1. Introduction

This work describes observations of velocity made at a nominal depth of 15 m with satellite-tracked drifters and provides new direct evidence of seasonal near-surface flow from the Philippine Sea into the South China Sea through the Luzon Strait (Fig. 1). Such flow is frequently described as a westward branch of the Kuroshio. Mainly hydrographic methods have provided evidence of the seasonal penetration of the Philippine Sea water into the South China Sea (e.g., Fang et al. 1998).

The surface circulation south and east of the Luzon Strait is dominated by strong and persistent subtropical current systems. At the surface, the yearly mean Pacific North Equatorial Current bifurcates at ∼13°N near the east coast of Luzon to form the northward-flowing Kuroshio and the southward-flowing Mindanao Current (Nitani 1972; Toole et al. 1990; Qu and Lukas 2003). The near-surface bifurcation latitude moves between 11°N in May and 14.5°N in November, and at depth it is even farther north of its surface expression (Nitani 1972; Qu and Lukas 2003). Qiu and Lukas (1996) have noted that the interannual variations of the bifurcation latitude can be successfully computed from the wind-driven vorticity dynamics of linear and nonlinear reduced-gravity circulation models.

At 18°N the Kuroshio is a well-formed, northward-flowing western boundary current concentrated entirely west of 124°E; its high-speed core is positioned at 123°E, and its baroclinic structure is evident in the upper 600 m (Toole et al. 1990; Qu et al. 1998). Before reaching Taiwan, the Kuroshio encounters the Luzon Strait, which is the deepest passage from the Pacific Ocean to the South China Sea. At the southern portion of the strait, the Kuroshio takes a westward set and makes a detour into the South China Sea through the deepest channels of the Luzon Strait: the Balintany Channel and south of Babuyan Island. West and north of Batan Island, the Kuroshio flows within the Bashi Channel until it reaches the southeast coast of Taiwan (Gilson and Roemmich 2002).

Between the north coast of Luzon and the southeast coast of Taiwan, the Kuroshio is occasionally referred to as a “loop current” because it makes a loop, or excursion, into the South China Sea. The largest loop occurs between October and January when the winds are dominated by the northeast monsoon. A description of the loop current can be found, for example, in Nitani's (1972) maps of the geomagnetic electrokinetograph data and dynamic heights compiled from several hydrographic surveys. Li et al. (1998) used hydrographic data to show the looping of the Kuroshio in the Luzon Strait in summer months in the upper 200 m. Between October and March the warm surface water of Philippine Sea origin contrasts with the surrounding colder South China Sea water. This condition was exploited by Farris and Wimbush (1996), who observed the occurrence of Kuroshio surface loops and intrusions in autumn and winter with sea surface temperature (SST) maps made from Advanced Very High Resolution Radiometer data. As defined by the SST, the loop intrudes most severely into the South China Sea during the October–January period, when significant amounts of Philippine Sea water are also found below the surface. For example, Shaw (1989) used conductivity–temperature–depth (CTD) profiles to demonstrate that Philippine Sea water extends down to 500 m on the continental slope west of the southern tip of Taiwan in the March–August period. Shaw (1991), by analysis of a larger hydrographic dataset, tracked water of Kuroshio origin as far as west of 115°E in the upper 250 m and along the continental slope of the northern South China Sea. Qu (2002) reached similar conclusions about intrusions upon inspection of the distribution of oxygen concentration. Although the inflow is apparent from watermass properties, diapycnal mixing in the upper South China Sea and the outflow needed to achieve a mass balance of the basin are not known well.

Multilayer ocean general circulation model simulations of the transport of upper-layer water through the Luzon Strait are discussed in Metzger and Hurlburt (1996, 2001). In these simulations, the inflow is a result of the seasonally varying basin-scale circulation of the entire western subtropical and tropical Pacific. These works associate the seasonal inflow with the lowering of sea level caused by the increased cyclonic circulation of the South China Sea that results from the positive wind stress curl of the northeast monsoon. Sheremet (2001) used a barotropic model to address the local parameter dependence of a western boundary current that encounters a gap along the coast. The parameters are the breadth of the gap relative to the width of the Munk viscous boundary current and the Reynolds number of the flow, defined as the ratio between the transport rate of mass per unit depth and the lateral eddy diffusion coefficient. The general theoretical results are that a strong current “leaps the gap” and a weak current loop propagates westward into the gap, forming eddies in the intrusion.

Even in the absence of direct evidence of the circulation patterns (but see Farris and Wimbush 1996), observations and theory are supporting the view that water of Philippine Sea origin can reach the interior of the South China Sea. In numerical studies of the circulation of the South China Sea, the processes that cause the inflow appear to be associated with, or produce, mesoscale eddies. Observational evidence for this mesoscale variability has recently become available from hydrography (Li et al. 1998).

In this work, we present and discuss observations of the surface circulation in the western tropical Pacific and in the South China Sea and we report on the observed flow from the Philippine Sea to the South China Sea.

2. The data

Direct velocity measurements in the surface mixed layer were obtained with Argos satellite-tracked drifters drogued at a nominal depth of 15 m. Between November of 1986 and May of 2002, 380 instruments were deployed or drifted into the region delimited by 105°–135°E and 10°–25°N. Only 14 deployments were made in the South China Sea. For a description of the drifters, wind slip correction, and data-processing methods see Niiler (2001). In this analysis, the positions of each drifter, interpolated to 6-h intervals, were used to construct the velocity vectors, which are called instantaneous velocities. The estimated accuracy of the velocity measurements in a 10 m s⁻¹ wind is 10⁻² m s⁻¹ (Niiler et al. 1995). The Lagrangian statistics and the ensemble mean velocity field were calculated in the region delimited by 120° and 135°E longitude and by 10° and 25°N latitude. For each drifter trajectory that was longer than 20 days, we computed the Lagrangian decorrelation timescale and length scale in the zonal and in the meridional directions (e.g., Freeland et al. 1975; Poulain and Niiler 1989). Nonconverging autocovariance functions (the ones without zero crossings) were discarded. Seasonal Lagrangian timescale and length scale were also computed. The results are summarized in Table 1.

The dimensions of each averaging cell were chosen to be the eastward and northward Lagrangian space scales, respectively (and up to 2 times these values for the sparser seasonal datasets). Within each cell, the independent estimates of the zonal and meridional velocities were computed by time averaging the 6-h velocity time series in bins with amplitude given by the upper bound of the largest Lagrangian timescale [T + σ(T) in Table 1]. The ensemble mean velocity and its standard error ellipse referred to the principal axes (Freeland et al. 1975) were calculated for each cell. The North Equatorial Current bifurcation latitudes in each season were chosen where the zonal average (between 127° and 130°E) of the seasonal ensemble mean northward velocity components was zero. Unless stated otherwise, all uncertainties are assumed to be 1 standard deviation.

3. Flow of the Philippine Sea water through the Luzon Strait

The ensemble of the individual drifter tracks (Fig. 2) shows how the North Equatorial Current intensifies to form the Kuroshio, made visible by instantaneous speeds in excess of 0.7 m s⁻¹, east of Luzon between 16.5° and 18.5°N. To the north, the drifters in this flow move northwest and enter the Luzon Strait principally through the channels north (Balitany Channel) and south of Babuyan Island. Two drifters moved very rapidly westward on the continental shelf north of Luzon. Some drifters within the South China Sea continued to exhibit instantaneous speeds in excess of 0.6 m s⁻¹ within the current system that appears to surround its entire perimeter.

The seasonal displacement diagram of the drifters (Fig. 3) shows that a westward flow through the Luzon Strait occurs only between October and March. Of the 29 drifters that crossed the meridional line at 120.8°E between Luzon and Taiwan at least once, 15, as shown in blue in Fig. 4, (5 of which had lost the drogue) moved to the interior of the South China Sea and 6 reached the broad continental-shelf area south of Vietnam. Daily average speeds in excess of 0.8 m s⁻¹ were observed when the drifters traveled southward over the narrow shelf area located along the southeast Vietnamese coast (marked “A” in Fig. 4). The drifters whose positions are plotted in red on Fig. 4 recrossed the strait eastward and moved north, some of them following a small loop pattern southwest of the south coast of Taiwan (marked “B” in Fig. 4). Five of these drifters had lost the drogue, implying that the position error accumulates at a rate of at most 10 km day⁻¹ downwind because of the slip of the float relative to a drogued drifter (Niiler 2001), an error that tends to push the tracks southwestward during the northeast monsoon. Some of the red drifters were trapped in a low-velocity region (marked “C” in Fig. 4) near the southeast tip of Taiwan before rejoining the northward Kuroshio flow. In the proximity of the Luzon Strait there were data in 13 of the 17 yr analyzed. In 10 of the 13 yr, the drifters crossed the strait between October and January and remained in the South China Sea (the blue drifters in Fig. 4). In 8 of the 13 yr, the drifters made the small loop across the Luzon Strait between October and March (the red drifters in Fig. 4). Five drifters (all with their drogues attached) reached the strait by traveling westward and crossing directly over the mean Kuroshio path. Most of the trajectories north of 15°N (Fig. 4) indicate that besides the seasonal westward current into the Luzon Strait there is, to the east of the strait, a vigorous eddy field.

The velocities at the crossing line are plotted in Fig. 5. The average speed computed only with data from the drifters that reached the interior of the South China Sea is 0.7 ± 0.4 m s⁻¹. The maximum, almost westward, speed of 1.65 ± 0.01 m s⁻¹ (the instrumental error is used here) was observed at 20.7°N, 120.8°E in December of 1997.

The ensemble mean velocity field, computed with a resolution of approximately 0.5° from all of the data east and southeast of the Luzon Strait (Fig. 6), shows that the Kuroshio system is relatively well mapped. The northward boundary current forms near the Philippines coast where the North Equatorial Current separates in two branches. The latitude of this bifurcation point, computed from the annual mean surface (15 m) velocity field, is located at 12.4° ± 0.3°N (dashed line in Fig. 6), and it varies seasonally by 2.6 ± 1.5°, reaching its northernmost position of 15.0° ± 0.8°N in the October–December period (not shown), in general agreement with the results of Qu and Lukas (2003). The Kuroshio accelerates as mass is added from the east, and, by 14.8°N, the average speed increases above 0.5 ± 0.2 m s⁻¹ (computed in box 1, Fig. 6). A second concentrated stream appears to join the Kuroshio from the east at about 20°N. The mean westward speed of this current is 0.20 ± 0.05 m s⁻¹ (computed in box 2, Fig. 6) and increases to 0.34 ± 0.06 m s⁻¹ if only autumn data are considered (average computed in a slightly larger box, not shown). This stream is mainly sampled by the drifters that crossed the Luzon Strait and moved into the South China Sea in the October–January period, as depicted and discussed in Figs. 3 and 4. This relatively strong current that crosses the Kuroshio path appears in both the global and autumn ensemble means. However, the interannual variability of this westward current structure is not well determined by the drifter data, implying that the estimate of the standard error is based on potentially time-aliased data and should be considered as a lower bound.

The mesoscale activity between 15 December 1997 and 18 January 1998 can be seen in the sea level anomaly of Centre National d'études Spatiales Archivage, Validation et Interprétation des Données des Satellites Océanographiques (CNES/AVISO) Ocean Topography Experiment (TOPEX)/Poseidon–European Remote Sensing Satellite (ERS) merged altimeter data of sea surface height anomaly (SLA) and the Tropical Rainfall Measuring Mission Microwave Imager (TMI) 3-day-averaged SST data (Fig. 7). During this period, one drifter crossed the Luzon Strait. The water warmer than 26°C from the Philippine Sea intruded north of 20°N and is visible in Fig. 7 because the surrounding South China Sea surface water was colder than 24°C. The most obvious intrusion of warm water along the northern boundary of the South China Sea extended westward to ∼117.5°E, with a weaker continuation east of ∼116°E. An anticyclonic eddy (A) was centered at approximately (21.6°N, 117.8°E) near the 200-m isobath. In the Luzon Strait and in the South China Sea, south of approximately 20.5°N, the sea level was lower than the average and a cyclonic eddy was located directly north of Luzon (B). The drifter track in Fig. 7b crosses the Luzon Strait westward along the border between the cold and the warm water that lies between the anticyclonic and cyclonic eddies at approximately 20.5°N. The trajectory in Fig. 7b is shown 5 days before and after the date of the SLA and SST observations. The average speed of the drifter for the first 5 days, west of ∼118°E, was 1.4 ± 0.2 m s⁻¹. For the last 5 days, east of ∼118°E, when the drifter began to leave the warm water, the average speed was 0.5 ± 0.2 m s⁻¹. Figures 7c and 7d show that the warm-water intrusion separated from the warm water along the southern coast of Taiwan by being entrapped in the westward-moving anticyclonic eddy. The drifter moved to the south and west in a cyclonic circulation around a cyclonic feature (C) that propagated westward in the northern part of the South China Sea basin.

4. Discussion and summary

Velocity observations made with drifters drogued to 15-m depth describe a seasonally recurring inflow of surface water from the Philippine Sea into the South China Sea through the Luzon Strait between October and January. The permanent looplike intrusion of the Kuroshio described, for example, by Nitani (1972) and, with numerical models, Metzger and Hurlburt (2001) does not appear in the drifter data between April and September. The loop is confined to the north of the Luzon Strait in the remaining months, when its westward extension reaches to 119.2°E. A net westward Ekman transport through the Strait of Luzon, computed from Special Sensor Microwave Imager (SSM/I)-derived winds (Fig. 8) exists only during the northeast monsoon, reaching a maximum of 0.6 Sv (1 Sv ≡ 10⁶ m³ s⁻¹). Figure 8 also shows that the drifters enter the South China Sea preferentially during the early part of the onset of the northeast monsoon. The drifters that crossed the strait were entrained by the Kuroshio east of Luzon or crossed the Kuroshio path inside the Luzon Strait within a strong westward jet. The ensemble mean velocity field computed between October and December (not shown) indicates that a nearly zonal westward flow of 0.34 ± 0.06 m s⁻¹ occurred east of the Luzon Strait at ∼20°N. The sampling of the drifters was not uniform in time; however, this jet appears between ∼18.5° and ∼20.5°N in all of the years (1989, 1992, 1994, and 1995) during which there were data in the region. The near-surface Ekman currents produced by 10 m s⁻¹ northeast monsoon winds would be at most 0.1 m s⁻¹ (Ralph and Niiler 1999). The much stronger surface currents observed with the drifters indicate that a deeper current system to the west must be present during the northeast monsoon [as hypothesized, e.g., by Shaw (1989, 1991)] and that, in the Luzon Strait region, entrains Kuroshio water. The study of Sheremet (2001) predicts that the Kuroshio can penetrate the South China Sea when its strength is reduced. This weakening occurs in the autumn months (see, e.g., Qu et al. 1998), when the North Equatorial Current system shifts northward. The depth of the penetration and the flow regime must also depend on the topography of the Luzon Strait, a problem not dealt with well by Sheremet's (2001) theory or the low-vertical-resolution models of Metzger and Hurlburt (1996, 2001).