshio Extension communicated via mesoscale eddies. [3] The Kuroshio south of Japan is known for the bimodality of its path, switching between a relatively straight and a meandering one [Masuda, 1982; Kawabe, 1985]. It is also known that the state of the path changes at an interdecadal timescale; for example, the Kuroshio preferred a large-meandering state during the mid-1970s and 1980s, whereas in the 1990s, Kuroshio preferred a nonlarge-meandering state [Kawabe, 1995; Qiu and Miao, 2000]. In a non-large-meandering state (Figure 1), a typical path of the Kuroshio is along the coast of Kyushu and Shikoku until it hits the Kii peninsula and separates from the coast [Yamagata and Umatani, 1989]. The flow then takes a cyclonic turn at the Izu ridge and reattaches to the coast between the Izu and the Boso peninsula (see the red line in Figure 1). The tendency of this cyclonic turn of the current as a result of baroclinic effect of the bottom pressure torque (JEBAR) has been discussed recently by H. Mitsudera et al. (Blocking of the Kuroshio meander by baroclinic interaction with the Izu ridge, manuscript in preparation, 2003). They suggest that the Izu ridge acts as a gate that will open and close depending on the bottom flow, unlike the wall at the Hachijo island suggested by Chao and McCreary [1982]. As a result of an interaction of the current with the sharp coastlines and shallow ridge, a semi-permanent recirculating region forms between the Kii peninsula and the Izu ridge at the Enshu-nada sea. As Mitsudera et al. [2001] showed, when the Kuroshio is perturbed by an anticyclonic eddy offshore, the size of the recirculating region can enlarge or reduce in a relatively short time (months). In their study, an active involvement of an anticyclonic eddy in inducing the Kuroshio variability was discussed using altimeter observation, current meter data, and the numerical simulation. The timescale of such an event is very similar to the one illustrated by Shoji as an N-B-C-D sequence describing the short-term variability of the Kuroshio south of Honshu [Shoji, 1972]. [4] The short-term Kuroshio meander formation during the non-large-meander state should be distinguished from the formation of the large meander. There are numerous studies conducted from both observational and theoretical grounds of the triggering of the large meander or of the transition from a non-large-meandering state to a largemeander state [Yamagata and Umatani, 1989; Akitomo et al., 1991; Kawabe, 1995; Masuda et al., 1999; Masuda and Akitomo, 2000; Endoh and Hibiya, 2001]. In the numerical study we conducted, the amplification of the meander is transient and will not lead to a formation of a long-lived meander. As illustrated in Figure 1, a strong anticyclonic eddy A propagates from the east and passes over the Izu ridge (IR) as observed by ADCP sections from the TOLEX repeated ferry cruise line [Hanawa et al., 1996; Ebuchi and Hanawa, 2000]. This eddy will then interact with the Kuroshio and eventually amplify the Kuroshio meander, forming a vortex pair with the cyclonic circulation inshore of the Kuroshio. As previous studies showed, the presence of this anticyclonic eddy may induce generation of trigger meander [Endoh and Hibiya, 2000], but in this study, this was not a necessary condition for the amplification of the meander. However, we will show that the presence of the trigger meander will induce the generation of cyclonic vorticity inshore of the Kuroshio and eventually lead to the formation of the meander. Unlike any of the earlier studies, we will show that this anticyclonic eddy can be detached from the Kuroshio and propagate toward the west, repeating another cycle of short-term meander formation. This will result in a Kuroshio-path evolution illustrated as a sequence of Phase 1, 2, and 3 in Figure 1 that will repeat itself as the anticyclonic eddy A makes a clockwise turn in the Shikoku recirculation gyre. [5] This study focuses on the discussion of the active role of the anticyclonic eddy offshore of the Kuroshio in triggering the short-term variation of the Kuroshio path, in its non-large-meander state. However, various analyses revealed that dynamics of the Kuroshio discussed in the context of large meander formation still apply during the short-term meander formation process. Those are the formation of a trigger meander [Akitomo et al., 1991; Endoh and Hibiya, 2000], a production of potential vorticity due to Kuroshio-topography interaction [Yamagata and Umatani, 1989; Akitomo et al., 1991], baroclinic instability causing the growth of the meander [Endoh and Hibiya, 2001], and release of cyclonic vorticity as the meander disappears [Qiu and Miao, 2000]. Particular attention is given to the role of the Kii peninsula in shedding cyclonic vortices inshore of the Kuroshio, a process which was first suggested by Yamagata and Umatani [1989]. [6] We will first present the observational results from the TOPEX/Poseidon altimeter and the SST from NOAA AVHRR. Results of the eddy-kuroshio interaction experiment are presented in section 3, and the discussion on specific issues is given in section 4. Conclusions follow. 2. Observations [7] Satellite observations have provided the oceanographic community an excellent tool to monitor mesoscale oceanic

3 WASEDA ET AL.: EDDY-KUROSHIO INTERACTION: MEANDER FORMATION 13-3 Figure 2. SSH dynamic height field from TOPEX/Poseidon altimeter data of Kuragano and Kamachi [2000]. The sequences are from 1997 to phenomena with uniform spatial and temporal coverage. With the TOPEX/Poseidon (T/P) altimeter data of this decade, we can now monitor the evolution of the Kuroshio path and various eddy activities in its vicinity. For example, the recent analysis of T/P altimeter data [Kuragano and Shibata, 1997; Kuragano and Kamachi, 2000] shows the interaction of eddies and the Kuroshio during 1997 to In Figure 2, the sea surface dynamic height field derived from the T/P altimetric data is shown for selected dates. On October 1, 1997, an anticyclonic eddy approaches from the east, crosses the Izu ridge, and strengthens on February 4, The eddy then propagates west and collides with the Kuroshio mean current near south of the Kyushu island on May 5, At this point, we can observe that a cyclonic eddy, the so-called trigger meander of the Kuroshio, is formed in reaction to the collision of the anticyclonic eddy with the Kuroshio. These two vortices propagate downstream, eventually triggering the formation of the short-term meander of the Kuroshio (July 4, 1998). Finally, the anticyclonic eddy detaches itself from the Kuroshio current and propagates toward the west as the meander amplitude decreases (September 4, 1998). [8] The Kuroshio path variation during this period as reported by the Oceanographic Prompt Reports of the Japan Marine Safety Agency is shown in Figure 3. This Kuroshio path was estimated from a combination of SST, sea surface velocity, temperature at 200 m, and satellite altimeter data if available. The agreement with the satellite-derived SSDH field (Figure 2) is quite good. [9] The interaction of an anticyclonic eddy and the Kuroshio is also evidenced by in situ observations showing an extraordinary strength of the anticyclonic eddy reported earlier by Mitsudera et al. [2001]. In this paper, and also in the work by Waseda et al. [2002], we have shown the close resemblance of the 1993 short-term meandering event and the 1998 short-term meandering event; both meanders were triggered by an anticyclonic eddy from the Kuroshio Extension region. An overview of various interactions of both anticyclonic and cyclonic eddies with the Kuroshio current as observed during 1993 to 1999 by T/P altimeter, including the case described here, has been reported by Ebuchi and Hanawa [2003]. They have shown from data that an anticyclonic eddy rotating clockwise in the Shikoku Recirculating Gyre (SRG) interacts with both anticyclonic and cyclonic eddies from the Kuroshio Extension. In the former case, the anticyclonic eddies coalesce and trigger the meander. In the latter case, when the cyclonic eddy precedes the anticyclonic eddy in circulating clockwise in the SRG, the meander was triggered, but not otherwise. [10] Another example of satellite observation that has captured the short-term meandering event of the Kuroshio during 1993 is the AVHRR SST image analyzed by Nishimura [1998] (a snapshot on April 19, 1993, Figure 4); the fine structure of the estimated surface velocity field within the SRG is visible. Unlike the majority of numerical simulations of large meander that do not show any flow features smaller than the scale of the SRG itself, an anticyclonic eddy of diameter 300 km is visible within the SRG. The image also indicates the strength of the surface velocity field of the anticyclonic eddy, which is consistent with the in situ observation of the eddy [see Mitsudera et al., 2001]. This observation is yet more evidence of a possible interaction of anticyclonic eddy and the Kuroshio current in triggering the short-term meander. Nishimura [1998] analyzed other satellite images and showed cases where a structure even finer is observed within the cyclonic eddy of the meander. Notably, he has seen a shedding of cyclonic eddy from the Kii peninsula that will eventually coalesce due to inverse cascading and

4 13-4 WASEDA ET AL.: EDDY-KUROSHIO INTERACTION: MEANDER FORMATION Figure 3. Kuroshio path from the Oceanographic Prompt Reports of the Japan Marine Safety Agency during enlarge the cyclonic eddy, that is, amplify the meander (Figure 5). We show from our model result an example of cyclonic eddy shed from the Kii peninsula that eventually will coalesce with the existing cyclonic circulation as Nishimura suggested. This scenario of inverse cascading is somewhat similar to the homogenization of inshore high potential vorticity as discussed by Yamagata and Umatani [1989]. They have shown numerically a possibility of local source of potential vorticity (the Kii peninsula) during the growing phase of the meander. [11] In the following sections, we will investigate the results of the numerical simulation of the short-term meandering event triggered by the anticyclonic eddy. We will also show in detail the processes such as collision of the Figure 4. Surface temperature field and the velocity field estimated from the SST observation by AVHRR, April 19, The image was captured and analyzed by Nishimura [1998].

5 WASEDA ET AL.: EDDY-KUROSHIO INTERACTION: MEANDER FORMATION 13-5 Figure 5. (a) Eddy shedding from the Kii peninsula (Cape Shiono-misaki) observed by AVHRR on April 25, 1983 [Nishimura et al., 1986]. (b) Schematic showing the shed eddies based on the AVHRR image. eddy with the Kuroshio, eddy shedding at the Kii peninsula, and detachment of the eddy from the Kuroshio. 3. Results 3.1. Numerical Model [12] We have used a GCM configured to mimic the Kuroshio/Oyashio circulation system. The model is a version of the s-coordinate primitive equation solver (POM [Blumberg and Mellor, 1983]) developed at JAMSTEC [Mitsudera et al., 1997] to cover the main Kuroshio stream along the southern Japan coast, the Oyashio current in the north, and the Kuroshio extension region; approximate domain is from 125 E to 170 E, 20 N to 52 N, configured to have a curvilinear coordinate system in which the horizontal axes follow the mean geometry of the Kuroshio stream. The horizontal resolution of the model varies between 1/6 degrees and 1/12 degrees within the model domain, and 32 sigma-levels are configured in depth. The model has successfully run in different settings, for example with various Kuroshio inflow rates. At high inflow rates, 35 Sv, the modeled Kuroshio demonstrated bimodality by taking a non-large-meandering or a large-meandering path, but at low inflow rates, 25 Sv, the Kuroshio did not meander. [13] For this study, we have used a model spun up at fixed 25 Sv Kuroshio inflow and a 20.5 Sv Kamchatka inflow condition, a realistic condition based on observations (e.g., Kawabe [1995] and Rogachev, [2000] for the Kuroshio and the East Kamchatka current, respectively), Hellerman-Rosenstein wind [Hellerman and Rosenstein, 1983], COADS heat flux [Slutz et al., 1985], and the Levitus monthly climatologies [Levitus, 1982] were used for the surface boundary conditions, the initial temperature and salinity fields, and the lateral boundary restorations. Within the 6 years of spin-up, a gentle recirculation of about 10 Sv forms within the Shikoku basin (the Shikoku Recirculation Gyre, SRG) (Figure 6) but is not strong enough to cause a stationary large meander. The variability at the extension is not large, and no eddies were shed from the extension region to the west up to year 6. Therefore the model state is calm, and meander does not occur naturally. [14] The detail of the eddy initialization is described by [Waseda et al., 2002], and so here we will just present a short summary. The eddy was initialized for 30 days at 140 E and 30 N by means of optimal interpolation nudging of the TOPEX/Poseidon sea level anomaly data [Le Traon et al., 1995; AVISO/Altimetry, 1996] that captured the same eddy observed during TOLEX by ADCP measurements in 1992 October. In addition to the SSH correction, a statistical correlation is used to map the surface elevation correction to the interior temperature similar to Ezer and Mellor [1994]. At day 30 the assimilation is stopped and the eddy starts to move westward. Its westward propagation speed is faster than that of the first baroclinic Rossby wave. During this westward propagation, the eddy creates barotropic tripolar vortices (from south to north, cyclonic, anticyclonic, and cyclonic) that rotate clockwise centered at the anticyclonic eddy; the energy transfers among these vortices, and eventually a trailing Rossby wave-like wake is left. [15] Understanding of the barotropic field was crucial. Among five tests conducted varying the strength of the assimilation, two of them propagated westward but the other three suddenly migrated south; the injected eddies had similar radius of about 180 km but the thermocline depth varied from 130 to 160 m. The distinct evolution of these eddies was explained as due to mixed effect of nonlinearity, dispersion, and barotropicity. In the current paper, we will analyze the selected result of the simulation that illustrates best how the anticyclonic eddy can trigger the short-term meander. [16] The initialized eddy of the meandering test cases had a diameter of about 360 km and a thermocline displacement of around 150 m (Figure 7). The corresponding Burger number was s = 0.049, and the Rossby number was = Velocities near the surface reached about 70 cm s 1. This initialized eddy is comparable in size to the energetic eddies observed in 1992 and 1997 near Izu ridge along TOLEX (Tokyo Ogasawara Line Experiment) [see Ebuchi and Hanawa, 2000]. These eddies have been identified as being responsible for the triggering of the short-term meandering events in

6 13-6 WASEDA ET AL.: EDDY-KUROSHIO INTERACTION: MEANDER FORMATION Figure 6. (left) Mean transport and (right) averaged potential vorticity plotted against transport of the spun-up model. The thick line in the left plot indicates a region of approximately even potential vorticity. The two thin lines in the left plot bound a region where the potential vorticity was averaged and 1998 [Yoshikawa et al., 1998; Mitsudera et al., 2001] An Overview of the Eddy-Kuroshio Interaction [17] We give an overview of the eddy-kuroshio interactions by a sequence of SSH images presented in Figure 8. In the initial state (day 0), the Kuroshio does not meander. The Kuroshio makes a turn around Yaku-shima and Tanegashima Islands (YT, see Figure 1), flows along Kyushu and Shikoku, detaches from the coast at Boso Peninsula (BP), and flows eastward as a free jet (Kuroshio extension). There is a cyclonic recirculation between Kii Peninsula and Boso Peninsula. The eddy inserted into this relatively calm state moves southwest and eventually collides with the Kuroshio at day 100 at around 28 N, a few degrees south of Yakushima and Tanega-shima Islands (YT). After the collision of the eddy with the Kuroshio, the eddy elongates meridionally due to strong shear of the Kuroshio and is advected downstream, causing the meander of the Kuroshio between the Kii Peninsula and the Izu ridge at around day 130 (not shown). The meander amplitude decreases as the eddy starts to detach itself from the Kuroshio at day 160. The detached eddy moves westward (see day 220) and collides with the Kuroshio again at day 230, slightly above Yaku-shima and Tanega-shima Islands (YT) around 30 N. The eddy is advected downstream, and once again, causes the Kuroshio to meander. [18] The described sequence has a striking resemblance to the observed short-term meandering event during 1998 presented earlier in section 2. In both cases, the latitude where the eddy first collided with the Kuroshio was lower than the second collision. It takes about 60 days from the first collision to a full development of the short-term meander in both observation and simulation. The interval between the first meander and the second meander is around 100 to 120 days. And finally, the amplitude of the meander seems very well reproduced by the simulation. We therefore consider that the simulation has successfully reproduced the observed short-term meandering event. The animated results are presented also in the virtual poster session (T. Waseda et al., Eddy-Kuroshio interaction, URI-ONR virtual poster session, 2000, available at waseda/) (hereinafter referred to as Waseda et al., virtual poster, 2000) Shedding of Cyclonic Eddy Due to Flow Separation [19] The short-term meander event can be viewed as a temporal increase of the size of a separation bubble between Kii Peninsula (KP) and Boso Peninsula (BP) (see Figure 1). A separation bubble refers to a bounded region of flow enclosed by a streamline and the coast between the separation and reattachment points of the Kuroshio from the coast, that is, a coastal region between the Kii peninsula and the Boso peninsula. The term separation bubble is used in analogy to the recirculating flow, for example, at a flow past a backward facing step [Chang, 1970]. The separation bubble can either be closed or opened and when it is opened Figure 7. (left) SSH and (right) temperature section taken along the cut shown in the left plot, of the initialized eddy.

7 WASEDA ET AL.: EDDY-KUROSHIO INTERACTION: MEANDER FORMATION 13-7 Figure 8. SSH sequence from the simulation. it forms a wake. We present here a possible scenario of the initial growth mechanism of this separation bubble, i.e., growth of the meander; we suggest that the growth is accompanied by supply of cyclonic vortices generated at sharp coastlines. [20] We now discuss flow near the two topographic features of interest; these are the islands south of Kyushu, Yaku-shima and Tanega-shima (YT), and the Kii Peninsula (KP). A small stationary cyclonic eddy (trigger meander) that is occasionally found south of the Yaku-shima and Tanega-shima islands (YT) did not form naturally in this simulation, since the flow rates were fixed and also since there were no cyclonic perturbations from the south. Only when an anticyclonic eddy approached, and the local flow rate increased, does the coastline appear sharp and appear to cause shedding of the cyclonic eddies. Between the Kyushu and the Kii Peninsula, there are two sharp capes, Cape Ashizuri-misaki (CA) and Cape Muroto-misaki (CM). These capes are, in this simulation, unnoticed by the Kuroshio main flow because of the recirculation that fills the void between these sharp topographic features creating a slip condition for an outer flow; that is, the coast line appears smooth for an outer flow. However, the Kii Peninsula (KP) is not unnoticed, and the Kuroshio impinges at different angles of attack on Cape Shiono-misaki, the southern tip of the Kii Peninsula and, as a result, cyclonic eddies are shed. [21] When the anticyclonic eddy collides with the Kuroshio south of the Kyushu Islands, the transport along the coast line increases and also directs slightly offshore (days , Figure 9). A number of cyclonic eddies are shed and eventually join to form a sizable cyclonic eddy, Cs in Figure 9. The Kuroshio takes a small meandering path east of Kyushu. The Cs and the anticyclonic eddy form a vortex pair that moves downstream of the Kuroshio. At day 120, the Cs is located just south of Shikoku. Because of Cs, the angle of attack of the Kuroshio on Cape Shionomisaki (KP) changes and a cyclonic eddy C1 is shed downstream. As the Cs passes through the Cape Shionomisaki (KP), both Cs and C1 merge with the pre-existing cyclonic eddy or the separation bubble, and the meander starts to grow. [22] In contrast, no separation eddies are shed at the southern coast of Kyushu at the second encounter of the anticyclonic eddy with the Kuroshio (see days , Figure 10). A notable difference from the first encounter is that the second encounter of the eddy with the Kuroshio occurs a few degrees farther north than the first encounter (compare day 105 and day 230 of Figures 9 and 10). This subtle change in the eddy-kuroshio encounter latitude at the second encounter seems to have oriented the Kuroshio to follow the topographic contour along Kyushu, and no trigger meander was formed. The anticyclonic eddy will then move along in the direction of the Shikoku recircula-

8 13-8 WASEDA ET AL.: EDDY-KUROSHIO INTERACTION: MEANDER FORMATION Figure 9. SSH sequence of the simulation from days Cs and C1 indicate cyclonic eddies formed inshore Kuroshio. tion gyre (SRG), and it eventually induces the growth of the meander forming a vortex pair like structure with the existing cyclonic circulation (day 255) Growth of the Meander [23] The evolution of the anticyclonic eddy during the meander growth stage is rather complicated. Here we compare the first meander between days 135 and 160 and the second meander between days 260 and 285. From the SSH field of days 135 and 155 in Figure 11, we see that as the meander grows, multiple neighboring SSH maxima appear. The peak shifts from one to the other; from A0 to A1 (day 135), and as A0 disappears, another peak appears (A2, see day 150) that would eventually survive (day 155). The eddy seems almost about to break up, but eventually it regains coherence around A2. The eddy centered around A2 will detach itself from the Kuroshio, migrate toward the west, and trigger the second meander (see day 230 of Figure 10). After triggering the second meander (see day 260 of Figure 12), once again, multiple SSH maxima are formed and the peak shifts among them; a new peak A4 emerges, but in this case, the original peak A3 does not disappear, and A3 and A4 coexist for quite a long time (see the sequence of Figure 12). Eventually, two distinct eddies centered at A3 and A4 appear and migrate in opposite directions, A3 westward and A4 eastward (sequence not shown). [24] Possible cause of this distinct evolution pattern of the first and the second meander is the size of the separation bubble. In the former case, the separation bubble remained closed ; that is, the reattachment point of the Kuroshio near the Boso peninsula remained attached to the coast while the eddy was detaching from the Kuroshio. Because the separation bubble remained finite in size, a cyclonic circulation of the inshore Kuroshio water was sustained. On the contrary, during the second meander event, the separation bubble opened and formed a wake (see day 265 to 275 of Figure 12). Simultaneously, the eddy has split into two parts that were centered around A3 and A4. The comparison suggests that when the anticyclonic eddy Figure 10. SSH sequence of the simulation from days

9 WASEDA ET AL.: EDDY-KUROSHIO INTERACTION: MEANDER FORMATION 13-9 Figure 11. SSH sequence of the simulation from days retained its coherence, so did the separation bubble, whereas when the anticyclonic eddy split, the separation bubble opened and formed a wake Detachment of the Eddy From the Kuroshio [25] Now we will switch our attention to the eddy detachment process during the first short-term meander event. The detachment occurs in two stages. First, after day 160 (see Figure 11), the shape of the eddy A2 becomes more circular as the meander amplitude reduces, and the eddy starts to migrate towards the west (see Figure 13). Next, after day 200 or so, the migration speed further increases. It is known that an isolated eddy will have its own westward drift, but in the vicinity of a strong jet, the migration of the eddy will be strongly influenced by the jet. As we have shown in our earlier report [Waseda et al., 2002], the motion of the initialized eddy in this numerical experiment behaved as a typical frontal geostrophic eddy that has an additional drift, in both the east-west and northsouth directions, to the regular westward drift due to beta effect. As quantified by Cushman-Roisin et al. [1990], the correction to the migration speed of an isolated eddy is a function of the amplitude of the displacement of the thermocline and the geostrophic velocity of the lower layer fluid. Thus two effects are important for increasing the westward migration speed of the eddy: one is to increase the amplitude of the displacement and, the other is to have baroclinic effects that will add westward advection speed of the upper layer eddy. The latter can also be interpreted as due to a shift of the baroclinic and the barotropic eddy centers. Figure 14 shows sequences of the thermocline displacements, and Figure 15 shows the stream lines of barotropic and baroclinic flows (the method used to estimate the barotropic and the baroclinic flow field is explained in detail by Waseda et al. [2002]). Prior to the first stage of detachment, the thermocline displacement increased gradually until it gave sufficient strength to the eddy s westward migration. Further acceleration occurs when the barotropic flow field shifts northward and the center of the baroclinic eddy lies on top of the barotropic westward flow. The dynamics of why such a shift occurred is not evident at this point, but we would like to emphasize Figure 12. SSH sequence of the simulation from days

10 13-10 WASEDA ET AL.: EDDY-KUROSHIO INTERACTION: MEANDER FORMATION Figure 13. SSH sequence of the simulation from days that such baroclinic structure of the flow field is important in the eddy-jet interaction. 4. Discussions 4.1. Eddy-Kuroshio Interaction [26] In the previous section, we have described the process of meander formation induced by the interaction of an anticyclonic eddy and the Kuroshio following a sequence of SSH snapshots. Here we attempt to give a more coherent picture of the eddy evolution by monitoring the following indices: (1) strength of the eddy; (2) meander amplitude; (3) barotropic velocity of the Kuroshio; and (4) baroclinic velocity of the Kuroshio. [27] The strength of the eddy is estimated by circulation of the eddy core, Z Z Z ¼ u dr C core ru da; ð1þ where the line integral of the depth-averaged velocity is taken along the SSH iso-line C. Note that C is not a material line but conveniently serves as an integration path to approximate the barotropic relative vorticity of the eddy core. As shown in Appendix A, the strength of the eddy defined as equation (1) will decrease monotonically as the integration path becomes larger. In an ideal situation of an isolated eddy, if the integration paths are sufficiently away from the eddy core, the circulations will reach an equilibrium because the core is the only source of relative vorticity. In a far field, regardless of its complexity, the eddy core (viscous core) appears as a single source of vorticity that can be integrated on any surrounding closed paths. In other words, the circulation does not depend on the choice of path. Thus, in Figure 16c, we have presented the circulation as an average of from 1.5, 1.6, and 1.7 m iso-ssh lines. Then the magnitude of the error bar indicates how far away the integration path was from the eddy core. In other words, the smaller the error bar, the more coherent the eddy core is. Figure 14. Evolution of the thermocline displacement prior to the detachment of the anticyclonic eddy from the meandering Kuroshio. The thermocline depth is defined as the depth of s = 26.5.

11 WASEDA ET AL.: EDDY-KUROSHIO INTERACTION: MEANDER FORMATION Figure 15. Sequence of barotropic streamlines (colored contours) and the baroclinic stream lines (white contours) during the acceleration of the westward migration of the anticyclonic eddy. [28] The amplitude of the Kuroshio was defined as the difference of the offshore distance of 0.7 m SSH isoline at the meander peak and at the downstream meander trough. Large value indicates meander. The barotropic velocity of the Kuroshio is defined as the maximum depth-averaged velocity off the Kii peninsula. At the same location, the difference of the surface velocity and the depth averaged velocity is computed and used as the baroclinic velocity of the Kuroshio. [29] These indices were monitored during the two cycles of meanders until the estimation no longer was possible for the circulation because of the eddy breakup (Figure 16c). In this figure, the strength of the eddy is plotted from days 30 to 260 at 5-day intervals. The circulation is defined to be negative for an anticyclonic eddy; thus, as the anticyclonic eddy strengthens, the value decreases. As the initialized eddy propagates west (between day 30 and 100), the strength remains more or less confined around m 2 s 1 (indicated by a horizontal line in Figure 16c). The strength was the same during the westward migration stage of the second meander event (after day 180), evidencing that we have successfully followed the evolution of the eddy core strength with the chosen SSH iso-lines as an integration path (examples of a 1.5-m SSH isoline are shown in Figures 16a and 16b). This is striking, since the evolution of the SSH alone would indicate that the eddy has almost lost its identity during the meander phase (see section 3.4). What appears to have happened is that the eddy that was on the verge of splitting somehow became coherent again at around day 180 and eventually detached itself from the Kuroshio. This did not happen during the second meander event; at day 260, the eddy broke up. Strictly speaking, the strength of the eddy reduces after day 210 or so (an overshoot is seen), and the implication of this will be explained later. [30] The eddy will be influenced by strong shear as it collides with the Kuroshio. While being advected along the Kuroshio in a clockwise direction, the eddy is elongated parallel to the Kuroshio flow, the circulation dramatically strengthens ( reduces), and the meander amplitude starts to grow (see the change in the strength of eddy after day 105 and day 225). The rapid increase of the cyclonic circulation (reduction of ) suggests some kind of an instability in action. Since the change of the area of the eddy and/or the latitude of the eddy did not seem correlated with the change in the circulation, the decrease of estimated from relative vorticity w indicates that the absolute vorticity w a = w +2 has also reduced accordingly. From the conservation of potential vorticity, d/dt (w a rr/r) = 0, this then implies that the vertical gradient of the density has increased, i.e., the shallowing of the thermocline depth (see Figure 17). If we interpret this as a rapid release of available potential energy, it likely suggests that the anticyclonic eddy went through a baroclinic instability while being advected by the Kuroshio current. As a result, the SSH signature showed multiple peaks (section 3.4). The increase of the kinetic energy of the barotropic (or the lower layer) flow has likely triggered the meander of the Kuroshio. The release of the available potential energy to the barotropic kinetic energy can be inferred from the velocity field at 1000 m depth (Figure 18). The flow orients southward (normal to the Kuroshio axis indicated by solid line) around day 120 when the meander amplitude is growing; thus the upper ocean Kuroshio is advected south. Prior to day 120, the flow is parallel to the Kuroshio axis, and the circulation is cyclonic inshore the Kuroshio. At day 120, the lower layer circulation reverses and triggers the meander.

12 13-12 WASEDA ET AL.: EDDY-KUROSHIO INTERACTION: MEANDER FORMATION Figure 16. SSH iso-lines at 1.5 m used to estimate the strength of the eddy ( ), are shown at 25-day intervals (a) between day 50 and day 150 and (b) between day 175 and day 250. (c) Evolution of the strength of the eddy. The horizontal line indicates the level of the initial strength of the eddy. Solid circles indicate the dates that the SSH iso-lines are shown in Figures 16a and 16b. Possible periods of the four stages are indicated as well. (d) Evolution of the meander amplitude defined by meander peak and downstream trough of SSH 0.7-m iso-line. (e) Evolution of the barotropic velocity of the Kuroshio defined as the maximum velocity off the Kii peninsula. (f) Evolution of the baroclinic velocity of the Kuroshio off the Kii peninsula, evaluated as the difference of surface velocity and the barotropic velocity at the same location as the barotropic velocity. [31] The timing issue can be better addressed by monitoring other indices. Prior to the increase of the meander amplitude (Figure 16d), the barotropic velocity of the Kuroshio starts to increase around day 100 (Figure 16e). Similarly but not as clear as the first meander, the second meander amplitude of the Kuroshio is preceded by the increase of barotropic velocity of the Kuroshio. Overall, the lag correlation of the Kuroshio amplitude and the barotropic velocity is largest with 28 days lag, with the latter leading the former (coherence ). Although this

13 WASEDA ET AL.: EDDY-KUROSHIO INTERACTION: MEANDER FORMATION Figure 17. Evolution of the thermocline displacement from day 105 to day 135. The thermocline depth is defined as the depth of s = may indicate that the meander growth is due to purely barotropic eddy-kuroshio interaction, a closer look at the evolution of the baroclinic velocity of the Kuroshio suggests otherwise. As shown in Figure 18f, the baroclinic velocity of the Kuroshio leads the barotropic velocity; the lag correlation with the barotropic velocity is largest with 21 days lag (coherence ). Therefore, in this simulation, the sequence of events is the following: first the baroclinic velocity of the Kuroshio increases; then the barotropic velocity of the Kuroshio increases; and, finally, the meander grows. To be more specific, the generation of the anticyclonic eddy of the lower ocean (see 1000-m velocity field of Figure 18, day 120) is a result of conversion of baroclinic energy to the barotropic energy, which may be interpreted as some kind of baroclinic instability (see rapid increase of velocity off Kii peninsula between day 100 and 120). Now, what caused the upper layer Kuroshio velocity to increase at day 80, much before the eddy core encounter with the Kuroshio (e.g., see Figure 9 day 105)? Presumably this is because the baroclinic perturbation off the Kii peninsula comes from the western edge of the eddy that encountered the Kuroshio at a much earlier time than the eddy core, around day 60 to 70 (figures are not shown in this paper; see the animation of Waseda et al. (virtual poster, 2000)). As can be seen from Figure 16f, the Kuroshio current is highly baroclinic. Like the Gulf Stream, Figure 18. Velocity field at 1000 m depth overlaid on top of the 1.0-m SSH contour indicating the location of the Kuroshio axis and the 1.5-m contour indicating the location of the anticyclonic eddy.

14 13-14 WASEDA ET AL.: EDDY-KUROSHIO INTERACTION: MEANDER FORMATION Figure 19. Production of high potential vorticity at the Kii peninsula, during days The colored contours indicate the relative vorticity contribution to the potential vorticity: red, ; and blue, m 1 s 1. The black contours indicate SSH at c.i. 0.1 m. both baroclinicity and barotropicity are dynamically significant, unlike the barotropic Agulhas Current [de Ruijter et al., 1999]. The timing issue of the meander growth is thus rather complicated, and the current analysis only suggests a possibility; by no means do we intend to conclude that all the meander evolution follows the same scenario. [32] We now address the process of meander disappearance. It is interesting that the available potential energy increases as the meander amplitude decreases; we have seen in section 3.5 that prior to the detachment of the eddy from the Kuroshio, the thermocline displacement increased (see Figure 14, day 155 to 170). That is, as the circulation increased (decrease of anticyclonic vorticity) with time, the available potential energy increased. Just prior to this (between day 135 and 155, Figure 11), the multiple peaks that appeared in the SSH signature merge. Simultaneously, the barotropic velocity of the Kuroshio starts to decrease (Figure 16e). It is difficult to think that such a process has resulted from the eddy s own dynamics. Rather, it is likely that the coalescence of the multiple peaks of SSH is a manifestation of the Kuroshio current. It is also worth noting that up until day 180 when the eddy had already started to migrate west, the cyclonic circulation or the Enshu-nada Sea separation bubble remained closed. At day 180, though, the separation bubble opens and the high potential vorticity is released to the Kuroshio extension region. [33] Production and release of the high potential vorticity inshore of the Kuroshio is best monitored by the change of relative vorticity contribution to the potential vorticity. In analyzing the GCM output, we have defined the potential vorticity as PV =(z + f )/h, where h s/(@s/@z), z is the relative vorticity, and s is the potential density. Thus the contribution of the relative vorticity to the potential vorticity is PV r = z/h. Figure 19 shows a production of high PV r at the Kii peninsula at s q = 25.2; see, for example, day 120. The produced high potential vorticity PV r will accumulate inshore of Kuroshio, and eventually the meander amplitude grows. Furthermore, a high potential vorticity is generated at the meander peak when the eddy detaches from the Kuroshio (see Figure 20, day 170). Unlike the case studied by Qiu and Miao [2000] in which the cyclonic ring is formed, the cyclonic vorticity is shed as a wake to the Kuroshio extension as the Kuroshio separates from the Boso peninsula. The separation bubble is now open and forms a wake. This process can be monitored by following the high PV r region in Figure 20. As this high potential vorticity region gets advected eastward, the meander amplitude reduces and the separation bubble closes again. [34] We can formulate the following possible sequence of the short-term meander: (1) an anticyclonic eddy migrates westward to meet the Kuroshio; (2) as the anticyclonic eddy gets advected by the Kuroshio, it experiences some kind of an instability releasing the available potential energy into barotropic kinetic energy and triggers the meander; (3) at the same time, flow separations are induced at sharp coastal topography generating high potential vorticity water inshore of Kuroshio; (4) the produced high potential vorticity water coalesces within the Enshu-nada Sea and the meander amplitude increases; (5) the seemingly splitting anticyclonic eddy merges and the thermocline deepens; (6) the anticyclonic eddy detaches from the Kuroshio while the Enshunada separation bubble opens up releasing the excess high potential vorticity into the Kuroshio Extension; (7) meander ceases; and cycles 1 to 7 repeat. [35] One of the key factors in this scenario is the cyclonic vorticity production at the sharp coastal topography [Yamagata and Umatani, 1989]. This is a mechanism to allow rapid increase of the cyclonic vorticity inshore of Kuroshio catching up with the rapid breakup of the anticyclonic eddy offshore due to baroclinic instability. The role of sharp topographic features in producing potential vorticity was discussed by Akitomo et al. [1991] and Yamagata and Umatani [1989], the former emphasizing the Kyushu and the latter emphasizing the Kii peninsula. In this study, we have shown examples of shedding of eddies from both the Kyusu and the Kii peninsula. However, more attention was given to the role of the Kii peninsula whose topographic character changes due to the Kuroshio path variations; that is, depending on the angle of attack of the Kuroshio, the Kii

15 WASEDA ET AL.: EDDY-KUROSHIO INTERACTION: MEANDER FORMATION Figure 20. Release of high potential vorticity to the Kuroshio extension, during days The colored contours indicate the relative vorticity contribution to the potential vorticity: red, ; and blue, m 1 s 1. The black contours indicate SSH at c.i. 0.1 m. peninsula appears as either a forward step (as in the study of Yamagata and Umatani [1989]) or a backward step. The resulting subtle changes in the occurrence of the eddy shedding seemed to affect the strength of the cyclonic circulation in the Kuroshio separation bubble that may eventually affect the merger of the split anticyclonic eddy. Overall, this sequence suggests that during the formation of the short-term Kuroshio meander, a conversion of the available potential energy to the barotropic kinetic energy was also accompanied by a cyclonic vorticity generation inshore of Kuroshio. When the meander amplitude reduced, available potential energy offshore increased due to eddy merger, while the cyclonic vorticity inshore reduced as a result of eddy shedding downstream towards the Kuroshio extension. A repetition of this cycle is assured if the total energy of the anticyclonic eddy is recovered each time but this is likely not the case. In Figure 16c, we do see an overshoot of the circulation strength after day 210 or so indicating loss of anticyclonic vorticity from the initial value. In such case, any additional supply of anticyclonic vorticity from an external source would help sustain the cycle, such as eddy coalescence as observed by Ebuchi and Hanawa [2003] Generation of Anticyclonic Eddy in the Kuroshio Extension [36] An interesting eddy detachment process was observed at the Kuroshio extension. We can only present a particular case observed in our model study and, based on this one example, we suggest a possible mechanism: an anticyclonic eddy generation south of a meandering jet. Note that because ring detachment will always generate a cyclonic eddy south of an eastward flowing jet, another mechanism is necessary for an anticyclonic eddy generation south of the jet. The example that we introduce here somewhat resembles the eddy detachment process near the Kuroshio meander; first, the anticyclonic eddy strengthens as a result of multiple eddy merger and, second, additional westward migration speed is provided by an advection by a barotropic flow field. At day 2410 (Figure 21), the SSH field shows that there is an anticyclonic eddy south of the meandering trough of the Kuroshio, pairing with a cyclonic eddy north of the trough. This anticyclonic eddy was generated, actually, as a result of a merger of two eddies, one in the south of the crest (west to the trough) and one in the south of the trough. The eddy merger process is evidenced in the stream line plot of the barotropic field as well (see Figure 22, day 2410). At day 2410, although it is not obvious from the SSH signature, the barotropic field indicates that there is still a strong circulation in the western eddy. As these two eddies merge, in particular the barotropic component, the eddy starts to migrate toward the west (day 2410 to day 2440) and eventually detaches itself from the Kuroshio extension around day At this time, one can see that there is a phase shift in the eddy center of the barotropic and the baroclinic components, indicating that the barotropic field is driving the upper ocean eddy to the west. 5. Conclusions [37] We have successfully simulated the eddy-kuroshio interaction observed in the satellite altimetric data (TOPEX/ Poseidon). The T/P altimetric data were assimilated into a GCM by means of optimal interpolation nudging to initialize the anticyclonic eddy west of the Izu Ridge. The successive evolution of the simulated eddy and its interaction with the Kuroshio resembled the observed short-term Kuroshio meanders in 1993 and 1998 that are known to be induced by an anticyclonic eddy. The simulated model outputs were analyzed to study the various stages of the interaction: (1) westward migration of the eddy; (2) advection of the eddy by the Kuroshio; (3) meander formation; and (4) detachment of eddy from the Kuroshio. This sequence may repeat several times in nature; in this simulation, the meander repeated twice. [38] Dynamically, two factors seem crucial to the formation of the meander. The first is the release of available potential energy of the anticyclonic eddy to the barotropic kinetic energy through baroclinic instability, and the second