1 Introduction

As a connection between the tropical and subtropical gyres in the North Pacific Ocean, the North Equatorial Current (NEC) is of great importance for the ocean and climate (e.g., Nitani 1972; Hu and Cui 1991). The NEC flows westward in between the tropical gyre and subtropical gyre, bifurcates into two branches as it encounters the Philippine coast, and feeds the northward-flowing Kuroshio and southward-flowing Mindanao Current (MC). Numerous studies examined the nature, variability, and climate effects of the NEC through observational, theoretical, and numerical approaches (e.g., Meyers 1975; Hu 1989; Hu and Cui 1989; Hu and Cui 1991; Taft and Kessler 1991; Qiu and Joyce 1992; Qiu and Lukas 1996; Qu et al. 1998; Wang and Hu 2006; Kashino et al. 2009; Qiu and Chen 2012; Yan et al. 2014).

The interannual variability of NEC is linked to the El Niño–Southern Oscillation (ENSO) cycle (e.g., Lukas 1988; Qiu and Joyce 1992). Toole et al. (1990) found that the NEC strength was enhanced in the spring of 1988 (La Niña phase) as compared with that in the fall of 1987 (El Niño phase) through analyzing hydrographic data from the United States–People’s Republic of China Cooperative Studies of Air–Sea Interaction in the Tropical Western Pacific. On the basis of onboard observations, Kashino et al. (2009) showed that the NEC under El Niño conditions (late 2006) was stronger than that under La Niña conditions (early 2008), and they suggested that the dynamic height rather than the local wind variability contributed to the interannual variation of NEC. The transport of the southern branch of NEC was found to be highly related to ENSO via upwelling/downwelling Rossby waves generated by wind stress curl anomalies in the tropical northwestern Pacific Ocean (Zhai and Hu 2013). Besides the observation studies, some numerical studies also focused on the interannual variability of the NEC transport. Qiu and Lukas (1996) suggested that the NEC bifurcated at a lower (higher) latitude during La Niña (El Niño) years, but the seasonal variability of the NEC transport near the Philippine coast was not significant. As Qiu and Lukas (1996) pointed out, prior to El Niño events, the positive wind stress curl shifted the zero wind stress curl line northward and hence induced the northward shift of the NEC bifurcation 1 year later. Through analyzing outputs from an ocean general circulation model, Kim et al. (2004) found that the interannual variability of the NEC bifurcation latitude was highly correlated to the NEC transport and was mainly induced by ENSO-related westward propagation of Rossby waves from the central equatorial Pacific Ocean. Previous studies facilitated our understanding of the NEC, but it seems that the interannual variability has little been examined on the basis of sustained direct velocity measurements, and the response of NEC to the ENSO cycle needs further research.

The climate system in the Pacific Ocean shifted after the early 1990s (e.g., Merrifield 2011; Qiu and Chen 2012). Although the atmospheric Walker circulation above the tropical Pacific Ocean was weakened over the past several decades, it has been strengthened since the early 1990s (Merrifield 2011; Tokinaga et al. 2012a, b). As a result, the NEC bifurcation shifts southward after the early 1990s (Qiu and Chen 2010; Chen and Wu 2012). Qiu and Chen (2012) examined the shift of the sea level trend using tide gauge records and satellite altimeter data and pointed out that the NEC transport had an increasing trend after the early 1990s. But the trend of NEC and the vertical structure of the NEC trend need to be further examined by direct observations.

The objectives of this paper are twofold. Firstly, we study the interannual variability of the NEC transport and its relationship with ENSO. Secondly, the trend of the NEC transport during 1993–2008 is investigated. Both the interannual variability and 16-year trend are compared with those inferred from hydrological and satellite data. The rest of this paper is arranged as follows. Section 2 introduces the data and corresponding processing methods; Sect. 3 specifies the climatology, variability, and trend of the NEC; and Sect. 4 summarizes the major results.

2 Data and method

2.1 Data sets

Oceanic current data along the 137°E section between 8 and 18°N are provided by the Japan Oceanographic Data Center (JODC)/the Hydrographic Department, Maritime Safety Agency (see http://www.jodc.go.jp/NEW_JDOSS_HP/FETI_vector_doc_e.html). All of the data were collected by shipboard acoustic Doppler current profilers (SADCPs). The quality of the original data is initially controlled by the JODC (the quality flags are stored within the data sets). We then process the SADCP data as below: firstly exclude the outliers including maximum and minimum and then remove the values whose anomalies are 2.5 times greater than their standard deviations at different depths. As a result, 38 quality-controlled SADCP sections across 137°E remain. Figure 1 shows the latitudinal–temporal distribution of the quality controlled data. Most of the oceanic current profiles have five vertical layers in the upper 200-m ocean with resolution of about 40 m. The SADCP records are temporally uneven with general temporal interval of half a year.

Fig. 1
figure 1

Latitudinal–temporal distribution of SADCP stations along 137°E (big/small cycles denote stations with 10/5 vertical layers in the upper 200 m)

Full size image

Outputs of the European Centre for Medium-Range Weather Forecasts (ECMWF) Ocean Analysis/Reanalysis System 3 (ORA-S3) including temperature, current, and wind stress are also used (Balmaseda et al. 2008). They cover the global ocean spanning 53 years from January 1959 to December 2011, with a horizontal resolution of 1° × 1°.

In addition, the monthly Niño 3.4 index (defined as the SST anomaly averaged over 5°S–5°N, 120–170°W) estimated from optimum interpolation (OI) sea surface temperature (SST) is provided by the Climate Prediction Center at National Oceanic and Atmospheric Administration (NOAA)/National Weather Service (http://www.cpc.ncep.noaa.gov/data/indices/). Here Niño 3.4 index is applied to examine the relationship between the NEC transport and ENSO.

2.2 Estimation of NEC transport

As a result of the limited vertical range of SADCP data, the NEC transport referred to in this paper is defined as the westward transport across 137°E above 200-m depth between 8 and 18°N:

$${\text{tran}}_{\text{SADCP}} (t) = - \int\limits_{{8^{ \circ } {\text{N}}}}^{{18^{ \circ } {\text{N}}}} {} \int\limits_{{0\;{\text{m}}}}^{{200\;{\text{m}}}} {u_{\text{NEC}} (y,z,t)} {\text{d}}y\;{\text{d}}z,$$
(1)

where u NEC(y,z,t) is the westward velocity (negative value) across 137°E. To examine the interannual variability of the NEC, tranSADCP is then linearly interpolated to form a monthly NEC transport tranmon. Since the NEC is defined in a fixed region, variabilities induced by the potential northward/southward shift of the current axis are included. But it seems that the variation due to the meridional movement of the NEC is small relative to the interannual variability of NEC strength (figure not shown). In addition, because these SADCP measurements provide only snapshots of the NEC, eddy activities and/or small scale processes might influence the results. To investigate these possible influences, we estimate both the westward transport defined in the above equation and net transport of NEC. It shows that the difference between the westward transport and net transport is relatively small and that the interannual variabilities of the westward and net transports are consistent with each other (not shown).

The bottom of the NEC is obviously deeper than 200 m. To illustrate the complete NEC structure at 137°E, we employ the ECMWF ORA-S3 data and plot Fig. 2a showing the climatology of zonal current and temperature. We find that the isoline of −0.05 m s−1 (westward) is much deeper than 400 m. Though the variability of NEC below 200 m is suggested to be of much importance, there are two reasons that we focus on the upper 200-m NEC besides the data limitation.

Fig. 2
figure 2

a Climatological temperature (contours with interval of 1 °C) superimposed on climatological zonal current (color, m s−1) across 137.1°E in ECMWF ORA-S3. The bottom of the western Pacific warm pool (28 °C isotherm) is highlighted by the thick black line. b Rates of westward current transports integrated from surface to different depths over that integrated from surface to about 535 m. c Standard deviation of 13-month running mean zonal velocities across 137°E in ECMWF ORA-S3

Full size image

First, the transport in the upper 200-m ocean contributes the largest part of the NEC transport and its interannual variability. We integrate the westward velocity from surface to different depths, to calculate the westward transport in different layers, and then estimate the rate of different layer transport over the whole NEC transport (upper 535 m). It is revealed that the transport in the upper 200 m exceeds 70 % of the whole NEC transport (Fig. 2b). The standard deviation of 13-month running mean velocity across 137°E over 1993–2008 is also estimated on the basis of the monthly ECMWF ORA-S3 data. It shows that the interannual variability in the upper 200 m is much stronger than that in the ocean deeper than 200-m depth (Fig. 2c).

Second, the upper 200-m NEC plays an essential role in the climate system. As shown in Fig. 2, the bottom of the western Pacific warm pool (WPWP) and mixed layer are completely enclosed in the upper 200-m ocean. The thermocline is also located at around 200-m depth (Fig. 2a). As an interface between the ocean and atmosphere, the mixed layer is of critical importance, and zonal current in the mixed layer plays an essential role in the variability of WPWP and ENSO cycle (e.g., Picaut et al. 1996; Guan et al. 2013).

3 Results

3.1 Climatology

The zonal velocity across the 137°E section is averaged over all the 38 cruises during 1993–2008 to demonstrate the vertical structure of climatological NEC. As shown in Fig. 3, the mean zonal current in the upper 200-m layer is completely westward. It is interesting that the NEC transport in the direct observation is relatively stronger than those derived from hydrographic data. Qu et al. (1998) claimed that the mean transport of the NEC is 41 Sv. Through analyzing 40-year hydrographic observations from the Japan Meteorological Agency (JMA), Zhai and Hu (2013) recently suggested that the mean transport of the NEC is 51 Sv. But in the present study, the climatological speed averaged over 8–18°N and 0–200 m is 27 cm s−1. The NEC transport of the mean flow (the average of all the SADCP sections shown in Fig. 3) is about 58 Sv and the mean transport of all the individual flows (i.e., the mean value of the observed NEC transports) is approximately 63 Sv with a standard deviation of about 19 Sv. So the NEC transport observed by the SADCP seems to be relatively larger than previous studies. But this difference between documented NEC transports and our results is probably induced by the inconformity of our definitions (they applied different depth/longitudinal domain and net transport).

Fig. 3
figure 3

Zonal component of SADCP velocities (m s−1) across 137°E averaged over 1993–2008. Negative denotes westward direction

Full size image

Qu et al. (1998) reported that the NEC across 130°E in the upper 200 m had several velocity cores, including two cores in the upper 100 m at 9.5°N/12.5°N and one core in between 100 and 200 m with maximum NEC velocity less than 35 cm s−1. Kashino et al. (2009) also mentioned the multi-core structure of the snapshots of NEC (December 2006 and January 2008) in the upper 200 m between 8 and 13°N. The maximum westward velocity of the snapshot NEC during December 2006 exceeded 60 cm s−1 (Kashino et al. 2009). Here, the maximal velocity in the direct observed climatological NEC is about 47 cm s−1 (Fig. 3). Two surface velocity cores at 9 and 10.5°N and another core at 14°N, 140 m are found. These velocity cores extend northward with increasing depth, which is consistent with previous studies (Qu and Lukas 2003).

3.2 Interannual variability related to ENSO

NEC transport is estimated cruise by cruise through Eq. (1). Kashino et al. (2009) estimated the NEC transport (Ekman transport plus geostrophic transport relative to 1000 db) and found that the NEC is 61.1 (44.0) Sv in the December 2006 (January 2008). We have no observation in the exact same dates, but over the time nearby, as shown in Fig. 4a, the NEC transports are 64.0 Sv in the January of 2007 and 50.0 Sv in the August of 2008, which are close to the results reported by Kashino et al. (2009). During the 38 cruises, the maximum and minimum monthly NEC transports are 101 and 26 Sv, respectively (Fig. 4a). While the temporal distribution is uneven, the annual mean NEC transport features significant and periodic interannual variability, which changes between 40.7 and 85.9 Sv with standard deviation of 11.7 Sv (Fig. 4b). This is in good agreement with previous results derived from hydrographic data. The maximum geostrophic NEC transport is less than 70 Sv and the minimum more than 30 Sv with standard deviation of 6–16 Sv (Qu et al. 1998; Zhai and Hu 2013). Figure 4c shows the standard deviations of zonal velocities among the 38 cruises. It suggests that the zonal velocity standard deviations are much stronger in the NEC bands south of 10°N and north of 15°N, i.e., it is relatively weaker inside the NEC bifurcation band around 12°N (Qiu and Chen 2012). In the latitude band 11–14°N, the minimums of velocity standard deviations extend northward with increasing depth. It should be noted that possible bias induced by the seasonality or sample frequency exists, but it is effectively excluded after low-pass filtering (annual mean or 13-month running mean in the present study).

Fig. 4
figure 4

SADCP measured upper 200-m NEC transports (Sv, a), annual means with error bars (red) defined as the standard deviation of the samples in a calendar year (b), and standard deviations of zonal velocities (m s−1) of the 38 cruises (c) (color figure online)

Full size image

Many foregone studies have linked the interannual variability of NEC to ENSO (e.g., Lukas 1988; Toole et al. 1990; Qiu and Joyce 1992; Qiu and Lukas 1996; Qu et al. 1998; Kim et al. 2004; Kashino et al. 2009; Zhai and Hu 2013). The power spectra of tranmon and Niño 3.4 index are shown in Fig. 5. The peak period of the tranmon is around 4–5 years and the Niño 3.4 index has an interannual period of 4–8 years, implying a clear overlap of their frequencies on an interannual time scale.

Fig. 5
figure 5

Power spectrum density (PSD) of NEC transport anomaly and Niño 3.4 index with 95 % confidence interval

Full size image

As mentioned above, Kashino et al. (2009) found that the NEC under El Niño conditions was stronger than that under La Niña conditions on the basis of two cases in December 2006 and January 2008. Here we focus on the differences of the upper 200-m NEC between El Niño and La Niña conditions in composited observations. Four well-defined El Niño events (1994/1995, 1997/1998, 2002/2003, and 2006/2007) and four La Niña events (1995/1996, 1998/1999, 1999/2000, and 2007/2008) occurred during the 1993–2008 period. Since the El Niño and La Niña events are phase locked to boreal winters, here we composite NEC transports during October to the following February (Table 1). It is concluded that NEC transport under El Niño conditions is generally greater than that under La Niña conditions except for the La Niña event in 1999/2000. This is consistent with the result presented by Kashino et al. (2009), but the mean NEC transports in both the El Niño phase and La Niña phase are relatively stronger than their results: here it is 68 Sv under El Niño conditions and 59 Sv under La Niña conditions.

Table 1 NEC transports during El Niño/La Niña events
Full size table

It may be interesting to examine the difference in the NEC’s vertical structure between the warm and cold phases of ENSO. Here we composite the velocity across the 137°E section over El Niño/La Niña events during 1993–2008. As shown in Fig. 6, velocities between 11 and 18°N are predominantly intensified under El Niño conditions relative to that under normal or La Niña conditions. But it indicates that the westward currents between 8 and 11°N, especially around 9.6–10.6°N, are significantly weakened during El Niño events. NEC under La Niña conditions between 9.6 and 11.4°N is much stronger than that under El Niño conditions. Thus, the response of NEC has clear latitudinal differences.

Fig. 6
figure 6

Vertical structure of upper 200-m NEC velocity across 137.1°E composited over El Niño events and La Nina events (m s−1)

Full size image

To further demonstrate the NEC–ENSO relationship, we plot in Fig. 7 the 11-month running mean tranmon and Niño 3.4 index. The interannual variability of NEC transport is by and large tied to the Niño 3.4 index with a simultaneous correlation coefficient of about 0.46 at the 99 % confidence level. Qiu and Chen (2012) suggested that the transport of the north branch of the NEC lags the Niño 3.4 index by 9 months with a maximum linear correlation of 0.61, but Zhai and Hu (2013) claim that the NEC transport leads the Niño 3.4 index by 1–2 months. Here we find that the maximum linear correlation coefficient between tranmon and Niño 3.4 index is 0.66 when the NEC transport lags the Niño 3.4 index by 6 months.

Fig. 7
figure 7

Time series of 11-month low-pass filtered Niño 3.4 and the NEC transport. Both the variables are normalized by their standard deviations. R 90 and R 00 denote the correlation coefficients between them during the 1990s (left) and 2000s (right). Student’s t test shows that in both the 1990s and 2000s, a correlation coefficient above 0.30 is of 99 % confidence level

Full size image

The relationship between NEC transport and ENSO also involves decadal difference. As shown in Fig. 7, the relationship is close after 2000 (2000s) with a high simultaneous correlation coefficient of 0.75. But the condition during 1993–1999 (1990s) is different: their correlation coefficient in the 1990s is 0.25 below the 99 % confidence level.

Why is the simultaneous NEC–ENSO relationship relatively poorer in the 1990s? As suggested in previous studies, the ENSO cycle exerts impacts on the interannual variability of the NEC by wind stress curl anomaly-induced upwelling/downwelling Rossby waves in the tropical Pacific Ocean (e.g., Zhai and Hu 2013). We thus hypothesize that the poor correlation between ENSO and NEC in the 1990s might be due to the change of the wind stress field in response to ENSO cycles from the 1990s to 2000s. Zhai and Hu (2013) suggested that the interannual variability of the NEC is controlled by the first mode baroclinic Rossby wave dynamics and can be explained by the sea surface height (SSH) variation. They further solved the vorticity equation in the framework of a 1.5-layer reduced gravity model and obtained the solution (Zhai and Hu 2013). On the basis of their solution, it is suggested that the interannual variability of the NEC at latitude y 0 is determined mainly by the wind stress curl variation (integration of the wind stress curl) along the same latitude.

To examine the response of the wind field to the El Niño events in the 1990s and 2000s, the wind stress curl anomalies are composited over the El Niño periods in the 1990s and 2000s. As shown in Fig. 8, significant positive wind stress curl anomaly occurs in the NEC region centered at about 15°N west of the dateline during the El Niño events in the 2000s (2002/2003, 2004/2005, and 2006/2007). However, this wind stress curl anomaly is relatively weaker during the El Niño periods in the 1990s (1994/1995 and 1997/1998) compared with that in the 2000s. One can expect that the response of the NEC to the El Niño events in the 2000s should be much more significant than that to the El Niño events in the 1990s. Thus, we suggest that the decadal difference in the interannual relationship between ENSO and NEC might result from the response difference in wind stress curl to ENSO between the past two decades.

Fig. 8
figure 8

Maps of wind stress curl anomaly (10−7 N m−3) composited over El Niño events in the 1990s (1994/1995 and 1997/1998) and 2000s (2002/2003, 2004/2005, and 2006/2007). The climatology is defined as the average of wind stress curl during 1993–2008

Full size image

3.3 Linear trend

A previous study reported that the geostrophic NEC had been strengthened after the early 1990s (Qiu and Chen 2012). Here we examine the linear trend of the upper 200-m NEC in the repeated SADCP measurements. As mentioned above, the discrete and temporal asymmetrical NEC transport is linearly interpolated into monthly time series of traninterp. The 13-month running mean series of traninterp is then estimated as shown in Fig. 9a. The linear trend of the NEC transport is calculated as well with a least-squares method. It is revealed that the NEC transport has been intensified at a rate of about 0.47 Sv year−1. In other words, the NEC transport has increased by about 7.46 Sv over the 16 years. This is similar to the result derived from altimeter data: Qiu and Chen (2012) suggested that the NEC transport over 1993–2009 showed an increase of 8.3 Sv, i.e., 0.49 Sv year−1. We also calculate the seasonal mean transports (summer, May–July; winter, October–December) and group them into summer and winter series of NEC transports. To exclude the influence of seasonality on the estimation of linear trend, we calculate the linear trends of summer and winter transports separately. As shown in Fig. 9, both the summer and winter transports have an increasing trend: the summer and winter transports increased by 0.24 and 0.39 Sv year−1 during 1993–2008.

Fig. 9
figure 9

a Monthly, b summer mean (May–July), and c winter mean (October–December) of interpolated SADCP observed upper 200-m NEC transports and linear trends over 1993–2008

Full size image

To examine the vertical patterns in the linear trend of NEC, we display the velocity as a function of depth and time, and the linear trend of velocity as a function of latitude and depth in Fig. 10. It is suggested that the NEC in the latitude bands including 8.0–12.0°N, 14.5–15.2°N, and 15.8–18.0°N is clearly enhanced. But the NEC in the latitude band 12.5°N–14.5°N is significantly weakened. The most intensification occurs in the 100–200-m layer around 8, 11, and 16.5°N. Qiu and Chen (2012) claimed that the increase of NEC transport is mostly due to the strengthening of the NEC south of 12.5°N. However, here we find that the NEC in the latitude band north of 15°N is also intensified significantly. The difference might be induced by the poor representation of the subsurface (here is 100–200 m) current trend in the surface geostrophic current.

Fig. 10
figure 10

a Meridional averaged zonal velocity across 137°E and b vertical structure of the linear trend of zonal velocities in the upper 200-m ocean

Full size image

The dynamics that are responsible for the SADCP observed strengthening of the NEC and the vertical structure of NEC’s linear trend might be the long-term change in the wind field. Figure 11 delineates the spatial pattern of wind stress difference between the 2000s and 1990s and implies that the easterly trade winds in the tropical Pacific Ocean, i.e., the sea surface branch of the atmospheric Walker circulation, have been sharply intensified. This is in concert with the linear trend of wind stress in the tropical Pacific Ocean (e.g., Merrifield 2011). As Qiu and Chen (2012) suggested, the increasing of NEC transport might be a result of the strengthened atmospheric Walker circulation.

Fig. 11
figure 11

ECMWF ORA-S3 wind stress vector (arrows) and its meridional components (color) in the 2000s minus that in the 1990s. The blue line marks the 137°E section (color figure online)

Full size image

However, this intensification is slightly canceled by the Ekman current anomalies in the Ekman layer. According to the Ekman theory (1905), the zonal component of the Ekman current velocity u E is determined by the meridional wind stress \(\tau_{y}\):

$$\int\limits_{{ - D_{\text{E}} }}^{0} {u_{\text{E}} (z){\text{d}}z} \approx \frac{{\tau_{y} }}{{\rho_{\text{E}} f}},$$
(2)

where D E is the Ekman layer depth (74 m at 15°N under 5 m s−1 wind speed), ρ E the density of seawater (typically 1025 kg m−3), and f the Coriolis parameter (\(3. 6\times 1 0^{ - 5} \;{\text{s}}^{ - 1}\) at 15°N). As shown in Fig. 11, the northward wind stress anomaly in the 2000s is stronger than that in the 1990s in the NEC region. At the 137°E section, this northward wind stress anomaly is about \(0.01\;{\text{N}}\;{\text{m}}^{ - 2}\). One can expect that the intensification of northward wind stress shall give rise to an enhanced eastward Ekman transport. According to Eq. (2), an anomalous τ y of about \(0.01\;{\text{N}}\;{\text{m}}^{ - 2}\) drives a change of the Ekman current velocity u E:

$$\int\limits_{{ - {\text{D}}_{\text{E}} }}^{0} {u_{\text{E}} (z){\text{d}}z} \approx 0.26\;{\text{m}}^{2} \;{\text{s}}^{ - 1} .$$
(3)

In other words, supposing the Ekman layer depth is 74 m, the strengthening of northward wind stress causes a vertically averaged eastward current anomaly of about 0.35 cm s−1 every 8 years (from the 1990s to 2000s). This partly explains the vertical difference in linear velocity trends between the 0–100-m layer and 100–200-m layer (Fig. 10b).

4 Summary and discussion

Variability of the Pacific NEC is intensively studied using satellite records, hydrographic data, and simulations. But the NEC variabilities in direct observations remain unclear. Based on 16-year repeated SADCP measurements, the present paper aims to characterize and understand the interannual variability and long-term trend of the NEC across 137°E during 1993–2008.

The vertical structure of the NEC was described in some previous studies with hydrographic measurements (e.g., Qiu and Joyce 1992; Qu et al. 1998; Zhai and Hu 2013) and instantaneous SADCP observations (Kashino et al. 2009). In this study, we examined the mean state of the repeated SADCP observations and found that the climatological NEC has a maximal velocity of about 47 cm s−1 and shows a multi-core structure: two surface velocity cores at 9 and 10.5°N and another core at 14°N, 140 m depth, in concert with previous studies (e.g., Qu et al. 1998; Kashino et al. 2009). The average NEC transport in the upper 200-m layer is 63 Sv and relatively larger than the documented results derived from hydrographic data.

NEC transport features significant and periodic interannual variability with a significant period of 4–5 years and varies between 26 and 101 Sv with standard deviation of 11.7 Sv. We further show the features of the latitudinal distribution of the zonal velocity variations. By performing composite analysis, we find that the average of NEC transport under the El Niño condition is 68 Sv and greater than that under the La Niña conditions which is 59 Sv. This is similar to the documented case study by Kashino et al. (2009).

Vertical velocity structures are composited over the warm and cold ENSO phases, and we find that the response of NEC features clear latitudinal differences. In terms of the lead–lag relationship between NEC and ENSO, controversy exists in previous studies. Qiu and Chen (2012) pointed out that the NEC transport lags the Niño 3.4 index by 9 months, but Zhai and Hu (2013) suggested that the NEC transport leads the Niño 3.4 index by 1–2 months. In the present paper, we find that the NEC transport lags the Niño 3.4 index by 6 months with a correlation coefficient of 0.66.

Although the interannual variability of NEC is closely related to the Niño 3.4 index, the NEC–ENSO relationship shows decadal difference between the 1990s and 2000s. The correlation coefficient is 0.75 (above the 99 % confidence level) in the 2000s but 0.25 (below the 99 % confidence level) in the 1990s. We examined the wind stress curl field under the El Niño conditions in both the 1990s and 2000s. It turns out that the response of wind stress curl to the El Niño in the 2000s is quite different from that in the 1990s: positive wind stress curl anomaly in the El Niño events in the 2000s is much stronger than that in the 1990s. This implies that the low-frequency variability in the wind field is important in regulating the NEC–ENSO relationship.

We then investigated the trend of the NEC velocity and its vertical pattern during 1993–2008 following Qiu and Chen’s work (2012). It is suggested that the NEC transport has been intensified with a linear trend of 0.24 Sv year−1 in the summer season and 0.39 Sv year−1 in winter during 1993–2008.

The linear trend of the NEC in this period shows significant meridional and vertical diversity. It is interesting that the strengthening of NEC in the 100–200-m layer is more intense than in the 0–100-m layer. In the latitude bands including 8–12°N, 14.5–15.2°N, and 15.8–18°N, the NEC has been clearly enhanced, but in the latitude band 12.5–14.5°N, it has been significantly weakened. Qiu and Chen (2012) pointed out that the increase of NEC transport is mostly due to the strengthening of the NEC south of 12.5°N, but we found that the NEC intensification is also significant north of 15°N.

To gain an insight into the dynamics responsible for the NEC’s linear trend, we finally checked the wind stress difference between the 2000s and 1990s and found that the strengthening of the NEC results from the strengthened atmospheric Walker circulation since the early 1990s. This is in agreement with the result reported by Qiu and Chen (2012). However, because the northward wind stress anomaly has increased clearly in the tropical Pacific Ocean, an eastward Ekman current anomaly occurs in the NEC region and slightly canceled the intensification of the NEC in the Ekman layer. This partly explains the vertical difference in NEC intensifications between the upper 0–100-m layer and bottom 100–200-m layer.

However, problems remain. First, the sparse SADCP observations presented here cannot capture the detailed evolution of the NEC following the ENSO cycle. The NEC and its variability of course cannot be fully captured here by the upper 200-m observations. Because of the limitation of the data (sample frequency and size), there is possible bias that we are unable exclude completely. Second, the linear trend during 1993–2008 is possibly influenced by the decadal or inter-decadal variabilities. But this issue is not further pursued here because of the limited temporal length of the available observation record. Finally, western boundary currents in the Philippine Sea show significant intraseasonal variability associated with mesoscale eddies (Hu et al. 2013) and the NEC intraseasonal variability might also be very important (Kashino et al. 2009). But the role of mesoscale eddies in the intraseasonal variability of NEC is unclear and will be considered in future work.