Global Biogeochemical Cycles

Decrease of dissolved oxygen after the mid-1980s in the western North Pacific subtropical gyre along the 137°E repeat section

Authors

  • Yusuke Takatani,

    Corresponding author
    1. Global Environment and Marine Department, Japan Meteorological Agency, Tokyo, Japan
      Corresponding author: Y. Takatani, Global Environment and Marine Department, Japan Meteorological Agency, 1-3-4 Otemachi, Chiyoda, Tokyo 100-8122, Japan. (y-takatani@met.kishou.go.jp)
    Search for more papers by this author
  • Daisuke Sasano,

    1. Global Environment and Marine Department, Japan Meteorological Agency, Tokyo, Japan
    2. Geochemical Research Department, Meteorological Research Institute, Tsukuba, Japan
    Search for more papers by this author
  • Toshiya Nakano,

    1. Global Environment and Marine Department, Japan Meteorological Agency, Tokyo, Japan
    2. Oceanographic Research Department, Meteorological Research Institute, Tsukuba, Japan
    Search for more papers by this author
  • Takashi Midorikawa,

    1. Geochemical Research Department, Meteorological Research Institute, Tsukuba, Japan
    2. Nagasaki Marine Observatory, Nagasaki, Japan
    Search for more papers by this author
  • Masao Ishii

    1. Global Environment and Marine Department, Japan Meteorological Agency, Tokyo, Japan
    2. Geochemical Research Department, Meteorological Research Institute, Tsukuba, Japan
    Search for more papers by this author

Corresponding author: Y. Takatani, Global Environment and Marine Department, Japan Meteorological Agency, 1-3-4 Otemachi, Chiyoda, Tokyo 100-8122, Japan. (y-takatani@met.kishou.go.jp)

Abstract

[1] The Japan Meteorological Agency has acquired dissolved oxygen (DO) concentration data each year since 1967 along the 137°E repeat section in the western North Pacific. In this data set we found significant regional temporal trends of decreasing or increasing DO concentrations on various isopycnal surfaces. DO decreases were particularly significant after the mid-1980s in the subtropical gyre; mean rates of DO change at 20–25°N for 1985–2010 were −0.28 ± 0.08 μmol kg−1 yr−1 on 25.5 σθ in North Pacific Subtropical Mode Water (NPSTMW), −0.36 ± 0.08 μmol kg−1 yr−1 on 26.8 σθ in North Pacific Intermediate Water (NPIW), and −0.23 ± 0.04 μmol kg−1 yr−1 on 27.3 σθ in the O2 minimum Layer (OML). The cause of DO decrease differed among isopycnal surfaces. On density surfaces shallower than 26.0 σθ (less than about 400 m), the deepening of isopycnal surfaces and decline of oxygen solubility due to ocean warming have had the greatest influence. In particular, between 25.2 σθ and 25.8 σθ near the NPSTMW their combined contributions accounted for >50% of the DO decrease. In the NPIW core at roughly 26.8 σθ (∼700 m), the decline in DO was attributable to the DO decrease in the formation region. In the OML between 27.0 σθ and 27.3 σθ (∼1000 m), the DO decrease likely resulted from an increase in westward transport of low O2 water due to strengthening of the subtropical gyre. The result of this study shows the importance of the long-term and high-frequency along the 137°E repeat section.

1. Introduction

[2] The variability of dissolved oxygen (DO) concentrations in the ocean is closely linked with physical and biogeochemical processes, including changes in air-sea interactions, ocean circulation, biological production near the surface, and degradation of organic matter in the interior of the ocean. It is critical to understand the causes of DO variability in the ocean and the relevant controlling processes in order to distinguish between increases of dissolved inorganic carbon (DIC) due to anthropogenic CO2 invasion and DIC variability caused by changes in ocean circulation or biological activity [Gruber et al., 1996; Wakita et al., 2010; Gruber, 2011].

[3] In the North Pacific, several authors in past decades have investigated the variability of DO or apparent oxygen utilization (AOU), i.e., saturation value of O2 (O*2) at in situ potential temperature and salinity minus measured DO (AOU = O*2 − DO). Most of these studies have indicated that DO concentrations are decreasing within and below the thermocline (potential density: σθ > 26.4) [e.g., Garcia et al., 2005; Whitney et al., 2007] and have suggested that a reduction of ventilation may have been responsible [Watanabe et al., 2001; Mecking et al., 2006, 2008]. Modeling studies have also tended to implicate physical rather than biological processes as the most important causes of the decreasing DO concentrations [Deutsch et al., 2005, 2006]. Yasuda [2004] has suggested that the low salinity-high DO water of the Okhotsk Sea is one of the sources of North Pacific Intermediate Water (NPIW). Some papers [e.g., Osafune and Yasuda, 2006; Nakanowatari et al., 2007] have suggested that the contribution of the Okhotsk Seawater to the formation of NPIW has been decreasing with bi-decadal oscillations related to the nodal tidal cycle. In subtropical regions, it was reported that AOU increased due to the reduction of ventilation or organic matter degradation [Emerson et al., 2004; Mecking et al., 2006]. The majority of previous studies on the variability of DO that used long-term time series record concerned about subarctic regions, and many of studies in the subtropical regions were based on comparisons of data from a few snapshots. Little is known about the short- to long-term variability of DO in subtropical regions.

[4] The Japan Meteorological Agency (JMA) has been conducting a series of hydrographic and hydrochemical observations along a meridional section at 137°E in the western North Pacific. This section extends from the tropics at 3°N off New Guinea across the subtropical gyre to 34°N near the southern coast of Japan (Figure 1). A total of 81 cruises were conducted from 1967 to 2010, with measurements being made each winter beginning in 1967 and twice a year in the winter and summer beginning in 1972. Data along the high-frequency repeat section acquired over the last more than 40 years is globally unique, and is important to be able to evaluate the behavior of water masses over wide range of timescales. Data along 137°E section have been used for many studies of changes in water masses [Kaneko et al., 1998] and air–sea CO2 interactions [e.g., Inoue et al., 1995; Midorikawa et al., 2010; Ishii et al., 2011] that are related to climate change.

Figure 1.

(left) The repeat hydrographic section along 137°E and (right) time-latitude distribution of sample collections. Closed circles in Figure 1, right, denote the collection of DO samples and open circles indicate that no DO samples were collected.

[5] There are several noteworthy water masses in this meridional section. In this study, we especially focused on North Pacific Subtropical Mode Water (NPSTMW), NPIW, and O2 minimum Layer (OML: DO < 70 μmol kg−1). NPSTMW is characterized by low potential vorticity (PV) and is usually observed at depth of around 100–400 m in the northwestern North Pacific subtropical gyre. Suga and Hanawa [1995a, 1995b] used DO concentration data to examine variability of NPSTMW. They found that the PV and AOU of NPSTMW were closely related to the wintertime monsoon index, and suggested that there was a strong relation between the wintertime cooling and the NPSTMW formation.

[6] NPIW is observed as the salinity minimum layer at around 26.8 σθ. Qiu and Joyce [1992] found that the size of the NPIW core, defined as the area of the cross-section where salinity (S) is lower than 34.25, showed large interannual variability between 1967 and 1988. They suggested that the variation was associated with meanders of the Kuroshio Current that blocked NPIW from advecting farther to the west. Shuto [1996] reported that the interannual variability of NPIW is closely related to the minimum (negative value) of the wind stress curl in the area south of Japan. Nakano et al. [2005] reported that interannual and decadal variability of the NPIW salinity minimum core along 137°E sections (S < 34.2) was related to the strength of the westward intrusion of a low salinity tongue and was connected to the variability of wind-forcing over the central North Pacific and heat flux fields in the region of the Kuroshio Extension. Moreover, Nakano et al. [2007] found a long-term trend of increase in the size of the salinity minimum core and a linear freshening with time of the salinity minimum layer compared to the thermocline in the subtropical gyre. They indicated that the freshening was caused by a warming in the upper layer and strengthening of the subtropical gyre.

[7] OML is observed at depths of around 900–1300 m (27.2∼27.4 σθ) in the northwestern North Pacific subtropical gyre. At the depth of near OML, it was suggested that high oxygen water from the South Pacific enters the northwestern North Pacific along its western boundary [Reid, 1997]. North of around 25°N, on the other hand, the deep recirculation flow of subtropical gyre carried low oxygen water, which originates in the subarctic gyre, and show a tongue-like distribution that extends to the west and reaches far into the central Pacific [Reid, 1965; Kaneko et al., 2001].

[8] The analysis presented here focused on long-term variability of DO in the high frequency repeat section along 137°E and in particular on the significant trend of DO decrease that has been observed in the mid-subtropics from 20°N to 25°N since the mid-1980s. The cause of the DO decrease is discussed in terms of warming in the upper layers of the ocean and changes in the formation of water masses and circulation in the subtropical cell.

2. Data

[9] The JMA has acquired DO concentration data from the repeat section along 137°E on a total of 81 cruises, including 43 cruises in winters (January–February) from 1967 to 2010 and 38 cruises in summers (June–August) from 1972 to 2010. Until 1989 discrete water samples had been collected with Nansen bottles fixed on a cable from nominal depths of 0, 10, 20, 30, 50, 75, 100, 125, 150, 200, 250, 300, 400, 500, 600, 700, 800, 900, 1000, and 1250 m (standard depths). Temperature and pressure at the sampling depths had been measured by means of a reversing mercury thermometer attached on each of the Nansen bottles. Since 1990 JMA has been using a CTD-rosette multisampler mounted with Niskin bottles to sample at the same depths, and discrete samples have also been taken from depths of 1500, 1750, and 2000 m and occasionally from 3000, 4000, 5000 m, and near the sea bottom.

[10] Analyses of DO were made by the Winkler titration method [Winkler, 1888] until January 1994. The end-point of the titration was visually determined by adding a small aliquot of starch solution. Beginning with the July–August 1994 cruise, which was conducted as a World Ocean Circulation Experiment (WOCE) Hydrographic Program (WHP) P9 one-time cruise, Carpenter's method [Carpenter, 1965, 1966] has been applied, and an automated titration system with photometric detection (ART-3, Hirama Laboratories Co., Ltd and DOT-01X, Kimoto Electric Co., Ltd) has been used. The standard deviation (SD) of measurements as inferred from the 2166 pairs of replicate analysis for samples taken from the same Niskin bottle was 0.60 μmol kg−1 after 2004. Before that time, no information is available for the results of replicate analyses from the same Niskin bottle, but SD of 244 analyses of paired samples taken from different bottles tripped at the same nominal depth was less than 1.6 μmol kg−1.

[11] There is no standardized way to discriminate bad data due to contamination and measurement error if it wasn't recorded at the time of sampling or measurement. In this work we removed outliers using a statistical method. Details of the method are described in Appendix A. We applied this quality control (QC) to winter and summer data sets separately because the seasonal variation is large in surface and subsurface layers. The data selected through the QC were interpolated to intervals of 10 dbar and 0.05 σθ for each station of each cruise with the Akima spline method [Akima, 1991].

[12] In the measurement of DO, it is necessary to evaluate the systematic error in the data. For example, Johnson et al. [2001] reported the property adjustments for WOCE data in the Pacific by using crossover analysis. In this study, however, it is difficult to estimate the adjustment value for each cruise along 137°E section since there aren't enough data in deep layers in all cruises. Therefore, according to Emerson et al. [2001], we assumed that DO concentrations at the depth of near 2000 m have not changed, and compared the data on 2000 m and 27.65 σθ surface (about 1850 m in the western North Pacific subtropical region) to estimate for systematic errors. The results at six latitudes are shown in Table 1. The standard deviations are mostly in the range between 2.5 and 3.0 μmol kg−1 except that at 30°N. The standard deviation at 30°N is larger due to the influence of the variability in the Kuroshio path. It is thought that the variability or change exceeding 3 μmol kg−1 is significant.

Table 1. Comparison of Deep DO Concentration on 2000 m and 27.65 σθ at Six Latitudes Among All Cruisesa
Latitude2000 m ( σθ = 27.66–27.68) (30°N: σθ = 26.64–27.67)27.65 σθ (∼1850 m)
  • a

    The number of measurements is given by n. Data are given in μmol kg−1.

29.5°–30.5°N104.7 ± 3.9 (n = 42)104.9 ± 3.8 (n = 73)
24.5°–25.5°N112.8 ± 2.5 (n = 66)106.1 ± 2.6 (n = 78)
19.5°–20.5°N114.1 ± 2.4 (n = 58)109.6 ± 2.5 (n = 75)
14.5°–15.5°N113.5 ± 2.7 (n = 56)109.1 ± 2.9 (n = 72)
9.5°–10.5°N115.7 ± 2.8 (n = 55)110.8 ± 2.8 (n = 74)
4.5°–5.5°N114.9 ± 2.6 (n = 61)110.8 ± 2.6 (n = 72)

3. Results

3.1. Variability in the Vertical Section of DO Concentrations at 137°E

[13] In the northern subtropics (20–30°N) there was a clear seasonality of DO in the upper layer (<25.0 σθ) due to seasonal vertical mixing and biological activity, but no remarkable seasonal differences were seen below the density of the winter mixed layer, where DO was lower than 200 μmol kg−1 (Figure 2). There was also no distinct seasonal difference below the subsurface (>24.0 σθ) of the southern subtropics and subtropical–tropical boundary (3–20°N). The oxycline, i.e., the layer in which DO decreases rapidly with depth or density, coincided with the permanent thermocline and was observed at 600–900 m (26.8–27.1 σθ) in the northern subtropics and at 100–200 m (26.0–26.6 σθ) in the subtropical–tropical boundary (5–10°N). The OML was observed near 1000 m (∼27.3 σθ) at 20–30°N below the NPIW and near 300 m (∼26.8 σθ) at 6–9°N (Figures 2a and 2b).

Figure 2.

Typical vertical sections of DO concentrations along the 137°E section on (a) pressure coordinate in winter, (b) pressure coordinate in summer, (c) potential density coordinate in winter and (d) potential density coordinate in summer. Thin (thick) contour lines denote DO concentrations with an interval of 10 (50) μmol kg−1 and shading denotes the magnitude of the standard deviation. Dotted lines in Figures 2a and 2b denote 25.5 σθ, 26.8 σθ and 27.3 σθ surfaces.

[14] In the pressure–latitude sections (Figures 2a and 2b) there were remarkable short-term variations of DO near the oxycline. The wobble of the subtropical gyre, that is, short-term dynamic changes of wind-driven circulation, and the passage of mesoscale eddies were probably responsible for these variations. In order to eliminate the effects of the resultant vertical water motion, we also examined the variability in the density–latitude DO sections (Figures 2c and 2d). Temporal variations of DO were also large above the 27.0 σθ density surface at 3–7°N and above the 26.6 σθ density surface at 3–20°N, where the horizontal gradient of DO on isopycnal surfaces was relatively large. North of 30°N near the Kuroshio there was a large temporal variation above the 26.5 σθ surface. These large temporal variations of DO at the southern and northern rims of the subtropical gyre indicate that the meridional drift of the subtropical gyre, i.e., changes in the paths of the North Equatorial Current in the south and the Kuroshio in the north, has a large impact on the temporal variability of DO on isopycnal surfaces. However, there are no reports of a long-term drift of these currents so far. In contrast, temporal variations of DO were relatively small on all isopycnal surfaces at 20–30°N in the northern subtropics, where the meridional DO gradient on isopycnal surfaces was small.

3.2. Long-Term DO Variability for 1967–2010

[15] To understand the long-term variability of DO in the subtropical gyre along the 137°E section, we calculated the rates of DO change for 20–25°N. Table 2 is a comparison of rates of DO change on 25.5, 26.8 and 27.3 σθ, which are isopynal surfaces of three major water masses (respectively NPSTMW, NPIW and OML) in the western North Pacific subtropical gyre. Trends were calculated by the slopes of linear least squares lines between start years ranging from 1967 to 1990 and end years ranging from 1980 to 2010.

Table 2. The Rates of DO Change for 20–25°N on 25.5 σθ, 26.8 σθ and 27.3 σθ in Various Periodsa
Start YearEnd Year
1980198519901995200020052010
  • a

    Each value is shown the rate of DO change from the start year to the end year. Data are given in μmol kg−1 yr−1. Bold indicates significance at the 95% confidence level, with confidence levels being estimated using a standard t-test.

25.5 σθ
19670.32 ± 0.140.13 ± 0.100.11 ± 0.07−0.07 ± 0.060.15 ± 0.050.13 ± 0.040.14 ± 0.03
19700.23 ± 0.200.05 ± 0.120.06 ± 0.080.13 ± 0.070.20 ± 0.050.17 ± 0.040.17 ± 0.03
19750.35 ± 0.45−0.06 ± 0.200.01 ± 0.110.24 ± 0.090.29 ± 0.060.22 ± 0.050.21 ± 0.04
1980-−0.02 ± 0.560.02 ± 0.200.39 ± 0.130.39 ± 0.090.26 ± 0.070.23 ± 0.05
1985--0.80 ± 0.410.97 ± 0.190.65 ± 0.120.33 ± 0.090.28 ± 0.08
1990---1.23 ± 0.550.50 ± 0.23−0.06 ± 0.14−0.09 ± 0.09
 
26.8 σθ
19670.20 ± 0.230.31 ± 0.140.39 ± 0.100.18 ± 0.080.07 ± 0.06−0.02 ± 0.05−0.04 ± 0.04
19700.02 ± 0.310.27 ± 0.170.39 ± 0.110.15 ± 0.090.03 ± 0.07−0.06 ± 0.05−0.07 ± 0.04
1975−0.92 ± 0.640.25 ± 0.280.45 ± 0.160.09 ± 0.11−0.05 ± 0.080.13 ± 0.060.13 ± 0.05
1980-1.66 ± 0.600.95 ± 0.270.08 ± 0.18−0.12 ± 0.120.21 ± 0.080.19 ± 0.06
1985--1.28 ± 0.730.52 ± 0.300.48 ± 0.160.45 ± 0.110.36 ± 0.08
1990---1.97 ± 0.460.75 ± 0.240.53 ± 0.150.30 ± 0.10
 
27.3 σθ
19670.22 ± 0.06−0.05 ± 0.070.04 ± 0.050.03 ± 0.03−0.04 ± 0.030.08 ± 0.020.08 ± 0.02
19700.27 ± 0.08−0.10 ± 0.080.03 ± 0.060.02 ± 0.04−0.05 ± 0.030.09 ± 0.020.09 ± 0.02
19750.02 ± 0.140.36 ± 0.13−0.02 ± 0.090.00 ± 0.050.09 ± 0.040.13 ± 0.030.12 ± 0.03
1980-−0.04 ± 0.410.35 ± 0.170.17 ± 0.08−0.03 ± 0.060.11 ± 0.040.10 ± 0.04
1985--0.52 ± 0.400.03 ± 0.130.21 ± 0.080.24 ± 0.050.23 ± 0.04
1990---−0.07 ± 0.170.34 ± 0.100.30 ± 0.060.17 ± 0.06

[16] On 25.5 σθ in the lower layer of the NPSTMW, DO change had decreased from 1970s to 2000s. Especially, DO decrease from 1980s to 2000s is significant. On 26.8 σθ in the salinity-minimum layer of the NPIW core, DO change was the increasing trend before 1990. In contrast, DO change has decreased from 1985–1990 to 2010. From these results, it is suggested that long-term variability of DO on this isopycnal surface has decadal cycles. On 27.3 σθ in the OML, although DO change had increased from 1967–1970 to 1980, DO change had decreased from 1970s to 2000s, as well as the change on 25.5 σθ. DO decrease from 1985–1990 to 2000s is significant. On each isopycnal surfaces, DO decrease after the mid-1980s is significant in the western North Pacific subtropical gyre.

3.3. Decrease of DO After the Mid-1980s

[17] From Result 3.2, for the time period 1985–2010 the tendency of DO to decrease with time was significant on each density in the subtropical gyre. We focused on the trend of DO decrease after the mid-1980s. Figure 3 shows the rates of DO change from 1985 through 2010 from the slopes of linear least squares lines fit to time series of DO anomalies from the mean seasonal values at intervals of 0.1 σθ in density and 1° in latitude. At 20–25°N in the mid-subtropics, DO decreased significantly in a vertical band over most density ranges from 25.3 σθ (∼250 m) in the NPSTMW through 27.3 σθ (∼1000 m) in the OML. This pattern is evident in Figure 4, which shows the time series of DO anomalies averaged for 20–25°N for 25.5 σθ, 26.8 σθ and 27.3 σθ together with the time series of physical and other oxygen parameters such as O*2 and AOU. On these density surfaces, the range of the variability of DO (the minimum is 4.5 μmol kg−1 on 27.3 σθ) is beyond systematic errors.

Figure 3.

Linear trends of DO from 1985 to 2010 with confidence greater than 95% on each isopycnal surface at intervals of 0.1 σθ. Solid contour lines indicate the mean salinity for 1965 to 2010 at intervals of 0.1. The area surrounded by a dotted line is where DO decreased significantly over most density ranges.

Figure 4.

Time series of physical and oxygen parameters averaged for 20–25°N on (a) 25.5 σθ, (b) 26.8 σθ, and (c) 27.3 σθ. Dotted lines indicate the linear trends with confidence greater than 95% from 1985 to 2010.

[18] DO concentrations on 25.5 σθ surface began to decrease in the late 1980s, the mean rate of DO change for 1985–2010 being −0.28 ± 0.08 μmol kg−1 yr−1 (Figure 4a). During the same period the position of this density surface was deepening (0.92 ± 0.33 m yr−1), potential temperature and salinity were decreasing (−0.0116 ± 0.0014 and −0.0035 ± 0.0004 yr−1, respectively), and O*2 and AOU were increasing (0.06 ± 0.01 μmol kg−1 yr−1 and 0.32 ± 0.08 μmol kg−1 yr−1, respectively). Similar changes of DO and physical parameters on this density surface also occurred at 26–30°N immediately offshore of the Kuroshio.

[19] On 26.8 σθ surface the DO anomaly was at maximum (+15 μmol kg−1) in 1989 and has decreased with time since then (Figure 4b). The mean rate of DO change from 1985 to 2010 (−0.36 ± 0.08 μmol kg−1 yr−1) was the highest among the density layers in this region. Potential temperature and salinity at 26.8 σθ were higher until the late 1980s, but neither parameter has evinced a significant trend with time since then. The increase of AOU since the mid-1980s has been the primary cause of the decrease of DO.

[20] On 27.3 σθ surface, the interannual variability of DO has been smaller than in the layers above, and the DO anomaly has been mostly negative since the mid-1990s. The mean rate of DO change has been −0.23 ± 0.04 μmol kg−1 yr−1 from 1985 to 2010 (Figure 4c). Long-term changes in potential temperature, salinity, and depth of this density surface are not significant. Instead, a significant increase in AOU has been responsible for the DO decrease.

[21] The latitude band from 16°N to 17°N in the southern part of the North Pacific subtropical gyre is another region where a significant DO decrease has occurred on most density surfaces deeper than 26.0 σθ (Figure 3). DO has declined most rapidly (−0.72 ± 0.26 μmol kg−1 yr−1) on 26.2 σθ above the salinity minimum layer of the NPIW at 16°N and has also decreased significantly on 26.6 σθ in the center of the salinity minimum (−0.41 ± 0.18 μmol kg−1 yr−1). The AOU increases of 0.77 ± 0.26 μmol kg−1 yr−1 on 26.2 σθ and 0.46 ± 0.15 μmol kg−1 yr−1 on 26.6 σθ have been primarily responsible for the DO decreases on these density surfaces.

4. Discussion

[22] In general, the variability of DO in the ocean is ascribed to the variability of biogeochemical and physical processes, including the release and consumption of DO by net community production, vertical and horizontal mixing, overturning circulation, and net sea–air O2 exchange [Keeling et al., 2010]. The variability of physical processes is driven by the undersaturation or supersaturation of O2 in surface waters due to the aforementioned processes as well as to changes in temperature and salinity. However, before considering the role of these processes in determining the trends in DO observed on various density surfaces from 25.2 σθ to 27.4 σθ, we first examine the apparent effects of isopycnal surface deepening and decreases of O*2, both of which are associated with long-term ocean warming.

4.1. Effects of Ocean Warming on DO Decrease

[23] Over the past 50 years there has been a rise in temperature in the upper layer of the ocean in the subtropical North Pacific as well as in other ocean basins worldwide [Levitus et al., 2005]. Along the 137°E section the temperature in the layers which are between the main thermocline and the salinity minimum layer has been increasing linearly with time at more than 0.01°C yr−1 from 1967 to 2005 [Nakano et al., 2007]. A large decadal variability in temperature has also been observed throughout the North Pacific, and this ocean warming has been particularly enhanced since the late 1990s. One of the consequences of ocean warming is a reduction of the solubility of oxygen in seawater. In addition, the lowering of water density associated with ocean warming causes isopycnal surfaces to deepen. Nakano et al. [2007] demonstrated that isopycnal surfaces have been deepening at a rate greater than 1.0 m yr−1 at 22–25°N above the salinity minimum layer of the NPIW (∼700 m) in the 137°E section. Because salinity decreases with depth in the layers above the salinity minimum layer, the warming causes isopycnals to be fresher and cooler, as evidenced, for example, on 25.5 σθ at 20–25°N (Figure 4a) and as shown schematically in Figures 5a, 5b, 5c and 5d. Since DO (O*2, AOU) decreases (increases) with depth in the layers above the minimum (maximum) layer as well as salinity, the deepening of isopycnal surfaces causes the DO (O*2, AOU) decrease (increase) on isopycnal surfaces (Figure 5).

Figure 5.

Schematic of temporal variation of DO on a typical isopycnal surface. Thin (dotted) line denotes the profile before (after) ocean warming.

[24] A comparison of pentadal mean vertical profiles of physical properties and DO concentrations with respect to depth at 20–25°N for 1986–1990 and 2006–2010 reveals that in the upper 450 m, the water in 2006–2010 was warmer and thus the depths of isopycnal surfaces deeper than in 1986–1990 (Figure 6). The comparison also reveals that DO and O*2 in the upper 450 m were 1.48–2.39 μmol kg−1 and 2.15–2.75 μmol kg−1 lower, respectively, in 2006–2010 than in 1986–1990. On the other hand, the change of AOU between these two pentads was small (0.26–0.72 μmol kg−1).

Figure 6.

Pentadal mean vertical profiles of physical and oxygen properties at 20–25°N in 1986–1990 (solid line) and in 2006–2010 (dotted line).

[25] Because DO = O*2 − AOU, the change of DO on an isopycnal surface (D(DO)/Dt) can be expressed by the following equation:

display math

The observed DO change on an isopycnal layer is to a certain extent attributable to deepening of the isopycnal due to warming. The change due to deepening of the isopycnal can be estimated by the multiplication of the slope of the vertical profile (∂X/∂z) and the change rate of the isopycnal (∂z/∂t). Therefore, the change of DO due to deepening of isopycnal layer can be calculated by ∂DO/∂z · ∂z/∂t ((ii) in Figure 5). The change of O*2 due to the change of solubility when water of the same density was last in contact with the atmosphere is expressed by the change of O*2 observed on an isopycnal layer ((iii) in Figure 5: DO*2/Dt) minus that caused by deepening of the isopycnal layer ((iv) in Figure 5: ∂O*2/∂z · ∂z/∂t). Similarly, the change of AOU is expressed by the change of AOU observed on an isopycnal layer ((v) in Figure 5: D(AOU)/Dt) minus that caused by deepening of the isopycnal layer ((vi) in Figure 5: ∂AOU/∂z · ∂z/∂t). As mentioned above, the change of DO isopycnal surface can be calculated by the following equation:

display math

[26] We calculated the contribution of each of the factors (ii, [iii – iv], and [v – vi]) to the DO decrease on each isopycnal layer (i) between 20°N and 25°N from the time series of physical and oxygen properties on each layer (see Figure 4) and their pentadal mean vertical profiles with respect to depth (see Figure 6) using equation (2) (Figure 7). On density surfaces shallower than 26.0 σθ (<450 m), the long-term DO decrease was attributable to the composite of all three factors (ii, [iii – iv] and [v – vi]). The total contribution from the deepening of isopycnal layers (ii) and the solubility reduction ([iii – iv]) accounted for more than 50% of the DO decrease on density surfaces above 25.8 σθ. This range of densities corresponds to the water that outcrops and subducts around the Kuroshio Extension in winter. Between 26.0 σθ, and 26.9 σθ the influence of isopycnal deepening is still discernible, but it is clear that increases in AOU are the dominant factors controlling DO decreases.

Figure 7.

Causes of DO decrease on each isopycnal surface. Black squares indicate the contribution of the change of the depth of the isopycnal surface, the change of O*2, and other factors estimated from the change of AOU, respectively, to the decrease of DO. Closed circles and error bars denote the mean and SD of the rate of DO decrease from 1985 to 2010 on each isopycnal surface.

[27] The warming of the NPSTMW illustrates ocean warming in the upper layer of the northwestern subtropics in the North Pacific. NPSTMW is formed immediately south of the Kuroshio and the Kuroshio Extension by convective mixing caused by the winter monsoon (Figure 8). The lower part of the deep winter mixed layer is retained as a pycnostad in warmer seasons and subducts and spreads over the northwestern subtropical gyre through the southwestward recirculation of the Kuroshio. It is observed as a low PV layer at depths around 100–400 m and at densities of 24.8 σθ–25.5 σθ [e.g., Suga et al., 1989; Suga and Hanawa, 1995a; Oka, 2009]. In the 137°E section the temperature in the core of NPSTMW has tended to increase with time since 1980 between 20°N and the Kuroshio (Figure 9, data available from http://www.data.kishou.go.jp/kaiyou_/shindan/b_1/stmw/stmw.html). A combination of processes are believed responsible for the interannual and decadal-to-interdecadal changes in temperature of NPSTMW including,

Figure 8.

Schematic diagram of formation and distribution of North Pacific Subtropical Mode Water (NPSTMW) and North Pacific Intermediate Water (NPIW). NPSTMW is formed immediately south of the Kuroshio and the Kuroshio Extension by convective mixing caused by the winter monsoon. It spreads in the northwestern subtropical gyre as the layer of Potential Vorticity minimum at 25.0–25.4 σθ [Hanawa and Talley, 2000]. NPIW is a subtropical gyre salinity minimum at 26.6–27.0 σθ. It is formed in the Kuroshio–Oyashio Interfrontal Zone east of northern Japan [Talley, 1993] from the Oyashio water that has its origins in Okhotsk Sea Mode Water (OSMW) and the Western Subarctic Gyre (WSAG) [Yasuda, 1997]. An additional source of NPIW has been found in the Alaskan Gyre (AG) in winter [You et al., 2000]. California Undercurrent (CUC) is thought to have a large contribution to the water in the Alaskan Gyre [Whitney et al., 2007].

Figure 9.

(top) Time series of temperature in the core of NPSTMW along the 137°E section (data available from http://www.data.kishou.go.jp/kaiyou/shindan/b_1/stmw/stmw.html). (bottom) Time series of O*2 calculated from temperature and salinity in the core of the NPSTMW.

[28] 1. Changes in heat loss at formation regions due to changes in the East Asian winter monsoon [Suga and Hanawa, 1995b; Yasuda and Hanawa, 1999],

[29] 2. Changes in the path of the Kuroshio Extension [Qiu and Chen, 2006; Oka, 2009] on shorter time frames, and

[30] 3. On longer time frames, changes in horizontal heat transport by the Kuroshio due to changes in the Westerlies over the central North Pacific through the spin-up of the subtropical gyre [e.g., Hanawa and Kamada, 2001; Yasuda and Kitamura, 2003].

[31] The reduction of oxygen solubility due to such warming of NPSTMW has been an important factor responsible for the decrease of DO in the upper layer of the 20–25°N latitudinal band in the 137°E section.

4.2. Influence of the Strengthening of the Circulation Field

[32] The distribution of DO in the ocean interior has potentially been altered over the past 40 years through strengthening of the North Pacific subtropical gyre due to a change in the magnitude of the Westerlies [e.g., Yasuda and Sakurai, 2006]. Examination of climatological fields of DO on 25.5 σθ, 26.8 σθ, and 27.3 σθ and the contour lines of acceleration potential in the North Pacific reveals that on the 25.5 σθ and 26.8 σθ surfaces, DO at 20–25°N on 137°E does not differ significantly compared to DO in the upstream regions in the central and eastern North Pacific (Figures 10a and 10b, data available from http://www.nodc.noaa.gov/OC5/WOA05/woa05data.html). Therefore, in these upper to intermediate layers, it is unlikely that strengthening of the subtropical gyre has been directly causing DO to decrease by transporting more water with a lower DO content.

Figure 10.

Climatological maps of DO concentrations (color) and acceleration potential (contours) on (a) 25.5 σθ, (b) 26.8 σθ, and (c) 27.3 σθ surfaces. Thick (thin) contours are drawn with intervals of 1 (0.2) m2 s−2. The meridional line denotes the 137°E section (data available from http://www.nodc.noaa.gov/OC5/WOA05/woa05data.html).

[33] In contrast, climatological fields of DO on density surfaces deeper than 27.0 σθ show that DO is lower in the eastern North Pacific than in the 137°E section in the western Pacific (Figure 10c). This distribution is seen by the influence of intrusion of high DO water from the South Pacific [Reid, 1997]. This distribution suggests that the DO decrease that has been observed in the OML around the 27.3 σθ surface in most regions of the 137°E section is attributable to enhanced transport of lower DO water from the eastern North Pacific due to strengthening of the subtropical gyre. To demonstrate the importance of enhanced transport from the east, we compared the σθ–DO diagrams from 20°N to 25°N observed in 1987–2010 along the 137°E section with similar observations in the same latitudinal band along a section at 165°E for August–October 1991 using data from a WHP P13C cruise on the R/V Hakuho-Maru (Figure 11, data available at http://cchdo.ucsd.edu/pacific.html). Values of the DO minimum at around 27.3 σθ in the 137°E section ranged from 62 to 64 μmol kg−1 in 1985–1995 and decreased to 58–61 μmol kg−1 in 1999–2010, whereas the comparable value in the 165°E section was 46 μmol kg−1 in 1991. It is evident that DO concentrations on density surfaces along 137°E have been decreasing with time, and the σθ–DO contours are approaching those observed at 165°E.

Figure 11.

Time-variation of the averaged σθ–DO curve for 20–25°N. Solid line denotes the σθ–DO curve for 20–25°N along the WHP-P13C (165°E) section.

4.3. Influence of DO Changes in the NPIW Formation Region

[34] In sections 4.1 and 4.2 we described a significant trend of DO decrease after the mid-1980s on density surfaces shallower than 27.5 σθ in the mid-subtropical (20–25°N) and in the southern subtropical (16–17°N) zones of the 137°E section (see Figure 3). The magnitude of the DO change was particularly large (−0.7 to −0.3 μmol kg−1 yr−1) on the intermediate density domain of 26.3–27.0 σθ across the NPIW and is ascribed neither to the direct effect of ocean warming nor strengthening of the subtropical gyre but rather to the net increase in AOU. The cause of the AOU increase on these density surfaces is not yet clear, but we presume that it is related to the changes in the sources of NPIW.

[35] NPIW is formed in the Kuroshio–Oyashio Interfrontal Zone east of northern Japan [Talley, 1993]. One of the main sources of NPIW is the Oyashio water that flows southwestward along the east coast of the southern Kuril Islands and Hokkaido. The Oyashio water has its origins in East Kamchatka Current Water from the Western Subarctic Gyre (WSAG) and Okhotsk Sea Mode Water (OSMW), which mix diapycnally through tidal processes in the vicinity of the Bussol Strait of the Kuril Islands [e.g., Yasuda, 1997; Nakamura et al., 2004; Ono et al., 2007]. The salinity of NPIW then increases and its DO decreases as it spreads over the intermediate layer of the subtropical–subarctic transition zone and the subtropical gyre in the North Pacific. We found a significant increase of AOU in the NPIW after the mid-1980s in the mid-subtropical zone (20–25°N) of the 137°E section (see Figures 3, 4b and 7), the same zone where the salinity of the NPIW was at minimum (see Figure 3). Nakano et al. [2005] consider that the NPIW observed within this zone has advected from the east of northern Japan through a more interior pathway of the subtropical gyre than is the case in the southern part of the North Pacific subtropical gyre and could be more sensitive to changes in the process of formation in the northwestern North Pacific (see Figure 8).

[36] Several studies have revealed a significant long-term decline and large interdecadal variability of DO in the Oyashio and NPIW formation regions [Ono et al., 2001; Takatani et al., 2007; Chiba et al., 2010]. The largest mean rate of DO change for the period 1972–2008 (−0.55 ± 0.18 μmol kg−1 yr−1) occurred on 26.7 σθ [Chiba et al., 2010]. A decrease of DO has also been reported in the WSAG [Andreev and Watanabe, 2002] and in the Okhotsk Sea [Nakanowatari et al., 2007]. It is possible that the DO decrease in these source waters has been the cause of the DO decrease in the Oyashio water and NPIW. Takatani et al. [2007] indicated that the PV of the Oyashio had been oscillating decadally and was increasing in the long term together with an oscillation and long-term decrease of DO. On the basis of this coupling between PV and DO changes in the Oyashio, they ascribed the decadal variability and long-term decline of DO in the Oyashio to changes in the mixing ratio of the East Kamchatka Current Water from the WSAG (lower DO, higher PV, higher salinity, and higher temperature) and OSMW (higher DO, lower PV, lower salinity, and lower temperature). The decadal change in the mixing ratio of the two source waters is thought to be controlled by the 18.6-year nodal tidal cycle [Yasuda et al., 2006; Osafune and Yasuda, 2006] that affects tidal mixing in the Bussol Strait and/or the change in circulation of the WSAG. The latter is associated with the change in the strength of the Aleutian Low as indicated by the North Pacific Index [Trenberth and Hurrell, 1994].

[37] The long-term rate of DO change that we found in NPIW in the mid-subtropical zone of the 137°E repeat section after the mid-1980s, −0.36 ± 0.08 μmol kg−1 yr−1, is similar to the mean rate of DO decrease in the Oyashio, and the decadal variability in this region is reasonable to that of previous studies about Oyashio or NPIW [Ono et al., 2001; Nakanowatari et al., 2007]. The increase of AOU in the NPIW (>0.5 μmol kg−1 yr−1) has also been found in the zonal section along 24°N around the date line between 1985 and 2005 [Kouketsu et al., 2010]. Because the trend of DO decrease that we found in NPIW was presumably related to the DO decrease in the Oyashio, it is important to understand how the signal of DO change in the Oyashio propagates into the intermediate layer of the North Pacific through the formation and circulation of NPIW.

[38] Around 26.5 σθ at the southern part of the North Pacific subtropical zone from 16°N to 17°N in the 137°E section is another zone where we found a significant AOU increase in NPIW after the mid-1980s (see Figure 3). There is a possibility that NPIW in this zone is different from that in the mid-subtropical zone. You et al. [2000] pointed out that there is an additional source of NPIW in the Gulf of Alaska, i.e., Gulf of Alaska Intermediate Water (GAIW). It contributes to NPIW in the eastern part of subtropical gyre east of date line, where it mixes with the NPIW of western origin, and spread into the southwestern subtropical gyre (see Figure 8). Whitney et al. [2007] reported that the 50-yr DO time series from Ocean Station P (50°N, 145°W) in the Gulf of Alaska shows long-term trend of DO decrease on isopycnal surfaces in GAIW. In the subtropical and the subtropical–subarctic transition zones in the eastern North Pacific along 152°W and 30°N, Emerson et al. [2001, 2004] and Mecking et al. [2008] discovered a large increase of AOU (>1 μmol kg−1 yr−1) near 26.6 σθ in the salinity minimum layer. It is suggested that the AOU increase at 16–17°N in the southern part of the North Pacific subtropical gyre along the 137°E section is related to the AOU increase in the eastern North Pacific.

5. Conclusions

[39] We found the decadal variability in the intermediate water and a significant trend of DO decrease after the mid-1980s over a broad expanse of the subtropical zone along the 137°E high frequency, long-term repeat hydrographic section in the western North Pacific Ocean, and we focused on the tendency of DO to decrease after the mid-1980s. When combined with earlier works, the results suggest that DO has been decreasing basin-wide in the North Pacific and not just in the subarctic zone and in the eastern tropical/subtropical Pacific zone [e.g., Watanabe et al., 2001; Mecking et al., 2008; Stramma et al., 2008]. The major controlling factors as well as the mean rate of DO decrease differ among density domains and regions, but there is compelling evidence that the decreasing trends in DO are a consequence of anthropogenic climate change over the North Pacific. The deepening of isopycnal surfaces and reduction of oxygen solubility due to ocean warming, and increased horizontal heat transport by the Kuroshio have all had significant impacts on the decrease in DO in the upper layers shallower than the 26.0 σθ density surface. The spin-up of the subtropical gyre due to strengthening of the Westerlies over the central Pacific not only increased heat transport by the Kuroshio but also increased the westward transport of lower DO water from the eastern North Pacific and thus reduced DO concentrations on 27.0–27.4 σθ density surfaces over the OML in the 137°E section. In the salinity minimum intermediate layers defined as NPIW, the mean rate of DO decrease has generally been the highest and appears to have been caused by an increase in AOU. The AOU increase in NPIW along 137°E is presumably linked to the AOU increase in the NPIW formation region and furthermore in Oyashio water, one of the main sources of NPIW. It was suggested that the formation rate of the subsurface water in the North Pacific and the reason for the DO (AOU) decrease (increase) in these regions were caused by the reduction of ventilation [Watanabe et al., 2001; Ono et al., 2001]. We hypothesize that this is due to an increase in the contribution of low-DO WSAG water to the Oyashio, with the strengthening of the Aleutian Low being the climate dynamical driver of the change in the WSAG circulation.

[40] Because the budgets of DO and carbon in the ocean are linked through biogeochemical processes and ocean circulation, understanding changes in DO in the ocean can facilitate understanding of carbon dynamics and associated controlling processes. Ocean monitoring of DO and carbon concentrations along with relevant hydrographic and hydrochemical properties for an extended period over appropriate spatial scales together with the development of computer simulations of these changes using ocean circulation/biogeochemical models is a key to a better understanding of changes in DO and carbon and their links to global warming, other climate-related ocean changes, and changes on shorter time scales.

[41] Finally, the result of this study which we found is a fruit by data along high-frequency repeat section acquired over the last more than 40 years by JMA. This long-term monitoring could be allowed to identify the signal-to-noise for DO change. Because it is unclear in which phase of decadal change was observed by snapshots such as WOCE hydrographic section, the uncertainty in the estimate of a long-term variability might be large. In contrast, the long-term monitoring of high-frequency repeat section can reduce its uncertainty. This study shows the importance of the long-term monitoring. It will be necessary to monitor on combining high-frequency repeat sections such as 137°E repeat section and a decadal survey of high-quality, high spatial and vertical resolution measurements such as WOCE hydrographic section.

Appendix A:: QC Method Using a Statistical Method

[42] We first calculated the mean and SD for each standard layer (0, 10, 20, 30, 50, 75, 100, 125, 150, 200, 250, 300, 400, 500, 600, 700, 800, 900, 1000, 1250, 1500, 1750 and 2000 m) and for each band of density at intervals of either 0.2 σθ or 0.05 σθ at horizontal intervals of 1° of latitude. In calculating the mean and SD at each 1° for the data QC, we used all data within ±2.5° of latitude to avoid removing the signal of natural variability due to meridional shifts of circulation. In surface and subsurface layers with lower density, smaller numbers of data were usually available in a unit interval of density (0.2 σθ or 0.05 σθ), and the concentration gradient with respect to density change was small. In contrast, there were larger numbers of data in a unit density interval, and the concentration gradient was steep in the deeper layers. We took these characteristics into account when we calculated the mean and SD by applying two kinds of potential density bands.

[43] We next interpolated each profile of mean, mean − 3SD, and mean + 3SD, respectively, with respect to pressure or density using the Akima spline method [Akima, 1991]. The data within ±0.5° of latitude that lay out of the interpolated curves of mean ±3SD were then rejected. However, for vertical profiles with respect to pressure, the outlying data above 2000 dbar were not rejected because there is significant variability in DO due to changes of ocean circulation such as the passage of mesoscale eddies. This statistical treatment was repeated three times until no data lay outside of the interval mean ±3SD. On the basis of this QC, 581 data (0.980%) out of 59,249 data were rejected (Figures A1 and A2). Figure A1 shows distributions of SD at each QC step in winter. The number of upper right denotes the number of the data within the interval mean ±3SD. For example, Figure A2 shows the result of QC at 25°N.

Figure A1.

Distributions of SD at each QC step on (left) pressure coordinates deeper than 2000 dbar and (right) potential density bands of 0.2 σθ in winter. (a, b) Before QC, (c, d) after 1st QC, (e, f) after 2nd QC, and (g) after 3rd QC.

Figure A2.

DO data at 25°N in winter after QC. (a) Pressure coordinate, (b) potential density bands of 0.2 σθ, and (c) potential density bands of 0.05 σθ. Closed circles denote data for 25°N ± 0.5°N; plus symbols denote data for 25°N ± 2.5°N; open circles denote data excluded by all QC, and open double circles denote data excluded by each QC. Gray line indicates the mean, and gray dotted line indicates the mean ± 3SD.

Acknowledgments

[44] We are grateful to the captains and crews of R/Vs Ryofu-Maru and Keifu-Maru for their laudable long-term observation efforts. We thank members of the Japan Meteorological Agency and the Meteorological Research Institute for helpful advice and discussions. We also thank K. B. Rodgers for useful suggestions. We are grateful to anonymous reviewers for valuable comments which helped to improve the paper. This research was partly supported by the Meteorological Research Institute's priority research fund for the study of ocean carbon cycle changes.