Regional Studies in Marine Science 18 (2018) 35–43

Contents lists available at ScienceDirect

Regional Studies in Marine Science

journal homepage: www.elsevier.com/locate/rsma

Analysis of seasonal characteristics of water exchange in Beibu Gulf

based on a particle tracking model
Lina Wang a,b , Weiran Pan a, *, Wei Zhuang a , Xiao-Hai Yan a,b, *, Victor V. Klemas b
a
State Key Laboratory of Marine Environmental Science, Xiamen University, Xiamen, 361005, China
b
Center for Remote Sensing, and Joint Institute for Coastal Research and Management, University of Delaware, Newark, DE, 19716, USA

highlights

• The residence time in the whole gulf does not vary much between summer and winter.
• The residence times in six sub-regions exhibit a large seasonal difference.
• Water movement in the gulf follows the circulation patterns.
• Currents from Qiongzhou Strait and the south opening help in refreshing the gulf.

article info a b s t r a c t
Article history: The seasonal characteristics of water exchange in Beibu Gulf are investigated based on a particle tracking
Received 19 May 2017 model. The gulf is divided into six sub-regions in order to better understand the exchange processes and
Received in revised form 29 December 2017 water movement in the gulf. The residence time is computed for each sub-region, and the results show
Accepted 29 December 2017
that the whole gulf has a small seasonal variation, with 66 days in winter and 71 days in summer, while
Available online 12 January 2018
the sub-regions exhibit a large seasonal difference with short residence times in summer. Water exchange
curves indicate water movement in the gulf follows the circulation patterns. In winter, the water particles
Keywords:
Water exchange move cyclonically and accumulate near the western coast of the gulf. The current flows entering the gulf
Beibu Gulf are dominated by the westward flow from the Qiongzhou Strait. The influence area of this flow can extend
Residence time to the Vietnam coast. In summer, water particles from the coastal area tend to move offshore and undergo
Particle tracking model strong mixing in the center of the gulf. The northward current flow from the south opening becomes the
Numerical simulation dominant flow, with a large influence area in the eastern part of the gulf.
Qiongzhou Strait © 2017 Published by Elsevier B.V.

1. Introduction distribution in the gulf, providing a greatly diverse marine life.

For example, the intrusion of high-salinity and high-temperature
Beibu Gulf is a semi-enclosed gulf located in the northwestern water from the SCS contributes to different migration pathways of
South China Sea (SCS), with an average depth of about 40 m and Paragyrops edita in Beibu Gulf (Chen and Qiu, 2005). Consequently,
a maximum depth less than 100 m. The total area of the gulf is the study of water renewal and related exchange processes can
approximately 128 000 km2 . Beibu Gulf is recognized as one of help us better understand the distribution of biological resources
the four largest fishing grounds in China, because of abundant and the formation of fishing grounds, and help provide scientific
marine life and year-round warm water temperatures. The gulf support for proper management of the ecosystem.
water is characterized by three major water masses (Chen, 1986), Lagrangian particle tracking methods are very useful tools to
investigate and model the transport pathways and mixing pro-
the coastal water supplied by nutrient-rich discharges from nu-
cesses (Dias et al., 2001; Perianez and Elliott, 2002; Perianez,
merous rivers along the northern and western coast, the mixed
2004, 2005; Ninto and Garcia, 1996; Wroblicky et al., 1998; Saxton
water formed by the current flow from the Qiongzhou Strait (QS),
and Jacobson, 1997). In particular, Random-Walk Particle Tracking
and the sea water coming from the south wide opening between
(RWPT) models that calculate particle pathways with advection
Hainan Island and Vietnam. Water exchange among these water
and random displacement are well suited for water exchange stud-
masses has a significant impact on the species composition and ies (Suh, 2006; Zheng et al., 2003; Bilgili et al., 2005; Dimou and
Adams, 1993; Signell and Butman, 1992; Liu et al., 2011; Oliveira
and Baptista, 1997). When studying large areas, the basic idea is
* Corresponding authors.
E-mail addresses: [email protected] (W. Pan), [email protected] (X.-H. Yan). to divide the study area into several zones, and estimate water

https://doi.org/10.1016/j.rsma.2017.12.009
2352-4855/© 2017 Published by Elsevier B.V.
36 L. Wang et al. / Regional Studies in Marine Science 18 (2018) 35–43

exchange within and between these zones (Thompson et al., 2002).

Here, we adopted this idea to split Beibu Gulf into six sub-regions
using the fuzzy C-means clustering technique (Bezdek, 2013). They
are coastal areas including Guangxi, Vietnam and Hainan Island,
and the interior gulf areas containing the northern, central, south-
ern gulf. The Random-Walk Particle Tracking (RWPT) model was
developed to simulate the exchange processes between the sub-
regions. The circulation patterns likely to affect water movement
(Gao et al., 2014) were calculated using a 3D hydrodynamic model
for particle tracking simulation. The circulation patterns in Beibu
Gulf have strong seasonality, driven by prevailing wind forcing, and
are affected by the exterior circulation from the SCS in the south as
well as outflow from the QS in the northeast (Wu et al., 2008; Gao
et al., 2013; Shi et al., 2002). The cyclonic circulation in winter has
been demonstrated in most observations (Xu et al., 1980; Yu and
Liu, 1993; Yuan and Deng, 1999). However, the circulation pattern
in summer remains controversial. Recent studies pointed out that
the circulation structure remains cyclonic in the summer despite
the monsoonal forcing that tends to drive it anti-cyclonically (Wu
et al., 2008; Gao et al., 2013; Chen et al., 2015; Bao et al., 2005).
Our model result of circulation patterns is consistent with recent
studies, and is discussed in Section 3.
The aim of this study is to analyze the seasonal characteristics Fig. 1. Topography of Beibu Gulf, with sampling stations for field observations
of water exchange between six sub-regions and water movement (units: m).
pathways in Beibu Gulf. Since the climate around the gulf is sub-
tropical and monsoonal, the southwest (SW) and northeast (NE)
monsoons are the two dominant regimes, with transition peri- forced at the open boundaries. The daily averaged elevation, tem-
ods in April and October (Zeng et al., 1989). We chose summer perature, and salinity are obtained from the global 1/12◦ Hybrid
(June, July, and August) and winter (December 2014, January and Coordinate Ocean Model + Navy Coupled Ocean Data Assimilation
February 2015) as two representative seasons for seasonal char- (HYCOM + NCODA) analysis. Daily averaged wind data is derived
acteristics analysis. The water renewal abilities of the sub-regions from the National Climatic Data Center, which is a product merged
were quantified based on the average residence time approach from multiple satellite scatterometers and model analysis on a
(Takeoka, 1984). Water exchange processes between sub-regions global 0.25◦ grid. The daily averaged heat flux is obtained from
were described using water exchange curves. Moreover, the influ- NCEP products with 1◦ resolution. The SST data set uses the Remote
ence areas of current flows from the QS and the south opening Sensing Systems (REMSS) SST product with 9 km and 1 day reso-
were examined via continuous particle release during winter and lution. The climatological monthly averaged freshwater discharges
summer. of eighteen rivers in Guangdong, Hainan, and Vietnam are collected
The rest of this paper is organized as follows: Section 2 de- from observations and published papers (Van Maren and Hoekstra,
scribes the 3D hydrodynamic model, Random-Walk Particle Track- 2004; Pruszak et al., 2005; Gao et al., 2013). The initial temperature
ing model, and definition of residence time used in this paper. Re- and salinity on December 1, 2013 are also derived from the 1/ 12◦
sults of circulation patterns, water exchange, and water transport HYCOM+NCODA analysis. The model is spun up with zero elevation
through QS and the south opening during winter and summer are and velocity, and the time step is set to 12 s. The model simulation
examined in Section 3. Section 4 presents a discussion and a set of period lasts through January 1, 2014 to December 31, 2015.
conclusions. The hydrodynamic model is validated against existing obser-
vations. Comparison of computed and measured tide amplitudes
2. Model, data, and method and phases at 14 locations is presented in Table 1. For the tidal
constituents M2 , K1 , O1 , the mean absolute errors in amplitude are
1.81, 2.9, and 3.4 cm, respectively, and the mean relative errors are
2.1. The 3D hydrodynamic model
7.1%, 4.9%, and 4.8%, respectively, and the phase mean absolute
errors are 4.6◦ , 5.3◦ , and 7.2◦ , respectively. If the mean relative
The MEC (Marine Environmental Committee) ocean model was
errors in amplitude are lower than 10% and the mean absolute
developed by the Japanese Society of Naval Architects and Ocean
errors are lower than 10◦ , the agreement between the model re-
Engineers (Marine Environment Committee, 2003). It has been
sults and observations are considered very good. The observed tidal
applied to solve problems involving the physical environment and elevation and currents measured at Tieshan Bay tide stations (15#
ecosystem in coastal seas and bays (Sato et al., 2006; Mizumukai et and 16# in Fig. 1) and Sanya Bay station (17# and 18# in Fig. 1) are
al., 2008; Kano et al., 2010). In this research, the MEC ocean model used for model calibration (Fig. 3). The root mean squared errors
was used to simulate the velocity fields to drive the RWPT model, (RMSE) for current velocity, current direction and water level in
as described below. Tieshan Bay station are 13.58 cm/s, 13.93◦ and 0.0871 m. The root
The model solves the hydrodynamic primitive equations us- mean squared errors (RMSE) for current velocity, current direction
ing the hydrostatic assumption in the vertical direction with the and water level in Sanya Bay station are 8.56 cm/s, 16.06◦ and
Boussinesq simplification for convective flows. Using Cartesian 0.0616 m. In general, the simulated and observed values are in a
coordinates, the model domain (Fig. 1) is divided into grids of 330 good agreement. Although the errors are not negligible for current
by 315, with a horizontal resolution of 2 km by 2 km, and 10 levels velocity and current direction at these two observational stations,
in the vertical. The depth field in the domain is extracted from they are still acceptable. On one hand, the Tieshan Bay and Sanya
ETOPO1 with a resolution of 1◦ by 1◦ . Bay are semi enclosed bays with complicated terrain; the input ter-
In this study, eight tidal harmonics (K1 , O1 , P1 , Q1 , M1 , S2 , N2 , rain used for model simulation may cause some errors that may in-
K2 ), extracted from the Oregon State University tidal model, are fluence the model results. On the other hand, the current observed
 L. Wang et al. / Regional Studies in Marine Science 18 (2018) 35–43 37

Table 1
Comparison between observation and simulation of harmonic constants.
No M2 K1 O1
Amplitude Phase Amplitude Phase Amplitude Phase
(cm) (◦ ) (cm) (◦ ) (cm) (◦ )
1# 22.1 319.9 33.0 332.8 30.6 285.2
(22.3) (319.9) (28.7) (333.3) (27.7) (283.2)
2# 17.2 58.2 52.5 66.4 59.7 2.2
(18.3) (59.7) (54.0) (71.2) (63.3) (10.3)
3# 19.4 133.9 71.0 83.1 77.8 16.2
(18.8) (130.2) (69.7) (85.7) (81.4) (19.4)
4# 23.6 165.1 66.9 81.0 74.3 14.0
(25.5) (174.7) (66.1) (88.2) (78.5) (29.4)
5# 22.9 162.59 69.6 81.6 76.8 14.6
(23.6) (159.8) (72.3) (90.7) (73.5) (28)
6# 41.0 173.7 86.4 92.0 92.2 24.2
(46.2) (180.3) (89.3) (97.3) (96.9) (34.7)
7# 32.1 177.2 83.4 95.0 89.4 27.2
(31.4) (180.8) (83.2) (98.9) (90.5) (37.6)
8# 36.1 168.3 81.9 89.7 87.9 22.1
(37.9) (171.3) (84.3) (92.9) (93.1) (30.7)
9# 21.2 176.9 80.7 95.9 86.8 28.1
(20.0) (179.0) (80.0) (85.0) (80.0) (33.0)
10# 20.7 180.7 79.2 98.1 85.4 30.2
(19) (179) (72) (96.0) (81.0) (30.0)
11# 22.4 171.9 75.5 96.3 81.9 28.5
(25.0) (174) (80.0) (83.0) (79.0) (32.0)
12# 9.3 115.1 63.4 92.2 70.3 25.0
(10.0) (129.0) (65.0) (95.0) (71.0) (26.0)
13# 23.0 20.5 52.3 105.8 60.1 37.6
(20.0) (26.0) (60.0) (107.0) (60.0.0) (52.0)
14# 25.9 22.1 47.3 108.6 55.2 40.1
(30.0) (31.0) (50.0) (102.0) (59.0) (45.0)

in these stations is influenced by many factors such as small river been introduced to quantify the exchange and transport processes
inputs that the model simulation did not take into consideration. and to assess the self-purification and assimilative capacity of a
The maps of monthly mean MODIS SST observations are selected water body. Bolin and Rodhe (1973) summarize the concepts on
to assess the model performance in simulating the temperature. the basis of ‘‘age’’, which is defined as the time that has elapsed
As Fig. 4 shows, the averaged model surface temperature in June since the material element entered the reservoir. Zimmerman
2014 and January 2015 over Beibu Gulf is similar to the satellite SST (1976) introduced the residence time as a complement to the age.
patterns. The modeled SST pattern in the coastal area in June is a The residence time for each material element is defined as the time
little different from the observations because the low resolution of
taken for the element to reach the outlet.
the satellite data may cause low-accuracy observations in coastal
Here we adopt the average residence time defined by Takeoka
areas. The elevation, current, and temperature results show that
(1984) as the renewal time for a given water parcel in this study.
the model is able to provide a hydrodynamic field for further study
of water exchange processes. Consider a parcel of material in a reservoir. Let the initial time be
t0 , the amount of the material at t0 be R0 , and the amount of the
2.2. Particle tracking model material which still remains in the reservoir at time t be Rt . Rt is
the amount of the material whose residence time is larger than t.
The Random Walk Particle Tracking (RWPT) model is used The function of the distribution of residence time can be defined
for the water exchange simulation. The RWPT model runs offline as:
with hydrodynamic model outputs of velocity to calculate particle 1 dRt
paths. The equation for particle locations is integrated using a τr′ = − . (2)
fourth-order accurate Runge–Kutta method, and can be described R0 dt
as advection because of the currents and random walk associated After an integration by parts, the average residence time τr is
with the sub-grid-scale diffusion (Tompson, 1993; Dimou and defined as:
Adams, 1993; Salamon et al., 2006). ∫ ∞ ∫ ∞
Rt
t +1t
√ τr = dt = r (t ) dt . (3)
X = X + U 1t + R (t )
t
2Kh 1t . (1) 0 R0 0

where X t and X t +1t are, respectively, the position vectors of a Here, r (t ) = Rt /R0 is called the remnant function. The function
passive particle at time t and t + 1t , U is the velocity vector of r (t ) denotes the decrease in the material considered and the
the model flow field, 1t is the random walk time step, R (t ) is exchange or transport of the material that is directly described by
the uniform distribution random number in the interval −1 to 1, this function.
here a uniform deviate given by the Fortran 90 random number
generator. Kh is the eddy diffusivity coefficient for the random 3. Circulation patterns
walk in the horizontal directions, derived from the hydrodynamic
model.
The mean circulation pattern in the gulf is one of the significant
2.3. Residence time calculation factors responsible for water exchange and water movement in the
gulf. We obtained the mean current fields in July and December,
Many concepts of time scales such as age, residence time, transit representing the summer and winter circulation patterns in 2014
time and flushing time (Monsen et al., 2002; Prandle, 1984) have (Fig. 5).
38 L. Wang et al. / Regional Studies in Marine Science 18 (2018) 35–43

The result shows that the cyclonic circulation patterns appeared

both in summer and winter, consistent with recent studies (Gao
et al., 2013; Chen et al., 2015). The cyclonic circulation forms
two main flow systems. One flows into the northern part of the
gulf from the QS and feeds the cyclonic circulation (Wu et al.,
2008). The other flows northward from the southern and western
coasts of Hainan Island. This northward current originates from
one portion of the SCS southwestward shelf current, most of which
turns southward and bypasses the gulf (Gao et al., 2013).
In the summer, the northward flow moves to the north of the
gulf and joins the westward flow coming from the QS. The joint
flow is northwestward, forming the western limb of the cyclonic
circulation. An anti-cyclonic flow appears in the northwestern
gulf, and an enclosed cyclonic circulation exits south of the gulf.
Driven by the summer monsoon, flows in the summer are mostly
eastward and northeastward. The maximum velocity occurs along
the Vietnam coast. The current velocity in coastal areas is larger
than in the outer sea.
In the winter, the westward flow from the QS is separated
into two branches. One branch flows northwestward along the Fig. 2. Sub-regions defined by fuzzy C-means clustering method. They are 1.
Guangxi coast and joins the gulf-wide cyclonic circulation. The The Guangxi Coast. 2. The West Coast of Hainan Island. 3. The Northern Gulf. 4.
The Central Gulf. 5. The Southern Gulf. 6. The Vietnamese Coast. (The rest is the
other branch flows along the northwestern Hainan Island coast and outer sea.) The color bar represents original nine groups divided by fuzzy C-means
joins the aforementioned northward current. The joint northward clustering method.
current moves westward across the gulf at ∼19.5◦ N. The enclosed
circulation in the south of the gulf is anti-cyclonic in winter. Flows Table 2
are mostly westward and southwestward under the northeasterly Average residence times in sub-regions (days).
monsoon forcing. The current velocity in the north of the gulf is Sub-regions Winter Summer
larger than in the south of the gulf. The maximum velocity appears Guangxi coast 33 49
in the QS. Hainan Island coast 5 22
Northern gulf 9 39
Central gulf 17 40
4. Water exchange Southern gulf 24 32
Vietnamese coast 71 49
The Beibu Gulf water is divided into nine groups (Fig. 2) Whole gulf 66 71
using the fuzzy C-means clustering technique (Bezdek, 2013),
based on the yearly averaged sea surface temperature (SST) and
Chlorophyll-a (Chl-a) concentration data which is derived from the 4.1. Residence times
Moderate Resolution Imaging Spectroradiometer (MODIS Aqua)
with 4 km spatial resolution. The groups (G7, G8 and G9 in Fig. 2) The residence times are computed for the whole gulf and each
are characterized by high Chl-a contents. The Group 6 (G6 in Fig. 2) sub-region. Table 2 shows that the residence time for the whole
has low SST with the mean value of 24.8◦ . Group 5 has high SST gulf has a small seasonal variation, 66 days in winter and 71 days
with the mean value of 27◦ . According to the topography, and in summer, while the residence times for the sub-regions have
physical parameters (SST and Chl-a contents) in each group, we larger seasonal variations. Generally, the sub-regions exhibit short
combine some groups and divide the gulf into six sub-regions residence times in winter, except for the Vietnamese coast. This
might be caused by the fact that movement of particles near the
(Fig. 2). The water in these sub-regions can be characterized with
Vietnamese coast is limited by the topography and coastal current,
coastal waters and mixed waters. The coastal waters in Guangxi
leading to a long residence time in winter. The regional residence
and Vietnamese coasts has high Chl-a contents may be contributed
times show a spatial difference. Among the six sub-regions, parti-
by nutrients from the river discharges into the gulf (Deetae and
cles off the west coast of Hainan Island have the shortest residence
Wisespongpand, 2001). The coastal water in the west coast of
times, and the Vietnamese coast has comparative long residence
Hainan Island has low surface temperature and High Chl-a con-
times, reaching 71 days in winter.
tent. This may be because of the local upwelling and tidal mixing
(Lü et al., 2008). The gulf mixed waters in the Northern, Central 4.2. Water exchange in summer
and Southern sub-regions are influenced by the water intrusion
through Qiongzhou Strait and the south opening. Contrary to the The water exchange curves show that about 40% of Guangxi
Central and the Southern Gulf sub-regions, the Northern Gulf ex- coast particles are exported to the central gulf (Fig. 6a, green line),
hibits low surface temperature characterization. while only 19% remains along the Guangxi coast by the end of the
In order to study the water exchange between these sub- simulation for the summer season (Fig. 6a, red line). The exported
regions, particles are released near the surface in each sub-region Guangxi coast particles work their way through the northern gulf
with a uniform distribution, and tracked with a 60-minute time (Fig. 6a, blue line) and central gulf (Fig. 6a, green line), eventually
step for 90 days in each season or until they cross the boundaries. reaching the southern gulf (Fig. 6a, pink line) as incrementally
In addition to analyzing the water exchange in the interior gulf, increasing fractions in the southern gulf.
the influence areas of current flows from the Qiongzhou Strait About 60% of the Hainan Island west coast particles are rapidly
(QS) and south opening are quantified. Particles are released across flushed out within the first 15 days of simulation (Fig. 6b, orange
transects of the QS and south opening during the winter and line). Most of the flushed particles move into the northern gulf
summer seasons (Fig. 8, red lines). (Fig. 6b, blue line) and remain there for a considerable amount of
 L. Wang et al. / Regional Studies in Marine Science 18 (2018) 35–43 39

Fig. 3. Comparison of observed and modeled water levels, current velocity, and current direction at Tieshan Bay station (a, b, c) and Sanya Bay station (d, e, f).

Fig. 4. Monthly mean SST (◦ C) calculated by model for (a) January 2015 and (c) June 2014; monthly mean satellite SST (◦ C) maps for (b) January 2015 and (d) June 2014.

time before being slowly flushed out to the central gulf in the final are found along the west coast of Hainan Island, they do commu-
20 days (Fig. 6b, green line). Although few Guangxi coast particles nicate with each other, as evidenced by a fraction of the Hainan
40 L. Wang et al. / Regional Studies in Marine Science 18 (2018) 35–43

Fig. 5. Vertically averaged circulation in July and December 2014.

Fig. 6. Graphs showing fraction of particles remaining in starting area or traveling to other areas during summer. Sub-regions are colored red, orange, light blue, green,
pink, gray, and black, representing Guangxi coast, west coast of Hainan Island, northern gulf, central gulf, southern gulf, Vietnamese coast, and outer sea, respectively. (For
interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Island west coast particles reached the Guangxi coast (Fig. 6b, red Its influence area shrinks to a narrow band along Guangxi and the
line). Another small fraction (about 20%) moves southward along Leizhou Peninsula near the shore.
Hainan Island to the outer sea (Fig. 6b, black line).
The northern gulf exports about 44% of its particles to the cen- 4.3. Water exchange in winter
tral gulf (Fig. 6c, blue line) by the end of the simulation. Meanwhile,
it has received 24% of the central gulf particles (Fig. 6d, blue line). During winter, a large percentage of the Guangxi coast and
The central gulf particles show a trend similar to the northern gulf northern gulf particles reach the Vietnamese coast by the end of the
particles. Roughly 30% of the particles are transported into the simulation (Fig. 7a and c, gray lines). The Guangxi coast particles
southern gulf, with only 17% retained (Fig. 6d, green line) by the travel through the northern and central gulfs into the region off
end of the simulation. In addition, a small portion of the central the Vietnamese coast. (Fig. 7a, blue line and green line), and the
gulf particles is found in the outer sea (Fig. 6d, gray line). As the northern gulf particles move through the Guangxi coast and central
gate to the South China Sea, the southern gulf and Vietnamese coast gulfs (Fig. 7c, red line and green line) to the Vietnamese coast
mainly export their particles to the outer sea (Fig. 6e and f). (Fig. 7c, gray line). The pathways suggest that the Guangxi coast
As Fig. 8a indicates, the current flow coming from the south and northern gulf particles have a tendency to follow the western
opening is dominant, and influences most of the eastern part of limb of the cyclonic circulation.
the gulf in summer. In contrast, the current flow from the QS moves The west coast of Hainan Island still has a high rate of water
cyclonically along the Leizhou Peninsula and Guangxi coast region. exchange, similar to summer (Fig. 7b, orange line). These flushed
 L. Wang et al. / Regional Studies in Marine Science 18 (2018) 35–43 41

Fig. 7. Graphs showing fraction of particles remaining in starting area or traveling to other areas during winter. Sub-regions are colored red, orange, light blue, green, pink,
gray, and black, representing Guangxi coast, west coast of Hainan Island, northern gulf, central gulf, southern gulf, Vietnamese coast, and the outer sea, respectively. (For
interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 8. Position of particles released to QS (green) and south opening (blue) in summer (a) and winter (b); red lines are original release transects. (For interpretation of the
references to color in this figure legend, the reader is referred to the web version of this article.)

particles move westward across the gulf, reaching the Vietnamese particles exported into this region (Fig. 7b, green line), but which
coast (Fig. 7b, gray line). The other exported particles move along are slowly flushed during the final 20 days to the Vietnamese coast
the near-shore region of Hainan Island to the sea, which can be (Fig. 7b, gray line).
seen from the increased fraction in the outer sea (Fig. 7b, black The southern gulf exports almost 90% of its particles to the outer
line). sea, and a small portion to the Vietnamese coast (Fig. 7e, black
Almost 80% of the central gulf particles are drained out within line and gray line). A small fraction of its particles that exist in the
the first 30 days of the simulation (Fig. 7d, green line), moving to- central gulf during the early simulation period indicates that the
ward the Vietnamese coast (Fig. 7d, gray line). The drained particles southern gulf particles move northward as well (Fig. 7e, green line).
eventually end up along the Vietnamese coast or are transported The majority of Vietnamese coast particles remain in the original
to the outer sea (Fig. 7d, gray line and black line). On the other region, while a small fraction (about 32%) are transported to the
hand, the central gulf shows the characteristics of a transitional South China Sea (Fig. 7f, black line).
region, where particles originating from adjacent sub-regions do The distribution of current flow through the QS shows a ten-
not reside for long, but move through to other regions (Fig. 7d). dency to move cyclonically, following the winter current pat-
This is obvious from the large number of Hainan Island west coast tern. The QS current flow has a great influence on gulf water
42 L. Wang et al. / Regional Studies in Marine Science 18 (2018) 35–43

Table 3
Monthly averaged water transport in QS in 2014 (Sv: 106 m3 s−1 ).
1 2 3 4 5 6 7 8 9 10 11 12
Q (u) 0.12 0.11 0.06 0.05 0.03 0.03 0.08 0.03 0.07 0.13 0.14 0.16
Q (v) 0.03 0.02 0.01 0.01 0.01 0.01 0.02 0.01 0.02 0.03 0.03 0.03

refreshment. However, the current flow from the south opening reach the gulf through the QS (Tang et al., 2003). Our result also
is blocked at the mouth of the gulf (Fig. 7b). shows a strong westward current through the QS and a northward
current along the west coast of Hainan Island that exceeds 1 m/s of
5. Discussion velocity, implying the current flow from these two openings have a
significant impact on gulf water refreshment. The water transport
The residence times identify the time taken by water particles to through the QS transect is estimated as 0.18 Sv (Sv ≡ 106 m3 s−1 ) in
leave the release areas. The residence time for the whole gulf varies summer and 0.52 Sv in winter based on the hydrodynamic model,
little between winter and summer. The comparatively longer res- consistent with previous studies (Shi et al., 2002; Chen et al., 2007).
idence time in summer may be caused by weak currents in the The mean westward water transport was 0.08 Sv in 2014, four
gulf and the strong northeast SCS current in the mouth of the gulf times larger than the mean southward water transport (Table 3),
limiting the particles’ transportation to the outer sea. The residence indicating cold and less saline water is transported into the gulf
time for each sub-region reveals strong seasonal variation. The sub- (Bao et al., 2005). Also, the strong current flow from the south
regions exhibit shorter residence times in winter. This might be opening, with its high temperature and high salinity, can provide
related to the circulation patterns in Beibu Gulf. Currents in the nutrients and a warm water environment for marine life in the gulf.
gulf are stronger in winter than in summer, forcing water parcels to
move out faster from the original sub-regions, resulting in shorter 6. Conclusion
residence times. The Vietnamese coast, however, is an exception,
with an extremely long residence time in winter. This can be The seasonal characteristics of water exchange and water
seen in Fig. 7f (gray line), where most of the Vietnamese coast movement in Beibu Gulf were investigated using the particle track-
particles are retained in this region. This retention may coincide ing method. Specifically, we paid attention to the water exchange
with the location of water pile-up, which is caused by the force of processes and water movement pathways of six sub-regions in the
the northeasterly monsoon (Gao et al., 2013). Since the prevailing gulf. We also quantified the influence areas of the current flows
northeasterly wind is favorable to downwelling, the coastal waters from the QS and the south opening.
are pushed close to shore, resulting in a water pile-up along the The residence time in the whole gulf was 66 days in winter and
western shore. In contrast, the upwelling-favorable southwesterly 71 days in summer. The regional residence times exhibit seasonal
monsoon wind causes the coastal water to spread further offshore and spatial differences. Except for the Vietnamese coast, other sub-
(Lü et al., 2008; Gao et al., 2014), leading to a short residence time. regions have shorter residence times in winter. The west coast
The spatial difference of the sub-regions may also be related to of Hainan Island had the shortest residence time, while the Viet-
local currents. For example, the shortest residence time (Table 2) namese coast had the longest residence time, which is associated
in both seasons, and the variable transported destinations (Figs. 6b with the topography and complex local currents.
and 7b), indicate the region off the west coast of Hainan Island has Water exchange analysis provides a description of how water
an active water exchange ability. This coincides with the strong parcels from different sub-regions interact with one another. The
current and upwelling along the west coast of Hainan Island (Hu water particle movement pathways suggest that the gulf water
et al., 2003; Lü et al., 2008; Kuo et al., 2000). moves cyclonically in winter. The water particles from coastal
The water exchange curves provide us clear insights into the areas have offshore movement and mix strongly in the gulf during
water exchange processes and water movement pathways in the summer. The current flows from the QS and the south opening
sub-regions. The journey of particles originating from the sub- were observed by continuously releasing particles across the two
regions suggests water movement in the gulf tend to follow the transects. The westward water transport from the QS influences
circulation patterns. In winter, the cyclonically alongshore move- the most northern part of the gulf in winter. The current flow from
ment is consistent with the gulf-wide cyclonic circulation pattern. the south opening moves along the west coast of Hainan Island and
In summer, it is observed that water particles released from coastal influences the most eastern part of the gulf.
sub-regions tend to move offshore and mix strongly with particles The Random Walk Particle Tracking model is useful for ex-
released in the center of the gulf. The offshore particles’ move- amining the regional water exchange in Beibu Gulf. Moreover, it
ment is consistent with the southwesterly wind-induced offshore provides a more intuitive viewpoint of how current flow influences
currents. Knowledge of how water interact between these criti- gulf water by releasing particles along the QS and south opening
cal areas is very important for understanding the distribution of transacts. In future research, it would be worthwhile to include
biological resources in the gulf. The freshwater river discharge in the spring and fall seasons in the simulations, to provide more
coastal areas, in some cases, carries abundant nutrient salts into details on seasonal variation. Also, the distribution of pollutants
the gulf (Lai et al., 2014), which is necessary for fish, shrimp, and and nutrients in the gulf requires further investigations, as we only
shellfish growth (Chen et al., 2009). In other cases, these discharges studied the potential pathways of pollutants and nutrients carried
may upset nutrient balances and bring toxic materials (Zheng et al., by water from different sources.
2012; Kaiser et al., 2015). The offshore water movement in summer
is favorable for spreading nutrients. The cyclonical alongshore Acknowledgments
water movement in winter is beneficial for water refreshing by the
outer sea. This study is supported by the National Key Research and Devel-
In addition, the Qiongzhou Strait and the south opening play opment Program of China [No. 2016YFA0601201], by the program
a significant role in gulf water refreshment. According to Shi et of Xiamen Southern Oceanographic Center [No. 15PZB009NF05],
al.’s (2002) study, the QS may cause up to 44% of gulf water by a joint effort of the University of Delaware and Xiamen Univer-
to be refreshed each season through the year-round westward sity’s Joint Institute for Coastal Research and Management (Joint
flow. Additionally, discharge from the Pearl River is reported to Institute of CRM), by the Natural Science Foundation of China
 L. Wang et al. / Regional Studies in Marine Science 18 (2018) 35–43 43

[41476007 and 41630963], and by the State Oceanic Administra- Ninto, Y., Garcia, M.H., 1996. Experiments on particle—turbulence interactions
tion of China Global Change and Air–Sea Interaction Project [GASI- in the near–wall region of an open channel flow: Implications for sediment
transport. J. Fluid Mech. 326, 285–319.
IPOVAI-01-04 and GASI-02-PAC-YGST2-02]. The authors would
Oliveira, A., Baptista, A.M., 1997. Diagnostic modeling of residence times in estuar-
like to thank Young-Heon Jo and Jinku Park for providing technique ies. Water Resour. Res. 33, 1935–1946.
support and useful comments. Perianez, R., 2004. A particle-tracking model for simulating pollution dispersion in
the Strait of Gibraltar. Mar. Pollut. Bull. 49, 613–623.
References Perianez, R., 2005. An operative Lagrangian model for simulating radioactivity
dispersion in the Strait of Gibraltar. J. Environ. Radioact. 84, 95–101.
Perianez, R., Elliott, A.J., 2002. A particle-tracking method for simulating the disper-
Bao, X., Hou, Y., Chen, C., Chen, F., Shi, M., 2005. Analysis of characteristics and
sion of non-conservative radionuclides in coastal waters. J. Environ. Radioact.
mechanism of current system on the west coast of Guangdong of China in
58, 13–33.
summer. Acta Oceanol. Sin. 24, 1–9.
Prandle, D., 1984. A modelling study of the mixing of 137Cs in the seas of the
Bezdek, J.C., 2013. Pattern Recognition with Fuzzy Objective Function Algorithms.
European continental shelf. Philos. Trans. R. Soc. Lond. Ser. A 310, 407–436.
Springer Science & Business Media.
Pruszak, Z., Van Ninh, P., Szmytkiewicz, M., Hung, N.M., Ostrowski, R., 2005.
Bilgili, A., Proehl, J.A., Lynch, D.R., Smith, K.W., Swift, M.R., 2005. Estuary/ocean
Hydrology and morphology of two river mouth regions (temperate vistula delta
exchange and tidal mixing in a Gulf of Maine Estuary: A Lagrangian modeling
and subtropical red river delta). Oceanologia 47, 365–385.
study. Estuar. Coast. Shelf Sci. 65, 607–624.
Salamon, P., Fernàndez-Garcia, D., Gómez-Hernández, J.J., 2006. A review and nu-
Bolin, B., Rodhe, H., 1973. A note on the concepts of age distribution and transit time
merical assessment of the random walk particle tracking method. J. Contam.
in natural reservoirs. Tellus 25, 58–62.
Hydrol. 87, 277–305.
Chen, B., 1986. Preliminary discussion on the formation and characteristic of the
Sato, T., Tonoki, K., Yoshikawa, T., Tsuchiya, Y., 2006. Numerical and hydraulic
water system in the Beibu Gulf. J. Guangxi Acad. Sci. 2, 92–95 (in Chinese).
simulations of the effect of density current generator in a semi-enclosed tidal
Chen, B., Yan, J.H., Wang, D.R., Shi, M.C., 2007. The transport volume of water
bay. Coastal Eng. 53, 49–64.
through the Qiongzhou Strait in the winter season. Period. Ocean Univ. China
Saxton, M.J., Jacobson, K., 1997. Single-particle tracking: Applications to membrane
37, 357–364.
dynamics. Annu. Rev. Biophys. Biomol. Struct. 26, 373–399.
Chen, Z., Qiao, F., Xia, C., Wang, G., 2015. The numerical investigation of seasonal
Shi, M., Chen, C., Xu, Q., Lin, H., Liu, G., Wang, H., Wang, F., Yan, J., 2002. The role of
variation of the cold water mass in the Beibu Gulf and its mechanisms. Acta
Qiongzhou Strait in the seasonal variation of the South China Sea circulation. J.
Oceanol. Sin. 34, 44–54.
Phys. Oceanogr. 32, 103–121.
Chen, Z.Z., Qiu, Y.S., 2005. Ecological distribution of Parargyrops edita Tanaka in the
Signell, R., Butman, B., 1992. Modeling tidal exchange and dispersion in Boston
Beibu Gulf. Mar. Fish. Res. 26, 16–21.
Harbor. J. Geophys. Res. 97, 591–615.
Chen, Z., Xu, S., Qiu, Y., Lin, Z., Jia, X., 2009. Modeling the effects of fishery man-
Suh, S.W., 2006. A hybrid approach to particle tracking and Eulerian–Lagrangian
agement and marine protected areas on the Beibu Gulf using spatial ecosystem
models in the simulation of coastal dispersion. Environ. Model. Softw. 21, 234–
simulation. Fisheries Res. 100, 222–229.
242.
Deetae, S., Wisespongpand, P., 2001. Sub-thermocline Chlorophyll maximum in the
Takeoka, H., 1984. Fundamental concepts of exchange and transport time scales in
South China Sea, area IV: Vietnamese waters. In: Proceedings of the SEAFDEC
a coastal sea. Cont. Shelf Res. 3, 311–326.
Seminar on Fisheries Resources in the South China Sea 4, pp. 251–264.
Tang, D., Kawamura, H., Lee, M.A., Van Dien, T., 2003. Seasonal and spatial distribu-
Dias, J.M., Lopes, J.F., Dekeyser, I., 2001. Lagrangian transport of particles in Ria de
tion of chlorophyll-a concentrations and water conditions in the Gulf of Tonkin,
Aveiro lagoon, Portugal. Phys. Chem. Earth Part B: Hydrol. Oceans Atmosphere
South China Sea. Remote Sens. Environ. 85, 475–483.
26, 721–727.
Thompson, K.R., Dowd, M., Shen, Y., Greenberg, D.A., 2002. Probabilistic characteri-
Dimou, K.N., Adams, E.E., 1993. A random-walk, particle tracking model for well-
zation of tidal mixing in a coastal embayment: A Markov Chain approach. Cont.
mixed estuaries and coastal waters. Estuar. Coast. Shelf Sci. 37, 99–110.
Shelf Res. 22, 1603–1614.
Gao, J., Shi, M., Chen, B., Guo, P., Zhao, D., 2014. Responses of the circulation and
Tompson, A.F.B., 1993. Numerical simulation of chemical migration in physically
water mass in the Beibu Gulf to the seasonal forcing regimes. Acta Oceanol. Sin.
and chemically heterogeneous porous media. Water Resour. Res. 29, 3709–
33, 1–11.
3726.
Gao, J., Xue, H., Chai, F., Shi, M., 2013. Modeling the circulation in the gulf of Tonkin,
Van Maren, D.S., Hoekstra, P., 2004. Seasonal variation of hydrodynamics and
South China Sea. Ocean Dyn. 63, 979–993.
sediment dynamics in a shallow subtropical estuary: The Ba Lat River, Vietnam.
Hu, J.Y., Kawamura, H., Tang, D.L., 2003. Tidal front around the Hainan Island,
Estuar. Coast. Shelf Sci. 60, 529–540.
northwest of the South China Sea. J. Geophys. Res.: Oceans 108, 3342.
Wroblicky, G.J., Campana, M.E., Valett, H.M., Dahm, C.N., 1998. Seasonal variation in
Kaiser, D., Hand, I., Schulz-Bull, D.E., Waniek, J.J., 2015. Organic pollutants in the
surface-subsurface water exchange and lateral hyporheic area of two stream-
central and coastal Beibu Gulf, South China Sea. Marine Pollut. Bull. 101, 972–
aquifer systems. Water Resou Res. 34, 317–328.
985.
Wu, D., Wang, Y., Lin, X., Yang, J., 2008. On the mechanism of the cyclonic circulation
Kano, Y., Sato, T., Kita, J., Hirabayashi, S., Tabeta, S., 2010. Multi-scale modeling of
in the Gulf of Tonkin in the summer. J. Geophys. Res.: Oceans 113, C09029.
CO 2 dispersion leaked from seafloor off the Japanese coast. Marine Pollut. Bull.
Xu, X., Qui, Z., Chen, H., 1980. Summary of the horizontal circumcurrent in south
60, 215–224.
china sea(in chinese with english abstract). In: Chinese Society of Oceanology
Kuo, N.J., Zheng, Q., Ho, C.R., 2000. Satellite observation of upwelling along the
and Limnology. Proceedings of the Hydrology and Meteorology. Science Press,
western coast of the South China Sea. Remote Sens. Environ. 74, 463–470.
Beijing, pp. 137–145.
Lai, J., Jiang, F., Ke, K., Xu, M., Lei, F., Chen, B., 2014. Nutrients distribution and trophic
Yu, M.G., Liu, J.F., 1993. The system and pattern of the South China Sea circulation
status assessment in the northern Beibu Gulf, China. Chin. J. Oceanol. Limnol. 32,
(in Chinese with English abstract). Ocean Predict. 10, 13–17.
1128–1144.
Yuan, S., Deng, J., 1999. A numerical study on circulation in the Beibu Gulf (in
Liu, W.C., Chen, W.B., Hsu, M.H., 2011. Using a three-dimensional particle-tracking
Chinese with English abstract). Nanhai Yanjiu Yu Fazhan 12, 41–46.
model to estimate the residence time and age of water in a tidal estuary.
Zeng, B., Zhang, J., Chen, G., 1989. Survey of Fishery Resources and Regional Fish-
Comput. Geosci. 37, 1148–1161.
eries Planning in South China Sea. Guangdong Scientific & Technological Press,
Lü, X., Qiao, F., Wang, G., Xia, C., Yuan, Y., 2008. Upwelling off the west coast of
Guangzhou (in Chinese).
Hainan Island in summer: Its detection and mechanisms. Geophys. Res. Lett.
Zheng, L., Chen, C., Liu, H., 2003. A modeling study of the Satilla River Estuary,
35, L02604.
Georgia. I: Flooding-drying process and water exchange over the salt marsh-
Marine Environment Committee, 2003. MEC Ocean Model Operation Manual Ver.
estuary-shelf complex. Estuaries 26, 651–669.
1.1. Soc. Naval. Archit., Japan (in Japanese).
Zheng, Q., Zhang, R., Wang, Y., Pan, X., Tang, J., Zhang, G., 2012. Occurrence and
Mizumukai, K., Sato, T., Tabeta, S., Kitazawa, D., 2008. Numerical studies on eco-
distribution of antibiotics in the Beibu Gulf, China: Impacts of river discharge
logical effects of artificial mixing of surface and bottom waters in density
and aquaculture activities. Mar. Environ. Res. 78, 26–33.
stratification in semi-enclosed bay and open sea. Ecol. Modell. 214, 251–270.
Zimmerman, J.T.F., 1976. Mixing and flushing of tidal embayments in the western
Monsen, N.E., Cloern, J.E., Lucas, L.V., 2002. A comment on the use of flushing time,
Dutch Wadden Sea part I: Distribution of salinity and calculation of mixing time
residence time, and age as transport time scales. Limnol. Oceanog. 47, 1545–
scales. Neth. J. Sea Res. 10, 149–191.
1553.