Abstract

Tide gauge and satellite data reveal an interannual oscillation of the ocean’s thermoclines east of the Philippines and Taiwan, forced by a corresponding oscillation in the wind stress curl. This so-called Philippines–Taiwan Oscillation (PTO) is shown to control the interannual variability of the circulation of the subtropical and tropical western North Pacific. The PTO shares some characteristics of known Pacific indices, for example, Niño-3.4. However, unlike PTO, these other indices explain only portions of the western North Pacific circulation. The reason is because of the nonlinear nature of the forcing in which mesoscale (ocean) eddies play a crucial role. In years of positive PTO, the thermocline east of the Philippines rises while east of Taiwan it deepens. This results in a northward shift of the North Equatorial Current (NEC), increased vertical shear of the Subtropical Countercurrent (STCC)/NEC system, increased eddy activity dominated by warm eddies in the STCC, increased Kuroshio transport off the northeastern coast of Taiwan into the East China Sea, increased westward inflow through Luzon Strait into the South China Sea, and cyclonic circulation and low sea surface height anomalies in the South China Sea. The reverse applies in years of negative PTO.

1. Introduction

East of the Philippines at 12°~13°N the North Equatorial Current (NEC) (see the  appendix for a list of acronyms) splits into the southward-flowing Mindanao Current and the northward-flowing Kuroshio, into roughly equal transports: 30 and 27 Sv, respectively (Sv ≡ 106 m3 s−1). The upper-layer (~1000 m) Kuroshio transport at this birthplace varies with season, controlled by the seasonal changes in the strength and bifurcation latitude of the NEC: it is weak in October (~23 Sv) and strong in March (~32 Sv) (Qu and Lukas 2003; Yaremchuk and Qu 2004). Downstream, the Kuroshio enters the East China Sea at 24°~25°N between Taiwan’s northeastern coast and the island of Ishigaki (Japan). Here, Johns et al. (2001) measured a 20-month mean transport of 21.5 Sv with 100-day fluctuations of ±10 Sv. Interannual variability of the Kuroshio at this important choke point between the subtropical and tropical western North Pacific has recently been studied by Chang and Oey (2011). The authors analyzed 29 years of sea level difference Δηik across the Kuroshio from tide gauge data at Keelung (northeastern Taiwan; 25.15°N, 121.75°E; see Fig. 1b for location) and Ishigaki (24.3°N, 124.15°E) as proxy for the Kuroshio transport and found interannual variations with a range of ±0.1 m [which is equivalent to a range in the proxy transport (Trik) of ±3.5 Sv]. The interannual fluctuations are not directly wind driven through linear Rossby wave and/or Sverdrup dynamics; rather, the Kuroshio strengthens in years of abundant eddies of the Subtropical Countercurrent (STCC), related to the current’s instability state (i.e., propensity to produce eddies) caused by the slow fluctuations of the large-scale wind stress curl in the western North Pacific (Qiu and Chen 2010b, hereafter QC2010b). Chang and Oey (2011) also examined the seasonal fluctuation of Δηik and found this to be much weaker (5~10 times weaker) than the interannual variation; the Δηik is a minimum in May and maximum in November. They argued that the seasonal transport fluctuation is also eddy forced, caused by seasonal variation of STCC instability due to seasonal surface cooling and heating (Qiu 1999). The weaker seasonal amplitude is then because the seasonal time scale is of the same order as the eddy-propagation time scale, and transport-producing eddy signals tend to overlap near the tide gauge stations east of Taiwan. A weak seasonality in the Kuroshio transport also exists at the PN line (28°~29°N) in the northern East China Sea, in which case anticyclonic eddy activity and the interaction between bottom slope and baroclinicity (JEBAR) play an important role (Kagimoto and Yamagata 1997).

Fig. 1.

(a) Colors and contours (interval 0.2, zero contour omitted) are correlations between Δηik and AVISO-SSHA (1993–2008), for values above the 95% significance level; from Chang and Oey (2011). The “eddy zone” is indicated by the parallelogram extended east-southeast off Taiwan where the correlations are generally high. (b) Correlation (color and contour interval 0.2, zero omitted) between 360-day low-pass Δηik and 5° × 5°-averaged × τ0 (ECMWF, leading by 8 months); the 95% significance level ≈ ±0.1; rectangles show regions used for the Taiwan and Philippines poles.

Fig. 1.

(a) Colors and contours (interval 0.2, zero contour omitted) are correlations between Δηik and AVISO-SSHA (1993–2008), for values above the 95% significance level; from Chang and Oey (2011). The “eddy zone” is indicated by the parallelogram extended east-southeast off Taiwan where the correlations are generally high. (b) Correlation (color and contour interval 0.2, zero omitted) between 360-day low-pass Δηik and 5° × 5°-averaged × τ0 (ECMWF, leading by 8 months); the 95% significance level ≈ ±0.1; rectangles show regions used for the Taiwan and Philippines poles.

Given the dominance of the major tropical climate mode, that is, the El Niño–Southern Oscillation in the western North Pacific (e.g., Tozuka et al. 2002; Tozuka and Yamagata 2003), one might presume that the Kuroshio transport proxy Δηik is closely related to ENSO. However, the interannaul Δηik is only weakly correlated with the Niño-3.4 index [Niño-3.4: corr(Δηik,Niño-3.4, −13) = 0.27], or with the El Niño Modoki index [EMI: corr(Δηik,EMI,−14) = 0.38; see  appendix Table A2].1 Clearly, a direct connection with ENSO does not exist, and a fresh look at alternative underlying link(s) that specifically account for the tropical–subtropical and Kuroshio system is necessary.

The Chang and Oey (2011) study suggests that nonlinear eddies exert a dominant influence on the behavior of the Kuroshio east of Taiwan, and understanding them and how they are forced may hold the key to an improved understanding of western North Pacific ocean dynamics.2 Given the ubiquitous abundance of westward propagating eddies and the absence of strong eastward currents in the tropical–subtropical western North Pacific (Chelton et al. 2011), eddies may have a significant impact on the regional circulation. Miyazawa et al. (2008) suggested that the northward propagation of a warm eddy that originated off Taiwan was responsible for the large Kuroshio meander south of Japan in 2004. Sheu et al. (2010) suggested that eddies can influence transport through Luzon Strait into the South China Sea. This work expands the Chang and Oey analysis using also wind, satellite altimetry, and tide gauge data. We make the reasonable assumption that the large-scale wind stress curl (WSC = × τ0; τ0 = kinematic wind stress) plays a fundamental role. However, given that the dominant oceanic response to WSC is not necessarily Rossby wave and/or Sverdrup type, the goal then is to identify key WSC and oceanic patterns that can explain the wide range of circulation processes of the western North Pacific. To incorporate the ocean’s influence, a proxy of the Kuroshio transport as monitored by tide gauge data northeast of Taiwan will be used (Chang and Oey 2011). Then we will show that wind stress curl and thermocline oscillations east of Philippine and Taiwan, which will be referred to as the “Philippines–Taiwan Oscillation” (PTO), account for a broad range of interannual variability of the western tropical and subtropical North Pacific Ocean, including the Kuroshio transport, eddies in the STCC, the bifurcation of the NEC, inflow into the South China Sea (SCS) and its circulation, and possibly the meander states of the Kuroshio south of Japan.

Section 2 describes the data. Section 3 presents the PTO, and section 4 the ocean responses due to PTO. Section 5 contains a discussion, and section 6 is a summary.

2. Data

Details of data sources and processing are given in Chang and Oey (2011); see also the  appendix. Monthly sea level data from 1980 to 2008 at Keelung and Ishigaki—west and east of the Kuroshio respectively off the northeastern coast of Taiwan—are used to compute sea level difference Δηik = SLIsh − SLKee. It is convenient (though not essential) to think of this sea level difference as a measure or proxy of the Kuroshio transport (Trik) into the East China Sea through the channel between Ishigaki and the northeastern coast of Taiwan. Johns et al. (2001) used direct current observations to obtain the linear regression Trik ≈ 35Δηik. Chang and Oey (2011) analyzed the ocean GCM for the OFES simulation (Masumoto et al. 2004) and showed that this (geostrophic) relation holds, though the proportional factor is larger: Trik ≈ 58Δηik (equivalent to a thicker upper layer in the model).3 Additional, longer time data from 1959 to 2008 at Guam (13.43°N, 144.65°E) and Midway (28.22°N, 177.37°W) will be used with the Ishigaki data to infer the tropic–subtropic thermocline oscillations. Satellite sea surface height (SSH) anomaly data (SSHA) from October 1992 to December 2008 on a ⅓° × ⅓° Mercator grid, and mean SSH at ¼° × ¼° are from Archiving, Validation, and Interpretation of Satellite Oceanographic data (AVISO; http://www.aviso.oceanobs.com/). Monthly-mean European Centre for Medium-Range Weather Forecasts wind stress data at 1° × 1° resolution is used to compute the corresponding WSC time series.

3. PTO

In addition to Chang and Oey (2011), two previous works are of fundamental importance to our goal of deriving the PTO. QC2010b demonstrate that interannual modulations of eddy kinetic energy (computed from satellite altimetry data) are caused by variations in the baroclinic instability growth rate of the STCC under different WSC conditions. They show that eddy-rich years lag, by about 9 months, periods of increased Ekman convergence caused by the strengthening of the WSC in the band 18°–25°N, 135°–170°E due to midlatitude westerlies and trade winds. On the other hand, Qiu and Chen (2010a, hereafter QC2010a) show that interannual variations of the NEC bifurcation latitude (NECBL) off the eastern coast of the Philippines depend on the WSC, especially that over the western North Pacific sector 12°–14°N, 140°–170°E. The NECBL shifts northward (southward) following a positive (negative) peak in the WSC, with a lag of approximately 5 months.

The STCC instability state depends on the vertical shear between the eastward-flowing STCC in the surface layer (approximately 100 m thick) and the northern, subsurface portion of the westward-flowing NEC beneath (QC2010b). Qu and Lukas (2003) found that the mean NECBL can migrate as far north as 18°N at 700 m, some 2.4° north of its surface position. It is reasonable to hypothesize then that the interannual variations of the NECBL can affect the state of eddies in the STCC. As seen above, both the STCC instability and NECBL depend on WSC, while the abundance (or scarcity) of STCC eddies directly affect the Kuroshio transport northeast of Taiwan (Chang and Oey 2011). Figure 1a shows the zero-lag correlation corr(Δηik,SSHA) between Δηik and the AVISO SSHA. This shows a positive region east of Taiwan in the STCC “eddy zone.” Chang and Oey show that the period of increased STCC eddy activity is followed by (with a small ~6-month lag) period of increased Kuroshio transport, and corr(Trik, EKEez,−6) = 0.83 (Table 1, EKEez is EKE averaged over the eddy zone).

In view of QC2010b,a and the Chang and Oey (2011) works, we search for a connection between WSC and Δηik. The ECMWF wind and tide gauges at Keelung and Ishigaki constitute two unique, relatively long-term (29 years), datasets that we now exploit to explore dynamics. Figure 1b shows the correlation map between Δηik and WSC: corr(Δηik,WSC,−8). The choice of 8 months (for WSC to lead Δηik) is based on examining several correlation maps, such as Fig. 1b, and choosing one that gives the largest area of |corr| > 0.2 west of the date line and for latitudes between the equator and 30°N. From Table 1, the 8-month lead is of the same order as the 11 months that PTO leads Δηik (defined shortly) and the 6 months that the EKEez leads Δηik, which seems reasonable. Figure 1b shows that WSC and Δηik are significantly correlated with a fairly large correlation coefficient (R > 0.4) in localized zones in the tropics and subtropics (5°~30°N). Two of these significant-correlation zones are particularly interesting: a southern zone with positive correlation east of the Philippines (7°~15°N, 135°~160°E) and a northern zone with negative correlation from east of Taiwan to the central Pacific (17°~28°N, 140°E~180°). We call these the Philippines and Taiwan poles, respectively. Since they are constructed based on two time series (WSC and Δηik), the two poles co-oscillate in time with the sea level difference between Ishigaki and Keelung; we will refer to the oscillation as the Philippines–Taiwan Oscillation (PTO). It is interesting that the Philippines and Taiwan poles are near the regions of WSC identified by QC2010a and QC2010b as relevant to the interannual variations of NECBL and STCC instability, respectively. There are also other significant correlation regions in Fig. 1b. However, by construction, they are part of the same WSC co-oscillations as the PTO. This can be verified by comparing them with the similar patterns seen in the first and second EOF eigenvectors of WSC, particularly south of 40°N ( appendix Fig. A1). The mode-1 EOF correlates with the Pacific decadal oscillation (PDO), corr(EOF1,PDO,3) = 0.81 ( appendix Table A1). The corresponding WSC anomaly has the same sign (negative WSC during the “warm phase” of PDO) extending southwestward from the Taiwan pole into South China Sea, Fig. A1, (Mantua et al. 1997; Zhang et al. 1997).4 This pattern agrees with Fig. 1b that the same (negative) sign of corr(Δηik,WSC) extends also from the Taiwan pole into the South China Sea. It means that years of positive Δηik (hence also years of abundant eddies in the STCC) coincide with years of negative WSC in the South China Sea. At latitudes of the South China Sea, Rossby waves propagate at speeds ≈ 7~9 km day−1 (Chelton and Schlax 1996), and the corresponding Rossby-wave-response time scales in the South China Sea (width ~1000 km) are 120~150 days. Therefore, a quasi-steady response would prevail at interannual time scales, and a negative WSC would tend to produce convergence, southward Sverdrup transport, and an anticyclonic gyre with a western boundary current and a positive SSHA in the interior, which contradicts Fig. 1a that Δηik and the SSHA are actually negatively correlated in the South China Sea. These results suggest that, at interannual time scales, the South China Sea circulation is not controlled by the local WSC. To make progress, it is necessary to quantify PTO.

Table 1.

Correlations/lags (months; >0 if column 1 leads) between column 1 and other columns; the dashes indicate insignificant correlation (at the 95% level). For each correlation, the 95% significance level is computed from 1 − (1 − 0.95)2/(F−1), where F is the degree of freedom calculated as N/τN, N the length of time series, and τN the dot product of the autocovariances of the two time series. Except for AVISO, all time series data are from 1980 to 2008.

Correlations/lags (months; >0 if column 1 leads) between column 1 and other columns; the dashes indicate insignificant correlation (at the 95% level). For each correlation, the 95% significance level is computed from 1 − (1 − 0.95)2/(F−1), where F is the degree of freedom calculated as N/τN, N the length of time series, and τN the dot product of the autocovariances of the two time series. Except for AVISO, all time series data are from 1980 to 2008.
Correlations/lags (months; >0 if column 1 leads) between column 1 and other columns; the dashes indicate insignificant correlation (at the 95% level). For each correlation, the 95% significance level is computed from 1 − (1 − 0.95)2/(F−1), where F is the degree of freedom calculated as N/τN, N the length of time series, and τN the dot product of the autocovariances of the two time series. Except for AVISO, all time series data are from 1980 to 2008.

The PTO index

We define the PTO index as the difference of WSCs averaged over the Philippines and Taiwan poles, respectively:

 
formula

The Taiwan pole is chosen to be (22°–27°N, 155°E–180°), while the Philippines pole is (8°–13°N, 130°–155°E) (Fig. 1b). To check that the result is not sensitive to this particular choice of the regions, we have calculated the PTO using two other choices: (i) the Taiwan pole is moved southwestward to (18°–23°N, 130°–155°E) and (ii) the two poles are calculated over regions where the absolute correlations east of Taiwan and the Philippines in Fig. 2a are > 0.2. The results are found to be insensitive to these choices because the WSC in these regions correlates with one another (figures not shown).

Fig. 2.

(a) JMA (see Fig. 1b for section location) temperature (solid contours) and zonal geostrophic velocity (color, dash is zero value) from QC2010b; locations of the Taiwan and Philippines poles where |corr(Δηik, × τ0)| > 0.2 in Fig. 1b are shown for PTO > 0; (b) seasonal (90-day low pass, gray) and interannual (360-day low pass, red) PTOs for 1980–2008, and seasonal ensemble PTO (dashed black), January–December, with standard error bars indicated only at ends, and PTOSSHA (blue dash-dot; max/min scale = ±0.3 m) = SSHA at south averaged over (8°–13°N, 130°–145°E) minus north averaged over the eddy zone.

Fig. 2.

(a) JMA (see Fig. 1b for section location) temperature (solid contours) and zonal geostrophic velocity (color, dash is zero value) from QC2010b; locations of the Taiwan and Philippines poles where |corr(Δηik, × τ0)| > 0.2 in Fig. 1b are shown for PTO > 0; (b) seasonal (90-day low pass, gray) and interannual (360-day low pass, red) PTOs for 1980–2008, and seasonal ensemble PTO (dashed black), January–December, with standard error bars indicated only at ends, and PTOSSHA (blue dash-dot; max/min scale = ±0.3 m) = SSHA at south averaged over (8°–13°N, 130°–145°E) minus north averaged over the eddy zone.

The Philippines and Taiwan poles oscillate 180° out of phase, that is, they see-saw with each other in time. Physically, Fig. 1b, the × τ0 near the two poles [where |corr(Δηik, × τ0)| > 0.2, say] co-oscillate with the sea level difference Δηik and, from the results presented in Chang and Oey (2011), also with Kuroshio transport northeast of Taiwan, as well as with the abundance of STCC eddies. This connection between wind, Kuroshio transport, and STCC eddies is detailed below; the connection makes the PTO unique and, as we will show, enables it to link various processes in the western North Pacific. Figure 2a shows the two poles displayed in a meridional sectional contour plot of temperature and zonal geostrophic velocity at 137°E. The WSC seesaw between the two poles produces, through Ekman pumping, oscillations in subsurface isopycnals also. The PTO therefore correlates inversely with the tropic–subtropic thermocline seesaw “PTOSSHA = SSHA|Philippines − SSHA|Taiwan.” This inverse relation is seen in Fig. 2b, which shows the PTO and PTOSSHA; the corresponding correlation corr(PTOSSHA,PTO,4) = −0.82. Figure 2b shows also the seasonal PTOseason, computed in the same way as the interannual PTO in Eq. (1) but using the 90-day (instead of 360 day) low-passed WSC time series. The PTOseason is positive (negative) in winter (summer), that is, the Asian monsoon, and will be discussed later for its remarkable analogy with the interannual PTO.

We extend the PTO to the longer 1959–2008 period that is available from the ECMWF dataset. Interannual fluctuations of the PTO are approximately 3 times weaker than seasonal but are also visible for the earlier, shorter period from 1959 to 1980 (Fig. 3a). The PTO tends to be more positive, and its amplitude and period tend to increase from 1976/1977—the time when climate shift occurred (Trenberth 1990)—to approximately 1997/1998. During that period, the easterly trade winds weakened (McPhaden and Zhang 2002) and the tropical Pacific Ocean is biased toward El Niño–like conditions (Vecchi et al. 2006). It will be seen below (Fig. 7) that a positive PTO indeed corresponds to a weakened equatorial trade wind. From approximately 1998 to present, Fig. 3a indicates another shift toward an even stronger amplitude and longer (decadal) PTO oscillation, but the data is too short to ascertain this. QC2010b found a similar trend toward a decadal period for the NEC bifurcation latitude and wind stress curl anomalies averaged over the region 12°–14°N, 140°–170°E (see their Fig. 8).

Fig. 3.

(a) Fifty-year (1959–2008) PTO, black (gray) is the 360-day (90-day) low-passed series, normalized by (bottom left) their own standard deviations are shown. (b) Normalized PTO and ΔηGM [sea level difference between Guam (13.43°N, 144.65°E) and Midway (28.22°N, 177.37°W)] sea levels from tide gauges. (c) Normalized PTO and ΔηGI [sea level difference between Guam (13.43°N, 144.65°E) and Ishigaki (24.3°N, 124.15°E)] sea levels. In (b) and (c), the max/min scale for Δη is ±0.3 m, their maximum lagged correlations and lags are shown in the upper left, and their curves have been shifted to the left by (b) one and (c) three months.

Fig. 3.

(a) Fifty-year (1959–2008) PTO, black (gray) is the 360-day (90-day) low-passed series, normalized by (bottom left) their own standard deviations are shown. (b) Normalized PTO and ΔηGM [sea level difference between Guam (13.43°N, 144.65°E) and Midway (28.22°N, 177.37°W)] sea levels from tide gauges. (c) Normalized PTO and ΔηGI [sea level difference between Guam (13.43°N, 144.65°E) and Ishigaki (24.3°N, 124.15°E)] sea levels. In (b) and (c), the max/min scale for Δη is ±0.3 m, their maximum lagged correlations and lags are shown in the upper left, and their curves have been shifted to the left by (b) one and (c) three months.

We check the inverse relation between tropic–subtropic thermocline and wind stress curl seesaws (i.e., between PTOSSHA and PTO) using extended time series with tide gauge data at Guam (13.43°N, 144.65°E), Midway (28.22°N, 177.37°W), and Ishigaki to compute PTOSSHA. The PTOSSHA is approximated as Guam – Midway and Guam – Ishigaki sea level differences: ΔηGM and ΔηGI, respectively. Figures 3b,c show that ΔηGM and ΔηGI are anticorrelated with the PTO: corr(PTO,ΔηGM,1) = −0.41 and corr(PTO,ΔηGI,3) = −0.58. The period tends to lengthen after 1980 especially for ΔηGI (Fig. 3c) but there is no obvious increase in amplitude. Nonetheless, the anticorrelative connection between PTO and tropic–subtropic seesaw of the thermocline seems clear.

4. Ocean responses to PTO

By construction, PTO correlates with Δηik and corr(PTO,Δηik,11) = 0.57 (Table 1). (We note that Δηik is poorly correlated with WSC averaged over the STCC eddy zone from 120° to 140°E east of Taiwan—see Chang and Oey 2011). Physically, Kuroshio transport Trik tends to increase approximately one year following a peak in PTO when the thermocline off the Philippines is uplifted and that off Taiwan deepens. Since the Kuroshio transport increases following increased STCC eddy activity (lag ≈ 6 months), PTO correlates well with EKEez, and leads it by about 9 months: corr(PTO,EKEez,9) = 0.76 (Table 1). We can confirm that periods of abundant STCC eddies coincide (with some lags ≈ 9 months) with periods of positive PTO, and vice versa for negative PTO. This is shown in Fig. 4, which plots 180-day high-passed SSHA2 from AVISO composited for positive and negative phases of PTO. To account for the slight lag, the AVISO time series has been shifted forward (“to the left”) by 9 months, and the high pass ensures that eddies are included in the composites. The stronger (weaker) eddy activity over the STCC during the positive (negative) phase of PTO is clearly seen in Fig. 4. We now discuss how PTO is related to other circulation processes.

Fig. 4.

Maps of the 180-day high-pass SSHA2 from AVISO data, composited for (top) positive and (bottom) negative PTOs. A weighted-composite method is used to reduce uncertainty in SSHA2 when the values of PTO are small—see the  appendix. Contours are 0.01 and 0.015 m2. A 9-month lead for PTO is used, as suggested by the lead in the correlation between PTO and eddy-zone EKE [corr(PTO, EKE, 9) = 0.76] in Table 1.

Fig. 4.

Maps of the 180-day high-pass SSHA2 from AVISO data, composited for (top) positive and (bottom) negative PTOs. A weighted-composite method is used to reduce uncertainty in SSHA2 when the values of PTO are small—see the  appendix. Contours are 0.01 and 0.015 m2. A 9-month lead for PTO is used, as suggested by the lead in the correlation between PTO and eddy-zone EKE [corr(PTO, EKE, 9) = 0.76] in Table 1.

a. South China Sea response

We pointed out before that the lower SSHA (<0) in the South China Sea when Δηik is positive (Fig. 1a) cannot be locally driven by the corresponding negative WSC anomaly (Fig. 1b). At interannual time scales, the South China Sea is driven in part by the stronger eddy activity in the STCC when PTO > 0, when a negative WSC anomaly extends from the Taiwan pole into the South China Sea, as previously discussed, and a positive WSC anomaly prevails east of Philippines. To demonstrate this, we plot in Fig. 5 maps of the correlation between AVISO SSHA and PTO and between SSHA and EKEez. The SSHA is lagged behind the PTO and the EKEez time series, and the lag (6 months) is taken to be the same as the lag between the tide gauge data Δηik and EKEez (see Table 1). The SSHA–PTO correlation shows opposite signs east of Taiwan and the Philippines (Fig. 5, top). A positive PTO corresponds to cyclonic WSC anomaly east of Philippines and anticyclonic WSC anomaly east of Taiwan, resulting in a lower SSHA off the Philippines and higher SSHA off Taiwan. The higher SSHA east of Taiwan moreover corresponds to stronger STCC eddy activity—that is, higher EKE or SSHA2 (Chang and Oey 2011, also Fig. 4). As a consequence, the SSHA–EKEez correlation shows the same, oppositely signed, dipole structure east of Taiwan and the Philippines (Fig. 5, bottom). However, unlike the positive WSC anomaly east of the Philippines when PTO > 0, the WSC anomalies in the South China Sea and east of Taiwan oscillate in phase, as mentioned previously (see  appendix Fig. A1); that is, they are both negative when PTO >0. Negative WSC is not likely to produce a lower SSHA indicated by the negative correlations seen in Fig. 5 between EKEez and SSHA (and also between PTO and SSHA) in South China Sea. Instead, the correlation map in Fig. 5 (lower) suggests that periods of high STCC eddy activity and lower SSH in the South China Sea coincide (with a small 6-month lag). We argue that the abundance of warm eddies in the STCC when PTO is positive then tends to force (by geostrophy) stronger westward flow through Luzon Strait, hence producing a cyclonic circulation anomaly in the South China Sea with a negative SSHA. Indeed, Table 1 shows that EKEez correlates significantly with the Luzon Strait transport TrLZ: corr(EKEez, TrLZ,4) = 0.53, where

 
formula

positive for westward flow into South China Sea, uAVg = zonal geostrophic velocity from AVISO, yLZ = Luzon, yTW = Taiwan, and HLZ ≈ 165 m to match the estimated mean transport of 3 Sv (Qu 2000).

Fig. 5.

Correlation maps (top) between AVISO SSHA and PTO and (bottom) between SSHA and EKE averaged over the eddy zone east of Taiwan (see Fig. 1a). In both cases, SSHA is lagged by 6 months, the same as for the lag between EKE and Δηik (Table 1). Contours are ±0.2, ±0.4, and ±0.6, etc; dark (light) shade is positive (negative); unshaded areas are not significant at 95% significance level.

Fig. 5.

Correlation maps (top) between AVISO SSHA and PTO and (bottom) between SSHA and EKE averaged over the eddy zone east of Taiwan (see Fig. 1a). In both cases, SSHA is lagged by 6 months, the same as for the lag between EKE and Δηik (Table 1). Contours are ±0.2, ±0.4, and ±0.6, etc; dark (light) shade is positive (negative); unshaded areas are not significant at 95% significance level.

In addition to STCC eddies, Sverdrup dynamics also contributes to flow through the Luzon Strait, and the flow is westward when PTO is positive:

 
formula

where USV (unit m2 s−1) is zonal volume transport per unit latitudinal distance y, and xE is the eastern boundary.5 Table 1 shows that the USV at Luzon Strait also correlates with TrLZ: corr(−USV,TrLZ,2) = 0.41, comparable to but weaker than corr(EKE,TrLZ,4) above.

In summary, both STCC eddies and Sverdrup dynamics contribute to transport through the Luzon Strait, TrLZ, and both are encapsulated in PTO—since PTO contains the ∂( × τ0)/∂y of Eq. (3) and it correlates well with EKE [corr(PTO,EKE,9) = 0.76, Table 1)]. The PTO therefore also correlates significantly with TrLZ: corr(PTO,TrLZ,6) = 0.74 (Table 1). To gain further insights into the Luzon Strait transport, we next examine PTO composites and their relation to NECBL.

1) Composites

Composites of SSHA and geostrophic current anomalies (Figs. 6a,b) show that for PTO > 0, SSHA is generally positive north of 18°~22°N, and negative to the south, and vice versa when PTO < 0. These broad patterns in part reflect the large-scale wind stress curl as shown by its corresponding composites for positive and negative PTO in Figs. 7a,b. The large-scale pattern is particularly clear east of the Philippines where one can see from Fig. 6 that the SSHA composites are relatively smooth. However, east of Taiwan, the SSHA composites (Fig. 6) show eddies that are moreover mostly confined between 122° and 150°E in the STCC, and the eddy zonal extent is notably smaller than that of the WSC (Fig. 7). As can be seen in the composite vectors, Fig. 6 confirms that the STCC eddies play an important role in driving flows through Luzon Strait. These flows result in generally negative SSHAs in the South China Sea, of opposite sign to that over the STCC (Fig. 1a). Centurioni et al. (2004) observed drifters to preferentially enter the South China Sea through Luzon Strait at approximately 20°N. This latitude coincides well with the westward flow driven by the abundance of STCC eddies and the existence of Sverdrup convergence during the positive phase of PTO (Fig. 6a).

Fig. 6.

The AVISO SSHA (shaded if positive and unshaded if negative, m) and geostrophic velocity (vectors, m s−1) composited for (a) positive and (b) negative PTOs with PTO leading by 9 months (Table 1), from 1993 to 2008.

Fig. 6.

The AVISO SSHA (shaded if positive and unshaded if negative, m) and geostrophic velocity (vectors, m s−1) composited for (a) positive and (b) negative PTOs with PTO leading by 9 months (Table 1), from 1993 to 2008.

Fig. 7.

ECMWF anomaly WSC (colors, N m−3) and wind stress (vectors, N m−2) in the North Pacific, composited for (a) positive and (b) negative PTOs from 1980 to 2008.

Fig. 7.

ECMWF anomaly WSC (colors, N m−3) and wind stress (vectors, N m−2) in the North Pacific, composited for (a) positive and (b) negative PTOs from 1980 to 2008.

2) Relation to the NECBL

For positive PTO, the anomaly circulation east of the Philippines is cyclonic and the NECBL shifts northward, and vice versa for negative PTO (Figs. 6a,b). Their correlation corr(PTO,NECBL,3) is 0.73 (Table 1), and both also correlate well with TrLZ. Inflow into the South China Sea increases (TrLZ > 0) when the NEC bifurcation latitude shifts northward: corr(NECBL,TrLZ,0) = 0.78. This relation is the same as for the seasonal time scale, which has been explained in the literature as being due to changes in Kuroshio strength east of Luzon and/or pressure setup across the Luzon Strait by the winter monsoon (Qu 2000; Yaremchuk and Qu 2004, and references therein). At interannual time scales, the dynamics are due to STCC eddies and Sverdrup dynamics [Eq. (3)] in combination with the NECBL. In models that simulate the NECBL, but which do not simulate STCC eddies, the South China Sea response differs from that observed in Fig. 1a and Fig. 6 (see  appendix Figs. A2 and A3). Despite the high corr(NECBL,TrLZ,0) in these models, the simulated TrLZ is too weak, and the NECBL alone is insufficient in reproducing the observed South China Sea response.

3) Luzon Strait transport and comparison with Qu et al. (2004)

Qu et al. (2004) used a near-global OGCM (¼° × ¼° and 55 z levels) driven by 17-yr ECMWF wind stress from 1982 through 1998 to examine the interannual variations of the Luzon Strait transport TrLZ. Their model shows an increased TrLZ during El Niño years when the NECBL shifts northward and the Kuroshio transport off Luzon (near 18°N) weakens (Masumoto and Yamagata 1991). Since El Niño years correspond approximately to the positive phase of PTO (see below), their results are consistent with Fig. 6a (and the above discussions on TrLZ) which indicates a weakened Kuroshio off Luzon by the presence of an anomalously low SSHA when PTO is positive. Equivalently, Fig. 1a shows that, in years when the Kuroshio transport northeast of Taiwan increases (as indicated by the proxy Δηik), the Kuroshio transport off Luzon weakens. Qu et al. (2004) attributed their increased TrLZ to a weakened Kuroshio off Luzon, based on Sheremet’s (2001) idea that the weakened inertia of the Kuroshio enables planetary beta to play a more dominant role in the potential vorticity (PV) balance, which may allow the Kuroshio to “bend” more westward through Luzon Strait. This mechanism may work at the seasonal time scale since the Kuroshio then varies over a large range (≈10 Sv), and Qu et al. found that the maximum TrLZ (i.e., maximum westward flow) lags the minimum Kuroshio by 2 ~ 3 months (see their Fig. 7).6 However, it is not entirely obvious that the mechanism will hold at interannual time scales since the Kuroshio fluctuates over a smaller range (≈3 Sv) and Qu et al. (2004) found that the maximum TrLZ actually leads the minimum Kuroshio (by 4 months, see Qu et al. Table 1, also their Fig. 9). The eddy-forced mechanism in combination with Sverdrup flow induced by the north – south difference in WSC, discussed above, offer an alternative explanation. The Kuroshio off Luzon still plays a role, however, since it serves as a PV barrier whose interannually varying strength during different phases of the PTO may then control the quantity of eddies crossing the Kuroshio into Luzon Strait (Sheu et al. 2010). More research, both observations and models, is clearly required to examine this interesting topic.

b. Seasonal versus interannual variations

Near the surface, currents east of Luzon Strait are dominated by the strong northward-flowing Kuroshio so that the above proposed zonal flow through the Luzon Strait most likely occurs at subsurface. Some evidence of this subsurface zonal flow at the seasonal time scale can be seen in Fig. 15 of Qu and Lukas (2003). Here, we compare responses due to PTO and PTOseason. In both cases, during the positive phase, the NECBL shifts northward and the Luzon Strait inflow into the South China Sea is strongest, and vice versa during the negative phase. [See corr(PTO,NECBL,3) = 0.73 in Table 1; see Qu and Lukas (2003) and Yaremchuk and Qu (2004) for descriptions of the seasonal responses.] The cause for the shift in NECBL for both PTO and PTOseason is the WSC. Despite different time scales between PTO and PTOseason, we argue that the mechanism with which STCC eddies contribute to Luzon (and Kuroshio, see Chang and Oey 2011) transports is actually remarkably similar. Because of seasonal cooling and heating, the mean state of the STCC–NEC system, hence its instability (i.e., propensity to produce eddies) varies, and the STCC EKE attains a maximum in May when PTOseason is negative and a minimum in December when PTOseason is positive (Qiu 1999, see his Fig. 10). The timing is opposite to the interannual PTO when EKE is maximum in positive PTO and minimum in negative PTO. However, the averaged time duration of 4~6 months [see e.g., corr(EKE,TrLZ) and corr(EKE, Trik) in Table 1] for eddies to arrive near Luzon and Taiwan (where they can alter the transports) is the same regardless of the time scales: seasonal and interannual. Therefore, at the seasonal time scale, the time when eddies are most effective is from September to November, which is also when the Luzon Strait transport is observed to increase (see Fig. 14 in Yaremchuk and Qu 2004). Chang and Oey (2011) note that the Kuroshio seasonal transport northeast of Taiwan is also consistent with this eddy-forced mechanism.

c. STCC eddies and NECBL

QC2010b show that Ekman convergence (∂υEk/∂y) over the region corresponding to the Taiwan pole defined here correlates well with EKE: corr(∂υEk/∂y,EKE,9) ≈0.77 (their Fig. 11). Table 1 shows that PTO is similarly effective in describing EKE: corr(PTO,EKE,9) = 0.76. Negative ∂υEk/∂y produces downward displacement of isopycnals in the “bulging” or southern side of the STCC front (near 21°N). A positive PTO additionally measures the upward movement of isopycnals in the updoming portion of the NEC front (near 10°N, Figs. 2a,b). A positive PTO, therefore, steepens both STCC and NEC fronts and shifts the NEC northward. Analogous to the seasonal northward shift of NECBL with depth in winter [i.e., when PTOseason > 0; see Qu and Lukas (2003), Figs. 8 and 16], we hypothesize that the NECBL also shifts more northward with depth when PTO > 0. Some evidence of this can be seen in the QC2010b  Figs. 8 and 9 in which the NEC–STCC system is vertically more sheared in “eddy-rich” years when PTO > 0. This northward, subsurface penetration of the NEC is predominantly due to the Philippines pole. The NEC penetration enhances STCC vertical shear, making the STCC more prone to baroclinic instability. This conclusion is supported by the fact that corr( × τ0|Philippines,EKE,5) is high, ≈0.74 (not shown).

Fig. 8.

Composites (1980–2008) of wind velocity (vectors) and wind stress curl (color, scale is ±2 × 10−8 N m−3) anomalies for the positive phases of various indicated indices and for PTO (from Fig. 7a). The top middle panel shows the regions used to define the indices: (WP) (Wallace and Gutzler 1981), (Niño-3.4) (Trenberth 1997), (PDO) (Mantua et al. 1997; Zhang et al. 1997), and (EMI) (Ashok et al. 2007). A table of maximum correlations/lags (months) between PTO and other indices is shown (positive lag means PTO leads) is shown below the middle panels; it also shows the variables used to define the indices—WSC: wind stress curl for PTO; SST: sea surface temperature for PDO, Niño-3.4, and EMI; and atmospheric pressure for WP.

Fig. 8.

Composites (1980–2008) of wind velocity (vectors) and wind stress curl (color, scale is ±2 × 10−8 N m−3) anomalies for the positive phases of various indicated indices and for PTO (from Fig. 7a). The top middle panel shows the regions used to define the indices: (WP) (Wallace and Gutzler 1981), (Niño-3.4) (Trenberth 1997), (PDO) (Mantua et al. 1997; Zhang et al. 1997), and (EMI) (Ashok et al. 2007). A table of maximum correlations/lags (months) between PTO and other indices is shown (positive lag means PTO leads) is shown below the middle panels; it also shows the variables used to define the indices—WSC: wind stress curl for PTO; SST: sea surface temperature for PDO, Niño-3.4, and EMI; and atmospheric pressure for WP.

Fig. 9.

Schematic summary of the PTO and its effects on the western North Pacific ocean responses. The wind stress curl dipole that defines the PTO is shown as dark-shaded circles: convergence (divergence) for negative (positive) curl. Smaller light-shaded circles with “w” are warm eddies, and arrows northeast of Taiwan (TW) denote the Kuroshio transport between the two black dots representing tide gauges at Keelung and Ishigaki. The mean bifurcation latitude of the NEC near the surface is shown at 13°N by the dotted line, and the westward inflow into the South China Sea through Luzon (LZ) Strait is represented by the dashed-curvy arrows. During the positive phase of PTO, both eddy activity in the STCC east of Taiwan and Sverdrup convergence due to ∂( × τ0)/∂y (i.e., due to the wind stress curl dipole) increase; the NEC also shifts northward. Together, these result in increased westward inflow through Luzon Strait and a cyclonic circulation anomaly in the South China Sea. Increased eddy activity in the STCC also increases the Kuroshio transport northeast of Taiwan. The reverse occurs during the negative phase of PTO. South of Japan, we hypothesize the likelihood of trigger for offshore meander for positive PTO as shown by a cyclonic low SSHA near the coast, and anticyclonic high SSHA for no meander when PTO turns negative.

Fig. 9.

Schematic summary of the PTO and its effects on the western North Pacific ocean responses. The wind stress curl dipole that defines the PTO is shown as dark-shaded circles: convergence (divergence) for negative (positive) curl. Smaller light-shaded circles with “w” are warm eddies, and arrows northeast of Taiwan (TW) denote the Kuroshio transport between the two black dots representing tide gauges at Keelung and Ishigaki. The mean bifurcation latitude of the NEC near the surface is shown at 13°N by the dotted line, and the westward inflow into the South China Sea through Luzon (LZ) Strait is represented by the dashed-curvy arrows. During the positive phase of PTO, both eddy activity in the STCC east of Taiwan and Sverdrup convergence due to ∂( × τ0)/∂y (i.e., due to the wind stress curl dipole) increase; the NEC also shifts northward. Together, these result in increased westward inflow through Luzon Strait and a cyclonic circulation anomaly in the South China Sea. Increased eddy activity in the STCC also increases the Kuroshio transport northeast of Taiwan. The reverse occurs during the negative phase of PTO. South of Japan, we hypothesize the likelihood of trigger for offshore meander for positive PTO as shown by a cyclonic low SSHA near the coast, and anticyclonic high SSHA for no meander when PTO turns negative.

5. Discussion

Since the PTO is based on wind stress curl that exhibits a Pacific-wide, nonlocal pattern, we compare its composite wind patterns (i.e., Figs. 7a,b) with those computed based on other indices such as WP, Niño-3.4, PDO, and EMI (Fig. 8). Figure 8 shows that the PTO shares various features of the other indices, though with subtle differences. West of the date line and south of (approximately) 30°N, the PTO pattern is most similar to EMI and Niño-3.4. However, in the regions east of the Philippines and Taiwan and to the central Pacific, PTO has much stronger and more distinct positive and negative × τ0 that define the Philippines–Taiwan oscillation and, in turn, make it the most relevant in influencing the western North Pacific ocean response. We can estimate its magnitude by computing the Ekman pumping velocities ≈ | × τ0|/f ≈ 0.1~0.2 m day−1, which give in 6 months a vertical excursion of the NEC and STCC isopycnals of about 20~40 m. Since the two poles act opposite to each other (see Fig. 2a), the net contribution to the vertical excursion of the NEC–STCC system is doubled, ≈40~80 m, which is substantial and contributes to eddy generation through baroclinic instability (cf. QC2010b). We can also compute the corresponding Sverdrup meridional transport (anomaly) per unit longitude as | × τ0|/β ≈ 1 m2 s−1, which gives a total meridional transport O(1~2 Sv) over the zonal extent 1000~2000 km. The resulting convergence (for positive PTO) at the general latitudes of Luzon Strait is therefore of O(2~4 Sv), which is also quite substantial for the transport through Luzon Strait. On top of this, one needs to also add the contribution from the STCC eddies. We also note that, as in the PTO, the “−WP” index displays a similar but weaker pattern of negative × τ0 east of Taiwan. This explains why −WP also correlates quite well with EKE (R ≈ 0.53, −WP leads by 10 months, Table A2, cf. QC2010b).

It is also instructive to compare the Philippines pole in PTO with the one for Niño-3.4 (and also for EMI, which has a similar wind stress curl structure as Niño-3.4 east of Philippines). The wind and wind stress curl composites for Niño-3.4 in Fig. 8 are very similar to the Wang et al (2000) El Niño composite [see their Fig. 4a, which the authors explained according to Matsuno (1966) and Gill’s (1980) models of Rossby wave response to heating (or cooling)]. The main difference between these Niño-3.4 patterns and those of the PTO (Figs. 7a or 8) is that, in the case of the PTO, the negative wind stress curl immediately southeast of the Philippines at 12°~13°N reverses sign to become strongly positive, and the south-southwesterlies from the South China Sea to southern Japan for El Niño are replaced by weak southerlies in the South China Sea and westerlies in the East China Sea and over Japan for positive PTO. These fine distinctions produce the Philippines–Taiwan dipole structure in the wind stress curl distributions that are crucial in generating the thermocline oscillations in PTO.

The unique aspect of PTO is the opposite-signed × τ0 dipole (Philippines and Taiwan) that, while it is approximately shared by Niño-3.4 and EMI, the latter indices do not uniformly explain several key aspects of the western North Pacific oceanic processes (see the  appendix and Table A2, and discussions). These processes can have far-reaching influences. The South China Sea throughflow modifies heat and salt transports of the Indonesian throughflow, which is affected by the NECBL through the latter’s influence on the Mindanao Current (Gordon 1986; Qu et al. 2006). Kuroshio transport and eddies affect watermass exchanges with marginal seas (Guo et al. 2006). Eddies and transport may affect Kuroshio meanders south of Japan. Indeed, regressing PTO with SSHA yields an anticyclone–cyclone pair south of Japan, indicative of conditions favorable for nonlarge or large meander patterns (see  appendix Fig. A4 and discussions). The corresponding correlations are ≈ ±0.6, which is significantly more robust than the ±0.2 values obtained using other indices. This interesting aspect of the PTO response is preliminary and clearly requires a much more in-depth study in the future. Through physical–biogeochemical coupling, PTO can affect primary productivity, hence the preservation and survival of fish larvae at interannual time scales. Through air–sea coupling, the PTO may also be related to the interannual shifting of the monsoon trough, hence typhoon intensities and genesis (Chen et al. 2006; Kim et al. 2010). The possible connection of the PTO with these and other processes (e.g. mode waters and their effects on the STCC fronts; Aoki et al. 2002) should be explored in future work.

6. Summary

The PTO is the difference in WSC anomalies east of Philippines and east of Taiwan. At the seasonal time scale, the PTO represents the East Asian monsoon oscillation. At interannual time scales, through Ekman pumping, the PTO also measures the meridional oscillation of isopycnals between approximately 10° and 25°N in the western north Pacific (120°E~180°). A positive PTO causes isopycnals east of the Philippines to rise and east of Taiwan to deepen. It results in a northward shift of the NEC, increased vertical shear of the STCC–NEC system, and eddy activity dominated by warm eddies in the STCC, increased Kuroshio transport northeast of Taiwan, increased westward inflow through Luzon Strait into the South China Sea by eddies and Sverdrup convergence, and cyclonic anomaly circulation with a low sea surface height anomaly in the South China Sea. The reverse applies when the PTO is negative. A schematic summary of the processes is given in Fig. 9.

Acknowledgments

We thank the three anonymous reviewers whose comments helped improve the manuscript. We benefited from discussions with Drs. Bo Qiu and Tang-Dong Qu of the University of Hawaii. We thank the following scientists for providing data: Dr. J. Potemra (ECMWF), Dr. B. Qiu (NECBL), Dr. H. Sasaki (OFES), and Dr. C.-R.Wu (tide gauge). This research was supported by Taiwan’s NSC Grant 100-2119-M-008-036-MY3.

APPENDIX

Acronyms, Data Processing, and Additional Discussions of PTO and Other Indices

a. List of acronyms

EKE Eddy kinetic energy

EKEez EKE averaged over the “eddy zone” east of Taiwan

EMI ENSO Modoki index

EOF Empirical orthogonal function

LZ Luzon

NEC North Equatorial Current

NECBL North Equatorial Current Bifurcation Latitude

Niño-3.4 El Niño index 3.4

PDO Pacific decadal oscillation

PTO Philippines–Taiwan oscillation

PV Potential vorticity

RG Reduced gravity

SCS South China Sea

SSHA Sea surface height anomaly

STCC Subtropical counter current

WP Western Pacific index

WSC Wind stress curl

b. Data and processing

Sea level data at Keelung (25.15°N, 121.75°E) and Ishigaki (24.3°N, 124.15°E), west and east of the Kuroshio, respectively, off the northeastern coast of Taiwan, were downloaded from the University of Hawaii Sea Level Center (http://uhslc.soest.hawaii.edu/). There are 29 years of data overlap, from 1980 to 2008. Data at Ishigaki is 99% complete, and missing data in 2007 (~1 month) and 2008 (~3 months) were filled using Keelung data by the regression: SLIsh = 1.099 + 0.749SLKee; R2 = 0.57, where R = correlation coefficient. Large data gaps exist at Keelung (84% complete) in 1998, 1999, and 2005 (~8 months each), and for 2.5 years from 2001 to July 2003. Data from a nearby station, Gangfeng (24.9°N, 121.85°E; source: the Taiwan Central Weather Bureau), were used to fill the Keelung missing data by the regression: SLKee = 1.14 + 1.083SLGan; R2=0.58. Filled hourly data were low passed using a 40-h Lanczos filter, subsampled daily, and then organized into monthly mean time series. Additional, longer-time data (1959–2008) at Guam (13.43°N, 144.65°E) and Midway (28.22°N, 177.37°W), as well as at Ishigaki were also downloaded and processed in a similar way.

Satellite sea surface height (SSH) anomaly data (SSHA) (Le Traon et al. 1998; Ducet et al. 2000) on a ⅓° × ⅓° Mercator grid were downloaded from http://www.aviso.oceanobs.com/es/data/index.html. The weekly data (October 1992–December 2008) is organized into monthly-mean data. The mean SSH field is from Rio et al. (2011). This has a resolution of ¼° × ¼° and was constructed by combining the Gravity Recovery and Climate Experiment (GRACE) geoid, drifting buoy velocities, profiling float and hydrographic temperature and salinity data.

Monthly-mean ECMWF wind stress data at 1° × 1° resolution was downloaded from the Asia–Pacific Data-Research Center (http://apdrc.soest.hawaii.edu/). The corresponding wind stress curl time series is then calculated.

c. Weighted composite

Positive or negative composite (C+ or C) of a variable υn (e.g., SSHA, wind stress curl, etc.) based on a time series wn (such as the PTO, PDO, etc.), n = time index, is calculated using the following formula:

 
formula

where Σ+ denotes sum over the positive wn only. A similar formula is used for C using Σ. The above formula gives a better composite than straight averaging since it reduces the influences of values of “υn” near small “wn,” which can potentially have larger uncertainty.

d. EOFs of wind stress curl

We show in Fig. A1 the two leading EOF modes of × τ0. For each mode, we calculated lagged correlations with major indices and show only the index that is most correlated with each mode: for mode 1 this is the PDO, and for mode 2 it is WP (plotted with a minus sign). The mode2/WP correlation is apparently due to the strong pole over the Kamchatka Peninsula. Table A1 below gives all correlations and lags. It is interesting that, while each of the major indices may correlate well (i.e., R > 0.47) with only one of the modes (PDO correlates with mode-1 EOF of the WSC, while −WP correlates with mode-2 EOF), the PTO correlates reasonably well with both modes.

Fig. A1.

(a) Mode-1 EOF of low-passed wind stress curl in the north Pacific: (top) eigenfunction, (middle) principal component (PC, solid line) and PDO index (dashed line), and (bottom) lagged correlation between PC and PDO, colored red if it is above the 95% significance level. (b) The mode-2 EOF and “−WP” index instead of PDO. Units are arbitrary.

Fig. A1.

(a) Mode-1 EOF of low-passed wind stress curl in the north Pacific: (top) eigenfunction, (middle) principal component (PC, solid line) and PDO index (dashed line), and (bottom) lagged correlation between PC and PDO, colored red if it is above the 95% significance level. (b) The mode-2 EOF and “−WP” index instead of PDO. Units are arbitrary.

Table A1.

Correlations/lags (in months) of WSC EOF modes 1 and 2 with major indices and with PTO. See Fig. A1 for the corresponding EOF modes.

Correlations/lags (in months) of WSC EOF modes 1 and 2 with major indices and with PTO. See Fig. A1 for the corresponding EOF modes.
Correlations/lags (in months) of WSC EOF modes 1 and 2 with major indices and with PTO. See Fig. A1 for the corresponding EOF modes.
e. OFES and reduced-gravity model results: Figs. A2 and A3

We use two sets of model results, one from an OGCM and the other from reduced gravity (RG), to check the role of STCC eddies in the interannual variability of the Luzon Strait transport. Details of the RG model are given in Chang and Oey (2011; online auxiliary materials). The OGCM is OFES (OGCM for the Earth Simulator; http://www.jamstec.go.jp/esc/ofes/eng/index.html; Masumoto et al. 2004) based on the Geophysical Fluid Dynamics Laboratory Modular Ocean Model, version 3 (MOM3), with a resolution of 0.1° × 0.1° and 54 vertical levels. The National Centers for Environmental Prediction monthly wind was used to force the model, and the downloaded data for analysis is from 1950 to 2006.

The mean SSH from OFES and AVISO have previously been compared in Chang and Oey (2011); a summary is given here. East and northeast of Taiwan, the OFES mean SSH contours are more tilted northeastward and there is a large anticyclone south of Kyushu at 29°N, 134°E. The general features are otherwise similar. In particular, both show NEC bifurcation latitude (NECBL) at around 13°N east of the Philippines and both show a generally cyclonic circulation in the South China Sea. The interannual fluctuations of NECBL also agree well. The tide gauge (Δηik or TG) time series from OFES and AVISO are very different, however, which suggests that each is driven by its own (uncorrelated) eddy field.

We next compare the eddy fields in Fig. A2, which show that AVISO and OFES STCC eddy fields are different. The OFES does not have an eastward extended eddy field over the STCC (east of Taiwan) as observed (cf. Figs. A2b,c). Figure A2e shows that OFES Δηik correlates positively in the southern South China Sea and insignificantly in the northern third with its own SSHA. These patterns differ from the observed correlation that shows a significant negative correlation throughout most of the South China Sea (Fig. A2d). Since the OFES NECBL matches well with satellite and since also the OFES PTO computed from NCEP (not shown) is very similar to the ECMWF PTO used in text to interpret observed responses, the discrepancy between the observed and OFES corr(Δηik,SSHA) in the South China Sea strongly suggests that interannual variation of NECBL alone is insufficient and the STCC eddy field also affects transport through Luzon Strait.

Fig. A2.

(left) Mean squared SSHA (〈η2〉, angle brackets represent time mean) in the indicated domains for (a),(b) AVISO and (c) OFES. (right) Correlation between sea level difference Δηik (Ishigaki minus Keelung) and SSHA using (d) observed tide gauge data and AVISO SSHA, and (e) OFES Δηik and SSHA. [(d),(e) are from Chang and Oey 2011).

Fig. A2.

(left) Mean squared SSHA (〈η2〉, angle brackets represent time mean) in the indicated domains for (a),(b) AVISO and (c) OFES. (right) Correlation between sea level difference Δηik (Ishigaki minus Keelung) and SSHA using (d) observed tide gauge data and AVISO SSHA, and (e) OFES Δηik and SSHA. [(d),(e) are from Chang and Oey 2011).

To further support the above inference, Figure A3 compares the SSHA and velocity composites from OFES (Fig. A3b) with those observed from AVISO data (Fig. A3a). The composites from a reduced-gravity (RG) model of the Pacific Ocean are also included in Fig. A3c. Like OFES, the RG model reproduces fluctuations of the NECBL very well, and the correlation with the observed NECBL from AVISO is high, ≈0.83 at zero lag (cf. QC2010b). It is clear from Fig. A3 that neither OFES nor the RG model reproduces the SCS response observed in AVISO.7 Since baroclinic instability is absent from the RG model, it does not have STCC eddies (QC2010b). During the positive phase of PTO, both models produce weak Luzon Strait inflow, weaker in RG than in OFES. OFES shows weak or slightly positive SSHA (Fig. A3b) in the SCS, and RG shows generally stronger positive SSHA (Fig. A3c), while the observed SSHA is predominantly negative (Fig. A3a). In the absence of STCC eddies in the RG model, the SCS response is governed by Rossby wave and Svedrup dynamics; at interannual time scales, under a negative wind stress curl when PTO > 0, the quasi-steady response is a clockwise gyre with a western boundary current (Fig. A3c). The OFES composite (Fig. A3b) is slightly closer to the observed (Fig. A3a) than the RG composite (Fig. A3c), which suggests that the multilevel model is able to simulate some STCC eddies (see Fig. A2, also Fig. A3b), though the distribution and strengths of the eddies differ from those observed (Chang and Oey 2011). These results again suggest that STCC eddies are necessary to generate a correct interannual response in SCS.

Fig. A3.

Composites of SSHA and velocity from (a) AVISO, (b) OFES, and (c) reduced-gravity Pacific Ocean model for (top) positive and (bottom) negative phases of the PTO.

Fig. A3.

Composites of SSHA and velocity from (a) AVISO, (b) OFES, and (c) reduced-gravity Pacific Ocean model for (top) positive and (bottom) negative phases of the PTO.

f. Comparisons of ocean responses due to PTO and other indices: Table A2

Table A2 lists the correlations of various indices with ocean variables discussed in text and compares them with the corresponding correlations based on the PTO. In general, Niño-3.4 and EMI explain well tropical responses such as the Luzon Strait transport and the NEC bifurcation latitude, but they do less well for eddy-zone SSHA and EKE and poorly for the Kuroshio transport northeast of Taiwan. This is consistent with the discussion in the text that these indices do not have a distinct wind stress curl dipole east of Philippines and Taiwan, making them poor predictors of STCC eddies, hence Kuroshio transport. As also mentioned in text, the WP correlates quite well with STCC EKE, but importantly it also correlates with TrLZ. Since negative WP produces convergence east of Taiwan, the correlation of WP with TrLZ reaffirms the relevance of STCC eddies in forcing TrLZ.

Table A2.

Correlations/lags (months, > 0 if column 1 leads) between indices in column 1 and ocean variables in columns 2 through 6 and with PTO (last column). Except for AVISO, all time series data are from 1980 to 2008.

Correlations/lags (months, > 0 if column 1 leads) between indices in column 1 and ocean variables in columns 2 through 6 and with PTO (last column). Except for AVISO, all time series data are from 1980 to 2008.
Correlations/lags (months, > 0 if column 1 leads) between indices in column 1 and ocean variables in columns 2 through 6 and with PTO (last column). Except for AVISO, all time series data are from 1980 to 2008.
g. PTO and Kuroshio south of Japan: Fig. A4

It is well known that the Kuroshio south of Japan exhibits a bimodal meander behavior (Kawabe 1985). In some years, the current stays close to the southern coast of Japan, sometimes with a small meander. In other years, the meander grows and the Kuroshio shifts seaward for hundreds of kilometers before meandering back to the coast and continuing its east-northeastward course. Various mechanisms have been proposed; an excellent review is given in Qiu and Miao (2000) who also proposed an internal mechanism for the bimodal meander behavior. On the other hand, Miyazawa et al. (2004, 2008) suggested that eddies and transports from farther south (off Taiwan and the Philippines) can affect the Kuroshio meander south of Japan. Figure A4 shows satellite SSHA regressed to the (normalized) PTO. This shows positive regression east of Taiwan and negative one east of the Philippines, as discussed in text, as well as a positive regression in the tropical Pacific. In addition, the plot also shows a prominent negative–positive dipolar structure, each about 300~400 km in diameter south of Japan [east of Kyushu and south of Honshu; correlations are approximately ±0.6 (figure not shown)], and large regression values extend farther east beyond the separation latitude of the Kuroshio. The dipole indicates a meandering of the Kuroshio that is tied to the PTO. A cyclonic anomaly south of Japan prevails during the positive phase of PTO, indicating a condition more conducive to triggering an offshore large meander, and vice versa for negative PTO. Miyazawa et al. (2008) show that the large meander in 2004 may be related to a warm eddy off Taiwan. This agrees well when the PTO changes from being negative to positive some 12 months earlier (Fig. 2b in text) and, also, with the time of increased Kuroshio transport (Chang and Oey 2011). Increased Kuroshio transport has also been attributed to initiation of the large meander (Saiki 1982; Akitomo et al. 1996). Since the PTO leads STCC eddies (SSHA and EKE) and Kuroshio transport by as long as one year (see text Table 1) and since also there is additionally a time lag of 3~6 months for eddy (and/or transport) perturbations east of Taiwan to reach southern Japan (Miyazawa et al. 2008), the development of a large meander can be some 18 months after a positive PTO. Therefore, since the PTO and El Niño are roughly in phase (Table A2 shows that PTO leads Niño-3.4 by a relatively short 5 months and is almost in phase with EMI), then, if large Kuroshio meanders develop, they tend to do so in La Niña years. This appears to be the case (e.g. Akitomo et al. 1996).

Fig. A4.

Regression of AVISO SSHA (1993–2008) against (normalized) PTO: contours intervals = 0.03 m added to show values that exceed the color scale (of ±0.1).

Fig. A4.

Regression of AVISO SSHA (1993–2008) against (normalized) PTO: contours intervals = 0.03 m added to show values that exceed the color scale (of ±0.1).

We have also conducted a similar analysis for the other four indices (Niño-3.4, EMI, PDO, and WP). The one that most resembles the PTO is Niño-3.4 (not shown), but its correlation with eddies south of Japan is weak, ≈±0.2.

Figure A4 also shows a north–south dipole at the Kuroshio separation latitude near 35°N. This suggests a more southerly separation of the Kuroshio during years of positive PTOs. This however may merely reflect large-scale wind conditions that are well correlated with PTO, rather than the PTO (or transport or eddies east of Taiwan) itself. More research is clearly necessary.

REFERENCES

REFERENCES
Akitomo
,
K.
,
M.
Doi
,
T.
Awaji
, and
K.
Kutsuwada
,
1996
:
Interannual variability of the Kuroshio transport in response to the wind stress field over the North Pacific: Its relation to the path variation south of Japan
.
J. Geophys. Res.
,
101
,
14 057
14 071
.
Aoki
,
Y.
,
T.
Suga
, and
K.
Hanawa
,
2002
:
Subsurface subtropical fronts of the North Pacific as inherent boundaries in the ventilated thermocline
.
J. Phys. Oceanogr.
,
32
,
2299
2311
.
Ashok
,
K.
,
S. K.
Behera
,
S. A.
Rao
,
H.
Weng
, and
T.
Yamagata
,
2007
:
El Niño Modoki and its possible teleconnection
.
J. Geophys. Res.
,
112
,
C11007
,
doi:10.1029/2006JC003798
.
Centurioni
,
L. R.
,
P. P.
Niiler
, and
D.-K.
Lee
,
2004
:
Observations of inflow of Philippine Sea surface water into the South China Sea through Luzon Strait
.
J. Phys. Oceanogr.
,
34
,
113
121
.
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
.
Chelton
,
D. B.
, and
M. G.
Schlax
,
1996
:
Global observations of oceanic Rossby waves
.
Science
,
272
,
234
238
.
Chelton
,
D. B.
,
M. G.
Schlax
, and
R. M.
Samelson
,
2011
:
Global observations of nonlinear mesoscale eddies
.
Prog. Oceanogr.
,
91
,
167
216
,
doi:10.1016/j.pocean.2011.01.002
.
Chen
,
T.-C.
,
S.-Y.
Wang
, and
M.-C.
Yen
,
2006
:
Interannual variation of the tropical cyclone activity over the western North Pacific
.
J. Climate
,
19
,
5709
5720
.
Ducet
,
N.
,
P. Y.
Le Traon
, and
G.
Reverdin
,
2000
:
Global high-resolution mapping of ocean circulation from TOPEX/Poseidon and ERS-1 and -2
.
J. Geophys. Res.
,
105
,
19 477
19 498
.
Gill
,
A. E.
,
1980
:
Some simple solutions for heat-induced tropical circulation
.
Quart. J. Roy. Meteor. Soc.
,
106
,
447
462
.
Gordon
,
A. L.
,
1986
:
Inter-ocean exchange of thermocline water
.
J. Geophys. Res.
,
91
,
5037
5050
.
Guo
,
X.
,
Y.
Miyazawa
, and
T.
Yamagata
,
2006
:
The Kuroshio onshore intrusion along the shelf break of the East China Sea: The origin of the Tsushima warm current
.
J. Phys. Oceanogr.
,
36
,
2205
2231
.
Johns
,
W. E.
,
T. N.
Lee
,
D.
Zhang
, and
R.
Zantopp
,
2001
:
The Kuroshio east of Taiwan: Moored transport observations from the WOCE PCM-1 array
.
J. Phys. Oceanogr.
,
31
,
1031
1053
.
Kagimoto
,
T.
, and
T.
Yamagata
,
1997
:
Seasonal transport variations of the Kuroshio: An OGCM simulation
.
J. Phys. Oceanogr.
,
27
,
403
418
.
Kawabe
,
M.
,
1985
:
Sea level variations at the Izu Islands and typical stable paths of the Kuroshio
.
J. Oceanogr. Soc. Japan
,
41
,
307
326
.
Kim
,
J.-H.
,
C.-H.
Ho
, and
P.-S.
Chu
,
2010
:
Dipolar redistribution of summertime tropical cyclone genesis between the Philippine Sea and the northern South China Sea and its possible mechanisms
.
J. Geophys. Res.
,
115
,
D06104
,
doi:10.1029/2009JD012196
.
Le Traon
,
P. Y.
,
F.
Nadal
, and
N.
Ducet
,
1998
:
An improved mapping method of multisatellite altimeter data
.
J. Atmos. Oceanic Technol.
,
15
,
522
534
.
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
.
Bull. Amer. Meteor. Soc.
,
78
,
1069
1079
.
Masumoto
,
Y.
, and
T.
Yamagata
,
1991
:
Response of the western tropical Pacific to the Asian winter monsoon: The generation of the Mindanao Dome
.
J. Phys. Oceanogr.
,
21
,
1386
1398
.
Masumoto
,
Y.
, and
Coauthors
,
2004
:
A fifty-year eddy-resolving simulation of the World Ocean—Preliminary outcomes of OFES (OGCM for the Earth Simulator)
.
J. Earth Simul.
,
1
,
35
56
.
Matsuno
,
T.
,
1966
:
Quasi-geostrophic motion in the equatorial area
.
J. Meteor. Soc. Japan
,
44
,
25
43
.
McPhaden
,
M. J.
, and
D.
Zhang
,
2002
:
Slowdown of the meridional overturning circulation in the upper Pacific Ocean
.
Nature
,
415
,
603
608
.
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
.
Qiu
,
B.
,
1999
:
Seasonal eddy field modulation of the North Pacific subtropical countercurrent: TOPEX/Poseidon observations and theory
.
J. Phys. Oceanogr.
,
29
,
2471
2486
.
Qiu
,
B.
, and
W.
Miao
,
2000
:
Kuroshio path variations south of Japan: Bimodality as a self-sustained internal oscillation
.
J. Phys. Oceanogr.
,
30
,
2124
2137
.
Qiu
,
B.
, and
S.
Chen
,
2010a
:
Interannual-to-decadal variability in the bifurcation of the North Equatorial Current off the Philippines
.
J. Phys. Oceanogr.
,
40
,
2525
2538
.
Qiu
,
B.
, and
S.
Chen
,
2010b
:
Interannual variability of the North Pacific Subtropical Countercurrent and its associated mesoscale eddy field
.
J. Phys. Oceanogr.
,
40
,
213
225
.
Qu
,
T.
,
2000
:
Upper-layer circulation in the South China Sea
.
J. Phys. Oceanogr.
,
30
,
1450
1460
.
Qu
,
T.
, and
R.
Lukas
,
2003
:
On the bifurcation of the North Equatorial Current in the Pacific
.
J. Phys. Oceanogr.
,
33
,
5
18
.
Qu
,
T.
,
Y.
Kim
,
M.
Yaremchuk
,
T.
Tozuka
,
A.
Ishida
, and
T.
Yamagata
,
2004
:
Can Luzon Strait transport play a role in conveying the impact of ENSO to the South China Sea?
J. Climate
,
17
,
3644
3657
.
Qu
,
T.
,
Y.
Du
, and
H.
Sasaki
,
2006
:
South China Sea throughflow: A heat and freshwater conveyor
.
Geophys. Res. Lett.
,
33
,
L23617
,
doi:10.1029/2006GL028350
.
Rio
,
M.-H.
,
P.
Schaeffer
,
G.
Moreaux
,
J.-M.
Lemoine
, and
E.
Bronner
, cited
2011
:
A new mean dynamic topography computed over the global ocean from GRACE data, altimetry and in-situ measurements
.
Saiki
,
M.
,
1982
:
Relation between the geostrophic flux of the Kuroshio in the eastern China Sea and its large meander in the south of Japan
.
Oceanogr. Mag.
,
32
,
11
18
.
Samelson
,
R. M.
,
1992
:
Fluid exchange across a meandering jet
.
J. Phys. Oceanogr.
,
22
,
431
440
.
Sheremet
,
V.
,
2001
:
Hysteresis of a western boundary current leaping across a gap
.
J. Phys. Oceanogr.
,
31
,
1247
1259
.
Sheu
,
W.-J.
,
C.-R.
Wu
, and
L.-Y.
Oey
,
2010
:
Blocking and westward passage of eddies in the Luzon Strait
.
Deep-Sea Res. II
,
57
,
1783
1791
,
doi:10.1016/j.dsr2.2010.04.004
.
Tozuka
,
T.
, and
T.
Yamagata
,
2003
:
Annual ENSO
.
J. Phys. Oceanogr.
,
33
,
1564
1578
.
Tozuka
,
T.
,
T.
Kagimoto
,
Y.
Masumoto
, and
T.
Yamagata
,
2002
:
Simulated multiscale variations in the western tropical Pacific: The Mindanao Dome revisited
.
J. Phys. Oceanogr.
,
32
,
1338
1359
.
Trenberth
,
K. E.
,
1990
:
Recent observed decadal climate changes in the Northern Hemisphere
.
Bull. Amer. Meteor. Soc.
,
71
,
988
993
.
Trenberth
,
K. E.
,
1997
:
The definition of El Niño
.
Bull. Amer. Meteor. Soc.
,
78
,
2771
2777
.
Vecchi
,
G. A.
,
B. J.
Soden
,
A. T.
Wittenberg
,
M. H.
Isaac
,
A.
Leetmaa
, and
M. J.
Harrison
,
2006
:
Weakening of tropical Pacific atmospheric circulation due to anthropogenic forcing
.
Nature
,
441
,
73
76
,
doi:10.1038/nature04744
.
Wallace
,
J. M.
, and
D. S.
Gutzler
,
1981
:
Teleconnections in the geopotential height field during the Northern Hemisphere winter
.
Mon. Wea. Rev.
,
109
,
784
812
.
Wang
,
B.
,
R.
Wu
, and
X.
Fu
,
2000
:
Pacific–East Asian teleconnection: How does ENSO affect East Asian climate?
J. Climate
,
13
,
1517
1536
.
Yaremchuk
,
M.
, and
T.
Qu
,
2004
:
Seasonal variability of the large-scale currents near the coast of the Philippines
.
J. Phys. Oceanogr.
,
34
,
844
855
.
Zhang
,
Y.
,
J. M.
Wallace
, and
D. S.
Battisti
,
1997
:
ENSO-like interdecadal variability: 1900–93
.
J. Climate
,
10
,
1004
1020
.

Footnotes

*

Additional affiliation: National Central University, Jhongli, Taiwan.

1

Here, the notation corr(A,B,lags) = maximum lagged correlation coefficient between A and B with lags in months, positive (negative) if A leads (lags) B. All correlations quoted in this paper are above the 95% significance level. For simplicity, we write corr(A,B) for the zero-lag correlation.

2

Nonlinear eddies are eddies whose rotational speed U is greater than translational speed C, or U/C > 1 (Samelson 1992; also R. M. Samelson 2011, personal communication). For the STCC eddies, the ratio is ~4 (Chelton et al. 2011).

3

By geostrophy, Trik ≈ 10−6 × (gH/f) Δηik, where g is gravity, f (6.2 × 10−5 s−1 at 25°N) is the Coriolis parameter, and H is the upper-layer depth scale of the transport. For Johns et al. (2001), H = 221 m, whereas for OFES, H = 366 m.

4

The word “anomaly” will be omitted unless the omission causes confusion.

5

A nearly identical result is obtained using USV calculated from a reduced-gravity model of the North Pacific forced by NCEP and satellite wind (Chang and Oey 2011).

6

On the other hand, the mechanism has not been conclusively demonstrated in the literature using observations and/or realistic models of Luzon Strait and the South China Sea.

7

B. Qiu (2011, personal communication) correctly pointed out that, since the RG model has no opening in the southeastern South China Sea (through the Sulu Sea, etc.) to the Philippines Sea in the equatorial Pacific, the SCS response may be anomalous because sea level set up by Kelvin waves is then blocked. However, the general similarity of the SCS response for both the RG model and OFES (which has the opening) suggests that the possible blockage of Kelvin waves is not crucial to the dynamics.