The satellite altimetry anomaly data between January 1, 1993 and December 31, 2016 in the SCS-CSR are analyzed by applying the Eulerian Eddy Detection and Tracking Method. There are 147 meso- and large-scale eddies identified inside the CSR, of which 70 and 77 of CEs and ACEs are found, respectively. The distributions, sizes and propagations of these eddies are estimated. It is revealed that eddies in the SCS are affected by Coriolis force, mean currents and topography, and propagate differently in different subdomains (Chen et al., 2011; Zheng et al., 2017).
The spatial distribution of the eddy generation is shown in Fig. 3. Almost all of the eddies in the CSR are generated between 17°N and 22°N in the SCS. There are quite a number of eddies originating from the vicinity of the Dongsha Islands (DS, 19.6°−21.3°N, 115.8°−118.2°E). The other important eddy generation region is located to the southwest of Taiwan (SWT, 19.7°−22°N, 118.3°−120.5°E) away from the margin of the CSR. In the SWT, both CEs and ACEs often coexist revealed by tracking eddy trajectories, and they are often deformed and stretched probably by eddy interaction.
Figure 3. The spatial distribution of locations where eddies are initially generated in the CSR. The two red boxes denote the region DS and SWT respectively. The triangle denotes the Dongsha Islands.
Eddy formation is very common in the vicinity of the DS region as mentioned above, which is consistent with Chow et al. (2008). There are always several eddies existing to the east of this region. Within the study period, a total of 51 eddies (about 35% of all) are counted, 27 of which are CEs and 24 of which are ACEs. In addition, the eddy generation does not show a clear seasonal cycle (Fig. 4). The results from the student t-test indicate that there is no significant discrepancy in eddy formation between the CEs and the ACEs.
Figure 4. Seasonal distributions of the number of eddies that are generated in the DS and subsequently enter the CSR.
Previous studies indicated that the SWT near the Luzon Strait is an essential place for eddy generation in the northern SCS. The ACEs are shed from the Kuroshio Current Loop (KCL) especially in winter, and almost all ACEs shedding from KCL are accompanied by a CE to the northeast (Nan et al., 2011b; Zhang et al., 2017). During the 24-year period of our census, there are 128 eddies generated in the SWT, including 58 CEs and 70 ACEs. Among these eddies, 13 CEs and 26 ACEs can climb up onto the CSR, which account for 22.4% and 37.1% of the total number of CEs and ACEs, respectively. The number of ACE is significantly higher than that of CE, suggesting that ACEs are more capable of moving onto the CSR than CEs. Results shown in Fig. 5 reveals that there is a remarkable seasonal variation in the eddy generation in the SWT. CEs move toward slope mostly in spring, while the ACEs mainly enter the CSR in winter. It is also revealed that the Kuroshio intrudes into the SCS through the Luzon Strait more frequently in winter than in other seasons, and the Kuroshio loop sheds more ACEs in the SWT during winter (Nan et al. 2011a). Thus, these results imply that the stronger Kuroshio intrusion might lead to more ACEs move onto the CSR in winter.
The relative vorticity is typically used to measure the strength of eddies. In this study, we define the vorticity value in the eddy center as the strength of the vortex following Dong (2015). To understand the spatial characteristics of eddy propagation, we average and then normalize the eddy relative vorticity into spatial grid bins (Fig. 6). The spatial patterns of eddy vorticity magnitudes have similar features: large values are found to the east of the Dongsha Islands and relatively small value are found in the middle and west parts of the CSR. This suggests that eddies lose their energy by friction as propagating westwards over the Dongsha Islands.
Moreover, the ACEs are relatively stronger than the CEs in the northeast part of the CSR, which is also mentioned in several previous studies (Fig. 6). For example, studies by both Zhang et al. (2013) and Huang et al. (2017b) indicated that southwest of Taiwan, the CE was weaker than the ACE in an eddy pair. This discrepancy may be caused by differences in the eddy formation mechanism, i.e. CEs are generated from barotropic instability of the northern branch of the KCL, while ACEs are shedding from the KCL and baroclinic instability, and barotropic instability provide energy for its generation and growth (Zhang et al., 2013, 2017).
Eddies will interact dramatically with the topography as moving over continental slope regions (Oey and Zhang, 2004). Many factors such as slope steepness, orientation and even the barotropicity and baroclinicity of water column will significantly influence the vortex motion (Jacob et al., 2002). Regardless of polarity, eddies in the study area have a self-induced westward moving tendency, which can be deduced by β-effect and conventional theories like potential vorticity conservation (Cushman-Roisin et al., 1990). Nevertheless, besides the pure westward motion in the global ocean, eddies also slightly migrate in the meridional direction based on global observed data (Chelton et al., 2007). In terms of the polarity, the CEs and ACEs have poleward and equatorward deflection respectively, which can be adequately explained by potential vorticity conservation as follows: in the northern hemisphere, assuming a β-plane, a relative large eddy could transport water parcels in the meridional direction, and secondary eddies will germinate on the flank. Then the rotation of secondary eddies in turn drive cyclone and anticyclone circulations that move poleward and equatorward respectively (Morrow et al., 2004).
To better understand the eddy propagation properties, the eddy propagation velocities along the trajectories were computed, and then averaged onto 0.25°×0.25° grid bins inside the CSR. Those bins that contain no more than 3 observations were removed for statistical robustness. Most eddies move southwestward along the continental slope. The mean zonal and meridional propagation velocities of the CEs are −5.50 and −6.13 cm/s, and those of ACEs −2.98 and −2.93 cm/s, respectively, which is consistent with earlier studies in the similar region (Chen et al., 2011). If subtracting the speed of mean flow, the mean zonal and meridional velocities of the CEs are changed to −1.16 and 6.50 cm/s, and those of ACEs are changed to 3.92 and −2.99 cm/s, respectively. The CEs and ACEs propagate in opposite meridional directions, with the CEs moving to the shallow region and ACEs to the deep region. This difference can be explained by the conservation of potential vorticity when a CE or ACE loses its energy and rotates more slowly during its movement, the relative vorticity decreases or increases. So that the CE or ACE will move to a shallower or deeper region due to the PV constraint. In general, the ACEs move slightly faster than CEs, which could be properly explained by the effect of the baroclinicity in a two-layer β-plane model proposed by Cushman-Roisin et al. (1990), that is, the discrepancy in the response of water in lower layer to eddies of different polarity leads to the acceleration of ACEs, and deceleration of CEs. There is no straight evidence suggesting any correlation between the propagation speed of Rossby waves and eddies in the CSR. Results based on satellite altimeter data suggested that anticyclonic eddies generated off the northwest coast of Luzon move southwestward along the continental slope are not free long Rossby waves (Yuan et al. 2007). Moreover, compared with the zonal speed of the first baroclinic Rossby wave calculated by Wang et al. (2017), eddies seem to move faster than the Rossby wave on the continental slope.
The mean propagation velocity field of CEs and ACEs in the CSR are elucidated in Fig. 7a and b, respectively. Both CEs and ACEs mainly translate southwestward steered by the topography of the slope. However, the motion directions of the ACEs are more concentrated to some extent, while CEs are inclined to move divergently, which is more intuitively reflected in Fig. 7c and d.
Figure 7. Mean eddy propagation velocities in the CSR. a and b are eddy velocity fields of CEs and ACEs, respectively; c and d are compass diagrams of CE and ACE propagation velocities, respectively.
Some eddies tend to move in and out of the CSR several times through their life cycle (not shown). However, these eddies will not stay for long (typically less than 3 days) if they leave the CSR and come back to the slope afterwards. Considering that some eddies only hover on the edge of the CSR or dissipate rapidly as soon as entering the CSR, the statistical standard is redefined as follows: the first day when the eddy appears in the interior of the CSR is set to be the starting point of our statistics, and only those eddies that stay inside the CSR longer than 7 days are considered in our statistical analysis. As a result, 103 eddies meet this standard, including 50 CEs and 53 ACEs. The statistics of the final destinations of those eddies are listed in Table 1. The CEs seem to have higher chance to leave the CSR because the number of CEs dissipated in the CSR is relatively low compared with that of ACEs. However, a number of those CEs just wander between the margin of the CSR and the deep basin (1 000−3 000 m).
Polarity Dissipating in the continental slope Leaving the continental slope Basinward (>2 000 m) Shelfward (<200 m) Others CE 26 (52%) 3 4 17 ACE 33 (62%) 5 6 9
Table 1. Statistics of the destination of eddies that stay more than 7 days in the CSR
The relationship between water depths of an eddy center and its life span by averaging all eddies of the same polarity is shown in Fig. 8a. Both ACEs and CEs cross isobaths onto shallow depth in the similar pattern for the first 6 days, and then move in significantly different patterns afterward. The depth variations of CEs are much larger than that of ACEs between 6 and 30 days when moving into the CSR These non-overlapping of confidence intervals indicate the significant difference in motion between these two types of eddies. Further, the depth variations of individual eddies indicate the similarity and differences between CEs and ACEs (Figs. 8b and c), that is, in the initial stage of 6 days, CEs generally spread across the isobaths in the similar range, between 6 and 30 days, the center depths of ACEs spread in a significant larger range than that of CEs, and in late stage longer than 30 days, more ACEs are inclined to move to the deeper basin while CEs continue to propagate along the edge of the CSR.
Figure 8. Water depth variations of eddy centers during propagation. Blue and red correspond to CE and ACE, respectively. In Fig. 8a, solid blue and red curves represent depth variations of synthetic eddy centers by averaging all eddies with the same polarity, the asterisk (*) represents the moment when the eddy leaves the CSR for the first time, and error bars the 95% confidence interval. In Figs 8b and c, blue and red curves represent eddy center depth variances of all CEs and all ACEs.
The seasonal forcing in the SCS is dominated by seasonal East Asian monsoon, which reverses the direction of the upper-layer circulation over the entire basin between seasons (Hu et al., 2000). In the CSR, the seasonal variation is reflected in the change in the strength of the along-slope current, i.e., the current is stronger in winter and weaker in summer (Fig. 9). During all seasons, the flow accelerates within the narrow slope corridor between 114.5°−115.5°E, which might be explained by the “venturi effect”-the expediting of the fluid when it flows through a constricted section of the terrain. Besides, in winter, the geostrophic currents in the CSR tend to bifurcate into two branches, one flowing to the region west of the Dongsha Islands and the other going down the slope, which could facilitate the cross-shelf transport in that area.
Figure 9. Seasonal patterns of geostrophic currents in the CSR in spring (a), summer (b), autumn (c), and winter (d).
To better understand the impact of mean flow on the eddy propagation in the CSR, we plot the eddy trajectories in each season (Fig. 10). Consistent with results shown in Fig. 9, the eddy propagation is in agreement with the along-slope current to a large degree. For example, in winter and autumn, eddies are propagating with the mean flow along the slope. In summer, eddies move randomly, and the trajectories diverge because of weaker wind and mean flows. More eddies tend to cross the mean flow and the isobaths so that cross-shelf transport is likely to take place in this season (He et al., 2016). In winter, eddies are prone to leave the CSR in the ambient region west of the Dongsha Islands in agreement with the downward slope current. Among those eddies, the ACEs are dominant both in quantity and propagation distance, and some of them can even reach south of 16°N.
The comparison in the direction and magnitude between the seasonal geostrophic current and eddy propagation velocities are shown in Fig. 11 and Table 2. The results show a good correlation between these two velocities in all seasons. In autumn and winter, geostrophic speed is slightly larger than the eddy propagation speed, and the direction of these two velocities almost overlap in winter. However, in summer, eddies propagate without strong constraints due to weak mean currents and even move a little faster than the ambient mean flow. In conclusion, there are no statistical differences between mean current and eddy propagation velocities, and the seasonal variations of the eddy behavior is probably owing to the seasonal change of the along-slope current.
Season Geostrophic velocity(GV) Eddy propagation velocity(EPV) Speed ratio
Velocity magnitude/m·s–1 Angel/(°) Velocity magnitude/m·s–1 Angel/(°) Spring 0.098 209.46 0.096 201.80 0.99 –9.785 Summer 0.059 196.95 0.062 201.45 1.05 5.709 Autumn 0.099 198.86 0.070 216.40 0.73 17.533 Winter 0.110 206.11 0.083 210.97 0.80 2.684
Table 2. Seasonal geostrophic current and eddy propagation velocities in the CSR
Statistical characteristics of mesoscale eddies on the continental slope in the northern South China Sea
- Received Date: 2018-11-15
- Available Online: 2020-04-21
- Publish Date: 2020-03-01
- the South China Sea /
- continental slope /
- Dongsha Islands /
- eddy statistics /
- eddy propagation
Abstract: The continental slope in the northern South China Sea (SCS) is rich in mesoscale eddies which play an important role in transport and retention of nutrients and biota. In this study, we investigate the statistical properties of eddy distributions and propagation in a period of 24 years between 1993 and 2016 by using the altimeter data. A total of 147 eddies are found in the continental slope region (CSR), including 70 cyclonic eddies (CEs) and 77 anticyclonic eddies (ACEs). For those eddies that appear in the CSR, the surrounding areas of Dongsha Islands (DS) and southwest of Taiwan (SWT) are considered as the primary sources, where eddies generated contribute more than 60% of the total. According to the spatial distribution of eddy relative vorticity, eddies are weakening as propagating westward. Although both CEs and ACEs roughly propagate along the slope isobaths, there are discrepancies between CEs and ACEs. The ACEs move slightly faster in the zonal direction, while the CEs tend to cross the isobaths with large bottom depth change. The ACEs generally move further into the basin areas after leaving the CSR while CEs remain around the CSR. The eddy propagation on the continental slope is likely to be associated with mean flow at a certain degree because the eddy trajectories have notable seasonal signals that are consistent with the seasonal cycle of geostrophic current. The results indicate that the eddy translation speed is statistically consistent with geostrophic velocity in both magnitude and direction.
|Citation:||Zi Cheng, Meng Zhou, Yisen Zhong, Zhaoru Zhang, Hailong Liu, Lei Zhou. Statistical characteristics of mesoscale eddies on the continental slope in the northern South China Sea[J]. Acta Oceanologica Sinica, 2020, 39(3): 36-44. doi: 10.1007/s13131-019-1530-3|