RESEARCH ARTICLE10.1002/2013JC009529

Interannual variations of Kuroshio transport in the East ChinaSea and its relation to the Pacific Decadal Oscillation andmesoscale eddiesEndro Soeyanto1, Xinyu Guo1,2, Jun Ono3,4, and Yasumasa Miyazawa2

1Center for Marine Environmental Studies, Ehime University, Matsuyama, Japan, 2Research Institute for Global Change,Japan Agency for Marine-Earth Science and Technology, Yokohama, Japan, 3National Institute of Polar Research,Tachikawa, Japan, 4Graduate School of Frontier Sciences, University of Tokyo, Kashiwa, Japan

Abstract Results of a data-assimilative ocean model (JCOPE2) from 1993 to 2012 were used to examinethe correlation between the Pacific Decadal Oscillation (PDO) index and interannual variations of the Kur-oshio transport in the East China Sea (ECS) and the influences of mesoscale eddies on this correlation. In aperiod from 1993 to 2002, the Kuroshio transport estimated from the JCOPE2 reanalysis has a positive corre-lation with the PDO index. This well-known correlation became weak or even disappeared when the analy-sis period was extended from 1993–2002 to 1993–2012. This occurs because the variation range of the PDOindex became small during enhanced mesoscale eddy activity southeast of Taiwan in years after 2002. Theeddies caused a larger variation in the Kuroshio transport in the years after 2002 than before 2002, andtherefore, changed the correlation between the PDO index and Kuroshio transport in the ECS. The influenceof mesoscale eddies on the Kuroshio transport has strong regional dependence: the Kuroshio transportfrom the area east of Taiwan to the midway along the shelf break in the East China Sea depends mainly oneddies arriving from southeast of Taiwan, while transport from the midway along the shelf break to theTokara Strait depends mainly on the eddies arriving from northeast of Okinawa Island. The combination ofPDO-related signals and eddy-related signals determines the interannual variations of the Kuroshio trans-port in the ECS and sufficient attention must be paid to the spatial dependence of the Kuroshio transport inthe ECS on eddies.

1. Introduction

The sea surface height (SSH) and Kuroshio transport in the East China Sea (ECS) is known to respond well toa basin-scale interannual oscillation, i.e., the Pacific Decadal Oscillation (PDO) [Mantua et al., 1997] that maybe used as a proxy for large-scale wind stresses curl forcing [Andres et al., 2009]. For example, Han andHuang [2008] reported a negative correlation between the interannual variations of SSH inside the ECS andPDO index by analyzing 11 years (October 1992 to July 2002) of altimetry data and sea level data at tidalgauges, and suggested a positive correlation between Kuroshio transport in the ECS and PDO index by fol-lowing an idea given by Gordon and Giulivi [2004]. Andres et al. [2009] confirmed this positive correlationusing the satellite altimetry-derived Kuroshio transport midway along the shelf break in the ECS.

On the other hand, mesoscale eddies also affect the Kuroshio transport in the ECS. Yang et al. [1999]reported the arrival of cyclonic and anticyclonic mesoscale eddies southeast of Taiwan with an interval of�100 days and their significant impacts on the Kuroshio transport east of Taiwan. Zhang et al. [2001] esti-mated the Kuroshio transport east of Taiwan from 20 months of mooring data and found a period of 100days in the variations of Kuroshio transport due to arrival of mesoscale eddies from the interior ocean.Hwang et al. [2004] found that most mesoscale eddies arriving from southeast of Taiwan had their origin inthe Subtropical Countercurrent (STCC) region and propagated westward along a zonal band near 22�N–24�N. In addition, mesoscale eddies also affect the Kuroshio transport through the Tokara Strait. Ichikawa[2001] reported that an eddy-related signal existing south of Okinawa 60 days prior may move to the Kur-oshio region in the ECS by passing through the Kerama Gap and finally arrive at the Tokara Strait.

With accumulation of altimetry data, the interannual variability of mesoscale eddies themselves was gradu-ally realized. Using the 16 years (1993–2008) of altimetry data, Qiu and Chen [2010] reported enhancededdy activity in 1996–1998 and 2003–2008 and weakened eddy activity in 1993–1995 and 1999–2002 in

Special Section:Western Pacific Ocean Circula-tion and Climate

Key Points:� ECS-Kuroshio transport from 1993 to

2012 is given by a reanalysis(JCOPE2)� ECS-Kuroshio transport is associated

with both PDO index and mesoscaleeddies� Mesoscale eddies southeast of

Taiwan were intensified after 2003

Correspondence to:X. Guo,guoxinyu@sci.ehime-u.ac.jp

Citation:Soeyanto, E., X. Guo, J. Ono, and Y.Miyazawa (2014), Interannualvariations of Kuroshio transport in theEast China Sea and its relation to thePacific Decadal Oscillation andmesoscale eddies, J. Geophys. Res.Oceans, 119, 3595–3616, doi:10.1002/2013JC009529.

Received 21 OCT 2013

Accepted 21 MAY 2014

Accepted article online 24 MAY 2014

Published online 9 JUN 2014

SOEYANTO ET AL. VC 2014. American Geophysical Union. All Rights Reserved. 3595

Journal of Geophysical Research: Oceans

PUBLICATIONS

the STCC region within 18�N–25�N in the western North Pacific. Recently, Qiu and Chen [2013] againreported the interannual modulation of mesoscale eddy activity and attributed it to the change in surfaceheat flux forcing. Since the mesoscale eddies in the STCC band eventually arrive at the Kuroshio region,their interannual variability is expected to result in interannual variation in the Kuroshio transport. Actually,Chang and Oey [2011] suggested that the eddy activity in the STCC region determines both the seasonaland interannual variations of the Kuroshio transport east of Taiwan.

As stated above, previous studies on the interannual variability in the Kuroshio transport in the ECS werebased on SSH data from altimetry and tide gauges or hydrographic data along one or two transects. Theinfluence of mesoscale eddies on the Kuroshio transport in the ECS has also been realized but the analysiswas carried out separately for the area east of Taiwan [Yang et al., 1999; Zhang et al., 2001; Chang and Oey,2011] and for the Tokara Strait [Ichikawa, 2001]. The length of the shelf break in the ECS is on the order of1000 km, which is much larger than the size of mesoscale eddies from the interior ocean [Yang et al., 2013].The pathway of eddies moving from south of Okinawa to the Tokara Strait [Ichikawa, 2001] is far from thearea east of Taiwan, indicating that the Kuroshio transport east of Taiwan and Tokara Strait are probablyaffected by different eddies. At present, we still do not have an understanding of the spatial variations inthe influences of eddies on the Kuroshio transport along its pathway inside the ECS.

In this study, we examine the interannual variations of Kuroshio transport in the ECS by using a reanalysisproduct from 1993 to 2012 produced by a data-assimilative ocean model that uses all the available observa-tion data and provides an optimum estimation for the Kuroshio region [Miyazawa et al., 2009]. We have twospecific purposes. The first is to revisit the correlation between interannual variability of the SSH in the ECSand PDO index reported by Han and Huang [2008] and the correlation between interannual variability ofKuroshio transport in the ECS and PDO index reported by Andres et al. [2009]. Our analysis shows that thepreviously reported correlations were not sustained throughout our analysis period (1993–2012). The sec-ond is to elucidate the influence of mesoscale eddies on the interannual variability of Kuroshio transport inthe ECS. Our analysis shows that the mesoscale eddies southeast of Taiwan strongly affect the Kuroshiotransport from east of Taiwan to midway along the shelf break (upstream region), while the mesoscaleeddies northeast of Okinawa strongly affect the Kuroshio transport from the midway along the shelf breakto the Tokara Strait (downstream region) in the ECS.

This paper is organized as follows. In section 2, the data-assimilative ocean model and analysis method aredescribed. Section 3 describes the comparison of model results with available observations of the Kuroshiotransport in the ECS and presents the correlation of Kuroshio transport in the ECS calculated from modelresults with PDO index. Section 4 demonstrates the influence of mesoscale eddies on the Kuroshio transportalong its pathway in the ECS with special attention on the regional dependence of the Kuroshio transporton eddies. Section 5 summarizes this study.

2. Reanalysis Product and Analysis Method

2.1. Data-Assimilative Ocean ModelThe reanalysis product used in this study was obtained using a data-assimilative ocean model developed inJapan Coastal Ocean Predictability Experiments 2 (JCOPE2) [Miyazawa et al., 2009]. The JCOPE2 model isbased on the Princeton Ocean Model with generalized coordinate system (POMgcs) [Mellor et al., 2002]. Itincludes two submodels that are connected by a one-way nesting method [Guo et al., 2003]. The innermodel (10.5�N–62�N, 108�E–180�E) has a horizontal grid interval of 1/12� in both meridional and zonaldirections and 46 levels in the vertical. The outer model (30�S–62�N, 100�E–90�W) has a horizontal gridinterval of 1/4� in both meridional and zonal directions and 21 levels in the vertical. The resolution of theinner model may directly resolve mesoscale eddies and has previously been used to study the interactionbetween mesoscale eddies and the Kuroshio large meandering south of Japan [Miyazawa et al., 2004].

Data assimilation in the JCOPE2 model uses all available observational data: (1) satellite altimetry data con-sisting of the TOPEX/Poseidon, ERS-1, Jason-1, and Geosat Follow-On for the SSH anomaly; (2) sea surfacetemperature obtained from the Advanced Very High Resolution Radiometer/Multi-Channel Sea SurfaceTemperature (AVHRR/MCSST) products; and (3) vertical profiles of temperature and salinity obtained fromthe in situ data archive of the Global Temperature-Salinity Profile Program (GTSPP). The details of the assim-ilation method and parameters used in it have been described by Miyazawa et al. [2009]. The present

Journal of Geophysical Research: Oceans 10.1002/2013JC009529

SOEYANTO ET AL. VC 2014. American Geophysical Union. All Rights Reserved. 3596

version of the reanalysis (FRA-JCOPE2) additionally assimilates the in situ data from the World Ocean Data-base (WOD) 2009 (http://www.nodc.noaa.gov/OC5/WOD09/pr_wod09.html) and an archive compiling thedata obtained around the Japan coasts provided by the Fishery Research Agency, Japan.

After a spin-up of 15 and 5 years for the outer low-resolution model and inner high-resolution model,respectively, reanalysis calculations from November 1992 to June 2013 were carried out for both modelswith the data assimilation and the external forcing including wind stress and heat flux fields calculatedfrom the six hourly National Centers for Environmental Prediction/National Center for Atmospheric Research(NCEP/NCAR) reanalysis data [Kalnay et al., 1996] using bulk formulae [Kagimoto et al., 2008], and sea surfacesalinity restored to the monthly mean climatology data [Conkright et al., 2002] with a time scale of 30 days.As the product of this reanalysis calculation, we saved the resultant SSH, two horizontal velocity compo-nents, water temperature, and salinity at 46 sigma levels at a 1 day interval.

2.2. Data Analysis MethodThe raw reanalysis data for this study are daily SSH and two horizontal velocity components from 4 Novem-ber 1992 to 30 June 2013. The procedures for extracting interannual signals are the following: (1) removingthe mean field and annual cycle from the daily data; and (2) applying an 81 day running mean to the dailydata after step 1. The annual cycle was represented by a cosine function with a period of 1 year. Its ampli-tude and phase were obtained by harmonic analysis of the daily raw data. After these two steps, weobtained interannual components of SSH and velocity with 1 day interval from 1 January 1993 to 31December 2012. This data set is referred to here as daily interannual components of SSH and velocity.

The SSHs at two grid points (Figure 1a) are presented in Figure 2 to demonstrate our data processing proce-dures. One grid point (Station CES in Figure 1a) is inside the ECS, i.e., on the shelf side of the Kuroshio whilethe other grid point (station PAC in Figure 1a) is on the ocean side of the Kuroshio. The daily raw data ofSSH at station CES (gray line in Figure 2a) have a mean value of 20.12 m and demonstrate an annual cyclewith an amplitude of 0.11 m (Figure 2a). At this station, the annual variation is a dominant component. The81 day running mean effectively removes the short-term variations (black line in Figure 2a) to produce thedaily interannual component (green line in Figure 2a). At station PAC, the mean SSH is 0.55 m and the

Figure 1. (a) Rainbow contours show mean sea surface height (SSH, m) of the JCOPE2 reanalysis from January 1993 to December 2012. Black thick lines A0, B0, C0, E0, F0, G0, and K1denote transects through which the Kuroshio transport was calculated. Two blue triangles, CES and PAC, denote the grid points whose SSH are presented in Figures 2 and 5. Two graysquares (ET, EO) denote the area used for the area-averaged vorticity given in Figure 15 and eddy kinetic energy in Figures 16a and 16b. (b) Standard deviation (m) of monthly interan-nual component (MIC) of SSH over entire model domain for January 1993 to December 2012.

Journal of Geophysical Research: Oceans 10.1002/2013JC009529

SOEYANTO ET AL. VC 2014. American Geophysical Union. All Rights Reserved. 3597

amplitude of the annual cycle is only 0.05 m (Figure 2b). The interannual signals are the dominant compo-nent in the variations of SSH there (green line in Figure 2b).

Following previous studies [Han and Huang, 2008; Andres et al., 2009], we also use the PDO index to repre-sent interannual variations in basin-scale wind stress curl forcing. The monthly mean PDO index wasacquired from the website of the Joint Institute for the Study of the Atmosphere and Ocean (JISAO) at theUniversity of Washington (http://jisao.washington.edu/pdo/PDO.latest). For comparison with monthly meanPDO index, we calculated the monthly mean from daily interannual components of SSH and velocity andhereafter refer to the month mean as the monthly interannual component (MIC). To separate the interan-nual variations that persist for less or more than 1 year, we applied a 13 month running mean to MIC of SSHand velocity and obtained their long-term monthly interannual component (LMIC). The difference betweenMIC and LMIC is defined as the short-term monthly interannual component (SMIC).

The significance of a correlation coefficient is examined by a criterion based on effective degree of freedom[Emery and Thomson, 2001]. The criterion is denoted by r0ðN; aÞ, where N is effective degree of freedom anda is significance level.

3. Results

3.1. Comparison of Interannual Variations of the Kuroshio Transport in the ECS Between the ModelResults and ObservationsThe dense contours in mean SSH from January 1993 to December 2012 (Figure 1a) indicate that the JCOPE2model reproduces a well-known path of the Kuroshio that starts from east of Luzon Island, passes the LuzonStrait, east of Taiwan, the ECS Shelf break, and south of Japan, and ends east of Japan around 35�N. Themean SSH is negative inside the ECS (20.2 to 20.1 m) but positive offshore of the Kuroshio (0.1–0.6 m).

Based on measured sea level difference between tide gauges of Keelung and Ishigaki (two ends of transectA0 in Figure 1a), Chang and Oey [2011] reported the interannual variations of Kuroshio transport east of

Figure 2. (a) Daily sea surface height (SSH, gray line), composition of mean and annual cycle of SSH (dash line), daily SSH anomaly afterremoving the mean and annual cycle (black thick line), and 81 day running mean of daily SSH anomaly (green line) at station CES. ‘‘Mean’’is mean value of SSH over entire period and ‘‘Amp’’ is amplitude of annual cycle. (b) The same as Figure 2a but for station PAC.

Journal of Geophysical Research: Oceans 10.1002/2013JC009529

SOEYANTO ET AL. VC 2014. American Geophysical Union. All Rights Reserved. 3598

Taiwan. Following their idea, we extracted the sea level difference between Keelung and Ishigaki from thereanalysis data of JCOPE2 and calculated the Kuroshio transport through transect A0 that connects Keelungand Ishigaki (Figure 1a). The correlation coefficient between MIC of sea level difference between Keelungand Ishigaki (red line in Figure 3a) and MIC of the Kuroshio transport through transect A0 (blue line in Figure3a) is 0.67 (>r0ð144; 95%Þ50:16), close to the value (0.72–0.78) from observations [Johns et al., 2001]. Thecorrelation coefficients between measured and modeled sea level differences between Keelung and Ishi-gaki for 1993–2008 are 0.42 (>r0ð101; 95%Þ50:20) for the 90 day low-passed data (Figure 3a) and 0.55(>r0ð18; 95%Þ50:46) for the 360 day low-passed data (Figure 3b).

Wei et al. [2013] reported the interannual variations of Kuroshio transport at PN line (equivalent to transectE0 in Figure 1a) that was calculated from in situ hydrographic data collected from 1955 to 2010 with aninterval of 6 or 3 months by the Japan Meteorological Agency. Although they reported the Kuroshio trans-port based on season of every year, the exact observation date for each cruise can be found in a report ofthe Japan Meteorological Agency. The direct comparison between the transport through transect E0 calcu-lated from the reanalysis data of JCOPE2 at the exact cruise date (black dots in Figure 4a) and thosereported by Wei et al. [2013] (green dots in Figure 4a) gives a correlation coefficient of 0.49(>r0ð68; 95%Þ50:24). Assuming that the Kuroshio transport reported by Wei et al. [2013] represented a sea-sonal mean, its correlation coefficient to the corresponding seasonal mean of model results (red dots in Fig-ure 4a) is 0.35 (>r0ð66; 95%Þ50:24). Apparently, the Kuroshio transport calculated from cruise data by Weiet al. [2013] includes both seasonal mean and high-frequency variations. The high-frequency variations inthe Kuroshio transport, which are demonstrated by the daily Kuroshio transport calculated from the reanaly-sis data of JCOPE2 (gray line in Figure 4a), act as a noise to the seasonal mean in the cruise data. This prob-lem becomes more serious for the case of yearly means (Figure 4b).

Based on the sea level difference across the Kuroshio and its relation with Kuroshio transport estimatedfrom in situ data [Andres et al., 2008], Andres et al. [2009] reported the yearly mean of Kuroshio transport cal-culated from satellite altimetry data with an interval of approximately 10 days. A comparison of their yearlymean (blue line in Figure 4b) and the yearly mean calculated from four or five data in the same year givenby Wei et al. [2013] (green line in Figure 4b) helps to understand the effects of high-frequency variations on

Figure 3. (a) Sum of annual cycle component and monthly interannual component of volume transport (Sv) through transect A0 (blueline), that of sea level difference (cm) across transect A0 (red line) from JCOPE2 reanalysis, 90 day low pass of sea level difference (cm)across the East Taiwan Channel (green line) [Chang and Oey, 2011]. (b) Long-term monthly interannual component of volume transport(Sv) through transect A0 (blue line), that of sea level difference (cm) across transect A0 (red line) from JCOPE2 reanalysis, 360 day low passof sea level difference (cm) across the East Taiwan Channel (green line) [Chang and Oey, 2011].

Journal of Geophysical Research: Oceans 10.1002/2013JC009529

SOEYANTO ET AL. VC 2014. American Geophysical Union. All Rights Reserved. 3599

the yearly mean of Kuroshio transport. In addition to these two data sets, we also calculated two yearlymeans from the reanalysis of JCOPE2: one calculated from four or five data as those in Wei et al. [2013](black line in Figure 4b, referred as pseudo yearly mean hereafter) and the other calculated from 365 or 366daily data within 1 year (red line in Figure 4b).

There are some consistent variations among the four time series shown in Figure 4b. For example, all showa positive anomaly from 1997 to 1999 and a negative anomaly from 2007 to 2008. Meanwhile, there arealso many inconsistent points. The yearly mean of Wei et al. [2013] (green line in Figure 4b) and the pseudoyearly mean of JCOPE2 (black line in Figure 4b) presents a large interannual variation range throughout theentire period. However, the yearly mean of Andres et al. [2009] (blue line in Figure 4b) and that of JCOPE2(red line in Figure 4b) show a reduction in variation range after 2003. The large variation range after 2003given by Wei et al. [2013] and pseudo yearly mean of JCOPE2 is likely caused by insufficient sampling num-ber and the presence of high-frequency variations in the Kuroshio transport. When the sampling interval isshortened to 10 days [Andres et al., 2009], the high-frequency variations in the Kuroshio transport are fil-tered out in its yearly mean. Therefore, a reasonable yearly mean must be based on data with a sufficientlyshort sampling interval.

3.2. Interannual Variations in SSH and Its Relation With the PDO IndexA general idea of the spatial variation in the interannual magnitude of SSH variations can be obtained fromthe distribution of standard deviation of its MIC (Figure 1b). Two well-defined zonal bands appear south ofJapan to the longitude of the Kuroshio Extension region along �34�N and southeast of Taiwan along �22�N.The northern band is related to the meandering of the Kuroshio path [e.g., Miyazawa et al., 2008] and the mes-oscale eddies separated from the Kuroshio Extension [e.g., Ebuchi and Hanawa, 2001]. The southern band isassociated with the eddy activity in the STCC region [e.g., Qiu and Chen, 2010]. The standard deviation is smallinside the ECS. As shown in Figure 2a, the major components in the variation of sea level inside the ECS areannual and short-term variations. However, both of these have been excluded in the MIC.

The expected negative correlation between the MIC of SSH at station CES and monthly PDO index is not sig-nificant and the correlation coefficient between the two time series for entire reanalysis period (1993–2012)

Figure 4. (a) Volume transport reported by Wei et al. [2013] with exact observation dates from the Japan Meteorological Agency (greendots), daily anomaly of volume transport through transect E0 (gray line) from JCOPE2 reanalysis, and those corresponding to the exactobservation date from the Japan Meteorological Agency for the volume transport reported by Wei et al. [2013] (black dots), seasonal meancalculated by daily anomaly of volume transport through transect E0 from JCOPE2 reanalysis corresponding to the season used by Weiet al. [2013] for their estimated volume transports (red dots). (b) Yearly anomaly calculated by the volume transport reported by Wei et al.[2013] (green line), by the daily anomaly of JCOPE2 reanalysis on the days corresponding to the volume transport reported by Wei et al.[2013] (black line), by the daily anomaly of JCOPE2 reanalysis every day in a year (red line), and those reported by Andres et al. [2009] (blueline).

Journal of Geophysical Research: Oceans 10.1002/2013JC009529

SOEYANTO ET AL. VC 2014. American Geophysical Union. All Rights Reserved. 3600

is 20.16 (<r0ð112; 95%Þ50:19; Figure 5a). The negative correlation has been reported for the period from1993 to 2002 [Han and Huang, 2008]. For a confirmation of our analysis, we calculated the correlation coeffi-cient for the period from 1993 to 2002 and obtained a value of 20.24 (<r0ð55; 95%Þ50:26; Figure 5a).Therefore, it is likely that the negative correlation between the MIC of SSH at station CES and monthly PDOindex decreased for the entire time series period (1993–2012). At station PAC, the correlation coefficientbetween the MIC of SSH and monthly PDO index is 0.33 (>r0ð53; 95%Þ50:27) for the entire period and 0.32(>r0ð93; 95%Þ50:20) for the period of 1993–2002.

The spatial distribution of correlation between the MIC of SSH and monthly PDO index demonstrates anapparent change inside the ECS and Japan Sea from the period of 1993–2002 to that of 1993–2012 (Figure6). In the period of 1993–2002, there is significant negative correlation in the two marginal seas except forseveral spots in the Japan Sea (Figure 6a). In the period of 1993–2012, the correlation coefficient decreasedover a large area in the two marginal seas (Figure 6b). In the areas outside the two marginal seas, the corre-lation coefficient is generally positive, independent of choice of period. The reduction of correlation coeffi-cient in the two marginal seas suggests that the relationship between the Kuroshio transport inside the ECSand the PDO index probably changes after 2003, which is examined directly from volume transport of Kur-oshio in the ECS in the next subsection.

3.3. Interannual Variations in the Kuroshio Transport and Its Relation With PDO IndexThe Kuroshio transports calculated from the MIC of velocity at transects A0 to G0, which are located alongthe pathway of the Kuroshio and are designed to span the Kuroshio (Figure 1a), present a range of the inter-annual variations larger than 10 Sv (black line in Figures 7a–7h). The standard deviation of the MIC of vol-ume transport at these transects is within a range of 2.17–4.28 Sv, which is significantly larger than theamplitudes of the annual cycle (Table 1).

To understand the cause of the interannual variation of Kuroshio transport in the ECS, we separate the MICof volume transport (black line in Figures 7a–7h) into two parts according to time scales of variation byusing a 13 month running mean. One is a relatively slow variation with a time scale longer than 1 year, i.e.,LMIC of volume transport (blue line in Figures 7a–7h) and the other is relatively fast variation with a timescale shorter than 1 year, i.e., SMIC of volumes transport (Figure 8), which is the difference of MIC and LMICof volume transport.

Figure 5. (a) Monthly interannual component (MIC) of sea surface height (SSH, green line) at station CES and monthly PDO index (blackline). ‘‘r(1993–2002)’’ and ‘‘r(1993–2012)’’ represent correlation coefficient between the two time series for a period of 1993–2002 and a period of1993–2012, respectively. (b) The same as Figure 5a but for station PAC.

Journal of Geophysical Research: Oceans 10.1002/2013JC009529

SOEYANTO ET AL. VC 2014. American Geophysical Union. All Rights Reserved. 3601

Figure 6. Distribution of correlation coefficient between monthly interannual component of sea surface height and monthly PDO index for (a) a period of 1993–2002 and (b) a period of1993–2012.

Figure 7. (a–h) Monthly interannual component of volume transport (black line) and its 13 month running mean (blue line) through transects A0-G0 whose positions are given in Figure 1a.(i and j) Monthly PDO index (black line). Shaded red and green areas represent the time for composite of long-term monthly interannual component of surface velocity shown in Figure 10.

Journal of Geophysical Research: Oceans 10.1002/2013JC009529

SOEYANTO ET AL. VC 2014. American Geophysical Union. All Rights Reserved. 3602

Transects A0, B0, and C0 are located along the upstream half of the Kuroshio in the ECS. The Kuroshio trans-port at these three transects presents a consistent variation in both LMIC (blue line in Figures 7a–7c) andSMIC (Figures 8a–8c), this is also confirmed by the high correlation between any two of them (Table 2). TheLMICs of volume transport at the three transects show a large peak in 1997, small peaks in 2001, 2003, and2008, and troughs in 1994, 2000, and 2002. The standard deviations of LMIC of volume transport at trans-ects A0, B0, and C0 are 2.00, 2.09, and 2.11 Sv in the period of 1993–2002 but 1.58, 1.62, and 1.65 Sv in the

Table 1. Mean and Amplitude of Annual Cycle of Volume Transport Through Transects A0-G0 for the Period From 1993 to 2012a

Transect

A0 B0 C0 K1 D0 E0 F0 G0

Mean value (Sv) 25.88 24.99 27.31 20.18 31.49 33.89 31.86 30.11Amplitude of annual cycle (Sv) 1.45 1.42 1.21 0.48 0.52 0.47 0.90 1.19SD of MIC-VT(Sv)

1993–2002 3.21 3.26 3.38 4.28 3.62 3.34 2.49 2.481993–2012 3.35 3.36 3.49 4.19 3.10 2.90 2.39 2.41

SD of LMIC-VT(Sv)1993–2002 2.00 2.09 2.11 2.30 2.03 1.95 1.37 1.381993–2012 1.58 1.62 1.65 2.04 1.77 1.75 1.33 1.34

SD of SMIC-VT(Sv)1993–2002 2.29 2.24 2.40 3.29 2.62 2.43 1.97 2.031993–2012 2.91 2.92 3.12 3.61 2.29 2.12 1.85 1.91

aStandard deviation (SD) of monthly interannual component of volume transport (MIC-VT), SD of long-term monthly interannualcomponent of volume transport (LMIC-VT), and SD of short-term monthly interannual component of volume transport (SMIC-VT)through transects A0-G0 in the period of 1993–2002 and that of 1993–2012.

Figure 8. Short-term monthly interannual component of volume transport through transects A0-G0. Red and blue crosses in (a) and (h)represent the times when the short-term monthly interannual component of sea surface height and velocity are shown in Figures 11–14.

Journal of Geophysical Research: Oceans 10.1002/2013JC009529

SOEYANTO ET AL. VC 2014. American Geophysical Union. All Rights Reserved. 3603

period of 1993–2012 (Table 1). Apparently, their variation range becomes smaller after 2003 (Figures 7a–7c).On the other hand, the SMIC of volume transport at three transects has a relatively large variation rangefrom 2003 to 2008 (Figures 8a–8c), resulting in an increase of standard deviations of SMIC of volume trans-port at transects A0, B0, and C0 for the period of 1993–2012 (Table 1).

Transects D0, E0, F0, and G0 are located along the downstream half of the Kuroshio in the ECS. The LMIC(blue line in Figures 7e–7h) and SMIC (Figures 8e–8h) of volume transport are spatially consistent at thefour transects and the correlation coefficient between any two of them is also high (Table 2). The reductionin the variation range in the upstream region that the LMIC of volume transport after 2003 is also apparentat the four downstream transects (Figures 7e–7h), but the SMIC of volume transport in the downstreamregion does not show any apparent change throughout the entire period (Figures 8e–8h).

Transect K1 is located at the Kerama gap (Figure 1a), through which the Kuroshio does not directly pass.Both the LMIC and SMIC of volume transport through transect K1 (Figures 7d and 8d, positive value is in thedirection from the Kuroshio region to the interior ocean) shows a positive correlation with the

Table 2. Correlation Coefficients Between a Pair of Transects for (a) Monthly Interannual Components of Volume Transport (MIC-VT), (b)Long-Term Monthly Interannual Components of Volume Transport (LMIC-VT), and (c) Short-Term Monthly Interannual Components ofVolume Transport (SMIC-VT) for the Period of 1993–2002 (Values Inside Left-Lower Triangle) and the Period of 1993–2012 (Values InsideRight-Upper Triangle)a

(a) MIC-VT

1993–2012

A0 B0 C0 K1 D0 E0 F0 G0

1993

----200

2A0 0.99 0.84 0.59 0.21 0.19 0.19 0.18B0 0.99 0.89 0.61 0.25 0.23 0.22 0.21C0 0.81 0.86 0.65 0.35 0.28 0.21 0.20K1 0.48 0.48 0.54 20.47 20.48 20.36 20.35D0 0.27 0.32 0.42 20.53 0.95 0.73 0.71E0 0.30 0.36 0.41 20.50 0.97 0.89 0.87F0 0.29 0.34 0.34 20.38 0.78 0.89 0.99G0 0.29 0.34 0.34 20.38 0.77 0.88 0.99

(b) LMIC-VT

1993–2012

A0 B0 C0 K1 D0 E0 F0 G0

1993

----200

2

A0 0.99 0.91 0.56 0.26 0.24 0.20 0.24B0 0.99 0.95 0.53 0.33 0.31 0.25 0.29C0 0.93 0.95 0.54 0.39 0.35 0.28 0.31K1 0.56 0.53 0.55 20.56 20.56 20.50 20.48D0 0.39 0.45 0.51 20.43 0.97 0.84 0.84E0 0.39 0.45 0.50 20.42 0.98 0.92 0.92F0 0.33 0.39 0.45 20.37 0.86 0.91 0.99G0 0.39 0.44 0.49 20.33 0.87 0.92 0.99

(c) SMIC-VT

1993–2012

A0 B0 C0 K1 D0 E0 F0 G0

1993

----200

2

A0 0.99 0.83 0.66 0.17 0.10 0.10 0.11B0 0.98 0.88 0.70 0.19 0.12 0.12 0.13C0 0.67 0.76 0.75 0.31 0.20 0.16 0.16K1 0.43 0.46 0.56 20.38 20.41 20.29 20.29D0 0.13 0.19 0.35 20.57 0.93 0.69 0.69E0 0.15 0.22 0.32 20.54 0.96 0.88 0.89F0 0.17 0.24 0.28 20.41 0.76 0.90 0.99G0 0.17 0.24 0.27 20.42 0.77 0.90 0.99

aA0-G0 are transects whose positions are given in Figure 1a. In period of 1993–2002, the effective degree of freedom is 45–64 forMIC-VT, 10–15 for LMIC-VT, and 38–87 for SMIC-VT; the corresponding criterion for significance at the 95% level is 0.24–0.29 for MIC-VT,0.52-0.61 for LMIC-VT, and 0.21-0.31 for SMIC-VT. In period of 1993–2012, the effective degree of freedom is 105–145 for MIC-VT, 23–30for LMIC-VT, and 70–161 for SMIC-VT; the corresponding criterion for significance at the 95% level is 0.16–0.19 for MIC-VT, 0.35–0.40 forLMIC-VT, and 0.15–0.19 for SMIC-VT. The significant correlation coefficients are given in bold font.

Journal of Geophysical Research: Oceans 10.1002/2013JC009529

SOEYANTO ET AL. VC 2014. American Geophysical Union. All Rights Reserved. 3604

corresponding component of volume transport through the transects at the upstream region and a nega-tive correlation with the corresponding component of volume transport through the transects at the down-stream region in the ECS (Table 2). Therefore, an increase of Kuroshio transport in the upstream region or areduction of Kuroshio transport in the downstream region is accompanied by an outflow of current fromthe ECS to the Pacific.

As with SSH inside the ECS, the extension of data from the period of 1993–2002 to the period of 1993–2012also reduces the correlation between the MIC of volume transport and the PDO index (Figure 9a). This fea-ture is particularly apparent in the upstream region (transects A0-C0, Figure 9a), where the correlation coef-ficients are above 0.20 in the period of 1993–2002, but decreases to the values less than 0.10 over theentire period (1993–2012). The LMICs of volume transport through transects A0-C0 in the upstream regionsgive a little high correlation coefficients with the PDO index in the period of 1993–2002 (Figure 9b). Theremoval of SMIC from MIC efficiently improves the correlation between LMICs and PDO index in theupstream region but not in the downstream region, suggesting a possibility that the SMIC behaves differ-ently in the upstream region compared to the downstream region.

To examine the cause for the difference in correlation between the Kuroshio transport and PDO index inthe upstream and downstream regions, we made composite maps of LMIC of currents based on the PDOindex (Figures 7i and 7j). We selected the period when the PDO index was larger than its mean value plusone standard deviation in the period of 1993–2002 and in the period of 1993–2012 (red area in Figures 7iand 7j); and the period when the PDO index was smaller than its mean value minus one standard deviationin the period of 1993–2002 and in the period of 1993–2012, respectively (green area in Figures 7i and 7j). A

Figure 9. (a) Correlation coefficients between monthly interannual component of volume transport (MIC-VT) through transects A0-G0 andmonthly PDO index for a period of 1993–2002 and a period of 1993–2012. (b) The same as Figure 9a but for long-term monthly interan-nual component of volume transport (LMIC-VT) and monthly PDO index. A0-G0 refer to the transects whose positions are given in Figure1a. In period of 1993–2002, the effective degree of freedom is 40–49 for MIC-VT and 19–23 for LMIC-VT; the corresponding criterion for sig-nificance at the 95% level is 0.28–0.30 for MIC-VT and 0.41–0.44 for LMIC-VT. In period of 1993–2012, the effective degree of freedom is78–128 for MIC-VT and 33–40 for LMIC-VT; the corresponding criterion for significance at the 95% level is 0.17–0.22 for MIC-VT and 0.31–0.34 for LMIC-VT. The significant correlation coefficients at the 95% level are denoted by red font.

Journal of Geophysical Research: Oceans 10.1002/2013JC009529

SOEYANTO ET AL. VC 2014. American Geophysical Union. All Rights Reserved. 3605

positive correlation between the Kuroshio transport in the ECS and PDO index means a strong Kuroshio dur-ing the period with positive PDO index but a weak Kuroshio during the period with negative PDO index. Inthe anomaly field of currents, i.e., MIC and LMIC, a strong Kuroshio is shown by a current in the same direc-tion as the Kuroshio, while a weak Kuroshio is shown as a current in the opposite direction to the Kuroshio.

Figures 10a and 10c are composites for positive PDO index in the periods of 1993–2002 and 1993–2012,respectively. Both show a current in the same direction as the Kuroshio. The magnitude of current is weakerin Figure 10c than in Figure 10a. Figures 10b and 10d are composites for negative PDO index in the periodsof 1993–2002 and 1993–2012, respectively. The current in the opposite direction to the Kuroshio is not con-sistent over the entire shelf break in both figures: it is strong in the upstream region but weak in the down-stream region; it also becomes weak when the analysis period is extended to 20 years.

The extension of the analysis period from 1993–2002 to 1993–2012 likely weakens the correlation of Kur-oshio transport in the ECS to the PDO index. This weakening is more apparent in the upstream region thanin the downstream region. Since only the SMIC of volume transport in the upstream region is stronger inyears after 2002 than before 2002 (Figure 8), it is possible that the regional difference in the correlation ofMIC or LMIC of volume transport with the PDO index is caused by the processes responsible for the SMIC ofvolume transport. In section 4, we demonstrate that the mesoscale eddies are responsible for the SMIC ofKuroshio transport in the ECS and their influences on the Kuroshio volume transport have a strong regionaldependence.

4. Influence of Mesoscale Eddies on the Kuroshio Transport and Its RegionalDependence

4.1. Eddy-Related Variations in the Kuroshio Transport in the ECSThe correlation between SMICs of volume transport among each pair of transects (Table 2) suggests a spa-tially consistent variation occurring in the upstream region and downstream region. To understand thecauses for the SMIC of volume transport in these two regions, we choose two maximum (December 1995;July 2008) and two minimum (June 1994; October 2008) SMICs of volume transport through transect A0(Figure 8a), as well as two maximum (June 1996; July 2011) and two minimum (October 1994; February2011) SMICs of volume transport through transect G0 (Figure 8h), and present horizontal and vertical distri-butions of the corresponding currents (Figures 11–14).

The maximum (minimum) SMIC of volume transport through transect A0 is closely associated with the pres-ence of an anticyclonic (cyclonic) eddy southeast of Taiwan (Figure 11). The presence of an anticycloniceddy induces a northeastward current along the inner side of the Kuroshio at transect A0 (Figures 11a and11b), while the presence of cyclonic eddy southeast of Taiwan induces a southwestward current along theinner side of the Kuroshio (Figures 11c and 11d). For the same anticyclonic eddies (Figures 11a and 11b) orthe same cyclonic eddies (Figures 11c and 11d), a stronger eddy (Figure 11b or 11d) induces a strongereddy-related current than the corresponding weak one (Figure 11a or 11c). Such an eddy-related current isevident in the upstream region of the Kuroshio, i.e., from east of Taiwan to the midway along the shelfbreak. Consequently, the Kuroshio in the upstream region is intensified (weakened) by the presence of ananticyclonic (cyclonic) eddy southeast of Taiwan. The downstream region of the Kuroshio in the ECS, i.e.,from the midway along the shelf break to the Tokara Strait, does not show a consistent response to theeddies southeast of Taiwan in the four cases shown in Figure 11.

The eddy-related current occupies entire depth of transect A0 and is strong in the upper layer and weak inthe lower layer (Figure 12). The arrival of a stronger eddy (Figure 12b or 12d), regardless of whether theeddy is cyclonic or anticyclonic, induces a stronger SMIC of velocity over the entire transect than that of aweak one (Figure 12a or 12c). This in turn increases the SMIC of volume transport through the transect.Therefore, the larger variation range in the SMIC of volume transport in the upstream region of the Kuroshioin the ECS in the years from 2003 to 2008 (Figures 8a–8c) is likely caused by the intensification of mesoscaleeddies southeast of Taiwan.

The maximum (minimum) SMIC of volume transport through transect G0 is associated with the presence ofan anticyclonic (cyclonic) eddy northeast of Okinawa Island (Figure 13). As an anticyclonic eddy is foundnortheast of Okinawa Island, a northeastward current is also found along the Kuroshio pathway from

Journal of Geophysical Research: Oceans 10.1002/2013JC009529

SOEYANTO ET AL. VC 2014. American Geophysical Union. All Rights Reserved. 3606

transects D0 to G0 (Figures 13a and 13b); as a cyclonic eddy is formed northeast of Okinawa Island, a south-westward current is found along the Kuroshio pathway from transect G0 to D0 (Figures 13c and 13d). Sucheddy-related currents are similar to those in the upstream region of the Kuroshio. However, an essential dif-ference in the eddy-related currents between the upstream region and the downstream region is also appa-rent. In the upstream region, a part of the eddy generally merges into the Kuroshio and actually becomes apart of the Kuroshio (Figure 11); in the downstream region, the eddy is far from the Kuroshio and thereforecannot merge into the Kuroshio main stream (Figure 13). The island chain from Okinawa Island to theTokara Strait is likely a barrier for the eddy northeast of Okinawa Island to approach the Kuroshio mainstream.

The eddy-related current is also strong in the upper layer and weak in the lower layer at transect G0 (Figure14). However, the eddy-related current at transect G0 is not as sensitive as that at transect A0 to the eddystrength. For example, the anticyclonic eddy is stronger in June 1996 (Figure 13a) than in July 2011 (Figure13b) but their corresponding currents at transect G0 are very similar (Figures 14a and 14b). This is also truefor the case of two cyclonic eddies shown in Figure 13.

Figure 10. (a) Composite of long-term monthly interannual component of surface current for the time when the PDO index is larger than its mean plus one standard deviation in aperiod of 1993–2002. (b) The same as Figure 10a but for the time when the PDO index is less than its mean minus one standard deviation in a period of 1993–2002. (c) The same as Fig-ure 10a but for a period of 1993–2012. (d) The same as Figure 10b but for a period of 1993–2012. The exact time for the composite is given in Figures 7i and 7j.

Journal of Geophysical Research: Oceans 10.1002/2013JC009529

SOEYANTO ET AL. VC 2014. American Geophysical Union. All Rights Reserved. 3607

To quantify the relationship between the mesoscale eddies and Kuroshio transport in the ECS, we calcu-lated an area-averaged vorticity from SMIC of surface velocity within two fixed areas, southeast of Taiwan(Zone ET, Figure 1a) and northeast of Okinawa Island (Zone EO, Figure 1a). The vorticity averaged in the twoareas well represents the strength of eddies southeast of Taiwan and northeast of Okinawa Island. Forexample, the two eddies southeast of Taiwan in December 1995 (Figure 11a) and July 2008 (Figure 11b)give two negative peaks of vorticity in Figure 15a. The vorticity magnitude of eddies is larger in July 2008than in December 1995, consistent with the current fields presented in Figures 11a and 11b. The eddiesshown in Figures 11c and 11d, as well as those in Figure 13, are also easily identified in Figures 15a and 15c.

The correlation between the mean vorticity within Zone ET and the Kuroshio transport in the ECS is as highas 20.70 (>r0ð88; 95%Þ50:21) at transect A0 but gradually decreases in magnitude from transect A0 totransect C0 and abruptly becomes insignificant from transect D0 to G0. Apparently, the impact of the meso-scale eddies southeast of Taiwan on the Kuroshio transport is limited to the upstream region of the Kur-oshio in the ECS. The significant negative correlation coefficient at transect K1 hints that the mesoscale

Figure 11. (a and b) Short-term monthly interannual component (SMIC) of surface velocity (arrows) and sea surface height (m, color tone) in the presence of an anticyclonic eddy east ofTaiwan. (c and d) The same as Figures 11a and 11b but for the presence of a cyclonic eddy east of Taiwan. The corresponding SMIC of volume transport through transects A0-G0 are pre-sented in Figure 8.

Journal of Geophysical Research: Oceans 10.1002/2013JC009529

SOEYANTO ET AL. VC 2014. American Geophysical Union. All Rights Reserved. 3608

eddy southeast of Taiwan is not completely absorbed by the Kuroshio main stream and the Kerama Gapprobably acts as a leakage pathway of the eddy from the ECS. This is probably why the impact of eddiessoutheast of Taiwan on the Kuroshio transport cannot be found in downstream region of the Kuroshio inthe ECS.

There is no significant change throughout the entire analysis period in the correlation between the meanvorticity within Zone ET and the Kuroshio transport at transects A0-C0 in the upstream region (Figure 15b).Therefore, the relation between the Kuroshio transport at transects A0-C0 and the mesoscale eddy south-east of Taiwan must be independent of the choice of period and the larger variation range of SMIC of vol-ume transport in the upstream region in from 2003 to 2008 (Figures 8a–8c) is surely caused byintensification of the eddies (Figure 11).

Figure 12. (a and b) Short-term monthly interannual component (SMIC) of velocity (color tone) normal to transect A0 in the presence of an anticyclonic eddy east of Taiwan. (c and d)The same as Figures 12a and 12b but for the presence of a cyclonic eddy east of Taiwan. Positive direction for the velocity normal to transect A0 is northward. Black contours show sumof SMIC and temporal mean of velocity normal to transect A0 over the entire period from 1993 to 2012. The SMIC of volume transport (SMIC-VT) is given inside each plot.

Journal of Geophysical Research: Oceans 10.1002/2013JC009529

SOEYANTO ET AL. VC 2014. American Geophysical Union. All Rights Reserved. 3609

The correlation between the area-averaged vorticity within Zone EO and the Kuroshio transport in the ECSis high at transects G0 and F0 and gradually decreases from the downstream region to the upstream region.The SMIC of volume transport in the downstream region of the Kuroshio in the ECS is affected more by theeddies northeast of Okinawa Island than by those southeast of Taiwan. This conclusion also does notdepend on choice of period.

4.2. Change of Intensity of Mesoscale EddiesAs shown in Figure 8, the SMIC of Kuroshio transport in the upstream region has a larger variation range inthe years from 2003 to 2008. According to the linear correlation between vorticity of mesoscale eddies andKuroshio transport (Figure 15), it is expected that the mesoscale eddies must be stronger during this period.

We use eddy kinetic energy (EKE) to elucidate the variation in eddy intensity. The EKE is defined asEKE5 1

2 u0 2

1v0 2

� �, where u’ and v’ are SMIC of surface velocity at a grid point.

The area-averaged EKE within Zone ET shows an apparent interannual variation, being relatively large in1993, 1995, 1998, 2000, 2004, 2006, 2008, and 2010 (Figure 16a). The largest value of EKE in Zone ET in 2006

Figure 13. The same as Figure 11 but for peaks and troughs of volume transport through transect G0 corresponding to the presence of (a and b) an anticyclonic eddy and (c and d) acyclonic eddy northeast of Okinawa Island.

Journal of Geophysical Research: Oceans 10.1002/2013JC009529

SOEYANTO ET AL. VC 2014. American Geophysical Union. All Rights Reserved. 3610

and 2008 is consistent with the largest variation range of SMIC of Kuroshio transport through transects A0,B0, C0, and K0 in the same year (Figure 8). Based on the EKE, we defined years with strong eddies when theEKE is larger than the sum of temporal mean and standard deviation of EKE over entire period (1993–2012).The temporal mean of the area-averaged EKE in Zone ET is 1:5431022 m2s22 for the years with strongeddies and 0:5531022 m2s22 for the other years, while the standard deviation of the area-averaged EKE inZone ET is 0:2931022 m2s22 for the years with strong eddies and 0:2331022 m2s22 for the other years.Apparently, the EKE southeast of Taiwan has not only a higher mean state but also a higher variability inthe years with strong eddies than in other years.

The area-averaged EKE within Zone EO also shows a large interannual variation, being large in 1995, 1996,2001, 2004, and 2006. Some consistent variations in EKE between Zone ET and Zone EO can be found in theyears such as 2004 and 2006; however, they are not always consistent. The correlation coefficient betweenthe two area-averaged EKEs shown in Figures 16a and 16b is 0.23 (>r0ð130; 95%Þ50:17) for the entireperiod.

Figure 14. The same as Figure 12 but for the velocity normal to transect G0 corresponding to the four cases shown in Figure 13.

Journal of Geophysical Research: Oceans 10.1002/2013JC009529

SOEYANTO ET AL. VC 2014. American Geophysical Union. All Rights Reserved. 3611

Since the minimum EKEs in Zones ET and EO did not change greatly throughout the entire period, theincrease of EKE in the years with strong eddies appears in both its temporal mean and standard deviation.We therefore present only the difference of temporal mean of EKE between the years with strong eddiesand the other years to observe where the eddy intensity increased (Figure 16c).

The enhancement of EKE at Zone ET in the years with strong eddies occurs over a large area southeast ofTaiwan outside the ECS and over the Kuroshio upstream region inside the ECS from northeast of Taiwan tothe midway along the shelf break (Figure 16c). The enhancement of EKE at Zone ET in the years with strongeddies is not found in the Kuroshio downstream region inside the ECS. The change of EKE in the years withstrong eddies is therefore different between upstream and downstream regions of the Kuroshio in the ECS.This difference is consistent with the SMIC of Kuroshio transport that shows an increase of variation rangein the years with strong eddies in the upstream region but almost the same variation range throughout theentire period in the downstream region (Figure 8). On the other hand, the enhancement of EKE at Zone EOin the years with strong eddies occurs southeast of Tokara Strait outside the ECS (Figure 16d). There is noapparent increase of EKE inside the Tokara Strait, suggesting again that the response of Kuroshio transportin the downstream region to the mesoscale eddies has a different mechanism from the upstream regionand is not as sensitive as that in the upstream region (Figures 11 and 13).

When enlarging the domain for Figure 16c, we found that the enhancement of EKE at Zone ET in the yearswith strong eddies reaches to the eastern boundary of the model domain, forming a zonal band within16�N–21�N from the central Pacific to the Kuroshio region covering a meridional band from east of LuzonIsland to east of Taiwan (figure not shown here). This is consistent with the report that most mesoscaleeddies arriving southeast of Taiwan originate in the STCC region [Hwang et al., 2004].

Qiu and Chen [2010] examined the interannual variation of EKE in the STCC band using satellite altimeterdata from 1993 to 2008 and found high eddy activity in some years (1998–1999 and 2003–2008) and loweddy activity in other years (1993–1995 and 1999–2002). Qiu and Chen [2013] reported concurrent

Figure 15. (a) Spatially averaged vorticity in Zone ET (122�E–126�E, 21�N–25�N, Figure 1a). The vorticity at each grid point within Zone ET was calculated from SMIC of surface velocity.(b) Correlation coefficient between the spatially averaged vorticity in Zone ET and SMIC of volume transport through transects A0-G0 in a period of 1993–2002 and a period of 1993–2012. (c) The same as Figure 15a but for Zone EO (128�E–132�E, 26�N–29�N, Figure 1a). (d) The same as Figure 15b but for the spatially averaged vorticity in Zone EO.

Journal of Geophysical Research: Oceans 10.1002/2013JC009529

SOEYANTO ET AL. VC 2014. American Geophysical Union. All Rights Reserved. 3612

occurrence of interannual variations of EKE over a broad band (18�N–28�N) and related this to the variationsin surface heat flux forcing. If we chose the same areas as those used by Qiu and Chen [2010, 2013], we canobtain the same interannual variations in the EKE from JCOPE2 reanalysis (figures not shown here) as thosereported by Qiu and Chen [2010, 2013]. This is not surprising because the data assimilation in the JCOPE2included the satellite altimeter data used by Qiu and Chen [2010, 2013].

4.3. Different Origin of Eddies Southeast of Taiwan and Northeast of Okinawa Island and DifferentMechanism for Them to Affect the Kuroshio in the ECSA lag correlation between the daily interannual component of volume transport through transect A0 andthat of SSH over the model domain provides the trajectory of eddies that strongly affect the Kuroshio trans-port east of Taiwan (Figure 17a). As time goes back, i.e., with a negative lag in Figure 17a, the area of highcorrelation moves eastward from the area east of Taiwan along 22�N–24�N. This means that eddies propa-gating westward along 22�N–24�N eventually arrive east of Taiwan and affect the Kuroshio transport eastof Taiwan. The propagation speed of this region of high lagged correlation is approximately 0.11 m s21. Thefeatures on the eddy propagation are very similar to the analysis by Hwang et al. [2004].

In contrast, the lag correlation between the daily interannual component of volume transport throughtransect G0 and that of SSH over the model domain does not suggests an apparent propagation pathwayof eddies found northeast of Okinawa Island (Figure 17b). The correlation coefficient is high locally north-east of Okinawa Island and is sustained for only 1 to 2 months. Ichikawa [2001] also reported a short lag

Figure 16. (a) Spatially averaged EKE for Zone ET (black line) calculated from short-term monthly interannual component (SMIC) of surface velocity. Shaded area represents the time forstrong eddies when the EKE is larger than the sum of its mean and standard deviation over entire period (1993–2012). (b) The same as Figure 16a but for Zone EO. (c) Difference of EKEbetween its temporal mean over the time with strong eddies with shaded area in Figure 16a and that over the other time without shaded area in Figure 16a. (d) The same as Figure 16cbut for Zone EO. The squares enclosed by gray lines in Figures 16c and 16d are for spatially averaged EKE shown in Figures 16a and 16b, respectively.

Journal of Geophysical Research: Oceans 10.1002/2013JC009529

SOEYANTO ET AL. VC 2014. American Geophysical Union. All Rights Reserved. 3613

time (1–2 months) with a significant correlation coefficient between the volume transport through theTokara Strait and SSH northeast of Okinawa Island. The locally high correlation coefficient, as well as a shortlife time, suggests a local intensification mechanism of eddy activity northeast of Okinawa Island. The anima-tion of SSH in this area actually shows combination of two or three eddies from different directions. Appa-rently, the eddies northeast of Okinawa Island have a different origin from the eddies southeast of Taiwan.

As shown in Figures 11 and 13, the physical processes involving the influence of mesoscale eddies south-east of Taiwan and that of the mesoscale eddies northeast of Okinawa Island on the Kuroshio are likely dif-ferent. The eddies southeast of Taiwan may partly merge into the Kuroshio main stream but the eddiesnortheast of Okinawa Island may not. Andres and Cenedese [2013] proposed a mechanism for the eddiesnortheast of Okinawa to remotely affect the Kuroshio main stream in the ECS through interaction betweenan eddy and an island. The relationship between the Kuroshio transport in its downstream region in theECS and eddies northeast of Okinawa Island given by this study is essentially consistent with the mecha-nism given by Andres and Cenedese [2013].

5. Summary

Using the JCOPE2 reanalysis from the past two decades, we demonstrated that both the PDO index andmesoscale eddies are closely related to the interannual variation of Kuroshio transport in the ECS. The

Figure 17. (a) Correlation coefficient between daily interannual component of volume transport through transect A0 and that of SSH at all the grid points ‘‘lag’’ days ago (lag 5 0–150).(b) The same as Figure 17a but for daily interannual component of volume transport through transect G0.

Journal of Geophysical Research: Oceans 10.1002/2013JC009529

SOEYANTO ET AL. VC 2014. American Geophysical Union. All Rights Reserved. 3614

positive correlation between the PDO index and the interannual variation of Kuroshio transport in the ECSis confirmed in a period from 1993 to 2002. However, this relation became weak as we extended the analy-sis to a period from 1993 to 2012.

Eddies southeast of Taiwan strongly affect the Kuroshio transport in the upstream region in the ECS andeddies northeast of Okinawa Island strongly affect the Kuroshio transport in the downstream region in theECS. The Kuroshio transport in the upstream region in the ECS is sensitive to the strength of the eddy south-east of Taiwan, while that in the downstream region is not sensitive to the strength of the eddy northeastof Okinawa Island. The merging of a part of the eddy east of Taiwan with the Kuroshio is the cause for a sen-sitive response of Kuroshio transport in the upstream region to the eddy strength southeast of Taiwan. Themerging of the eddy with the Kuroshio cannot occur in the downstream region because of the presence ofislands between Okinawa Island and Tokara Strait. The remote impact of the eddy on the Kuroshio transportin the downstream region is the cause of insensitive response of Kuroshio transport in the downstreamregion to the eddy strength northeast of Okinawa Island.

The PDO index showed a larger variation range in the period before 2002 (>2.0) than after 2002 (<2.0, Fig-ure 7a). Consequently, the PDO-related signal in the interannual variation of Kuroshio transport is strongerin the period before 2002 than after 2002. The eddies southeast of Taiwan and those northeast of OkinawaIsland were stronger in the period of 2003–2008 than in other years. As a result of the sensitive response tothe eddies southeast of Taiwan, the eddy-related variation of Kuroshio transport in the upstream region inthe ECS was larger during years of 2003–2008 than in other years; as a result of insensitive response to theeddies northeast of Okinawa Island, the eddy-related variation range of Kuroshio transport in the down-stream region in the ECS was approximately the same throughout the entire period. The intensification ofthe eddies southeast of Taiwan and the lower PDO index variability are the reasons for the reduction of cor-relation values between the Kuroshio transport and the PDO index in the entire period (Figure 9a).

The intensification of eddies in the STCC region can also be confirmed in the satellite altimeter data [Qiuand Chen, 2010]. Due to assimilation of satellite altimeter data in the JCOPE2 reanalysis, we cannot furtheranalyze the dynamics responsible for the intensification of the eddies in the STCC region [Qiu and Chen,2013]. However, the reanalysis of JCOPE2 provides new insight on the response of the Kuroshio transport inthe ECS to the intensification of eddies in the STCC region and how the response modifies the relationbetween the Kuroshio transport and PDO index. We conclude that the combination of PDO-related signalsand eddy-related signals determines the interannual variations of the Kuroshio transport in the ECS. In thefuture, sufficient attention must be paid to the spatial dependence of Kuroshio transport in the ECS on themesoscale eddies.

ReferencesAndres, M., and C. Cenedese (2013), Laboratory experiments and observations of cyclonic and anticyclonic eddies impinging on an island,

J. Geophys. Res. Oceans, 118, 762–773, doi:10.1002/jgrc.20081.Andres, M., M. Wimbush, J.-H. Park, K.-I. Chang, B.-H. Lim, D. R. Watts, X. Zhu, H. Ichikawa, and W. J. Teague (2008), Observation of Kuroshio

flow variations in the East China Sea, J. Geophys. Res., 113, C05013, doi:10.1029/2007JC004200.Andres, M., J.-H. Park, M. Wimbush, X. Zhu, H. Nakamura, K. Kim, and K.-I. Chang (2009), Manifestation of the Pacific Decadal Oscillation in

the Kuroshio, Geophys. Res. Lett., 36, L16602, doi:10.1029/2009GL039216.Chang, Y.-L., and L.-Y. Oey (2011), Interannual and seasonal variations of Kuroshio transport east of Taiwan inferred from 29 years of tide

gauge data, Geophys. Res. Lett., 38, L08603, doi:10.1029/2011GL047062.Conkright, M. E., et al. (2002), World Ocean Database 2001, vol. 1, Introduction, NOAA Atlas NESDIS 42, vol. 1, edited by S. Levitus, 167 pp.,

U.S. Gov. Print. Off., Washington, D. C.Ebuchi, N., and K. Hanawa (2001), Trajectory of mesoscale eddies in the Kuroshio recirculation region, J. Oceanogr., 57, 471–480.Emery, W. J., and R. E. Thomson (2001), Data Analysis Methods in Physical Oceanography, 2nd and revised edition, 638 pp., Elsevier Sci.,

Amsterdam.Gordon, A. L., and C. F. Giulivi (2004), Pacific decadal oscillation and sea level in the Japan/East Sea, Deep Sea Res., Part I, 51, 653–663, doi:

10.1016/j.dsr.2004.02.005.Guo, X., H. Hukuda, Y. Miyazawa, and T. Yamagata (2003), A triply nested ocean model for simulating the Kuroshio—Roles of horizontal

resolution on JEBAR, J. Phys. Oceanogr., 33, 145–169.Han, G., and W. Huang (2008), Pacific decadal oscillation and sea level variability in the Bohai, Yellow and East China Seas, J. Phys. Ocean-

ogr., 38, 2772–2783, doi:10.1175/2008JPO3885.1.Hwang, C., C.-R. Wu, and R. Kao (2004), TOPEX/Poseidon observations of mesoscale eddies over the Subtropical Countercurrent: Kinematic

characteristic of an anticyclonic eddy and a cyclonic eddy, J. Geophys. Res., 109, C08013, doi:10.1029/2003JC002026.Ichikawa, K. (2001), Variation of the Kuroshio in the Tokara Strait induced by meso-scale eddies, J. Oceanogr., 57, 55–68, 2001.Johns, W. E., T. N. Lee, D. Zhang, and R. Zantopp (2001), The Kuroshio east of Taiwan: Moored transport observations from the WOCE PCM-

1array, J. Phys. Oceanogr., 31, 1031–1053.

AcknowledgmentsThis work was supported by JSPSKAKENHI (22106002) and the NationalBasic Research Program of China(2012CB956004). We thank L.-Y. Oeyfor providing the digital data of sealevel difference between the tidegauges of Keelung and Ishigaki.Comments from two anonymousreviewers were helpful in improvingthe original manuscript.

Journal of Geophysical Research: Oceans 10.1002/2013JC009529

SOEYANTO ET AL. VC 2014. American Geophysical Union. All Rights Reserved. 3615

Kagimoto, T., Y. Miyazawa, X. Guo, and H. Kawajiri (2008), High resolution Kuroshio forecast system—Description and its applications, inHigh Resolution Numerical Modeling of the Atmosphere and Ocean, edited by W. Ohfuchi and K. Hamilton, pp. 209–234, Springer, N. Y.

Kalnay, E., et al. (1996), The NCEP/NCAR 40-year reanalysis project, Bull. Am. Meteorol. Soc., 77, 437–471.Mantua, N. J., S. R. Hare, Y. Zhang, J. M. Wallace, and R. C. Francis (1997), A Pacific interdecadal climate oscillation with impacts on salmon

production, Bull. Am. Meteorol. Soc., 78, 1069–1079.Mellor, G. L., S. Hakkinen, T. Ezer, and R. Patchen (2002), A generalization of a sigma coordinates ocean model and an intercomparison of

model vertical grids, in Ocean Forecasting: Conceptual Basis and Applications, edited by N. Pinardi and J. D. Woods, pp. 55–72, Springer,N. Y.

Miyazawa, Y., X. Guo, and T. Yamagata (2004), Roles of mesoscale eddies in the Kuroshio Paths, J. Phys. Oceanogr., 34, 2203–2222.Miyazawa, Y., T. Kagimoto, X. Guo, and H. Sakuma (2008), The Kuroshio large meander formation in 2004 analyzed by an eddy-resolving

ocean forecast system, J. Geophys. Res., 113, C10015, doi:10.1029/2007JC004226.Miyazawa, Y., R. Zhang, X. Guo, H. Tamura, D. Ambe, J.-S. Lee, A. Okuno, H. Yoshinari, T. Setou, and K. Komatsu (2009), Water mass variability

in the Western North Pacific detected in a 15-Year eddy resolving ocean reanalysis, J. Oceanogr., 65, 737–756.Qiu, B., and S. Chen (2010), Interannual variability of the North Pacific Subtropical Countercurrent and its associated mesoscale eddy field,

J. Phys. Oceanogr., 40, 213–225, doi:10.1175/2009JPO4285.1.Qiu, B., and S. Chen (2013), Concurrent decadal mesoscale eddy modulations in the Western North Pacific subtropical gyre, J. Phys. Ocean-

ogr., 40, 344–358.Wei, Y., D. Huang, and X.-H. Zhu (2013), Interannual to decadal variability of the Kuroshio Current in the East China Sea from 1955 to 2010

as indicated by in-situ hydrographic data, J. Oceanogr., 69, 571–589, doi:10.1007/s10872-013-0193-5.Yang, G., F. Wang, Y. Li, and P. Lin (2013), Mesoscale eddies in the northwestern subtropical Pacific Ocean: Statistical characteristics and

three-dimensional structures, J. Geophys. Res. Oceans, 118, 1906–1925, doi:10.1002/jgrc.20164.Yang, Y., C.-T. Liu, J.-H. Hu, and M. Koga (1999), Taiwan Current (Kuroshio) and impinging eddies, J. Oceanogr., 55, 609–617.Zhang, D., T. N. Lee, and W. E. Johns (2001), The Kuroshio east of Taiwan: Modes of variability and relationship to interior ocean mesoscale

eddies, J. Phys. Oceanogr., 31, 1054–1074.

Journal of Geophysical Research: Oceans 10.1002/2013JC009529

SOEYANTO ET AL. VC 2014. American Geophysical Union. All Rights Reserved. 3616