1. Introduction
The South China Sea (SCS) is the largest semienclosed marginal sea in the northwestern Pacific Ocean. The Luzon Strait (LS), with a sill depth of approximately 2400 m, is the only passage connecting the deep SCS with surrounding oceans. The water characteristics in the deep SCS resemble those in the northwest Pacific at approximately 2000 m (Qu 2002). A persistent pressure gradient below approximately 1500 m between the two sides of the LS drives the deep-water overflow (Qu et al. 2006a), through which the cold and saline North Pacific Deep Water (NPDW) penetrates the deep SCS, driving abyssal cyclonic circulation (Lan et al. 2013; Wang et al. 2011; Yuan 2002). Abyssal currents in the LS and SCS seemingly play important roles in deep-water renewal, heat budget, and sediment distributions in the deep SCS, and have potential impacts on regional and global climate (Chen et al. 2013; Liu et al. 2010; Lüdmann et al. 2005; Qu et al. 2006b; Qu et al. 2009; Shao et al. 2007).
Deep-water overflow in the LS transports approximately 0.7–2.5 Sv (1 Sv ≡ 106 m3 s−1) of NPDW into the SCS according to previous studies using diagnostic analysis (Qu et al. 2006a; Song 2006; Wang 1986), hydrographic data (Tian et al. 2006; Yang et al. 2010; Zhao et al. 2014), and in situ observations (Chang et al. 2010; Liu and Liu 1988; Ye et al. 2019; Zhao et al. 2016; Zhou et al. 2014). Hydraulic theory provided an overflow transport of 2.5 Sv under the assumption of zero potential vorticity and a flat bottom (Qu et al. 2006a). According to full-depth profile measurements at 15 stations, approximately 2 Sv of NPDW penetrates the SCS below 1500 m (Tian et al. 2006). Based on in situ observations and numerical modeling results, Zhao et al. (2014) provided the vertical structure of the deep-water overflow. The first long-term observation in the deep LS estimated transport of approximately 1.2 Sv using a single current meter (Liu and Liu 1988). Nine-month observations found that the majority of NPDW enters the LS through the Bashi Channel rather than the Taltung Canyon with intraseasonal variations (Chang et al. 2010). In the last decade, a more detailed and accurate spatiotemporal structure of the overflow was identified using several long-term mooring observations in the LS (Ye et al. 2019; Zhou et al. 2014). Deep-water overflow in the LS is strong in late autumn but weak in spring and shows intraseasonal variations with dominant periods of 20–60 days and approximately 100 days (Ye et al. 2019; Zhao et al. 2020; Zhou et al. 2014; Zhu et al. 2017a).
The classical Stommel–Arons abyssal circulation theory predicts abyssal cyclonic circulation with a strong deep western boundary current (DWBC) in the SCS (Stommel and Arons 1960a; Stommel and Arons 1960b). This circulation has been presented by water property analyses (Li and Qu 2006; Qu et al. 2006a) and numerical simulations (Gan et al. 2016; Lan et al. 2013; Xie et al. 2013; Yuan 2002). Deep overflow through the LS is considered to be the primary force of such circulation, and the potential vorticity budget is important for circulation formation (Cai and Gan 2019; Gan et al. 2016; Lan et al. 2013). However, observations of abyssal circulation and the DWBC in the SCS remain sparse. Using the monthly climatology database from the U.S. Navy Generalized Digital Environment Model, Wang et al. (2011) described the abyssal circulation and indicated that the DWBC transport reaches approximately 3.0 Sv. Observations from six moorings deployed near the Zhongsha Islands showed a narrow southwestward DWBC with a volume transport of 1.65 Sv and a core velocity of 2.0 cm s−1 (Zhou et al. 2017). NPDW from the LS deep overflow can reach the upper ocean and affect the upper circulation through upwelling along the continental shelf and near islands with rough topographies (Shu et al. 2014; Wang et al. 2012) due to enhanced diapycnal mixing (Alford et al. 2011; Tian et al. 2009; Yang et al. 2016).
The abyssal current in the SCS showed significant temporal variations. Modeling studies have indicated that abyssal circulation intensified in summer but weakened in winter, which is associated with LS deep overflow (Gan et al. 2016; Lan et al. 2015). The DWBC near the Zhongsha Islands intensified in spring and summer but weakened in autumn and winter, according to Zhou et al. (2017). Observations from 11 current- and pressure-recording inverted echo sounders (CPIESs) revealed that the abyssal current in the northern SCS was stronger in summer and autumn but weaker in winter and spring (Zheng et al. 2021a). Upper-ocean perturbations such as mesoscale eddies (Shu et al. 2018; Sun et al. 2020; Zhang et al. 2013; Zhang et al. 2016) and near-inertial waves (Xiao et al. 2016) can reach the deep ocean and impact the abyssal current in the SCS. Intraseasonal variations in the deep SCS vary from ∼10 to ∼120 days, possibly related to topographic Rossby waves (Shu et al. 2016; Shu et al. 2022; Wang et al. 2019; Zheng et al. 2021b; Wang et al. 2021), deep-water overflow variations, and deep eddies (Shu et al. 2022; Zhou et al. 2020).
Although several studies have focused on the abyssal current in the LS, a detailed and accurate spatiotemporal three-dimensional structure of abyssal circulation remains uncertain. Flows after the NPDW entering the SCS through the west gap of the LS have never been described in detail using direct observations. Discussion of the influence of the upper perturbations on the abyssal current also remains sparse. Moreover, the reason for seasonal variations in the abyssal current showing different patterns in the LS (intensifies in late autumn) and SCS (generally intensifies in summer) also remains unclear. This study addresses these knowledge gaps by focusing on the spatial structure and seasonal variations in the abyssal circulation west of the LS, utilizing 27 CPIESs deployed in the region.
The remainder of this paper is organized into the following: data and methods (section 2), spatial structure and seasonal variations (section 3), discussion (section 4), and summary (section 5).
2. Data and methods
a. CPIES measurements
A total of 28 CPIESs were deployed west of the LS between June 2018 and July 2019 for over 400 days, aiming to clarify the spatiotemporal evolution of the current system near the LS in the SCS (Fig. 1). Of these, 27 were successfully recovered except for C19. The station number started at C13, since stations C1–C12, which were northeast of the LS in the Pacific Ocean, were not used in the study. CPIES is an inverted echo sounder (IES) attached to the seafloor equipped with an Aanderaa Doppler current sensor positioned approximately 50 m above the IES.
Map of the study region in the South China Sea. Current and pressure-recording inverted echo sounder (CPIES) sites are indicated. The circle at each site indicates the data coverage of the near-bottom current (black), round-trip acoustic travel time (τ; red), and bottom pressure (blue). Depth in meters (text) and CPIES observed average currents (cyan arrows) are also shown.
Citation: Journal of Physical Oceanography 52, 9; 10.1175/JPO-D-21-0284.1
Fundamental ocean properties, including round-trip acoustic travel time (τ), bottom pressure (Pbot), and near-bottom current (u, υ), were collected during the CPIES operation. The IES transmitted a 12.0-kHz pulse and subsequently recorded τ 24 times each hour. The standard deviation of a 24-ping sample, which followed a Rayleigh-shaped distribution, was typically less than 2.2 ms. The Pbot was recorded every 10 min with an absolute accuracy of ±0.6 dbar (1 dbar = 10 kPa). Zonal (u) and meridional (υ) components of near-bottom current vectors were observed at hourly intervals with an absolute accuracy of ±0.15 cm s−1. Most CPIESs recorded data covering the entire observational period; however, data loss occurred at C24, C32, and C34 (Fig. 1). Following Kennelly et al. (2007), all records measured by CPIESs were preprocessed (e.g., despiked, detided, dedrifted) and resampled hourly. Finally, all records were 72-h low-pass filtered with a fourth-order Butterworth filter and resampled at 12-h intervals.
The CPIES observations were further used to map the abyssal currents in the study region in three dimensions. The geostrophic current fields were estimated based on τ; the barotropic current fields in the deep ocean were estimated from Pbot and near-bottom currents.
b. Gravest empirical mode
The gravest empirical mode (GEM) method was applied to estimate temperature, salinity, and specific volume anomaly (δ) profiles based on a two-dimensional (τ and pressure) lookup table from historical hydrocast data (Watts et al. 2001a). To further estimate the geostrophic current, 635 conductivity–temperature–depth (CTD) profiles and 961 Argo profiles in the northeastern SCS (Fig. 2) were used to establish the GEM field for δ (Fig. 3a). The error between the GEM-estimated section and the CTD/Argo section was 3.19 × 10−7 m3 kg−1 in the upper 300 m and 5.18 × 10−8 m3 kg−1 for the full depth (Fig. 3b). Moreover, the seasonal model for the upper ocean was established for τ and δ utilizing historical profiles (Watts et al. 2001a).
Spatial and temporal distribution of conductivity–temperature–depth (CTD) and Argo profiles. (a) Spatial, (b) depth, (c) annual, and (d) monthly distributions.
Citation: Journal of Physical Oceanography 52, 9; 10.1175/JPO-D-21-0284.1
(a) Specific volume anomaly (δ) gravest empirical mode (GEM) and (b) δ GEM root-mean-square error (RMSE).
Citation: Journal of Physical Oceanography 52, 9; 10.1175/JPO-D-21-0284.1
Following the procedures of Donohue et al. (2010), the contribution of mass-loading-induced pathlength was removed, and τ was converted to a latitudinal-independent dynamic τ (τdyn). Moreover, the seasonal cycle was removed from τdyn, and CTDs taken near CPIES sites during the study period were used for calibration. The offset was considered as the average value if there were multiple calibration casts. The τ from the calibration casts (τcast) were converted to τ at the IES depth (τies) based on historical empirical relationships for when the casts were shallower than the IES. Here, we only used casts with a depth difference less than 600 m from the IES. Finally, using the empirical relationship, τdyn was converted to τindex, which was further applied to the GEM lookup table. Here, τindex was chosen as the round-trip travel time between 0 and 1200 dbar (τ1200) to capture the pycnocline and retain a large number of historical hydrographic profiles. Considering that deep profiles were scarce, and the deep vertical gradient was weak, travel time below 3700 dbar was assumed to relate only to pathlength. Under this assumption, τ from IES deeper than 3700 dbar were first converted to τ at 3700 dbar (τ3700) using the averaged sound speed profile.
The offset calibration and τ1200 conversion of C24 are shown as representative examples (Fig. 4). The recorded CTD profile reached 3214 dbar, 78 dbar shallower than the IES, and the τ at 3214 dbar from the profile was converted to τ at the depth of C24 (pentagram in Fig. 4b) using the empirical relationship shown in Fig. 4a. The offset was then calculated, and the τ series was calibrated (Fig. 4b). Finally, the calibrated τ series was converted to τ1200 based on the polynomial curve fit between τies and τ1200 from the hydrographic profiles (Figs. 4c and 4d). The root-mean-square error (RMSE) between τies and τ1200 varied from 0.1274 to 0.2405 ms among sites, with an average value of 0.1933 ms. The series in Fig. 4d was used for further study using the GEM method.
The offset calibration and τ1200 conversion process of C24. (a) Polynomial curve fit (black line) between τcast and τies from hydrographic profiles (gray dots). (b) τies before (gray) and after (black) calibration. The pentagram indicates the τ used for calibration. (c) Polynomial curve fit (black line) between τies and τ1200 from hydrographic profiles (gray dots). (d) Time series of τ1200. RMSEs are also indicated in (a) and (c).
Citation: Journal of Physical Oceanography 52, 9; 10.1175/JPO-D-21-0284.1
c. Mapping τ1200 and geostrophic shear
Maps of τ1200 were produced based on τ1200 at mooring sites using the optimal interpolation (OI) method (Bretherton et al. 1976). The noise-to-signal ratio (Vn), which is determined by the ratio of the square of the total error and the record variance, is required for the OI method. The errors from 1) observation uncertainty in 72-h low-pass-filtered record was (2.2 ms) × (24 × 72)−1/2 = 0.05 ms, 2) mass loading was ∼0.07 ms, 3) the conversion to dynamic τ was ∼0.03 ms, 4) seasonal model was ∼0.28 ms, 5) the conversion to τ1200 was ∼0.19 ms, and 6) the calibration was ∼0.03 ms. The total error was ∼0.35 ms, the average variance of τ1200 among sites was 0.82 ms2, and Vn = 0.15. The correlation length used in the OI method was 105 km, as determined by least squares fitting a Gaussian function to the τ1200 correlation between site pairs (Fig. 5a). The mapping error also resulted from the OI method, with the mapping errors typically less than 12% when using τ1200 records at 27 sites (Fig. 5b). The raw and mapped τ1200 at C24 is displayed as an example to examine the credibility of the OI method (Fig. 5c). The mapped (red) and raw (black) records agreed well within 0.262 ms. It should be noted that even if the raw record at C24 was not used in the OI method, the mapped (blue) and raw (black) records agreed within 0.437 ms, indicating that the OI method works well.
(a) Gaussian correlation function determined by the correlation coefficient between τ1200 pairs. Gray dots are the averaged correlation coefficient in 20-km bins, with error bars showing standard deviations. The Gaussian correlation function fit from gray dots is indicated by a black line. (b) Contour map showing mapping error (%) when using τ1200 records at current and pressure-recording inverted echo sounder (CPIES) locations (red dots). (c) Comparison between raw and mapped τ1200 at C24. The black, red, and blue lines indicate the raw data, mapped τ1200 based on all records, and mapped τ1200 based on all records except C24, respectively. RMSE between black and either red or blue data is indicated by red and blue text.
Citation: Journal of Physical Oceanography 52, 9; 10.1175/JPO-D-21-0284.1
d. Mapping abyssal current
Multivariate OI was used to obtain the deep velocity field at the reference level by constraining Pbot, u, and υ records constrained under the geostrophic balance assumption (Watts et al. 2001b). As we focused on the abyssal current, only records from CPIESs deeper than 2500 dbar were used. In a previous study, Vn was determined by overlapping Pbot records from paired IESs deployed at one site (Donohue et al. 2010). This analysis was not possible here because of the lack of overlapping records from the same site; however, the error was estimated by the smallest RMSE of data between site pairs (0.004 dbar). The Vn (0.06) was determined by dividing the square of the smallest RMSE by the averaged variance of the pressure records, implying that Vn was overestimated as the error caused by the location difference was included. The common mode in Pbot, related to the barotropic response to large-scale atmospheric forcing, was first removed (Donohue et al. 2010). The Pbot values from different sites were remarkably similar, indicating a similar large-scale prominent fluctuation related to the common-mode (Fig. 6a). Such substantial variability can be removed by subtracting a referenced pressure record, chosen as the Pbot from one site (Watts et al. 2021). Here, the referenced pressure was chosen as Pbot at C34 (Fig. 6a) because of the presence of fewer small-scale distractions, such as fluctuations caused by topography. Subsequently, the correlation length used for current and pressure OI was 120 km, as determined from the Pbot correlation between site pairs with distances larger than 30 km (Fig. 6b).
(a) Pbot records at sites deeper than 2500 dbar (gray lines). Data for C34 are indicated by a black line. (b) Gaussian correlation function determined by the correlation coefficient between Pbot record pairs. Gray dots are the averaged correlation coefficient in 20-km bins, with error bars displaying the standard deviations. Gaussian correlation function fit from gray dots is indicated by a black line.
Citation: Journal of Physical Oceanography 52, 9; 10.1175/JPO-D-21-0284.1
The abyssal current discussed in this study was at depths below 2500 dbar. Data from C16 and C33 were not used because they may be influenced by regional topography, as C16 was located near the western gap of the LS, and C33 was located on the northeast flank of a seamount. Considering the complex topography of the study region, three reference levels were chosen at 3700, 3300, and 2800 dbar to meet the CPIES deployed at different depths. Observations of Pbot, u, and υ at C17, C18, C26, C27, C34, C35, and C40 were combined using the multivariate OI method to map the velocity field at 3700 dbar under the assumption of weak deep thermal wind shear. As introduced in the previous section, geostrophic shears above the reference level were determined using the GEM method. Absolute velocities were considered the sum of the referenced velocity field and geostrophic shears above the reference level. Mapped absolute velocities at 3300 dbar from C17, C18, C26, C27, C34, C35, and C40 along with observed Pbot, u, and υ from C22–C25 and C28 were used to create the referenced velocity field at 3300 dbar. Similarly, the referenced velocity field at 2800 dbar was created using Pbot, u, and υ from C15, C21, C30–C32, and C39 combined with the mapped absolute velocities from other sites. The velocity mapping error at each reference level using the multivariate OI method was less than 15%. According to the comparison between the current meter observed and mapped records (Fig. 7), although mapped velocities (red) were smoother than raw velocities (black) as the OI method is a smoothing process (Watts et al. 2001a,b), records matched well before and after OI. The mapping results, including (red) and excluding (green) observed data from C24 during the mapping process, are compared in Fig. 8. Both spatial structure (Fig. 8a) and time series (Figs. 8b and 8c) of velocities at C24 matched well with an RMSE of less than 0.4 cm s−1, indicating that the OI method is effective in the array.
Mapped (red) and observed (black) current meter velocity records from June 2018 to July 2019. The white lines are zero lines.
Citation: Journal of Physical Oceanography 52, 9; 10.1175/JPO-D-21-0284.1
Comparison between mapping results, including (red) and excluding (green) observed data from C24 during the mapping process. (a) Temporal averaged spatial velocity structure at 2500 dbar. The data from sites indicated by white dots were used for mapping. (b) Time series of zonal component of velocity (u) at C24 at different depths. (c) Time series of meridional component of velocity (υ) at C24 at different depths.
Citation: Journal of Physical Oceanography 52, 9; 10.1175/JPO-D-21-0284.1
3. Results
a. Cyclonic circulation west of the LZ
The directly observed near-bottom averaged currents from current meters below 2500 dbar showed robust cyclonic abyssal circulation west of the LS (cyan arrows in Fig. 1). Sites near the eastern boundary showed northward velocities, whereas those near the western boundary showed southward velocities. Histograms of the velocity components at sites deeper than 2500 dbar display the probability distribution of the current (Fig. 9). The meridional component (υ) shows positively and negatively skewed distributions near the eastern and western boundaries, respectively. The zonal component (u) showed a positively skewed distribution in the southeast and generally showed a negatively skewed distribution in the north. The combination of these indicates cyclonic circulation. The medians of both u and υ are located from −2 to +2 cm s−1 at most sites; however, sites near the steep boundary (i.e., C28, C18, C17, C35, C39) showed greater velocities (Fig. 9). The near-bottom velocities exhibit maximum values at C28 (3.28 ± 1.84, 3.38 ± 2.26) cm s−1, C18 (1.58 ± 1.45, 2.77 ± 1.79) cm s−1, C17 (−0.33 ± 1.21, 2.42 ± 2.64) cm s−1, C35 (−1.74 ± 4.85, −2.16 ± 2.38) cm s−1, and C39 (−3.50 ± 2.38, −1.37 ± 1.78) cm s−1. During the entire observation period, the percentages of northward current along the eastern boundary totaled 93.9%, 96.0%, and 83.2% at C28, C18, and C17, respectively, and the percentages of southward current along the western boundary totaled 80.5% and 84.6% at C35 and C39, respectively (Fig. 7). Although near-bottom currents at other sites with subdued topography showed weaker continuity during the study period, the cyclonic circulation was conspicuous (Fig. 1).
Histograms of velocity components. Blue and red bars represent zonal (u) and meridional (υ) velocities, respectively.
Citation: Journal of Physical Oceanography 52, 9; 10.1175/JPO-D-21-0284.1
The abyssal current field west of the LS was identified by combining the GEM method and the multivariate OI method. For both the current meter observed (black) and mapped (red) results, the yearly averaged abyssal current from July 2018 to June 2019 showed obvious cyclonic circulation from 2500 dbar to the seafloor (Fig. 10). The current showed an average velocity greater than 4.0 cm s−1 near the eastern boundary above 3000 dbar, and the largest averaged velocity reached 6.4 cm s−1 near C28. The current became weaker in the deeper ocean. The NPDW penetrated the deep SCS through the western gaps of the LS and then flowed northward. When reaching the northern boundary, the current turned anticlockwise and flowed southwestward along the western boundary. The abyssal current was strong and narrow near the steep eastern boundary with a width of ∼70 km at 2500 dbar (blue lines in Fig. 11); however, it became weaker and wider (∼180 km) near the subdued western boundary (purple lines in Fig. 11). The yearly averaged velocity at 2500 dbar across the blue line along the eastern boundary was ∼2.3 cm s−1, while that across the purple line along the western boundary was ∼1.5 cm s−1. The yearly averaged volume transport below 2500 dbar reached ∼1.21 ± 0.93 Sv along the eastern boundary (across the blue lines in Fig. 11) and ∼1.59 ± 0.95 Sv along the western boundary (across the purple lines in Fig. 11).
Three-dimensional view of yearly averaged abyssal current from July 2018 to June 2019. Current meter observed (black) and mapped (red) results are indicated. Observed values are moved to the nearest layers (2500, 2800, 3100, 3400, or 3700 dbar) above the current meter.
Citation: Journal of Physical Oceanography 52, 9; 10.1175/JPO-D-21-0284.1
(a)–(l) Monthly averaged abyssal currents from July 2018 to June 2019 at 2500 dbar. Black dots indicate sites used for mapping the abyssal current. Note that the current in (a) is from June 2019, while currents in (b)–(l) are from July 2018–May 2019, respectively. Blue and purple bars indicate the total transported volumes across the northeastward and southwestward sections, respectively. The sections are represented by the thinner blue and purple lines oblique to the current direction. Cyan arrows are the averaged velocity components across the lines. Note that cyan and red arrows use different scales. (m) Volume transported across the blue and purple sections below 2500 dbar.
Citation: Journal of Physical Oceanography 52, 9; 10.1175/JPO-D-21-0284.1
Although the water exchange at the northernmost gap is not included in our representation of the western boundary, this gap is shallower than 2500 dbar and the volume transport there is small. According to previous observations, the volume transport intrusion from the LS and in the interior of the SCS were ∼1.18 ± 0.10 and ∼1.65 ± 0.48 Sv, respectively (Ye et al. 2019; Zhou et al. 2017), which is comparable with our measurements. The volume transport along the eastern boundary was less than that along the western boundary, implying that the DWBC is not only from the deep-water overflow but is also supplied by the interior of the SCS (Zhou et al. 2017).
b. Variability of abyssal circulation
Monthly patterns in abyssal circulation from July 2018 to June 2019 at 2500 dbar are shown in Fig. 11. Although monthly cyclonic circulation was identified, its strength and structure varied. The narrow current along the eastern boundary was strong in autumn and early winter, with the monthly maximum value of 8.1 cm s−1 reached in December near C28. On a spatial average, the monthly current at 2500 dbar near the eastern boundary (across the blue line) reached its seasonal maximum of 3.4 cm s−1 in December, and it was weakest in February (1.2 cm s−1) and May (1.3 cm s−1). The southwest current across the purple lines at 2500 dbar was strongest in September (2.0 cm s−1) and March (2.6 cm s−1). In addition, the entire cyclonic circulation was extremely weak in February, and the southwestward current was weakest in August.
The near-bottom current meters along the eastern boundary (C17, C18, C28) recorded strong and stable northward currents; however, C16 showed a stable southwestward current due to the surrounding topography (Fig. 12a). Substantial seasonal variations were identified from monthly averaged velocities. In general, the current along the eastern boundary was stronger in late autumn and early winter (November–January) but weaker in late winter and spring (February–May) (Figs. 12b–e). This corresponded with seasonal variations in the deep-water overflow measured in the Luzon Trough and western gaps of the LS (Zhou et al. 2014; Ye et al. 2019), suggesting that the deep-water overflow has a large impact on seasonal variations along the eastern boundary. When the boundary current was weak, the current direction at C16 turned much more westward, especially in March. The current reached the seasonal maximum in September–December at C18 (>4 cm s−1), in October–January at C17 (>4 cm s−1), but in December–January at C16 (>6 cm s−1). Similarly, the current reached the seasonal minimum in January–May at C18, in February–June at C17, but in March–April at C16. This temporal deviation implies the propagation of seasonal variations.
(a) Yearly averaged near-bottom current (red arrows) near the eastern boundary. The gray contour lines from light to dark indicate 1900-, 2200-, and 2500-m isobaths. Monthly averaged near-bottom velocities at (b) C16, (c) C17, (d) C18, and (e) C28. Tail orientations show current directions, with length showing average current velocity. Gray and cyan shading indicate the periods with stronger and weaker currents, respectively.
Citation: Journal of Physical Oceanography 52, 9; 10.1175/JPO-D-21-0284.1
The monthly volume transports along the eastern (across the blue lines in Fig. 11) and western boundaries (across the purple lines in Fig. 11) were calculated. Volume transport along the eastern boundary reached its maximum in late autumn and early winter and reached its minimum in late winter and spring (Fig. 11), controlled by seasonal variations of deep-water overflow. When ignoring the abrupt value in August, September, and March, the volume transports along the western boundary showed a trend of larger values in May–September (>1.6 Sv, except in August) and smaller values in January–April (<1.5 Sv, except in March) (Fig. 11m). This is different from the seasonal variations of deep-water overflow and current near the eastern boundary but is similar to seasonal variations in the deep basin of the SCS (Zheng et al. 2021a), implying that the seasonal variations of abyssal currents in the interior of the SCS do not synchronously respond to those of deep-water overflow (Zhu et al. 2017b).
The asynchronous pattern is attributed to the propagation of seasonal variations along boundary sites. The currents at C17, C21, C25, C28, and C35 are shown in Fig. 13. All five sites showed current along the boundary, indicating cyclonic circulation following the topography (Fig. 13a). According to the monthly averaged value, the current reaches its maximum in August–December at C28 (Fig. 13b), in October–January at C17 (Fig. 13c), in January–March at C21 (Fig. 13d), in May–September at C25 (Fig. 13e), and in June–November at C35 (Fig. 13f). The maximum velocity at C21 lagged that at C17 by ∼2.5 months and corresponded to a propagation velocity of ∼2.8 cm s−1. In contrast, the maximum velocity at C25 lagged that at C21 by ∼5 months, which corresponded to a propagation velocity of ∼1.4 cm s−1. These values were close to the average abyssal current velocities shown in Fig. 13a.
(a) Yearly averaged near-bottom current (red arrows) at C17, C21, C25, C28, and C35. The gray contour lines from light to dark indicate 1900-, 2200-, and 2500-m isobaths. Monthly averaged near-bottom velocities at (b) C28, (c) C17, (d) C21, (e) C25, and (f) C35. Gray shading shows the periods with strongest consecutive currents.
Citation: Journal of Physical Oceanography 52, 9; 10.1175/JPO-D-21-0284.1
4. Discussion
Although the abyssal current west of LS shows conspicuous cyclonic circulation on a monthly average, it was extremely weak in August 2018 and February 2019 but strengthened in the following months of September 2018 and March 2019 (Fig. 11). Previous studies suggest that upper-ocean perturbations in the SCS, such as eddies, could reach the deep ocean (Shu et al. 2018; Sun et al. 2020; Zhang et al. 2013; Zhang et al. 2016). Deep-water overflow variations and deep eddies may contribute to the abyssal current variations in the SCS (Zhou et al. 2020).
The current weakness along the eastern boundary in February was related to the deep-water overflow and is supported by previous studies (Zhou et al. 2014; Ye et al. 2019). However, the cause of the weak current along the western boundary in August 2018 and February 2019 remains unknown. Weak abyssal currents in August 2018 were also recorded by the mooring used by Zhou et al. (2020). To diagnose whether such weakness was related to upper-ocean perturbations, the local surface relative vorticity was determined using surface geostrophic current data from the Copernicus Marine Environment Monitoring Service (CMEMS). An extremely weak southwestward abyssal current along the western boundary occurred in August 2018 and February 2019 (purple line in Fig. 14a). Coincidentally, a strong surface relative vorticity was observed during a similar period, which may be related to the upper-ocean cyclonic eddy (gray shading in Fig. 14b). This observation suggests that the weak current along the western boundary in August 2018 and February 2019 may be related to the upper-ocean eddies.
(a) Time series of volume transport along the western (solid purple line) and eastern (dashed blue line) boundaries below 2500 dbar. The solid blue line is a 228-day offset of the dashed line to show the lag of the flow. (b) Time series of average surface relative vorticity (see magenta boxes in Figs. 15b and 15e). Gray shadings indicate August 2018 and February 2019 when the volume transport along the western boundary was extremely weak. Horizontal dashed lines represent the mean plus and minus the standard deviation.
Citation: Journal of Physical Oceanography 52, 9; 10.1175/JPO-D-21-0284.1
The maps of two weak western boundary abyssal currents from 20 August 2018 and 16 February 2019 are displayed in Fig. 15 to identify the detailed structure of the current patterns. In August, a cyclonic eddy appeared in the deep ocean (magenta box in Fig. 15a), with the southeastern part of the deep eddy causing an inverted abyssal current, decreasing the southwestward volume transport and the monthly averaged boundary current in August 2018 (Fig. 11). A similar situation occurred in February 2019, as shown in Fig. 15d. During these two periods, strong cyclonic eddies were also observed in the upper ocean, northwest of Luzon Island, according to surface geostrophic current data obtained from CMEMS (Figs. 15b and 15e), implying that these two surface eddies might stretch to the ocean bottom and influence abyssal currents. The eddies show a northwestward vertical tilt with increasing depth in the deep ocean, possibly related to the sloping regional topography (Zhang et al. 2016). During these two events, anticyclonic and cyclonic eddy pairs existed in the northern SCS (Figs. 15b and 15e). The region southwest of Taiwan shows abundant mesoscale eddies shed from the Kuroshio loop, originating from the North Pacific, or locally generated by wind (Zhang et al. 2013). Anticyclonic eddies are frequently generated there in autumn during the monsoon transition (Zu et al. 2013). However, anticyclonic eddies were not observed in the deep ocean because they were located north, outside the array.
(a) Currents (arrows) and pressure work (color) at 2500 dbar on 20 Aug 2018. (b) Surface geostrophic currents (arrows) and absolute dynamic topography (color) from CMEMS on 20 Aug 2018. (c) Time series of average pressure work for the purple line in (a). (d)–(f) As in (a)–(c), but for 16 Feb 2019. Magenta boxes indicate cyclonic eddies. Gray shadings indicate the same period as those in Fig. 14. Pressure work is integrated from 3700 to 2500 dbar.
Citation: Journal of Physical Oceanography 52, 9; 10.1175/JPO-D-21-0284.1
Variations in the deep-water overflow are another candidate for variations in the SCS abyssal current. Although the time series of volume transport along the western and eastern boundaries are uncorrelated in real time, the volume transport along the western boundary matched well with that along the eastern boundary with a 228-day time lag (Fig. 14a). This is because abyssal water took ∼7.5 months to travel from the LS to the interior of the SCS (as shown in Fig. 13, from C17 to C21 to C25). The correlation coefficient between the purple and blue solid lines in Fig. 14a was 0.67. This indicates that the remarkably weak (strong) southwest current in February (March) can be directly related to the weak (strong) current upstream in late June and early July (late July and early August) near the LS, which may further result from the LS overflow. However, the limited time span of the study meant that the extreme values of volume transport along the western boundary in August and September could not be tracked back to variations near the LS.
Two possible explanations for the weak southwestward current along the western boundary in August 2018 and February 2019 are provided above. The volume transport along the western boundary is remarkably similar to that along the eastern boundary, with a time lag of ∼7.5 months. Meanwhile, the weak southwestward current events in August 2018 and February 2019 may be related to the upper-ocean mesoscale eddies through pressure work. Although ΔQp for the purple line was negative during these two cases (Figs. 15c and 15f), the time series of ΔQp, and volume transport did not match well at other times. For instance, the remarkable negative value in March 2019 (Fig. 15f), possibly related to the strong surface relative vorticity (Fig. 14b), did not result in a weak current along the western boundary (Fig. 14a). The weak volume transport in January 2019 (Fig. 14a) did not correspond to an abnormal surface relative vorticity (Fig. 14b). Consequently, although upper-ocean perturbations may have contributed, the variation in the deep-water overflow was considered the dominant factor for the weak current along the western boundary in August 2018 and February 2019.
5. Summary
Based on a CPIES array deployed in the northern SCS near the LS for over 400 days, which provided unprecedented coverage of abyssal circulation observations in the northeast SCS, the detailed three-dimensional structure and the spatiotemporal variations of the conspicuous cyclonic circulation west of the LS have been observed for the first time. The NPDW flows northward along the boundary after entering the SCS through the western gap of the LS and then turns anticlockwise following the topography, forming cyclonic circulation. The yearly averaged current along the steep eastern boundary was narrow and strong (∼70 km, ∼2.3 cm s−1 at 2500 dbar), while that along the subdued western boundary was wide and weak (∼180 km, ∼1.5 cm s−1 at 2500 dbar), and the maximum current (∼6.4 cm s−1) appeared near C28 close to the gap of the LS. The volume transport below 2500 dbar along the eastern and western boundaries was ∼1.21 ± 0.93 and ∼1.59 ± 0.95 Sv, respectively, indicating that the water of DWBC originated from both the deep-water overflow and the interior ocean. Asynchronous seasonal variation was observed near the LS and interior SCS. The abyssal current near the LS was stronger in late autumn and early winter but weaker in late winter and spring, following the seasonal variation of the deep-water overflow. However, the southwestward current in the interior SCS was strongest in summer and early autumn and weakest in late winter and early spring, which lags the seasonal variation near the eastern boundary ∼7–11 months. This can be attributed to the propagation of seasonal variations along the boundary. The weak southwestward current along the western boundary in August 2018 and February 2019 was seemingly dominated by variations in the deep-water overflow. However, cyclonic eddies west of the LS may also contribute to this variation. Detailed upper-ocean circulation in this region using CPIES data will be presented in a subsequent study.
This study provides important observational evidence for the existence of SCS abyssal cyclonic circulation and has strong implications for the research of the current system in the SCS. Knowledge of abyssal current patterns could help understand water renewal, energy budget, and sedimentary processes in the deep ocean and track the transport of dissolved elements, minerals, or pollutants. As an important part of the SCS throughflow, the study of abyssal current in the SCS may also contribute to understanding the Indonesian Throughflow and global climate change.
Acknowledgments.
The School of Oceanography, Shanghai Jiao Tong University, State Key Laboratory of Satellite Ocean Environment Dynamics, Second Institute of Oceanography, Ministry of Natural Resources, and the Southern Marine Science and Engineering Guangdong Laboratory (Zhuhai) contributed equally to the work and should be regarded as co-first institutional affiliations. This study was sponsored by the National Natural Science Foundation of China (Grants 41920104006 and 41906024), the Scientific Research Fund of Second Institute of Oceanography, MNR (Grants JZ2001 and QNYC2102), the Project of State Key Laboratory of Satellite Ocean Environment Dynamics, Second Institute of Oceanography (SOEDZZ2106 and SOEDZZ2207), the Oceanic Interdisciplinary Program of Shanghai Jiao Tong University (Project SL2021MS021), the Innovation Group Project of Southern Marine Science and Engineering Guangdong Laboratory (Zhuhai) (311020004), and the Global Climate Changes and Air-Sea Interaction Program (GASI-02-PAC-ST-Wwin).
Data availability statement.
Bathymetry was obtained from ETOPO1 (doi:10.7289/V5C8276M). Surface geostrophic currents and absolute dynamic topography were obtained from the CMEMS (http://marine.copernicus.eu/). Access to the analyzed moored data should contact the corresponding author. Argo profiles were obtained from ftp://ftp.argo.org.cn/pub/ARGO/global/.
REFERENCES
Alford, M. H., and Coauthors, 2011: Energy flux and dissipation in Luzon strait: Two tales of two ridges. J. Phys. Oceanogr., 41, 2211–2222, https://doi.org/10.1175/JPO-D-11-073.1.
Bretherton, F. P., R. E. Davis, and C. B. Fandry, 1976: A technique for objective analysis and design of oceanographic experiments applied to MODE-73. Deep-Sea Res. Oceanogr. Abstr., 23, 559–582, https://doi.org/10.1016/0011-7471(76)90001-2.
Cai, Z., and J. Gan, 2019: Coupled external‐internal dynamics of layered circulation in the South China Sea: A modeling study. J. Geophys. Res. Oceans, 124, 5039–5053, https://doi.org/10.1029/2019JC014962.
Chang, Y.-T., W.-L. Hsu, J.-H. Tai, T.-Y. Tang, M.-H. Chang, and S.-Y. Chao, 2010: Cold deep water in the South China Sea. J. Oceanogr., 66, 183–190, https://doi.org/10.1007/s10872-010-0016-x.
Chen, H., X. Xie, D. Van Rooij, T. Vandorpe, L. Huang, L. Guo, and M. Su, 2013: Depositional characteristics and spatial distribution of deep-water sedimentary systems on the northwestern middle-lower slope of the northwest sub-basin, South China Sea. Mar. Geophys. Res., 34, 239–257, https://doi.org/10.1007/s11001-013-9191-7.
Donohue, K. A., D. R. Watts, K. L. Tracey, A. D. Greene, and M. Kennelly, 2010: Mapping circulation in the Kuroshio extension with an array of current and pressure recording inverted echo sounders. J. Atmos. Oceanic Technol., 27, 507–527, https://doi.org/10.1175/2009JTECHO686.1.
Gan, J., Z. Liu, and C. R. Hui, 2016: A three-layer alternating spinning circulation in the South China Sea. J. Phys. Oceanogr., 46, 2309–2315, https://doi.org/10.1175/JPO-D-16-0044.1.
Kennelly, M. A., K. L. Tracey, and D. R. Watts, 2007: Inverted echo sounder data processing manual. University of Rhode Island GSO Tech. Rep. 2007-02, 89 pp.
Lan, J., N. Zhang, and Y. Wang, 2013: On the dynamics of the South China Sea deep circulation. J. Geophys. Res. Oceans, 118, 1206–1210, https://doi.org/10.1002/jgrc.20104.
Lan, J., Y. Wang, F. Cui, and N. Zhang, 2015: Seasonal variation in the South China Sea deep circulation. J. Geophys. Res. Oceans, 120, 1682–1690, https://doi.org/10.1002/2014JC010413.
Li, L., and T. Qu, 2006: Thermohaline circulation in the deep South China Sea basin inferred from oxygen distributions. J. Geophys. Res., 111, C05017, https://doi.org/10.1029/2005JC003164.
Liu, C.-T., and R.-J. Liu, 1988: The deep current in the Bashi Channel. Acta Oceanogr. Taiwan., 20, 107–116.
Liu, Z., and Coauthors, 2010: Clay mineral distribution in surface sediments of the northeastern South China Sea and surrounding fluvial drainage basins: Source and transport. Mar. Geol., 277, 48–60, https://doi.org/10.1016/j.margeo.2010.08.010.
Lüdmann, T., H. K. Wong, and K. Berglar, 2005: Upward flow of North Pacific Deep Water in the northern South China Sea as deduced from the occurrence of drift sediments. Geophys. Res. Lett., 32, L05614, https://doi.org/10.1029/2004GL021967.
Oey, L.-Y., 2008: Loop Current and deep eddies. J. Phys. Oceanogr., 38, 1426–1449, https://doi.org/10.1175/2007jpo3818.1.
Qu, T., 2002: Evidence for water exchange between the South China Sea and the Pacific Ocean through the Luzon Strait. Acta Oceanol. Sin., 21, 175–185.
Qu, T., J. B. Girton, and J. A. Whitehead, 2006a: Deepwater overflow through Luzon Strait. J. Geophys. Res., 111, C01002, https://doi.org/10.1029/2005JC003139.
Qu, T., Y. Du, and H. Sasaki, 2006b: South China Sea throughflow: A heat and freshwater conveyor. Geophys. Res. Lett., 33, L23617, https://doi.org/10.1029/2006GL028350.
Qu, T., Y. T. Song, and T. Yamagata, 2009: An introduction to the South China Sea throughflow: Its dynamics, variability, and application for climate. Dyn. Atmos. Oceans, 47, 3–14, https://doi.org/10.1016/j.dynatmoce.2008.05.001.
Quan, Q., Z. Cai, G. Jin, and Z. Liu, 2021a: Topographic rossby waves in the abyssal south china sea. J. Phys. Oceanogr., 51, 1795–1812, https://doi.org/10.1175/JPO-D-20-0187.1.
Quan, Q., and Coauthors, 2021b: Influence of the Kuroshio intrusion on deep flow intraseasonal variability in the northern South China Sea. J. Geophys. Res. Oceans, 126, e2021JC017429, https://doi.org/10.1029/2021JC017429.
Shao, L., and Coauthors, 2007: Deep water bottom current deposition in the northern South China Sea. Sci. China, 50D, 1060–1066, https://doi.org/10.1007/s11430-007-0015-y.
Shu, Y., and Coauthors, 2016: Persistent and energetic bottom-trapped topographic rossby waves observed in the southern South China Sea. Sci. Rep., 6, 24338, https://doi.org/10.1038/srep24338.
Shu, Y., and Coauthors, 2022: Deep-current intraseasonal variability interpreted as topographic rossby waves and deep eddies in the xisha islands of the South China Sea. J. Phys. Oceanogr., 52, 1415–1430, https://doi.org/10.1175/JPO-D-21-0147.1.
Shu, Y., H. Xue, D. Wang, F. Chai, Q. Xie, J. Yao, and J. Xiao, 2014: Meridional overturning circulation in the South China Sea envisioned from the high-resolution global reanalysis data GLBa0.08. J. Geophys. Res. Oceans, 119, 3012–3028, https://doi.org/10.1002/2013JC009583.
Shu, Y., J. Chen, S. Li, Q. Wang, J. Yu, and D. Wang, 2018: Field-observation for an anticyclonic mesoscale eddy consisted of twelve gliders and sixty-two expendable probes in the northern South China Sea during summer 2017. Sci. China Earth Sci., 62, 451–458, https://doi.org/10.1007/s11430-018-9239-0.
Song, Y. T., 2006: Estimation of interbasin transport using ocean bottom pressure: Theory and model for Asian marginal seas. J. Geophys. Res., 111, C11S19, https://doi.org/10.1029/2005JC003189.
Stommel, H., and A. B. Arons, 1960a: On the abyssal circulation of the world ocean—I. Stationary planetary flow patterns on a sphere. Deep-Sea Res., 6, 140–154, https://doi.org/10.1016/0146-6313(59)90065-6.
Stommel, H., and A. B. Arons, 1960b: On the abyssal circulation of the world ocean—II. An idealized model of the circulation pattern and amplitude in oceanic basins. Deep-Sea Res., 6, 217–233, https://doi.org/10.1016/0146-6313(59)90075-9.
Sun, Z., and Coauthors, 2020: Three-dimensional structure and interannual variability of the kuroshio loop current in the Northeastern South China Sea. J. Phys. Oceanogr., 50, 2437–2455, https://doi.org/10.1175/JPO-D-20-0058.1.
Tian, J., Q. Yang, X. Liang, L. Xie, D. Hu, F. Wang, and T. Qu, 2006: Observation of Luzon Strait transport. Geophys. Res. Lett., 33, L19607, https://doi.org/10.1029/2006GL026272.
Tian, J., Q. Yang, and W. Zhao, 2009: Enhanced diapycnal mixing in the South China Sea. J. Phys. Oceanogr., 39, 3191–3203, https://doi.org/10.1175/2009JPO3899.1.
Wang, J., 1986: Observation of abyssal flows in the northern South China Sea. Acta Oceanogr. Taiwan., 16, 36–45.
Wang, Q., and Coauthors, 2019: Energetic topographic rossby waves in the northern South China Sea. J. Phys. Oceanogr., 49, 2697–2714, https://doi.org/10.1175/JPO-D-18-0247.1.
Wang, G., S.-P. Xie, T. Qu, and R. X. Huang, 2011: Deep South China Sea circulation. Geophys. Res. Lett., 38, L05601, https://doi.org/10.1029/2010GL046626.
Wang, G., R. X. Huang, J. Su, and D. Chen, 2012: The effects of thermohaline circulation on wind-driven circulation in the South China Sea. J. Phys. Oceanogr., 42, 2283–2296, https://doi.org/10.1175/JPO-D-11-0227.1.
Wang, J., and Coauthors, 2021: Observed variability of bottom‐trapped topographic Rossby waves along the slope of the northeastern South China Sea. J. Geophys. Res. Oceans, 126, e2021JC017746, https://doi.org/10.1029/2021JC017746.
Watts, D. R., C. Sun, and S. Rintoul, 2001a: A two dimensional gravest empirical mode determined from hydrographic observations in the subantarctic front. J. Phys. Oceanogr., 31, 2186–2209, https://doi.org/10.1175/1520-0485(2001)031<2186:ATDGEM>2.0.CO;2.
Watts, D. R., X. Qian, and K. L. Tracey, 2001b: Mapping abyssal current and pressure fields under the meandering Gulf Stream. J. Atmos. Oceanic Technol., 18, 1052–1067, https://doi.org/10.1175/1520-0426(2001)018<1052:MACAPF>2.0.CO;2.
Watts, D. R., M. Wei, K. L. Tracey, K. A. Donohue, and B. He, 2021: Seafloor geodetic pressure measurements to detect shallow slow slip events: Methods to remove contributions from ocean water. J. Geophys. Res. Solid Earth, 126, e2020JB020065, https://doi.org/10.1029/2020JB020065.
Xiao, J., and Coauthors, 2016: On the near-inertial variations of meridional overturning circulation in the South China Sea. Ocean Sci., 12, 335–344, https://doi.org/10.5194/os-12-335-2016.
Xie, Q., J. Xiao, D. Wang, and Y. Yu, 2013: Analysis of deep-layer and bottom circulations in the South China Sea based on eight quasi-global ocean model outputs. Chin. Sci. Bull., 58, 4000–4011, https://doi.org/10.1007/s11434-013-5791-5.
Yang, Q., J. Tian, and W. Zhao, 2010: Observation of Luzon Strait transport in summer 2007. Deep-Sea Res. I, 57, 670–676, https://doi.org/10.1016/j.dsr.2010.02.004.
Yang, Q., W. Zhao, X. Liang, and J. Tian, 2016: Three-dimensional distribution of turbulent mixing in the South China Sea. J. Phys. Oceanogr., 46, 769–788, https://doi.org/10.1175/JPO-D-14-0220.1.
Ye, R., and Coauthors, 2019: Variability in the deep overflow through the Heng-Chun ridge of the Luzon Strait. J. Phys. Oceanogr., 49, 811–825, https://doi.org/10.1175/JPO-D-18-0113.1.
Yuan, D., 2002: A numerical study of the South China Sea deep circulation and its relation to the Luzon Strait transport. Acta Oceanol. Sin., 21, 187–202.
Zhang, Z., W. Zhao, J. Tian, and X. Liang, 2013: A mesoscale eddy pair southwest of Taiwan and its influence on deep circulation. J. Geophys. Res. Oceans, 118, 6479–6494, https://doi.org/10.1002/2013JC008994.
Zhang, Z., J. Tian, B. Qiu, W. Zhao, P. Chang, D. Wu, and X. Wan, 2016: Observed 3D structure, generation, and dissipation of oceanic mesoscale eddies in the South China Sea. Sci. Rep., 6, 24349, https://doi.org/10.1038/srep24349.
Zhao, W., C. Zhou, J. Tian, Q. Yang, B. Wang, L. Xie, and T. Qu, 2014: Deep water circulation in the Luzon strait. J. Geophys. Res. Oceans, 119, 790–804, https://doi.org/10.1002/2013JC009587.
Zhao, X., C. Zhou, W. Zhao, J. Tian, and X. Xu, 2016: Deepwater overflow observed by three bottom-anchored moorings in the bashi channel. Deep-Sea Res. I, 110, 65–74, https://doi.org/10.1016/j.dsr.2016.01.007.
Zhao, X., C. Zhou, X. Xu, R. Ye, and W. Zhao, 2020: Deep circulation in the South China Sea simulated in a regional model. Ocean Dyn., 70, 1461–1473, https://doi.org/10.1007/s10236-020-01411-2.
Zheng, H., C. Zhang, R. Zhao, X. H. Zhu, Z. N. Zhu, Z. J. Liu, and M. Wang, 2021a: Structure and variability of abyssal current in northern South China Sea based on CPIES Observations. J. Geophys. Res. Oceans, 126, e2020JC016780, https://doi.org/10.1029/2020JC016780.
Zheng, H., X. H. Zhu, C. Zhang, R. Zhao, Z. N. Zhu, and Z. J. Liu, 2021b: Propagation of topographic rossby waves in the deep basin of the South China Sea based on abyssal current observations. J. Phys. Oceanogr., 51, 2783–2791, https://doi.org/10.1175/JPO-D-21-0051.1.
Zhou, C., W. Zhao, J. Tian, Q. Yang, and T. Qu, 2014: Variability of the deep-water overflow in the Luzon strait. J. Phys. Oceanogr., 44, 2972–2986, https://doi.org/10.1175/JPO-D-14-0113.1.
Zhou, C., W. Zhao, J. Tian, X. Zhao, Y. Zhu, Q. Yang, and T. Qu, 2017: Deep western boundary current in the South China Sea. Sci. Rep., 7, 9303, https://doi.org/10.1038/s41598-017-09436-2.
Zhou, M., G. Wang, W. Liu, and C. Chen, 2020: Variability of the observed deep western boundary current in the South China Sea. J. Phys. Oceanogr., 50, 2953–2963, https://doi.org/10.1175/JPO-D-20-0013.1.
Zhu, Y., J. Sun, Y. Wang, Z. Wei, D. Yang, and T. Qu, 2017a: Effect of potential vorticity flux on the circulation in the South China Sea. J. Geophys. Res. Oceans, 122, 6454–6469, https://doi.org/10.1002/2016JC012375.
Zhu, Y., J. Sun, Z. Wei, Y. Wang, G. Fang, and T. Qu, 2017b: A fresh look at the deepwater overflow in the Luzon strait. Acta Oceanol. Sin., 36, 1–8, https://doi.org/10.1007/s13131-017-1057-4.
Zu, T., D. Wang, C. Yan, I. Belkin, W. Zhuang, and J. Chen, 2013: Evolution of an anticyclonic eddy southwest of Taiwan. Ocean Dyn., 63, 519–531, https://doi.org/10.1007/s10236-013-0612-6.
