1 Introduction

The South China Sea (SCS) is a semi-enclosed marginal sea in the tropical western Pacific Ocean, and it connects to the East China Sea, the Pacific Ocean, the Sulu Sea and the Java Sea mainly through Taiwan Strait, Luzon Strait, Mindoro Strait and Karimata Strait, respectively. Under the influence of the East Asian Monsoon, the basin scale upper layer circulation in the SCS is cyclonic in winter and anticyclonic in summer, basically follows the wind-forced Rossby wave mechanism (Wyrtki 1961; Liu et al. 2001, 2008). The most prominent features of the summer upper layer circulation are the eastward offshore jet around 12°N against the Vietnam coast and the associated dipole structures (Metzger and Hurlburt 1996; Fang et al. 1998; Wu et al. 1998; Fang et al. 2002; Wang et al. 2006c). The eastward jet is also called the Vietnam Coastal Current, and the dipole structure is also phrased as double gyre or eddy pair in some literatures. The upwelling and advection accompanied by the summer circulation causes an obvious semiannual cycle and interannual variation of the sea surface temperature (SST) (Xie et al. 2003; Wang et al. 2006a), influences the upper layer thermohaline structure (Chen et al. 2010), reduces the local wind speed at the sea surface and enhances the air-sea interaction (Xie et al. 2007), and affects chlorophyll-a concentrations (Kuo et al. 2000; Tang et al. 2004; Liu et al. 2012; Chen et al. 2014). Thus, extensive researches have been carried out to study the formation and variability of the SCS summer circulation, because of its pronounced effects on the oceanic and atmospheric environment.

The formation of the eastward jet and dipole structure is mainly due to the orographic local wind stress curl by Annam mountain in Vietnam (Xu et al. 2008), and their variations are sensitive to the strength, direction and zonal asymmetry of the wind jet in an idealized rectangle basin (Wang et al. 2006c). While considering realistic topography, the interaction between wind-driven coastal currents and coastal promontory is another key factor, instead of wind stress curl alone, that controls the coastal jet separation around Vietnam at 12°N (Gan and Qu 2008). According to the western boundary current separation formula of Marshall and Tansley (2001), the critical separation velocity in this region is 0.29 m/s, which only occurs during strong southwesterly summer monsoon (Dippner et al. 2013). Besides local wind and topography, the encounter of two buoyancy-driven coastal current and tidal-rectification near Vietnam coast might also be favorable factors for the offshore current separation (Chen et al. 2012), and nonlinear advection would affect the strength and separation location of the current (Cai et al. 2007) as well.

The eastward jet also shows distinct variability from intra-seasonal to inter-decadal time scale, because of its sensitivity to the SCS summer monsoon. Each year, the development of the jet usually experiences 2–3 intra-seasonal cycles instead of simply persists through the summer monsoon, which is possibly owing to its response to Madden–Julian Oscillation (Xie et al. 2007). On interannual time scale, the stronger/weaker and southward/northward migration of the eastward jet at the developing/decaying stage of an El Niño event and the disappearance of the eddy pair during the strong El Niño to La Niña transition year are found to be mainly related to the southwesterly/northeasterly anomaly of the SCS summer monsoon (Chen and Wang 2014; Li et al. 2014; Chu et al. 2017), while the oceanic internal variability (such as the eddy-mean flow interaction) is found to be less important (Yang et al. 2015) and is estimated to contribute about 20% of the variation (Li et al. 2014). The inter-decadal variability of the eastward jet is closely related to the pattern and strength of the Asian summer monsoon (Wang et al. 2010) modulated by Pacific Decadal Oscillation (PDO) and Atlantic multidecadal oscillation (AMO) (Fang et al. 2014; Fan et al. 2018), and it is also influenced by the remote signals via Kelvin waves from the Western tropical Pacific through Sulu Sea (Fang et al. 2014).

Nevertheless, most previous studies focused on the role of wind on the change of the SCS upper layer circulation in summer, and there were few discussions on the contribution of Luzon strait transport (LST) (i.e. South China Sea Through Flow, SCSTF), despite that its variation is closely related to ENSO and PDO (Qu et al. 2004; Wang et al. 2006b; Yu and Qu 2013). Recent studies show that the LST is equally important compared to the local wind for the formation of the SCS western boundary current (SCSWBC) (Chen and Xue 2014), and the interannual variation of SCSWBC is related to the LST (Quan et al. 2016). Moreover, the advection of planetary vorticity through lateral boundaries (mainly through Luzon strait) is found to be an important vorticity source of the upper layer cyclonic circulation, especially for the northern SCS (NSCS) in summer (Gan et al. 2016; Zhu et al. 2017; Quan and Xue 2018). The lateral boundary forcing is shown to have greater influence on the cyclonic circulation in summer than that in winter (Chu et al. 1999; Gan et al. 2016). Thus, when the change of Kuroshio intrusion and LST contributes obviously to the intensified interannual variation of the circulation in the southern SCS (SSCS) in winter through SCSWBC advection (Zu et al. 2019), we would suggest that it will also influence the interannual variation of the SCS upper layer circulation in summer.

It is known that SCS summer monsoon (SCSSM) (Wang et al. 2009; Zhou and Chan 2007) and the LST (Qu et al. 2004; Wang et al. 2006b) are greatly modulated by the El Niño-Southern Oscillation (ENSO), and the interannual variation of the summer circulation is significantly influenced by the SCSSM (Chen and Wang 2014; Li et al. 2014; Chu et al. 2017), but whether the LST plays an important role and how much contribution it does to the interannual change of the SCS summer circulation remains largely unclear.

In this paper, the characteristics of interannual variation of the summer circulation are re-explored, a prominent anti-correlation between an eastward jet (eastward branch: EB for short) and a northward branch (NB for short) along the western boundary is revealed, and their variation with the SCS summer monsoon and LST are discussed. The remainder of the paper is organized as follows. Section 2 gives a description on the data and model utilized, Sect. 3 shows the interannual variation of the summer circulation, Sect. 4 explores the relative contribution of local wind and lateral boundary advection and their responses to ENSO complexity, Sect. 5 summarize the findings.

2 Data and model

2.1 Data

Two types of observed current data are used to study the interannual variation of the summer circulation in the SCS. One is surface geostrophic current derived from 1/4 × 1/4°gridded daily sea surface height (SSH) data over 20 years (1993–2016). The SSH data is merged from multi-altimetry products of TOPEX/Poseidon (T/P), Jason-1, Jason-2, ENVISAT, GFO, and ERS-1/2, with tidal corrections and mean dynamic topography added [provided by the Archiving, Validation, and Interpretation of Satellite Oceanographic data (AVISO)]. And the surface geostrophic currents are calculated using the horizontal gradients of SSH as the following,

$$(u_{g} ,v_{g} ) = \frac{g}{f}\left( { - \frac{\partial SSH}{\partial y},\frac{\partial SSH}{\partial x}} \right)$$
(1)

in which g is the gravitational acceleration, and f is the Coriolis parameter.

The other type of current data is velocity profile obtained from two Acoustic Doppler Current Profilers (ADCP) mooring stations from 2007 to 2016. One of the station is located in the NSCS (at 110.3°E, 17.1°N) on the path of northern branch of the western boundary current, and the other one is in the SSCS basin (at 113°E, 9.7°N) on the path of the eastward jet. The velocities from the moorings are smoothed by using a window of 30-days to remove high frequency oscillations. The velocity from mooring station in NSCS reveals the year to year variability of the SCSWBC current, especially in the summer (Shu et al. 2016). By comparing the simultaneous observation from the mooring station in the SSCS, we find an anti-correlated interannual variation between these two stations, indicating a competitive relationship between the NB and EB in summer.

The 1/4 × 1/4°, daily averaged wind field at 10 m from the Cross-Calibrated Multi-Platform (CCMP) ocean surface wind product (Atlas et al. 2011), and three dimensional daily current product of Hybrid Coordinate Ocean Model + Navy Coupled Ocean Data Assimilation (HYCOM + NCODA) global 1/12°Analysis (GLBu0.08, http://www.hycom.org/dataserver/gofs-3pt0/reanalysis) are used in this study to explore the effects of local wind and LST on the interannual variation of the summer circulation.

2.2 Model

The Advanced Taiwan Ocean Prediction (ATOP) model (Oey et al. 2013) is utilized in this paper to explore the influence of LST to the anti-correlated interannual variation of the EB and NB. ATOP is developed from Princeton Ocean Model (POM), which covers the entire North Pacific Ocean and SCS, with a horizontal resolution of 0.1° × 0.1° and 41 vertical sigma levels. It has been successfully used to study the interaction between the SCS through flow (SCSTF) and the Indonesian through flow (ITF) (Wei et al. 2016; Li et al. 2019), and detailed description of the model setup could be found in Oey et al. (2013).

To evaluate the effects from LST, numerical sensitivity experiment by closing the LS was carried out. The sensitivity experiments were restarted from Jan. 2004 and run for additional 8 years, and the monthly mean results of July and August in 2009 and 2010 are utilized for analysis.

3 Result

3.1 General feature of the summer circulation

The monthly climatology mean surface geostrophic currents and wind stress curl from June to September (Fig. 1) show the general evolution of the summer circulation in response to the summer monsoon in the SCS. After the break of the SCS summer monsoon in May, a northward flow is formed along the western boundary of the SCS in May. The magnitude of the flow increases from June to August as the SCS summer monsoon is enhanced, and an eastward jet forms around 12°N against the Vietnam coast and becomes stronger since July. In September, when the southwesterly summer monsoon starts to change into northeasterly winter monsoon in the NSCS, the circulation pattern varies with that from June to August. The flow along the west boundary in the NSCS turns to southward or southwestward, and a well-developed cyclonic circulation appears north to the eastward jet. Together with the anticyclonic circulation south of it, they form the dipole structures. Similarly, the monthly climatology temperature and salinity (data from SCSPOD15, Zeng et al. 2016) at 30 m depth in September also show different structure with those from June to August. There is colder and saltier upwelling water along the western boundary north of 11°N near Vietnam coast from June to August, but an obvious dipole structure in September, with colder and saltier core in the north, and warmer and fresher core in the south of the eastward jet. Thus, the average of the circulation in July and August are used to represent the mature pattern of the SCS summer circulation in this paper.

Fig. 1
figure 1

Monthly climatology temperature (upper row in color shading: °C) at 30 m depth, superimposed with monthly climatology surface geostrophic current (current in the SCS basin where water depth is shallower than 75 m is not shown, vectors: m/s) and monthly climatology salinity (lower row in color shading) at 30 m depth, superimposed with monthly climatology wind stress at 10 m height from June to September. The temperature and salinity data is from SCSPOD15, which is obtained by merging Argo float, WOD94 and cruise observations. The surface geostrophic currents are obtained from AVISO. The wind stress is obtained from CCMP. TS, LS, MS, and KS in a stand for Taiwan Strait, Luzon Strait, Mindoro Strait, and Karimata Strait respectively. The white lines in a indicate the − 200 and − 1000 m isobaths

Full size image

3.2 Variability of the summer circulation

The climatology of the mature pattern of summer circulation, averaged by the monthly mean surface geostrophic current of July and August from 1993 to 2016, shows clearly a bifurcation of the current near the Vietnam coast, with a strong eastward jet separating from the northward boundary current and a weaker branch continuing flowing northward (Fig. 2a). A part of the eastward jet flows southward as an anticyclonic recirculation that joins the WBC in the SSCS. There are two strong negative vorticity cores in the anticyclonic circulation in the SSCS, while there is no evident cyclonic circulation or positive vorticity core to the north of the eastward jet. The first EOF mode of the summer surface kinetic energy (Fig. 2b) and velocity speed (supporting information Fig. S1) both show distinct competition between the NB and EB (eastward jet), and they account for 28.5% and 22.5% of the total variance, respectively. This pattern indicates that when the NB is enhanced, the EB and anticyclonic circulation in the SSCS is weakened, and vice versa. Besides, it also implies that the variation of the boundary current along the western SCS in summer is mainly regulated by the competitive bifurcation of the WBC against the Vietnam coast around 12°N. Likewise, the first EOF mode of the summer surface velocity speed from HYCOM simulation GLBu0.08 show similar pattern as those derived from AVISO, and it accounts for about 28% of the total variance (Fig. 6c). The anti-correlation of the strength of the NB and EB is also clearly observed from the spatial correlation of their mean speed with the whole SCS region (supporting information Fig. S1).

Fig. 2
figure 2

a Climatology surface geostrophic current (vectors: m/s) and related vorticity (color shading: s−1) in summer, b the first EOF mode of summer surface kinetic energy per unit mass, and it accounts for 28.51% of the total variance, c the normalized temporal coefficient of the first EOF mode (PC-1: black lines) superimposed on Nino3.4 index, x-axis is the summer of each year. The red bars indicate index greater than 1 and the blue bars indicate the index less than − l. In a and b, the surface geostrophic current and the first EOF mode of the summer surface kinetic energy in the SCS basin where water depth is shallower than 75 m are not shown. The red squares in a show the locations of the two ADCP moorings in Fig. 3, and the red lines show the locations of the section for velocity transport along the pathways of the EB, NB and WBC in the SSCS in Fig. 4 The pink lines in a depict the main branches (eastward and northward) of climatology surface geostrophic flow in summer. The mean summer flow and kinetic energy are obtained from the monthly surface geostrophic current from AVISO in July and August

Full size image

The temporal coefficient of the first EOF mode (PC1) of the summer surface kinetic energy shows correlation with the Nino3.4 index (Fig. 2c). They have better negative correlation at strong ENSO cycles (i.e. an El Niño event followed by a La Niña event in 1994–1995, 1997–1998, 2009–2010), which suggests that the EB is enhanced during El Niño developing phase, while the NB is enhanced during El Niño decaying and La Niña developing phase. Previous results also pointed out a stronger/weaker and southward/northward movement of the eastward jet during developing/decaying stage of an El Niño event (Chen and Wang 2014; Li et al. 2014), or even disappeared in transition summer from strong El Niño to La Niña, such as in 2010 (Chu et al. 2017). However, the PC1 does not show clear relationship with the Nino3.4 at the second El Niño or La Niña developing phase during multiyear El Niño or La Niña events (i.e. 1999, 2011, 2015). This suggests that strong ENSO signals play important roles on the interannual correlation of the eastward and northward branches of the summer circulation, but the year to year variations of summer circulation are not solely influenced by ENSO. Li et al. (2014) and Yang et al. (2015) pointed out that besides southwesterly summer monsoon, ocean internal variability also contributes to the variation of the eastward jet, and this together with other factors will be further discussed in Sect. 4.

The anti-correlated interannual variations of the EB and NB are also demonstrated by the current observations from ADCP mooring station N and S (their locations are shown in Fig. 2a). Station N is on the path of western boundary current in the NSCS, the positive along-isobath velocity in station N represents northeastward flow while the negative value represents southwestward flow (left column in Fig. 3). Station S is on the path of the eastward jet in the SSCS. The negative value indicates southward extension associated with the jet. At station N, a stronger northeastward flow exists in 2007, 2010, and 2013, vertically it ranges from surface to more than 300 m deep, and its magnitude could reach over 0.5 m/s at 50 m depth, while in 2009 and 2012, the northeastward flow is greatly weakened to less than 0.15 m/s or even changes to southwestward flow sometimes. On the contrary, at station S no obvious southward flow appears in 2007, 2010, and 2013, but strong southward flow occurs in 2009 and 2012 with magnitude greater than 0.5 m/s and depth less than 300 m. Thus, it is evident that when the NB is stronger, the EB is weaker, and vice versa. Summer mean surface geostrophic currents from AVISO, upper layer circulation from ATOP and HYCOM simulations in these years all show similar anti-correlation between the EB and NB that is revealed by the ADCP mooring stations (supporting information Fig. S2–S4). Stronger NB appears in 2007, 2010 and 2013, while stronger EB and associated anticyclonic circulation in the SSCS exists in 2009 and 2012. Although the N and S ADCP mooring stations are only two points, they are quite representative in demonstrating the variation of EB and NB, as they are located in the regions of significant difference between EB and NB (Fig. 5c).

Fig. 3
figure 3

Interannual variation of the along-isobath component (positive is northeastward) of the current (m/s, left column) observed at the ADCP mooring station N in the NSCS and the meridional component (positive is northward) of the current (m/s, right column) observed at the ADCP mooring station S in the SSCS. (The locations ofthe stations are marked in Figs. 2a and 5c). The x-axis denotes the time from June to September. The y-axis is the water depth. The blank areas show the time and depth without observation. White contour lines indicate the velocity from − 1 to 1 m/s at an interval of 0.05 m/s

Full size image

Additionally, the relationship between the monthly mean velocity transport integrated for the upper 100 m at different sections (locations shown on Fig. 2a) along the pathways of the EB, NB and WBC also show a clear anti-correlation between the EB and NB, a positive correlation between the EB and WBC, while a weak negative correlation between the NB and WBC (Fig. 4). Their correlations suggest that this feature is partly related to the anticyclonic recirculation on the south of the EB, which joins and reinforces the WBC and consequently the EB, thus keeps the negative vorticity in the SSCS and is not favorable of the formation of NB. Although strongly related with the EB, the NB is weakly correlated with the WBC, and this indicates that the NB is not only determined by its upstream strength, other processes, such as LST and eddy activities in the NSCS might affect it.

Fig. 4
figure 4

Relationship between the monthly mean transports integrated at different sections for the current in the upper 100 m obtained from HYCOM GLBu0.08. a The transports of the northward branch and the eastward branch, b the transports of the eastward branch and the WBC in the SSCS, c the transports ofthe northward branch and the WBC in the SSCS

Full size image

3.3 Composite patterns of eastward and northward flow

Surface geostrophic currents averaged in summer (July and August) in years with larger positive/negative value of normalized temporal coefficients of the first EOF mode (according to Fig. 2c) are selected for a composite analysis. Same years would be selected out for composite, if we use 1 standard deviation of the PC-1 as the threshold. Current composited by 1994, 1997, 2009, 2012, 2014 (with normalized PC-1 smaller than − 1) show stronger EB (Fig. 5a), and those composited by 1995, 1998, 2007, 2010 (with normalized PC-1 larger than 1) show stronger NB (Fig. 5b). There is a prominent anticyclonic circulation in the SSCS for the EB pattern, but no evident boundary current in the NSCS. However, for the NB pattern, a continuous western boundary current flowing from the southern basin to the northern shelf is obvious, and the eastward jet together with the anticyclonic circulation in the southern basin are greatly weakened, instead a clear negative vorticity core appears around 16°N in the NSCS. The differences between EB and NB (NB-EB) that past the t test (Fig. 5c) show a significant westward flow towards Vietnam coast between 11 and 12°N, with cyclonic circulations south of it, and anticyclonic ones north of it. The most prominent variation is along the western boundary of the SCS basin.

Fig. 5
figure 5

Composite of the summer surface geostrophic current (vectors: m/s) and related vorticity (color contours: s−1) derived from altimetry data from AVISO. a For eastward branch and b for northward branch. c The difference between the northward and eastward branch that past the t-test (shaded region). Eastward branch is the composite of currents in 1994, 1997, 2009, 2012, 2014, and northward branch is the composite of currents in 1995, 1998, 2007, 2010. Selected years have large positive/negative values of the temporal coefficients of the first EOF mode of the KE as shown in Fig. 2c. Pink lines in a and b depict the major flow pattern in eastward and northward dominant patterns, respectively. The green squares in c show the locations of the two ADCP moorings in Fig. 3

Full size image

Composite currents from the HYCOM GLBu0.08 simulation also clearly show the EB and NB patterns at different depths from 0 to over 200 m below the surface (here, only currents at 100 m depth are shown in Fig. 6). An anticyclonic circulation with strong eastward jet south of 12°N is the most prominent feature of the summer circulation in EB pattern, while a prolonged northward western boundary current from 8 to 18°N exists in NB pattern. The magnitude of the flow in NB is smaller than that in EB at the bifurcation region southwest of Vietnam.

Fig. 6
figure 6

Composite of the summer current (vectors: m/s) and current speed (color contours: m/s) at 100 m depth (a for eastward branch pattern and b for northward branch pattern) and the first EOF mode of summer surface velocity speed (c) from HYCOM simulation GLBu0.08

Full size image

4 Factors contributing to the interannual variation of the summer circulation

Early studies (e.g. Wyrtki 1961; Shaw and Chao 1994) pointed out that monsoon wind is the dominant forcing for the seasonal variation of the upper layer circulation in the SCS, and thus a basin-scale anticyclonic circulation is the main feature in summer. Subsequent studies with more observations and higher resolution numerical models (Cai et al. 2002; Chern and Wang 2003; Chu et al. 1999; Fang et al. 1998; Gan et al. 2006; Hu et al. 2000; Qu et al. 2004; Wang et al. 2003, 2008; Xue et al. 2004; Yang et al. 2002) revealed the importance of the Kuroshio intrusion and LST on the complexity of the circulation in the NSCS, and some studies showed cyclonic circulation in the NSCS in summer. Recent studies (Gan et al. 2016; Zhu et al. 2017) suggested that the cyclonic upper layer circulation in the NSCS in summer is mainly induced by the advection of planetary vorticity through LST. And the influence from Kuroshio intrusion into the SCS is not limited in the NSCS, it could affect the circulation and water mass properties in the SSCS through western boundary current advection (Zu et al. 2019).

In Sect. 3, observations and numerical simulations all show that on interannual time scale, the summer SCS circulation could be sometimes a basin scale anticyclonic circulation, and sometimes a strong anticyclonic circulation in the SSCS with either a weak cyclonic or anticyclonic circulation in the NSCS. In this section, we would further explore under what conditions will these patterns appear and the relative contributions of the SCS summer monsoon and LST to the variability of upper layer circulation in summer.

4.1 South China Sea summer monsoon

The spatial correlation between the kinetic energy of the summer surface circulation (derived from HYCOM GLBu0.08 current at10 m depth) and the South China Sea summer monsoon index (SCSSMI) (Li and Zeng 2002) shows that a larger value exceeding the 90% confidence level is mainly along the western boundary current and the eastward jet (Fig. 7a). The SCSSMI is positively correlated to the southern segment of the western boundary current and the EB, while negatively correlated to the NB. This is similar to the first EOF mode of the surface velocity speed and kinetic energy that the NB and EB are anti-correlated (Figs. 2b, 6c and supporting information Fig. S1). It also indicates that when the South China Sea summer monsoon (SCSSM) is strengthened, the magnitude is increased for EB but decreased for NB, when the SCSSM is weakened, the contrary situation would appear. The temporal coefficient of first EOF mode (PC1) is negatively related with SCSSMI and the correlation coefficient is about − 0.79 (Fig. 7e). When the influence of the LST is excluded, their partial correlation (Fig. 7c) shows similar spatial pattern as Fig. 7a, although the areas of the significant correlation is reduced. Greater difference between the results of the partial correlation and correlation mainly appears in the NB region, suggesting that SCSSM has larger impact on the eastward jet and related anticyclonic circulation in the SSCS, whereas LST have larger impact on the northward branch of the WBC in the NSCS. Composites of the wind vectors and their vorticity in Fig. 8 show stronger southwesterly SCS summer monsoon in EB pattern, but a weaker one in NB pattern. In EB pattern years, a strong northeastward wind jet (6–8 m/s) is obvious between 9 and 15°N with positive vorticity in the north and negative vorticity in the south. Whereas, in NB pattern years, the strength of the wind jet is greatly weakened (3–4 m/s) and the coverage of positive vorticity in the NSCS is reduced to a small region along the Vietnam coast. The difference of wind vectors between EB pattern and NB pattern (NB-EB) is a strong easterly anomaly from western tropic Pacific to the SCS and the Bay of Bengal (Fig. 8c), which suggest the strong influence from ENSO. The anomaly easterly wind jet in the SCS is around 9–10°N, and it also induces a negative curl in the NSCS and a positive curl in the SSCS, thus greatly weaken the EB and the anticyclonic circulation in the SSCS in the NB pattern years.

Fig. 7
figure 7

Correlations (a and b) and partial correlations (c and d) of the South China Sea upper layer circulation between South China Sea summer monsoon index (SCSSMI) and the Luzon strait transport (LST). a Spatial distribution of correlation coefficients of the surface kinetic energy anomaly (current at 10 m depth derived from HYCOM GLBu0.08) between SCSSMI that exceeding the 90% confidence level. b Similar to a but for the anomaly of the surface kinetic energy and LST (derived from HYCOM GLBu0.08). c and d are similar to a and b but excluding the influences from LST and SCSSM, respectively. e Time series of the normalized temporal coefficient of the first EOF mode of the surface kinetic energy anomaly from HYCOM GLBu0.08 (PC1: black line), SCSSMI (red), and normalized LST (blue). The correlation coefficient between PC1 and SCSSMI is − 0.79 and is 0.61 between PC1 and LST

Full size image
Fig. 8
figure 8

Composite of the summer wind vectors (m/s) and related vorticity (s−1) from CCMP data for a eastward branch, b northward branch. c The differences of wind vectors and vorticity between eastward and northward branch patterns (northward–eastward) that past the t-test (shaded region). Eastward branch is the composite of currents in 1994, 1997, 2009, 2012, 2014, and northward branch is the composite ofcurrents in 1995, 1998, 2007, 2010

Full size image

4.2 Luzon strait transport (SCSTF)

The spatial correlation between the SCS surface kinetic energy and the LST (both of them are derived from HYCOM GLBu0.08) in Fig. 7b also has larger and more significant correlation coefficient along the western boundary current and eastward jet. The LST (negative: net transport into the SCS, positive: net transport out of the SCS) is positively related to the NB, while negatively related to the EB and the southern segment of the western boundary current. It suggests that when the LST into the SCS increases, the eastward jet and the Mindoro Strait transport out of the SCS is enhanced (i.e. an increased SCSTF), whereas when the LST into the SCS decreases, the NB is enhanced and the EB is weakened. The LST is positively correlated with the PC1 (Fig. 7c), but the value of correlation coefficient is 0.61, smaller than that with SCSSMI. This implies that both SCSSM and the LST could affect the competitive interannual variation the EB and NB, and SCSSM might play a more important role than the LST. When the influence of the SCSSM is excluded, their partial correlation (Fig. 7d) shows similar spatial pattern as Fig. 7b, with the areas of the significant correlation being reduced. Major difference between the results of the partial correlation and correlation appears in the EB region and anticyclonic circulation region in the SSCS. This result again suggests that local wind stress curl has larger impact on the eastward jet and related anticyclonic circulation in the SSCS, whereas remote wind forcing and LST have larger impact on the northward branch of the WBC in the NSCS.

The LST and Mindoro Strait transport (MST) from April to October in EB and NB pattern years together with their composites in Fig. 9 show that the value of net LST into the SCS and the net MST out of the SCS (i.e. a branch of the SCSTF) are generally larger in EB years than in NB years. In July and August, the average difference could reach 3–4 Sv in both LS and MS. The greatest difference of about 6–7 Sv is between the extreme El Niño in 1997 and the extreme La Niña in 2010.

Fig. 9
figure 9

Time series of the a Luzon strait, and the b Mindoro strait transports (m3/s) from HYCOM simulation result GLBu0.08. The negative value of LST is into the SCS, while the negative value of MST is out of the SCS. Color lines are for different years, the blank solid line indicates the composite of the northward branch years, and the black dash line indicates the composite of the eastward branch years

Full size image

Sensitivity experiments by closing the Luzon Strait in ATOP simulation clearly show a weakened EB and associated double gyres in 2009, while an enhanced NB in 2010 (Fig. 10). It confirms that the net LST from Pacific into the SCS (advection of planetary vorticity) favors a tendency of cyclonic circulation in the NSCS, which would strengthen the EB but weaken the NB in summer.

Fig. 10
figure 10

Surface current in July from ATOP control run (left column) and sensitivity case with Luzon Strait closed (right column) in 2009 and 2010. Color shading is the speed of the current (m/s)

Full size image

4.3 Vorticity contribution from local wind stress and lateral flux

Besides the aforementioned comparisons, we have also made a quantitative estimate of the vorticity input from local wind stress curl and the advection of planetary vorticity fluxes through the straits, as they are the major external vorticity sources for the cyclonic and anticyclonic seasonal circulations in the SCS (Xu and Oey 2014; Gan et al. 2016; Zhu et al. 2017; Quan and Xue 2018). In addition, the standard deviation of the vorticity of the surface geostrophic current in summer is larger in both magnitude and area next to the Vietnam coast along the western boundary of the SCS (supporting information Fig. S5). This indicates that the upper layer integrated vorticity variation in the SCS domain could well reflects the changes of vorticity and circulation in this region.

Domain and upper layer integrated vorticity budget Eq. (2) is written similar as that in Gan et al. (2016), Quan and Xue (2018).

$$\varOmega_{ACC} + \varOmega_{ADV} + \varOmega_{COR} + \varOmega_{PGF} + \varOmega_{VVIS} + \varOmega_{HDIF} = 0$$
(2)

The terms in Eq. (2), from left to right are the curls of the acceleration ΩACC (tendency of relative vorticity), horizontal and vertical (vortex tilting) advection ΩADV, Coriolis force (vortex stretching) ΩCOR, pressure gradient force ΩPGF, vertical viscosity ΩVVIS and horizontal diffusion ΩHDIF, respectively. After integration, in the upper layer, ΩVVIS could be expressed as ΩVVIS  =  ΩUSTR + ΩLSTR (the sum of wind stress curl in the upper interface and frictional stress curl in the lower interface). The Coriolis force term could be written as \(\Omega _{COR} = {\oint }U_{n} fdl\) (integral of planetary vorticity flux along the lateral boundary), where \(U_{n}\) is the vertically integrated velocity component perpendicular to the lateral boundary, and \(f\) is the Coriolis parameter.

Xu and Oey (2014), Gan et al. (2016), Zhu et al. (2017) and Quan and Xue (2018) all show that the major external vorticity input terms in the SCS upper layer circulation are wind stress curl from atmosphere (ΩUSTR) and the lateral flux of the planetary vorticity from open-ocean through major straits (ΩCOR). In summer, both terms together balance the ΩPGF, and other terms play minor roles on the balance (Quan and Xue 2018). Therefore, we only compare the contributions from these two terms. The ΩUSTR is calculated as the areal integration of the composite wind stress curl (Eq. (3)) in the region between 107–121°E and 3–25°N, where water depth is greater than 5 m.

$$\Omega _{USTR} = {\iint }curl_{z} \left( {\frac{{\tau_{a} }}{{\rho_{0} }}} \right)dxdy$$
(3)

In which \(\tau_{a}\) is wind stress, \(\rho_{0}\) is the reference density of seawater, and the operator is defined as \(curl_{z} \left( {\vec{A}} \right) = \frac{{\partial A_{y} }}{\partial x} - \frac{{\partial A_{x} }}{\partial y}\). The ΩCOR is calculated by the multiples of the composite transport and the Coriolis parameter \(f\) at the major straits.

$$\Omega _{COR} = \mathop \sum \limits_{i = 1}^{N} Q_{i} f_{i}$$
(4)

In which \(Q_{i}\) and \(f_{i}\) are the volume transport and Coriolis parameter at the ith strait. Karimata Strait is not considered here as the \(f\) is too small and consequently its contribution to the flux of the planetary vorticity is negligible. The roughly estimates of the relative contributions of these two terms are summarized in Table 1. In NB years, when the southwesterly summer monsoon is weakened, the magnitude of both positive and negative wind stress curl inputs into the SCS are smaller than those in EB years. But the net wind stress curl input of the whole SCS basin is about − 20 m3 s−2 in NB years, which is twice of that in EB years. The difference between NB and EB years (NB minus EB) is around − 10 m3 s−2. The negative wind stress curl input favors an anticyclonic circulation in the SCS, thus the wind stress curl input contributes to a stronger anticyclonic circulation in the SSCS with eastward jet in EB years while a stronger NB and a basin scale anticyclonic circulation in NB years. Regarding to the flux of the planetary vorticity through the major straits (positive value means gaining positive vorticity, while negative value means loosing positive vorticity), Luzon Strait (LS) transport favors cyclonic circulations, whereas Taiwan Strait (TS) transport favors anticyclonic circulations. Mindoro Strait (MS) transport favors anticyclonic circulations in EB years, when the SSCS anticyclonic circulation and eastward jet are strong. The net flux of planetary vorticity is about − 73 m3 s−2 in NB year, and − 56 m3 s−2 in EB year. The difference between them is − 17 m3 s−2, and it has a similar magnitude to the difference between the wind stress curl in NB and EB years. This suggests that the change of local wind stress and major strait transport contribute equally. Although the planetary vorticity flux at TS is large, on interannual variations, the change at LS is the major contributor to the negative difference of planetary vorticity flux between NB and EB years.

Table 1 Estimations of the domain integrated wind stress curl (WSC) and planetary vorticity flux through Luzon Strait (LS), Mindoro Strait (MS) and Taiwan Strait (TS) in the SCS in summer (July and August) of northward branch (NB) and eastward branch (EB) years. (Unit: m3 s−2)
Full size table

4.4 Other factors that add to the complexity of the interannual variations

The interannual variation of the summer circulation in the SCS is influenced by the SCSSM and LST, which are modulated by the ENSO events. However, ENSO complexity might lead to complicated responses of the SCSSM and LST to ENSO events,and consequently a more complicated interannual variation of the summer circulation in the SCS. According to Yu and Fang (2018), there are ENSO cycles (one event followed by an opposite phase event), episodic events (one event preceded and followed by a neutral year), and multiyear El Niño or La Niña events. The charged-discharged (CD) mechanism tends to make regular and symmetric ENSO cycles, while seasonal footprinting (SF) mechanism tends to produce multiyear La Niña events and contributing to El Niño-La Niña asymmetries. From Fig. 2c, it is clear that during ENSO cycles (i.e. 1994–1995, 1997–1998, 2009–2010), there would always be obvious EB pattern in El Niño followed by strong NB pattern in La Niña. Nevertheless, the EB or NB pattern would not necessarily appear during episodic events (such as in 2002, 2007, 2015). Additionally, NB pattern will not appear in the second La Niña year during multiyear La Niña events (i.e. 1999, 2011). This suggests that the alternate appearance of the EB and NB pattern during ENSO cycles is a relatively deterministic response that is mainly related to the tropical ocean and atmosphere signals of CD mechanism, whereas the subtropical atmosphere–ocean coupled forcing of SF mechanism add uncertainty and complexity to the appearance of EB and NB patterns. However, to further understand the complexity of the appearance of the two kinds of circulation patterns, we need to explore the responses of the SCSSM and LST to the tropical and subtropical atmosphere–ocean coupled forcing, which is beyond the present work but deserves future efforts.

In addition to the ENSO complexity, ocean internal variability (OIV, through eddy-mean flow interactions) will also contribute to the interannual variation of the summer circulation. Numerical experiments forced with monthly climatologic atmospheric fields show that the OIV accounts for about 20% of the interannual variability of the strength of the eastward jet and dipole structure at Vietnam coast (Li et al. 2014). When studying the interannual variation of the South Vietnam Upwelling (SVU), Da (2018) also shows that OIV could account for about 40% or even more of the observed interannual variation of the magnitude of the SVU, through numerical experiments with climatological daily atmospheric field and higher spatial resolution model. Both of them confirm the importance of OIV to the interannual complexity of the magnitude of the eastward jet and dipole structure and related upwelling. However, none of them shows visible influences of OIV on the displacements of the upwelling center or jet separating from the Vietnam coast. This implies again that interannual change of the local atmospheric fields and oceanic forcing play more significant roles on the variability of the major flow patterns in the SCS in summer.

5 Summary

By analyzing the current derived from altimetry data, ADCP moorings and numerical simulations, we revisited the interannual variation of the SCS upper layer circulation in summer, and provided another perspective (EB and NB patterns) to understand the change of the basin scale circulation in the SCS in summer. Besides the interannual variation of the eastward jet (EB) pointed out by many previous studies, we also revealed its anti-correlation with the NB. The results could be summarized by a sketch in Fig. 11. During El Niño developing phase, three major processes are in favor of a stronger EB: firstly, the stronger southwesterly summer monsoon with the zero wind stress curl across the SCS basin located more southward, is favorable for a separation of boundary current at a lower latitude, secondly, the stronger LST into the SCS with larger positive planetary vorticity flux is favorable for a cyclonic circulation in the NSCS which would facilitate boundary current separation and enhance the EB, thirdly, the recirculation associated with anticyclonic eddy on the south of the EB would strengthen the WBC and lead to a more robust EB. While during the El Niño decaying phase and La Niña developing phase, the situation is in favor of a NB pattern, a weaker summer monsoon with a northward and counter-clockwise shift of the zero wind stress curl line in the SCS basin is unfavorable for the boundary current separation, besides the decreased LST with greatly reduced positive planetary vorticity flux would not support for a cyclonic circulation in the NSCS. Basically, the shift between the EB and NB patterns is related to the interannual variation of the SCSSM and the LST. The contribution of the major strait transport to the interannual variation of the SCS summer circulation is quantitatively evaluated for the first time, which show that the vorticity input from planetary vorticity flux through these major straits modulated by the remote wind forcing is as important as that from local wind stress curl. Moreover, results show that ENSO complexity also influences the interannual variation of the summer circulation. Further analysis of the responses of the SCSSM and LST to the ENSO complexity is needed to better understand and predict the interannual variation of the SCS summer circulation.

Fig. 11
figure 11

Sketch of the two distinct patterns of the upper layer circulation in the SCS in summer superimposed on the topography map in this region. Red lines indicate the situation in EB pattern years, and blue lines indicate the situation in NB pattern year. The semitransparent solid lines show the main pathways of the circulation, and the semitransparent dash line shows the unstable flow or the tendency of the cyclonic circulation. The opaque lines indicate the zeros wind stress curl lines that across the SCS basin

Full size image