RESEARCH ARTICLE New Insight Into the Onshore Intrusion of the Kuroshio

10.1029/2020JC016248
Into the East China Sea
Key Points:
Xuan Cui1,2,3,4 , Dezhou Yang1,2,3,4 , Chaojiao Sun5 , Xingru Feng1,3,4 ,
• T o study the Kuroshio intrusion into
Guandong Gao1,3,4, Lingjing Xu1,4 , and Baoshu Yin1,2,3,4
the East China Sea, we obtained
a new formula using a two- 1
CAS Key Laboratory of Ocean Circulation and Waves, Institute of Oceanology, Chinese Academy of Sciences, Qingdao,
dimensional momentum balance
• This new formula addresses the China, 2College of Earth and Planetary Sciences, University of Chinese Academy of Sciences, Beijing, China, 3Function
mechanism of the Kuroshio Branch Laboratory for Ocean Dynamics and Climate, Qingdao National Laboratory for Marine Science and Technology,
Current (KBC) recirculation, Qingdao, China, 4Center for Ocean Mega-Science, Chinese Academy of Sciences, Qingdao, China, 5CSIRO Oceans &
whereas the previous one does not
Atmosphere, Crawley, Western Australia, Australia
• The new formula explains the
seasonal variation in the KBC and
may potentially be applied to other
similar regions Abstract  The Kuroshio Current, a vigorous western boundary current in the Pacific, is responsible
for bringing high temperature, high salinity, and phosphate-rich open ocean water into the East China Sea
(ECS) when it flows past Taiwan and generates a branch current, known as the Kuroshio Branch Current
Correspondence to:
D. Yang and B. Yin
(KBC), which intrudes onto the continental shelf of the ECS. Based on observations and numerical
[email protected]; modeling, a 1.5-layer reduced-gravity model is evaluated and modified to simulate the flow pattern of the
[email protected] Kuroshio intrusion (KI) northeast of Taiwan. We revised the analytical solution derived by Hsueh (1992),
https://doi.org/10.1029/92JC01401 to conserve the horizontal momentum balance in the along-shelf
Citation: and cross-shelf directions. The new analytical solution not only explains the KI angle but also addresses
Cui, X., Yang, D., Sun, C., Feng, X., the KBC recirculation mechanism. To sustain the steady-state momentum balance in the area adjacent
Gao, G., Xu, L., & Yin, B. (2021). New
insight into the onshore intrusion
to the shelf break northeast of Taiwan, there must be a recirculation that plays the same role in vorticity
of the Kuroshio into the East China dynamics as the combination of the joint effect of baroclinicity and relief, bottom friction, and nonlinear
Sea. Journal of Geophysical Research: advection. Unlike previous studies, our results agree well with the observations and could account for
Oceans, 126, e2020JC016248. https://
doi.org/10.1029/2020JC016248
seasonal variations in the KI direction. Moreover, the new model predicts a reasonable velocity maximum
(0.42 m s−1) over the shelf break. Satellite data and numerical results confirm that the new solution
Received 29 APR 2020 reflects a reasonable estimation for KI direction. This theory may potentially be applied to regions with
Accepted 31 DEC 2020 trenches or valleys with completely different flow patterns by using similar analytical methods.

Plain Language Summary  The northeastward-flowing Kuroshio Current is one of the

most important western boundary currents in the global ocean. One branch of the Kuroshio Current
enters the East China Sea transporting heat and nutrients after encountering the zonally running shelf
break northeast of Taiwan. We have developed a new theoretical model to explain the mechanism of
the Kuroshio intrusion into the East China Sea. The new model provides a unified understanding of the
behaviors of the Kuroshio intrusion: its angle of intrusion, recirculation, and seasonal variation. The
analytical solution from the new model is highly consistent with the results from a state-of-the-art ocean
general circulation model and is examined using satellite data and the World Ocean Atlas 2018.

1. Introduction
The Kuroshio Current plays a crucial role in the subtropical gyre of the Pacific circulation by compensating
for southward transport in the ocean interior induced by the negative wind stress curl; this current brings
nutrients, high salinity, and high-temperature water downstream (Andres et al., 2015, 2017; Isobe, 2008).
This vigorous poleward western boundary current creates a complex ocean current system when it flows
over steep topography northeast of Taiwan where canyons often cut through the continental slope (e.g., X.
© 2021. The Authors.
Guo et al., 2006, 2003; Hsueh et al., 1993, 1997, 1992; Tang et al., 2000; Yang et al., 2012, 2018, 2011). Moreo-
This is an open access article under ver, this current system becomes even more complex due to the energetic interactions between the Kuroshio
the terms of the Creative Commons and westward propagating mesoscale eddies that originate from the tropical North Pacific (e.g., Vélez-Belchí
Attribution NonCommercial License,
which permits use, distribution and
et al., 2013; Y. Zhang et al., 2020).
reproduction in any medium, provided
the original work is properly cited and The Kuroshio intrusion (KI) onto the shelf northeast of Taiwan has been a focus of intensive research over
is not used for commercial purposes. the past 3 decades. Both observations and modeling results have shown that as the Kuroshio mainstream

CUI ET AL. 1 of 19
 Journal of Geophysical Research: Oceans 10.1029/2020JC016248

flows past the island of Taiwan, a branch of the current, the Kuroshio
Branch Current (KBC), intrudes onto the continental shelf of the East
China Sea (ECS). After reaching the ECS continental shelf, the KBC turns
southeastward and rejoins the mainstream of the Kuroshio, forming the
KBC recirculation. Qiu and Imasato (1990) demonstrated the existence
of the KBC using long-term geomagnetic electrokinetograph measure-
ments, the flow pattern of which is shown in Figure 1. Reportedly, there
are two major locations where Kuroshio waters intrude onto the ECS
shelf: the KI northeast of Taiwan at approximately 26°N and a Kurosh-
io branch that separates southwest of Kyushu (X. Guo et al., 2006). The
KI onto the ECS continental shelf was first directly observed by acoustic
Doppler current profiler (ADCP) data, which showed an abrupt change
in the upper-layer flow field (Tang & Yang, 1993). Subsequently, season-
al and intraseasonal variations in the Kuroshio migration east of Tai-
wan have been investigated (Doong et al., 2019; Wang & Oey, 2016; Wu
et  al.,  2014). It was observed that the Kuroshio current east of Taiwan
migrates shoreward in winter and seaward in summer (Tang et al., 2000)
and exhibits large variability (Jan et al., 2015). Several studies have report-
ed that the 100-day fluctuation in the Kuroshio volume transport mainly
Figure 1.  The Kuroshio Current northeast of Taiwan, the KBC, and its
results from impingements of eddies from the open ocean (e.g., Chern
recirculation. Topography and bathymetry are adopted from the ETOPO-1
data set of the National Geophysical Data Center (Amante & Eakins, & Wang, 2005; Johns et al., 2001; D. Zhang et al., 2001). In contrast, Wu
2009). KBC, Kuroshio Branch Current. et al. (2017) argue that effects from westward propagating cyclonic eddies
that originate from the interior Pacific Ocean have significant interan-
nual variability. Based on long-term satellite observations and high-res-
olution global ocean circulation model data, the interannual variations
in the KI volume and structure have been attributed to cyclonic eddies immediately west of Kuroshio, and
these eddies are induced by the seaward deflection of the Kuroshio axis (C. Liu et al., 2014). The KI is closely
related to the cold dome northeast of Taiwan and the migration of the Kuroshio path (Tang et al., 2000). The
interannual variation in the Kuroshio path northeast of Taiwan is also strongly related to the strong wind
stress curl and surface heat flux in this region (Wang & Oey, 2014). Such a variation can be established or
sustained by a positive feedback related to the joint effect of baroclinicity and relief (JEBAR), as well as the
upwelling dynamics (Oey et al., 2018).

Several studies have been carried out to explain the KI mechanism northeast of Taiwan. Qiu and Imasa-
to (1990) concluded from numerical model experiments that the planetary  -effect and the island of Tai-
wan are necessary for the formation of the KBC. The  -effect, which balances the vortex stretching effects
caused by the shelf break northeast of Taiwan, drives part of the Kuroshio to flow southwestward. The
blocking of this flow by Taiwan Island causes it to intrude onto the continental shelf until it gradually turns
to the southeast and rejoins the mainstream of the Kuroshio Current. A 1.5-layer reduced-gravity model was
proposed by Hsueh et al. (1992, 1993), and an analytical solution for the angle of the KI was obtained by
integrating the along-shelf momentum flux. Ichikawa and Beardsley (1993) suggested that 75% of the Kuro-
shio volume transport in the ECS is barotropic, which supports the assumption in Hsueh's model (1992)
that vertical shear in the shallow regions is not considered. Using a three-dimensional, primitive equation
model, Chao (1990) attributed the enhancement of the KI northeast of Taiwan to the shoreward Ekman
drift induced by the northeasterly monsoon. Based on Chao's model, the intensifications of upwelling and
intrusion northeast of Taiwan are carefully examined (K. K. Liu et al., 1992). X. Y. Guo et al. (2003, 2006)
established a triply nested model and found that the JEBAR term, which is related to the variation in the
isopycnal layer (Mertz & Wright, 1992), is the predominant term controlling transport across the 200-m iso-
baths in the ECS. However, Oey et al. (2010) noted that the KI caused by winter cooling alone is comparable
with the observations (see their Equation 2), which provides evidence for the conclusion of Chuang and
Liang (1994) that cooling, rather than the direct wind forcing, is responsible for the winter intrusion. Both
numerical models and observations indicate that there are two Kuroshio branches in the lower layer of the
ECS (Yang et al., 2011, 2012), the nearshore KBC and the offshore KBC. The two branches turn anticyclon-
ically from the bottom to the surface, forming an anticyclonic spiral, which is known as the topographic

CUI ET AL. 2 of 19
 Journal of Geophysical Research: Oceans 10.1029/2020JC016248

beta spiral. This theory could account for the vertical structure of the intrusion path into the ECS (Yang,
Yin, et al., 2018). Therefore, it can be concluded that the unique topography, the planetary  -effect, and the
winter cooling northeast of Taiwan are essential for the KI. The water and material exchange between west-
ern boundary currents and marginal seas is attracting immense research interests because it significantly
influences coastal hydrodynamic environments and ecosystems. Because of the steep topography associated
with shelf breaks and the strong nonlinearity of currents, the interaction between continental shelf breaks
and circulation in the open ocean remains a great challenge (Lentz et al., 1999). Northeast of Taiwan, the
KBC and its recirculation play an important role in the material exchange between ECS shelf water and the
Kuroshio Current. However, few of the abovementioned studies have focused on KBC recirculation and the
reason it could rejoin the Kuroshio mainstream remains unclear.

In this study, we propose a modified 1.5-layer reduced-gravity model in which the recirculation of the KBC
is properly considered and we obtain a new formula for regulating the onshore intrusion of the Kuroshio
Current northeast of Taiwan based on Hsueh's 1.5-layer model from 1992. The rest of the paper is organized
as follows. We will provide a brief analysis and a review of the previous model in Section 2. The modified
model and the new formula are described in detail in Section 3. Section 4 provides a brief introduction to
the numerical model and data used to examine our theory and the results. Section 5 discusses the connec-
tion between our model and the JEBAR effect, as well as its possible application. The paper ends with a
conclusion in Section 6.

2.  Hsueh's Theory

2.1.  The Framework and Its Bottleneck

To investigate the scenario of a strong current meeting a shelf break, Hsueh et al. (1992) proposed a 1.5-layer
model and obtained a formula for regulating the KI direction northeast of Taiwan. The schematic of this
model is shown in Figures 2a–2b. The model is based on a reduced-gravity model with a motionless lower
layer and a rigid lid. However, we find that one of the assumptions in this model is not reasonable and that
the flow pattern in this model is not accurate enough to represent the KI pattern northeast of Taiwan due to
limited observations and numerical studies at that time (to be discussed later). Hereafter, a brief description
of this model is present.

In Figure 2b, Area 1 represents the region of inflow away from the collision area, while Area 3 is the outflow
region blocked by the step. Areas 2, 4, and 5 are the regions located over the step. A crucial consideration
in this model is that the currents in the far field of the collision area are geostrophic and conserve potential
vorticity (PV), which has the following expression,

 V / n  f  / h f / H
(1)

 fV  gh / n

where V is the magnitude of velocity, n is the normal direction at the right side of velocity, f is the Coriolis
parameter, g'=(ρ2−ρ1)/ρ1 is the reduced gravity, h is the upper-layer thickness, and H is the undisturbed
upper-layer depth in the far field (see Figure 2a). Equation 1 suggests that the upper-layer thickness h varies
exponentially with n in Areas 1–3 with a decaying scale of the baroclinic deformation radius R  gH / f ,
while a relatively shallow constant depth H0 is set on the step so that the currents in Areas 4 and 5 are as-
sumed to be barotropic with a rigid lid. Negative relative vorticity is induced (Areas 4–5) as a result of vortex
stretching (PV conservation). As shown in Figure 2e, the density stratification in the World Ocean Atlas
2018 (WOA18) data is similar to that of Areas 1 and 3 in Figure 2a, meaning that it is relatively appropriate
to divide the water column in this domain into two layers. Please see Appendix A for the description of this
model and refer to Hsueh et al. (1992) for the derivation.

Based on the momentum balance in the along-shelf (x) direction, Hsueh et al. (1992) calculated the volume
integral of the x-momentum in domain ABCDEMF (Figure 2b). The following formula is obtained for the
KI angle  in terms of the Kuroshio incidence angle α and the depth ratio H0/H. It is an indirect method
used to investigate the KI circumventing the complicated dynamics around the impingement point with the
shelf break northeast of Taiwan.

CUI ET AL. 3 of 19
 Journal of Geophysical Research: Oceans 10.1029/2020JC016248

Figure 2.  Oblique view and top view of Hsueh's model (a) and (b) and our model (c) and (d). (e) The density slice
northeast of Taiwan calculated from WOA18 data with an idealized topography. The red shading in (a) and (c) denotes
the upper layer with a ρ1 density, the blue shading denotes the lower layer with a ρ2 density. The black dashed line
denotes the step. Areas 1–5 are marked in black rectangular boxes. The red dotted line separates Areas 2 and 4, and the
black dashed line separates Areas 3 and 5. The thin black arrows denote the normal direction of the flow along which
the normal distance n increases.

1
1     H 
2 2
4 H0  H0   H0 H  1
(2)
cos    1   1    3  cos     1  0    0
2  H  H   H    H  12 H 
 
However, as mentioned above, the flow pattern implied in his formula is not accurate enough to represent
the KI pattern northeast of Taiwan. Since the flow over the step in Area 4 (Area 5) is northwestward (west-
ward) in his model, the KBC recirculation is not included. Therefore, the flow pattern in this region is in-
consistent with the current observations and numerical results. Most importantly, the assumption implied
in this model that only the along-shelf-direction momentum is conserved is unreasonable. Thus, we present
the specific calculation process and address why and how both the along-shelf-direction momentum and
the cross-shelf-direction momentum are conserved at the same time as follows.

The time-derivative momentum equation yields the following expression:

d     
(3)
dt

 V d  
 t
V d  
 V Vn ds   ,    
 
where d is the element volume, V is the velocity, ρ is the ocean density, ds is the element area and F is
the resultant force applied to the fluid. The domain ABCDEMF in Figure 2b is specified with black dotted

CUI ET AL. 4 of 19
 Journal of Geophysical Research: Oceans 10.1029/2020JC016248

lines to evaluate the momentum balance in this region. In a steady state, the local time-derivative term
 
  
 t V d  = 0 and Equation 3 can be written in an x-y coordinate frame, where x points to the east with
its origin placed at the front and y is normal to x, which points northward, with its origin placed at the step.
In such a case, Equation 3 leads to the following expressions:
B C D E M
x
  u1v1h1dx   u3u3h3dy   u5u5h5dy   u4 v4 h4 dx   u2 v2 h2 dx
(4a)   Fx ,
A B C M F 1
B C D E M y
  v1v1h1dx   v3u3h3dy   v5u5h5dy   v4 v4 h4 dx   v2 v2 h2 dx
(4b)   Fy ,
A B C M F 1

where ℱ/ρ1 denotes the resultant force exerted on the upper-layer fluid. The left-hand side (LHS) of Equa-
tion 4 denotes the momentum flux through the domain boundaries. The x-momentum (y-momentum) is
conserved when Fx (Fy) equals zero.

When the right-hand side (RHS) of Equation 4a equals zero, it will yield the following expression of the
intrusion angle  .
1
 
2
H   4 H0     H0  
2
 cos   1  0  1 
(5)
cos    1   1    .
  H   H    H  
   
In Hsueh's model, he assumes that the ocean is in a steady state and that both Fx and Fy are zero. It is easy to
verify that Equation 5 satisfies Equation 4a but it does not satisfy Equation 4b at the same time. This result
conflicts with his assumption that both Fx and Fy are zero. This conflict can be shown in another way. If
both the x-momentum and the y-momentum are assumed to be conserved (Fx = Fy = 0), there will be two
linearly independent equations for  and hence two different solutions for  . That is, his assumption leads
Equation 4 to be an overdetermined system for  .

The fact that Equation 5 does not satisfy Equation 4b, means that Fy  0. In other words, this system feels
a force, and then its total momentum will change with time. Thus, it conflicts with his assumption that the
ocean is in a steady state. In the next section, we will show some of his model results in detail.

2.2.  Hsueh's Model Results and Disadvantages

Hsueh's model results are presented as follows. From Equation 4a, the intrusion angle  is expressed as a
function of the incidence angle α and the depth ratio H0/H (see Figures 2a–2b for details):
1
 
2
H   4 H0     H0  
2
 cos   1  0  1 
(6)
cos   Fx / S   1  1    ,
  H   H     H  
   
where S = R'H(R'f)2/6. It is obvious in Figure 3a that  increases as α increases, meaning that α needs to
be large enough for the intrusion to occur (e.g., the intrusion angle   45 , an angle between the 200-m
isobath in the ECS and the east direction). The Kuroshio current east of Taiwan mainly flows along its
eastern coast; thus,   70 in Hsueh’s paper is a suitable angle for the intrusion to occur. Figure 3b shows
that  decreases as the depth ratio increases before reaching 1. This finding means that steep topography is
favorable for the intrusion.

However, when the LHS of Equation 4b is substituted with A8 and Fy = 0, Equation 4b will yield a different
expression of  ,
1
  H  
2
sin  sin   1  1  0   .
(7)
  H  
 

CUI ET AL. 5 of 19
 Journal of Geophysical Research: Oceans 10.1029/2020JC016248

Figure 3.  Solution for Equations 6 and 7 with Fx/S = 0 and Fy/S = 0. (a) The solution of θ against the incidence angle α in Equation 6. (b) The solution of 
against the depth ratio H0/H. (c) Same as Figure 3a but for Equation 7. (d) Same as Figure 3b but for Equation 7. The black dashed lines in Figure 3a and 3c
denote the incidence angle α = 90°, while the black dashed lines in Figure 3b and 3d denote the depth ratio H0/H = 1.

The solutions for Equation 7 are shown in Figures 3c and 3d. As shown in Figure 3, it is worth noting that the
trends of these two  are opposite to each other as α or H0/H varies. It is physically unreasonable that there are
two different intrusion angles for one incidence angle α. This inconsistent phenomenon comes from the fact
that Equation 5 cannot satisfy Equations 4a and 4b at the same time. The reason is that the momentum balanc-
es yield two linearly independent equations Equations 4a and 4b for only one variable ( ). Equations 4a and 4b
requires one more variable in addition to  to make it a well-posed equation set. Physically, this model requires
an extra momentum flux to reach a steady state. A detailed analysis of this model is presented in the appendix.

3.  New Theory


To change Equation 4 into a well-posed system,
 we introduce a new variable s which exactly is the missing
V
recirculation. This supplementary current (Vs ) is added to Hsueh’s model in Areas 4–5 (see the green arrows
in Figures 2c–2d). This current represents the KBC recirculation and flows in the opposite direction to the
 
flows V4 and
 5 . Since we are more interested in the magnitude and direction than the structure of the recir-
V
culation, Vs is considered to be constant without horizontal or vertical shear throughout Areas 4 and 5. The
current flows into this domain
 from Area 4 and out of this domain from Area 5, thus the volume in this do-
main is conserved. The Vs plays a crucial role in conserving momentum in the steady state. Simultaneously,
it makes sense of the flow field in terms of the KBC and its recirculation.

The PV is still conserved in the new model. With the KBC recirculation included, it is closer to the realistic
flow pattern northeast of Taiwan. The momentum equations in both the along-step and cross-step direc-
tions yield the following expressions:

CUI ET AL. 6 of 19
 Journal of Geophysical Research: Oceans 10.1029/2020JC016248


 
2
1     
2
 cos     sin  2   2
    

(8)

2     cos  2     cos   
  
1/ 2
Vs   H   Rf
  6H 
 0 

and

cos     
  cos1
(9) ,
 

where  is a nondimensional parameter representing the squared ratio of Vs to the incoming Kuroshio
2 2
 H0   4 H0   H0 
current velocity,  1   1  ,  1  1   . Please refer to Appendix C for the detailed
 H   H   H 
derivation and calculation.

When compared with Hsueh’s model, it is important to note that the introduction of Vs leads to two addi-
tional terms:

D E
  u5su5s H 0 dy   u4s v4s H0 dx  S    1  cos 
C M
 
(10)
E ,
  v v H dx  S    sin 
 M 4 s 4 s 0

where u and v are the x and y components respectively, and the subscripts 4s and 5s denote the supple-
mentary current in Areas 4 and 5, respectively. It is exactly the absenceof these two terms that leads to the
unbalanced momentum in Hsueh's model. From another perspective, Vs can be regarded as a steady  current
induced by the unbalanced force exerted on the domain ABCDEMF in Hsueh's model. When Vs is zero, our
model degenerates into Hsueh's model. Most importantly, the new system is in a steady state, conserving
both the x-momentum and the y-momentum at the same time. Since Fx and Fy are zero, this system still
conserves the PV (f/H). The flow patterns of the Kuroshio Current and KI northeast of Taiwan (e.g., X. Y.
Guo et al., 2003, 2006; Qiu & Imasato, 1990; Yang et al., 2011, 2012, 2018) are well explained by the modified
model.

It is clearly shown in Figures 4a–4b that there is an asymptote for each H0/H or α selected when the de-
nominator of the RHS in Equation 8 reaches 0. When the solutions approach these asymptotes,  becomes
unreasonably large. Such solutions cannot be applied in the real ocean. Notably, the new solution predicts 
to be 140° when it is given the same H0/H and α as Hsueh et al. (1992). Such a result is consistent with the
conclusion in his research ( is 137°, which is defined as the orientation of the observed 21° isotherm). In
the next section, α and H0/H will be determined more rigorously. In the climatological mean state, α in this
region is approximately 60°, and H0/H is 0.56; thus, the 0.09 value of  seems reasonable (see Figures 4a–4b,
the yellow lines). In this case, Vs can be calculated to be 0.42 m.s−1 from Equation 10 upon taking g' to be
2.64  10 2 m s−2 (from the ROMS model results), and H0 and H are 140 m and 250 m, respectively (to be dis-
cussed later). Since Vn5 in Area 5 ranges from R'f(H0/H−1) to 0 (see Appendix A), 0.42 m.s−1 is the maximum
value of the recirculation speed when Vs is added to Vn5.

4. Validation
To verify the validity of our analytical solutions, we compare them with the results obtained from a state-
of-the-art ocean general circulation model, the absolute dynamic topography (ADT) data from Archiving,
Validation and Interpretation of Satellite Oceanographic (AVISO) database ranging from 1993 to 2015 and
the climatological data from the WOA18.

CUI ET AL. 7 of 19
 Journal of Geophysical Research: Oceans 10.1029/2020JC016248

Figure 4.  Solutions for Equations 8 and 9 compared with Hsueh's solution. (a) The solution of  against α in Equation 8. (b) The solution of  against the depth
ratio H0/H in Equation 8. (c) The solution of  against α in Equation 9. (d) The solution of  against the depth ratio H0/H in Equation 9.

4.1.  Data and Numerical Model

The temporal variation in the incidence angle is derived from the surface geostrophic velocities along the
upstream path of the Kuroshio using merged ADT data with spatial and temporal resolutions of 0.25 × 0.25°
and 1 day, respectively (Wu et al., 2017). The daily ADT data are converted to monthly mean values with 276
timestamps. The temperature data from WOA18 are utilized to analyze the temporal variation in the up-
per-layer thickness. The WOA18 data have a 0.25 × 0.25° horizontal resolution and a 57-level vertical resolu-
tion (Locarnini et al., 2018). Using the temperature data from WOA18, the depth of the 18°C isotherm (D18)
at approximately 24°N, 123°E is calculated to represent the upper-layer thickness (to be discussed later).

The Regional Ocean Model System (Shchepetkin & McWilliams, 2005) (ROMS) is a free-surface, terrain-fol-
lowing, primitive equation ocean model widely used by the scientific community for research on circulation
in estuary areas and open oceans (Dinniman et al., 2003; Marchesiello et al., 2003; Peliz et al., 2003). The
application of vertical S-coordinates makes it possible to maintain high resolution in the surface Ekman
layer and bottom Ekman layer (Song & Haidvogel, 1994). Details of the ROMS computational algorithms
are described in Shchepetkin and McWilliams (2005).

A state-of-the-art ocean general circulation model based on the ROMS has been developed for the China
Sea. The model has a fine horizontal resolution of 3  3 and a vertical resolution of 31 S-coordinate layers
and is well established to examine the flow pattern of the Kuroshio current and KI into the shelf of the ECS
(Yang et al., 2011, 2012, 2018). Extending from 105°E to 136°E and from 15°N to 41°N (see Figure 5), the
model domain covers most of the ECS and part of the South China Sea (SCS). The model is forced by clima-
tological data, and the model output from the 14th model year was used to examine the theory above. The
KI angle is calculated directly from the velocity field of the ROMS model and compared with the analytical

CUI ET AL. 8 of 19
 Journal of Geophysical Research: Oceans 10.1029/2020JC016248

solution using the AVISO data and WOA18. Please refer to Yang, Yin,
et al. (2018) for details on the model configuration.

Figure 5a shows the modeled annual mean currents at the 10-m depth.
Yang, Yin, et al. (2018) demonstrated that this model can reproduce the
seasonal patterns of ocean circulation over the ECS shelf, such as the
Kuroshio current, the Min-Zhe coastal current off the coast of Zhejiang,
and the Taiwan warm current. Figure 5b shows the trajectory of the La-
grangian drifters released at the Kuroshio mainstream around 24°N in
the ROMS model from 10 to 100 m depth at a 10-m interval in January. It
is a typical KBC pattern after meeting with the zonal shelf break. Moreo-
ver, as shown in Figure 6, the seasonal variation in the KI pattern is quan-
titatively similar to both the observations (e.g., Figures 2 and 8 in Tang
et al., 2000) and the numerical model results (e.g., X. Y. Guo et al., 2003,
2006). Therefore, the results from this numerical model together with
data from AVISO and WOA18 are used to examine the new formula and
explore the seasonal variation in the KI direction.

4.2.  Validation of the New Theory

Figure 6 depicts the monthly mean ocean currents in January, April, July,
and October at 10-m depth in the 14th model year. The Kuroshio Current
has the strongest intrusion in January. Part of the current turns north-
west after colliding with the zonal-running shelf break and intrudes onto
the ECS shelf. Subsequently, the intruded water turns southeast and then
east, forming the KBC (Qiu & Imasato, 1990). The intrusion angle from
the ROMS results ( ROMS) is calculated by averaging the barotropic veloci-
ty over the green box in Figure 6 (similar to Wu et al., 2017) and is shown
in Figure  7e (the purple line with triangles). The seasonal variation in
the modeled KI direction is significant with a larger θ in winter (from
December to February) and a relatively small θ in spring and autumn
(from March to May, September to October). The annual mean value is
Figure 5.  (a) ROMS model domain, bathymetry, and annual mean ocean 83.7° and the standard deviation (STD) is 4.3°. The monthly mean results
current at 10-m depth; the filled color denotes the model bathymetry, show that December has the largest θ (89.3°) while the smallest θ occurs
and the blue and purple dashed lines denote the 150 and 500 m isobaths, in April (76.3°); the change range reaches 13.0°. The ADT-derived veloc-
respectively. The black arrows are drawn every five grid points, and the red ity over the shelf break is not utilized because of the limited resolution
arrows represent velocities larger than 0.3 m·s-1 at every eight grid points.
and the inconsistency between the surface geostrophic currents and the
(b) The trajectory of the Lagrangian drifters released in January in the
ROMS model. ROMS, Regional Ocean Model System. observed flow field from the ADCP (Jan et al., 2011; Tang et al., 2000; Wu
et al., 2014, 2017). Instead, the numerical models behave better in terms
of reproducing the flow pattern in this region (Guo et  al,  2003, 2006;
Yang, Yin, et al., 2018).

The depth ratio H0/H is determined as follows. According to the ETOPO-1 data, H0, the mean water depth
over the step in the study region, is set as 140 m. We define the upper-layer thickness H as the depth of the
largest vertical temperature gradient in the region far offshore of Taiwan (the yellow star in Figure 7a). As
shown in Figure 7b, the annual mean H is taken as 250 m, which is nearly D18, and the annual mean of
H0/H is 0.56. Similar results are obtained at other latitudes (figures not shown). As shown in Figure 7d, the
climatological monthly mean H shows a weak temporal variation with a 6.8-m STD and remains almost
constant. For example, H only shows a 21-m change at maximum in climatology that cannot be resolved
by WOA18 standard levels (the vertical resolution is 25 m at this depth). Thus, it is reasonable that H0/H is
set to 0.56.

For the incidence angle, the mean current velocity is first computed from the monthly mean AVISO data
(from 1993 to 2015, 276 months in total, see Figure 7a). Then the observed incidence angle (αobs) is defined as
the direction of the ADT-derived Kuroshio axis (the track of the largest current speed at each zonal section

CUI ET AL. 9 of 19
 Journal of Geophysical Research: Oceans 10.1029/2020JC016248

Figure 6.  Monthly mean ocean currents at 10 m in January (a), April (b), July (c), and October (d). The black arrows
are drawn every five grid points, and the red arrows are drawn for velocities larger than 0.3 m·s−1 at every seventh grid
points. The blue and purple solid lines denote the 150 and 500 m isobaths, respectively. The green box and the green
line denote the domain and transect that are utilized to compute the intrusion angle and the upper-layer thickness.

along the upstream Kuroshio, shown in Figure 7d). Consistent with previous studies (e.g., Tang et al., 2000;
Wang & Oey, 2016), the Kuroshio generally shifts shoreward in autumn and wintertime, resulting in a large
incidence angle (64.9° at maximum in December). In spring and summertime, it shifts offshore with a small
incidence angle (58.6° at a minimum in October). The annual mean incidence angle is 60.3° with a 4.1° STD
(from the 23-year monthly data shown in Figure 7d, the blue line with circles, with ±2 STDs in each month).

Then, the predicted intrusion angle ( pred ) is calculated according to Equation 9 from αobs with a constant
H0/H (0.56). Interestingly, the solution would collapse if the incidence angle is below 53.7°. However, it
is shown in Figure  7d that most αobs values in the range of ±2 STDs are above this threshold and have
the corresponding  pred . Notably, the correlation coefficients between  pred and  ROMS reaches 0.70 at the
0.05 significance level, and the explained variance is 74.4%, meaning that on the seasonal timescale, over
two-thirds of the total variance in  ROMS could be explained by the variation in αobs using the new solution.
Therefore, the physical framework described in the new model could be a dominant process in regulating
the seasonal variation in the intrusion direction. It is also interesting to note that when the incidence angle
is small,  (the squared ratio of Vs to the incoming Kuroshio current velocity, see Figure 4a) is almost zero.

CUI ET AL. 10 of 19
 Journal of Geophysical Research: Oceans 10.1029/2020JC016248

Figure 7.  (a) The multiyear mean surface geostrophic currents (ADT-derived) northeast of Taiwan and the path of the upstream Kuroshio. The red line with
triangles denotes the Kuroshio mainstream along which the incidence angle is computed. (b) The annual mean vertical temperature gradient at approximately
24°N east of Taiwan from WOA18. (c) Same as Figure 7b, but for the temperature distribution. The dashed line denotes the 18°C isotherm. (d) The monthly
climatological mean of D18 at approximately 24°N, 123°E (the yellow star in Figure 7a, in red) and the incidence angle α from AVISO (with ±2 STDs, in blue)
computed along the mainstream in Figure 7a. The blue dashed line is the critical line below which the solution for  would collapse when H0/H = 0.56. (e) The
predicted intrusion angle  from AVISO according to Equation 9 (in yellow) and the modeled  from the ROMS results by averaging the barotropic velocity in
the green box in Figure 6. ADT, absolute dynamic topography; AVISO, Archiving, Validation and Interpretation of Satellite Oceanographic.

Such a result could also explain why the recirculation is not obvious in weak intrusion situations, such as
those in the summertime.

5. Discussion

In this section, we first present the physical interpretation of the supplementary current Vs. This current
plays the same role as the combined effect of the JEBAR term, the bottom stress term, and the horizontal
advection term (the second, fourth, and fifth terms on the RHS) in the vorticity equation below. This equa-
tion is obtained by vertically averaging the primitive momentum equations and cross differentiating them
(e.g., X. Guo et al., 2006).

 τ   τ 

(11)
t
 M
 H
 f 
 M     J , H
curlz   
H
1
 
 0 H   0 H 
 A
H
D
 curlz  a   curlz  b   curlz    curlz  
H

 
In Equation  11, J  , H 1   / x  H 1 / y   / y  H 1 / x is the JEBAR term, where
0 0
   gz /  0 dz is the potential energy, curlz is the curl in the z-direction, M     ui  v j  dz is a volume
H H
 f 
transport vector, M     is the advection of the potential vorticity term (APV), τ a is the wind stress, τ b is
H
the bottom friction, and A and D are the vertically integrated nonlinear advection term and the horizontal
viscosity term from the primitive momentum equations, respectively. In a quasi-steady state, the time-deriv-
ative term is neglected, thus, Equation 11 is a diagnostic tool for analyzing the motion in this area.

Figure 8 shows the annual mean distributions and the temporal variations in the terms in Equation 11 from
the monthly mean ROMS output. The area northeast of Taiwan is divided into two regions. The first region

CUI ET AL. 11 of 19
 Journal of Geophysical Research: Oceans 10.1029/2020JC016248

Figure 8.  Annual mean distributions and the temporal variations in the terms in Equation 11. (a) The spatial distributions of the APV term, the JEBAR
term, the wind stress term (WIND), the bottom friction term (BOT), the horizontal viscosity term (HVISC), and the advection term (ADV). (b) The temporal
variations in these terms averaged over the intrusion region (the solid rectangular box in Figure 8a. (c) Same as Figure 8b, but for the KBC recirculation region
(the dashed rectangular box in Figure 8a). KBC, Kuroshio Branch Current; APV, advection of the potential vorticity; JEBAR, joint effect of baroclinicity and
relief.

(the solid rectangular box in Figure 8a) mostly includes the main branch of the Kuroshio Current and the
KI area, while the second region (the dashed rectangular box in Figure 8a) is mainly the KBC recircula-
tion area. As shown in Figure 8b, in the Kuroshio mainstream area and the KI area, the motion is mainly
a balance between the APV term and the JEBAR term. In the f-plane, when the APV term is negative, it
represents the motion that is on-slope, which is northward or northwestward in this area. The wind stress,
bottom friction, or any other terms do not contribute significantly to the balance. However, in the KBC recir-
culation region, the APV term is balanced by the JEBAR term, the bottom friction term, and the advection
term (Figure 8c). The APV term is positive, representing an off-slope motion in the f-plane. The direction of
this off-slope motion is eastward or southeastward, which is exactly the direction of the KBC recirculation.
The physical interpretation of the anticyclonic pattern of the KBC is similar to that of Oey et al. (2018) in
terms of the different signs of the APV term in these two regions.

 f 
 M      J  , H 1
(12a)
H
 
 τ 
(12b)
 f 
H
   H
 0 
 A
    J  , H 1  curlz  b   curlz  
M
H
 
Among all the terms to balance the APV terms in Figure 8c, the JEBAR term is the largest, while the bottom
friction and the advection terms contribute less than half of the JEBAR term. However, the bottom friction

and advection terms cannot be neglected because the water is shallow in this region. Therefore, Vs reflects

CUI ET AL. 12 of 19
 Journal of Geophysical Research: Oceans 10.1029/2020JC016248

the part of motion balanced by these terms (there is no JEBAR term in this model because of the flat topog-
raphy H0 over the step). The bottom friction and nonlinear advection are neglected in this 1.5-layer model,
and the added flow also reflects the influence of these processes to a certain extent (in a quasi-steady state
such as monthly mean or annual mean state).

This formula may potentially be applicable in other similar regions such as the Charleston Bump in the
Gulf Stream area (Gula et al., 2015). Therefore, much work on its application is needed in future research.
Moreover, although this formula is obtained where the depth ratio < 1, it applies to regions where the depth
ratio > 1. If the incidence current runs down into a trench or valley, intrusion might still occur, although the
flow pattern is altered (see Figure 4). This scenario is beyond the scope of this study and may be a subject
of future research.

6. Conclusion
A theoretical 1.5-layer model is modified and improved based on observations and numerical model results.
The new model explains the seasonal variation in the intrusion direction northeast of Taiwan and the mech-
anism for the formation of the KBC, especially the KBC recirculation. That is, an off-slope-direction current
balanced by the JEBAR term, the bottom friction term, and the nonlinear advection term in the vorticity
equation. This current is necessary to conserve the momentum balance in this domain. In the long-term
mean state, the flow field northeast of Taiwan could reach a steady state only if the KBC recirculation is
considered. This theory predicts that the maximum speed of the flow over the zonal-running shelf break is
0.42 m s−1, which is comparable to the numerical model results and observations in previous studies. Based
on the ROMS model results and the AVISO data, it is shown that the KI direction is highly correlated with
the incidence angle on the seasonal timescale, which provides evidence for the new formula from the mod-
el. The flow pattern from the ROMS output is well represented by the modified model in terms of the KI and
its recirculation. Through a diagnostic analysis, this modified model behaves well in terms of simulating
the KI pattern northeast of Taiwan because it could reflect the influence of the topography and baroclinic
processes, bottom friction, and nonlinear advection on the mean state.

The water and material exchange between the Kuroshio Current and marginal seas significantly influences
the coastal hydrodynamic environment and ecosystem. Northeast of Taiwan, the KBC and its recirculation
play an important role in the material exchange between the ECS shelf water and the Kuroshio Current. The
dynamic process of the KI northeast of Taiwan could be much more complicated than the quasi-steady state
investigated here. However, this model suggests some key dynamic factors by which the KI and its recircu-
lation could be estimated using this new formula. Moreover, when the Kuroshio mainstream varies to cor-
respond to different scenarios such as global warming, typhoons, and mesoscale eddies, this simple formula
could predict an intrusion angle and estimate volume or material flux instead of running a time-consuming
numerical model.

Appendix A:  Description and Calculation of Hsueh's Model

For 0  n1  ,

n1 x  sin 

  n 
 1
h1  x   H 1  e R 
  
  

(A1) n
 1
u1  x   g h1 cos    Rf   e R  cos 
 f n1
 n
 g h1  1
v1  x   sin    Rf   e R  sin 
 f n1

CUI ET AL. 13 of 19
 Journal of Geophysical Research: Oceans 10.1029/2020JC016248

Area 2:
 H 
For 0  n 2   R ln 1  0 ,
 H 

n2 x  sin 

  n 
 2
h2  x   H 1  e R 
  
  
(A2)
 n
 2
u2  x   g h2 cos    Rf   e R  cos 
 f n2
 n
 g h2  2
v2  x   sin    Rf   e R  sin 
 f n2

Area 3:
For 0  n 3  ,

n3   y

  n 
 3
 n 
 H   3
h3  y   H0   H  H0  1  e R   H 1  1  0  e R 
     H  
(A3)
    
 n
g h3  H0   R3
u3  y f n   Rf   1  H  e
 3  
v y  0
 3  

Area 4:
 H   H 
For  R ln 1  0   n 4  R  R ln 1  0 ,
 H   H 

 n4 x  sin 

h4  y   H0
  
Vn   Rf   1  H0   1  ln 1  H0   n4 
 4 H  H R  
    
(A4)

u y  V  cos   Rf  1  H0  1  ln 1  H0   n4   cos 
 4  n4    H  
 


 
H  R 

  H0    H0  n4 
v4  y   Vn4  sin    Rf   1   1  ln 1     sin 
  H    H  R 

Area 5:
For 0  n 5  R,

n5  y

h5  y   H0
  H  n 
Vn5  Rf    1  0  1  5 
(A5)
  H  R 
  H  n 
u5  y  Vn5    Rf    1  0  1  5 
  H  R 
v y  0
 5  

CUI ET AL. 14 of 19
 Journal of Geophysical Research: Oceans 10.1029/2020JC016248

where f is the Coriolis parameter; g' = g(ρ2−ρ1)/ρ1 denotes the reduced gravity for a two-layered ocean with
upper and lower layer densities given by ρ1 and ρ2, respectively; H is the undisturbed upper-layer thickness;
R  gH / f is the baroclinic deformation radius; H0 is the step depth (see Figure 4 in Hsueh et al., 1992);
α is the incidence angle;  is the intrusion angle; n is the normal vector of the front; V is the magnitude of
the current velocity; u is the magnitude of the zonal velocity; v is the magnitude of the meridional velocity;
and the subscripts 1 to 5 represent n, V, u and v in Areas 1–5, respectively (Figure 2b).

The pressure field at different points can be calculated as PA = PF = 0, PM  = g'H0, PD = PE = 1/2g'(H + H0), 
PB = g'H, PC = g'H0, where P = p/ρ1 and the subscripts of P represent different points (Figure 2b). Based on
the conservation of volume transport, the volume transport across different sections can be calculated as
follows:

Tin  T  T1  T2  T3  T4  Tout ,
(A6)
2 2
where T  gH / 2 f denotes the volume transport across section AB', T1  gH0 / 2 f denotes the vol-
ume transport across section FM, T2  gH 0  H  H0  / 2 f the volume transport across section ME,
 
T3  gH0  H 0  H  / 2 f is the volume transport across section C'D, and T4  g H 2  H0 2 / 2 f is the vol-
ume transport across section B'C'. Notably, the positive direction of T is taken as the normal direction of sec-
tion AB', which points inside domain AB'C'DEMF, while the positive directions of T1, T2, T3 and T4 represent
the direction out of domain AB'C'DEMF.

After a fair amount of algebra, the terms in Equation 3 can be calculated as follows:

B
  u1v1h1dx  S  cos 
A
2
C  H   2 H0 
  u2u2 h2 dy  S  1  0  1  
B  H   H 
D 2
  u u h dy  S  1  H0  2 H0
(A7)
C 5 5 5
H  H
 
2
E  H0  2 H0
  u4 v4 h4 dx  S  1   cos 
M  H  H
M  2 
  u2 v2 h2 dx  S  1  H0  1  2 H0   1 cos 
F  H   H  
  

B
  v1v1h1dx  S  sin 
A
C
  v2u2 h2 dy  0
 BD
  v u h dy  0
5 5 5
(A8)
C
E 2
  v4 v4 h4 dx  S  1  H0  2 H0 sin 
M  H  H

 M  
2
H0   2 H0  
 v
 2 2 2v h dx  S    1   1    1 sin 
 F  H   H  
 

where S  RH  Rf  / 6 is a constant number related to the baroclinic deformation radius, the undisturbed
2

upper-layer thickness H, and the Coriolis parameter f.

CUI ET AL. 15 of 19
 Journal of Geophysical Research: Oceans 10.1029/2020JC016248

Appendix B:  Detailed Analysis of Hsueh's Model Results

In the main text, we concluded that Hsueh’s model is not in a steady state: there must be external forces
applied to this system, or this system is not self-sustainable. In other words, part of the momentum is miss-
ing from the model. The analysis above indicates three possible situations. In the first case, y-momentum
is conserved, while x-momentum is not self-sustained (Fy equals 0, while Fx does not). In the second case,
x-momentum is conserved, while y-momentum is not self-sustained (Fx equals 0, while Fy does not). In the
last case, both x-momentum and y-momentum are not self-sustained (neither Fx nor Fy equals zero).

In the first situation, the LHS of Equation 4a yields the following expression:


 cos       sin    sin  ,  90
LHS / S   ,
  
(A9)
 cos       sin    sin  ,  90

  
2 2
 H0   4 H0   H0 
 , and S  RH  Rf  / 6 . The LHS of Equation 4b in the
2
where   1   1  ,  1  1 
 H   H   H 
second situation yields the following expression:

 sin     cos      cos    .

LHS / S 
(A10)   
For the Kuroshio east of Taiwan, the incidence angle α is approximately 70°, and the depth ratio H0/H is ap-
proximately 0.5, taking H0 to be 150 m and H to be 300 m (same as Hsueh et al. based on observations result-
ing from a hydrographic survey conducted during 16–21 April 1989). However, when H0/H is 0.5, there is no
solution corresponding to α = 70° (see Figure A1a and Figure A1b, the yellow lines). Such a result implies
that the y-momentum is not conserved by all means in this system. The solution shown in Figure A1c taking
α to be 70° and H0/H to be 0.5 seems reasonable (the yellow line); however, there are inherent problems in
this model, as described below.

The first problem is that there is a discontinuity at the step (see Figure 2b, between Area 3 and Area 5) that,
from the viewpoint of Hsueh, is an artifact of inviscid formulation (the conservation of PV). More importantly,
the current over the step in Area 5 flows westward, while the current in Area 3 flows eastward (see Figure 2b,
the red arrows and blue arrows). At this step, the velocity magnitude from Area 3 is  Rf 1  H0 / H , which
is identical to that from Area 5, but its direction is opposite to that in Area 5. This finding is inconsistent with
the current knowledge of this area since the flow over the step is mainly eastward (e.g. Yang, Yin, et al., 2018);
therefore, the flow pattern in Area 5 is not accurate enough. The second problem is that this model fails to ex-
plain the KBC recirculation, which flows southeastward as part of the KBC (Qiu & Imasato, 1990). The third
problem is the most serious one: this system is not in a steady state. Thus, this model needs to be modified.

Appendix C:  Derivation and Calculation of the Modified Model


The velocity of the supplementary current Vs in Area 4 and Area 5 can be expressed as follows:
u4 s Vs  cos 

v4 s Vs  sin 
(A11)

u5s  Vs
v5s  0


and the momentum flux Vs generated in the x and y directions can be expressed as follows:

D E

C M
2
 
  u5su5s H0 dy   u4s v4s H0 dx  H0Vs R 1  cos   S    1  cos   
(A12)
E
  v v H dx  H V 2 R sin   S    sin 
 M 4 s 4 s 0 0 s

CUI ET AL. 16 of 19
 Journal of Geophysical Research: Oceans 10.1029/2020JC016248

Figure A1.  The solutions for Equation A9 and (A10). (a) The solution for Equation A9 against α when   90. (b) Same as (a) but for   90. (c) The solution
for Equation A10 against α.

6 H0 Vs 2  
where   , Vs  Vs is the magnitude of Vs , and the subscripts u and v denote the supplementary
H  Rf  2

velocities in Areas 4 and 5, respectively. Assuming that the momentum in the modified model is conserved,

we have the following expression:

 B C D E
   u1v1h1dx   u2u2 h2 dy   u5u5h5dy   u4 v4 h4 dx 
 A B C M
M D E

  u2 v2 h2 dx   u5su5s H0 dy   u4s v4s H0 dx 0
F C M
(A13)
 B C D E
   v v h dx   v u h dy   v u h dy   v v h dx 
 A 1 1 1 2 2 2 5 5 5 4 4 4
B C M
 M E
  v2 v2 h2 dx   v4 s v4 s H0 dx 0
 F M

Substituting the terms in Equation A13 with Equation A7, (A8) and (A12), we obtain the expression of the
cosine and sine of the intrusion angle  :

Acknowledgments
 cos     
cos  
This study was supported by the   
National Natural Science Foundation (A14)
 .
sin   sin 
of China (NSFC) (No. 41876019), the
Strategic Priority Research Program of   

the Chinese Academy of Sciences (No.
XDB42000000 and XDA19060203),
the National Key Research Devel-
Thus, the equation set (A14) is closed, and we obtain Equations 8 and 9 as the solution.
opment Program of China (No.
2017YFC1404000, 2016YFC1401601 and
2017YFA0604102), the NSFC-Shandong
Joint Fund for Marine Science Research Data Availability Statement
Centers (No. U1406401), and the project
“A Forecasting System to Provide Early
The ETOPO-1 data are accessible from the National Geophysical Data Center (https://maps.ngdc.noaa.gov/
Warnings for Coastal Ecological Disas- viewers/wcs-client/). The satellite data from AVISO can be downloaded from the website (https://www.
ters (harmful algal blooms and harmful aviso.altimetry.fr). The WOA18 data are freely available (https://www.nodc.noaa.gov/OC5/woa18).
jellyfish blooms” jointly funded by the
Chinese Academy of Sciences and the
Commonwealth Scientific and Industri-
al Research Organization (CSIRO). This References
study was also supported by the High
Performance Computing Center at the Amante, C., & Eakins, B. W. (2009). ETOPO1 1 arc-minute global relief model: Procedures, data sources and analysis. NOAA Technical
Institute of Oceanology Chinese Acade- Memorandum NESDIS NGDC-24. NOAA: National Geophysical Data Center.
my of Sciences (IOCAS) and the Youth Andres, M., Jan, S., Sanford, T., Mensah, V., Centurioni, L., & Book, J. (2015). Mean structure and variability of the Kuroshio from north-
Innovation Promotion Association CAS. eastern Taiwan to southwestern Japan. Oceanography, 28(4), 84–95. https://doi.org/10.5670/oceanog.2015.84

CUI ET AL. 17 of 19
 Journal of Geophysical Research: Oceans 10.1029/2020JC016248

Andres, M., Mensah, V., Jan, S., Chang, M. H., Yang, Y. J., Lee, C. M., et al. (2017). Downstream evolution of the Kuroshio's time-varying
transport and velocity structure. Journal of Geophysical Research: Oceans, 122, 3519–3542. https://doi.org/10.1002/2016jc012519
Chao, S. Y. (1990). Circulation of the East China Sea, a numerical study. Journal of the Oceanographical Society of Japan, 46(6), 273–295.
Chern, C. S., & Wang, J. (2005). Interactions of Mesoscale Eddy and Western Boundary Current: A Reduced-Gravity Numerical Model
Study. Journal of Oceanography, 61(2), 271–282. https://doi.org/10.1007/s10872-005-0037-z
Chuang, W. S., & Liang, W. D. (1994). Seasonal variability of intrusion of the Kuroshio water across the continental shelf northeast of
Taiwan. Journal of Oceanography, 50(5), 531–542.
Dinniman, M. S., Klinck, J. M., & Smith, W. O. (2003). Cross-shelf exchange in a model of the Ross Sea circulation and biogeochemistry.
Deep Sea Research Part II: Topical Studies in Oceanography, 50(22–26), 3103–3120. https://doi.org/10.1016/j.dsr2.2003.07.011
Doong, D. J., Peng, J. P., & Babanin, A. V. (2019). Field investigations of coastal sea surface temperature drop after typhoon passages. Earth
System Science Data, 11(1), 323–340. https://doi.org/10.5194/essd-11-323-2019
Gula, J., Molemaker, M. J., & Mcwilliams, J. C. (2015). Gulf stream dynamics along the southeastern U.S. seaboard. Journal of Physical
Oceanography, 45(3), 690–715.
Guo, X. Y., Hukuda, H., Miyazawa, Y., & Yamagata, T. (2003). A triply nested ocean model for simulating the Kuroshio - Roles of hori-
zontal resolution on JEBAR. Journal of Physical Oceanography, 33(1), 146–169. https://doi.org/10.1175/1520-0485(2003)033<0146
:Atnomf>2.0.Co;2
Guo, X., Miyazawa, Y., & Yamagata, T. (2006). The Kuroshio onshore intrusion along the shelf break of the East China Sea: The origin of
the Tsushima Warm Current. Journal of Physical Oceanography, 36(12), 2205–2231. https://doi.org/10.1175/jpo2976.1
Hsueh, Y., Chern, C. S., & Wang, J. (1993). Blocking of the Kuroshio by the continental shelf northeast of Taiwan. Journal of Geophysical
Research, 98(C7). https://doi.org/10.1029/93jc01075
Hsueh, Y., Schultz, J. R., & Holland, W. R. (1997). The Kuroshio flow-through in the East China Sea: A numerical model. Progress in Ocean-
ography, 39(2), 79–108. https://doi.org/10.1016/s0079-6611(97)00008-6
Hsueh, Y., Wang, J., & Chern, C. S. (1992). The intrusion of the Kuroshio across the continental shelf northeast of Taiwan. Journal of Geo-
physical Research, 97(C9). https://doi.org/10.1029/92jc01401
Ichikawa, H., & Beardsley, R. C. (1993). Temporal and spatial variability of volume transport of the Kuroshio in the East China Sea. Deep-
Sea Research, Part I (Oceanographic Research Papers), 40(3), 583–605.
Isobe, A. (2008). Recent advances in ocean-circulation research on the yellow sea and east china sea shelves. Journal of Oceanography,
64(4), 569–584.
Jan, S., Chen, C. C., Tsai, Y. L., Yang, Y. J., Wang, J., Chern, C. S., et al. (2011). Mean structure and variability of the cold dome northeast
of Taiwan. Oceanography, 24(4), 100–109.
Jan, S., Yang, Y. J., Wang, J., Mensah, V., Kuo, T. H., Chiou, M. D., et al. (2015). Large variability of the Kuroshio at 23.75°N east of Taiwan.
Journal of Geophysical Research: Oceans, 120, 1825–1840.
Johns, W. E., Lee, T. N., Zhang, D., Zantopp, R., Liu, C. T., & Yang, Y. (2001). The Kuroshio east of Taiwan: Moored transport observations
from the WOCE PCM-1 array. Journal of Physical Oceanography, 31, 1031–1053.
Lentz, S., Guza, R. T., Elgar, S., Feddersen, F., & Herbers, T. H. C. (1999). Momentum balances on the North Carolina inner shelf. Journal
of Geophysical Research: Oceans, 104(C8), 18205–18226. https://doi.org/10.1029/1999jc900101
Liu, K. K., Gong, G. C., Shyu, C. Z., Pai, S. C., Wei, C. L., & Chao, S. Y. (1992). Response of Kuroshio upwelling to the onset of the northeast
monsoon in the sea north of Taiwan: Observations and a numerical simulation. Journal of Geophysical Research, 97(C8), 12511–12526.
https://doi.org/10.1029/92jc01179
Liu, C., Wang, F., Chen, X., & von Storch, J.-S. (2014). Interannual variability of the Kuroshio onshore intrusion along the East Chi-
na Sea shelf break: Effect of the Kuroshio volume transport. Journal of Geophysical Research: Oceans, 119, 6190–6209. https://doi.
org/10.1002/2013jc009653
Locarnini, R. A., Mishonov, A. V., Baranova, O. K., Boyer, T. P., Zweng, M. M., Garcia, H. E., et al. (2018). World Ocean Atlas 2018. A. Mis-
honov. In Temperature (1). NOAA Atlas NESDIS in preparation.
Marchesiello, P., McWilliams, J. C., & Shchepetkin, A. (2003). Equilibrium structure and dynamics of the California Current System. Jour-
nal of Physical Oceanography, 33(4), 753–783. https://doi.org/10.1175/1520-0485(2003)33<753:Esadot>2.0.Co;2
Mertz, G., & Wright, D. G. (1992). Interpretations of the JEBAR term. Journal of Physical Oceanography, 22(3), 301–305. https://doi.
org/10.1175/1520-0485(1992)022<0301:Iotjt>2.0.Co;2
Oey, L. Y., Hsin, Y. C., & Wu, C. R. (2010). Why does the Kuroshio northeast of Taiwan shift shelfward in winter? Ocean Dynamics, 60(2),
413–426. https://doi.org/10.1007/s10236-009-0259-5
Oey, L. Y., Wang, J., & Lee, M. A. (2018). Fish catch is related to the fluctuations of a western boundary current. Journal of Physical Ocean-
ography, 48(3), 705–721. https://doi.org/10.1175/jpo-d-17-0041.1
Peliz, A., Dubert, J. S., & Haidvogel, D. B. (2003). Subinertial response of a density-driven eastern boundary poleward current to wind
forcing. Journal of Physical Oceanography, 33(8), 1633–1650. https://doi.org/10.1175/2415.1
Qiu, B., & Imasato, N. (1990). A numerical study on the formation of the Kuroshio Counter Current and the Kuroshio Branch Current in
the East China Sea. Continental Shelf Research, 10(2), 165–184.
Shchepetkin, A. F., & McWilliams, J. C. (2005). The regional oceanic modeling system (ROMS): A split-explicit, free-surface, topogra-
phy-following-coordinate oceanic model. Ocean Modelling, 9(4), 347–404. https://doi.org/10.1016/j.ocemod.2004.08.002
Song, Y., & Haidvogel, D. (1994). A semi-implicit ocean circulation model using a generalized topography-following coordinate system.
Journal of Computational Physics, 115(1), 228–244.
Tang, T. Y., Tai, J. H., & Yang, Y. J. (2000). The flow pattern north of Taiwan and the migration of the Kuroshio. Continental Shelf Research,
20(4), 349–371.
Tang, T. Y., & Yang, Y. J. (1993). Low frequency current variability on the shelf break northeast of Taiwan. Journal of Oceanography, 49(2),
193–210.
Velez-Belchi, P., Centurioni, L. R., Lee, D. K., Jan, S., & Niiler, P. P. (2013). Eddy induced Kuroshio intrusions onto the continental shelf of
the East China Sea. Journal of Marine Research, 71(1–2), 83–107. https://doi.org/10.1357/002224013807343470
Wang, J., & Oey, L. Y. (2014). Inter-annual and decadal fluctuations of the Kuroshio in east china sea and connection with surface fluxes
of momentum and heat. Geophysical Research Letters, 41, 8538–8546.
Wang, J., & Oey, L. Y. (2016). Seasonal exchanges of the Kuroshio and shelf waters and their impacts on the shelf currents of the east China
sea. Journal of Physical Oceanography, 46(5), 160229122404004.
Wu, C. R., Hsin, Y. C., Chiang, T. L., Lin, Y. F., & Tsui, I. F. (2014). Seasonal and interannual changes of the Kuroshio intrusion onto the
East China Sea Shelf. Journal of Geophysical Research: Oceans, 119, 5039–5051. https://doi.org/10.1002/2013jc009748

CUI ET AL. 18 of 19
 Journal of Geophysical Research: Oceans 10.1029/2020JC016248

Wu, C. R., Wang, Y. L., Lin, Y. F., & Chao, S. Y. (2017). Intrusion of the Kuroshio into the South and East China Seas. Scientific Reports, 7(1),
7895. https://doi.org/10.1038/s41598-017-08206-4
Yang, D., Huang, R., Yin, B., Feng, X., Chen, H., Qi, J., et al. (2018). Topographic beta spiral and onshore intrusion of the Kuroshio current.
Geophysical Research Letters, 45(1), 287–296. https://doi.org/10.1002/2017gl076614
Yang, D., Yin, B., Chai, F., Feng, X., Xue, H., Gao, G., & Yu, F. (2018). The onshore intrusion of Kuroshio subsurface water from February to
July and a mechanism for the intrusion variation. Progress in Oceanography, 167, 97–115. https://doi.org/10.1016/j.pocean.2018.08.004
Yang, D., Yin, B., Liu, Z., Bai, T., Qi, J., & Chen, H. (2012). Numerical study on the pattern and origins of Kuroshio branches in the bottom
water of southern East China Sea in summer. Journal of Geophysical Research, 117(C2). https://doi.org/10.1029/2011jc007528
Yang, D., Yin, B., Liu, Z., & Feng, X. (2011). Numerical study of the ocean circulation on the East China Sea shelf and a Kuroshio bottom
branch northeast of Taiwan in summer. Journal of Geophysical Research, 116(C5). https://doi.org/10.1029/2010jc006777
Zhang, D., Lee, T. N., Johns, W. E., Liu, C. T., & Zantopp, R. (2001). The Kuroshio east of Taiwan: Modes of variability and relationship to
interior ocean mesoscale eddies. Journal of Physical Oceanography, 31, 1054–1074.
Zhang, Y., Zhang, Z., Chen, D., Qiu, B., & Wang, W. (2020). Strengthening of the Kuroshio current by intensifying tropical cyclones. Science,
368, 988–993.

CUI ET AL. 19 of 19