Observation of deep currents around a seamount between the Japan and Kuril trenches

Abstract

Hydrographic observations and two-year current measurements revealed the flow around the Erimo Seamount between the Japan and Kuril trenches. Northward currents were observed to flow both around the seamount and on the eastern flank of the trench. Since the dissolved oxygen concentration of the former is higher than that of the latter, these currents correspond to the western and eastern branch currents of the deep western boundary current in the North Pacific, respectively. The western branch current forms a Taylor cap over the seamount, although during certain periods, it joins the eastern branch current. During such periods, a strong jet forms around the seamount and the Taylor cap is washed away. Such alternately intensified currents are consistent with previous observations of the Japan Trench. On the western flank of the trench, a southward current flows near the seafloor, while an eastward current is formed at a depth of 3360 m due to the confluence of the currents from north and south.

1 Introduction

Deep western boundary currents (DWBCs) transport new water masses from their formation regions at high latitudes to deep oceans worldwide. The North Pacific Ocean has the most abundant deep waters in the world and stores large amounts of heat and materials. Determining the pathway and water properties of its DWBC is necessary to understand climate change and geochemical cycles.

Kawabe and Fujio (2010) described the DWBC pathways in the Pacific Ocean. The origin of deep water is the Lower Circumpolar Deep Water (LCDW), which separates from the Antarctic Circumpolar Current and is transported northward by the DWBC. In the Central Pacific Basin, the DWBC bifurcates into a western branch current (WBC) and an eastern branch current (EBC) because of the bottom topography. In the Northwest Pacific Basin, the WBC flows along the western Izu-Ogasawara and Japan trenches, and the EBC flows on the eastern side, turning west near the southern Shatsky Rise. After joining near 38$^\circ$N in the Japan Trench, the branch currents flow further north into the Kuril Trench.

Details of the current behavior at 38$^\circ$N in the Japan Trench were reported by Fujio and Yanagimoto (2005) using hydrographic observations and current measurements. Since the western water on the eastern flank of the trench was colder, more saline, and more oxygen-rich than the eastern water, the two branch currents flowed side by side following different isobaths, conserving the upstream water properties at least within the Japan Trench (Fig. 1a). The branch currents have also been reported at 40$^\circ$N by Ando et al. (2013). While the WBC was observed to flow steadily along deep isobaths (at depths $\ge$ 6000 m), the EBC was observed to change its pathway between the 5500-m and 6000-m isobaths over a 6-month period. A side-by-side structure occurred when the EBC flowed along the 5500-m isobath. At 42$^\circ$N, in the Kuril Trench, Owens and Warren (2001) reported a northward current with large velocity fluctuations of 10 cm s⁻¹ or more. Whether the temporal variations related to the branch currents continue to the Kuril Trench has not yet been determined.

Fig. 1

a and b Topography of the observation region. The isobaths were drawn at intervals of 250 m (thin lines) and 1000 m (thick lines). The mooring stations and conductivity–temperature–depth (CTD) stations are indicated by red and black crosses, respectively. The red lines in panel (a) depict the schematic pathways of the western branch current (WBC) and eastern branch current (EBC) south of 40$^\circ$N. c Vertical cross section of the topography along the survey line. Current meters and CTD stations are indicated by red circles and black inverted triangles, respectively

In contrast to the northward DWBC on the eastern flanks of the trenches, a southward current flows on the western flanks. Owens and Warren (2001) depicted this current as a recirculation of the DWBC over the western North Pacific trenches from the Aleutian to the Izu-Ogasawara trenches.

Very few observational studies have been conducted on trench junctions; therefore, details of the current fields at the junctions remain unclear. One notable feature at the junction between the Japan and Kuril trenches is the presence of the Erimo Seamount with a radius and height of approximately 10 km and 2000 m, respectively, rising from the seafloor at a depth of 6000 m (Fig. 1). Seamount in a steady current can form a clockwise current, that is a Taylor cap, in a stratified ocean.

Huppert (1975) described the quasi-geostrophic theory of Taylor caps. A Taylor cap is a current that forms a closed streamline and traps water particles over a convex bottom topography. It is considered to be a superposition of a uniform impinging current and a secondary rotating current over the topography. At depths where the latter is stronger than the former, closed streamlines are formed. The vertical e-folding scale of the upward-decaying rotating current depends only on the Burger number. The strength of a rotating current is inversely proportional to the Rossby number, which is defined using the impinging current. In other words, as the Rossby number increases, the water particles in the upper part of the water column are washed away by the impinging current as it becomes stronger than the rotating current. If the Rossby number exceeds a critical value, the Taylor cap can no longer exist, and all the water particles over the seamount are washed away. Chapman and Haidvogel (1992) verified Huppert’s (1975) theory using a primitive equation model. They examined the critical Rossby numbers by varying the height of the seamount and found that a tall seamount creates large vertical velocities since the water column is strongly compressed or stretched, allowing water particles to readily move over the seamount and reducing the critical Rossby number from that predicted by the quasi-geostrophic theory.

A Taylor cap was reported by Owens and Hogg (1980), who conducted hydrographic observations and current measurements on a bump with a radius and height of approximately 25 km and 400 m, respectively, rising from the seafloor at a depth of 5000 m. They demonstrated uplifted isopycnals and closed streamlines, which were evidence of a Taylor cap. If an impinging current covers the Erimo Seamount, a Taylor cap can be observed; however, this has not yet been reported.

Another notable feature of the junction between the Japan and Kuril trenches is that it corresponds to the boundary between currents with different structures on the western flanks of the Japan and Kuril trenches. Recirculation currents typically exhibit an upward-decaying horizontal velocity structure. However, the velocity structure in the Japan Trench observed by Ando et al. (2013) was more strongly restricted to the seafloor than that in the Kuril Trench observed by Owens and Warren (2001). Furthermore, Ando et al. (2013) observed a weak northeastward current at a depth of 3440 m (above the recirculation current) while Owens and Warren (2001) observed a weak southwestward current at a depth of 3006 m (in the upper part of the recirculation current). The current generated when these currents meet at the junction still remain unknown.

This study aimed to examine the details of the current field at the junction between the Japan and Kuril trenches. The questions addressed include whether the DWBC and its temporal variation observed in the Japan Trench continue at the trench junction and further into the Kuril Trench, whether Taylor caps form over the Erimo Seamount, and what would be the type of current field formed by the currents on the western flanks of the trenches that meet this junction.

2 Observations and data processing

We conducted current measurements at four mooring systems (EM1–EM4) on a survey line across the Erimo Seamount near 41$^\circ$N between the Japan and Kuril trenches (Fig. 1). The moorings were deployed during the Shinsei Maru KS-20–18 cruise in August 2020 and were recovered during the Hakuho Maru KH-22–8 cruise in October 2022. Table 1 presents the details of the mooring observations. At each mooring, we installed current meters near depths of 3500, 4500, and 5500 m and measured the horizontal velocities at 30-min intervals. The current meters installed near a depth of 4500 m at EM2 and EM4 were single-point Aquadopp Deepwater meters manufactured by the Nortek Group (Sandvika, Norway), and the other 10 current meters were RCM11 meters manufactured by Aanderaa Instruments (currently part of Xylem Analytics, Norway).

Table 1 Mooring configurations, instrument details, and sampling specifics

Occasionally, the EM2 current meters were significantly tilted by strong currents. The accompanying large pressure changes were also problematic (e.g., up to 800 dbar on the current meter at a depth of 4800 m). Therefore, we manually removed the data for which pitch and roll exceeded the limit of 20$^\circ$. Ratios of the amount of data removed to the amount of data recorded are listed in Table 1. RCM11 stopped because of battery exhaustion prior to recovery; therefore, we analyzed the data up to April 20, 2022, when all the current meters were still active. We applied a fifth-order Butterworth filter with a cut-off period of 3 days to the velocity data after August 19, 2020, both forward and backward, to prevent a phase shift, and then removed the last 3 days of data. Figure 2 shows the resulting daily time series.

Fig. 2

Stick diagrams of velocities measured by current meters at a EM1, b EM2, c EM3, and d EM4. The upward direction of the sticks indicates the major axis direction (the northeast is positive) of the velocity ellipse near a depth of 3500 m at each mooring, and the lower left arrow in each panel indicates the north direction and magnitude. Rectangles with dashed lines indicate jet periods 1 (JP1) and 2 (JP2)

We conducted conductivity–temperature–depth (CTD) observations at five stations (EC2–EC6) on the survey line of the moorings (Fig. 1b) during the recovery cruise in October 2022. The CTD system was an SBE911plus system manufactured by Seabird Scientific (Bellevue, WA, USA). CTD salinity and dissolved oxygen (DO) data were calibrated using seawater samples. We used the Thermodynamic Equation of SeaWater 2010 (TEOS-10, McDougall and Barker 2011) to calculate the Absolute Salinity and other quantities.

For bathymetry, we used the GEBCO2021 15-arc-second gridded dataset (GEBCO Compilation Group 2021). Our shipboard multi-beam survey results around the survey line matched the bathymetry data.

3 Results

3.1 Mean currents

Figure 3 and Table 2 present the mean current statistics. Statistics are not shown for EM2 at a depth of 5800 m owing to the large percentage of missing data. The standard error was calculated as ${\left(2{\upsigma }^{2}\uptau /t\right)}^{1/2}$, where ${\upsigma }^{2}$ is the variance, $\uptau$ is the integral time scale, and $t$ is the data length (Dickson et al 1985). The stability of the current direction was calculated by the ratio of vector mean to scalar mean of velocity.

Fig. 3

Mean velocity vectors from August 19, 2020, to April 20, 2022. The colors of the vectors indicate the approximate depths of the current meters. Isobaths were drawn at intervals of 500 m (thin lines), and gray areas indicate areas deeper than 6500 m

Table 2 Statistics of the velocities from August 19, 2020 to April 20, 2022. The mean velocities are shown along with the standard errors

First, at EM4 on the eastern flank of the trench, the direction and stability of mean velocities did not vary significantly with depth and the velocities were oriented northeastward along the isobath. The major axis direction of the standard deviation ellipse (hereafter, the velocity ellipse) indicated similar features. Moreover, the ellipse had a major radius comparable to the mean velocity and high oblateness. Therefore, this current was likely controlled by the seafloor. Furthermore, the magnitudes of the velocities at depths of 4470 and 5470 m were greater than those at 3470 m. These features were consistent with the current measurements on the eastern flanks of the Japan and Kuril trenches by Ando et al. (2013) and Owens and Warren (2001), respectively.

Second, at EM3 and EM2 on the seamount flanks, the mean velocities near depths of 3500 and 4500 m were stable and northeastward (Fig. 3; Table 2). However, at EM3 on the eastern flank, the mean velocity at a depth of 5450 m was unstable, close to zero, and clearly different from that at shallower depths. Instead, the velocity ellipse had a large major radius and high oblateness, indicating a strong oscillating current. The mean velocities on the seamount flanks were greater to the west (EM2) than to the east (EM3). This asymmetry suggested a Taylor cap on the seamount, hence explaining the weak mean velocity at a depth of 5450 m at EM3 owing to the downward-amplifying rotating current. The Taylor cap is discussed further in Sect. 4.2. In any case, the observed velocities arise from the superposition of impinging and rotating currents. Accordingly, we estimated the mean velocities of the impinging current at the upper two depths by averaging the mean velocities at EM2 and EM3. The values were approximately 5.5 cm s⁻¹ and 8.0 cm s⁻¹ near depths of 3500 and 4500 m, respectively, being 1–2 cm s⁻¹ greater than those at EM4.

Finally, at EM1 on the western flank of the trench, the mean velocities rotated clockwise from depths of 3360 m to 5360 m (Fig. 3; Table 2). However, the mean velocities did not adequately capture the actual current structure. Figure 4 shows the direction of mean velocities and the rose diagrams expressing the intensity and occurrence frequency of the velocities. At a depth of 5360 m, the directional distribution was unimodal, and there was a sharp directional mode in the south–southwest direction, which contained the direction of its mean velocity. However, at depths of 3360 and 4360 m, the distributions were bimodal, and there were relatively weak modes in the east and from south to south–southwest, neither of which contained the direction of their mean velocities. At depths of 3360 and 4360 m, in addition to the downward-decaying eastward current, an upward-decaying south–southwestward current was present in the time series, although the former current did not reach 5360 m (Fig. 2a). This led to the discrepancy between the mode and mean of the current directions at the upper two depths, but not at a depth of 5360 m. The predominant current at a depth of 3360 m was found to be eastward, in contrast to the mean current.

Fig. 4

Rose diagrams of EM1 current data at depths of a 3360, b 4360, and c 5360 m from August 19, 2020, to April 20, 2022. Lengths of the bars formed a histogram of the current directions with a bin width of 22.5$^\circ$. The colors in each bar indicate the proportion of the velocity magnitude included in the bins, as shown in the lower left legend. The red line at each depth indicates the direction of mean velocity

The south–southwestward current at a depth of 5360 m at EM1 flowed in an intermediate direction between the axes of the Japan and Kuril trenches (Fig. 3). Since the velocity ellipse had a major radius comparable to the mean velocity and high oblateness (higher than that at shallower depths and comparable to that at other stations) (Table 2), this current was likely controlled by the seafloor. Similar southward currents had been observed by Ando et al. (2013) and Owens and Warren (2001). Conversely, the eastward current at a depth of 3360 m flowed along neither the isobath nor the trench axes. Such currents have not been observed previously between depths of 3000 and 3500 m. The eastward current is discussed further in Sect. 4.3.

3.2 Temporal variation

Figure 5 shows the time series of velocity along the major axis of the velocity ellipse at a depth of 4800 m at EM2. Jets that significantly exceeded the mean velocity were observed occasionally. Extremely strong jets that exceeded the mean by twice the standard deviation were observed continuously from December 2020 to March 2021 and from February 2022 to May 2022. Hereafter, we refer to these periods as jet periods 1 (JP1) and jet periods 2 (JP2), respectively.

Fig. 5

Time series of the major axis component of velocity at a depth of 4800 m at EM2 (the northeast is positive). Horizontal lines indicate the mean (solid red), mean plus standard deviation and twice the standard deviation (dotted black). The black dashed rectangles indicate the jet periods, JP1 and JP2

During the jet periods, the northeastward velocities were strong not only at a depth of 4800 m at EM2 but also at depths of 3360 and 4360 m at EM1, 3800 m at EM2, and all three depths at EM3 (Fig. 2a–2c). The velocities at a depth of 5800 m at EM2 were frequently lost during the jet periods owing to the large tilt; therefore, the velocity should be extraordinarily strong at this location (Fig. 2b). In contrast to the velocities around the seamount (EM1–EM3), those on the eastern flank of the trench (EM4) were weak at all three depths during the jet periods, while being strong during other periods (Fig. 2d). Such variations related to jet periods were not observed at a depth of 5360 m at EM1 (Fig. 2a). To visualize this tendency better, Fig. 6 shows the mean velocities for JP1 and JP2 and the periods before and after them. An alternating intensification of the velocities around the seamount during the jet periods and on the eastern flank of the trench during the non-jet periods was evident.

Fig. 6

Similar to Fig. 3, a before JP1 (August 19, 2020 to October 31, 2020), b JP1 (November 1, 2020 to March 31, 2021), c between JP1 and JP2 (April 1, 2021 to January 31, 2022), d JP2 (February 1, 2022 to May 31, 2022), and e after JP2 (calculated from current meters with data for more than four months)

3.3 Hydrography

CTD observations were conducted after JP2 when the mean northeastward current was weak at EM2 and strong at EM4 (Fig. 6e). Accordingly, we first checked whether the density field was representative of the mean current field for the non-jet periods (Fig. 6a, 6c, and 6e). Figure 7 shows the vertical section of the potential density. The LCDW with a density higher than 45.850 ${\sigma }_{4}$, as defined by Ando et al. (2013), was deeper than approximately 3500 dbar. The depth of this isopycnal does not change significantly if calculated using EOS-80 as in Ando et al. (2013) or using TEOS-10. Using these density profiles, the solid lines in Fig. 8 show the geostrophic velocities referenced to the current meters from EM2 to EM4. Since only two current meters were operational when the CTD observations were conducted, we used the mean velocities between JP1 and JP2 when the velocities were weak at EM2 and strong at EM4 (Fig. 6c). We selected the reference velocities near a depth of 4500 m, since this was the central depth of the LCDW. The vertical shears of theses geostrophic and mean velocities were roughly qualitatively consistent as they reduced the northeastward velocity downward below a depth of 4500 m between EC4 and EC5 and increased elsewhere. Therefore, the observed density field was consistent with the mean current field.

Fig. 7

Vertical section of the potential density ${\sigma }_{\theta }$ above 2000 dbar and ${\sigma }_{4}$ below 2000 dbar. Current meters and CTD stations (EC2–EC6) are indicated by red circles and black inverted triangles, respectively

Fig. 8

Vertical profiles of mean velocity from current meters and geostrophic velocity from CTD data. The mean velocities are shown as a component perpendicular to the survey line (the northeast is positive) and were calculated from the data between JP1 and JP2 (circles). The geostrophic velocities of EC2–EC3, EC4–EC5, and EC5–EC6 were referenced to the mean velocity near a depth of 4500 m in nearby moorings (solid). The geostrophic velocities of EC3–EC4 were calculated using the reference level at 3500 dbar (dashed-dotted line)

The dash-dotted line shows the geostrophic velocity between EC3 and EC4 calculated with 3500 dbar as the reference level since there was no nearby mooring that could be referenced (Fig. 8). This profile indicated a vertical shear that reduced the northeastward velocity downward, similar to that between EC4 and EC5, which was the result of an uplifted isopycnal above the seamount summit (Fig. 7). Therefore, the two profiles on the eastern flank of the seamount indicated the influence of a rotating current near the seafloor. Since the CTD pair straddled the seamount, further discussion of the geostrophic velocity would not be appropriate.

Next, we checked the water properties to estimate the spatial structure of the current field in more detail. Figure 9 shows the DO concentrations of isopycnals near the seafloor. Since the DWBC transports oxygen-rich water, the DO concentration on the eastern flank of the trench (EC6) should be higher than that around the seamount (EC2–EC5) owing to its relatively high velocity. However, the DO concentration was the lowest on the eastern flank of the trench (EC6) and high around the seamount (EC3 and EC4). Here, the precision of the DO difference between stations could be evaluated by the scattering range of the DO profiles since our CTD measurement was performed with a single CTD system with the same sensors and calibration factors (Ando et al. 2013). Since the DO difference between EC6 and other stations was sufficiently larger than the scattering range at each station, the difference was significant. The DO distribution is discussed in detail in Sect. 4.1, in relation to the branch currents.

Fig. 9

Vertical profiles of the dissolved oxygen (DO) concentration versus the potential density ${\sigma }_{4}$

4 Discussion

4.1 The DWBC variability

Ando et al. (2013) reported an intensification of northward velocities on a time scale similar to ours on the eastern flank of the Japan Trench. During the first 6 months of their current measurements (referred to as P1), weak and strong mean velocities occurred on the isobath near depths of 6000 and 5500 m, respectively, on the eastern flank of the trench. During the next 12 months (referred to as P2 and P3), the northward velocity became stronger on the isobath near a depth of 6000 m than near a depth of 5500 m. Considering the locations of the strong northward velocities, we associated P1 and P2–P3 with our non-jet and jet periods, respectively, which are plotted together in Fig. 10. As our observation sites were close to and downstream of the sites of Ando et al. (2013), the observed variations were possibly identical to the DWBC variations reported by Ando et al. (2013).

Fig. 10

Average mean velocities near depths of 3500 and 4500 m calculated from the velocities during the a non-jet and b jet periods. The mean velocities of (a) P1 and (b) P2–P3 from Table 2 of Ando et al. (2013) were added to the results. Isobaths at 1000-m intervals from the surface (solid lines) and at a depth of 5500 m (dashed lines) are shown

Ando et al. (2013) conducted CTD observations for the periods of Fig. 10a. They reported two separate oxygen-rich waters on the eastern flank of the trench. Near the seafloor, the DO concentration in the western water was approximately 2 μmol kg⁻¹ higher than that in the eastern water. According to our observation, the DO concentration around the seamount was approximately 2 μmol kg⁻¹ higher than that on the eastern flank of the trench (Fig. 9). Although these values cannot be directly compared, since accuracy of the bottle measurements of both cruises was not assessed, the CTD measurements during each cruise were consistently performed, and their results were similar to ours in terms of the DO distribution, which was higher to the west than to the east. Ando et al. (2013) further discussed the transport of two oxygen-rich waters by the currents using the silica concentrations and showed the currents to be the WBC and EBC. Therefore, the currents around the seamount and on the eastern flank of the trench should also be the WBC and EBC, respectively.

Overall, we speculated that during non-jet periods, the relatively strong EBC and weak WBC flowed side by side on the eastern flank of the trench and around the seamount, respectively. Conversely, during the jet periods, the EBC flowed along the deeper isobaths and joined the WBC, which then flowed strongly together around the seamount. The temporal variation related to the branch currents in the Japan Trench should continue into the Kuril Trench.

4.2 Taylor caps

First, we examined the presence of a Taylor cap in the study area. The isopycnals were uplifted close to the seamount summit and the peak was above the western flank (Fig. 7). The peak on the left-hand flank toward the downstream is typical of a Taylor cap with an impinging current (Chapman and Haidvogel 1992). Since isopycnals in trenches are generally depressed along the trench axis by cyclonically opposite geostrophic currents on the trench flanks (Johnson 1998), the uplifted isopycnal should be locally caused by the influence of the seamount. In addition, between JP1 and JP2, the mean velocity at a depth of 5450 m at EM3 was southwestward, suggesting a closed streamline, unlike that at shallower depths (Figs. 6c). Therefore, a Taylor cap, restricted to the area near the seafloor, existed during this period.

Next, we examined the vertical structure of the Taylor cap in more detail. The theoretical vertical scale of a rotating current is $fL/N$, where $f$ is the Coriolis parameter, $L$ is the seamount radius, and $N$ is the buoyancy frequency (Owens and Hogg 1980). We used $N=4.8\times {10}^{-4} {\text{ s}}^{-1}$; this is the mean value of $N$ from all the CTD data between depths of 3800 and 4800 m (near the depths of the summit and the seafloor recorded at EC3 and EC4, respectively). Given $f=9.6\times {10}^{-5} {\text{ s}}^{-1}$ and $L=10\text{ km}$, $fL/N$ was 2000 m. The vertical scale can also be estimated from the current meter data. Figure 11 shows the vertical profiles of the velocities of impinging and rotating currents, which were decomposed from the velocities at EM2 and EM3. These velocities were estimated as the average of the velocities at EM2 and EM3 and as the difference from the impinging current, respectively. The vertical scale of the rotating current was estimated as 941 m by fitting the velocity profile to an exponential function. Although this value was approximately half of that estimated from stratification, it was of the same order and hence acceptable.

Fig. 11

Vertical profiles of mean velocities of the impinging and rotating currents decomposed from the mean velocities at EM2 and EM3 (see text). The solid curve represents an exponential function fitted to the velocity of the rotating current. These are shown as components perpendicular to the survey line (the northeast is positive) and were calculated from the data between JP1 and JP2

Further, the shallowest depth of the Taylor cap that forms closed streamlines was determined by the depth at which the velocity of the impinging current is equal to that of the rotating current. Due to the coarse vertical resolution of the current meter data, the exact depth could not be estimated; however, it is likely to be deeper than 5000 m. The height of the Taylor cap was considerably smaller than the water depth, and water particles trapped by the seamount were likely limited near the seafloor. Conversely, water particles at shallower depths were likely to be washed away by the impinging WBC, as indicated by the velocity of the impinging current being greater than that of the rotating current near depths of 3500 and 4500 m.

Finally, we examined the temporal variations in the Taylor cap. The orange and green lines in Fig. 12 show the low-pass filtered velocities along the major axis of the velocity ellipse at depths of 4800 m at EM2 and 5450 m at EM3. Since the velocity at EM2 was frequently lost at a depth of 5800 m, we considered the velocity at a depth of 4800 m. During the non-jet period between JP1 and JP2, when the Taylor cap was formed, the long-term fluctuations in the velocities at EM2 and EM3 were out of phase. Conversely, during the jet periods, strong northeastward velocities were occasionally observed at EM3, and the fluctuations tended to be in phase, indicating the disappearance of the Taylor cap.

Fig. 12

Time series of the velocities at depths of 4800 m at EM2 (orange) and 5450 m at EM3 (green), and the velocity of the impinging current (black), which was estimated by averaging the velocities at depths of 4800 m at EM2 and 4450 m at EM3. Each time series was low-pass filtered with a cut-off period of 30 days. The red line indicates the critical velocity of the impinging current (${U}_{c}$) that forms the Taylor cap. Dashed rectangles indicate jet periods, JP1 and JP2

Chapman and Haidvogel (1992) explained the conditions for the formation of Taylor caps by impinging currents using the Rossby number $Ro=U/fL$, the Burger number $Bu=N{H}_{m}/fL$, and the fractional height of the seamount $\delta ={H}_{m}/H$, where $U$ is the velocity of the impinging current, ${H}_{m}$ is the seamount height, and $H$ is the water depth. Given ${H}_{m}=2000 \text{ m}$ and $H=6000\text{ m}$, we obtain $Bu=1.0$ and $\delta =0.33$ for the Erimo Seamount. Therefore, on the basis of their numerical results with $Bu=1.0$ (their Fig. 20), the critical Rossby number for forming a Taylor cap was ${Ro}_{c}=0.13$ and a Taylor cap will form only when $U\le {U}_{c}={Ro}_{c}fL=12 \text{ cm} {\text{s}}^{-1}$, where ${U}_{c}$ is the critical velocity of the impinging current.

We estimated $U$ by averaging the velocities on the western and eastern flanks of the seamount. The black line in Fig. 12 shows the estimated $U$ near a depth of 4500 m. When the northeastward velocity at a depth of 5450 m at EM3 intensified during the jet periods, $U$ exceeded ${U}_{c}$, which occurred four times. Therefore, the Taylor cap was fully washed away when the EBC was speculated to have flowed along the isobath near a depth of 6000 m and joined the WBC, forming an intensified impinging current around the seamount.

Our results can potentially be applied to other seamounts in the Pacific trenches, such as the First Takuyo Seamount at 41.5$^\circ$N, 146$^\circ$E, which has a similar scale and shape as the Erimo Seamount, given the information on background impinging currents (see Fig. 1 for location).

4.3 Another local current

The south–southwestward current at a depth of 5360 m above the western flank of the trench (Fig. 3) is evidence of a recirculation current passing through the trench junction, as proposed by Owens and Warren (2001). Conversely, the eastward current at a depth of 3360 m, which did not flow along the isobaths, was unique. The intensified currents at depths of 3360 and 5360 m extended downward and upward, respectively, without significant rotation (Fig. 4b). Temporal variation of the eastward current was correlated to that of the currents around the seamount, whereas that of the southward current was not (Fig. 6). These results indicated that the eastward current was different from the lower recirculation current.

The eastward current at a depth of 3360 m at EM1 was likely the result of the confluence of the previously observed currents from the north and south. If this is the case, the eastward direction approximately perpendicular to the meridionally aligned isobaths was determined probably because the confluence cannot flow toward the onshore direction owing to mass conservation and must escape in the offshore direction.

Notably, when the eastward current at a depth of 3360 m intensified, that is, during the jet periods, its directional stability increased significantly. Before JP1 and between JP1 and JP2, the stability was 0.68 and 0.50, respectively, whereas during JP1 and JP2, it increased to 0.84 and 0.92, respectively. Stability during the jet periods approached those at EM2 and EM3 (Table 2) although stability values at EM2 and EM3 also increased to approximately 0.95 and 0.91, respectively. This high stability suggested that the current may have followed some type of geostrophic contour. Since the western flank of the trench is steep, the eastward current quickly carries water particles away from the seafloor, which can weaken the topographic control on the particles. Therefore, $f$ contours may be the contour along which the eastward current flows.

Part of the water transported southward near a depth of 3360 m on the western flank of the Kuril Trench was, therefore, speculated to change direction eastward at the trench junction. This water then joins the northeastward current around the seamount and flows back into the Kuril Trench. This is a new finding and suggests the existence of similar local currents at other trench junctions as well.

5 Conclusions

We conducted hydrographic observations and 2-year current measurements to reveal the current field between the Japan and Kuril trenches at 41$^\circ$N.

Mean velocities obtained from the current meters over the seamount and on the eastern flank of the trench indicated directionally stable northward currents. The currents intensified alternately between the jet and non-jet periods. The CTD data obtained when the velocities intensified on the eastern flank of the trench showed a higher DO concentration around the seamount than on the eastern flank of the trench. The results were consistent with those of Ando et al. (2013); hence, the currents were speculated to be the western and eastern branch currents, which bifurcate in the Central Pacific and re-join near 38$^\circ$N in the Japan Trench. However, when the velocities intensified around the seamount, confluence of the branch currents likely flowed strongly around the seamount. Temporal variation related to the branch currents observed in the Japan Trench should continue into the Kuril Trench.

A Taylor cap formed when the current against the seamount was weak (i.e., during non-jet periods). Height at which the water particles were trapped was considerably smaller than that of the entire water column, and the water particles above the Taylor cap were washed away by the impinging WBC. When the impinging current became stronger due to the confluence of the branch currents (i.e., the jet periods), the Taylor cap was fully washed away. The periods coincided with those during which the local Rossby number exceeded the critical value calculated from a previous study.

The mean velocities on the western flank of the trench indicated a directionally stable southward recirculation current just above the seafloor. Conversely, an eastward current was observed at a depth of 3360 m, the direction of which was along neither the isobath nor the trench axes. Since the eastward and southward currents extended downward and upward, respectively, without significant rotation, and temporal variation in the eastward current was correlated with that of the current around the seamount, although the southward current was not, these currents were different. The eastward current was locally caused by the confluence of the previously observed northward current in the Japan Trench and southward current in the Kuril Trench.