Journal of Geophysical Research: Oceans

Revisit the interannual variability of the North Equatorial Current transport with ECMWF ORA-S3

Authors

  • Fangguo Zhai,

    1. Key Laboratory of Ocean Circulation and Waves, Chinese Academy of Sciences, Qingdao, China
    2. Institute of Oceanology, Chinese Academy of Sciences, Qingdao, China
    3. Third Institute of Oceanography, State Oceanic Administration, Xiamen, China
    Search for more papers by this author
  • Dunxin Hu

    Corresponding author
    1. Institute of Oceanology, Chinese Academy of Sciences, Qingdao, China
    • Key Laboratory of Ocean Circulation and Waves, Chinese Academy of Sciences, Qingdao, China
    Search for more papers by this author

Corresponding author: D. Hu, Key Laboratory of Ocean Circulation and Waves, Institute of Oceanology, Chinese Academy of Sciences, 7 Nanhai Road, Qingdao, 266071, China. (dxhu@qdio.ac.cn)

Abstract

[1] The interannual variability of the North Equatorial Current (NEC) transport in the tropical northwestern Pacific Ocean is investigated with the output from ECMWF Ocean Analysis/Reanalysis System 3 (ORA-S3). The results show that the amplitude and root mean square (RMS) of interannual NEC transport anomalies increase from about 3.0–4.0 Sv and 2.0 Sv at 170°E to above 5.0 and 3.4 Sv at 135°E, respectively. The NEC transport variation agrees well with the variation of the sea surface height (SSH) anomaly difference between the southern and northern boundaries of the NEC region. Further analysis near the Philippine coast suggests that their good agreement mainly comes from the agreement of the NEC transport and SSH variations south of the gyre boundary. Around the bifurcation point off the Philippine coast, the southern branch of the NEC transport is highly related to El Niño–Southern Oscillation (ENSO) events. During El Niño/La Niña years, westerly/easterly wind anomalies and positive/negative wind stress curl anomalies develop in the tropical northwestern Pacific Ocean south of 20°N before the mature phase. The wind forcing center moves eastward with time and reaches the easternmost position around 170°E several months before the mature phase. This wind forcing generates upwelling/downwelling Rossby waves, which propagate westward to result in negative/positive SSH anomalies, hence inducing a cyclonic/anticyclonic gyre anomaly, which is responsible for the increase/decrease of the NEC transport. The northern branch of the NEC transport near the Philippine coast has no significant simultaneous relation with ENSO events.

1 Introduction

[2] The North Equatorial Current (NEC) in the North Pacific Ocean is located between the subtropical and tropical gyres and bifurcates into the northward flowing Kuroshio Current (KC) and southward flowing Mindanao Current (MC) near the Philippine coast; the above three currents comprise the NEC-MC-KC (NMK) current system. The NMK current system has been heavily investigated based on in situ observations [e.g., Toole et al., 1988; Hu and Cui, 1989; Toole et al., 1990; Hu and Cui, 1991; Qiu and Joyce, 1992; Gu, 1996; Qiu and Lukas, 1996; Qu et al., 1997; Qu et al., 1998; Qu and Lukas, 2003; Yaremchuk and Qu, 2004; Wang and Hu, 2006; Kashino et al., 2009; Xie et al., 2009; Qiu and Chen, 2010b; Zhai and Hu, 2012] and numerical simulations [e.g., Qiu and Lukas, 1996; Wang et al., 2002; Kim et al., 2004]. Among the NMK current system, the NEC is the root and plays an important role in mass and heat transport to the western Pacific warm pool [e.g., Qu et al., 1997] and global thermohaline circulation via the Indonesian throughflow [e.g., Gordon, 1986]. Its transport variations on both seasonal and interannual time scales have been paid much attention [e.g., White and Hasunuma, 1980; Qiu and Joyce, 1992; Gu, 1996; Qiu and Lukas, 1996; Qu et al., 1998; Wang et al., 2002; Kim et al., 2004; Yaremchuk and Qu, 2004; Zhou et al., 2006; Kashino et al., 2009; Zhai and Hu, 2012]. It is well known that the interannual NEC transport variations are closely associated with El Niño–Southern Oscillation (ENSO) events, increasing during El Niño events and decreasing during La Niña events. This is mainly caused by the westward propagating baroclinic Rossby waves generated by anomalous winds in the western-central tropical North Pacific Ocean [e.g., Qiu and Joyce, 1992; Qiu and Lukas, 1996; Qu et al., 1998; Kim et al., 2004; Kashino et al., 2009; Zhai and Hu, 2012]. However, all the previous studies have focused on the NEC transports only in the far tropical northwestern Pacific Ocean near the Philippine coast but without any information about the picture in the east. Furthermore, though Kim et al. [2004] pointed out that all the extremes in the NEC transport follow the mature phase of ENSO events based on the output of a high-resolution Ocean General Circulation Model (OGCM), it is shown from observations by Qiu and Joyce [1992] and Zhai and Hu [2012] that not all the extremes are related to ENSO events. Therefore, the exact relationship between the interannual NEC transport variability and ENSO events should be reexamined with focus on the underlying dynamics based on long-term observations or data assimilation. Recently, Zhai and Hu [2012] pointed out that during ENSO events, the sea surface height (SSH) in the tropical northwestern Pacific Ocean falls/rises, giving rise to a cyclonic/anticyclonic gyre anomaly, which should be responsible for the increase/decrease in the NEC transport and northward/southward shift of the NEC bifurcation latitude. This should be further confirmed with long-term three-dimensional ocean variables. Using long time series of the output of European Centre for Medium-Range Weather Forecasts (ECMWF) Ocean Analysis/Reanalysis System 3 (ORA-S3) [Balmaseda et al., 2008], the current study intends to identify the following aspects: (1) the interannual variability of the NEC transport near and in the east away from the Philippine coast, (2) its relationship with ENSO events, and (3) the exact mechanisms at work. The rest of the paper is starting with section 2 describing the data and methods, followed by section 3 specifying the characteristics and exploring the underlying dynamics of the interannual NEC transport variability, and ending with section 4 giving the summary.

2 Data and Methods

2.1 ECMWF ORA-S3 and Its Validation in the Tropical North Pacific Ocean

[3] Monthly mean output of ECMWF ORA-S3 adopted in this study covers the global ocean with a horizontal resolution of 1° × 1° and spans a period of 51 years from January 1959 to December 2009. There are 29 vertical levels un-uniformly extending from 5 to 5250 m. This data set has some innovative features, including assimilation of salinity on temperature surfaces and also assimilation of altimetry observations and global sea level trends. One can refer to Balmaseda et al. [2008] for more detailed description of the data set, including the ocean assimilation scheme, the wind forcing, data sets assimilated, and comparisons with observations and previous versions. Before further analysis, it is necessary to check the performance of the reanalysis data in the tropical North Pacific Ocean. Here are some comparisons of the reanalysis data with observations in the tropical North Pacific Ocean with focus on the western part. Figure 1 shows the modeled annual mean velocity in ECMWF ORA-S3 and observed annual mean geostrophic velocity calculated from World Ocean Atlas 2009 [WOA09; Antonov et al., 2010; Locarnini et al., 2010] at 100 m in the North Pacific. In Figure 1b, the geostrophic velocities south of 6°N are not calculated. It can be seen that the modeled velocity is generally consistent with the observation except in the Kuroshio region, where the former is slightly stronger than the latter. The SSH anomalies in ECMWF ORA-S3 also show good agreement with tidal gauge sea level observations in the tropical northwestern Pacific Ocean (Figure 2; see Figure 1a for locations of the stations). Different in magnitude, their simultaneous correlation coefficients are all higher than 0.70 above 95% confidence level. Figure 3 gives the annual mean zonal geostrophic velocities at 137°E calculated with temperature and salinity from ECMWF ORA-S3 and hydrographic observations obtained by Japan Meteorological Agency (JMA), respectively. Here the observed temperature and salinity are meridionally interpolated onto integer latitudes and smoothed with a Gaussian filter of about 150 km e-folding scale [Qu and Lukas, 2003], intending to remove contributions from transient processes such as internal waves, eddies and other small-scale motions so as to highlight the long-term variations. The meridional structures of the zonal velocities in both two data sets are in quite good agreement with each other. The NEC transport in ECMWF ORA-S3 also agrees well with observations (Figure 3c); the linear correlation coefficient between them is about 0.7 above 95% confidence level. The modeled and observed mean transports of the NEC are 52.9 Sv (1 Sv ≡ 106 m3 s−1) and 51.0 Sv with root mean square (RMS) of 6.1 and 6.7 Sv, respectively. Then we have the conclusion that the output of the ECMWF ORA-S3 is valid to be used for the present study.

Figure 1.

(a) Modeled annual mean velocity at 100 m in ECMWF ORA-S3. (b) Observed annual mean geostrophic velocity at 100 m relative to 1500 m calculated from WOA09. Solid dots in Figure 1a denote tidal gauge stations.

Figure 2.

Normalized time series of SSH anomalies (SSHA, gray lines) from ECMWF ORA-S3 and observations (black lines) at tidal gauge stations Guam (13.43°N, 144.65°E), Legaspi (13.15°N, 123.75°E), Malakal (7.33°N, 134.46°E), Saipan (15.23°N, 145.74°E), Truk (7.45°N, 151.85°E), and Yap (9.51°N, 138.13°E). See Figure 1a for locations of these six stations. The number in each panel indicates their simultaneous correlation coefficient above 95% confidence level.

Figure 3.

Annual mean zonal velocities at 137°E calculated with temperature and salinity observed by JMA (a) and modeled by ECMWF ORA-S3 (b). In Figures 3a and 3b, the gray lines indicate the 26. 7σθ-isopycnal surface. (c) The bottom panel compares the (gray line) observed and (black line) modeled NEC transports.

2.2 Extraction of Interannual Signals

[4] For monthly variables, monthly mean anomalies (MMAs) are obtained by subtracting the monthly climatology from the original monthly means. Then a 13-month running mean filter is applied to MMAs to exclude subannual signals. At last, a high-pass filter is applied to remove decadal variations longer than 8 years.

3 Interannual Variability of the NEC Transport

3.1 Longitudinal Distribution

[5] As shown in Figures 3a and 3b, the main body of the NEC climatology along 137°E is confined between 7°N and 21°N and above 26.7σθ-isopycnal surface. Its maximum westward velocity larger than 5 cm s−1 shifts from about 11°N near the sea surface northward with increasing depth to about 20°N at 300 m. We also examine the time variability of the boundary and depth of the NEC. The results show that the main body of the NEC transport is confined above 26.7σθ-isopycnal surface south of 21°N. In order to include the main body of the NEC transport and compare with previous results [e.g., Qiu and Joyce, 1992; Qiu and Lukas, 1996; Qu et al., 1998; Kashino et al., 2009], the NEC transports (TNEC) at integer longitudes from 135°E to 170°E are calculated through integrating the westward velocity vertically from the sea surface downward to 26.7σθ-isopycnal surface and meridionally from 7°N to 21°N. Qu et al. [1998] and Kashino et al. [2009] have also chosen 26.7σθ-isopycnal surface as the lower boundary. For the southern and northern boundaries, Qu et al. [1998] and Kashino et al. [2009] have defined them at 8°N and around 18°N along 130°E, respectively. Qiu and Joyce [1992] and Qiu and Lukas [1996] have defined them at 7°N and 25°N along 137°E and at 8°N and 20°N along 135°E, respectively. According to the zonal velocity distribution along 137°E (Figures 3a and 3b), 18°N is too south to include the main body of the NEC transport, while 25°N is so north that part of the Kuroshio countercurrent may be included in the NEC transport. As the total NEC transport mainly comes from the southern part, the influence of choosing a fixed southern boundary at 7°N on the interannual NEC transport variability will be further analyzed later. The calculations show that the mean NEC transport increases from 39.4 Sv at 170°E to 53.4 Sv at 135°E. Their interannual anomalies are estimated according to the method described in section 2.2. Figure 4 shows the NEC transport anomalies at different longitudes from 137°E to 170°E as examples (black lines). As discussed in the following sections, accompanying the interannual variations of the NEC transport, the significant circulation changes mainly occur in the southern NEC region, implying large northward/southward shifting of the southern boundary. To examine the impact of the meridional shift of the southern boundary, we also show in Figure 4 the transport anomalies (red lines) calculated with a variable boundary of zero velocity following the tropical gyre center along longitudes west of 160°E. Along longitudes in the east, the southern boundary of the NEC is nearly always north of 7°N. These two kinds of anomalies are nearly equal to each other, indicating that using a southern boundary fixed at 7°N or varying with time makes no difference to the interannual NEC transport anomalies. The amplitude and RMS of the transport anomalies increase from about 3.0–4.0 Sv and 2.0 Sv at 170°E to above 5.0 and 3.4 Sv at 135°E, respectively. Their RMS is comparable with that of the monthly anomalies. Temporally, the transport anomalies in the west lag behind those in the east. The maximum time lag of 1 month takes place between transport anomalies at 170°E and 135°E.

Figure 4.

NEC transport anomalies at different longitudes. Black lines are those calculated with a fixed southern boundary defined at 7ºN, while the red lines are those calculated with a time-varying southern boundary.

[6] As the NEC is the boundary current between the subtropical and tropical gyres, its variability is related to that of the broader-scale gyre circulation. Figures 5a and 5b give the differences of the SSH and horizontal velocity averaged from 50 to 120 m between years when the NEC transport anomalies at 135°E are larger than 3.0 Sv and smaller than −3.0 Sv, respectively. In the years of large NEC transport anomalies, the SSH in the North Pacific tropical gyre largely falls and the circulation significantly strengthens. In the western part, especially west of 160°E, the meridional scale of the circulation anomaly significantly broadens, which may result in the NEC bifurcating at further north latitudes near the Philippine coast. As discussed in the following subsections, the meridional broadening of the western part of the tropical gyre is related to ENSO events. On the other hand, the SSH and horizontal currents in the North Pacific subtropical gyre show no significant changes, implying that associated with the interannual NEC transport variations, largest circulation changes mainly occur in the southern NEC region. This can also be seen from the standard deviation of the zonal velocity anomalies along 137°E. As shown in Figure 5c, large standard deviation mainly appears in the upper layer south of 10°N, corresponding well with that of horizontal circulation anomalies.

Figure 5.

(a) Differences of the SSH (cm) between years with large and small interannual anomalies of NEC transport at 135°E (see text for definition). (b) Same as Figure 5a but for the depth-averaged horizontal velocity (cm s−1) from 50 to 120 m. (c) Standard deviation of the interannual anomalies of zonal velocity at 135ºE.

3.2 inline image Layer Reduced Gravity Model

[7] As the ocean is basically dominated by the first baroclinic mode [e.g., Wunsch, 1997], the variability of the upper layer ocean transport can be examined by a inline image layer reduced gravity model, which is expressed as

display math(1)

where h is the upper layer thickness (equivalent to the pycnocline depth), y1 and y2 are the southern and northern latitudes for transport integration, and u is the upper layer zonal velocity and can be calculated with the SSH η (equivalent to the depth-integrated sea surface dynamic height) or h as

display math(2)

[8] In (1), the negative sign is chosen to make the NEC transport westward positive. In (2), g and g′ are the gravity and reduced gravity, respectively, and f is the Coriolis parameter.

[9] The pycnocline depth h is chosen as has highest overall correlation coefficient with η, because in the reduced gravity model, h is related to η following  = gh + C with C being constant. The overall correlation coefficients between h and η within the latitudinal band of 7°N–21°N are near 1.0 with confidence level above 95%. The reduced gravity g′ and C are estimated through fitting η to h. The estimated g′ generally decreases eastward from about 0.044 m s−2 at 135°E to about 0.036 m s−2 at 170°E, indicating eastward decreasing density difference across the base of the main thermocline. Oppositely, the C generally increases eastward from about −1.4 m2 s−2 at 135°E to about −0.6 m2 s−2 at 170°E. For comparison with the depth-integrated transport, we define y1 = 7°N and y2 = 21°N in (1). With the above parameters, the NEC transport in the inline image layer reduced gravity model is highly correlated with the depth-integrated transport in ECMWF ORA-S3 (the correlation coefficients are higher than 0.75 with confidence level above 95%). The above approves that the inline image layer reduced gravity model well reproduces the variability of the depth-integrated NEC transports and therefore will be used in the following to simplify the discussion.

[10] The transport anomalies in the inline image layer reduced gravity model can be expressed as

display math(3)

[11] Here we define inline image and inline image to denote the transport anomalies induced by upper layer thickness anomalies and zonal velocity anomalies, respectively. To evaluate contributions of these two kinds of anomalies to the total NEC transport anomalies, the correlation coefficients between the depth-integrated NEC transport anomalies (TNEC) and Treg, inline image, and inline image both on monthly and interannual time scales are calculated and shown in Figures 6a and 6b. The correlations between TNEC and inline image on all longitudes are generally negative and low on both time scales. The negative correlations are because, as discussed in the following sections and also by Zhai and Hu [2012], associated with positive/negative NEC transport anomalies are negative/positive SSH anomalies in this region. On the contrary, the correlation coefficients between TNEC and inline image are all positive and high and nearly equal to those between TNEC and Treg. On the interannual time scale, the correlation coefficients between TNEC and inline image range from 0.75 to 0.89 and generally increase westward. Moreover, the RMS of inline image is larger than that of inline image and comparable with that of the total NEC transport anomalies (Figure 6c). This suggests that inline image contributes and correlates the most to TNEC. In other words, the NEC transport anomalies are mainly induced by the zonal velocity anomalies instead of fluctuations of the upper layer thickness. Therefore, the Treg can be well represented by inline image. In inline image, inline image equals to the mean upper layer thickness inline image, and it multiplying g/f basically decreases northward from 9 × 107 m2 s−1 to about 4.5 × 107 m2 s−1 over the NEC region. The ∂ η′/∂ y denotes the meridional gradient of SSH anomalies and ranges from −3 × 10−7 to about 3 × 10−7 over the NEC region. Therefore as a zeroth order approximation, the product of inline image and g/f can be assumed to be constant over the NEC region and the expression of Treg could then be further reduced to be

display math(4)

where C0 is constant. The above relation implies that the variability of the depth-integrated NEC transport is related to the SSH variations at the northern and southern boundaries. Instead of using y1 = 7°N and y2 = 21°N, we calculate the simultaneous linear correlations between monthly anomalies of the SSH and TNEC at each integer longitude and define the latitudes of the negative and positive maximum correlations along the same longitude as y1 and y2, respectively. The results show that y2 is around 20°N and y1 increases from 7°N at 135°E to about 9°N at 170°E, corresponding well to the North Pacific tropical gyre center.

Figure 6.

(a) Correlation coefficients of TNEC and Treg (solid black line), inline image (dashed black line), and inline image (gray line) for monthly anomalies. (b) Same as Figure 6a but for interannual anomalies. (c) Standard deviations of monthly anomalies of Treg (solid black line), inline image (dashed black line), and inline image (gray line). (d) Predictive skills of TNECp to TNEC in ECMWF ORA-S3 on (black line) monthly and (gray line) interannual time scales.

[12] Using above parameters, the proxy NEC transport anomaly can be constructed with the SSH anomaly following inline image (equation ((4))). The longitude-dependent a can be obtained through fitting the depth-integrated NEC transport monthly anomaly to the SSH monthly anomaly difference between y1 and y2 at the same longitude. The fitted a slightly increases eastward from about −40 Sv m−1 at 135°E to about −30 Sv m−1 at 170°E. As a is related to the product of inline image and g/f, its absolute value decreasing eastward results from the eastward shoaling of the thermocline. Figure 6d shows the longitudinal distribution of the predictive skills of the proxy NEC transport anomaly to the depth-integrated NEC transport anomaly on both monthly and interannual time scales; the predictive skill is defined as

display math(5)

[13] The longitudinal distributions of the correlation coefficients between TNEC and TNECp on monthly and interannual time scales are similar to those of the predictive skill and not shown. On monthly time scale, the predictive skill is about 0.4 to 0.6, implying linear correlation coefficient of about 0.6 to 0.7. On interannual time scale, the predictive skill is much higher, about 0.6 to 0.8, implying linear correlation coefficient of about 0.8 to 0.9. Therefore along the same longitude, the interannual variability of the depth-integrated NEC transport can be well represented by that of the SSH anomaly difference between the southern and northern boundaries.As the y1 and y2 well defined above are in the tropical gyre and subtropical gyre of the North Pacific Ocean, respectively, it is interesting to test whether the two terms inline image and inline image can serve as the proxy NEC transport anomaly south and north of the gyre boundary near the Philippine coast, respectively. For this test, we define the total NEC transport TPNEC and NEC transports entering the tropical gyre TPNEC-S and subtropical gyre TPNEC-N near the Philippine coast, respectively, as

display math(6)

where inline image defines the depth of 26.7σθ-isopycnal surface, Yb(z,t) defines the depth-time varying NEC bifurcation latitude, and the negative signs are chosen to make the transport westward positive. As the streamlines south and north of the bifurcation latitude enter the tropical gyre and subtropical gyre, respectively, Yb gives the gyre boundary near the Philippine coast. To capture the broad-scale circulation changes, Yb(z,t) is defined to be where the mean meridional velocity averaged within a 5°-longitude band east of the Philippine coast equals to 0 [Qu and Lukas, 2003] for each month and each depth layer and the presented transport values are all zonally averaged over Lx from 135°E to 140°E. TPNEC-S and TPNEC-N have mean values of 23.5 and 27.0 Sv, respectively. The mean value of TPNEC-S agrees well with the MC transport of ~27 Sv derived from repeated hydrographic sections at 8°N by Qu et al. [1998]. However, the mean value of TPNEC-N is much larger than the KC transport of ~14 Sv at 18°N derived by them. The reason may be that we define the northern boundary for TPNEC-N at 21°N, which is 3° north of 18°N. The monthly anomalies of TPNEC have correlation coefficients of 0.5 and 0.7 with those of TPNEC-S and TPNEC-N, respectively. However, its interannual anomalies have correlation coefficients of 0.8 and 0.6 with those of TPNEC-S and TPNEC-N, respectively. Moreover, the standard deviation of the interannual anomalies of TPNEC-S is 2.4 Sv, larger than that of TPNEC-N, which is 2.1 Sv. It seems that on the interannual time scale, the variation of the total NEC transport is more related to its southern branch. The linear correlation map of the monthly anomalies of TPNEC and SSH is shown in Figure 7a. With y1 and y2 shown as the black dots around 135°E, the proxy NEC transport anomaly is obtained as

display math(7)
Figure 7.

(a) Linear correlation map of the monthly anomalies of TPNEC and those of the SSH. (b) Time series of interannual anomalies of (solid line) TPNEC and (gray line) proxy NEC transports. (c) Anomalies of the (black line) TPNEC-N and (gray line) inline image. (d) Anomalies of (black line) TPNEC-S and (gray line) inline image.

[14] The time series of the interannual anomalies of TPNEC and TNECp are shown in Figure 7b. These two time series have a linear correlation coefficient of 0.85 with the confidence level above 95%, which means 72.3% of the interannual variance of TPNEC can be presented by the proxy NEC transport.

[15] The linear correlation coefficient between the interannual anomalies of TPNEC-N and inline image is about 0.56 (Figure 7c), while that between the interannual anomalies of TPNEC-S and inline image is about 0.82 (Figure 7d) both with the confidence level above 95%. Actually, the maximum correlation coefficient between the interannual anomalies of TPNEC-S and inline image is 0.84 when the former leads the latter by about 1 month. Clearly, on the interannual time scale, the SSH signals in the North Pacific tropical gyre center can serve as good transport proxy of the time-varying southern branch of the NEC near the Philippine coast. The relative low correlation between the interannual anomalies of TPNEC-N and inline image may be due to the complexity of the vertical ocean structure in the northern part of the NEC region [e.g., Qiu and Chen, 2010a]. In the following subsection, only the dynamics responsible for the interannual variability of the NEC transport to the tropical gyre near the Philippine coast is discussed.

3.3 Dynamics

[16] As discussed in the former subsection, the interannual variation of the NEC transport to the tropical gyre can be well explained by the SSH variation around the tropical gyre center in the framework of the inline image layer reduced gravity model. Therefore, the dynamics for the interannual variation of the NEC transport to the tropical gyre are those governing the interannual SSH variation around the tropical gyre center at the same longitude. In the tropical Pacific Ocean, the low frequency variability of the SSH is mainly induced by the sea surface wind forcing and can be quantified by the inline image layer reduced gravity model [Meyers, 1979; Kessler, 1990; Qiu and Joyce, 1992; Capotondi et al., 2003; Qiu and Chen, 2010b; Zhai and Hu, 2012]. Under the long wave approximation and assumptions of low frequency and quasi-geostrophy, the linear vorticity equation for the SSH anomaly is the following

display math(8)

[17] where CR = βλ2 is the speed of the long first mode baroclinic Rossby waves, β is the meridional gradient of the Coriolis parameter f, λ is the baroclinic Rossby radius, inline image is the wind stress vector anomaly, and ε is the Newtonian dissipation rate. For longitude-dependent ε(x) and CR(x), the solution of equation ((8)) can be obtained through integrating it along the Rossby wave characteristics from the eastern basin boundary

display math(9)

[18] where x = xe denotes the eastern basin boundary and inline image is the negative transit time needed for the first mode baroclinic Rossby waves generated in the east propagating westward to the target point x. inline image gives the ratio of the interannual anomaly of wind-induced Ekman pumping velocity to CR and will be denoted as Vr in the following discussion. The solution due to the eastern basin boundary is ignored in (9) because its influence is confined only a few Rossby radii away from the boundary [e.g., Fu and Qiu, 2002; Qiu and Chen, 2010b]. The baroclinic Rossby wave speed CR is calculated with the baroclinic Rossby radius λ derived by Chelton et al. [1998]. Monthly wind stress vector anomalies are from ECMWF ORA-S3 [Balmaseda et al., 2008]. With the above given parameters, the modeled η′ is a function of g′ and ε, which are empirically determined through comparing modeled SSH anomalies and those in the ECMWF ORA-S3. In (8), as the Newtonian dissipation is used to parameterize the momentum dissipation, which is assumed in the form of horizontal eddy diffusion [Qiu et al., 1997], ε is a function of longitude. However, for the purpose of simplicity and being easy to handle, ε is usually chosen as a constant in the literature [e.g., Capotondi et al., 2003; Qiu and Chen, 2010b; Hsin and Qiu, 2012; Zhai and Hu, 2012]. Though it is possible to obtain an optimal ε for each longitude, it is quite time consuming. In the current study, we try to obtain an optimal ε for one part of the North Pacific Ocean and another optimal ε for the other part. The optimal g′ is estimated to be 0.06 m s−2, while ε is 0 west of 180°E and 1/4 per month in the east. The estimated optimal ε east of 180°E is consistent with that derived by Hsin and Qiu [2012]. The relatively low ε estimated here indicates that the Rossby waves responsible for the interannual SSH variations in the tropical northwestern Pacific Ocean experience little dissipation along their way propagating westward. Figures 8a and 8b show the time-longitude plots of interannual SSH anomalies in ECMWF ORA-S3 and modeled ones within the latitude band of 7°N–9°N from January 1960 to December 2008. Overall, the linear vorticity equation well reproduces the low frequency variability of the SSH in ECMWF ORA-S3 in the tropical northwestern Pacific Ocean. As shown in Figure 9a, the correlation coefficients between the modeled SSH anomalies and those in ECMWF ORA-S3 decrease eastward from 0.80 or more at 137°E to 0.10 or less east of 110°W and are generally larger than 0.60 in the western-central part. The relatively low correlation coefficients in the eastern Pacific Ocean are probably because the influence from the eastern boundary is neglected in the model results.

Figure 8.

(a) Time-longitude plot of SSHA (cm) derived from ECMWF ORA-S3 in the latitude band of 7°N–9°N from January 1960 to December 2008. (b) Same as Figure 8a but for modeled SSHA (cm). (c) Same as Figure 8a but for Vr (×10−6). (d) Time series of Niño-3.4 SSTA (°C). In Figures 8a–c, the gray lines indicate the Rossby wave characteristics as examples.

Figure 9.

(a) Correlation coefficients between the modeled SSHA and those in ECMWF ORA-S3. (b) Explained percent of variance of the modeled SSHA by the cumulative wind forcing from (black line) 137ºE and (gray line) 160ºE to X.

[19] Given the good performance of the wind-induced inline image layer reduced gravity model in reproducing the interannual SSH variability, it is helpful to clarify the relative contributions of the wind forcing over different longitude bands. For that purpose, we examine the cumulative SSH variance S(xT,X) explained by the wind forcing as a function of longitude X for each integer longitude xT from 135°E to 170°E. Only results at 137°E and 160°E are displayed in Figure 9b as examples. S(xT,X) is defined as

display math(10)

[20] where inline image denotes the SSH anomaly at xT forced by the wind anomaly to the west of X. The results show that much of the interannual SSH variance is caused by the wind forcing in the western-central tropical North Pacific Ocean. In the far western part, for example, at 137°E, over 90% of the interannual SSH variance is caused by the wind forcing west of the date line, consistent with studies of Zhai and Hu [2012]. We then give the time-longitude plot of Vr in Figure 8c. It can be easily seen from Figure 8c that positive/negative SSH anomalies in the tropical northwestern Pacific Ocean always correspond to negative/positive Vr in the western-central Pacific Ocean. This is easy to understand because the downward/upward Ekman pumping always induces deepening/shoaling of the thermocline and rising/falling of the SSH, and thus triggers downwelling/upwelling Rossby waves propagating westward along its characteristics (gray lines in Figures 8a–8c) and eventually contributes to the SSH variability in the far western region.

[21] To further demonstrate the influence and relationship of the wind forcing at different parts of the North Pacific Ocean and to the SSH variability in the far western region, we calculate the lagged correlations between SSH anomalies averaged over 135°E–140°E and wind-induced Ekman pumping velocity anomalies averaged over different longitude bands in the east within the latitude band of 7°N–9°N (Figure 10). On monthly time scale, the SSH anomaly has maximum negative correlations with the wind forcing through the ocean basin by a time lag of about 5 months. The correlations are even higher in the western ocean basin. Around 160°E–190°E, the lagged time of the maximum negative correlations well corresponds to the time needed for the first mode baroclinic Rossby waves propagating to 135°E, implying a potential causal relationship [e.g., Kashino et al., 2011]. In the eastern ocean basin, however, the correlations may result from a shared seasonal variation instead of a causal relationship. On interannual time scale, the situation is more interesting that significant correlations only occur in the western ocean basin and the SSH anomalies lag the wind forcing. The maximum negative correlation coefficients range from −0.60 to −0.80 west of 180°E but are usually higher than −0.60 in the east. This is consistent with that the interannual SSH variability in the tropical northwestern Pacific Ocean is mainly controlled by the wind forcing in the western-central Pacific Ocean. We then define TLag to denote the lagged time for maximum negative correlations and also represent the time span from the wind forcing peak to the SSH anomaly peak. As shown in Figure 10, the TLag first decreases eastward from about 7 months over 135°E–140°E to about 4 months over 165°E–170°E and then increases eastward again to about 6 months over 175°E–180°E. Moreover, the TLag is generally larger than the time needed for the first mode baroclinic Rossby waves to propagate westward to impact the SSH fluctuations over 135°E–140°E. The exception takes place around 170°E where they equal to each other. This eastward decreasing TLag west of 170°E suggests that the center of the anomalous wind forcing move eastward in the tropical northwestern Pacific Ocean before the SSH gets its extreme value. This can also be seen in Figure 8c. For example, in 1988 and 1997, the centers of the significant negative and positive Ekman pumping velocity anomalies moved eastward with time before the SSH got maximum and minimum, respectively. That the TLag increases again from 165°E–170°E to 175°E–180°E may suggest that the easternmost position of the anomalous wind forcing center is around 170°E. East of 170°E, TLag increases eastward again; this implies another anomalous wind forcing center moves westward in this region. Indeed, there exists strong interannual wind forcing variability east of 200°E. However, as shown before and by the relatively low correlations, this wind forcing contributes little to the interannual SSH variability in the western part of the Pacific Ocean. As discussed in the following subsection, the interannual SSH variability in the far tropical northwestern Pacific Ocean and its relations to wind forcing throughout the Pacific Ocean basin are related to the interannual variability of the tropical ocean-atmosphere processes.

Figure 10.

Lagged correlations between SSH anomalies averaged over 135°E–140°E and wind-induced Ekman pumping velocity anomalies averaged over different longitude bands in the 7°N–9°N latitude band for (solid line) monthly and (dashed line) interannual anomalies. The gray lines indicate the time needed for the first mode baroclinic Rossby waves propagating westward to 135°E, and the black dots denote the lead-lag time for the maximum correlation.

3.4 Relation to ENSO Events

[22] As described in section 1, the interannual NEC transport variability is pointed out to be related to ENSO events by several previous studies [e.g., Qiu and Joyce, 1992; Qiu and Lukas, 1996; Kim et al., 2004; Zhai and Hu, 2012]. Because of the low temporal-spatial resolutions of observations or non-assimilation of observations into the models adopted by the previous studies, it is important to reexamine the connection of the interannual NEC transport variability to ENSO events using the current results. Here we adopt the sea surface temperature anomalies (SSTA) averaged over Niño-3.4 region (5°S–5°N, 120°–170°W) as the oceanic Niño index. The lead-lag correlation coefficients of the NEC transport anomalies with the Niño-3.4 SSTA are shown in Figure 11a. The maximum correlation coefficients are relatively low and decrease eastward from 0.45 at 135°E to about 0.10 at 170°E. The NEC transport anomalies lead the Niño index by 1–2 months over 138ºE to 156ºE, lag the latter by ~1 month east of 164ºE, and are in phase with the latter at the other longitudes. Though the correlation coefficients are low, the NEC transport in the far tropical northwestern Pacific Ocean tends to be larger/lower than the average level in most El Niño/La Niña years with the extreme values occurring in different phases for different events (Figure 11b). Thus, the low correlation coefficients between them may result from the impact of small NEC transport anomalies in non-ENSO years. On the other hand, not all the extremes in the transport anomalies are related to the El Niño/La Niña events, e.g., the extreme small transports during 1968 and 1978. However, there are indeed extreme large or small values in the Niño-3.4 SSTA during those periods.

Figure 11.

(a) Lead-lag correlation coefficients of the NEC transport anomalies with the Niño-3.4 SSTA. (b) Time series of NEC transport interannual anomalies at 137ºE and Niño-3.4 SSTA. (c) Depth-lag section of the cross correlation between the NEC BLA and the Niño-3.4 SSTA. (d) Lagged correlation coefficients between the Niño-3.4 SSTA and transport anomalies of the NEC entering the (solid line) North Pacific tropical and (dashed line) subtropical gyres at the Philippine coast. Positive (negative) values of the Y axis in Figure 11a and X axes in Figures 11c and 11d denote the time lagged (led) by the Niño-3.4 SSTA. In Figure 11a, the black line indicates the lead-lag time for the maximum correlation.

[23] As discussed in section 3.2, near the Philippine coast, the northern and southern branches of the NEC transport have different contributions to the total NEC transport variability. It is meaningful to clarify their separate connections to the Niño-3.4 SSTA. Figure 11c shows the correlation coefficient between the NEC bifurcation latitude anomaly (BLA) and the Niño-3.4 SSTA as a function of depth and time lag. Using the output of a high resolution OGCM, Kim et al. [2004] pointed out that the interannual variations of the NEC bifurcation latitude are highly correlated with Southern Oscillation Index (SOI), especially within the main thermocline, and the maximum correlation occurs with the SOI leading by about 1 month. However, our current results reveal that the maximum correlation occurs at much shallower depths with the Niño-3.4 SSTA slightly lagging. Figure 11d gives the correlations of the Niño-3.4 SSTA with the NEC transport anomalies entering the subtropical and tropical gyres near the Philippine coast, respectively. The results show that the interannual variability of the NEC southern branch is highly correlated with the Niño-3.4 SSTA. The correlation coefficient reaches 0.64 with the former leading the latter by about 1–2 months. The lead time is similar to that the total NEC transport anomalies lead the Niño-3.4 SSTA. On the contrary, the variation of the NEC northern branch seems to have no significant simultaneous correlation with the Niño-3.4 SSTA. The different connections of the NEC transport variability south and north of the gyre boundary to the Niño-3.4 SSTA can also be seen from the different connections of the SSH anomalies at 20°N and around the tropical gyre center to it (Figure 12). At 20°N, the interannual SSH variations have no significant correlation with the Niño-3.4 SSTA. However, around the tropical gyre center, the interannual SSH variations are highly correlated with the Niño-3.4 SSTA. Their correlation coefficients are lower than −0.80 in most regions west of 170°E, and the SSH anomaly extremes occur around the peak of the Niño-3.4 SSTA. Moreover, the linear correlation coefficient of the NEC BLA with the NEC transport anomaly entering the subtropical gyre is −0.40, while its correlation coefficient with that entering the tropical gyre is 0.60, both of which with confidence level above 95%. The high correlations between the Niño-3.4 SSTA and NEC BLA, the southern branch of the NEC transport, and the SSH all suggest that the tropical ocean-atmosphere processes have important influence on the interannual variability of the circulation in the tropical northwestern Pacific Ocean, which is more prominent during El Niño/La Niña events.

Figure 12.

Lagged cross correlation between the Niño-3.4 SSTA and SSHA around (solid line) the North Pacific tropical gyre center and (dashed line) at 20°N. Positive (negative) values of the X axis indicate the time lagged (led) by the Niño-3.4 SSTA. The gray dots denote the lead-lag time for the maximum correlation.

[24] In the above subsection, we have demonstrated that much of the interannual SSH variance in the far western part of the tropical gyre center is caused by anomalous wind forcing over the tropical northwestern Pacific Ocean. We will then test the relations of the anomalous wind forcing to ENSO events. As shown in Figure 13, the interannual variations of the wind forcing west of 200°E and near the eastern ocean basin boundary have closer correlations with the ENSO events than those around 200°E–230°E. The maximum correlations near the eastern basin boundary occur around the mature phase of the ENSO events, implying that the wind forcing there gets strongest accompanying the mature of the ENSO events. In the western ocean basin, however, the situation is quite different. The maximum correlation coefficients occur several months before the mature of the ENSO events. The time led by the wind forcing first decreases from about 8 months over 135°E–140°E to about 4 months over 165°E–170°E and then increases again to about 7 months over 195°E–200°E. These lagged times by the Niño-3.4 SSTA are well consistent with those by the SSH anomalies in the far western tropical gyre center as shown in Figure 10. The high positive correlations in the western ocean basin imply that there are positive/negative Ekman pumping velocity anomalies in this region during El Niño/La Niña years, which can be clearly seen from Figures 8c and 8d. The eastward decreasing lead time by the wind forcing from 135ºE–140ºE to 165ºE–170ºE suggests that the anomalous wind forcing center moves eastward in this region before the mature of ENSO events (also see Figures 8c and 8d).

Figure 13.

Lagged correlations of the Niño-3.4 SSTA with wind-induced Ekman pumping velocity anomalies over different parts of the North Pacific Ocean. Positive (negative) values of the X axis indicate the time lagged (led) by the Niño-3.4 SSTA. The gray dots denote the lead-lag time for the maximum correlations.

[25] Actually, though the interannual variations of the NEC transport entering the tropical gyre can be explained by those of the SSH around the tropical gyre center based on the inline image reduced gravity model, their exact modulations result from the SSH variations between the tropical gyre center and gyre boundary. This is mainly caused by westward propagating Rossby waves generated in the east by anomalous wind forcing related to ENSO events [e.g., Zhai and Hu, 2012]. The lagged correlations of Niño-3.4 SSTA with SSH anomalies at latitudes between the tropical gyre center and 19°N in the tropical northwestern Pacific Ocean are similar to its lagged correlations with SSH anomalies around the tropical gyre center (figures are not shown). Therefore, it is useful to clarify the spatial-temporal patterns of the anomalous wind forcing associated with the ENSO events that are responsible for the interannual SSH variations in the tropical northwestern Pacific Ocean. The upper six panels in Figure 14 display the regressed patterns of the wind stress vector anomalies and associated Ekman pumping velocity anomalies to the normalized Niño-3.4 SSTA with different months lagged by the Niño-3.4 SSTA. In the present study, the regressed pattern S(x,y,Δt) of a variable A(x,y,t) to a normalized time series B(t) at lead time Δt by the former is calculated as S(x,y,Δt) = A(x, y, t − Δt)/B(t) with the least squares fitting method. With B(t) being a normalized time series, S(x,y,Δt) has the unit of A(x,y,t). The regressed patterns of the anomalous wind forcing are similar to the composite results of Wang [1995] and Zhai and Hu [2012]. During the development and transition phases (about 3–9 months before the mature phase, see Wang and Weisberg [2000]) of El Niño/La Niña events, the tropical northwestern Pacific Ocean experiences developing of westerly/easterly wind anomalies and positive/negative wind stress curl anomalies north of the equator and eastward moving of the wind forcing center. Several months before the mature phase, the wind forcing center reaches its easternmost position around the 170°E, which is also indicated by the above lagged correlations between the wind forcing and the Niño-3.4 SSTA. As discussed before, these westerly/easterly wind anomalies along with the positive/negative wind stress curl anomalies in the north induce anomalous upward/downward Ekman pumping, which generates upwelling/downwelling Rossby waves that propagate westward to shoal/deepen the pycnocline and lower/elevate the SSH. As the westerly/easterly wind anomalies mainly develop south of about 15°N, the positive/negative wind stress curl anomalies in the north are mainly confined south of 20°N, and thus, the influence of El Niño/La Niña events is limited to low latitudes. During the left phases (the mature and decay phases), an anticyclone/cyclone forms near the Philippine coast in the far tropical northwestern Pacific Ocean, develops, and extends eastward [e.g., Wang et al., 1999; Wang and Weisberg, 2000; Wang, 2001a, 2001b; Wang and Picaut, 2004]. This anticyclone/cyclone and downwelling/upwelling Rossby waves generated about 1 year before in the far eastern tropical Pacific Ocean tend to deepen/shoal the pycnocline and elevate/lower the SSH in the far tropical northwestern Pacific Ocean.

Figure 14.

(upper six panels) Regressed patterns of the wind stress vector anomaly (N m−2) and associated Ekman pumping velocity anomaly (m s−1) to normalized Niño-3.4 SSTA with different months lagged by the Niño-3.4 SSTA. (lower six panels) Same as upper six panels but for the SSH anomalies (cm) and the horizontal current anomalies around 100 m deep (m s−1).

[26] With the above patterns of anomalies of the wind stress vector and wind-induced Ekman pumping velocity, the corresponding variations of the SSH and horizontal currents over much larger regions are clarified through regressing their anomalies to normalized Niño-3.4 SSTA. As shown in the lower six panels in Figure 14, significant SSH falling in the tropical northwestern Pacific Ocean can be seen about 6 months before the mature phase of El Niño events. With the development of positive Ekman pumping velocity anomalies, the SSH in the tropical northwestern Pacific Ocean south of 20°N and west of 170°E continues to fall and gets lowest around the mature phase. Accompanying the SSH falling, a cyclonic gyre anomaly forms around the center of SSH anomalies and strengthens with time. Though the SSH falling, which means thermocline shoaling at the same time, tends to reduce the NEC transport, the westward zonal velocity strengthens and thus induces increase in the NEC transport and northerly NEC bifurcation at the Philippine coast. During La Niña years, the situation is reversed.

4 Summary

[27] With the help of the ocean reanalysis data from ECMWF ORA-S3, the interannual variability of the NEC transports in the tropical northwestern Pacific Ocean is investigated. In general, both the amplitudes and RMS of the interannual transport anomalies increase from the east to the west. Temporally, the transport anomalies in the west slightly lag behind those in the east. It is shown that the NEC transport variation can be well represented by variations of the SSH anomaly difference between the southern and northern boundaries of the NEC region. The proxy transport anomaly constructed from the SSH anomaly difference is highly correlated with the depth-integrated transport anomaly and can be used as a predictor. Further analysis near the Philippine coast suggests that the above correspondence largely comes from that between the NEC southern branch and the SSH anomaly around the tropical gyre center. Using the linear vorticity equation governing the inline image layer reduced gravity model and appropriate parameters, much of the interannual SSH variance in the western North Pacific tropical gyre is found to be caused by the wind forcing over the tropical northwestern Pacific Ocean.

[28] The correlations between the Niño-3.4 SSTA and the NEC transport anomalies are relatively low and decrease eastward from 0.45 at 135°E to 0.10 at 170°E. During the development and transition phases of El Niño events, westerly wind anomalies between the equator and 10°N along with the positive wind stress curl anomalies south of 20°N develop and strengthen in the tropical northwestern Pacific Ocean. This anomalous wind forcing induces anomalous upward Ekman pumping, which generates upwelling Rossby waves propagating westward to shoal the pycnocline and lower the SSH. Several months before the mature phase, the wind forcing center reaches its easternmost position around 170°E. The cumulative effect of the westward propagating upwelling Rossby waves results in the lowest SSH around the mature phase mainly south of 20°N and west of 170°E. The SSH falling induces a cyclonic gyre anomaly around the center of SSH anomalies, which results in the increase of the NEC transport and northward shift of the NEC bifurcation near the Philippine coast. During the La Niña years, the situation is reversed.

Acknowledgments

[29] We are much obliged to Kentaro Ando for providing the Web link to the hydrographic observations at 137ºE obtained by Japan Meteorological Agency. We are also grateful to Bo Qiu for valuable discussion and Cong Yin for her efforts on the early manuscript. Detailed comments from two anonymous reviewers helped improve an early version of the manuscript. The present study is sponsored by the National Natural Science Foundation of China Major Project (grant 40890151) and the National Basic Research Program of China (grant 2012CB417401).

Ancillary