As an indicator of the Kuroshio Extension (KE) path, the KE northern boundary (KENB) was detected based on the position of the strong winter sea surface temperature (SST) gradient between 142° and 155°E, using high spatial resolution satellite-derived SST for the 30 winters (January to March) from 1982 to 2011. The KE path showed meridional movement with a period of 10–15 yr and an amplitude of about 2° latitude. The changes in latitudinal position of the KE path were initiated by a north–south shift of the Aleutian low (AL). Negative wind stress curl anomalies around 35°N in the eastern North Pacific associated with a northward shift of the AL induced a deepening of the main thermocline depth, and then this deepening signal propagated westward, reaching the KE region after about 3 yr, where it caused the KE path to move northward. The path state of the KE (straight path/convoluted path) modulated on a time scale of 8–12 yr, but this was not significantly correlated with the meridional movement of the KE path. The anticyclonic eddies containing warm-salty water that detached northward from the convoluted KE exerted a strong influence on oceanic conditions in the Kuroshio–Oyashio Confluence (KOC) region. The changes in path state of the KE were related to the path of the Kuroshio south of Japan over the long term; a convoluted (straight) KE path was associated with the Kuroshio taking the offshore nonlarge (nearshore nonlarge or typical large) meander path.
The Kuroshio is the western boundary current of the wind-driven subtropical gyre of the North Pacific. The Kuroshio separates from the coast of Japan and turns eastward at about 35°N. This separated eastward current is called the Kuroshio Extension (KE). The mean KE path is characterized by two quasi-stationary meanders with ridges located around 144° and 150°E (Fig. 1).
The KE path exhibits low-frequency north–south movement (Qiu and Chen 2005, 2010; Joyce et al. 2009; Frankignoul et al. 2011; Qiu et al. 2014), with a meridional displacement of about 200 km (Qiu and Chen 2005, 2010). The meridional shift of the KE path causes significant changes in the near-surface synoptic-scale atmospheric field (Joyce et al. 2009). Observational and numerical studies have established that the north–south movement of the KE path is probably forced by decadal fluctuations in wind stress curl (WSC) over the central North Pacific (Miller et al. 1998; Deser et al. 1999; Seager et al. 2001; Schneider et al. 2002; Qiu 2003; Qiu and Chen 2005, 2010; Taguchi et al. 2005, 2007; Nonaka et al. 2006; Ceballos et al. 2009; Qiu et al. 2014). Negative WSC anomalies in the central North Pacific cause a deepening of the main thermocline, and then the wind-induced deepening signals propagate westward and reach the KE region after a delay of a few years, leading to the northward movement of the KE path, and vice versa. The WSC field in the North Pacific predominantly reflects variations of the Aleutian low (AL) located in the central North Pacific (Ishi and Hanawa 2005). Two dominant time scales have been identified for the AL: an interdecadal-scale (about 20 yr) intensity variation (Minobe 1999) and a decadal-scale (about 10 yr) north–south shift (Sugimoto and Hanawa 2009). Previous studies have pointed that the meridional movement of the KE path responds to changes in WSC associated with the intensity variations in the AL (Qiu and Chen 2005; Taguchi et al. 2007) and its AL north–south shift (Seager et al. 2001; Kwon and Deser 2007).
Qiu and Chen (2005, 2010) showed two dominant states in the KE path, using high spatial resolution satellite altimetry data: an unstable state when the meanders become more obscure in 1995–2001 and 2006–09, and a stable state with two quasi-stationary meanders, observed in the intervening years. Some studies pointed to a relationship between the path state of the KE and the path of the Kuroshio south of Japan (Qiu and Chen 2005; Sugimoto and Hanawa 2012). When the Kuroshio takes the nearshore typical large meander (tLM) path or the nearshore nonlarge meander (nNLM) path and flows through the deeper channel (about 2500 m) of the Izu Ridge (Fig. 2), the KE path adopts a relatively stable state, but when the Kuroshio takes the offshore nonlarge meander (oNLM) path, passing over the shallower part of the ridge (about 1000 m), the KE path tends to be convoluted (i.e., an unstable state). Recently, it has been reported that the path state of the KE exerts a strong influence on ocean–atmosphere conditions in the Kuroshio–Oyashio Confluence (KOC) region. In the unstable state, warm eddies detach northward from the KE (Itoh and Yasuda 2010), resulting in increased sea surface temperature (SST) (Sugimoto and Hanawa 2011; Kouketsu et al. 2012) and upward heat release in winter (Hanawa et al. 1995; Tanimoto et al. 2003; Sugimoto and Hanawa 2011), and a deepening of the winter ocean mixed layer (Kouketsu et al. 2012; Oka et al. 2012).
To gain a better understanding of the air–sea coupled system in the western North Pacific, we investigate long-term modulation of the KE path in winter. Many studies have detected the KE path using satellite altimetry data (Qiu and Chen 2005, 2010; Sugimoto and Hanawa 2012) or in situ temperature profiles, based, for instance, on the 15°C isotherm at 200 m (Kawai 1972) or the 14°C isotherm at 200 m (Joyce et al. 2009; Frankignoul et al. 2011). However, satellite altimetry data have only been available since October 1992, which is too short a time series to identify long-term features of the KE path, and in situ temperature profiles are sparse, making it difficult to detect the spatial features of the KE path (i.e., stable/unstable states).
In winter, two strong SST fronts, which reflect the subsurface oceanic conditions, are found in the western North Pacific (Fig. 1): the subarctic front (SAF) at about 40°N, and a second front at about 36°N, along the KE northern boundary (KENB). In this study, we attempt to deduce KE path perturbations by detecting the KENB from the winter SST gradient, using the satellite-derived SST data of Reynolds et al. (2007) with a high spatial resolution of ¼° (longitude) × ¼° (latitude). This approach provides a new perspective on the long-term KE path variations because the SST product is available from September 1981, more than 11 yr prior to satellite altimetry data. The remainder of this paper is organized as follows. Section 2 outlines the datasets and processing procedures. Section 3 detects the KENB based on the SST gradient and then examines whether or not the KENB reproduces the KE path perturbations in terms of latitudinal position and path state. Section 4 investigates the long-term meridional movement of the KE path and identifies its cause with particular reference to variations in the intensity and meridional position of the AL. Section 5 explores long-term changes in the KE path state and investigates its influence on oceanic conditions in the KOC region. We reconsider the relationship between the path state of the KE and the path of the Kuroshio south of Japan using the longer-term dataset. Section 6 describes our summary and conclusions.
2. Dataset and processing procedures
We detect the KENB using the monthly National Oceanic and Atmospheric Administration (NOAA) optimum interpolation monthly SST product (OISST), which is based on Advanced Very High Resolution Radiometer (AVHRR) infrared satellite data (AVHRR-only product), with a high spatial resolution of ¼° (longitude) × ¼° (latitude) (Reynolds et al. 2007). The analysis period of this study is 30 winters (January–March) in the period of 1982–2011 because the strong SST front around the KENB is observed among the three months (Fig. 1).
To examine the influence of cloud cover on the detection of the KENB from infrared SST data, we compare our results with the monthly SST product based on the Advanced Microwave Scanning Radiometer for the Earth Observing System (AMSR-E) provided by Remote Sensing Systems (http://www.ssmi.com) in 9 winters during 2003–11, with a spatial resolution of ¼° (longitude) × ¼° (latitude).
We use the time series of the KE axis of 1993–2011, which is defined as the 110-cm sea surface height (SSH) contour reconstructed by adding the satellite altimetry dataset from the Archiving, Validation, and Interpretation of Satellite Oceanographic (AVISO) data (Ducet et al. 2000) to the mean dynamic topography (http://www.aviso.oceanobs.com); the 110-cm SSH contour is consistently located at, or near, the maximum north–south gradient of SSH. We use the information published by the Japan Coast Guard in the “Quick Bulletin of Ocean Conditions” for the Kuroshio path south of Japan; this is available twice a month until September 2002 and weekly thereafter. The axis has been estimated comprehensively using data of expendable bathythermograph (XBT) data, surface velocity data obtained by ship-borne acoustic Doppler current profilers (ADCP), and the satellite-derived SST and SSH data.
We use temperature (T) and salinity (S) data archived in the World Ocean Database 2009 (WOD09; Boyer et al. 2009) and profiles from Argo floats (Oka et al. 2007). To control data quality, we first removed profiles duplicated in different data sources. For each profile, we then compared the measured data with all values measured in the same month within a 1° × 1° box, and excluded data if they fell outside of three standard deviations of the mean. After quality control, we vertically interpolated T–S profiles at a 10-dbar interval using the Akima (1970) scheme and calculated potential temperature (θ) and the potential density (σθ). The main thermocline depth (MTD) is taken to be the depth of the θ = 12°C isotherm, which is located in the middle portion of the main thermocline in the western part of the North Pacific subtropical gyre (Uehara et al. 2003).
We use the WSC, calculated as the spatial derivative of wind stress from the Japanese 55-yr Reanalysis (JRA-55) (Ebita et al. 2011), to investigate the influence of atmospheric forcing. As the derivative operation tends to exaggerate small-scale features of the WSC field compared with the atmospheric pressure field, we highlight the large-scale variations, by smoothing the WSC field using a Gaussian filter with an e-folding scale of 200 km. Sugimoto and Hanawa (2009) showed that a meridional movement of the AL is linked to the west Pacific (WP) teleconnection pattern (Fig. 3a), and a change in magnitude of the AL is associated with the Pacific–North American (PNA) teleconnection pattern (Fig. 3b). These two teleconnection patterns are identifiable as the first two leading modes extracted by performing an empirical orthogonal function (EOF) analysis for the WSC field (Ceballos et al. 2009). As an indicator of the AL, we use the WP and PNA indices from the NOAA Climate Prediction Center (NOAA/CPC; http://www.cpc.ncep.noaa.gov), given as rotated empirical orthogonal functions (REOFs) of monthly 700-hPa geopotential height anomalies, obtained based on a methodology based on Barnston and Livezey (1987).
We use a 10% significance level for all correlation coefficients in this study, based on Student’s two-sided t test in which the degrees of freedom are estimated by dividing the data length by the characteristic time scale (a zero-cross scale) calculated according to a lagged autocorrelation function.
3. Detection of the KENB
We detect the KENB from the latitudinal position of the SST front, defined as the latitude of the maximum horizontal SST gradient between 32° and 37°N at each longitudinal grid point from 142° to 155°E, for each month from January to March. Mesoscale eddies and the SAF, which also have strong SST gradients, are occasionally found in this latitudinal band. We inspected the SST front visually, checking both the SST and its anomaly field, and modified any apparent erroneous data points associated with eddies and the SAF. In this study, we use winter mean values.
We examine the influence of cloud cover on the detection of the KENB based on the infrared satellite SST data, by comparing it with the SST based on microwave observation (AMSR-E) of 2003–11. Figure 4 displays time series of the latitudinal position of KENB, zonally averaged between 142° and 155°E, and the pathlength of the KENB from 142° to 155°E, which is a measure of the stability of the KE path (Qiu and Chen 2005). The result shows that the KENB extracted from the two SST products are very similar.
Next, we examine whether or not the KENB reproduces the KE path perturbations, by comparison with the KE axis based on SSH for 1993–2011. The KENB is always located north of the KE axis as expected, by about 1° in latitude (Fig. 4a). The two time series are significantly correlated (R = 0.77). The pathlength of the KE axis is very similar to that of the KENB (Fig. 4b). The significant correlation coefficient is obtained between the two time series (R = 0.77).
These results show that cloud cover has an insignificant effect on the detection of the KENB defined in this study and that the KENB closely reproduces the temporal features of the KE path in terms of latitudinal position and path state. In the following sections, we refer to the KENB as the KE path.
4. Meridional movement of the KE path
a. Long-term variation and its influence on winter SST in the KE region
We investigate long-term variations in the latitudinal position of the KE path and its influence on the winter SST field. The latitudinal position time series (Fig. 4a) shows a north–south movement of up to 2° in latitude with a long-term period of 10–15 yr (Fig. 5a). The KE path is located to the north in the early 1990s and early 2000s, and to the south in the late 1990s and late 2000s. Figure 6a displays a map of correlation coefficients between the time series of latitudinal position and the winter SST field. The meridional movement affects the SST in a narrow latitudinal band between 34° and 36°N.
b. Impact of Rossby waves excited by meridional movement of the AL on the KE path position
Numerous authors have concluded that the meridional movement of the KE path is caused by wind-induced Rossby waves formed in the central North Pacific (Qiu and Chen 2005, 2010; Nonaka et al. 2006; Taguchi et al. 2007; Ceballos et al. 2009). Previous studies have pointed out that the Rossby waves are attributable to the changes in magnitude and meridional movement of the AL (Qiu and Chen 2005; Kwon and Deser 2007; Taguchi et al. 2007; Ceballos et al. 2009).
In this subsection, we investigate the long-term behavior of Rossby waves, using variables smoothed by a 3-yr running mean filter to highlight the low-frequency time scale. To detect oceanic Rossby waves, we specifically focus on MTD variations. Figure 7a displays the MTD anomaly averaged over the meridional band of 33°–35°N, where the zonal mean KE axis is located (Fig. 4a). Decadal-scale variations in the western region are found, and most signals can be traced from the eastern region. These temporal–spatial features are consistent with the SSH anomalies (Fig. 7b), indicating that the MTD anomaly field used in this study closely follows the behavior of oceanic Rossby waves.
We examine the relationship between the meridional movement of the KE path and Rossby waves. The latitudinal position time series is significantly correlated with the MTD anomaly around the KE region (Fig. 8a). Changes in the latitudinal position of the KE path appear to be initiated by Rossby waves, which are generated in the central North Pacific around 165°W (Figs. 7a and 8a), and take about 3 yr to reach the KE region.
To determine what atmospheric forcing generates Rossby waves in the central North Pacific, we first explore the relationship between the latitudinal position time series and the winter WSC field with a lead lag of 3 yr. Large negative signals are obtained around 35°N in the central North Pacific (Fig. 9), which corresponds approximately to the latitudinal position of the KE path. The spatial pattern closely resembles the regression pattern of the WSC against the WP index (Fig. 3a), which is linked to AL meridional movement. Indeed, the latitudinal position time series (Fig. 4a) and the WP index with a lead lag of 3 yr are significantly correlated (R = 0.58). The WP index is significantly correlated with the MTD variations around the Rossby wave formation region (R = 0.57 at 165°W). In contrast, the AL intensity (PNA index) has no significant correlation with either the MTD at 165°W (R = 0.09) or the latitudinal position of the KE path with a delay lag of 3 yr (R = 0.10).
The series of results shows that the negative WSC anomalies around 35°N in the central North Pacific, which are associated with the northward shift of the AL, induce a deepening of the MTD. The deepening signals then propagate westward, reaching the KE region after about 3 years, leading to the northward movement of the KE.
5. Changes in path state of KE
a. Long-term behavior
The pathlength time series (Fig. 4b) shows low-frequency activity, with minima (stable states) in the early 1990s and early 2000s, and maxima (unstable states) in the late 1990s and late 2000s. The time series has a marked period of 8–12 yr since about 1990 (Fig. 5b), which suggests a lengthening of the pathlength during unstable states in recent years.
b. Influence of oceanic conditions in the KOC region
We explore the influence of path state changes on oceanic conditions. First, we investigate the relationship between path state changes and the SST field. The changes in path state are significantly correlated with the SST anomalies in the KOC region, with SST increases occurring when the KE path is in the unstable state; however, no such correlation is observed in the KE region (Fig. 6b). Recent studies using satellite altimetry data reported that the positive SST anomalies in the KOC region are generated by warm eddies detached northward from the KE (Itoh and Yasuda 2010; Sugimoto and Hanawa 2011). We examine relationships between the SST anomalies in the KOC region, the KE path state, and eddy activity over 30 years. Here, as an indicator of eddy activity, we use the regional standard deviation of daily SST anomalies (Reynolds et al. 2007), according to the method proposed by Sugimoto and Hanawa (2011): the SST anomaly field becomes spatially inhomogeneous (homogeneous) with more (fewer) eddies, and this results in a large (small) regional standard deviation. Figure 10 displays the winter eddy activity index in the KOC region (KOCR), specifically for the area displayed in Fig. 6b (36°–39°N, 143°–151°E). It shows a clear decadal-scale variation, and that the values in the high activity period have increased in recent years. These temporal features closely resemble those of the pathlength time series (Fig. 4b), and a significant correlation coefficient is obtained (R = 0.56). In addition, the eddy activity index is significantly correlated with the SST in the KOCR (R = 0.55). These findings suggest that over the long term, positive SST anomalies in the KOC region are related to the high eddy activity associated with the unstable state of the KE path.
We investigate the vertical structure of the ocean to reveal the contribution of eddies to oceanic conditions in the KOC region. We define two categories of KE path years; unstable (stable) state years are those during which the KE pathlength time series (Fig. 4b) exceeds plus (minus) one standard deviation. Then, we prepare meridional cross sections for February of unstable and stable state years on a 50-km latitude grid × 10-dbar interval, applying a Gaussian filter with an e-folding scale of 55 km to θ–S profiles within the longitudinal band of 143°–151°E.
In unstable state years (top panels in Fig. 11), warm-salty water is distributed widely and deeply around 250–300 km north of the KENB. The bowl-shaped structure in θ and S is expected to correspond to an anticyclonic eddy. This structure is not observed in stable state years (middle panels). The θ–S fields in the KOC region for unstable state years are much warmer (>3°C) and saltier (>0.3 psu) than those of stable state years (bottom panels), and the water properties at the 100-dbar depth at the central position (300 km north of the KENB) are θ = 13.0°C, S = 34.4 psu, and σθ = 25.9 kg m−3, corresponding to “Kuroshio water” as defined by Hanawa and Mitsudera (1986). This shows that eddies originating from the KE are primarily responsible for oceanic conditions in the KOC region.
c. Relation with the Kuroshio path state south of Japan
As described above, the changes in path state of the KE are not significantly correlated with its north–south shift. Figure 8b shows that the KE pathlength variations have a delay lag of 1–2 years relative to the arrival of Rossby waves into the KE region, implying that changes in path state are not affected directly by oceanic Rossby waves. A few studies have noted the relationship between the path of the Kuroshio south of Japan and the path state of the KE, based on short-period satellite altimetry data (e.g., Sugimoto and Hanawa 2012). We examine the long-term relationship over 30 yr, using the Kuroshio path from the “Quick Bulletin of Ocean Conditions.” The Kuroshio takes three typical paths (Kawabe 1995): the oNLM, the nNLM, and the tLM (Fig. 2). We defined the paths as follows: the nNLM is located north of 32°N at 136°E, 33°N at 138°E, and 33.5°N at 140°E; the oNLM is situated south of 33.5°N at 140°E; and the tLM is located south of 32°N in the region of 136°–139°E and north of 33.5°N at 140°E. The Kuroshio sometimes takes an undulating path, and such cases are classified as “others.” We classify each winter based on the most frequent path type that occurred; the number of winters for nNLM, oNLM, tLM, and others are 9, 10, 6, and 5, respectively.
Figure 12a displays a histogram of the number of winters with each path plotted against the KE pathlength. Over the 30-yr period, the KE path is more variable when the Kuroshio takes the oNLM, and more stable when the Kuroshio takes the tLM or the nNLM. Figure 12b displays a time series of Kuroshio path types. Interestingly, the Kuroshio path switched frequently between the oNLM and tLM during the 1980s. On the other hand, after 1990, the nNLM and oNLM paths alternated, but less frequently, and there is just one tLM event (in 2005). The long-term modulation of the Kuroshio paths is similar to the decadal-scale temporal features observed since about 1990 (Fig. 5b). Therefore, we suggest that changes in the path state of the KE are related to the paths of the Kuroshio south of Japan.
6. Summary and conclusions
We detected the position of the KENB during the 30 winters from 1982 to 2011 based on the position of the strong SST gradient from 142° to 155°E, using high spatial resolution satellite-derived SST data. The SST field closely reflects subsurface oceanic conditions during the winter, while the detection of the KENB from SST data is more difficult in other seasons. We found that cloud cover had an insignificant effect on the detection of the KENB in this study, based on a comparison with the SST product derived from microwave observations. In addition, the KENB defined in this study effectively reproduced the temporal variation in latitudinal position and path state (stable/unstable state) of the KE axis based on the SSH product, although the KENB was located about 1° in latitude north of the KE axis.
The KE path showed low-frequency north–south movement with a period of 10–15 yr, with a meridional displacement of about 2° in latitude. The meridional movement of the AL initiated the north–south shift of the KE path. Negative WSC anomalies around 35°N in the central North Pacific, associated with the northward shift of the AL, induced a deepening of the MTD, and the deepening signal propagated westward and reached the KE region after about 3 yr, leading to the northward movement of the KE path. The key contribution of this study is the identification of the strong influence of meridional movement of the AL on changes in latitudinal position of the KE path, with a time lag of about 3 yr.
The path state of the KE was modulated on a time scale of 8–12 yr. We pointed that changes in path state had no significant correlation with the meridional movement of the KE path. The long-term (30 yr) dataset showed that the KE path in the unstable state has tended to be longer (more convoluted) in recent years. The path state exerted a strong influence on the oceanic conditions in the KOC region; the unstable state of the KE path was associated with large positive anomalies in both θ and S, not only at the sea surface, but also below 400 dbar. The water properties (θ = 13.0°C, S = 34.4 psu, and σθ = 25.9 kg m−3) were those of the Kuroshio water. The warm-salty Kuroshio water in the KOC region was supplied by anticyclonic eddies, which detached northward from the KE in association with the convoluted path of the KE, as reported previously (Itoh and Yasuda 2010; Sugimoto and Hanawa 2011). Over the long term, the KE path states were related to the paths of the Kuroshio south of Japan. The KE path adopted a relatively stable state when the Kuroshio took the nNLM or tLM path through the deeper channel of the Izu Ridge (about 2500 m), but tended to be convoluted (unstable) when the Kuroshio took the oNLM path over the shallower part of the Izu Ridge (about 1000 m). The results presented here provide more evidence to suggest that the Kuroshio path type strongly influences oceanic conditions in the KOC region.
In winter, decadal-scale SST variations are dominant in the KE/KOC region (Kwon et al. 2010) and several coupled climate models have shown that the KE/KOC region is a key area for understanding decadal-scale ocean–atmosphere interaction (e.g., Kwon and Deser 2007). Therefore, we expect that the Kuroshio/KE path perturbations are responsible for the decadal-scale variations. Satellite altimetry data for the period since October 1992 have greatly improved our understanding of the KE path. However, the short time series of data has restricted our understanding of ocean–atmosphere interaction on the decadal time scale. We believe that our KE path indicators for the longer period from 1982 are a step toward better understanding the decadal-scale air–sea coupled system in the western North Pacific.
The authors thank members of the Physical Oceanography Group at Tohoku University for useful discussions. The Kuroshio axis data in the “Quick Bulletin of Ocean Conditions” were digitized and provided by the Marine Information Research Center. AMSR-E data are produced by Remote Sensing Systems and sponsored by the NASA Earth Science MEaSUREs DISCOVER Project and the NASA AMSR-E Science Team. Three anonymous reviewers provided useful and constructive comments in revising the manuscript. The second author (SS) was partly supported by a Grant-in-Aid for Scientific Research on Innovative Areas (Grant 25106702, “A “hot spot” in the climate system: Extra-tropical air-sea interaction under the East Asian monsoon system”) from the Ministry of Education, Culture, Sports, Science and Technology of Japan, and by a Grant-in-Aid for Young Scientists (B) (Grant 23740348) from the Japan Society for the Promotion of Science.