1 INTRODUCTION

The East China Sea (ECS) is one of the marginal seas in the western Pacific Ocean. The marine and ecological environments of the ECS are influenced strongly by the Taiwan Warm Current (TWC), Changjiang (Yangtze) River, and Kuroshio Current (KC). Previous studies have shown that the Kuroshio subsurface water (KSSW) transports nutrients in huge quantities to the ECS continental shelf through oceanic upwelling to the northeast of Taiwan, China (Chen, 1996; Chen and Wang, 1999; Liu et al., 2000). The KSSW can reach as far as the coastal area of Zhejiang Province in China (Fig. 1a) (Wang et al., 2016), and the nutrients it carries promote plankton growth in this area that contributes to large-scale harmful algal blooms (Yang et al., 2013). As these algal blooms affect both the coastal ecology and the nearby fisheries (Yang et al., 2013; Hu and Wang, 2016), it is important to study the dynamics of both the KSSW and the branches of the KC intrusion onto the ECS shelf.

Many previous studies focused on the intrusion pathways and the spatial distribution of the KSSW. For example, based on a passive tracer experiment using a nested ocean model, Guo et al. (2006) suggested that the KSSW dominates the bottom-layer waters of the ECS continental shelf throughout the year. Kondo (1985) highlighted that the KC collides with the continental shelf break to the northeast of Taiwan, China, and forms a flow that runs almost northward. This current, known as the Kuroshio Branch Current northeast of Taiwan, China, moves northward at 123.5°E over the southern part of the ECS continental shelf, just to the east of Hangzhou Bay, China (Fig. 1a). However, Ichikawa and Beardsley (2002) demonstrated that the Kuroshio Branch Current northeast of Taiwan, China, exists only to the south of 29°N over the ECS shelf. Using long-term (25 years) satellite data and other observational information, Wang and Oey (2016) found that the KSSW could intrude as far north as 30°N over the ECS continental shelf. Based on summer cruise observations and Regional Ocean Modeling System (ROMS) simulations, Yang et al.(2011, 2012) reported that there are two branches of the KSSW in the ECS (Fig. 1a): the Nearshore Kuroshio Branch Current (NKBC), which can intrude into the Zhejiang coastal area and reach as far north as 31°N, and the Offshore Kuroshio Branch Current (OKBC), which flows along the 100-m isobath. Using available hydrological data combined with chemical indicator observations of the KSSW, Zhou et al. (2017) found that the KSSW moves in the North-West (NW) direction and intrudes onto the ECS shelf northeast of Taiwan, China, during spring, and that this intrusion consists of a nearshore branch and an offshore branch at 27.5°N–28.5°N. These branches exhibit marked seasonal variation. For example, Xu et al. (2018) revealed that the KSSW intrudes into the Zhejiang coastal area in June rather than in December, while Zhou et al. (2018) highlighted that the extent of intrusion of the nearshore Kuroshio branch strengthens (weakens) in spring and summer (autumn and winter).

Several factors influence the intrusion of the KSSW onto the ECS continental shelf. Jacobs et al. (2000) revealed that a strong TWC could generate a strong bottom Ekman flow, which drives the northward Kuroshio branch current beneath the TWC. Similarly, wind stress during winter over the ECS continental shelf could increase the Kuroshio flow into the Yellow Sea. Oey et al. (2010) found that winter cooling could explain the anticyclonic branch of the Kuroshio intrusion onto the ECS. Yang et al. (2018a) proposed a topographic beta-spiral theory to explain the vertical current structure of the KSSW intrusion to the northeast of Taiwan, China, during summer. Xu et al. (2018) determined that bottom Ekman current variation is the fundamental process regarding the intrusion of the KSSW into the Zhejiang coastal area, and highlighted that seasonal variation of KSSW intrusion is induced primarily by the different wind stress between summer and winter.

Although it is widely accepted that the NKBC exists and that it can intrude into both the Zhejiang coastal area and the Hangzhou Bay area, the dynamic mechanism of the seasonal variation of the NKBC remains unclear. The question of how the regional wind, TWC, and KC might influence the NKBC also requires further study. In this study, we successfully simulated the circulation structure of the ECS using the Regional Ocean Model System (ROMS). By defining the cross-shore range and the intensity of the NKBC, we investigated the reasons for the seasonal variation of the NKBC. In addition, a series of sensitive experiments were performed to assess the relative influence of the regional wind, TWC, and KC on the cross-shore range and the intensity of the NKBC and its dynamic process.

2 OBSERVATION AND NUMERICAL MODEL 2.1 Observation

Cruise observations acquired in the ECS during 19–20 May 2019 and 12–25 September 2019 were used to determine the spatial distribution of the Kuroshio subsurface water on the ECS shelf in spring and autumn, respectively. As shown in Fig. 1a, the cruise observations were conducted along two transects (DH2 and DH3) aligned perpendicular to the coast of Zhejiang Province. In the spring cruise (Fig. 1b), transect DH2 comprised DH2-1, DH2-2, DH2-3, DH2-4, DH2-5, DH2-6, and DH2-7 and transect DH3 comprised DH3-1, DH3-2, DH3-3, DH3-4, DH3-5, and DH3-6. In the autumn cruise (Fig. 1c), transect DH2 comprised DH2-1, DH2-2, DH2-3, DH2-4, and DH2-5 and transect DH3 comprised DH3-1, DH3-2, DH3-3, DH3-4, and DH3-5. The number and spacing of the stations on transect DH2 and transect DH3 were slightly different between the spring and autumn cruises; however, the overall position and scope of the transects were consistent. Water temperature and salinity observations were collected at each station using Sea-Bird conductivity-temperature-depth sensors.

2.2 Numerical model 2.2.1 Model description

The simulations in this study were performed using the ROMS, which is based on the primitive equations for a hydrostatic Boussinesq fluid, is an example of a three-dimensional, free-surface, and terrain-following numerical model (Haidvogel et al., 2008), and is widely used in the study of regional oceanography. The model uses the vertical mixing parameterization introduced by Mellor and Yamada (1982). Details regarding the computational algorithms of the ROMS can be found in Shchepetkin and McWilliams (2005).

The model domain (22°N–42°N, 117°E–135°E; Fig. 1a) included the entire ECS and the northwestern Pacific Ocean. Guo et al.(2003, 2006) reported that model resolution should be < 10 km to resolve the steep shelf break of the ECS along which the KC flows. Therefore, the model in this study had horizontal resolution of 1/18°×1/18°ϕ (where ϕ is latitude). In the vertical direction, the model had 28 nonlinear terrain-following coordinate (s-coordinate) layers, with finer grid spacing near the sea surface and the bottom (Song and Haidvogel, 1994; Haidvogel et al., 2008). Bottom topography was extracted from the General Bathymetric Chart of the Oceans.

2.2.2 Climatology runs

The climatological monthly mean temperature and salinity in January derived from WOA13 (http://www.nodc.noaa.gov/OC5/woa13/) were used to initialize the model. Additionally, the initial current velocity of the model and the sea surface height were all set to zero. The model was driven by climatological monthly mean wind, net surface longwave radiation fluxes, precipitation rate, surface solar shortwave radiation fluxes, surface air pressure, relative humidity, surface air temperature, sea surface temperature, and sea surface salinity data, which were all introduced as 2011-2019 averages derived from NCEP CFSv2. Climatological monthly mean temperature, salinity, velocity, and sea surface height data, which were taken as averages from January 2015 to December 2019 from HYCOM 3.1 analysis data, were used for lateral boundary conditions. The model was also forced using the climatological monthly freshwater discharge of the Changjiang River and eight tidal constituents (M₂, S₂, N₂, K₂, K₁, O₁, P₁, and Q₁) extracted from TPXO7 (Egbert and Erofeeva, 2002). The model was integrated for 20 years to spin-up.

2.2.3 Hindcast runs

After the climatological runs stabilized, the results for the final month of the climatological runs were used as the initial field of the hindcast runs. During the hindcast duration from January 2017 to December 2019, the model was forced by the daily mean wind, net surface longwave radiation fluxes, precipitation rate, surface solar shortwave radiation fluxes, surface air pressure, relative humidity, surface air temperature, sea surface temperature, and sea surface salinity data, which were derived from NCEP CFSv2. The freshwater discharge of the Changjiang River and the eight tidal constituents were consistent with the climatology runs. The monthly mean temperature, salinity, velocity, and sea surface height data from HYCOM 3.1 analysis data from January 2017 to December 2019 were used for lateral boundary conditions. The time evolution of the volume-averaged kinetic energy of the model is shown in Fig. 2.

2.2.4 Index of the KSSW intrusion

To intuitively show the NKBC, we adopted a passive-tracer modeling method (Hu et al., 2020). We distinguished Kuroshio water from ECS shelf water by releasing a passive tracer "dye" vertically (from the sea surface to the bottom) along the entire Pacific boundary (east of Taiwan, China, along the southern boundary and south of Japan along the eastern boundary) of the model domain (Fig. 1a). The concentration of passive tracer was fixed at 100% at the release location from the sea surface to the bottom, whereas it was fixed at zero at the other boundaries.

3 RESULT 3.1 Observed result

The observed distributions of temperature and salinity along transects DH3 in May 2019 are shown in Fig. 3a & c and those of DH2 in Fig. 3b & d, respectively. There are remarkable differences in the distributions of temperature and salinity between the two transects. Along DH3, which is to the south of DH2, the bottom layer is occupied by waters with high salinity (> 34.5) and low temperature (< 22 ℃), whereas along DH2, there are two obvious cores of high salinity and low temperature in the bottom layer. The nearshore core (salinity: > 34.5, temperature: < 20 ℃) is located near the 60-m isobath, while the offshore core is located near station DH2-7.

The observed distributions of temperature and salinity along transect DH3 in September 2019 are shown in Fig. 4a & c, respectively. In comparison with May 2019, the temperature of the water in the upper layer is higher, while in the bottom layer, the water with high salinity (> 34.2) and low temperature (< 22 ℃) exists only in the area west of station DH3-2. The observed distributions of temperature and salinity along transect DH2 in September 2019 are shown in Fig. 4b & d, respectively. In comparison with May 2019, the nearshore core of high salinity in the bottom layer is unclear and the salinity is low, while the offshore core located near station DH2-5 in autumn remains distinct. The water with high salinity and low temperature in the bottom layer of the ECS during spring and autumn has been proven to originate from the KSSW (Zhou et al., 2017; Yang et al., 2018b). The nearshore core of high salinity and low temperature in the bottom layer of DH2 is caused by the NKBC, whereas the offshore core of high salinity and low temperature in the bottom layer of DH2 is caused by the OKBC. The substantial difference between the two cruises indicates that the intensity of the NKBC in September 2019 is much weaker than in May 2019.

3.2 Model validation

To investigate the reasons for the difference in the NKBC between spring and autumn, we simulated the circulation structure of the ECS using the ROMS. Details of the model settings are described in Section 2.2. To validate model reproducibility, the annual mean volume transport through both the east of Taiwan (ET) section and the Taiwan Strait (TS) section (see Fig. 1a for the range and position of each section) were compared with the results of previous studies (Table 1). The annual mean volume transport through the TS and ET sections based on the results of the final modeled year of the climatology runs were 1.62 and 23.54 Sverdrup (1 Sverdrup (Sv)=1×10⁶ m³/s), respectively, which are close to the values estimated by Wang et al. (2003), Lee and Chao (2003), and Guo et al. (2006). Moreover, the seasonal variation of the volume transport through section TS from the climatology runs (Fig. 5b) is also consistent with the observation by Chen et al. (2016), i.e., the monthly mean volume transport through section TS is greater in spring and summer than in autumn and winter. Additionally, the seasonal variation of the volume transport through section ET from the climatology runs (Fig. 5a) is also consistent with the observation by Lee et al. (2001), i.e., the monthly mean volume transport through section ET is large in spring and summer and small in autumn and winter.

The profiles of simulated salinity and temperature of transects DH3 and DH2 for May 2019 (Fig. 6) and September 2019 (Fig. 7) were also compared with the aforementioned observations. In May 2019, the temperature (salinity) decreases (increases) with depth along DH3 (Fig. 6a & c), and the entire bottom layer is occupied by water with low temperature and high salinity, consistent with the observed data. Along DH2 (Fig. 6b & d), while the salinity of the upper layer is higher than observed and the lowest temperature is two degrees lower than observed, two cores of water with low temperature and high salinity in the bottom layer are reproduced clearly in the model results. In the hindcast results for September 2019, the distribution of temperature is generally reproduced well by the model along DH3, whereas the salinity is higher than observed, especially in the upper layer (Fig. 7a & c). This discrepancy is considered attributable to the difference between the climatological monthly freshwater discharge of the Changjiang River used in our hindcast model and the actual freshwater discharge of the Changjiang River in 2019. However, in the bottom layer, the distribution of water with low temperature and high salinity in September 2019 is reproduced well in comparison with May 2019, when the nearshore core of high salinity in the bottom layer is unclear and the salinity is low. This means that the weak NKBC in autumn 2019 was simulated successfully in our model. The successful reproduction of the difference in the NKBC between May 2019 and September 2019, which was the focus of this study, means that the model could be considered suitable for examination of the NKBC.

3.3 The cross-shore range and the intensity of the NKBC

The monthly mean current velocities in the surface and bottom layers in May 2019 and September 2019 from the results of the hindcast runs are shown in Fig. 8. The model simulates the circulation structure of the ECS and its adjacent sea areas well. In the upper ocean, after impinging on the shelf break northeast of Taiwan, China, the KC turns northeastward and flows along the shelf break. The intrusion of the KC onto the ECS shelf from northeast Taiwan, China, is weak. Meanwhile, the TWC flows strongly into the ECS. In the bottom layer, the KC can intrude into the ECS from the seas off northeastern Taiwan, China. In May 2019, the Kuroshio intrusion in the area (Fig. 8b) divides into two branches: the NKBC moving northeastward along the 60-m isobath and the OKBC that turns eastward. In September 2019, the NKBC is almost disappeared, whereas the OKBC remains as obvious as in May 2019.

In this work, to distinguish the NKBC and the OKBC, a passive tracer "dye" was used to represent the concentration of the Kuroshio water. The concentration of "dye" along DH3 (Fig. 9a & c) shows that the Kuroshio intrusion exists mainly in the bottom layer of the ECS. Therefore, the spatial concentration of "dye" in the bottom layer of the ECS during May 2019 and September 2019 is shown in Fig. 9b & d, respectively. In May 2019, after the Kuroshio intrusion reaches the DH3 section, it is evident that it divides into the NKBC and the OKBC. However, in September 2019, after the Kuroshio intrusion reaches the DH3 section, the NKBC almost disappears. This means that the DH3 section is the principal section at which to assess and quantify the NKBC. Therefore, the DH3 section is used for further analysis in this paper.

We oriented the currents of the ECS with respect to transect DH3, which was almost perpendicular to the coast of Zhejiang Province. The x-axis was taken as positive in the offshore direction, and the y-axis was taken as positive in the poleward direction (Fig. 1b). When the Kuroshio intrusion reaches the DH3 section, if its alongshore velocity is positive and the cross-shore velocity is onshore, then the Kuroshio intrusion would form the NKBC and intrude northward along the coastline; conversely, the Kuroshio intrusion would form the OKBC and move offshore. Therefore, the direction of the cross-shore velocity along DH3 is an important index with which to determine the cross-shore range of the NKBC. Moreover, the magnitude of the alongshore velocity in the cross-shore range of the NKBC can be used to represent the intensity of the NKBC. The alongshore velocity (V) and cross-shore velocity (U) along DH3 in May 2019 and September 2019, derived from the monthly mean results of the hindcast runs, are shown in Fig. 10. In May 2019, except for some areas to the west of station DH3-3, the cross-shore velocity is offshore. The onshore velocity is distributed mainly in the bottom area to the west of station DH3-3, and the range where the cross-shore velocity is onshore is largest in the lowest layer. As for the alongshore velocity, the overall performance is northward and it is largest at station DH3-2. In September 2019, the cross-shore velocity and alongshore velocity are markedly different from those in May 2019, i.e., the onshore velocity is disappeared and the northeastward alongshore velocity is decreased substantially.

As the NKBC is at the bottom of the ECS, and the range where the cross-shore velocity is onshore is largest in the lowest layer in May 2019, we used the cross-shore velocity and alongshore velocity in the bottom layer along the DH3 section to study the range and intensity of the NKBC.

The alongshore velocity and cross-shore velocity along DH3 in the bottom layer in May 2019 are shown in Fig. 11. According to the difference in sign of the onshore velocity in the bottom layer along DH3 (Fig. 11b), the section can be divided into three parts: region A (122.0°E–122.2°E), region B (122.2°E–122.75°E), and region C (122.75°E–123.6°E). In region B, the cross-shore velocity is onshore, while the cross-shore velocity is offshore in both region A and region C. The opposite cross-shore velocity in regions B and C leads to the formation of the NKBC and OKBC. The offshore velocity in region A suggests that the NKBC in the bottom layer of the ECS cannot reach region A. Therefore, we define region B as the cross-shore range of the NKBC during May 2019. The strength of the alongshore velocity in the bottom layer at the cross-shore range of the NKBC is used to represent the intensity of the NKBC. In September 2019 (Fig. 12), the alongshore velocity is positive only to the east of 122.5°E, while the cross-shore velocity along the bottom s-coordinate layer is always positive. It means that the cross-shore range of the NKBC in September 2019 is nonexistent. Therefore, the nearshore core of low temperature and high salinity in DH2 during September 2019 is very weak.

The primary issue was to determine the mechanism that controls the cross-shore range and the intensity of the NKBC. To resolve this problem, we resorted to the momentum conservation equations (Eqs.1 & 2). Here, the equations are decomposed into geostrophic and ageostrophic terms to examine the mechanisms governing the formation of the NKBC and its movement in the bottom layer of the ECS:

where u and v represent the cross-shore and alongshore currents, respectively, f is the Coriolis parameter, ν_H and ν_V are the horizontal and vertical eddy viscosity coefficients, respectively. All terms were estimated diagnostically from the monthly mean simulated results for May and September 2019. Here, V_accel and U_accel represent the contributions of acceleration, V_adv and U_adv represent the contributions of advection, V_prsgrd and U_prsgrd represent the contributions of the pressure gradient(the geostrophic current), V_hvisc and U_hvisc represent the contributions of horizontal eddy diffusion, and V_vvisc and U_vvisc represent the contributions from the Ekman flow in Eqs.1 & 2.

The alongshore velocity near the bottom can be decomposed into five terms (Figs. 11a & 12a). With the exception of the nearshore part of transect DH3 having an alongshore velocity that is affected by both the acceleration term V_accel and the advection term V_adv, the alongshore velocity near the bottom of transect DH3 is determined mainly by the cross-shore pressure gradient V_prsgrd and the Ekman flow V_vvisc. The cross-shore velocity and velocity contributions along the bottom layer of DH3 in May 2019 and September 2019 are shown in Figs. 11b & 12b, respectively. It can be seen that the contribution of the acceleration term U_accel, advection term U_adv, and horizontal eddy diffusion term U_hvisc are relatively small. Thus, the movement of bottom water shoreward and offshoreward is governed mainly by the pressure gradient term U_prsgrd and Ekman flow term U_vvisc.

According to the velocity contributions in the bottom layer along DH3, the shoreward bottom Ekman current, caused by the northeastward alongshore velocity and bottom friction, supports the NKBC onshore intrusion. However, the offshoreward geostrophic flow diminishes the NKBC onshore intrusion. This ongoing battle between the bottom Ekman current and the geostrophic current determines the cross-shore range of the NKBC. In comparison with May 2019, the combination of a weakened shoreward bottom Ekman current and an increased offshoreward geostrophic flow makes the cross-shore range of the NKBC disappear in September 2019. Meanwhile, the weakened northeastward alongshore geostrophic flow in September 2019 also weakens the intensity of the NKBC.

4 DISCUSSION

To investigate the cause of the marked difference in the NKBC in May 2019 and September 2019, we conducted sensitivity experiments in which the KC velocity, TWC velocity, and wind speed of May 2019 were replaced in turn with those of September 2019 (Table 2). The time series of the volume transport through section ET and section TS based on the results of the hindcast runs during 2019 (Fig. 5) show that the KC velocity and the TWC velocity in May are much greater than in September. The average wind speed for May and September 2019 derived from the NCEP CFSv2 is shown in Fig. 13. The control run was performed under normal conditions. Each experiment ran to May 2019 from January 2019. The monthly averaged results in May of each experiment were used for comparison.

The cross-shore velocity and the alongshore velocity along the DH3 section at the bottom layer in each sensitivity experiment was compared with the results of the hindcast model in May 2019 and September 2019 (Fig. 14). Along DH3 to the west of 122.5°E, the cross-shore velocity in the bottom layer is influenced mainly by the wind, whereas to the east of 122.5°E, the bottom cross-shore velocity is controlled by the wind, TWC, and KC. Both weak KC velocity and weak TWC velocity can lead to the eastern boundary of the NKBC moving shoreward. A strong southwestward wind can make the eastern (western) boundary of the NKBC move shoreward (offshoreward) (Fig. 14a). As the southwesterly wind is so strong in September, the east and west boundaries of the NKBC move closer together until the NKBC disappears. The alongshore velocity in the bottom layer can also be influenced by the wind, TWC, and KC. A strong southwestward wind can lead to reduced northeastward velocity. In particular, to the west of 122.4°E along DH3, the alongshore velocity is controlled primarily by the wind. However, to the east of 122.4°E in the bottom layer along DH3, weak TWC (KC) velocity can decrease (increase) the northeastward velocity. Although the wind, TWC, and KC can affect the NKBC, it is evident that the difference in the NKBC between May and September 2019 was caused mainly by the difference of the wind.

To further explore the dynamic process of the influence of the wind, TWC, and KC on the NKBC, the bottom velocity and its main velocity contributions (pressure gradient term and bottom Ekman flow term) in each sensitivity experiment were compared with the results of the hindcast model for May 2019.

The differences in bottom cross-shore velocity and its main velocity contributions along transect DH3 between the Wind-case run and the control run are shown in Fig. 15a. When the wind was replaced by a more southwesterly wind in September 2019, the wind could make the western boundary of the NKBC move offshoreward by increasing the offshore geostrophic current in the area near the western boundary of the NKBC in May 2019 (122.0°E–122.6°E, maximum difference of the geostrophic current: 0.045 m/s). Furthermore, the wind could also make the eastern boundary of the NKBC move shoreward by decreasing the onshore bottom Ekman current in the area near the eastern boundary of the NKBC in May 2019 (122.3°E–123.1°E, maximum difference of the bottom Ekman current: 0.05 m/s). Additionally, the southwesterly wind could decrease the northeastward alongshore velocity by reducing the northeastward geostrophic current (maximum difference of the geostrophic current: 0.25 m/s) (Fig. 15b), which then reduces the intensity of the NKBC.

The differences in bottom cross-shore velocity and its main velocity contributions along transect DH3 between the TWC-case run and the control run are shown in Fig. 16a. When the TWC velocity is replaced by that of September 2019, the weakened TWC velocity of September 2019 (Fig. 5b) can make the eastern boundary of the NKBC move shoreward by decreasing the onshore bottom Ekman current in the area near the eastern boundary of the NKBC in May 2019 (122.5°E–123°E, maximum difference of the bottom Ekman current: 0.025 m/s). Under the influence of a weakened northeastward geostrophic current (maximum difference of the geostrophic current: 0.04 m/s), the intensity of the NKBC is weakened in the region near the eastern boundary of the NKBC (Fig. 16b).

Additional experiments in which the KC velocity was replaced (Table 2) were used to diagnose the influence of KC velocity on the NKBC. The differences in bottom cross-shore velocity and its main velocity contributions along transect DH3 between the KC-case run and the control run are shown in Fig. 17a. When the KC velocity was replaced by that of September 2019, the movement of the western boundary of the NKBC is determined mainly by the increased offshore geostrophic current in the area near the eastern boundary of the NKBC in May 2019. Under the influence of an enhanced northeastward geostrophic current, the intensity of the NKBC is increased (Fig. 17b).

5 CONCLUSION

Observed temperature and salinity data revealed that the Kuroshio intrusion comprised the NKBC and the OKBC at the bottom of the ECS in May 2019, and that the NKBC was weak in September 2019. Our high-resolution model reproduced the hydrographic characteristics on the ECS continental shelf during 2019. By analyzing the flow field and the distribution of "dye" that was used to reflect the Kuroshio water, we found that the velocity along the DH3 section in the bottom layer is the key index with which to assess and quantify the NKBC. After the bottom currents of the ECS were oriented in a cross-shore and alongshore coordinate system with respect to transect DH3, according to the cross-shore velocity, the cross-shore range of the NKBC was defined as the region where the cross-shore velocity was shoreward in the bottom layer of transect DH3. The strength of the alongshore velocity in the bottom layer at the cross-shore range of the NKBC was used to represent the intensity of the NKBC.

Analyses of the momentum balances indicated that the boundary of the NKBC in the bottom layer is determined by a combination of the geostrophic flow and the bottom Ekman current. In comparison with May 2019, a weakened shoreward bottom Ekman current and an increased offshoreward geostrophic flow caused the cross-shore range of the NKBC to disappear in September 2019. Meanwhile, a weakened northeastward alongshore geostrophic flow in September 2019 also diminished the intensity of the NKBC. Therefore, the NKBC was weak in September 2019. Based on sensitivity experiments, we found that a strong southwestward wind can push the western (eastern) boundary of the NKBC offshoreward (shoreward) by increasing (decreasing) the offshore geostrophic flow (the bottom Ekman current). A weak TWC can move the eastern boundary of the NKBC shoreward by decreasing the onshore bottom Ekman current; however, it has little effect on the western boundary of the NKBC. A weak KC can move the eastern boundary of the NKBC shoreward by increasing the offshoreward geostrophic flow. Meanwhile, a strong southwestward wind, weak TWC, and strong KC can diminish the intensity of the NKBC; conversely, the intensity of the NKBC would be increased. Results indicate that the wind plays the major role in influencing the NKBC. This work furthered the study of the dynamic mechanism of the seasonal variation of the NKBC, and revealed the relative influences of the wind, TWC, and KC on the cross-shore range and the intensity of the NKBC and its dynamic process.

6 DATA AVAILABILITY STATEMENT

The WOA13 and NCEP CFSv2 data is maintained and distributed by National Oceanic and Atmospheric Administration (NOAA) at http://www.nodc.noaa. gov. The HYCOM 3.1 data are available at https://www.hycom.org/. The bottom topography is extracted from the General Bathymetric Chart of the Oceans (1/120°) at https://www.gebco.net/data_and_products/gridded_bath-ymetry_data/.

7 ACKNOWLEDGMENT

Data and samples were collected onboard of R/V Xiang Yang Hong 18 implementing the open research cruise NORC2019-02 supported by NSFC Shiptime Sharing Project (project number: 41749902). The study was also supported by the High Performance Computing Center at IOCAS.