To investigate the currents (TWC, Kuroshio Current, ZMCC, etc.) and their synoptic fluctuations on the ECS shelf, a three-dimensional (3-D) unstructured-grid (Fig. 2, left panel) FVCOM is developed for the entire Bohai, Yellow and East China Seas (part of the Japan/East Sea, and part of the Pacific Ocean). A regional refinement of the resolution (approximately 3 km) is specified around the ECS shelf break at the 200 m isobaths, where a strong excursion of the Kuroshio Current also occurs. The General Bathymetric Chart of the Oceans (GEBCO) provides high-resolution (approximately 1 km) bathymetric data (Smith and Sandwell, 1997). In all, 20 vertical layers with 76 954 triangle cells were specified in the water column in a sigma-stretched coordinate system.

The driving forces of the numerical simulation include tides, river discharge, surface heat fluxes, wind and open-boundary conditions. Harmonic constants of 11 major tidal constituents (M2, S2, N2, K2, K1, O1, P1, Q1, M4, MS4 and MN4) were used; these are based on the Oregon State University global inverse tidal model TPXO.7.0 (Egbert et al., 1994; Egbert and Erofeeva, 2002). The daily mean river discharge of the Changjiang and Huanghe was taken from publicly available observation data at the Datong hydrometric station (http://yu-zhu.vicp.net/). Other rivers were not included because of their small discharges; e.g., the Qiantang River, with the largest runoff from the Zhejiang coast, has a climatological mean discharge in winter of about 230 m3 s-1, which is nearly negligible compared to the Changjiang winter discharge of about 11 500 m3 s-1. The daily mean heat fluxes were from the objectively analyzed air–sea fluxes (Yu and Weller, 2007), and the 3-hourly wind stress and 10 m wind speed data were from the ERA-Interim re-analysis (Dee et al., 2011). The open-boundary conditions, including daily temperature, salinity and fluxes at the Taiwan Strait, the western Pacific Ocean and the Japan/East Sea, were obtained from the Hybrid Coordinate Ocean Model (Bleck, 2002) and interpolated onto the FVCOM model grid points. The temporal resolution of all the driving force fields is better than or equal to 1 day, which is essential to resolve synoptic fluctuations.

The hindcast outputs of sea surface height, temperature, salinity and velocities for the 5 years of simulation from 2009 to 2013 are used, following three spin-up years (2006–2008) initiated with the temperature and salinity taken from the Hybrid Coordinate Ocean Model and velocity set to zero. The initial conditions are ramped-up over a period of 30 days and at the lateral boundaries a sponge layer was used with the same method as Chen et al. (2008). The model time step was 15 seconds for the 2-D barotropic mode and 90 seconds for the 3-D baroclinic mode. All of the output fields were processed with a tidal filter (Godin, 1972) to remove tidal oscillations (considering that the major timescale of synoptic fluctuations in this study area is 3–15 days).

Since the currents in 2009 could partly be validated by means of available observational data (see Sect. 2.2), the currents from 1 January to 28 February 2009 were selected for analysis of the wintertime TWC.

The mean currents, e.g., the Kuroshio Current, the TWC, and the ZMCC, were calculated by averaging the outputs of January and February 2009. We validated the mean currents in terms of circulation structure, boundary fluxes, and coastal currents.

The FVCOM has reproduced almost all of the known circulation structure in the ECS in winter. The surface mean currents (Fig. 2) show three major currents: the Kuroshio Current, the TWC and the ZMCC. The Kuroshio Current, with a speed of about 1 m s-1, enters the ECS just northeast of Taiwan and flows along the shelf break up to the northern area and ultimately leaves the ECS through the Tokara Strait. Both the route and strength of the Kuroshio are comparable with those reported in the literature (Guan, 1978; Qiu and Imasato, 1990). The TWC has two northeastward branches, one inshore (between the 30 and 100 m isobaths) and another offshore (between the 100 and 200 m isobaths), which is consistent with Su and Pan (1987). The southwestward-directed ZMCC in the nearshore area from the Changjiang Estuary to the Taiwan Strait agrees well with that reported in previous studies (Guan and Mao, 1982; Zeng et al., 2012).

The simulated volume transports across the Taiwan Strait, the East Taiwan Channel, the Tsushima Strait, the Tokara Strait and the shelf break of the 200 m isobath were validated using results from the literature (Table 1). The simulated transports were accurate enough to reproduce volume transport (1.22 Sv) through the Taiwan Strait, which is closer to the observation value (1.20 Sv) from Isobe (2008) than former model results. The volume transports across the Taiwan Strait and the Tokara Strait, as well as the cross-shore exchange, affected the path and magnitude of the TWC. The annual-mean transport across the 200 m isobath toward the shelf is 1.66 Sv, which is balanced by the inflow from the Taiwan Strait (1.22 Sv) and the outflow through the Tsushima Strait (2.85 Sv).

Figure 3 shows a comparison between simulation and observation results for the alongshore currents and the cross-shore currents on the ECS shelf. The observational data were obtained from four mooring surveys (Fig. 2, red stations) off the Zhe-Min coast (Zeng et al., 2012). The observed and simulated currents were both averaged for the observational period, which was from 1 January to 28 February 2009. Using the same method as in Huang et al. (2016), we defined the positive alongshore current direction as from the southwest (218∘) to the northeast (38∘), which is the mean tangential direction of the isobaths on the southwestern shelf of the ECS. The positive cross-shore direction is from the northwest (308∘) to the southeast (128∘), normal to the isobaths. The alongshore components (Fig. 3a and b) show that the ZMCC flows southwestward parallel to the coast in winter, with a maximum speed of 0.15 m s-1 along the 30 m isobath. The TWC flows northeastward with a speed of 0.05 m s-1, and the core is located in the lower layer at about 50 m at station 4. The cross-shore component (Fig. 3c and d) is much weaker than the alongshore components, and it shows a complex spatial pattern. It flows offshore in the upper layer and onshore in the lower layer at station 1. Moreover, it mainly flows onshore at station 2, and it flows offshore in the entire water column at stations 3 and 4. Altogether, the simulated pattern and magnitude both of the alongshore and cross-shore components are in good agreement with the observations. However, there are some differences between the observed and simulated results; for example, the simulated ZMCC occupies a broader space than that in the observations. This may have been caused by the relatively low number of observational stations.

Synoptic fluctuations of the TWC inshore branch during January and February 2009 were also validated against the mooring results (Fig. 4). Since the TWC shows a strong signature at station 4, the time series of the alongshore currents and cross-shore currents in the whole water column of station 4 were used for the validation. To eliminate the influence of local effects, the simulated currents were averaged in a 10 × 10 km2 area around station 4. Both the observed and simulated results show that the TWC fluctuates with a period of 3–15 days. The simulated TWC (Fig. 4a, warm color) appeared stronger (> 0.1 m s-1) on 7, 12, 18, 21, 26 and 29 January, and 10, 14, 19, 22 and 25 February, which agrees well with data from the observations (Fig. 4b). The time series of the simulated cross-shore component (Fig. 4c) are virtually in phase with the observations (Fig. 4d). The magnitude of the cross-shore fluctuations is comparable to the alongshore fluctuations. This is different to the anisotropic characteristic of the mean currents (Fig. 3), for which the alongshore component is nearly 1 order of magnitude larger than the cross-shore component.

The empirical orthogonal function (EOF) method (Emery and Thomson, 2001), as a statistical method, has been used to understand synoptic fluctuations of the wintertime TWC. The simulated currents from 1 January to 28 February 2009 were selected and their anomalies were calculated. Then, using the Matlab EOF function, the current vectors were separated into several orthogonal modes to show the spatial and temporal variations. Because the first two leading modes explain 91 % of the total variance, only these two modes were used for the analysis.

The spatial distributions of the two leading EOF modes were used to analyze the regional difference of the synoptic fluctuations. To investigate the driving force of the two EOF modes, the temporal variation was compared to the potential influence factors, such as wind, upstream currents and net surface heat flux.

The driving mechanisms of the synoptic fluctuations were further analyzed using the momentum equation. First, the momentum balance as implemented in FVCOM (Chen et al., 2003) is shown in Eq. (1). The three terms on the left-hand side represent local acceleration, Coriolis acceleration and advection, and the three terms on the right-hand side represent pressure gradient, friction and diffusion. ∂V∂t-2Ω×V+(V⋅∇V)=-1ρ0∇P+∂∂zKm∂V∂z+F, where V is velocity, Ω is the Earth's rotation angular velocity, ρ0 is the average density, P is pressure, Km is the vertical eddy viscosity coefficient and F is horizontal diffusion.

Second, according to the hydrostatic approximation used in FVCOM (as shown in Eq. 2), the pressure gradient is given as the product of density times the gravitational acceleration. This results in Eq. (3), which indicates that pressure gradient can be decomposed into the effects of the barotropic and baroclinic components, as shown in Eq. (4). ∂P∂z=ρg,Pz=∫zηρgdz=∫zηρ0+ρ′gdz=ρ0g(z+η)+∫zηρ′gdz,∇P=ρ0g∇η+∇∫zηρ′gdz, where ρ is density, ρ′ is density anomaly, g is the gravitational acceleration and η is sea surface height.

Finally, the momentum equation is vertically integrated to estimate momentum balance for the water column. Since the horizontal diffusion is a comparatively small term, it is neglected for simplicity. ∫-H0∂V∂t︸Acceleration+∫-H0-2Ω×V︸Coriolis+∫-H0(V⋅∇V)︸Advection=-gH∇η︸Barotropic-∫-H0∇∫zηρ′gdz︸Baroclinic︸Total Pressure+ρaCD|U|U︸τa-kb|Vb|Vb︸τb, where τa is wind stress and τb is bottom stress, ρa is the density of air, U is the wind speed at 10 m above sea surface, CD is a drag coefficient at the sea surface (which varies with wind speed U), kb is a bottom friction coefficient (kb = 0.005) and Vb is the simulated velocity at the bottom.

Since the observational results (Su and Pan, 1987; Zeng et al., 2012) show that both branches of the wintertime TWC are flowing in the subsurface, we use the vertical maximum velocity (VMV) and its corresponding depth as two indices to quantify the strength of the subsurface currents (Fig. 5).

As stated above, the distribution of the VMV shows two branches of the TWC (Fig. 5a). The inshore branch (Fig. 5a, blue arrow of IB), which was located between the 30 and 100 m isobaths, followed a straight route from the northwest of Taiwan to the northern ECS shelf. The offshore branch (Fig. 5a, blue arrow of OB) existed near the 100 m isobath and had two meanders. The two meanders turn to the cross-shore direction along latitudes 26.5 and 28∘ N. These two branches are further illustrated in the distributions of current speed along the six TWC cross sections (S1–S6), which were located at critical points in the two meanders (Fig. 6). From the VMV structure, it can be inferred that the intrusions of the TSC and the Kuroshio Current both affected the origin of the offshore branch (Fig. 6, S1–S3).

We further examined the subsurface current core using the depth of the VMV (Fig. 5b). We found that the VMV of the TWC was located 40–60 m below the surface at the inshore branch and 20–40 m below the surface at the offshore branch. Figure 6 shows the VMV positions in the subsurface layer; furthermore, it illustrates that the depth of the subsurface VMV in the inshore branch was deeper than that in the offshore branch. The difference can be explained by the combined effects of baroclinicity and wind friction. Assuming a relatively spatially homogeneous heat loss, different cooling occurs, due to the smaller heat capacity of the shallow coastal water compared to the deeper offshore waters, hence generating a northwestward horizontal density gradient leading to a northeastward thermal current (vertical current shear) according to the thermal wind relationship, resulting in an upward-increasing northeastward flow. The northeasterly wind in winter weakens the northeastward TWC, particularly in the upper layer, which leads to the formation of the subsurface VMV. Therefore, the fact that the depth of the subsurface current core in the inshore branch is greater than that in the offshore branch indicates weaker baroclinicity or stronger wind friction on the inshore branch than on the offshore branch.

The magnitude of the wintertime TWC was obtained by flux analysis. Two dividing lines (Fig. 5a, red lines) were defined as the boundaries for the ZMCC: the TWC inshore branch, and the TWC offshore branch, which had the weakest flows. The flux of each branch (Fig. 5c) was calculated using the horizontal integration between the boundaries and the vertical integration in the water column. The inshore branch intensifies along its way and becomes significant north of 26.5∘ N, showing particularly strong flow velocities between 27.5 and 28.0∘ N. In this area, the subsurface current was much stronger from S4 to S5 than in the other areas (Fig. 6). The flux in the entire offshore branch was large, particularly north of Taiwan.

The observations (Fig. 4) have demonstrated that the synoptic fluctuation in the TWC inshore branch (near 121.5∘ E, 27.0∘ N) is significant. We further investigated the regional difference of fluctuations in the two TWC branches in winter 2009 using the following three steps: (i) two regions with significant fluctuations are identified by the current standard deviations of the VMV (Fig. 7) and the corresponding temporal variation of vertical structures at their extremes (Fig. 8); (ii) each of the two significant fluctuations is decomposed into EOF components (Fig. 9); and (iii) the influence factors, such as wind, upstream currents, and net surface heat flux, are investigated by examining their correlations with the first two leading EOF components (Figs. 10 and 11).

The current standard deviations (Fig. 7) show that prominent fluctuations occurred in two regions: north of Taiwan and the inshore area. The standard deviations of VMV at the two regions were larger than 0.1 m s-1 (comparable to the mean currents). In the area north of Taiwan, the fluctuation was located in the origin area of the TWC offshore branch. The fluctuation in this region was in phase with the fluctuation in the Taiwan Strait, indicating that the TSC played an important role in generating the fluctuation north of Taiwan (to a greater extent than did the Kuroshio intrusion). The TWC fluctuation had a strong cross-shore component, which means the fluctuation transported the water north of Taiwan to both the inshore and offshore branches. In the inshore area, the fluctuations were influencing a wide region between the 30 and 100 m isobaths, with a magnitude that was sometimes larger than the mean flow (Fig. 5a). These strong fluctuations led to an episodic occurrence of the TWC inshore branch, as observed at the site off the Zhe-Min coast (Fig. 4, red color). When the TWC inshore branch was weakened due to these fluctuations, the ZMCC might even dominate a wide region out to the 100 m isobath, especially at the surface (Fig. 4, blue color).

The vertical structures of the fluctuations north of Taiwan and in the inshore area at two representative points and their relation with the upper mixed layer depth are further analyzed (Fig. 8). The major component (the alongshore current) of the TWC in each of the two regions (P1 and P2, Fig. 7) is used to show the vertical structure of the fluctuation. The depths of the upper mixed layer were determined by a Richardson number criterion (Mellor and Durbin, 1975; Grachev et al., 2013; Richardson et al., 2013), i.e., where the critical Richardson number equals 0.25 in this paper (as in Xuan et al., 2012b). The mean depth of the upper mixed layer north of Taiwan (20 m) was much shallower than the mean depth of the inshore area (42 m). However, the TWC (Fig. 8, warm color) fluctuated with significant variations of the upper mixed layer depth (Fig. 8, gray lines) in both areas. When the upper mixed layer deepened, the northeastward TWC (Fig. 8, warm color) was weakened or even replaced by the southwestward ZMCC, and vice versa. Wind and surface cooling, which both drive the mixed layer depth, can affect the TWC fluctuation.

The TWC fluctuations were further decomposed into EOF modes. The first two leading EOF modes account for 54 and 37 % of the total variances (Fig. 9), associated with the two prominent fluctuations north of Taiwan and in the inshore area (Fig. 7). Both EOF modes had a maximum fluctuation larger than 0.2 m s-1 (comparable to the mean currents). The spatial pattern of the first EOF mode (EOF1; Fig. 9a) shows that the fluctuation continued from the Taiwan Strait to the area north of Taiwan, indicating that the fluctuation north of Taiwan was related to the TSC and not to the Kuroshio Current. The alongshore component also showed a strong fluctuation in the Taiwan Strait, which means that the TSC episodically intruded the shelf. The cross-shore component revealed a fluctuation north of Taiwan that was larger than 0.1 m s-1. This cross-shore fluctuation impacted on the trajectory of the Taiwan Strait (TWS) water, synoptically flowing into the TWC inshore branch, offshore branch or Kuroshio Current.

The spatial pattern of the second EOF mode (EOF2; Fig. 9b) shows a synoptic fluctuation in the inshore area. The area with alongshore fluctuation (Fig. 9d) larger than 0.1 m s-1 was located between the 30 and 100 m isobaths, which demonstrates that the TWC could episodically affect this area. In addition, there were cross-shore fluctuations in the inshore area (Fig. 9f), mostly along the latitudes 26.5 and 28∘ N. The latitudes of larger cross-shore fluctuations agreed well with the latitudes where the TWC offshore branch of the mean currents (Fig. 5a) turned to the cross-shore direction. This indicated that the cross-shore transports were most significant at the latitudes 26.5 and 28∘ N, according to both the mean currents and the synoptic fluctuations.

Figure 10 shows the temporal variation of EOF1 and its relation to the north–south component of wind speed, net surface heat flux, the TSC and the Kuroshio Current. We found a close correlation between EOF1 and TSC (R = 0.86), demonstrating that the TSC played the most important role in generating the TWC fluctuation north of Taiwan. The EOF1 and TSC were positively correlated, meaning that a larger TSC intrusion north of Taiwan leads to a cross-shore current from the coastal area to the offshore area and that a weak TSC intrusion causes a cross-shore current from offshore to inshore north of Taiwan.

Figure 11 shows the temporal variation of EOF2 and its relation with the north–south component of wind speed, net surface heat flux, the TSC and the Kuroshio Current. It can be seen that EOF2 and wind are well correlated (R = 0.89), indicating the important role of wind in generating the TWC fluctuation in the inshore area. The northeasterly monsoon would greatly enhance the southwestward ZMCC, which would then replace the northeastward TWC in the inshore area.

The wintertime (January and February 2009) mean of the water column momentum balance (Fig. 12) is used to show the overall distribution of the fundamental forces over the ECS shelf. The Coriolis force (Fig. 12a) is mainly balanced by the total pressure (Fig. 12b) in both branches, indicating the dominant role of geostrophic balance in the wintertime TWC. However, the wind-induced surface friction plays an important role in the TWC, especially in the inshore area and the Taiwan Strait (Fig. 12c). The bottom friction has an impact north of Taiwan and in the shallow Taiwan Strait, in particular when significant Kuroshio intrusion enhances the bottom flow (Fig. 12d). The effects of advection and acceleration are predominantly local indicated by mostly incoherent small-scale distributions (Fig. 12e and f); therefore, they can be ignored when studying the large-scale current of the wintertime TWC.

The variation of the driving forces at two representative points P1 and P2 was used to analyze the dynamics of synoptic fluctuations north of Taiwan and in the inshore area. Regarding the results from the EOF analysis, the three force terms, namely, Coriolis, total pressure and wind (Fig. 13), were selected to investigate the effect of the TSC on the fluctuation north of Taiwan (Fig. 9a) and the effect of wind on the fluctuation in the inshore area (Fig. 9b).

In the area north of Taiwan, the cross-shore fluctuations were induced by the TSC intrusion. The variation of alongshore Coriolis force (Fig. 13a, black line) was much greater than the cross-shore Coriolis force (Fig. 13b, black line), which means that the fluctuation north of Taiwan was mainly in the cross-shore direction. The Coriolis force (Fig. 13a, black line) was mainly balanced by the total pressure (Fig. 13a, blue line), which means the currents fluctuations north of Taiwan are dominated by geostrophic balance. As mentioned in Sect. 3.2, the TWC fluctuation north of Taiwan was associated with the TSC rather than with the Kuroshio Current. Therefore, in the shallow coastal area the TSC mainly caused variations in the depth-independent barotropic pressure gradients, which further generated the cross-shore fluctuation. The mechanism can be interpreted as follows. When a larger TSC intrusion occurred, the isobaric slope tilted downward from south to north, generating a cross-shore current from the coastal area to the offshore area. On the contrary, when the TSC intrusion was weak, the Kuroshio intrusion from offshore to inshore dominated north of Taiwan.

Wind friction (Fig. 13c and d) was a fundamental factor in generating the fluctuations in the inshore area. Although the geostrophic balance dominated in the inshore branch for most of the time, the episodically strong winter monsoon had an important role in generating the TWC fluctuations. The northwestward direction Coriolis force (Fig. 13c, black line) shows that the southwestward ZMCC occurred on 12, 22 January and 14 February 2009, and was associated with a northeasterly wind (Fig. 13c, red line). It indicates that strong northeasterly monsoon in winter can reduce or even stop the northeastward TWC in the inshore area, causing the intermittency of the TWC inshore branch.

In the area north of Taiwan, the TSC intrusion generated strong fluctuations of the TWC in the cross-shore direction (Fig. 9a). When a larger TSC intrusion occurred, the isobaric slope tilted downward from south to north, generating a cross-shore current from the coastal area to the offshore area. Compared to the reported summer route that transports Taiwan Strait water to the inshore area between the 30 and 100 m isobaths (Guan, 1978; Fang et al., 1991; Isobe, 2008; Yang et al., 2011, 2012), our results showed that most Taiwan Strait water was transported to the TWC offshore branch and to the Kuroshio area as a result of the cross-shore fluctuations induced by the synoptic TSC intrusion.

A numerical tracer simulation was used to analyze the role of the cross-shore fluctuation in the transport of the TSC water and the Kuroshio water north of Taiwan. In order to demonstrate the characteristics of the flow patterns more clearly, artificial tracers are released in the model domain and transported by the velocity field provided by the FVCOM simulation. The tracer running was part of the FVCOM simulation; therefore, all the abovementioned dynamics were involved, e.g., tide, wind and boundary forces. The release location and start date of the particles were configured as follows. Two sections, one in the Taiwan Strait (Fig. 15a, black dots) and another in the East Taiwan Channel (Fig. 15b, black dots), were selected as the source locations for the water masses of the TSC and the Kuroshio, respectively. The particles were released on 1 January 2009 and tracked until 31 March 2009 (a total of 90 days).

Figure 15a shows the traces originating from the TSC area. Unlike the traditional route, where the TSC water flows from the Taiwan Strait to the inshore area between the 30 and 100 m isobaths, most particles (Fig. 15a, gray lines) were concentrated in the offshore branch under the effect of cross-shore fluctuation. Two particles were selected to show the inshore route (Fig. 15a, red line) and offshore route (Fig. 15a, blue line), with both passing the area north of Taiwan. When the two particles arrived at the area north of Taiwan, the behavior of the tracers, according to specific velocity conditions (Fig. 15c), was very different; a northwestward transport occurred on 25 January for the inshore particles (Fig. 15c) and a northeastward transport occurred on 12 February for the offshore particles (Fig. 15c). The velocity conditions in the area north of Taiwan corresponded to the variation of the Taiwan Strait flux (Fig. 10), which shows that the Taiwan Strait flux on 12 February was much greater than on 25 January. Therefore, it can be concluded that the TSC intrusion induced an offshore transport north of Taiwan.

Figure 15b shows the traces originating from the Kuroshio area. In the same way as the TSC water, the Kuroshio water was also transported to the northern shelf via both the inshore branch and the offshore branch. The separation of the two branches north of Taiwan was caused by cross-shore fluctuations of the currents. When the two particles arrived at the area north of Taiwan, a northwestward transport occurred on Feb. 2 for the inshore particles (Fig. 15c) and a northeastward transport occurred on Feb. 12 for the offshore particles (Fig. 15c). This means that the offshore transport induced by the TSC also had an effect on the distribution of Kuroshio water north of Taiwan. Liu et al. (2016) showed that the winter TSC originated from a small branch of Kuroshio intrusion into the Luzon Strait. Our results complement this picture, since they show that most TSC particles flow into the TWC offshore branch under the influence of cross-shore fluctuation.

Our results may underestimate the impact of Kuroshio intrusion on the fluctuation of the TWC northeast of Taiwan, especially at the seasonal and interannual timescales. Wei et al. (2013) demonstrated that the annual and interannual variations of the Kuroshio volume transport are large. In addition, Zhou et al. (2015) pointed out that the annual and interannual variations of the Kuroshio intrusion northeast of Taiwan are prominent. X. Liu et al. (2014b) presented supportive evidence that the Kuroshio intrusion, from east of Taiwan to the onshore area north of Taiwan, is closely related to the Kuroshio volume transport. This relation between the Kuroshio intrusion and the Kuroshio volume transport had been interpreted by Su and Pan (1987) as the β effect because of the sudden change in topography northeast of Taiwan. Our results show that the intraseasonal variation of the Kuroshio intrusion and the Kuroshio volume transport was negligible compared with the TSC variation at the same timescale, indicating that the synoptic fluctuation of TWC north of Taiwan is mainly induced by the TSC. However, because FVCOM uses sigma coordinates in the vertical that are prone to errors in regions of steep topography, our results may underestimate the fluctuations at the shelf break, in particular to the northeast of Taiwan where Kuroshio intrusion occurs.

In the inshore area, the synoptic fluctuations of the TWC (Fig. 9b) caused by wind were generally strong in the alongshore direction and regionally important (along the latitudes 26.5 and 28∘ N) in the cross-shore direction. The alongshore fluctuations showed that the TWC inshore branch occurred episodically. This episodic occurrence of the TWC agrees with the results from a previous study based on four mooring surveys off the Zhe-Min coast (Zeng et al., 2012). The mechanism of the episodic occurrence of the TWC was mainly associated with the winter monsoon, which agrees with the analysis of observational data by Huang et al. (2016). However, the overall magnitude of the TWC fluctuation, and its role in the cross-shore flux, is still not fully understood due to the short-term nature of the observational data.

We investigated the magnitude of TWC fluctuation, and its role in the water exchange, in the inshore area. Previous studies (Su and Pan, 1987; Zeng et al., 2012) showed that the TWC flows between the 50 and 100 m isobaths, whereas the ZMCC water dominates the coastal area west of the 50 m isobath in the surface layer. As mentioned when discussing Fig. 9d, the strongest TWC could reach the coastal area as close as the 30 m isobath, being stronger than those reported in the literature. Moreover, the area with large fluctuations spanned the area between the 30 and 100 m isobaths (Fig. 9b), indicating that water between the 30 and 100 m isobaths may be either ZMCC or TWC water.

The episodic occurrence of the TWC inshore branch is directly related to the relative importance of the southwestward ZMCC (Fig. 16, blue arrows) and the northeastward TWC (Fig. 16, red arrows). In this paper, only wind-induced synoptic fluctuations are considered, not short-term extreme storm events. When the winter monsoon (the northeasterly wind) prevails, the ZMCC occupies most of the inshore area and the TWC inshore branch weakens (Fig. 16a). On the contrary, the TWC inshore branch can intrude into the near-coast area under southwesterly wind conditions (Fig. 16b). The boundary between the coastal current and the TWC may shift from the 100 m isobaths to the 30 m isobath in the cross-shore direction, covering the entire area of the TWC inshore branch.

Our results further reveal that strong wind-induced cross-shore fluctuations occur in the inshore area (Fig. 9f). This cross-shore fluctuation has a significant ecological impact because of the connected nutrient transport (Zhao and Guo, 2011). Ren et al. (2015) observed a cross-shore flux in the inshore area, which was triggered by the transition of northeasterly to southwesterly winds. Their observed features can be further interpreted with our result that wind-induced fluctuations can affect the cross-shore water transport in the inshore area.

The largest cross-shore fluctuations were located at the latitudes 26.5 and 28∘ N (Fig. 9f), which agreed well with the latitudes where the TWC offshore meanders occurred in the mean currents (Fig. 5a). Thus, the offshore transports were most significant along the latitudes 26.5 and 28∘ N according to both the mean currents and the synoptic fluctuations. The offshore transport may be associated with the offshore-penetrating fronts of coastal water in the ECS. Many remote sensing images (He et al., 2010; Bai et al., 2013) have exhibited offshore-penetrating fronts that crossed the 70 m isobath and played an important role in cross-shore material exchange, but the mechanisms of the offshore-penetrating fronts are still under debate. Yuan et al. (2005) pointed out that both downwelling- and upwelling-favorable winds are associated with the occurrence of the offshore-penetrating front. Ren et al. (2015) suggested that the penetrating front is generated by the transition of northeasterly to southwesterly winds. Wu (2015) suggested that the offshore-penetrating front is the response of buoyant coastal water to an along-isobath undulation of the ambient pycnocline, which is controlled by a temperature stratification of the water column. Our study offers a new interpretation; i.e., the penetrating front is generated through the wind-induced fluctuations and the TWC offshore meanders.