Deep intense vertical mixing in the Bussol' Strait



[1] To reveal the intensity of deep mixing in the Bussol' Strait, we performed the direct observations of turbulence for the first time. Vigorous vertical mixing with a turbulent energy dissipation rate ε above 10−6 W/kg and diapycnal diffusivity Kρ above 103 cm2/s were observed at depths of 600–1300 m (potential densities of 27.0–27.5 σθ) at the western gap of the Bussol' Strait, possibly affecting water-mass transformation and thermohaline circulation. The strong vertical mixing with a diurnal period occurred when the diurnal tidal current flowed toward the Okhotsk Sea, corresponding to a period of enhanced shear consisting of the mean flow with a two-layered structure and the large-amplitude diurnal tidal current with a vertical structure explained by a low-vertical mode topographically trapped wave. The empirical relationship ofε = 10−1.46±1.91 × S2.16±0.72 is obtained between ε and the shear Sreconstructed from the 50-m-scale mean, diurnal, and semi-diurnal tidal currents.

1. Introduction

[2] The Kuril Straits, which connect the Sea of Okhotsk and the North Pacific (Figure 1a), are a possible site of the deep mixing that maintains and controls global thermohaline circulation [e.g., Kawasaki and Hasumi, 2010]; indeed, numerical models suggest the occurrence of strong vertical turbulent mixing [Nakamura and Awaji, 2004; Tanaka et al., 2010, hereinafter T10]. Strong vertical mixing due to interaction between complicated topography and strong diurnal tidal currents [Katsumata et al., 2004, hereinafter K04] may influence the water masses in the Okhotsk Sea and the North Pacific [e.g., Tatebe and Yasuda, 2004], and influence bi-decadal water-mass and climate variability through the 18.6-year nodal tidal cycle [Osafune and Yasuda, 2006; Yasuda et al., 2006; Hasumi et al., 2008].

Figure 1.

Bathymetric maps of (a) the area of the Kuril Islands and (b) the Bussol' Strait. The red circle and red arrows denote the location of the BF1 station and the directions of along-strait flowuand across-strait flowv, respectively. (c) Vertical cross-section of the Bussol' Strait, along the black curve in Figure 1b, showing the calculated amplitude of normalized across-strait velocity of a topographically trapped wave with a diurnal period.

[3] The Bussol' Strait is the deepest and widest of the Kuril straits (Figure 1b), and the main exchange gate of water masses between the Okhotsk Sea and the North Pacific; the net transport is from the former to the latter with a total transport of 8.2–8.8 Sv (1 Sv = 106 m3s−1) and 4 Sv at densities of 27.0–27.5 σθ (K04). The occurrence of a peculiar, deep low-temperature water-mass at densities of 27.3–27.6σθ, which cannot be explained by isopycnal mixing of surrounding water-masses, is suspected to be caused by intensive diapycnal mixing [Ono et al., 2007]. Previous numerical studies reported strong vertical diffusivity Kρ over 200 cm2s−1 in the Bussol' Strait [Nakamura and Awaji, 2004; T10].

[4] Although previous studies have suggested strong turbulent mixing, there have been few direct observations of turbulence in the Kuril Straits. Turbulence data have been reported only for the shallow Urup Strait [e.g., Itoh et al., 2010]; turbulence has not been directly observed in the deep straits. To reveal the intensity of deep mixing and its relationship with the strong diurnal tidal current in the deep Kuril straits, we performed the direct observations of turbulence in the Bussol' Strait for the first time.

2. Data and Method

[5] Two sets of repeated observations over a 24-hour period were carried out at station BF1 (46°26.6′N, 151°7.4′E) on 19–20 August 2006 and on 13–14 August 2007 from aboard the R/VProfessor Khromov (Far Eastern Hydrometeorological Research Institute, Russia). Station BF1 is located at the western gap of Bussol' Strait, with a depth of 1600 m (Figure 1), where K04observed a large-amplitude diurnal tidal current in August–September 2001.

[6] Casts of a conductivity–temperature–depth (CTD) profiler (SBE 9plus, SeaBird Electronics) with an attached lowered acoustic Doppler current profiler (LADCP; 300 kHz Workhorse, Teledyne RD Instruments) were made alternately with casts of a vertical microstructure profiler (VMP2000, Rockland Scientific International). The VMP2000 is a tethered free-falling vertical profiler with two shear probes, and pumped standard CTD sensors (SBE-3 and SBE-4, Sea-Bird Electronics) provided vertical profiles of turbulent energy dissipation rateεVMPand density simultaneously. The sampling rates of the shear sensors and SBE-CTD are 512 and 64 Hz, respectively. The influence of instrumental noise and disturbance of the main body were removed from the shear data, which were then divided into 16-second segments (approximately equivalent to 10-m depth segments) overlapping by 14 seconds.εVMPdata at every 1-m interval was calculated by integrating the shear spectrum in each segment.

[7] Velocity data at 50-m intervals were derived from the LADCP, processed by conventional methods [Fischer and Visbeck, 1993] with box-averaged at 50-m depth intervals. Measured velocities were rotated such thatuis the along-strait flow andvis the across-strait flow with a rotation angle of −18.8°, followingK04 (Figure 1b). A harmonic analysis with least-squares fitting was performed to estimate the tidal components of the diurnal and semidiurnal periods, and mean velocity foru and v, as in the work of K04. The reconstructed velocities U and Vcomposed of the mean flow, diurnal and semi-diurnal tidal currents were used as the velocity in this study. This approach enables us to fill the lack of current data in the case where the turbulence was measured only by VMP. Horizontal velocity shearS is estimated as S2 = Uz2 + Vz2, employing the central difference method for U and V at an interval of 50 m.

[8] To supplement the direct turbulence measurements, we made indirect estimates of the dissipation rate εCTD based on density inversions observed in density profiles (Thorpe [1977], with the quality control tests of Galbraith and Kelley [1996]). This method assumes a linear relationship between the density inversion length scale LT [Thorpe, 1977] and the Ozmidov scale LO = (εN−3)1/2 (where N is the buoyancy frequency) [Ozmidov, 1965], as follows: LO = cLT. This linear relationship was confirmed by simultaneously obtaining LO from directly measured εVMP and obtaining LT from density inversion, yielding c = 0.67 and a 95% confidence range of c = 0.60–0.74 according to the bootstrap method [Efron and Gong, 1983]. Using the calibrated coefficient c, we obtained εCTD = c2LT2N3.

3. Results

[9] Vigorous turbulent mixing was observed at depths below 600 m and at potential densities of 27.0–27.5 σθ during the period when the current was toward the Okhotsk Sea at these depths (Figure 2). Both the turbulent intensity εand the cross-strait reconstructed currentV showed remarkable diurnal variability. Strong vertical mixing (ε > 1.0 × 10−6 W/kg and diapycnal diffusivity Kρ > 103 cm2/s) was observed at depths of 600–1200 m at 02:00–9:00 GMT on 19 August 2006 (Figures 2e and 2g) and at 09:00–16:00 GMT on 13 August 2007 (Figures 2f and 2h), corresponding to the strong, deep Okhotsk-ward current with a maximum speed above 1.5 m/s (Figures 2a and 2b). The strong turbulence also corresponded to a large velocity shear S (Figures 2c and 2d).

Figure 2.

Time series of reconstructed across-strait velocityV [m/s], shear S2 [s−2], turbulent energy dissipation rate ε [W/kg], and vertical diffusivity Kρ [cm2/s] (a, c, e, and g) on 19–20 August 2006 and (b, d, f, and h) on 13–14 August 2007. In Figures 2e–2h, columns with gray shading represent ε and Kρ estimated indirectly from density inversions with calibration (εCTD) and other columns represent ε and Kρ measured directly by shear probe (εVMP); gray shading indicates regions in which density inversions were not detected or did not pass the quality test. Black contours and numerals represent potential density σθ.

[10] Figure 3 shows the decomposed vertical profiles of V: the mean flow (Figure 3a) and the amplitude of the diurnal and semidiurnal tides (Figure 3b). The phases of these tides are vertically nearly uniform (not shown). The mean flow shows a two-layer structure, with the upper layer flowing toward the Pacific and the lower layer toward the Okhotsk Sea, with the zero velocity at a depth of around 800 m. The amplitude of the diurnal tidal current increased with depth, with a maximum of 1.3 m/s (1.6 m/s) at depths of 1200–1300 m in 2006 (2007). These vertical structures of the mean flow and the diurnal tide are similar to those observed byK04. The amplitude of the semidiurnal tidal current was 0.3–0.4 m/s (Figure 3b), much less than the diurnal amplitudes.

Figure 3.

Vertical profiles of the decomposed (a) mean velocity and (b) amplitudes of diurnal and semi-diurnal tidal components (denoted by M2 and K1, respectively) of the across-strait velocityVin 2006 and 2007. The green curve in Figure 3b indicates the normalized across-strait velocity amplitude of the diurnal topographic trapped wave at BF1, which is shown inFigure 1c. The vertical profiles of Vat the maxima of the diurnal tide show that the shear at the diurnal tidal current toward the Okhotsk Sea (O-ward) is two orders of magnitude greater than that toward the Pacific (P-ward). Vertical profiles of 1-day-averaged (d)εShear and (e) KρShear in 2006 and 2007 (blue and red lines).

[11] Statistically significant relationships were found between directly measured εVMP and the Richardson number Ri = N2/S2 (Figure 4a), and between εVMP and reconstructed current shear S2 (Figure 4b). Their least square fits are ε = aSb with a = 10−4.68 ± 0.93 and b = 0.96 ± 0.32, where the error bars denote 95% confidence intervals and ε = dRie with d = 10−7.22 ± 0.13 and e = −0.35 ± 0.14; the correlation coefficients are +0.40 and −0.38, respectively. Focusing on large shear (S > 10−6 s−2), the relationship of ε = fSg with f = 10−1.46 ± 1.91 and g = 2.16 ± 0.72 yields a better correlation coefficient of +0.48 (green line in Figure 4b), similarly to Itoh et al. [2010]. In contrast to Ri and S, the correlation between εVMP and Nis weakly negative (correlation coefficient: −0.14, p-value: 0.08 andεN−0.48 ± 0.54) (Figure 4c).

Figure 4.

Logarithmic scatter plots of εVMP versus (a) the Richardson number Ri, (b) shear S2, and (c) buoyancy frequency N2. Averaged profiles (blue lines) and regression lines (red lines) are also shown with their 95% confidence interval (dash-dotted lines). Green line shows the regression line ofεVMP on limited S2 over 10−6 s−2.

[12] By using εShear defined by the empirical relationship εShear = fSg for S > 10−6 s−2 and εShear = aSb for S < 10−6 s−2, we further examine why the mixing had a diurnal period and why vigorous turbulence occurred only during the Okhotsk-ward current. Since the diurnal tidal current and its shear have two maxima per day, the diurnal tide should cause strong vertical mixing with a semi-diurnal period. By combining the mean and diurnal tidal currents and considering the corresponding vertical shear changes, the diurnal mixing can be explained.Figure 3c shows current profiles at times of maxima in the diurnal tidal current, for the diurnal current against the mean flow in the lower layer (dashed curve in Figure 3c) and for the diurnal current in favor of the mean flow in the lower layer (solid line). It is clear that the vertical gradient of velocity superimposed of the mean flow and the Okhotsk-ward (Pacific-ward) diurnal tidal current was strengthened (weakened) (Figure 3c). This influence causes a large shear around mid-depths during the period of the Okhotsk-ward diurnal current.

[13] Using an empirical relation and reconstructed velocity fields, we estimated vertical profiles of 1-day average 〈εShear(z)〉 in 2006 and 2007 (Figure 3d). 〈εShear(z)〉 has a characteristic vertical structure with a relatively low 〈εShear (z = 100–500 m, σθ = 26.6–26.9)〉, large 〈εShear (z = 600–1200 m, σθ = 27.0–27.5)〉, and a vertical maxima of O (10−7 W/kg) at 27.2–27.4 σθ (930–1100 m in 2006; 800–1000 m in 2007). The vertical maxima correspond to steep gradients in both the mean current and the diurnal current amplitude (at around 1000 m in 2006 and 800 m in 2007; Figures 3a and 3b). Column-averaged 〈εShear〉 is 1.8 × 10−7 W/kg in 2006 and 1.1 × 10−7 W/kg in 2007. Vertically integrated 〈εShear(z)〉 is 0.29 W/m2 in 2006 and 0.18 W/m2 in 2007.

[14] The 1-day averaged vertical profiles of vertical diffusivity 〈KρShear(z)〉 with KρShear = 0.2 εShear/N2 [Osborn, 1980] are shown in Figure 3e. 〈KρShear(z)〉 is 10 cm2/s for depths of 100–500 m, increases with depth from 600 to 800 m, and takes a maximum of 442 cm2/s (211 cm2/s) around 27.3–27.4 σθfor depths of 900–1200 m in 2006 (2007). Column-averaged 〈KρShear〉 is 101 cm2/s in 2006 and 44 cm2/s in 2007.

4. Discussion

[15] The vertical structure of the diurnal tidal current may be explained by topographically trapped waves (TTW) with a diurnal period around the seamount in the Bussol' Strait, as suggested by T10. One possible example of the velocity distribution is shown for the first-mode TTW for the western gap in the Bussol' Strait (Figure 1c). The eigenvalue problem of the TTW model (T10) was solved for the observed stratification and bottom topography. The velocity amplitude of this wave increases with depth (green curve in Figure 3b) and is in good agreement with observed profiles of the diurnal internal tide (red and blue curves in Figure 3b). These results support the proposal that the diurnal internal tidal current is composed of lower-mode TTW with a diurnal period (T10). This kind of low-mode, diurnal, internal tidal current could be better reproduced by future numerical models with more accurate bathymetric data, because the bathymetry and TTW at BF1 in the numerical model (T10) is different from observations.

[16] The vertical structure of the mean flow in the present study is an important factor for inferring the mixing intensity via the empirical relationship between εShear and S. The mean flow structure is similar among the four sets of 24-hour observations in the present study andK04, and may be permanent, at least in summer. Mean flow structures might be explained by wind-driven or tidal residual flow [Nakamura and Awaji, 2004; T10]. Further studies are necessary because seasonal [e.g., Ohshima et al., 2010] and year-to-year variability of the exchange flow may alter the mean flow structure.

[17] The deep strong mixing with a maximum at around 27.3–27.4 σθ (Figure 3e) confirms the occurrence of strong vertical mixing from the peculiar low-temperature water mass at 27.3–27.6σθ in the Bussol' Strait [Ono et al., 2007]. This vertical structure may also explain the greater isopycnal thickness at 27.2–27.4 σθ on the North Pacific side of the Kuril Straits compared with the Okhotsk Sea side [Yasuda, 1997]. The difference in thickness arises because the water mass converges toward the density with maximum ε and water in this density range eventually flows out to the Pacific Ocean through the Bussol' Strait, mainly through the eastern part of the strait (K04). The observed turbulence intensity is large at depths from 700 m to 1400 m, near the bottom; that is, the intense turbulence extended upward to 800 m from the bottom. This extension (800 m) is much greater than that 200 m obtained from numerical models (T10). This discrepancy may be ascribed to differences in the bathymetry and/or the mean flow.

[18] The vertical profile of the dissipation rate ε is essential in determining the thermohaline circulation, because diapycnal velocity is formulated as w* = 0.2εzN−2 [Kunze et al., 2006]. The observed dissipation profile 〈εShear(z)〉 shows that large ε (>10−7 W/kg) occurs at depths greater than 700 m. The results of numerical simulations indicate that the Pacific thermohaline circulation is strongly dependent on the vertical profile of vertical diffusivity in the Kuril Straits [Kawasaki and Hasumi, 2010], where large diffusivity in the deep layer results in downward circulation in strong-mixing regions, whereas vertically homogeneous diffusivity (i.e.,ε decreases with depth, corresponding to decreasing N2) results in upward circulation. It is noted that the deep-level strong diffusivity reported byKawasaki and Hasumi [2010] was based on the present observations (in fact, the authors cited the M.Sc. thesis of the first author).

[19] The observed mixing in the deep layer is relatively strong, reaching Kρ ≥ 103 cm2/s, which is 104times the background diffusivity. The vertically averaged 1-day mean estimated turbulent intensity 〈KρShear(z)〉 is ∼101 cm2/s, which is larger than the value of 10–50 cm2/s reported by Nakamura and Awaji [2004], and the depth-integrated energy dissipation rate of ∼0.29 W/m2 is less than the value of ∼2 W/m2 calculated by numerical modeling (T10). These differences between observations and numerical models may be ascribed to differences in bathymetry and the mean flow. The mixing estimates obtained from advanced numerical models and the empirical εS relationship, as obtained in the present study, could markedly improve our knowledge of turbulence around the Kuril Strait.


[20] The authors greatly appreciate T. Nakatsuka, J. Nishioka, J. Volkov, and A. Scherbinin for organizing the joint cruise. We also thank the captain, officers, and crew of the R/V Professor Khromov, and all the shipboard scientists for various observations and analyses. Thanks are extended to S. Itoh, S. Osafune, H. Nagae, H. Kaneko, K. Ono, Y.W. Watanabe, K. Kuma, and Y. Tanaka for observations and useful discussions. This work was supported by the Ministry of Education, Science, Sports and Culture of Japan via a Grant-in-Aid for Scientific Research (KAKENHI 20221002).

[21] The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.