1. Introduction

A shallow eastward current accompanied by a subsurface temperature/density front has been recognized in the central to southern latitudes of the North Pacific subtropical gyre as the subtropical countercurrent (STCC; Uda and Hasunuma 1969), where a general flow toward the west is expected according to the classical wind-driven circulation theory. The STCC runs from 122°E to at least 160°E nearly along the tropic of Cancer and shifts slightly to the north downstream. Another eastward current along 18°N was reported by Nitani (1972) who investigated statistically the distribution of zonal components of the surface geostrophic flow with respect to latitudes along 15 meridional sections across the western subtropical gyre. The two eastward currents, the STCC and the other to the south of it, were confirmed by Hasunuma and Yoshida (1978) based on both the synoptic and the long-term mean maps of the dynamic topography in the western North Pacific.

In spite of the early recognition, these two eastward currents have not been further clarified from the observational point of view. This is at least partly because of the large variability in the upper-flow field there. For example, satellite altimetry reveals a regional maximum of the sea surface height variability in the zonal band from 19° to 25°N, which extends from the western boundary region to the Hawaiian Islands (Qiu 1999; Kobashi and Kawamura 2001), indicating that energetic mesoscale eddies dominate this region. These eddies make it difficult to interpret an eastward flow in a given meridional section as one of the eastward currents reported earlier. Consequently, the overall distribution of the shallow currents in the central and southern part of the western subtropical gyre is still not understood satisfactorily.

On the other hand, as the STCC has been regarded as one of the permanent features of the upper flow in the subtropical gyre, quite a few authors have theorized about formation mechanisms of STCC and the accompanying front. Yoshida and Kidokoro (1967a,b) indicated that a trough in the wind stress curl found in the lower subtropical latitudes could cause the eastward current. Several authors relate the STCC and its accompanying front to Ekman convergence induced by the prevailing westeries and the trade winds (e.g., Roden 1975; Welander 1981; Cushman-Roisin 1981). However, neither of these mechanisms were essential for formation of the STCC in Takeuchi's (1984b) numerical experiment. He succeeded in reproducing the eastward current in a multilevel numerical model driven either by the wind stress curl without a trough or by a hypothetical meridional wind stress whose curl field is similar to that of the actual wind stress.

Alternatively, geostrophic current convergence has been considered as a possible cause of the countercurrent and the front by several authors (Cushman-Roisin 1984; Takeuchi 1984a; Dewar 1992; Kubokawa 1995, 1997). These theories show that the subtropical frontlike structure can occur through the geostrophic current convergence if some specific density structure is imposed on a boundary, that is, a surface boundary, an eastern boundary, or a western boundary. While the earlier theories by Cushman-Roisin (1984) and Takeuchi (1984a) depend on the rather artificial eastern boundary ventilation, a theory excluding the eastern boundary ventilation was firstly proposed by Dewar (1992). He used a two-layer planetary geostrophic ventilated thermocline model to discuss the frontogenesis caused by nonlinear Rossby waves and produced the subtropical countercurrent/front if the outcrop line slants northeastward. Kubokawa (1995) expanded Dewar's theory and showed that the stationary Rossby wave emanating from the western boundary region can also cause the subtropical frontlike structure using a two-layer model, although a somewhat unrealistic density structure is required for the western boundary to be a source of midocean fronts. Subsequently, however, Kubokawa (1997) formulated a two-level model that can represent the density and flow structure in the northwestern subtropical gyre more realistically. It was then shown that, if the northwestern region has a small north–south gradient of surface density and weak vertical shear, the stationary Rossby wave generated from this region possibly produces the subtropical countercurrent/front. Such structure in the northwestern region compares well with Takeuchi's (1984b) numerical experiment.

Meanwhile, Kubokawa and Inui (1999) discussed the generation mechanism of the STCC reproduced in an ocean general circulation model with simple geometry, driven by surface wind stress and surface buoyancy forcing. According to their analysis, the low potential vorticity (PV) fluid formed around the intersection of the outcrop line and the mixed layer front defined as the narrow transition zone of the mixed layer depth in the northern subtropical gyre. When advected southwestward, the low PV fluids on different isopycnals tend to converge in the horizontal plane and pile up vertically. As a result, a thick subsurface low PV pool is formed in the central subtropical gyre. This thick layer elevates the base of the surface layer, and a countercurrent occurs along the southern edge of this pool. Using an ideal fluid thermocline model, Kubokawa (1999) examined the influence of the mixed layer depth distribution on the structure of the ventilated thermocline and supported Kubokawa and Inui's scenario.

From the observational point of view, the low PV fluid north of the STCC has indeed been noticed. Uda and Hasunuma (1969) reported that the STCC flows along the southern edge of the North Pacific Subtropical Mode Water (STMW), which is the vertically homogeneous water mass in the northwestern subtropical gyre (Masuzawa 1969). Suga et al. (1989) further suggested the importance of STMW in maintenance of the STCC by analyzing repeat hydrographic data along the 137°E meridian. While these early observational views were based on rather limited data, the low PV water north of the countercurrent/front both in the ocean general circulation models by Takeuchi (1984b) and Kubokawa and Inui (1999) and in the analytical models by Kubokawa (1997) and Kubokawa (1999) resembles STMW to some extent.

It should be noted, however, that the model setup and the interpretation of the low PV are somewhat different among those models. Especially, Kubokawa (1997) modeled the low PV water as originating near the northwestern corner of the subtropical gyre and showed that its southward/southwestward intrusion maintained the STCC. On the other hand, Kubokawa and Inui (1999) found low PV water forming along the mixed layer front extending well into the Sverdrup interior. The STCC in their model was maintained by the stack of the low PV water with different densities from different locations along the mixed layer front. In order to examine whether the STCC in those models are related to that in the real ocean, we need a better description of the STCC, the accompanying front, and the associated low PV water based on observation. Moreover, although none of those models distinguish the two separate countercurrents/fronts, better observational description of these frontal features would help to relate more clearly some models to one front and others to the other.

The present paper reexamines the above observational notions in a more systematic way, using more comprehensive hydrographic data, to clarify the relationship between the countercurrent/front and the low PV water. We confirm the existence of two distinct subsurface temperature/density fronts accompanied by the respective low PV water to the north: the northern front corresponding to the STCC front of Uda and Hasunuma (1969) and the southern one corresponding to the front reported by Nitani (1972) mentioned above. These two fronts are referred to as the northern and southern subsurface subtropical fronts in the present study. Since our analysis is confined in the central and southern part of the subtropical gyre, we do not intend to test the entire scenario of STCC generation proposed by Kubokawa (1997) or Kubokawa and Inui (1999). Our purpose here is to examine whether the countercurrent/front is an inherent boundary in the ventilated thermocline suggested by those authors.

The remainder of this paper is organized as follows. The data and analytical methods used in the present study are explained in section 2. In section 3, the repeat hydrographic sections in the western part of the subtropical gyre are analyzed to identify the northern and southern subsurface subtropical fronts and extract the mean, or typical, thermocline structure associated with those fronts. In section 4, the two fronts are identified also in the recent high-resolution hydrographic sections over the North Pacific to demonstrate their basinwide characters. In section 5, we discuss implications of our observations of the subsurface subtropical fronts and their associated thermocline structure relative to the previously proposed theories about the subtropical countercurrents/fronts, followed by concluding remarks.

2. Data and processing

Three datasets are used in the present study. The first comes from several repeat hydrographic sections meridionally crossing the western North Pacific, maintained by the Japan Meteorological Agency (JMA) and the Hydrographic Department of Japan Coast Guard (HD/JCG) as listed in Table 1. The station spacing of these repeat sections is close to one degree in latitude. The second set consists of available data from the World Ocean Circulation Experiment (WOCE) Hydrographic Program (WHP) listed in Table 2. Finally, we prepared an isopycnally averaged climatology by using North Pacific HydroBase (Macdonald et al. 2001). While the original climatology by Macdonald et al. is based on their quality-controlled version of observed bottle data from World Ocean Atlas 1994 (Levitus 1994), and part of the WHP data, we updated the climatology by adding other WHP data and the CTD data from NOAA Pacific Marine Environmental Laboratory (PMEL) cruises (Johnson and McPhaden 1999).

The hydrographic data from the Japanese repeat sections are low-resolution profiles consisting of discrete samples. Each profile of temperature, salinity, or oxygen is interpolated onto a vertical grid with a 10 dbar interval using a shape-preserving local spline (Akima 1970) after visual quality check. The WHP data are high-resolution profiles. To reduce small-scale variability, individual profiles are smoothed with the 10 dbar half-width Hanning filter and subsampled at 10-dbar intervals. The resultant 10-dbar gridded profile data both from the repeat section and WHP are used to calculate several derived properties at 10-dbar intervals. The resultant profiles of the observed and derived properties are further interpolated linearly onto a series of isopycnal surfaces with a 0.05σ_θ interval.

The main tools for delineating the ventilated thermocline structure associated with the subsurface subtropical fronts in our analysis are PV and apparent oxygen utilization (AOU). A finite difference form of PV,

View Expanded

is calculated at each depth level using the fixed depth increment Δz = 60 m, where Δσ_θ is the variable potential density increment, f is the Coriolis parameter, and ρ is the in situ density. AOU is the difference in oxygen concentration of a water parcel from its saturation value. AOU has a tendency to increase due to the consumption of oxygen by organic process after isolation from the atmosphere.

Geostrophic velocities are estimated relative to 1000 dbar for each station pair of the repeat sections using the 10-dbar gridded profiles. The total eastward transports associated with the subsurface subtropical fronts are evaluated by integrating all the eastward speeds over the sectional area from the sea surface to 1000 dbar within a 5° latitudinal band centered at the front.

3. Mean structure of the thermocline associated with the subsurface subtropical fronts

Our purpose is to examine “typical” thermocline structure associated with the subsurface subtropical fronts. This is not, however, an easy task because mesoscale features and/or other spatiotemporal variability tend to hide such typical structure. In fact, it is often difficult even to identify the countercurrent/front relevant to the large-scale current structure. In order to extract the robust thermocline structure associated with the fronts, we construct mean sections along several meridians with the use of the repeat hydrographic observations (Table 1). For preserving the frontal structure in the mean fields, we calculate the means with respect to the frontal coordinate system where the front in a given individual section is regarded as the origin of the along-section coordinate for that particular section (e.g., Hanawa and Hoshino 1988). The actual procedure to construct the mean sections is described below, followed by presentation of the typical or robust thermocline structure associated with the fronts.

An example of the synoptic section of potential temperature θ along 155°E is shown in Fig. 1a. This example is the case where we can identify the two subsurface fronts unambiguously. The northern front is found at 24.5°N in the depth range of 50–300 m, and the southern one at 18.5°N in the depth range 100–250 m. When we averaged all nine available θ sections along 155°E with respect to the geographical coordinate rather than the frontal coordinate, the two fronts were still discernible but considerably broadened and smoothed (not shown). The section of the accompanying standard deviations of θ (Fig. 1b) captures the maxima near the two fronts (Fig. 1a), illustrating that the fronts change their positions from one section to another and thus the frontal coordinate is beneficial.

To apply the frontal coordinate, we have to identify the relevant fronts in individual sections. After some trials, we chose the criteria to detect the fronts as follows. The northern (southern) front is defined as the maximum in the magnitude of meridional gradients of 100–200 m depth averaged θ in the latitudinal range 20°–26°N (14°–20°N). An example is given in Fig. 2 to demonstrate how these criteria perform for the same θ section as shown in Fig. 1a. The meridional gradients are evaluated between each adjacent pair of stations. Note that temperature fronts are characterized by large negative gradients since the meridional axis is positive northward. The northern (southern) maximum magnitude is found at the same latitude as that of the northern (southern) front identified by visually inspecting the θ section as described above. Furthermore, since the meridional gradients of vertically averaged θ are virtually comparable to the vertical shear of zonal velocity, the maximum magnitude of negative gradients corresponds with the upper eastward velocity maximum (Fig. 2a). It should be also noted that similar criteria based on σ_θ give practically the same results as those using θ (Fig. 2c) because salinity contributes little to the upper ocean density field in the subtropics. Finally, we exclude from the calculation of the means a few sections that have only subtle gradient maxima with magnitudes smaller than 1°C/deg lat in order to ensure that the resulting means adequately represent the frontal structure. The mean properties of the fronts identified along each repeat section such as their latitudes, magnitudes of meridional gradients of the vertically averaged θ, and associated total eastward transports are summarized in Tables 3 and 4. Since the whole procedure described above is fairly mechanical, we may occasionally fail to detect the right front among other mesoscale features. However, we emphasize that even the analysis based on these simple criteria of “mere thermal fronts” reveal a robust thermocline structure associated with the fronts.

The mean sections with respect to the northern front along 155°E (Fig. 3) show the front as a sharp northward shoaling of the upper isopycnals above 250 m, which accounts for the maximum eastward geostrophic velocity, greater than 20 cm s^–1 relative to 1000 dbar. A relatively thick layer lies beneath the upper isopycnals to the north of the front. This vertical PV minimum layer is characterized by PV less than 2 × 10^–10 m^–1 s^–1 at 25.3–25.6 σ_θ centered at 25.4 σ_θ. This low PV water corresponds to STMW formed in the Kuroshio Extension region to the north of this site (Suga and Hanawa 1995). That is, the early notion of the correspondence between STCC and the southern edge of STMW (Uda and Hasunuma 1969; Suga et al. 1989) is confirmed in terms of the mean fields.

The AOU distribution in the mean section (Fig. 3d) further supports this view. The sharp downward increase of AOU below 26.3–26.5 σ_θ suggests that this is the base of the ventilated thermocline. The general southward increase of AOU along each isopycnal indicates that the more aged water is in the south, as expected from the ventilated thermocline theory (e.g., Luyten et al. 1983; Huang and Russell 1994). As for the STMW layer, the southward increase of AOU is not monotonic but considerably sharper just north of the front, indicating the southern edge of the newer STMW there. In addition, it is not surprising that AOU at the isopycnals above the STMW layer rapidly decreases upward because those isopycnals are close to the sea surface.

Similar to the northern front, the mean sections along 155°E with respect to the southern front (Fig. 4) show a sharp northward shoaling of the upper isopycnals above 250 m. The eastward velocity of the countercurrent reaches a similar value, greater than 20 cm s^–1, to that associated with the northern front. Thicker water lies beneath the shoaling isopycnals north of the front, compared with the water south of the front in the same density range. The PV section captures the remarkable character of the thermocline associated with the southern front. That is, the southern front coincides with the sharp front of the isopycnal PV at 25.3–26.3 σ_θ and the southern edge of the lower PV water. The lower PV water north of the front is above the sharp downward increase in AOU at 26.3 σ_θ that suggests the base of the ventilated thermocline. The noticeably sharp southward increase in AOU especially around 25.5–26.3 σ_θ suggests that the front marks the boundary between the differently ventilated regions. It should be noticed that the lowest portion, around 26.0–26.3 σ_θ, of the water north of the southern front corresponds to North Pacific Central Mode Water (CMW: Nakamura 1996; Suga et al. 1997) whose origin is near the northern edge of the subtropical gyre.

Although the correspondence between the southern front and the southern edge of the CMW is similar to that between the northern front and the southern edge of STMW, there is a marked difference between the two. The thick water responsible for the shoaling of the upper isopycnals north of the southern front consists of not only CMW but also the low PV fluids of a fairly wide σ_θ range, while the thick water north of the northern front consists solely of STMW within a narrow σ_θ range. The stack of low PV fluids in a rather wide σ_θ range is similar to that in the scenario proposed by Kubokawa and Inui (1999).

The mean sections with respect to the northern and southern fronts at 130°E, 137°E, and 144°E are constructed in the same manner as for the 155°E section. They reveal essentially the same thermocline structure associated with the fronts as found in the 155°E section. The northern front is characterized as the southern boundary of the thick and thus low PV water corresponding to STMW within a relatively narrow σ_θ range, while the southern front is characterized as the southern boundary of the low PV water spanning a wider σ_θ range. The σ_θ ranges of the low PV water north of the two fronts at each repeat section are summarized along with the other mean properties in Tables 3 and 4.

Since the periods over which the mean values were calculated are different from one section to another, it is difficult to deduce a meaningful mean zonal variation in frontal properties, so we discuss this matter only briefly. STMW north of the northern front appears to diffuse from east to west; its σ_θ range becomes wider and the minimum PV at its core increases to the west (not shown), as partly captured by its wider σ_θ range in the 137°E and 130°E sections (Table 3). The westward weakening of the low PV signature is consistent with Suga and Hanawa's (1995) observation that STMW dissipates considerably along its southwestward/westward advective path within a year from its formation. In contrast to the westward weakening of STMW, the eastward transport increases to the west, while the magnitude of the 100–200 m averaged temperature gradient at the front shows no systematic zonal trend. The larger transport in the west may be an artifact due to the higher eddy activity there (e.g., Aoki and Imawaki 1996). If there are more eddies in the west, we may inadequately regard the northern flank of anticyclonic eddy as the front more frequently there. Since anticyclonic eddies tend to have more intense “eastward currents” than the actual countercurrent, the mean transport can be overestimated in the west. Nevertheless the mean eastward transport range of 8 to 13 Sv (Sv ≡ 10⁶ m³ s^–1) corresponds moderately to the 8–18 Sv estimated by Uda and Hasunuma (1969). On the other hand, the mean properties of the southern front appear much the same for all four repeat sections (Table 4).

Finally, two other features are revealed by the frontal composites. The first is the difference in vertical structure of the two fronts. The eastward geostrophic velocity at the northern front at 155°E remains greater than 1 cm s^–1 down to 500 m and close to zero below (roughly captured in Fig. 3b) as implied by the isopycnals shoaling northward at 250–500 m and flattened below that (Fig. 3a). On the other hand, the isopycnals at the southern front are almost flattened at 400 m and deepened northward below that (Fig. 4a). As a result, the eastward flow associated with the front is confined above 200 m and the flow underneath is westward. The westward flow has a maximum speed of 4 cm s^–1 at 400 m (not shown). A similar difference in the vertical structure of the flow associated with the two fronts is observed for all four repeat sections (not shown). This feature may reflect a difference in the formation mechanisms of the two fronts.

The second point is that the southern (northern) frontal features appear on the composites of the northern (southern) front. For example, the PV and AOU gradients at −6° latitude in Figs. 3c and 3d are associated with the southern front, and the PV minimum north of 6° centered at 25.4–25.5 σ_θ (Fig. 4c) is associated with the northern front. A similar tendency is observed along the other repeat sections (not shown). The concurrent appearance of the two fronts on the composites indicates that the meridional excursions of the two fronts tend to preserve the relative distance between the two fronts. These two features are interesting but beyond the scope of the present study and thus left for future work.

In conclusion, based on the mean sections, we found two subsurface subtropical fronts in the western subtropical gyre. The northern and the southern countercurrents/fronts are distinctive in the sense that their thermocline structures differ. Both fronts are zonally connected features, within the study area from 130°E to 155°E.

4. Basinwide distribution of the subsurface subtropical fronts

The repeat hydrographic sections captured the northern and southern subsurface subtropical fronts and the accompanying thermocline structure for the western part of the subtropical gyre. Our purpose in the present section is to clarify the spatial extent of these two fronts over the North Pacific. More specifically, we aim to determine how far east we can identify these two fronts accompanied by the same thermocline structure specified in the preceding section. While it is often difficult to identify the subtropical fronts in a given synoptic section, as mentioned above, the typical thermocline features associated with the fronts provide criteria to identify those fronts to a certain degree. We examine a series of WHP meridional sections (Table 2). Every frontal feature in the thermocline in each θ section was checked to see if it was the southern boundary of low PV/AOU water. If it was, the feature was identified as the subsurface subtropical front. The locations of the northern and southern fronts identified in the WHP sections are shown in Fig. 5 along with their mean positions in the repeat sections. The σ_θ ranges of the low PV/AOU water north of the fronts, which identify those fronts in each section, are listed in Table 5.

The northern front, characterized by a thick layer of STMW to the north, is detected on P9 (137°E), P10 (149°E), P13C (165°E), and P14N (179°E) but not on the other sections farther east. The PV and AOU features in the thermocline similar to those depicted in the mean sections are found in the P9 and P10 sections, which are within the longitudinal range covered by the repeat lines (figures not shown). Similar features characterized by the thick STMW of low PV and low AOU also appear north of the front in the P13C section (Fig. 6a), which is located outside the area covered by the repeat sections. As for the P14N section (Fig. 6b), the signature of STMW in the PV and AOU fields is less marked but still apparent north of 25°N. We thus regard the front at the STMW's southern edge as the northern front. STMW is absent from the sections farther east (Figs. 6c,d), which is consistent with the previous description of STMW (e.g., Masuzawa 1969; Hanawa 1987; Talley 1988; Suga et al. 1997; Yasuda and Hanawa 1997). Therefore it is reasonable to conclude that the eastward extent of the northern front characterized as the southern limit of STMW is somewhere between the international date line and 165°W.

Kaneko et al. (1998) analyzed the P9 section and identified three eastward currents at 18°N, 22°N, and 24°N. Looking at temperature at 100–200 m, they regarded the eastward current at 22°N as the STCC defined by Uda and Hasunuma (1969). The 22°N front also has the deepest signature in the thermocline among the three, reaching to 1000 m. On the other hand, we chose the 24°N front as it corresponds with the southern limit of STMW. We do not think, however, that the difference in the interpretation is critical because there may be some eddies to hide the “typical” structure.

The southern front, characterized as the southern edge of the lower PV/AOU with a relatively wide range of σ_θ, is detected at P9 (137°E), P10 (149°E), P13C (165°E), P14N (179°E), and P15N (165°W) but not in the other sections farther east. As in the case of the northern front, the two sections (P9 and P10) within the longitudinal range covered by the repeat sections show the features of PV and AOU similar to those found in the mean sections associated with the southern front (figures not shown). The southern front is also clearly identified along 165°E (Fig. 6a) and 179°E (Fig. 6b) accompanied by low PV/AOU to the north. Especially the southern front at 179°E corresponds to a remarkable front in AOU, reflecting the southern front's role as the southern boundary of well-ventilated water in the thermocline.

The PV distribution associated with the southern front along 179°E is noticeably different from that along the other sections to the west. Unlike the other western sections, the 179°E does not show the broad meridional PV minimum extending from the southern front to the north on the isopycnals centered at 26.0–26.3 σ_θ. The PV at these isopycnals along 179°E monotonically decreases to the north, which means the low PV water corresponding to CMW apparently recedes to the north. As a result, the pile of the low PV waters of different σ_θ is not as vertically aligned as observed in the western section.

This tendency for the low PV waters of different σ_θ to stack unevenly becomes much more enhanced along 165°W (Fig. 6c). While the PV between 25.6 σ_θ and 26.3 σ_θ gradually decreases from 15°–17°N to the north, we identify the southern front at 24.5°N where the relatively sharp northward decrease of isopycnal PV and AOU at 25.6–26.3 σ_θ occurs. The sections farther to the east do not show the sharp meridional changes in the isopycnal PV that characterize the southern front (Fig. 6d). We thus conclude that the eastward extent of the southern front is somewhere between 165°W and 153°W.

The locations of the two fronts identified above are plotted on the climatological maps of isopycnal PV on 25.4 σ_θ for the northern front and on 26.0 σ_θ for the southern front (Fig. 7). The former isopycnal corresponds with the core of STMW. The northern front is consistently located along the southern boundary of the low PV region. The latter isopycnal corresponds to the central isopycnal of the rather wide σ_θ range of the low PV/AOU water and also to the upper portion of CMW. The southern front is also consistently located along the southern edge of the low PV region. The correspondence between the fronts and the two dimensional extent of low PV region generally supports the eastern limits of the two fronts inferred above.

5. Discussion and concluding remarks

The two previously documented subsurface temperature/density fronts and their accompanying eastward currents in the central and southern North Pacific subtropical gyre are identified using repeat hydrographic sections in the western part of the gyre and the WHP sections over the whole gyre. The northern countercurrent corresponds to the STCC reported by Uda and Hasunuma (1969). The countercurrent/front is located typically near 24°N and extends from the western boundary to the international date line. The early suggestion that the STCC is located at the southern edge of the STMW is confirmed. The front appears as the southern boundary of the lower PV/AOU on the isopycnals within the STMW layer. The southern countercurrent corresponds to the eastward current reported by Nitani (1972). The countercurrent/front is located near 18°N west of the date line, but shifts northward east of the date line. Its eastern limit is around 165°W near 25°N. The southern front corresponds to the southern boundary of the lower PV/AOU waters of a wide σ_θ range. While CMW contributes significantly to this low PV/AOU layer, the contribution from other waters is still significant.

We now discuss what these observations imply for previously proposed theories about the formation mechanism of these fronts. Although the theories and numerical simulations focusing on the STCC have been aimed at the single countercurrent/front noted by Uda and Hasunuma (1969), the present study captures two distinct countercurrents/fronts with associated features in the thermocline. This does not mean, however, that the previous theories are unrelated to the fronts in the real ocean.

The southern countercurrent/front, corresponding to the eastward current noted by Nitani (1972), is located along the southern edge of low PV/AOU water vertically spanning a range of σ_θ, as in Kubokawa and Inui's (1999) numerical simulation and Kubokawa's (1999) analytical model. In their models, the low PV fluids of different densities subduct from the northern zone of the subtropical gyre; those on lighter isopycnals originate from the western region and those on denser isopycnals from the eastern region. Advected along the anticyclonic gyre, the low PV fluid on each isopycnal overlaps that on another isopycnal and makes a thick low PV pool in the western central subtropical gyre. The low PV fluids push up the isopycnals above, causing the eastward current at the southern edge of the pool. We also observe the low PV pile is less vertically coherent in the WHP sections upstream of the west/southwestward flow, which is also favorable for their scenario. However, since we examine only the southern and central part of the subtropical gyre, we cannot tell whether the low PV fluids are formed at the intersection of the mixed layer front and the outcrop line.

The northern countercurrent/front is located along the southern edge of the thick STMW layer. STMW has much more pronounced vertical homogeneity compared with the low PV water associated with the southern front. Its spatial distribution is more confined in the northwestern part of the gyre compared with the southern counterpart. These features match the countercurrent/front in Kubokawa's (1997) two-level model. Kubokawa modeled STMW as the dense and vertically and horizontally homogeneous fluid in the northwestern corner of the subtropical gyre, from which a stationary Rossby wave emanates to produce the countercurrent. This homogeneous fluid resembles the observed STMW much better than the low PV fluids with wide σ_θ range found in Kubokawa and Inui's (1999) simulation. Our observations thus suggest that the formation mechanism proposed by Kubokawa (1997) is more likely for the northern front.

The observed features of the two fronts and the related theories of the frontogenesis suggest that both fronts are lateral boundaries in the ventilated thermocline. It is thus implied that variability of these fronts can be related to variations in the thermocline circulation, the subduction or water mass formation processes associated with the wintertime air–sea interaction, etc. It is also implied that numerical models of the upper circulation and water properties should reproduce these fronts in order to simulate the real ocean better. As reviewed in Hanawa and Talley (2001), STMW exists in all the subtropical gyres of the World Ocean. Therefore, it is expected that all the subtropical gyres may have an STCC. Whether it is true or not is another good subject to be addressed in future study.

The present results are mostly based on meridional sections. Further clarification of the formation mechanisms of the fronts may be possible through analysis of the three-dimensional PV/AOU structure of the ventilated thermocline. Such analysis with the full use of the North Pacific HydroBase is now under way.