Decadal variability of the Subtropical Front of the western North Pacific in an eddy-resolving ocean general circulation model

Authors


Abstract

[1] We examined decadal variability of the Subtropical Front (STF) of the western North Pacific by using a North Pacific ocean general circulation model (OGCM), comparing the results of three simulations with different horizontal resolutions. In the long-term mean fields, the eddy-resolving model (10 km) was able to simulate the distributions of the STF and the associated Subtropical Countercurrent (STCC) between 20°N and 30°N better than either the non-eddy-resolving model (100 km) or the eddy-permitting model (20 km) because it simulated the Kuroshio recirculation gyre more realistically. The simulated STF intensity exhibited significant decadal-scale variations: it was stronger in the late 1970s and weaker in the early 1990s. During these two periods, the simulated mode waters showed corresponding differences in their potential vorticity minima density and paths. We also investigated the relationship between the decadal-scale STF variability and atmospheric forcing. The results suggest that the decadal-scale STF variability can be largely explained by changes in the mode waters formed in the western North Pacific and advected to north of the STF by the subtropical gyre in response to a change in surface westerly winds that occurred in the mid-1970s.

1. Introduction

[2] The North Pacific sea surface temperature (SST) fluctuates with a period of more than 10 years, as well as with a period of 2 to 6 years in relation to El Nino-Southern Oscillation events [Nitta and Yamada, 1989]. Observational studies have shown that SST decadal-scale variability in the North Pacific is associated with two major oceanic fronts, namely the Subarctic Front (SAF) and the Subtropical Front (STF) [Nakamura et al., 1997] (Figure 1). Near the SAF, the SST variability is negatively correlated with that in the Alaskan Gyre, whereas near the STF, the SST variability exhibits a strong negative correlation with tropical SST variability [Nakamura et al., 1997]. Although there have been many studies on SST variability in the SAF region, especially in relation to the decadal variability in the mid-latitude atmosphere–ocean system [e.g., Latif and Barnett, 1994], few studies have focused on decadal-scale SST variability in the STF region, probably because of the scarcity of observed data.

Figure 1.

Schematic illustration of the near-surface current, front, and water-mass structures in the western North Pacific. KE: Kuroshio Extension, SAF: Subarctic Front, KBF: Kuroshio Bifurcation Front, STF: Subtropical Front, STMW: Subtropical Mode Water, CMW: Central Mode Water.

[3] The STF is defined as a temperature/density front in central to southern latitudes of the North Pacific subtropical gyre (Figure 1). In the western North Pacific, the STF is accompanied by a shallow eastward current, referred to as the Subtropical Countercurrent (STCC) [Yoshida and Kidokoro, 1967], suggesting a thermal wind relation between the STF and the STCC [Takeuchi, 1984]. The STF is also found in the eastern North Pacific, east of Hawaii, where it is identified as a temperature/salinity front during winter and a salinity front during summer [Roden, 1974, 1975, 1980; Van Woert, 1982; Niiler and Reynolds, 1984; Lynn, 1986; Kazmin and Rienecker, 1996; Dinniman and Rienecker, 1999]. However the dynamics of the eastern STF could be very different from the STF in the western North Pacific, given that the STF forms the northern boundary of the eastern Pacific gyre [Sverdrup et al., 1942; Munk, 1950; Kenyon, 1975].

[4] In the western North Pacific, the STF is located around 20°N–25°N between the Subtropical Mode Water (STMW) [Masuzawa, 1969] on the north and the Tropical Water on the south [Uda and Hasunuma, 1969]. In long-term mean fields, it has been pointed out that mode waters characterized as a pycnostad, such as the STMW and the Central Mode Water (CMW) [Nakamura, 1996], play a great role in the formation and maintenance of the STF by theoretical [Kubokawa, 1997, 1999] and simulation [Takeuchi, 1984; Kubokawa and Inui, 1999] studies. Recent observational studies also confirmed the relationship between the STF and the mode waters [Aoki et al., 2002; Kobashi et al., 2006]. These mode waters are formed in the northwestern subtropical gyre during winter and are spread to the north of the STF by subtropical gyre circulation [Suga et al., 2004]. These findings indicate that the ocean subsurface variability associated with water mass formation and ocean circulation may influence STF temporal and spatial variability.

[5] In this study, we examined decadal-scale STF variability of the western North Pacific and investigated the mechanism for the variation. Both observational [Yasuda and Hanawa, 1997; Qiu and Chen, 2006] and simulation studies [Xie et al., 2000; Hosoda et al., 2004] have shown that the properties of mode waters change on a decadal timescale, suggesting that corresponding changes in STF intensity might also occur on a decadal timescale. Thus we focused on the relationship between the STF and mode waters on a decadal timescale. In addition, we discuss the effect of atmospheric forcings, which may induce STF variability via changes in surface momentum and heat fluxes.

[6] Our analysis is based on hindcast simulations by a North Pacific eddy-resolving ocean general circulation model (OGCM) driven by atmospheric reanalysis data. The eddy-resolving OGCM can be expected to elucidate oceanic processes that non-eddy-resolving OGCMs cannot resolve. Past OGCM studies on decadal-scale variability in the North Pacific have used non-eddy-resolving models [Xie et al., 2000; Hosoda et al., 2004] or, if an eddy-resolving model was used, have focused on variability in the SAF region [Nonaka et al., 2006]. Hence this is the first study to use an eddy-resolving model to investigate decadal-scale variability in the North Pacific STF region.

[7] In section 2, we explain the model and experimental procedure. In section 3, we describe the features of the simulated STF and its decadal variability. In section 4, we investigate the thermocline structure associated with the STF, and examine the relationship between the STF and mode waters in the model on a decadal timescale. In section 5, we discuss the effect of atmospheric forcing on STF variability. Section 6 is a summary.

2. Model and Experiments

[8] For this study, we used the Meteorological Research Institute Community Ocean Model (MRI.COM) [Ishikawa et al., 2005]. The model includes a domain of the North Pacific between 15°S to 65°N and 100°E to 75°W. Its horizontal resolution is 1/12° (approximately 10 km) in longitude and 1/12° in latitude and its vertical resolution is 62 levels (27 levels in the top 500 m).

[9] This model solves the primitive equation system in spherical coordinates under the Boussinesq and hydrostatic approximations. The model has a free surface and a σ − z hybrid coordinate in the vertical direction. The generalized Arakawa scheme [Ishizaki and Motoi, 1999] is used for momentum advection to conserve quasi-enstrophy. We adopted Smagorinsky-like biharmonic horizontal mixing of momentum [Griffies and Hallberg, 2000]. For horizontal and vertical advection of tracers, the Uniformly Third-Order Polynomial Interpolation Algorithm (UTOPIA) [Leonard et al., 1994] and the Quadratic Upstream Interpolation for Convective Kinematics with Estimated Streaming Terms (QUICKEST) [Leonard, 1979] schemes, respectively, were adopted. The turbulent closure scheme of Mellor–Yamada level 2.5 was used for vertical mixing of momentum and tracers [Mellor and Yamada, 1982]. Vertical mixing parameterization based on St. Laurent et al. [2002] was also included to represent the effect of tidal mixing.

[10] In the model, surface heat fluxes are calculated with the bulk formula reported by Kara et al. [2000] from atmospheric variables based on the National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis data set [Kalnay et al., 1996]. The freshwater flux is evaluated from daily precipitation rates taken from the reanalysis data, under the constraint that sea surface salinity is restored to its monthly climatology on a timescale of 3 days. The climatology is based on World Ocean Atlas 1998 (WOA98) data [Antonov et al., 1998]. At the model's artificial boundaries at 15°S and 65°N, temperature and salinity are restored to their local monthly climatology (WOA98) with a timescale of 20 days at the boundaries increasing to infinity in the interior region.

[11] The model was integrated using NCEP-Department of Energy (DOE) Atmospheric Model Intercomparison Project (AMIP-II) [Kanamitsu et al., 2002] climatological (1979–2003) forcing for 50 years from the initial condition, which was determined by a coarser resolution model (approximately 20 km) driven by the same forcing for 20 years. The model was then driven by daily wind stresses and atmospheric state data based on the NCEP/NCAR reanalysis during the period from 1949 to 2005. This simulation is referred to as NPERM5 hereafter. Basic settings of the NPERM5 simulation are summarized in Table 1.

Table 1. Features of NPERM5
DomainPacific Ocean between 15°S to 65°N and 100°E to 75°W. Temperature and salinity are restored to World Ocean Atlas 1998 at the southern wall, and other side boundaries are closed.
Resolution1/12° (zonal) × 1/12° (meridional), 62 levels (vertical)
Surface forcing1949–2005 atmospheric state from NCEP/NCAR reanalysis; bulk formula from Kara et al. [2000]
Momentum advectionGeneralized Arakawa scheme [Ishizaki and Motoi, 1999]
Horizontal viscosityBi-harmonic Smagorinsky-like viscosity [Griffies and Hallberg, 2000]
Horizontal diffusionBi-harmonic
Initial stateIntegration for 70 years from WOA98

[12] To examine the effect of horizontal resolution on the simulated results, two additional simulations with different horizontal resolutions were performed using the same boundary conditions (Table 2). The NPERM15 simulation used a horizontal resolution of 1/4° in longitude and 1/6° in latitude (approximately 20 km), and the NPELM60 used a horizontal resolution of 1° in longitude and 1° in latitude (approximately 100 km).

Table 2. Experiments
Experiment NameResolution
NPERM51/12° (zonal) × 1/12° (meridional), 62 levels
NPERM151/4° (zonal) × 1/6° (meridional), 62 levels
NPELM601° (zonal) × 1° (meridional), 62 levels

[13] A finite difference form of potential vorticity (PV), as defined by

equation image

was calculated from densities between adjacent standard depths, and then was interpolated to standard potential density (σθ) levels by 0.05 kg m−3, where f is the Coriolis parameter, ρ the density, Δσθ the potential density difference, and Δz the depth interval. The subsurface PV minimum is a useful tracer for STMW and CMW because the deep convection that forms those mode waters is a source of low PV and water conserves PV along its advection path if source and dissipation are negligible [e.g., Tally and McCartney, 1982]. In the following analysis, we use the monthly mean model outputs, with the resolution reduced to 1° × 1°, for our analytical convenience.

3. Simulated Subtropical Front

3.1. Features of the STF

[14] Figure 2 shows meridional sections of water temperature during June 2004, based on hydrographic observations by the Japan Meteorological Agency and the NPERM5 simulation. In the observed data (Figure 2a), the STF is found between 23°N and 25°N at 50 to 200 m depth, corresponding to a maximum of meridional temperature gradient. In the summertime, the STF is hardly seen at the surface, because the development of the seasonal thermocline hinders a clear relationship between the surface and the subsurface. The Kuroshio Extension front is found at around 34°N, corresponding to a subsurface maximum of meridional temperature gradient between 300 m and 500 m. In the area north of the STF and south of the Kuroshio Extension front, there is a relatively thick water layer with a temperature of 16 to 18°C, that is, the STMW. Although the NPERM5 simulation displays a slightly warmer bias (for example, the simulated STMW is warmer than the observed), it reasonably captures the subsurface temperature distributions associated with the STF and the latitudinal positions and the vertical distributions of the two fronts (Figure 2b).

Figure 2.

Latitude-depth sections of temperature at 165°E during June 2004, based on the hydrographic observation by the Japan Meteorological Agency (a) and the NPERM5 simulation (b). Contour intervals for temperature fields are 1°C; meridional temperature gradient [°C (100 km)−1] is shown with shading to indicate the subtropical frontal zone and the Kuroshio Extension front.

[15] Figure 3a shows the NPERM5 simulated sea surface height distribution and zonal geostrophic currents at the surface relative to 400 dbar in the long-term mean field. Apart from the Kuroshio Extension area, two eastward flows are seen in the region of the subtropical gyre as follows. The eastward flow between 23°N and 28°N from 140°E to 160°W corresponds to the simulated STCC. The STCC is associated with the STF, which is recognized as a large meridional temperature gradient at 100 m depth (Figure 3b). The location of the STF coincides with a large meridional density gradient in the western North Pacific, where a thermal wind relation between the STF and the STCC plays an important role in the maintenance of the STF [Takeuchi, 1984]. Although the STF also corresponds to the salinity front (not shown), the meridional salinity gradient acts to weaken the meridional density gradient, implying that salinity plays a minor role in the maintenance of the STF at least in the western North Pacific. Comparison with observation data [Kobashi et al., 2006] showed that the model was able to represent the long-term mean distribution of the STF, except that the simulated STF is not clear west of 140°E.

Figure 3.

(a) Results of NPERM5 simulation of surface zonal geostrophic velocity [10−2 ms−1] relative to 400 dbar (colors) and sea surface height [cm] (contours). (b) Same as (a) but for meridional temperature gradient [°C(100 km)−1] (contours) and meridional potential density gradient [10−6 kg m−4] (colors) at 100 m depth. The contour intervals are every 0.1 °C(100 km)−1 for the meridional temperature gradient greater than 0.5 °C(100 km)−1 and less than 1.0 °C(100 km)−1. Closed triangles denote the location of the model STF. Results of the simulations of surface zonal geostrophic velocity relative to 400 dbar (colors) and sea surface height (contours) by (c) NPERM15 and (d) NPELM60.

[16] The eastward flow near 18°N–20°N west of 160°W corresponds to the Hawaiian Lee Countercurrent (HLCC) [Xie et al., 2001]. Compared with observation data [Yu et al., 2003], the simulated HLCC is farther west of the International Date Line, and its magnitude is too strong. Formation of the HLCC is considered to be related to wind stress curl near the Hawaiian Islands [Xie et al., 2001]. The NCEP/NCAR reanalyzed wind fields that we employed as boundary conditions seem to show a wind wake behind the Hawaiian Islands that is somewhat stronger than that in other atmospheric reanalyses, such as the ECMWF 40-year reanalysis data set [Uppala et al., 2005] and the Japanese 25-year reanalysis data set [Onogi et al., 2007]. Also, the small viscosity used in the eddy-resolving model was apparently not sufficient to damp the westward extension of the HLCC. Both of these factors may have contributed to the exaggerated HLCC in the model.

[17] To understand the effect of horizontal resolution on the simulation results, the results of the eddy-resolving model were compared with those of the non-eddy-resolving models (NPERM15 and NPELM60) driven by similar surface boundary conditions (Figures 3c and 3d). Among the three models, the Kuroshio recirculation gyre, recognized as the westward tilt of the surface height contours, was better simulated by the eddy-resolving model (NPERM5). Because the separation latitude of the Kuroshio and the strength of the recirculation gyre strongly depend on the magnitude of vorticity advection [Nakano et al., 2008], this improvement may result from the increased nonlinear effect in the eddy-resolving model. The better simulation of the Kuroshio recirculation gyre suggests that the distributions of the mode waters may also be simulated better by the NPERM5 model.

[18] The simulated HLCC is still clear in the NPERM15 simulation, whereas its spatial extent is substantially reduced in the NPELM60 simulation. The larger viscosity of the eddy-less model (NPELM60) compared with the eddy-permitting (NPERM15) and the eddy-resolving (NPERM5) models may have led to the damping of the westward extension of the HLCC. The coarser horizontal resolution of NPELM60, which does not resolve well the wind wake behind the Hawaiian Islands, may also have contributed to this difference.

[19] These results indicate that the NPERM5 model could successfully reconstruct the oceanic structure related to the STF in the long-term mean field. As for the seasonal variation of the observed STF in the western North Pacific, White et al. [1978] reported that the STF is strongest in spring and that the STF is located at a higher latitude from winter to spring in the region close to the western boundary. The seasonal variation of the STF simulated by the NPERM5 model was in good agreement with observations (not shown). In the following analysis, winter (January–March) mean data are used.

3.2. Decadal STF Variability

[20] Figure 4a shows 5-year running mean time series of the simulated STF intensity in the central North Pacific (160°E–180°E) at the surface and the subsurface. In this study, the STF intensity is defined as the meridional temperature gradient maximum in the area north of the HLCC and south of the Kuroshio Extension between 140°E and 160°W. The simulated STF intensity at all depths shows a marked decadal-scale variation; in particular, it increased in the late 1970s, but decreased in the early 1990s. A similar tendency was also found in the observed SST (COBE-SST) [Ishii et al., 2005] gradient, although differences in the observed STF intensity at the surface were relatively small because of its lower horizontal resolution arising from data averaging.

Figure 4.

(a) Time series of the simulated STF intensity, as indicated by the meridional temperature gradient [°C (100 km)−1] at the surface and at various depths in the central North Pacific (averaged over 160°E–180°E) and the observed intensity at the surface (red) during winter (January–March). The strong and weak STF periods are shaded in blue and pink, respectively. A 5-year running mean is applied to all time series. (b) Wintertime (January–March) mean water temperature at 100 m depth in the North Pacific during 1975–1979, based on NPERM5 (contours). Color shading indicates the difference between that mean field and the five-winter mean during 1990–1994. Closed triangles denote the location of the STF.

[21] Note that the STF variability in the subsurface does not necessarily correspond to that at the surface, because the temporal subsurface variability can emerge more clearly at deeper levels. For example, the STF magnitude at 150 m shows marked fluctuation with a period of 10 years from the mid-1980s, but similar fluctuations are not clear at the surface. Nevertheless, the enhancement of the STF in the late 1970s and its weakening in the early 1990s are robust at both the surface and subsurface.

[22] Figure 4b shows the temperature difference at 100 m depth between the five-winter means during 1975–1979 and those during 1990–1994. These two periods correspond to well-defined STF-strong and STF-weak phases, respectively, associated with the decadal-scale variability in STF intensity (Figure 4a). In the late 1970s, consistent with the enhancement of the STF at that time, positive temperature anomalies are simulated southeast of the STF, and negative ones northwest of the STF. The STMW distributes in the Kuroshio recirculation gyre region, which is to the northwest of the STF, suggesting a relationship between the mode water distribution and the STF. Next, we focus on variability in the oceanic interior related to the STF variability.

4. Mode Waters and Their Relation to STF

4.1. Relationship Between the STF and Mode Waters

[23] In the western North Pacific, the STF appears as the southern edge of vertically thick layers of low-PV water (Figure 5). The structure of the STF in the subsurface is related to the thermocline structure below the STF, and thus the analysis using PV is convenient for the understanding of the STF formation and maintenance. First, we show how the PV distribution is related to the STF. The following analysis procedure is based on Kobashi et al. [2006].

Figure 5.

Schematic meridional density section across the STF. ρb, reference density, corresponding to the lower pycnocline; ρ, the density beneath the STF; Z0 and Z, the depths of the isopycnal surfaces, ρb and ρ, respectively.

[24] Assuming that relative vorticity is negligible, PV (q) is given, using a vertical coordinate of density ρ, by

equation image

where Z is the depth of isopycnal surface (measured positive under the sea surface), f is planetary vorticity, and ρ0 is the reference density. Solving for Z and taking the meridional derivative yields,

equation image

where Z0 is the depth of the reference isopycnal surface ρb (≥ρ) and the subscript ρ denotes the partial derivative taken at a constant ρ surface. The left-hand side of equation (3) is the meridional slope of the isopycnal surface, and the first term on the right-hand side is the vertical integration of the deviation of the meridional PV gradient from the ambient vorticity gradient β. Thus equation (3) represents the relationship between the meridional slope of an isopycnal surface and the PV gradient below that surface.

[25] In the southern part of the western North Pacific subtropical gyre, where the lower isopycnal surfaces deepen northward (Figure 5), the meridional slope (∂Z0(ρb)/∂y)ρ is positive. In order for (∂Z(ρ)/∂y)ρ to be negative, which is necessary for STF formation and maintenance, the PV gradient deviation ((f/q)(∂q/∂y)ρβ) must be sufficiently negative between the upper and lower pycnoclines. This can be satisfied by a large negative PV gradient at the southern edge of the low-PV water mass north of the shoaling upper pycnocline, and suggests a relationship between the STF and mode waters north of the STF.

[26] Figure 6 shows the meridional distributions of the PV gradient deviation in the simulated long-term mean field. The large meridional density gradient between 25°N and 30°N around 24.0 σθ corresponds to the STF region. Below the STF, the PV gradient deviation is characterized by large negative values, suggesting that the negative PV gradients at the STF are closely associated with the STMW and CMW. This situation is quite different from that at the HLCC front around 20°N, because no mode water is found north of the front. These results indicate that the model captures well the relationship between the STF and mode waters recognized in the observed long-term mean field [Kobashi et al., 2006].

Figure 6.

Meridional sections of the northward density gradient (a) and potential vorticity (PV) (b) averaged over 160°E–180°E, contoured every 0.025 × 10−5 kg m−4, and every 0.5 × 10−10 m−1 s−1, respectively. The PV gradient deviation ((f/q)(∂q/∂y)ρ=constβ) is shown by color shading in both panels. Areas where the PV gradient deviation is positive are orange, and those where it is less than −2 × 10−11 m−1 s−1 and less than −4 × 10−11 m−1 s−1 are light and dark blue, respectively.

4.2. Features of the Mode Waters

[27] Before examining the features of the simulated mode waters in relation to the variation in the STF, we examined the general features of the simulated mode waters. In this study, we defined a mode water as a water mass with a low PV, that is, one that has vertically uniform properties.

[28] First, to identify the mode waters, we examined the low-PV water volume distribution in the density ranges of 24.5 to 26.5 σθ between 10°N and 40°N in the North Pacific (Figure 7a). The highest volume of low-PV water was in the PV range of 2 to 2.5 × 10−10 m−1 s−1. Therefore, we defined waters with PV < 3 × 10−10 m−1 s−1 as mode waters in this study.

Figure 7.

Frequency distribution of PV (top) and the relationship between the number of grid points with vertical PV minima of less than 3 × 10−10 m−1 s−1 and potential density (bottom).

[29] Next, the density distribution of the low-PV water was investigated to identify the mode water cores. We define a mode water core as water with a PV value of less than 3 × 10−10 m−1 s−1 and a vertical PV minimum in the density range of 24.5–26.5 σθ. We chose this density range because it is the surface density in the STMW formation area during winter and the base of the CMW. Figure 7b shows the mode water cores. Two clear peaks are apparent. The peak at 25.0 to 25.2 σθ corresponds to the simulated STMW core, and the other peak at 25.6 to 25.8 σθ corresponds to the simulated CMW core. Because a gap appears around 25.5 σθ, we distinguished the simulated STMW from the simulated CMW at this density. We defined the simulated STMW and CMW in the density ranges of 24.9 to 25.5 σθ and 25.6 to 26.2 σθ, respectively. The observed STMW and CMW are in the density ranges of 24.8 to 25.7 σθ and 25.8 to 26.5 σθ [Kobashi et al., 2006]; thus the simulated mode waters have lighter densities by 0.2 to 0.3 σθ than the observed mode waters.

[30] Figure 8a shows the long-term mean surface density and mixed layer depth (MLD) during winter simulated by NPERM5. Consistent with observations, the simulated MLD increases from the Kuroshio to the Kuroshio Extension in the northwestern subtropical gyre. Both the STMW and CMW form in the western North Pacific east of Japan, where the winter mixed layer is deep because of intense surface cooling acting on warm advection by the Kuroshio Extension. Moreover, an isopycnal PV minimum forms where the outcrop line intersects the MLD front [Kubokawa, 1999] between the deep mixed layer in the Kuroshio Extension and the shallow one in the rest of the subtropical gyre.

Figure 8.

(a) Simulated distributions of wintertime surface density (contours) and MLD (colors). The bottom of the mixed layer is defined as the depth where σt first exceeds its surface value by 0.05 kg m−3. (b) Simulated distribution of vertical PV minima in the density range 24.9–25.5 σθ (STMW layer). (c) Same as (b) but for the density range 25.6–26.2 σθ (CMW). In (b) and (c), a thick curve denotes the position of the simulated Kuroshio Extension front represented by the 13°C isotherm at 400 m depth. Closed triangles denote the location of the STF.

[31] Figures 8b and 8c show the horizontal distributions of the STMW and CMW layers, which are defined as the vertical PV minima in the density ranges of 24.9 to 25.5 σθ, and 25.6 to 26.2 σθ, respectively. The STMW and CMW have different geographical distributions and are separated by the Kuroshio Extension front. It is noted that the STF is located along the PV fronts of the STMW and CMW, with large PV gradients. These simulated geographical distributions of the STF and the mode waters agree well with the observed distributions [Suga et al., 1997; Kobashi et al., 2006].

4.3. Decadal Variability of Mode Waters

[32] We also investigated changes in the distribution of the mode waters between the two periods 1975–1979 and 1990–1994 (see Figure 4a). Figure 9 shows the meridional distribution of PV in relation to density during these two periods. The intrusion of low-PV waters was more significant in the density range of 25.5 to 26.2 σθ in the 1970s. In the 1990s, the intrusion of low-PV waters was in the density range of 25.4 to 25.7 σθ and the PV value increased by 0.5 × 10−10 m−1 s−1. In the difference between the two periods (Figure 9c), the meridional gradient of low-PV waters below the STF was negative, indicating that the enhancement of the STF in the 1970s was closely related to the low-PV water below the STF.

Figure 9.

Meridional sections of five-winter (January–March) mean PV averaged in 160°E–180°E for 1975–1979 (a), for 1990–1994 (b), and for the changes in five-winter mean PV from 1975–1979 to 1990–1994 (c).

[33] Figure 10 contrasts isopycnal PV distributions at 25.4 σθ and 26.0 σθ, corresponding to the STMW and CMW densities, respectively, between the two periods. In the 1970s, the STMW and CMW were formed in 160°E–180°E and 170°E–170°W, and were distributed west of 170°E and west of 170°W, respectively. In the 1990s however their formation areas and their paths shifted eastward. A similar path changes of mode water in the 1980s is simulated in an OGCM forced by observed wind stress [Xie et al., 2000]. It is also noted that PV value of the low-PV waters increased at both densities during the two periods. In particular, the increase in their formation regions implies that the deep convection that forms those mode waters was relatively weak in the 1990s. In fact, the temperature of the mode water core increased between the two periods. Figure 11 shows the distributions of the mode water core densities between the two periods. It is found that the mode water core density decreased from 1975–1979 to 1990–1994. The increased paths associated with the eastward shift in the 1990s can also cause changes in mode water properties through diabatic processes and dissipation.

Figure 10.

Map of five-winter (January–March) mean PV on the 25.4 σθ (a), and 26.0 σθ (b) isopycnal surfaces for 1975–1979 (color), for 1990–1994 (contour). Closed triangles denote the location of the STF.

Figure 11.

The relationship between the number of grid points with vertical PV minima <3 × 10−10 m−1 s−1 and potential density for 1975–1979 (top) and for 1990–1994 (bottom).

[34] Figure 12 shows meridional sections of the northward density gradient averaged over 160°E–180°E, along with the PV gradient, for 1975–1979 and for 1990–1994. The latitudinal position of the STF, which is associated with a relatively large PV gradient below the STF, shifted southward from 27°N in 1975–1979 to 25°N in 1990–1994. As shown in Figure 9, the large PV gradient below the STF is associated with the intrusion of low-PV waters from the north. Corresponding to the enhancement of the low-PV waters, the large PV gradient below the STF was found at 25 to 26 σθ during 1975–1979, while it was found at 24.5 to 25.5 σθ during 1990–1994, resulting in a weaker STF intensity. The latitudinal variation of the STF is also connected to the distribution and circulation of the mode waters (Figure 10). In the early 1990s, low-PV waters originating from their formation areas were advected eastward and then to the southwest by the enhanced subtropical gyre, showing a large eastward shift of the low-PV tongue. As a result of this southward shift of the mode water distribution, the latitudinal location of the STF also shifted southward in the early 1990s. These results indicate that water properties of the mode waters significantly changed along with the eastward shift of the mode water paths, and that the decadal variability of the simulated STF is closely related to that of the mode waters.

Figure 12.

Meridional sections of the northward density gradient [10−5 kg m−4] averaged over 160°E–180°E for 1975–1979 (a) and for 1990–1994 (b). The PV gradient deviation ((f/q)(∂q/∂y)ρ=constβ) is shown by color shading in both panels. Areas where the PV gradient deviation is positive are orange, and those where it is less than −2 × 10−11 m−1 s−1 are blue colors.

5. Discussion: Effect of Atmospheric Forcing

[35] So far we have examined the relationship between the STF and mode water variations on a decadal timescale. In this section, we discuss the effects of atmospheric forcing on the decadal variability of the mode waters.

[36] It is well known that in the mid-1970s, the surface winds in the North Pacific experienced a major regime shift, with the westerly wind shifting its axis southward and intensifying [e.g., Nitta and Yamada, 1989]. Figure 13a shows latitude-time sections of zonal wind stress anomalies averaged between 180°E and 140°W. The westerly wind anomalies between 30°N and 40°N show the strengthening of the westerly beginning in the mid-1970s. Then, after a brief period of relaxation in the 1980s, the westerly was again enhanced in the mid-1990s.

Figure 13.

(a) Latitude-time sections of the zonal wind stress [10−1 Nm−2] anomalies averaged over 180°E-140°W. (b) Longitude-time sections of the MLD [m] (colors) and surface density [kg m−3] (contours) averaged over 35°N–40°N. (c) Same as (b) but for the depth of the 12°C isotherm [m] averaged over 30°N–35°N. A 5-year running mean filter is applied.

[37] The wind regime shift in the mid-1970s can cause changes in the formation region of the mode waters. Figure 13b shows longitude-time sections of the MLD averaged between 35°N and 40°N. This area is chosen because the MLD develops during winter (see Figure 8a), and thus it corresponds to the formation area of the mode waters. The MLD increased in the mid-1970s in response to the change in the westerly wind. A similar deepening of the mixed layer following the wind regime shift is observed in the central North Pacific [Deser et al., 1996]. After the end of the 1970s, the region of deep MLD gradually moved eastward, consistent with the eastward shift of the mode water paths (Figure 10). During periods when the westerly wind shifted southward (1985 and 1995), the MLD deepened and the surface density increased.

[38] Changes in the westerly wind can also cause changes in the ocean circulation. Figure 13c shows depth anomalies of the 12°C isotherm averaged between 30°N and 35°N. This isotherm is chosen because it is located almost in the middle of the main thermocline. The averaged area approximately corresponds to the center of the North Pacific subtropical gyre, and thus its depth variability can be an indicator of the strength of the subtropical gyre. Negative anomalies of the 12°C isotherm dominated west of 180°E in the 1970s, indicating that the subtropical gyre was relatively weak during that period. In response to the gradual enhancement of the westerly in the mid-1970s, the anomalies became positive around 1980, about five years after the wind shift, and the subtropical gyre gradually intensified after the 1980s. The time lag of five years probably reflects the propagation of the first baroclinic Rossby wave [Seager et al., 2001; Schneider et al., 2002].

[39] In contrast to the major role of surface winds, the role of surface heat fluxes in the central North Pacific is not clear (Figure 14). In most cases, SST anomalies and surface heat flux anomalies are negatively correlated, suggesting that the variation of the surface heat flux is not the cause of the SST variation but its result. Thus the surface heat flux probably does not greatly influence the STF variation. Beginning in the mid-1990s, the surface heat flux anomalies were negative, reflecting the warming SSTs.

Figure 14.

Latitude-time sections of (a) surface heat flux [Wm−2] and (b) SST [°C] anomalies averaged between 160°E and 140°W. A 5-year running mean filter is applied.

[40] From these results and those of previous studies, we suggest the following mechanism to explain the decadal variation of the STF (Figure 15). In response to the strengthening of the westerly winds in the mid-1970s (Figure 13a), the MLD south of the Kuroshio Extension markedly increased at around 170°E (Figure 13b). As a result of intense cooling as a large amount of heat was released from the ocean surface and the increased southward Ekman transport of cold water owing to the intensification of the westerly in the formation region of the mode waters [Yasuda and Hanawa, 1997], the volume of the mode waters increased and the core layer temperature decreased (Figure 11). The resulting enhanced southward intrusion of the mode waters strengthened the STF during 1975–1979.

Figure 15.

Schematic diagram depicting the decadal variation of the STF related to mode waters for (top) 1975–1979 and for (bottom) 1990–1994.

[41] Beginning in the 1980s, the spin up of the North Pacific subtropical gyre resulting from the strengthening of the westerly wind (Figure 13c) caused increased warm water transport by the Kuroshio [Hanawa and Kamada, 2001]. This increased warm water transport increased the core layer temperature in the mode waters (Figure 11). Associated with the increased strength of the subtropical gyre, the formation area of the mode waters shifted eastward, the mode water paths changed (Figure 10), and the PV of the mode waters increased. As the result, the STF intensity weakened and the latitudinal location of the STF shifted southward during 1990–1994.

[42] In the present study, decadal variations in the STF were largely explained by those in the mode waters, suggesting that the STF is controlled mainly by internal oceanic processes. The effect of atmospheric forcing via changes in the sea surface heat fluxes is not important (Figure 14). Instead, SST variations in the tropics may have indirectly changed the formation rate of the mode waters via changes in the midlatitude westerly wind field, for example, as a result of the “atmospheric bridge” mechanism [Lau and Nath, 1996].

6. Summary

[43] We examined the decadal variability of the STF of the western North Pacific using North Pacific OGCMs, focusing especially on the relationship between the STF and the mode (low-PV) waters, STMW and CMW. In the long-term mean field, the distributions of the STF and STCC simulated by a North Pacific eddy-resolving model (NPERM5) were more consistent with the observed distributions than those simulated by non-eddy-resolving models. The Kuroshio recirculation gyre was well simulated by NPERM5, which contributed to the better STF and STCC simulations. The simulated STF intensity exhibited significant decadal-scale variations: it was stronger in the late 1970s and weaker in the early 1990s. During these two periods, the simulated mode waters showed corresponding differences in their PV minima density, and paths. We also investigated the relationship between the decadal-scale STF variability and atmospheric forcing. The results suggest that the decadal-scale STF variability can be largely explained by that of the mode waters caused by changes in the surface westerly in the mid-1970s.

[44] Warming of the North Pacific subtropical gyre since the late 1990s (Figure 14b) not only weakened the north–south SST gradient, but also caused the warming of the mode water core temperature. We speculate that the relationship between the STF and the mode waters has probably become less defined on a decadal timescale. Elucidation of the mechanism of these long-term variations of the North Pacific subtropical gyre remains for a future study.

[45] It has been pointed out that the decadal variability of the SST in the STF region seems to be correlated with the SST variability of the tropical Pacific [Nakamura et al., 1997]. Variations in the tropics may affect the formation rate of the mode waters via changes in the westerly wind, because the change in the westerly wind in the mid-1970s is considered to be of tropical origin [Yasunaka and Hanawa, 2003]. Hence we suggest that conditions in the tropical Pacific do not play a direct role but indirectly cause variations in the STF.

[46] This study showed the importance of mode waters in causing STF variability. At the same time, the STF might actively influence the variability of mode waters through eddy activities. It is known that significant meso-scale variability occurs in the STF region [Kobashi and Kawamura, 2002]. Since the decadal variability of these eddies may depend on the STF intensity, it may affect northward heat transport in the subtropical gyre thus changing the features of the mode waters. Further study is needed to verify this scenario.

Acknowledgments

[47] The authors would like to express heartfelt thanks to Kimio Hanawa, Toshio Suga, Tamaki Yasuda, and the members of the Oceanographic Research Department of Meteorological Research Institute for valuable comments and fruitful discussions. Comments and suggestions from the editor and two anonymous reviewers were very helpful for the improvement of this paper. This work was funded by the Meteorological Research Institute. Partial support by MEXT Grant-in-Aid for Scientific Research (18540436) was greatly acknowledged.

Ancillary