Using Argo float data, this study examined the formation region, spatial distribution, and modification of transition region mode water (TRMW), which is a recently identified pycnostad in the subtropical–subarctic transition region of the North Pacific, the basin-scale boundary region between subtropical and subarctic water masses. Analyses of the formation fields of water masses within and around the transition region reveal that TRMW forms in a wide area from the western to central transition region and is separated from the denser variety of central mode water (D-CMW) to the south by a temperature and salinity front. TRMW has temperatures of 4°–9°C and salinities of 33.3–34.0, making it colder and fresher than D-CMW. TRMW has a density range of 26.3–26.6 σθ, and thick TRMW is widely distributed in the transition region. However, the range of the T–S properties at TRMW cores is substantially reduced downstream within 10°–20° longitude from the formation region by gradually losing its fresh and cold side. It is also demonstrated that a major part of TRMW of 26.4–26.6 σθ is entrained into the mixed layer in the following winter. Quasi-Lagrangian observation by an isopycnal-following Argo float demonstrates that the double-diffusive salt-finger convection plausibly causes not only rapid erosion of the TRMW pycnostads but also an increase of salinity and temperature at the TRMW cores, at least to some degree. It is demonstrated that strong salt fingering within TRMW is probably caused by geostrophic currents with vertical shear crossing the density-compensating T–S front that brings warm and saline water to the upper TRMW and creates instability in the salinity stratification. This modification process could explain why water that is subducted from the transition region and constitutes the pycnocline of the subtropical gyre in the North Pacific has different T–S properties from the winter mixed layer of the transition region. This knowledge about the modification process of subducted water in the transition region would help to model the permanent pycnocline structure more realistically and to clarify how large signals of decadal and multidecadal variability of sea surface temperature in this region are propagated into the ocean interior.
The subtropical–subarctic transition region in the North Pacific is the basin-scale boundary region between subtropical and subarctic waters where the upper region forms a transition from warm and saline subtropical water to cold and fresh subarctic water. This transition region has been described in regard to fisheries oceanography using isohaline or isotherm criteria (e.g., Favorite et al. 1976). The northern boundary of the subtropics (the southern boundary of the transition region) is called the subarctic boundary (SAB), defined as the near-surface salinity front represented by the 34.0 isohaline. The southern boundary of the subarctics (the northern boundary of the transition region) is called the Subarctic Front (SAF), defined as the location where the 4°C isotherm is almost vertical below a depth of 100 m. On climatological wintertime mixed layer property maps (e.g., Suga et al. 2004), the western and central part of the transition region is delineated by the 4°C isotherm (SAF) and the 34.0 isohaline (SAB) from the coast of Japan to 170°W (Fig. 1). A clear intermediate layer of the salinity minimum appears to the south of the SAB, which is one of main characteristics of subtropical stratification, and a temperature inversion appears predominantly to the north of the SAF, which is one of main characteristics of subarctic stratification (e.g., Favorite et al. 1976). In the synoptic field, the transition region contains several coupled fronts of temperature and salinity (T–S) where the density changes caused by temperature and salinity change across these fronts nearly compensate each other (Roden 1972; Yuan and Talley 1996). This characteristic originates from the “water mass front” formed in the transition region in which water masses with different T–S properties lie adjacent to each other on the same density surface (Zhang and Hanawa 1993). The Kuroshio–Oyashio mixed water region approximately corresponds to the westernmost part of the transition region.
Recently, a new type of low potential vorticity (PV) water or pycnostad was identified in the western transition region from a research survey near 40°N, 160°E (Saito et al. 2007, hereafter SSHW). The potential temperatures, salinities, and densities of the low PV water were 5°–7°C, 33.5–33.9, and 26.5–26.7 σθ and different from those of any other known mode waters in the North Pacific. The low PV water extended into a relatively large area, ranging over 650 km in the zonal direction and 500 km in the meridional direction, located between SAB and SAF. High concentrations of oxygen indicated that the low PV water originated in the deep mixed layer that developed to more than 200 m in depth during the preceding winter. Since these characteristics are known as those of mode water (Hanawa and Talley 2001), SSHW regarded this colder and fresher low PV water as a new type of mode water in the North Pacific and termed it transition region mode water (TRMW).
The possible formation region of TRMW appeared in the climatology of winter mixed layer depth (MLD) examined by Suga et al. (2004), suggesting that TRMW is not a transient phenomenon but rather forms regularly. In their climatology, the region of large winter MLD (>150 m) extends widely from the Japanese coast to near 160°W in the zonal direction and 35°–45°N in the meridional direction (Fig. 1). The T–S properties of this wide area of large winter MLD suggest that both TRMW and North Pacific Central Mode Water (CMW) (Nakamura 1996; Suga et al. 1997, 2004) are formed in this area as follows. Suga et al. (2004) defined typical T–S properties of CMW as those of the low PV core on each of isopycnals corresponding to CMW over the subtropical gyre in the isopycnally averaged climatology: 8°–13°C and 34.0–34.5 (see their Fig. 6). They then argued that CMW is formed mostly in the southern half of this area of large winter MLD by identifying the grid points where the winter mixed layer has the T–S properties of CMW (see their Fig. 7). This possible formation region of CMW is approximately delineated by the 26.0-σθ isopycnal, the 34.0 isohaline, and the winter MLD greater than 150 m in Fig. 1. On the other hand, SSHW argued that TRMW is possibly formed in the northwestern part (39°–43°N, 152°–162°E) of this area by comparing the wintertime mixed layer climatology with their synoptic observation of TRMW and the flow field.
Subduction of TRMW and CMW into the permanent pycnocline is demonstrated by Suga et al. (2008), whereas they did not recognize TRMW as a type of mode water. They calculated the annual subduction rate of the North Pacific based on isopycnally averaged hydrographic climatology, high-resolution winter mixed layer climatology, and various wind stress climatologies, following the method using Lagrangian coordinates introduced by Huang and Qiu (1994). Their maps of the subduction rate overlaid on the winter mixed layer properties (their Fig. 4) show that most of the water in the possible formation region of TRMW mentioned above is subducted into the permanent pycnocline at moderate rates. On the other hand, the same maps show that the water in the CMW formation region described above is subducted into the permanent pycnocline at much lower rates, except the water mostly lighter than 26.2 σθ in the easternmost part at 180°–160°W, which is bounded by a sharp winter MLD front. It is noteworthy that the water denser than 26.4 σθ subducted into the permanent pycnocline is mostly fresher than 34.0 and thus apparently dominated by TRMW (Suga et al. 2008), whereas the low PV water, actually spreading widely over the subtropical permanent pycnocline denser than 26.4 σθ, which is the densest part of CMW, has salinity greater than 34.0 (Suga et al. 2004).
The following questions are posed by the apparent mismatch mentioned just above between the properties of the water subducted from the mixed layer indicated by the annual subduction rate estimation (Suga et al. 2008) and those of the water spreading into the subtropical permanent pycnocline in the climatological fields (Suga et al. 2004) for the density greater than 26.4 σθ. Is the mismatch artifact due to smoothing applied to produce the climatological field? For example, since the transition region is the frontal zone and has large variability in the water mass distribution (e.g., Suga et al. 2003), the climatology possibly does not represent actual distribution of the winter mixed layer properties and, thus, actual properties of the subducted water. If the estimation of the subduction rate by Suga et al. (2008) represents reality, what is the fate of the water denser than 26.4 σθ and fresher than 34.0 subducted from the transition region? Since the annual subduction rate corresponding to this water is largest on the isopycnals denser than 26.4 σθ, it could show up as low PV water: namely, TRMW on these isopycnals. However, the isopycnal climatology does not show a notable low PV signature with salinity lower than 34.0 in either the subtropics or the subarctics (Suga et al. 2004). Is there any mechanism that modifies and/or erodes TRMW more rapidly than CMW? These questions have to be answered to understand how the lower part of the subtropical permanent pycnocline is ventilated because the isopycnals forming the lower permanent pycnocline mostly outcrop in the transition regions (Fig. 1).
There is a clue to the question about possible rapid modification of TRMW. Talley and Yun (2001) showed that salt-finger convection actively occurred in the upper part of the intrusion of Oyashio source water that substantially affected the properties of newly formed NPIW in the mixed water region. In the eastern subtropical region of the South Pacific, which is also the boundary region between different water masses, South Pacific Eastern Subtropical Mode Water (SPESTMW) formed there is influenced by salt-finger convection (Wong and Johnson 2003). Johnson (2006) found that T–S anomalies in SPESTMW decay significantly within 6 months by salt-finger convection. In the transition region where water masses with different T–S properties but similar densities are adjacent, intrusions across the T–S front in the nearly isopycnal layers would cause double-diffusively unstable stratification. Therefore, TRMW may also be affected by double-diffusive convection as in the case of NPIW and SPESTMW.
The purpose of the present paper is to answer the aforementioned questions posed by the apparent mismatch between the annual subduction rate estimation and the climatological property distribution in the lower part of the subtropical permanent pycnocline. For that purpose, we investigate the formation, distribution, and modification of water masses within and around the transition region, especially focusing on those of TRMW, with sufficient amount of unsmoothed profile data both of temperature and salinity supplied by Argo floats in this region. By clarifying the formation of the mode water in the transition region and its fate, the present study will contribute to understanding roles of this complicated T–S frontal region in the ventilation of the deepest part of the subtropical permanent pycnocline in the North Pacific.
The remainder of this paper is organized as follows: Section 2 describes the procedure and data. Section 3 illustrates the formation of water masses within and around the transition region. Section 4 presents the wide distribution of TRMW in the North Pacific as well as its T–S properties and possible modification. The vertical structure of TRMW is also examined. Section 5 quantitatively examines the modification of TRMW through salt fingering using quasi-Lagrangian observation provided by an isopycnal Argo float. All available Argo float data for the study region were used to provide evidence of modification in the transition region. In section 6, the mechanism of modification is discussed based on Argo float data and climatological data. Finally, section 7 presents a discussion and conclusions.
2. Data and processing
Argo floats operating in the midlatitude North Pacific (30°–50°N, 130°E–120°W) from January 2001 to December 2005 collected 14 940 hydrographic vertical profiles. The Argo floats usually drift at pressures of 1000 or 2000 dbar and rise to the surface every 10 days to observe pressure, temperature, and salinity. The Argo floats include eight isopycnal floats designed to follow specific isopycnal surfaces. The data obtained by the isopycnal floats are regarded as quasi-Lagrangian observations of the water surrounding the target density and are used in section 5.
Argo floats provide vertical profiles of temperature and salinity typically every 5–10 dbar in the upper 200 m, 10–25 dbar between 200 and 1000 dbar, and 50–100 dbar below 1000 dbar. Oka et al. (2007) quality controlled the data used here. They removed defective profiles through several quality checks; for example, they examined whether the vertical resolution was lower than that for typical Argo observations; checked sensor drifts of salinity and flags of temperature, salinity, or pressure; and removed obviously bad data by visual inspection. Each profile of temperature and salinity was then interpolated to every 10 dbar by the Akima spline method (Akima 1970). Potential temperature and potential density were subsequently calculated. Potential vorticity Q was calculated by the following equation when the relative vorticity was negligible and hydrostatic approximation was appropriate:
where f is the Coriolis parameter, g is the gravity acceleration, ρθ is the potential density, and p is the pressure. The vertical gradient of the potential density was calculated from the linear fitting using the least squares method with nine vertical grid points. This method filters out fluctuations with a vertical scale less than 80 dbar and reduces small-scale noise as much as possible. This process helped us detect the depths of the vertical PV minima and thus accurately obtain pycnostad properties.
Pycnostads were detected as the low PV layers forming the vertical PV minima based on two conditions: 1) the PV at the core layer (the vertical minimum of PV) was less than 1.5 × 10−10 m−1 s−1 and 2) the thickness of the low PV layer with PV less than 1.5 × 10−10 m−1 s−1 was greater than 50 dbar. The second condition was imposed to capture the low PV water with a vertical scale greater than 50 dbar. Otherwise, the transient low PV water with small scale caused by events such as internal waves or wave breakings is captured and the signal of mode water pycnostad is buried. The T–S properties at the pycnostad core were analyzed as the representative properties of the pycnostad. The core should be the least eroded by surrounding water and should mostly conserve the properties of the pycnostad. The pycnostad thickness was defined as the width of the layer with a density difference of 0.1 σθ that spans ±0.05 σθ from the potential density at the core.
The vertical density ratio Rρ(z) and Turner angle in the vertical direction Tu(z) were calculated as indices of double-diffusive mixing according to the following equations:
where α and β are the thermal expansion coefficient and the haline contraction coefficient, respectively, calculated every 10 dbar. For the Rρ(z) calculation, the vertical gradients of T and S were evaluated as the difference between T and S at p + 10 dbar and p − 10 dbar at each pressure level p to obtain the minimal possible gradient of the vertical scale for measurement of the potential for double-diffusive convection.
The time evolution of salinity caused by salt fingering was estimated using the method of Johnson (2006). The diapycnal salt diffusion Ks was parameterized as a function of Rρ(z),
for 1 < Rρ(z) < 2.05 [71° < Tu(z) < 90°] by fitting to the observations of St. Laurent and Schmitt (1999). In their observations, Rρ(z) was calculated on a vertical scale of 5 m using data obtained by a high-resolution profiler. This vertical scale is significantly larger than the microscale at which salt fingering occurs, similar to our calculation of Rρ(z) based on Argo float data. Thus, it seems proper to adopt this parameterization of Ks in our calculation. Salinity change due to vertical diffusion, including the effect of salt fingering, was simulated by integrating the following one-dimensional diffusion equation for each density surface:
The density surface properties were prepared by linearly interpolating the pressure surface properties to every 0.01-σθ surface.
3. Formation field of water masses within and around the transition region
The formation field of water masses was deduced from the field of the late winter mixed layer, which was obtained from individual Argo float observations in March (Fig. 2). A large area of MLD >150 dbar extended from 140°E to 160°W near 40°N with a range of 25.7–26.6 σθ, corresponding to the density ranges of CMW and TRMW. The large winter MLD region corresponds well with the region deduced from the climatological winter MLD and contains the formation region of both CMW and TRMW, judging from the density and salinity of the mixed layer.
In the potential density histogram, the deep mixed layer with a range of 25.7–26.6 σθ was divided into two separated peaks at 25.7–26.2 σθ and 26.2–26.6 σθ (Fig. 3). Oka and Suga (2005), from analysis of synoptic survey data along 165°E, reported that CMW could be divided into lighter (L-CMW) and denser (D-CMW) varieties separated by the Kuroshio Bifurcation Front (KBF). The two modes of the deep mixed layer in Fig. 3 correlated with the density ranges of L-CMW and D-CMW. The main outcrop area of 25.7–26.2 σθ, which is the density range of L-CMW, was found in the central North Pacific at 33°–39°N, 180°–160°W (stars in Fig. 2). This formation area of L-CMW corresponds to the classic formation area of CMW, as estimated by previous studies (e.g., Suga et al. 1997).
The outcrop area of 26.2–26.6 σθ was further divided into two regions with salinities of >34.0 and <34.0 by the salinity front with a minimal increase of 0.1 (light gray and dark gray circles in Fig. 2). The former large winter MLD region with high salinity was judged to be the formation region of D-CMW from its T–S properties. The latter large winter MLD region with low salinity was determined to be the formation region of TRMW. Although SSHW roughly estimated the TRMW formation region as 150°–160°E near 43°N based on limited data, the actual formation region extended in a wider area from 143°E to 180° near 43°N in the transition region with an area comparable to the formation region of D-CMW. Because TRMW and D-CMW were divided by the salinity gap, including 34.0, it is appropriate to classify pycnostads with S < 34.0 as TRMW. This means that D-CMW has salinity higher than 34.0, which is consistent with the typical properties of CMW identified by Suga et al. (2004) based on the isopycnally averaged climatology.
TRMW can be distinguished as a different group from D-CMW in the T–S diagram of the deep winter mixed layer (Fig. 4). In the T–S field with salinities greater than 34.0, the properties of the deep mixed layer coincided well with CMW properties reported by previous studies (e.g., Suga et al. 2004) (crosses in Fig. 4). Conversely, there were many profiles with significantly deep mixed layers over 200 dbar in the field of salinities less than 34.0. These profiles had properties closer to those of the TRMW reported by SSHW (gray circles in Fig. 4) than to those of CMW. The observed TRMW had potential temperature and salinity ranges of 4°–9°C and 33.3–34.0, showing wider ranges than those of the TRMW T–S properties reported by SSHW. However, this wide range falls within the traditional range of T–S properties in transition region upper water; the transition region is a frontal region and thus has a wide range of T–S properties (Yuan and Talley 1996). TRMW apparently occupied a large amount of the water mass with a density range of 26.2–26.6 σθ (Fig. 3). In particular, almost TRMW only was formed in the densest layer, 26.5–26.6 σθ since this density range outcrops only in the transition region in the open ocean of the North Pacific. This is essentially what was observed in the climatological winter mixed layer, as described in section 1, and thus demonstrates that the climatology represents the actual distribution of the winter mixed layer properties.
4. TRMW Distribution
a. TRMW spatial distribution and evidence of its rapid modification
As illustrated in section 3, the TRMW formation region extends zonally to 180°. Although TRMW extends next to the formation region of D-CMW, which is formed in a similar density layer, these two mode waters have distinct water properties. In this section, we describe the spatial distribution of TRMW and its interrelationship with the D-CMW distribution on isopycnal surfaces.
The distribution of pycnostads in the range of 26.3–26.6 σθ shows that the thick TRMW was distributed from the formation region (green crosses) eastward along the streamline representing isopycnal geostrophic flow relative to 2000 dbar for about 10°–20° longitude of each density layer (Fig. 5). For the entire density range of 26.3–26.6 σθ, thick TRMW covered a wide area of the transition region and was confirmed to be the substantial water mass prevailing in the transition region. On the other hand, D-CMW tended to be distributed southeastward from its formation region across the streamline. D-CMW was relatively thin compared with TRMW but broadly distributed in the subtropical gyre.
The spatial distribution of the two types of mode water for each density layer is characterized as a more limited TRMW distribution and wider distribution of D-CMW, despite the latter’s formation in a small area disconnected to its distribution as follows. The density of the TRMW formation region became lower to the east; the densest part of TRMW reached 26.5–26.6 σθ and was formed in the westernmost region near 160°E. The densest surface widely outcropped in the western transition region with especially large winter MLD (Figs. 4 and 5c). Thick TRMW with the density of 26.5–26.6 σθ was widely distributed along the streamline to 175°E where it suddenly disappeared, as discussed in the next subsection. As observed by SSHW, the thick TRMW near 160°E was also within this density layer. Although only TRMW was formed in this density range of 26.5–26.6 σθ, thin D-CMW was also widely distributed on this density surface, mainly to the south of 40°N. On the other hand, there was a large amount of thick D-CMW on the 26.3–26.4-σθ density layer (Fig. 5a), although the D-CMW formation region (blue crosses) was minor in this density layer and found only near 40°N, 150°E. The distribution of thick D-CMW began close to the formation region of TRMW (green crosses).
Figures 6 and 7 show the longitudinal dependences of the T–S properties at the pycnostad cores in each density layer. In the all density layers, T–S properties of TRMW cores showed a tendency to gradually lose their fresh and cold portion from the formation region toward the east. As a result, while the pycnostad cores had wide range of T–S properties in the west, the range appeared to be considerably reduced downstream to the east within 10°–20° longitude from the formation region. Considering the geostrophic current speed in this region, the substantial reduction of the T–S property range of TRMW cores is expected to occur within the time span of a year. This reduction implies that TRMW with fresher and colder core properties has been eroded away faster than that with more saline and warmer ones and/or that the fresher and colder core properties of TRMW have been modified to be more saline and warmer. The latter process, that is, fairly rapid modification of the pycnostad core properties, may sound peculiar. However, especially in the densest layer (Figs. 6c and 7c), where almost all of the deep mixed layer had low salinities and temperatures, pycnostads with high salinity and temperature cannot be formed directly in the mixed layer, suggesting that such modification actually occurs to some extent. We will argue that it can happen in section 5.
b. Vertical distribution of TRMW
Here, we analyze interrelations between TRMW in vertically adjacent density layers. On a particular isopycnal surface, downstream, eastward spreading of thick TRMW was confined up to about 10° longitude from the formation region and then almost disappeared beyond that point (Fig. 5). On the densest surface, disappearance of TRMW was particularly clear and confirmed by sufficient observations (Fig. 5c). This region corresponded to an area where the density layer immediately above, 26.4–26.5 σθ, outcropped in winter with the deep mixed layer reaching 200 dbar (Fig. 5b). The sudden absence of thick TRMW was possibly associated with the development of this deep mixed layer with slightly low density.
Figure 8 shows the vertical overlapping of TRMW in adjacent density layers. While the cores of thick TRMW occurred in nearly the same depth range (100–250 dbar) for all density ranges, their longitudinal distributions differed. The large winter MLD (>200 dbar) region in the density range of 26.4–26.5 σθ overlaps with the region of the thick TRMW in the density range of 26.5–26.6 σθ. Since the thick TRMW of 26.5–26.6 σθ lay in the depth range of about 50–300 dbar, judging from its core depth and thickness, the winter mixed layer with density of 26.4–26.5 σθ developed to the depth over 200 dbar entrained a substantial part of the thick TRMW of 26.5–26.6 σθ. The similar phenomenon was observed for the thick TRMW of 26.4–26.5 σθ, which apparently has been entrained into the mixed layer of 26.3–26.4 σθ. This phenomenon is confirmed in a quasi-Lagrangian depth–time section of density, introduced in section 5a, which captured that the upper part of TRMW of 26.3–26.5 σθ formed around 42°N, 175°E was entrained into the winter mixed layer of 26.3 σθ around 41°N, 170°W, although the lower part of TRMW (26.4–26.5 σθ) remained as relatively thin pycnostad of about 75 dbar.
As shown in Fig. 8, most pycnostads within the 26.4–26.6-σθ layer seemed to have been entrained into the mixed layer developed in the downstream region, although the lower parts of those pycnostads seemed to remain as relatively thin pycnostads with thickness lower than 75 dbar near the 300-dbar depth (Fig. 8). In contrast, pycnostads in the 26.3–26.4-σθ layer were not entrained into the mixed layer and were advected to the east of 160°W, then to the southwest (Fig. 5a). This finding can be explained by the distribution of large winter MLD, which terminated near 170°W to the north of 40°N (Fig. 2). Since the 26.3–26.4-σθ layer outcropped at the eastern end of the large winter MLD region (Fig. 5a), the pycnostads formed there were no longer entrained into the deep mixed layer in the downstream region. Consequently, the subduction of pycnostads formed within and around transition region would mainly occur in this eastern end of the large winter MLD on the density surface of 26.3–26.4 σθ, where the horizontal gradient of winter MLD combined with geostrophic current becomes maximum in this large winter MLD region. On the other hand, there would be moderate subduction in the western to central parts of the transition region on the density surface of 26.4–26.6 σθ where winter MLD decreases moderately to the east from 200–300 dbar in 140°–160°E to 150–250 dbar in 160°E–180°. These speculations are fairly consistent with the annual subduction rate estimated by Suga et al. (2008), which shows a zonal band of moderate subduction rate at 39°–45°N, 150°E–155°W with significantly large subduction rate greater than 75 m yr−1 near the eastern end of the band around 42°N, 175°W and around 40°N, 165°W (their Fig. 3d).
5. TRMW modification through double-diffusive convection
a. Quasi-Lagrangian observation of TRMW
This section investigates the mechanism of TRMW modification based on quasi-Lagrangian observations of TRMW by an isopycnal Argo float. First, an example TRMW vertical profile is presented to illustrate the possibility of TRMW modification through double-diffusive convection. Figure 9 shows vertical profiles of various properties observed at 41.5°N, 179°W by the isopycnal Argo float WMO 2900220. This profile captured TRMW with a core density of 26.4 σθ, temperature of 8.5°C, and salinity of 33.95. A thick and vertically homogeneous pycnostad was found in the density profile at 140–250 dbar. However, temperature and salinity profiles were fairly stratified inside this TRMW. The temperature was stratified to stabilize the density stratification, whereas the salinity was stratified to destabilize it. The effects of temperature and salinity on the density stratification nearly compensated each other. Consequently, the vertical Turner angle Tu(z) was quite high (over 77°) throughout the entire TRMW pycnostad centered around its core. This phenomenon is also seen in pycnostads contained in the warm core ring in the mixed water region, the most western part of the transition region (Talley and Yun 2001, Fig. 6).
Schmitt (1981) showed that salt-finger convection was active when Tu(z) exceeded 77°. Vertically homogeneous Tu(z) inside the TRMW pycnostad also provides evidence that active salt-finger convection had occurred, because this characteristic is achieved after active salt fingering takes place (Schmitt 1981). Locally increased Tu(z) leads to more active salt-finger mixing, which relieves the instability, hastening the smoothing of the vertical Tu(z) profile.
From this profile, we suspect that the vertical salt flux caused by salt fingering is the possible cause of the increase of salinity and temperature at the core of TRMW suggested in section 4a. We examined the durability of this situation and its possible impact on the T–S properties of TRMW based on an example of quasi-Lagrangian observations by the isopycnal Argo float WMO 2900220 (Fig. 10). This float followed the 26.5-σθ layer and was advected just north of the Kuroshio Bifurcation Front around 180°.
The time evolutions of density and Tu(z) observed by the float are shown in Fig. 11. The float observed the formation of TRMW of 26.40 σθ and the thick (>150 dbar) TRMW until the following winter. The observation of this TRMW pycnostad was expected to be nearly Lagrangian because this float followed the 26.5-σθ surface. The Tu(z) calculated from the profiles showed values over 77° within the entire TRMW layer and especially high values near its core. Remarkably, this situation continued from early spring until late autumn. The very active salt fingering presumably transported salt from the upper layer downward.
To estimate salinity change in the TRMW core due to salt fingering, the vertical salt flux caused by salt-finger convection was calculated by estimating the vertical salt diffusivity from Eq. (2.4). Then, the salinity change due to vertical diffusion was simulated by integrating the one-dimensional diffusion Eq. (2.5). Figure 12 shows the simulated salinity change on the isopycnal surface near the TRMW pycnostad core compared with the observed salinity change. The estimated downward salt flux was convergent near the pycnostad core during the simulation period, causing the increase of salinity. A gradual but large salinity change over 0.1 during the year was simulated and corresponded well with the observed salinity change. Specifically, simulated salt fingering best explained the observed salinity increase near the TRMW core during the year, indicating a high likelihood of the core property modification due to salt fingering, although other explanations for the increased salinity, such as deviation from the “true Lagrangian” measurement, were not completely excluded. The rate of salinity increase was largest in spring, followed by a moderate rate throughout the rest of the year. On the other hand, short-term large variability (less than 10 days) was also observed, which was possibly due to cross-frontal isopycnal intrusions.
b. TRMW modification over a wide area of the transition region
The previous subsection used the quasi-Lagrangian observation to demonstrate the likely rapid modification of T–S properties of TRMW core due to salt fingering within a pycnostad. The modification process was particularly intense when Tu(z) exceeded 77°. This subsection uses all available Argo float data to examine whether the entire transition region is likely to undergo the same phenomenon. Figure 13 shows the Tu(z) distribution at the cores of all detected pycnostads. Throughout the year, Tu(z) > 77° appeared in a wide area of the transition region from the western part near 160°E to the central part near 180°. The conditions for active salt fingering were thus widely met over the transition region. Therefore, the case presented in the previous subsection is not unusual, and modification by salt fingering can be expected within the pycnostads of this wide region. The widest area of large salt-fingering potential [i.e., where Tu(z) > 77°] occurred in spring. This region narrowed in summer. In fall, much fewer Tu(z) values over 77° appeared in this region, although high angles of Tu(z) were continuously observed in some locations.
To illustrate the development of the modification process over a wide area, Fig. 14 shows the time evolution of the salinity distribution at pycnostad cores in the density range of 26.3–26.6 σθ. In April, pycnostads were observed immediately after their formation. The spatial salinity distribution showed that TRMW (pycnostads of S < 34.0) was distributed mostly north of 40°N (Fig. 14a) and D-CMW (pycnostads of S > 34.0) was found mostly to the south of 40°N. D-CMW was formed as a narrow band northward of the Kuroshio Extension Front in the region 140°E–180° and a wide band in 180°–150°W. In the salinity histogram of the pycnostad cores, TRMW formed at 33.3–34.0 with a relatively flat distribution throughout the salinity classes, whereas D-CMW peaked around 34.0–34.2 (Fig. 14b). There were 1.8 times as many TRMW pycnostads as D-CMW pycnostads, which suggest that the volume of TRMW formed in spring is almost twice the amount of D-CMW. In about 30% of these pycnostads Tu(z) > 77° was observed, indicating that active salt fingering occurred; they were presumably modified by heat and salt from the overlying layer.
In May, the salinity distribution at the pycnostads changed from that in April, generally becoming saltier. For example, the salinity symbols in Fig. 14c turn to green (33.6–33.8) northward of 42°N, whereas they were blue (<33.6) in April in the same region. The symbols turn red (34.0–34.2) near 40°N, 170°E–170°W, although they were yellow (33.8–34.0) in April. Since the higher salinity proceeded northward across the streamline, the salinity change at the core was not caused by advection. The histogram in Fig. 14d also confirms the salinity change. The TRMW peak at 33.4–33.6 in April had decreased by May, and the peak shifted to the higher side of 33.6–33.8. On the other hand, a D-CMW peak in May was slightly higher than that in April. The occurrence rate of 60° < Tu(z) < 70° increased, especially in D-CMW pycnostads from April to May. This feature can be regarded as a trace of modification by salt fingering; that is, salt fingering lessened the strong instability in salinity and the moderately unstable remaining salinity stratification no longer yielded active convection (Schmitt 1981).
After May, modification by salt fingering apparently occurred continuously until late autumn, although the most dramatic change occurred in early spring. The region of high salinity indicated by yellow (33.8–34.0) increased in 40°–42°N near 165°E in July. In the histogram, the salinity peak shifted to 33.7–33.9, and Tu(z) > 77° was still frequently observed, especially in the TRMW pycnostads. Salinity symbols in the region near 43°N, 165°E–180° turned to yellow (33.8–34.0) in September. Red symbols expanded northward of 42°N, near 175°W. The fresh part of TRMW with 33.3–33.7 diminished from September to November. The number of D-CMW pycnostads became similar to the number of TRMW pycnostads in November. The occurrence of Tu(z) of 60°–70° in D-CMW pycnostads reached 50%.
6. Onset and maintenance mechanism of salt fingering within TRMW
a. Mechanism of salt-fingering maintenance
Why is salt fingering so active in the transition region? The transition region contains a background field in which the horizontal density ratio is nearly equal to one and contains strong temperature and salinity fronts, as mentioned in section 1. In this situation, flow with vertical shear or isopycnal intrusion that crosses the density-compensating T–S front can yield a double-diffusively unstable condition. Furthermore, in a vertically homogeneous layer, the vertical density ratio nearly equals one because the density stratification becomes extremely weak. In such a condition, a double-diffusive instability easily develops. Nevertheless, it is unclear why such strong salt fingering occurs coherently over this wide area. This section investigates the above question and presents a mechanism that would maintain active salt fingering in this region. The series of profiles obtained by the isopycnal Argo float discussed in section 5a were examined in detail, with specific focus on the onset of salt fingering in early spring.
Figure 15 shows potential density, potential temperature, salinity, and Tu(z) [only for Tu(z) > 60°] profiles obtained by the float WMO 2900220 from February to May 2003. In late winter, both density and salinity profiles were vertically uniform for the upper 200 dbar due to winter mixing (profiles 1–4). Salinity and temperature in the upper 200 dbar drastically changed within 10 days in March, although there was no significant change in the density profile (profile 5). Simultaneously, Tu(z) became greater than 77° in the upper 200 dbar; this was an inevitable consequence because the vertical salinity gradient was unstable while the density was vertically homogeneous. The unstable salinity stratification within the pycnostad and consequent high values of Tu(z) continued even after the TRMW pycnostad subducted under the seasonal pycnocline in late April (profile 9). The TRMW pycnostad becomes smooth and loses the sharp vertical homogeneity as the salt-fingering convection goes by because double-diffusive convection works to stabilize the density. However, its PV is still lower than 1.5 × 10−10 m−1 s−1, so the TRMW is recognized as a pycnostad. In this observation, the onset of salt fingering rapidly appeared within 10 days in early spring.
In this time series, the sudden stratification of salinity and temperature in early spring was the key phenomenon that generated a high value of Tu(z) and active salt fingering within the TRMW pycnostads. The next question is what factors would cause such sudden stratification of salinity and temperature.
b. Sudden stratification of salinity and temperature in early spring
Surface fluxes and advection are two possible factors that could account for the sudden stratification of salinity and temperature profiles. Generally, mode waters are thought to be capped by the formation of a seasonal thermocline due to positive net heat flux in spring (Hanawa and Talley 2001). However, the typical annual cycle of net heat flux in the formation region of TRMW and D-CMW showed almost zero net heat flux in March and April (Fig. 16). Therefore, surface heating cannot account for the sudden stratification of temperature in early spring. Salinity also changes at the same time, completely compensating the temperature change in terms of density. Such completely compensating fluxes of heat and salt are quite unnatural.
Therefore, the remaining plausible cause of the sudden stratification is advection. Two conditions may produce the sudden stratification of temperature and salinity within the pycnostad. First, there should be a flow crossing the density-compensating T–S front. Second, the flow must have strong vertical shear in the upper few hundred meters. Flow that satisfies both of these conditions would advect warm and saline water from the south more strongly in the upper part of the pycnostad, resulting in a density-compensating vertical gradient of temperature and salinity. In this subsection, the existence of such flow is investigated based on synoptic and climatological fields.
Figure 17 shows an example of synoptic observations along 162°E; this line was also analyzed by SSHW. The density section (Fig. 17a) indicates that the Kuroshio Bifurcation Front was located around 40°N. To determine whether vertical shear occurred in the Kuroshio Bifurcation Current (KBC), geostrophic current speeds were calculated in reference to 300 dbar. Strong vertical shear was identified and reached 10 cm s−1 at 100 dbar (relative to 300 dbar). Although this was a zonal flow crossing the meridional section, this density section clearly illustrated that the KBC had a vertical shear structure in the upper 300 dbar. Therefore, the second condition was satisfied. In addition, the meridional gradients of temperature and salinity were higher in the upper layer (0–150 dbar in Figs. 17b,c), which would yield larger advection of temperature and salinity in the upper layer. These combined effects could produce vertical gradients of temperature and salinity if the flow crosses horizontal gradients of temperature and salinity.
We roughly estimated the advection term. Over 10 days in late March, approximately 0.1 of salinity change was observed (Fig. 15). The meridional salinity gradient at the salinity front associated with the KBC in the synoptic section was approximately 0.5 over 30 × 103 m at 100 dbar (Fig. 17c). The meridional component of geostrophic speed needed to cause a salinity change of 0.1 was 7.0 cm s−1 relative to 300 dbar, where the salinity change was small. This value is plausible as strong baroclinicity was confirmed in the KBC (Fig. 17a).
Figure 18 shows the relationship between the geostrophic current direction and the horizontal distributions of temperature and salinity at 100 dbar based on the annual-mean HydroBase climatology (Suga et al. 2004). The geostrophic current clearly crossed the zonal fronts of temperature and salinity in a wide region north of 40°N. Therefore, the first condition was also satisfied. The horizontal advection terms of temperature and salinity were calculated based on the annual-mean HydroBase climatology. Figure 19a shows a horizontal field of geostrophic current at 100 dbar relative to 300 dbar. Considerable baroclinic flow was widely recognized in the current field of the Kuroshio Extension and the Kuroshio Bifurcation Front in the annual-mean climatology. Therefore, the above example of synoptic vertical sections is not an unusual case that would be affected by short-term phenomena such as mesoscale eddies. Figure 19b shows the distribution of the salinity advection term at 100 dbar, simulated from geostrophic current speeds relative to 300 dbar (Fig. 19a) and the horizontal salinity distribution (Fig. 18b). Large salinity advection was recognized northward of 40°N, although current speeds were relatively small in this region. This means that the current field in the transition region generally crossed the strong T–S front, which consistently produced the subsurface vertical gradients of temperature and salinity in the wide area. The region of high salinity advection coincided approximately with the region of high Tu(z) (Fig. 13), indicating that the stratification of temperature and salinity by the advection was probably the main cause of the high Tu(z) in this wide area.
From these findings, the plausible modification mechanism of TRMW is summarized as follows. Baroclinic flow advects warm and saline water across density-compensating salinity and temperature fronts in the transition region. This flow would lead to upper stratification of temperature and salinity throughout the year; meanwhile, strong vertical mixing due to large heat loss tends to create vertical homogeneity in the upper surface layer in winter. In early spring, the net heat flux would become nearly zero and strong vertical mixing would cease, such that sudden stratification of temperature and salinity would occur in the subsurface pycnostads owing to advection. Then, double-diffusive instability would take place within TRMW and the active salt fingering would cause modification of TRMW core properties over a wide area of the transition region. The double-diffusive instability would gradually become weak as pycnostads get stabilized by the convection.
7. Conclusions and discussion
This study has investigated the formation region and spatial distribution of TRMW using Argo float data and revealed rapid modification and erosion of TRMW. Examination of the formation fields of water masses within and around the transition region identified density modes at 25.7–26.2 σθ for L-CMW and 26.2–26.6 σθ for D-CMW. In the denser mode, TRMW could be identified as a distinct water mass with colder and fresher properties (4°–9°C and 33.3–34.0) than those of D-CMW. TRMW forms in an extensive area over the western and central transition region, separated from the formation region of D-CMW to the south by density-compensating salinity and temperature fronts. The formation volume of TRMW is suggested to be twice as much as that of D-CMW.
Analysis of the water mass distributions within and around the transition region indicated that thick TRMW is distributed widely over the entire transition region. However, examination of the distribution of TRMW core properties along the density surface showed that the range of the core T–S properties is substantially reduced downstream within 10°–20° longitude from the formation region (or, equivalently, within the time span of a year) by gradually losing its fresh and cold side. It is also demonstrated that a major part of TRMW with 26.4–26.6 σθ is entrained into the mixed layer in the following winter. Quasi-Lagrangian observation by the isopycnal-following Argo float demonstrated that the double-diffusive salt-finger convection plausibly causes not only the rapid erosion of the TRMW pycnostads but also the increase of salinity and temperature in the TRMW cores. Analysis of time-dependent synoptic fields obtained by the Argo floats indicated that the similar modification is likely to take place in a large area over the western to central transition region. Further analysis of climatology suggested that the cause of strong salt fingering in the wide area is the geostrophic current with vertical shear crossing the density-compensating T–S front that brings warm and saline water of similar density to the upper part of TRMW.
Although there may be an interannual variation in T–S properties of pycnostads for the analyzed period spanning 2001–05, it would not affect the conclusion substantially. Concerning the distribution of T–S properties of TRMW cores, described above (Figs. 6, 7), it is difficult to attribute these systematic changes to the interannual variation. The same can be said for the seasonal evolution of the distribution of salinity and Tu(z) (Figs. 13, 14). Even if there are interannual variations in salinity and Tu(z), it is difficult to interpret the systematic seasonal changes as consequences of their interannual variations.
Concerning the quasi-Lagrangian observation by the isopycnal float (Figs. 11 and 12), the amount of possible deviation from the “true Lagrangian” observation from March 2003 to November 2003 is estimated as follows. The meridional and zonal distances between the first position detected after surfacing and the last position before leaving the surface are regarded as surface drift components. The drift components that the float would experience, if it stayed at the parking depth for the surfacing period, typically 8 h, are estimated by extrapolating the subsurface drift before surfacing. The differences between the surface drift components and the virtual subsurface drift component are regarded as the deviation of the quasi-Lagrangian observation from the true Lagrangian observation. The total meridional and zonal deviations are about 13 km to the south and about 21 km to the east, respectively. These deviations seem fairly small and would not affect our conclusions on the mechanism of the TRMW modification substantially.
As most of the TRMW of 26.4–26.6 σθ was entrained into the mixed layer during the following winter, TRMW formed in the western region near 160°E would critically affect the formation of TRMW in the downstream region and, consequently, the structure of the entire transition region. For example, the transition region is known to have weak stratification; the main pycnocline nearly disappears, a phenomenon referred to as the stability gap (Roden 1972; Yuan and Talley 1996). Since TRMW of 26.3–26.6 σθ is distributed downstream, overlying onto each density layer, the stratification of the entire density layer becomes weak across the transition region. This process likely plays an important role in the formation of the stability gap in the transition region.
As the D-CMW is distributed widely despite its limited formation region, especially in the density layer 26.5–26.6 σθ, it is suggested that at least some part of TRMW is modified into the warmer and more saline D-CMW. This implies that D-CMW and L-CMW are not only divided in the density histogram but are also fundamentally different from each other in terms of their formation mechanisms. T. Tokieda (2005, personal communication) noted that the two types of CMW have distinct chemical properties, with D-CMW containing some amount of subarctic water, which supports the result of this paper.
Because of the large volumes and subduction rates of mode waters, these waters have a large impact on the permanent pycnocline (Talley 1988; Suga et al. 2004, 2008). Suga et al. (2008) found that T–S properties differed between the mixed layer and the water distributed in the lower permanent pycnocline (their Fig. 8d). The modification process of TRMW described in this study can partly explain this difference in T–S properties. Previously, it was thought that water maintained its T–S properties acquired in the mixed layer through subduction into the permanent pycnocline (Iselin 1939) due to minor vertical diffusions. However, in the case of the transition region of the North Pacific, the vertical diffusion coefficients of salinity and temperature are large because of the favorable conditions for producing salt-finger convection in a wide area. Therefore, the water subducted into the permanent pycnocline tends not to conserve the water properties in the surface mixed layer.
The modification process of TRMW is also regarded as the reduction process of the low PV signature imposed on the lower permanent pycnocline. This may at least partly explain the difference between the low PV signatures in the high-resolution ocean general circulation models (Qu et al. 2002, Fig. 10; Tsujino and Yasuda 2004, Fig. 4) and in the observational climatology (e.g., Suga et al. 2004, Fig. 5) pointed out by Suga et al. (2008); it is much more intense in the former. Judging from their mixed layer depth and property distributions, the models do not reproduce the formation of TRMW effectively, which may result in long-lasting low PV signature in the lower permanent pycnocline.
The knowledge about the modification process of the lowest part of ventilated pycnocline in the North Pacific obtained in this study is important to understanding the physical phenomena and distribution of chemical properties of this density layer. The transition region is known to have large signals of decadal and multidecadal variability of sea surface temperature (Tanimoto et al. 1993; Hasegawa and Hanawa 2003). How such signals that propagate into the permanent pycnocline is an interesting question, and the modification process discussed here must be taken into account. Because the significant part of TRMW seemed to be entrained to the winter mixed layer, TRMW would play some role to integrate the interannual variability of T–S properties, which is given to surface mixed layer in this region through the forces from atmosphere and might be one factor of long-term variability of the sea surface temperature in this area. The results of this study should be also helpful in further analyses of CMW (Suga et al. 2003). In any such analysis, D-CMW would have to be distinguished from L-CMW because these waters presumably have different formation mechanisms. Effects of the transition region must also be considered to understand the temporal variability of D-CMW.
The modification process with strong diffusion may also affect ventilation in the lower part of the permanent pycnostad in the North Pacific. The oxygen saturation rate in the upper ocean is generally known to have vertical discontinuity between the ventilated layer (high oxygen saturation rates) and the unventilated layer (low oxygen saturation rates). However, recent hydrographic surveys around the transition region, such as the survey described by SSHW (their Fig. 2), have revealed a three-layer structure of the vertical discontinuity of the oxygen saturation rate: 1) an upper pycnocline with high oxygen saturation rates, 2) a lower pycnocline with 60%–80% oxygen saturation, and 3) an unventilated layer with oxygen saturation rates lower than 40%. This structure can now be understood from the findings of the current study. The middle layer is possibly ventilated by TRMW. TRMW is influenced by the water brought by geostrophic current with vertical shear, which is not originated in surface mixed layer and thus has a low oxygen saturation rate, through the strong double diffusion of subducted water and eventually has medium saturation rate of oxygen. A similar structure would be expected for other chemical substances that originate in the atmosphere such as CO2 or chlorofluorocarbons.
We are grateful for Prof. Dean Roemmich and Dr. Nicholus Schneider for their helpful discussions and kind encouragement. We wish to express our sincere thanks to the members of the Physical Oceanography Group at Tohoku University for their help. Dr. Eitarou Oka kindly provided the quality-controlled Argo float data. Comments from two anonymous reviewers and the editor, Prof. Lynne Talley, greatly helped us to improve the presentation. The first author (HS) was supported by the 21st Century COE program “Advanced Science and Technology Center for Dynamic Earth” (E-ASTEC) at Tohoku University. The second author (TS) was partially supported by funds from the Japanese Society for Promotion of Science [Grants-in-Aid for Scientific Research (B) 13440138, 16340135, and 21340133 and (C) 22540455]; from the Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan, for the “Western Pacific Air-Sea Interaction Study” (W-PASS; Grants 19030004 and 21014004); and from the Agriculture, Forestry and Fisheries Research Council (AFFRC), Japan, which sponsored the Studies on Prediction and Application of Fish Species Alteration (SUPRFISH) project.