As the only deep-water connection between the South China Sea(SCS) and the northwestern Pacifi c Ocean, and the main wide gap in the Pacifi c western boundary between the isl and s of Luzon and Taiwan Island , the Luzon Strait has been a hotspot of scientifi c research for several decades. There is a consensus of opinion that the northern SCS is dominated by a large cyclonic gyre and characterized as a northward current west of Luzon toward Taiwan Island , which is paralleled by the Kuroshio mainstream east of Luzon. The complex current structure and water exchange in the Luzon Strait have attracted much attention from oceanographers. The circulation pattern and multiple time scales variabilities of the Kuroshio intrusion in the SCS have been investigated in several major ways, such as from in-situ data(Chen and Huang, 1996; Chu and Li, 2000; Qu et al., 2000), satellite data(Centurioni et al., 2004; Ho et al., 2004) and modeling results(Qu et al., 2004). Observati indicate that Kuroshio intrusions into the SCS may have different appearances, and several potential forcing mechanisms control the intrusion events(Hu et al., 2000; Liu et al., 2008). Among these mechanisms, the Kuroshio loop current enclosed by both the inflow and outflow of the Kuroshio through the Luzon Strait has often been quoted(Caruso et al., 2006) and reproduced in several numerical models(Farris and Wimbush, 1996; Xue et al., 2004). Yuan et al.(2006)demonstrated that the anticyclonic intrusion of the Kuroshio is a transient phenomenon rather than a persistent circulation pattern in the Luzon Strait. Nan et al.(2011a)defi ned the Kuroshio SCS Index for the fi rst time to quantitatively distinguish the Kuroshio intrusion as three main types: looping, leaking and leaping using altimetry data and HYCOM model results. At the present time, arguments are unresolved regarding how the Kuroshio intrudes into the SCS and what mechanisms control the variability of intrusion patterns.

Mesoscale eddies are rather active in the SCS, changing dynamic conditions within the ocean, playing an essential role in the transport of water mass, heat, salt and chemical substances, and affecting temperature and salinity variations in the water column. The distribution of eddy occurrences in the SCS has great regional differences. The area southwest of Taiwan Island is one of the highest eddy kinetic energy centers in the SCS(Chen et al., 2009); high mesoscale eddy activity being repeatedly reported in this area(Jia and Liu, 2004; Yuan et al., 2007; Wang et al., 2008a ; Sheu et al., 2010; Nan et al., 2011b). The mechanisms of eddy formation, polarity and development in the SCS differ from one place to another and one time to another. From statistical results and correlation analysis, wind stress curl(Xiu et al., 2010), the orographic wind jet(Wang et a l., 2008b) and background flow strength(Chen et al., 2011a)are all considered to be an important, but none is a sole explanation. To further underst and the problem of currents and eddies and their interactions, long time series moored array observations and fi nescale three-dimensional numerical models with eddy dynamics are urgently needed.

Restricted by insuffi cient in-situ data, especially vertical profi le data from inside an eddy or along a current, studies of the spatial and temporal evolution of eddies and currents are not common. In this study, we examine one anti-cyclonic eddy and one cyclonic eddy southwest of Taiwan Island and the surface flow field in the Luzon Strait using in-situ hydrographic data, an Argo profi le and trajectory data and the multisource satellite-derived data collected from May to August 2009. The thermohaline structures and water mass characteristics are the focus of the study. The rest of the paper is organized as follows. Section 2 introduces the datasets as well as the method of surface current calculation. Section 3 represents the results of observations in the Luzon Strait and ambient area. Section 4 is a discussion and Section 5 the summary.

Conductivity-Temperature-Depth(CTD)profi le data were obtained in the northeastern SCS and Luzon Strait on R/V Kexue No . 1 from June 22 to July 5, 2009. The CTD survey included 40 stations from 118°E to 122°E over the Luzon Strait area(Fig. 1a), with approximate spatial intervals of 0.5° in the meridional(longitudinal)direction and 1° in the zonal(latitudinal)direction. The observational depth of CTD is about 10 m above the seabed, and the maximum depth of the CTD instrument is fi xed at 1 550 m if the bottom depth is greater than 1 550 m. The vertical resolution of all the CTD data is set to 1 m. Based on CTD data, water masses in this region can be identifi ed by their temperature-salinity(T-S)characteristics.

Fig. 1 Field experiment over the Luzon Strait area a. CTD stations (black solid dots) acquired during late June and early July 2009 with bathymetry (grey dashed lines denote 50, 200, 1 000 and 3 000 m isobaths, as indicated) from ETOP02 digital topography; b. temperature profiles at K801 (black) and K601 (red); c. salinity profiles at K801 (black) and K601 (red); d. composite T-S scatter-diagrams for six CTD observations (grey curves are contour lines of isopycnic surface σθ , units are kg/m3 ); e. trajectories of Argo floats (after quality control) used in this study are indicated by green (2901169), blue (2901180) and red (2901183) solid lines; the grey dots are locations of CTD survey shown in Fig.1a.

Figure options

The global Argo project has grown to be a major component of the ocean observing system. The broadscale array of temperature and salinity profi ling floats collects data that can be used in innovative ways to solve large-scale three-dimensional problems related to ocean circulation and climate change. To balance spatial layout and prolong observation life, Argo deployments avoid strong current areas and marginal seas. Generally, an Argo has a sampling range in upper 2 000 m of the ocean, and it freely drifts at its parking depth with neutrally buoyant by having a density equal to the ambient pressure. When working, the Argo float fi rst dives to the setting depth(usually 2 000 m, referred to as the deepest depth) and then autonomously ascends to sea surface recording the T-S relationship of the water. These profi le data are transmitted to satellites during surface drifting, and then the Argo sinks to its parking depth and drifts until the cycle is repeated.

The real-time Argo data used in this paper are from floats 2901169, 2901180 and 2901183(hereinafter referred to as A69, A80 and A83)obtained from the China Argo Project. In October 2008(A69) and July 2009(A80 and A83), these three Argos were deployed on the west side of the Kuroshio mean axis in the Luzon Strait. The A69 and A83 deployments were at roughly the same location(120.7°E, 20.0°N), and the A80 was deployed a spherical distance of 43 km east of the fi rst two floats. However, their surface trajectories were rather different from their deployment to death(Fig. 1e). A83(green)migrated eastward for a short term, and then turned north out of the Luzon Strait and finally stopped at the northern SCS continental shelf. A69(blue) and A80(red)were seen to move northward in the Luzon Strait. When passing through 21°N, A69 turned to the northwest into the northern SCS and rotated to the southwest of Taiwan Island ; A80 turned to the northeast out of the Luzon Strait at the southern tip of Taiwan Island . Although the survey of these three Argos has spatiotemporal asynchronism during the observational period, they were all positioned in the Luzon Strait area from May to August in 2009. This provides an opportunity to investigate the complex circulation structure and water exchange using Argo data from the strong western boundary current(Kuroshio)area.

The three Argos all collected data in a similar fashion but differed somewhat in their design characteristics. The parking depths were set at 1 000 m and vertical observing intervals varied with depth(10 m intervals from the surface to about 400 m, and 50 m intervals below 400 m). However, while the profi le cycle interval of A80 and A83 was typically 10 days, the cycle time of A69 was only 3 days. Highfrequency profi le collection can provide more T-S information; however the higher proportion of residence time for A69 at the sea surface in comparison to its parking depth shows that its trajectory was much more affected by the surface flow field than the other two floats.

The primary satellite data used to identify the current-eddy structure in the Luzon Strait area are the global merged absolute dynamic topography(ADT) and sea level anomaly(SLA)data sets produced by the AVISO data project. The products are available every week with 1/4° resolution. The ADT data are used to calculate the horizontal geostrophic velocity(u_g, v_g)by the geostrophic balance: fu_g =-g· ∂(ADT)/ ∂y, fv_g=g·∂(ADT)/∂x. We also get the surface horizontal geostrophic velocities anomaly (u ′, v ′)from SLA maps: fu′=0g· ∂(SLA)/∂y, fv′=g·∂(SLA)/∂x, where f is the local Coriolis parameter, and g is the gravitational constant. We consider an objective criterion of the Okubo-Weiss parameter(W = α² + η² − ω²)as an eddy identifi cation(Isern-Fontanet et al., 2003; Xiu et al., 2010), where α=∂u′/∂x−∂v′/∂y and η=∂v′/∂x+∂u′/∂y are stretching and shearing deformations, and ω=∂v′/∂x−∂u′/∂y is the vorticity of the geostrophic current field. Under the assumption that strain and vorticity are slowly varying, the vorticity-dominant region(W ＜-0.2 σ_w, where σ_w is the st and ard deviation of W obtained from each time step of the entire domain of interest)is designated as the eddy. In addition, altimetry data are still aliased by tides and internal waves over the shelf, although they have had tidal and sea level pressure corrections incorporated. Following the conventional processing method, altimetry data for areas shallower than 200 m are excluded.

Monthly Moderate Resolution Imaging Spectroradiometer(MODIS)4 km sea surface temperature(SST)data were also acquired to corroborate exchanges between the Kuroshio and SCS waters in the Luzon Strait. Although MODIS measures SST by two b and s: thermal infrared at 11 μm and mid-infrared(mid-IR)at 4 μm, the differences are indistinctive in the study, so we only present the 4-μm SST product.

Wind is an important external forcing factor on the sea surface, and the monthly QuikSCAT wind field product at 1/4° resolution is used to examine the potential dynamics of eddies southwest of Taiwan Island . The Sea-Winds instrument onboard QuikSCAT is considered very accurate in the 3–20-m/s range. Ocean near-surface 10-m scatterometer winds are converted to wind stress(τ)with a bulk formula used by Oey et al.(2006). The drag coeffi cient(C d)is that proposed by Large and Pond(1981)for low-tomoderate winds and by Powell et al.(2003)for high winds. The density of air is fi xed as a constant of 1.29 kg/m³ . The wind stress curl only selects the vertical component and is calculated as(∂τ_y/∂x-∂τ_x/∂y)using central difference.

Satellites or GPS track the positions of a float as it drifts for about a dozen hours on the surface before descending to its parking depth. Using these data, we can estimate the surface current containing not only the geostrophic component but also the nongeostrophic component. So the Argo trajectoryderived surface current is more credible in the tropical oceans where the rotation effect f is not signifi cant. Based on the least squares method, Park et al.(2004, 2005)demonstrated an estimation of the inertial current from Argo surface trajectories. This method required at least six surface positions(after quality control)to obtain stable results. Xie and Zhu(2008)presented a Kalman Filter method for calculating the current and extrapolating the trajectory.

In this study, we use a simple and fast way to determine the surface current from Argo trajectories, without regard for the position number and trajectory forecasting. First, we assume that the surface flow changes slowly so that the non-stationary nature of surface motions is excluded. Second, for each duration time at the surface, we obtain fi tting lines in both the X- and Y-directions(Fig. 2). Denoted as(x_i, y_i)are the spatial locations on ocean surface of the 4^th surface drifting of A80(the subscript i is the number of logged fi xes; e.g., in the 4^th A80 cycle, i =12). Therefore, the X- and Y-direction fi tting lines are y=k_zonal·x+b_zonal and y=k_merid·x+b_merid, where x_i(y_i)is an independent variable in k_zonal(k_merid) and b_zonal(b_merid)but is a dependent variable in k_merid(k_zonal) and b_merid(b_zonal). Third, we calculate the distances from each location point(black circles)to the X-(black line) and Y-direction(red line)fi tting lines, and then sum them as the indexes ΣX and ΣY. The low value of the ΣX or ΣY decides the direction of the flow field. In Fig. 2, the values of ΣX and ΣY are 0.29 and 0.41 km, and we therefore chose the zonal direction as the drifting direction. Following this method, the spherical displacement of the flow field is about 101.4 km from(x 1, y 1)to(x 12, y 12), and the direction of flow field is 79.5°(with due north at 0°).

Fig. 2 Locations during the 4th surface drifting of A80 from 16:37 UTC on August 7 to 7:19 UTC on August 8, 2009 The black circles mark 12 GPS points. The black and red lines are linear least-square fits in the zonal (latitudinal) and meridional (longitudinal) directions,respectively.

Figure options

The Luzon Strait is a natural barrier between the northeastern SCS and the northwestern Pacifi c Ocean; hydrographic conditions differ greatly between SCS water and Kuroshio water. The T-S profiles from cruise stations K801 and K601(separated by a spherical distance of 232 km and 66 h in sampling time)show typical water characteristics for the two areas and their marked differences(Fig. 1b and c). A relatively high-salinity, high-temperature water mass was observed to exist in the upper 300 m at station K801. An alternating maximum(upper layer) and minimum(intermediate layer)in the vertical salinity profi le is shown in Fig. 1c. The Kuroshio water was observed to be more uniform in the upper mixing layer, which extends to a depth of up to 100 m under serene sea state conditions.

The T-S scatter-diagrams for the southernmost six stations are plotted in Fig. 1d. The water mass at station K801 had a maximum salinity >35.0 with a temperature of 24.8°C on the 23.4- σ_θ surface at a depth of 150 m and minimum salinity ＜34.2 near 5.8°C on the 27.0- σ_θ surface beneath 670 m. The only source of high-salinity water is the northwestern Pacifi c, and the Kuroshio carries the high-salinity water northward along the east coast of Luzon Island into the Luzon Strait. Meanwhile, both the high- and low-salinity characteristics are associated with two well-defi ned water masses: the high-salinity North Pacifi c Tropical Water and the low-salinity North Pacifi c Intermediate Water, as compiled by Qu et al.(1999). Compared with previous in-situ data from the same station in August 2008(Liu et al., 2010), the salinity values seem to be little changed, suggesting that the interannual variability of the vertical salinity of the Kuroshio water is not signifi cant. The T-S diagram for a zonal section along 18.5°N(K401, K508, K509 and K601)shows consistent water mass characteristics. The maximum salinity was less than 34.6 with a temperature of 18.5°C on the 24.8- σ_θ surface at about 160 m, but the minimum salinity reached 34.4 near 9.0°C on the 26.6- σ_θ surface at 670 m. This is typical of SCS water(Shaw, 1991; Chen et al., 2011b). However, the water mass at station K701(cyan curve in Fig. 1d)had a medium maximum salinity value of 34.8 between the Kuroshio water and the SCS water, suggesting that the water has more than one source above 25.7 σ_θ and that strong mixing occurs there. Although K701 is very close to Luzon Island , the water exchange between the Kuroshio and SCS waters had begun to emerge over the 200-m deep water column.

There are three distinct branches of the salinity maximum on the isopycnic surfaces from 23.0 σ_θ to 25.0 σ_θ . The Kuroshio and SCS waters form curves at the two boundaries, and water with mixed characteristics occupies the middle portion. The three branches in the T-S diagram also indicate the presence of complex water mass distributions from the Kuroshio, the SCS, and their mixing within the Luzon Strait in late June and early July, 2009. When below 25.7 σ_θ, all of the water masses, except K801, had the characteristics of SCS water. The intermediate water from the SCS were slightly warmer and more saline than that from the Kuroshio at a depth below 300 m, and SCS water showed a tendency to flow into the northwestern Pacifi c along the north shore of Luzon Island , at least occupying the areas west of 121°E. Because of the scarcity of the in-situ data from K801 to Luzon Island , the ability of the SCS water to flow into the northwestern Pacifi c across 122°E needs more observations from future studies.

Using the A69 float information collected during the summer of 2009, an anti-cyclonic structure was identifi ed southwest of Taiwan Island is an area of known mesoscale eddies(Chen et al., 2011a; Nan et al., 2011b). To examine the spatial and temporal evolutions of an eddy, we obtained surface trajectories with objective-defi ned velocities for the 3 months of A69.

The Argo surface track for May is shown in Fig. 3a. During this period, A69 presented a perfect clockwise movement straddling the shelf slope southwest of Taiwan Island with the central point at(120°E, 22°N)near Profile 69. Strong flows were all concentrated in deep water(about 2 000 m). The maximum velocity appeared at Profile 69 with a speed of 90.7 cm/s; a secondary maximum reached 90.3 cm/s at Profile 71.The two maxima had opposite orientations across the local isobaths. Shallow water flows are always weak, and the mean speed was only 32.2 cm/s between Profiles 75 and 79. The flow directions near the shore of Taiwan Island were basically along isobaths. The surface trajectory in June(Fig. 3b)indicates northward propagation of A69 and looks like an elongated semianticyclonic structure. The velocities were generally low with a mean speed of 35.4 cm/s. Profiles 81 and 89 still show high speeds of up to 80.8 and 56.2 cm/s, respectively. The surface track from July(Fig. 3c)huddles over the 1 000-m isobath, and the mean speed was only 27.4 cm/s with a maximum velocity reaching 54.7 cm/s at Profile 91 in early July. Summarizing the trajectories of A69 from May to July, the strength of anti-cyclonic eddy was observed to gradually decrease, as seen from surface velocity weakening and the shrinking area occupied by the eddy.

Fig. 3 A69 information for May through July, 2009 Surface currents (a. in May; b. in June; c. in July) southwest of Taiwan Island. Numbers indicate the serial numbers of the profiles. The length of blue arrow is proportional to the magnitude of the velocity (in cm/s). Vertical profiles of d. temperature (in °C) and e. salinity (in) against pressure (in m).

Figure options

Argo floats also give us a chance to examine T-S characteristics along the trajectory. According to the time series of salinity profiles collected by A69(Fig. 3e), the maximum salinity reached 34.8 in the high-salinity cores. Relatively high-salinity(>34.6)patches were embedded in the vicinity of the 200-m layer from May to June but disappeared in July, 2009. High-salinity water in the upper layer could only be derived from Kuroshio water that intrudes into the sea southwest of Taiwan Island across the Luzon Strait; however, the temperature around the high-salinity patches is not very high(~20°C). So the lowtemperature, high-salinity water mass is unlikely to have originated only from Kuroshio leaking(i.e., a direct Kuroshio branch into the northern SCS)or looping(i.e., both inflow and outflow of the Kuroshio through the strait)in the Luzon Strait. The mixedwater characteristics observed between the Kuroshio and SCS waters are prominent. At the same time, we also found that low-salinity cores(＜34.4)appear around the 500-m layer. The low-salinity value was a little lower than the SCS water but was much higher than the Kuroshio water in the intermediate layer, whereas the temperature maintained a stable level of 9°C, which is typical of SCS water. Water with hightemperature and low-salinity expresses that the SCS water has a signifi cant advantage in the middle layer and the infl uence of the Kuroshio water seems very weak. Based on the T-S evolution shown in Fig. 3d and e, we believe that the water in the anti-cyclonic eddy is not from a single source and the mixing process only happens in the shallow area.

From Fig. 3d, isotherms in the upper layer were observed to have an uplifting tendency, especially in July 2009. The 24.8°C curve slopes upward from 110 m in May to 50 m in July, and the 18.5°C contour jumps from 190 m in May to 120 m in July. In addition to the effect of surface heating from solar radiation in summertime, the variation of isotherms also suggests a new dynamic process, such as a cyclonic eddy, induced uplifting(discussed further in Section 4.1).

In addition to the Argo data, we can also obtain a snapshot of vertical salinity structure based on the results of hydrographic observations. The salinity distribution along the 21°N section can be examined in Fig. 4. At K801, a salinity core higher than 34.8 was found roughly at the 200-m layer, indicative of typical mixing between Kuroshio and SCS waters in the subsurface layer. Above the 200-m layer, a highsalinity belt(>34.6)extended from K805 to K503. Two low-salinity cores(＜34.4)are located beneath the 400-m layer. One is a small plaque at 119°E from 400 to 550 m, the other one is a great group extending from 120°–122°E and covering depths from 400 to 850 m. Obviously, the high-salinities observed in the subsurface layer and low-salinities seen in the middle layer are both infl uenced by Kuroshio water. However, the origins of these salinities are different as shown by the in-situ salinity profiles. The high-salinity belt in the subsurface layer looks like a natural extension of the Kuroshio salty water into the northeastern SCS, indicating the continuous horizontal diffusion of the salt from heavy to low concentrations represented as the Kuroshio front intrusion. The low-salinity core at K503 comes from the well-mixed water that forms in the lower latitudes of the Luzon Strait and then is entrained by SCS circulation northward. The Kuroshio mainstream does not directly flow into the SCS across the Luzon Strait in the middle layer, staying on the eastern side of the strait at 122°E or even in more eastern sea areas. From the salinity distribution in section 21°N, we consider that the dynamic mechanisms of the flow structure at the subsurface and in the middle layers are different, and the horizontal gradient effect controlled by the west wing of the Kuroshio front is more signifi cant at shallow depths.

Fig. 4 Distribution of salinity (in) along the 21°N section

Figure options

Floats A83 and A80 were deployed on July 5 and July 7, 2009, respectively. Unlike the short profi ling cycle of A69, both A83 and A80 had 10-day observing intervals, thereby providing more information on the effect of the flow field at the parking depth. From July to August, the trajectories of these two floats revealed that the surface flow field in the Luzon Strait is rather complicated. In Fig. 5a, A83 presents an unexplained east-west orientated shuttle run in the middle of the Luzon Strait, whereas A80 followed the Kuroshio leap current, which is traditionally defi ned as a clockwise bend into(not through)the Luzon Strait without any direct penetration of Kuroshio water west of 120.5°E(Caruso et al., 2006). The mean speed of A80 reached 79.1 cm/s, but that of A83 was only 44.9 cm/s, just over half the speed of A80. The maximum velocity of A83 was 79.1 cm/s in Profile 3, and the alternating eastward and westward trajectories also happened in Profile 3 at 121.6°E. The opposite drift paths of A83 from Profiles 1 to 6 suggest that a weak SCS outflow was obstructed by the solid flowing boundary of the Kuroshio and the float was pushed back to the SCS by strong Kuroshio inflow. As for A80, all of its profiles have high speeds in the surface layer except Profile 6. The highest velocities appeared in Profiles 3 to 5 with a mean speed of 104.5 cm/s, just like an outflow from the SCS into the northwestern Pacifi c to finally rejoin the Kuroshio mainstream; the maximum velocity achieved was 122.1 cm/s to the northeast in Profile 5.

Fig. 5 Same as Fig.3, except showing data for A80 (circles) and A83 (squares) during July and August, 2009 a. surface currents in the Luzon Strait; profi les of b. temperature and d. salinity measured by A83, and of c. temperature and e. salinity by A80.

Figure options

Each float obtained six profiles in the north-central region of the Luzon Strait during the 2-month observational period. A83 and A80 have different time series of T-S characteristics(Fig. 5b–e), indicating a sophisticated water exchange and circulation structure along the 121°E section that has been repeatedly described(Hu et al., 2000; Tian et al., 2006; Yuan et al., 2006; Liang et al., 2008; Liu et al., 2008; Chen et al., 2011b, c). One salty core higher than 34.8 is found on Profile 3 from A83 at 200 m, and two highsalinity(>34.7)plaques were located on Profiles 1 and 2 from 160 to 230 m and on Profile 4 above 150 m. The locations of the high-salinity regions on Profile 3 from A83 and Profile 4 from A80 are considered telling features of both the inflow and outflow of the Kuroshio in the Luzon Strait. The water masses sampled by these profiles are notably different from the waters of the SCS and Kuroshio, implying vigorous mixing between the Kuroshio water and SCS water in the central Luzon Strait and at the southern tip of Taiwan Island .

Using Argo trajectory data from May to July 2009, we have characterized the surface flow field through the Luzon Strait from the perspective of Lagrangian flow measurements. In this section, we will confi rm these with Eulerian flow measurements by using satellite-derived data, such as altimetry SLA and geostrophic currents, infrared SST and scatterometer near-surface winds.

The images of the weekly SLA and geostrophic currents during May and June 2009 are shown in Fig. 6. An anti-cyclonic eddy is clearly visible southwest of Taiwan Island by positive SLA contours and a clockwise geostrophic current field(see the red rectangular area in Fig. 6). This eddy is very close to the 200-m isobath and also to the isl and of Taiwan Island , so we cannot observe completely closed eddy formation from the SLA data. From the incomplete SLA contours, we are still able to determine that the eddy is elliptical, with a diameter of about 160–280 km, similar to statistical data for Luzon warm eddies in summer-fall(Yuan et al., 2007)but much larger than the data for eddies southwest of Taiwan Island (Nan et al., 2011b). The maximum SLA of the eddy ranged from 16 to 23 cm in May and June, and the W value over the same period maintained a low level that was much less than -0.2 σ_w, indicating a strong eddy existence. The period of this anti-cyclonic eddy event is about 2 months, but this eddy remained in place instead of propagating westward during its lifetime. The geostrophic currents express a continuous clockwise movement around the high value at the center of the SLA feature. As for the typical Kuroshio intrusions into the SCS mentioned by Caruso et al.(2006) and Nan et al.(2011a), neither the Kuroshio leaking nor looping events occurred in the geostrophic current data during May and June over the Luzon Strait. Meanwhile, the red rectangular region shown in Fig. 6 was indeed an area where eddies shedding from the Kuroshio bend are often observed(Li et al., 1998; Jia and Liu, 2004). If this eddy was really generated from the Kuroshio, the T-S characteristics inside the eddy should have moved with the eddy because of advective trapping of interior water parcels(Dong et al., 2014). According to the information from float A69, the maximum salinity value only reached 34.8, which is much fresher than typical Kuroshio salinity(Fig. 3). So we speculate that the anti-cyclonic eddy southwest of Taiwan Island during May and June 2009 originated from the me and ering front of the Kuroshio in the Luzon Strait, which has been detected by the analysis of historical observational data by Shaw(1991).

Fig. 6 Weekly mean SLA (white-and-grey) and geostrophic current (blue vectors) from May and June 2009 over the Luzon Strait area Data from a. May 6; b. May 13; c. June 17; d. June 24. Units are cm for the sea levels and cm/s for the currents. Contour interval for sea level is 5 cm, and negative SLAs are shaded. The red rectangular frame delineates the area where an anti-cyclonic eddy occurs. Altimetry data over the shelf shallower than 200 m are masked out.

Figure options

MODIS monthly SST and near-surface wind data were also collected for the same period. The Kuroshio mainstream and the SCS water can be identifi ed from Fig. 7a, and the boundaries are indicated by the 28°C contours. Both the Kuroshio and the SCS water had warm tongue features in May 2009, extending northeast and northwest of Luzon Island , respectively, and then converging at 20°N in the central part of the Luzon Strait where the alternating trajectory was captured by float A83. The signifi cant anti-cyclonic eddy was evidenced by the offshore extension of two warm SST fi laments which were much warmer than ambient SST field southwest of Taiwan Island . In addition, there was a cold SST region(i.e., ＜27.5°C)in the interior of the two warm tongues, indicative of a blocking effect on water-mixing by the isl and chain. Well–mixed waters, like the results observed at station K701(see Fig. 1d), were not seen in May 2009. In Fig. 7b, the warm tongue SST feature around Luzon was shown to be eclipsed by the spatially uniform surface heating that occurs during the summer. However, the warm fi lament is still signifi cant when compared with the relatively cold background SST.

Fig. 7 MODIS/Aqua monthly average SST over the Luzon Strait area Data from a. May 2009; b. June 2009.

Figure options

The monthly mean wind stress curl over the Luzon Strait was strongly negative southwest of Taiwan Island in May 2009. The core of the stress curl, with an absolute value higher than 30×10 ⁻⁸ N/m³, is in excellent agreement with the location of the anticyclonic eddy confi rmed by both the trajectory of float A69 and altimetry data. Theoretically, the negative wind stress curl could easily produce or preserve anticyclonic eddies through linear dynamics without considering the lateral stresses. The wind stress curl weakened in June(in Fig. 8b) and became almost zero during July and August(not shown). Consequently, the anti-cyclonic eddy southwest of Taiwan Island was unable to be maintained under this condition, resulting in the eddy atrophy shown in Fig. 3c.

Fig. 8 Spatial distribution of the wind stress curl (in 10-8 N/m3 ) within the study area Data from a. May 2009; b. June 2009. Contour interval for wind stress curl is 10×10-8 N/m3 , and negative wind stress curls are shaded.

Figure options

In July 2009, a cyclonic eddy appeared in the position where the anti-cyclonic eddy was located in May(see the red rectangular area in Fig. 9). The eddy seemed weak, and the structure was a quasi-circle with closed negative SLA contours. The maximum SLA of the eddy varied from -5 to -8 cm during its lifetime, and the diameter was about 90–170 km, which is much smaller than the anti-cyclonic eddy. Cyclonic eddies are able to move deeper cold water upward and make the thermocline depth shallower, a process which coincided with variations of the temperature profi le of float A69 in July(Fig. 3e). According to seasonal eddy occurrence patterns(Nan et al., 2011b), summer is season with most cyclonic eddy occurrences southwest of Taiwan Island . However, the wind stress curl is not strong in summer, and this inconsistency suggests that the wind stress curl cannot be the determining factor in the eddy formation in this area. Also, the Kuroshio bend or loop current always generates shedding anti-cyclonic eddies but not cyclonic eddies themselves. Therefore, the dynamics of this cyclonic eddy formation is still uncertain and requires further study.

Fig. 9 Same as Fig.6, except for July 2009 Data from a. July 8, b. July 15, c. July 22, d. July 29. The red rectangular frame shows the area where a cyclonic eddy occurs.

Figure options

Chen et al.(2011b)previously computed the geostrophic flow field from the same hydrographic CTD data of this paper using the thermal wind relation. They found that the Kuroshio axis follows a ε-shaped path slightly east of 121°E in the upper layer and the SCS water flows into the Philippine Sea through the Luzon Strait via three paths: to the north of Luzon Island and to the south of Taiwan Island in the upper layer, and to the north of Luzon Island in the deep layer. Based on the analysis of hydrographic observations, Argo float trajectories, and multi-source satellite remote sensing data acquired during July and August 2009, we can preliminarily identify the main features of the surface current and sketch out a general structure of the flow field from another point of view.

Although there are differences between the Lagrangian and Eulerian measurements in how they portray the flow field, they can be combined to characterize mean flow under time-averaged conditions. There are two current systems that migrate northward to be encountered in the middle of the Luzon Strait. Float A83 followed an outflow from the SCS and an inflow from the Kuroshio at the 20°N section, and float A80 followed a Kuroshio bend into the Luzon Strait but did not cross 120.5°E. When the Kuroshio flows out of the Luzon Strait just off the southern tip of Taiwan Island , the water masses from K805 to K808 become less salty than the water of the Kuroshio source east of Luzon Island from K801 to K804 in the upper layer above the 25.0 σ_θ isopycnic surface(Fig. 10). There is no directed branch of the Kuroshio into the SCS, but the Kuroshio frontal intrusion in the subsurface layer could extend to 119°E or even farther west. The high-energy Kuroshio frontal eddy located southwest of Taiwan Island maintained a stable position, and the strongly negative wind stress curl also induced the formation of an anticyclonic eddy there.

Fig. 10 Composite T-S scatter-diagrams along the 122°E section Grey curves are isopycnic surfaces.

Figure options

In-situ hydrographic data, Argo profi le and trajectory information, and multi-source satellitederived data have been used to examine the formation of mesoscale eddies southwest of Taiwan Island and the surface flow field structure in the Luzon Strait area from May to August 2009. Water mass characteristics and mixing can be assessed with T-S scatter diagrams from the CTD field experiment and by time series of T-S profiles from three Argo floats along their trajectories. Using satellite positioning information for the Argo floats, surface velocities can also be calculated by virtue of an objective defi nition, which can be collocated well with altimetry SLA and geostrophic current data over the same period. Monthly data from MODIS SST and QuikSCAT nearsurface wind provide good data sets for observing the dynamics of eddy formation.

An anti-cyclonic eddy was found southwest of Taiwan Island in May and June 2009, where both the kinetic energy and occurrence frequency of eddies is very high(Wang et al., 2003; Chen et al., 2009; Nan et al., 2011b). The eddy, straddling the shelf and slope down to the 1000-m isobath remained mostly stationary during its lifetime; the eddy was strong in May and June and then lost its intensity in July. Float A69 identifi ed high-salinity anomalies in the vicinity of the 200-m layer, which is very different from typical SCS water. The in-situ CTD salinity distribution along the 21°N section suggests that the Kuroshio frontal intrusion through the Luzon Strait led to the formation of this eddy and the frontal infl uence occupied the subsurface layer in the form of continuous horizontal diffusion extending west to at least 119°E. A warm SST fi lament, negative SLA and clockwise geostrophic current are all supported the existence of the eddy. Meanwhile, the great wind stress curl was also a favorable factor for the anticyclonic eddy formation.

As to water exchange in Luzon Strait, water mixing appears to be very energetic in the upper layer, at least between station K706 and the entire strait. SCS water has a strong tendency to flow out of the strait until it is finally blocked by the Kuroshio mainstream wriggling past 122°E. Floats A83 and A80 also depict a complex surface flow field during July and August in Luzon Strait. The coexistence of both outflow and inflow appears at 20°N section. The water masses in the outflow have a low-salinity feature round the 200- m layer that probably has an affi nity to SCS water; whereas the water in the inflow has a high-salinity core that is saltier than 34.8 and refl ects the Kuroshio me and ering northwestward in the middle of the Luzon Strait. Float A80 indicates a complete Kuroshio bend over the north-central area of Luzon Strait but it does not cross 120.5°E. The outflow past the southern tip of Taiwan Island has a strong surface current feature, with the mean velocity reaching 104.5 cm/s. A less salty outflow implies that a possible SCS water remnant is carried by the Kuroshio bend out of the strait.

The authors are grateful for the comments of the anonymous reviewer. And we also would like to thank Captain ZHANG Zhiping and the crew of R/V Kexue No 1 for enthusiastic support of the project. This study is as a partial for fulfi llment for Liu’s Ph. D. requirement at IOCAS.