Using recently collected current and hydrographic data, we provide a high resolution picture of the subinertial flow and estimate the volume transport through the Luzon Strait. The distribution of the subinertial flow shows a strong westward flow around 100 m in the northern part of the Luzon Strait, while the eastward flow is confined to the deeper layers, mostly at depths around 1000 m. The total volume transport is estimated to be 6 ± 3 Sv during the period of observations from October 4 to 16, 2005. The observations also confirm that the Luzon Strait transport has a sandwiched vertical structure. The net westward volume transport in the deep (>1500 m) layer of the Luzon Strait reaches 2 Sv.
 The Luzon Strait is the only deep channel that connects the South China Sea (SCS) with the Pacific. The transport through the Luzon Strait, called the Luzon Strait Transport (LST) hereinafter, is an important process influencing the circulation and heat and fresh water budgets of the SCS [Qu et al., 2004]. Evidence also exists to suggest that the LST conveys Pacific variability further southward to the Java Sea and has a notable impact on the Indonesian Throughflow [Qu et al., 2005], implying a potentially important role in the world's climate.
 The LST has been extensively studied in past decades. In the upper layer (above 300 m), Wyrtki  first noted that water enters the SCS in winter and flows back to the Pacific in summer, with an annual average inflow of ∼0.5 Sv from the Pacific to the SCS. In the intermediate layer, between 350 and 1350 m, recent studies suggested an outflow to compensate the inflow of the Pacific water in the surface layer [Chen and Huang, 1996]. In the deep layer (below 2000 m), recent observations showed that there is a deep ventilation of the SCS by water from the Pacific through the Luzon Strait [Nitani, 1972; Broecker et al., 1986] and the abyssal water transport reaches as high as 0.7–2.5 Sv [Wang, 1986; Liu and Liu, 1988; Qu et al., 2006]. Existing studies of LST yield mean transport estimates ranging from 0.5 to 10 Sv [Wyrtki, 1961; Metzger and Hurlburt, 1996; Qu et al., 2000; Lebedev and Yaremchuk, 2000; Chu and Li, 2000; Yaremchuk and Qu, 2004]. However, because of the lack of high-resolution, full-depth current observations, these studies failed to provide a detailed picture of the velocity structure in the Luzon Strait. In addition, most of the earlier studies were based on hydrographic data and the resulting estimates depended on the assumption of geostrophy. This study uses recently collected current and hydrographic data to examine the subinertial flow structure and the total volume transport through the Luzon Strait.
2. Experimental Setting and Data Processing
2.1. Experimental Setting
 A field experiment was conducted using R/V. Dongfanghong No.2 to measure the flow and temperature-salinity-pressure profiles across the Luzon Strait from October 4 to 16, 2005 (Figure 1). The experiment consisted of 15 stations in the Luzon Strait, with a station spacing of about 20 Km. Station locations were carefully selected, well to the west of the main axis of the Kuroshio Current, to minimize the transport leakage near the coasts. In order to effectively filter out tidal currents and internal waves, we also conducted repeat measurements at 5 stations where the data records were typically 30 hours.
 A total of 65 full-depth profiles of horizontal velocity (u, v), temperature (T), salinity (S), and pressure (P) were collected at all 15 stations, including 55 profiles at 5 repeat-occupation stations. The flow measuring instrument was a 300-kHz LADCP manufactured by RD Instruments, Inc. The vertical bin size was set to 8 m, the number of layers was set to 13, and the sampling frequency was set to 1 Hz. The instrument measuring temperature, salinity and pressure was an SBE 911 Plus CTD manufactured by SeaBird Electronics, Inc. We used an acoustic altimeter mounted on the CTD package to enable all profiles to extend to within 30 m of the ocean bottom.
2.2. Data Processing
 The relative horizontal velocity was measured throughout the water column with an estimated uncertainty of 1 cm s−1, and the absolute velocity was derived from the relative velocity by a linear inverse method [Visbeck, 2002], with an uncertainty of 1.5 cm s−1. Ship GPS positions and bottom track velocities from the LADCP itself are essential to reference the relative velocity profile when using the inverse method. The uncertainty of barotropic velocity reaches 2.0 cm s−1 at depths below 2000 m.
2.2.1. Extraction of Barotropic Motions
 There were 11 profiles of absolute velocity, temperature, salinity and pressure at each repeat-occupation station, and the time interval between 2 successive profiles was about 3 hours. In the SCS, the principal tidal constituents are semidiurnal and diurnal tides, and near-inertial motions are very energetic [Liang et al., 2005]. Thus, in the following we focus our analysis on the semidiurnal tides, diurnal tides and near-inertial motions. In order to obtain the barotropic subinertial flow, the velocity profiles u(zh, ti) were averaged (velocities involved here are all for zonal flow unless specified otherwise) over depth at each station that yields barotropic motions
We assume the following form for barotropic flow
where the symbol ‘bt’ indicates barotropic motions, m, k and f semidiurnal tides, diurnal tides and near-inertial motions respectively, i = 1, 2, 3, …, 11 the successive profiles at each repeat-occupation station, and Abtj, ωbtj, ϕbtj, (j = 1, 2, 3) the amplitude, frequency, initial phase of barotropic semidiurnal tides, barotropic diurnal tides and barotropic near-inertial motions, respectively. Here, the semidiurnal tide is M2, the diurnal tide K1, and the period of near-inertial motion is approximately 35 hours. For the near-inertial motion, the amplitudes of u and v are equal and the difference of initial phase between them is 90 degrees, which corresponds to the clockwise circulation in the Northern Hemisphere. The remaining coefficients are determined with the least square method, thereby providing solutions for the velocity of barotropic subinertial flow ubt0, the velocity of barotropic semidiurnal tides ubtm, the velocity of barotropic diurnal tides ubtk, and the velocity of barotropic near-inertial motions ubtf. Our calculation suggests that the amplitudes of semidiurnal tides are typically 6–10 cm s−1, diurnal tides 4–10 cm s−1, and near-inertial motions 1–4 cm s−1, in good agreement with earlier numerical studies [Yanagi and Takao, 1998].
2.2.2. Extraction of Baroclinic Motions
 Separating barotropic motions from velocity profiles u(zh, ti) yields baroclinic motions at each repeat-occupation station
We assume the following form for baroclinic flow at a fixed depth zh
where the symbol ‘bc’ indicates baroclinic motions, i = 1, 2, 3, …, 11 the successive profiles at each repeat-occupation station, and Abcj, ωbcj, ϕbcj, (j = 1, 2, 3) the amplitude, frequency, initial phase of baroclinic semidiurnal tides, baroclinic diurnal tides, and baroclinic near-inertial motions, respectively. The above coefficients are determined with the least square method, which then yields solutions for the velocity of baroclinic subinertial flow ubc0(zh), the velocity of baroclinic semidiurnal tides ubcm(zh), the velocity of baroclinic diurnal tides ubck(zh), and the velocity of baroclinic near-inertial motions ubcf(zh). The results reveal that the amplitudes of semidiurnal and diurnal tides range between 5 and 15 cm s−1 intensively, and near-inertial motions between 3 and 8 cm s−1, which agree with Niwa's three-dimensional numerical simulation results [Niwa and Hibiya, 2004]. The rms subinertial flow velocities are typically 10–15 cm s−1with uncertainties of ±1.5 cm s−1.
2.2.3. Subinertial Flow
 In order to get subinertial flow velocity across the Luzon Strait, we first interpolate the amplitudes and phases of tidal currents and near-inertial motions extracted at repeat-occupation stations to single-occupation stations. Then we obtain the flow velocities of barotropic tides, baroclinic tides, barotropic near-inertial motions, and baroclinic near-inertial motions at each station, which are symbolized as follows, respectively:
where s = 1, 2, 3, …, 15 indicates station sequence. Finally, the velocities of barotropic and baroclinic subinertial flow are obtained for each station
where at the repeat-occupation stations, ubtmes = ubt0, ubcmes = ubc0; at the single-occupation stations, ubtmes = ∫u(zh)dzh, and ubcmes(zh) = u(zh) − ubtmes. The subinertial flow discussed below equals ubts plus ubcs(zh).
3.1. Velocity Structure
3.1.1. Subinertial Flow
 The subinertial velocity field (Figure 2) shows that westward and eastward flows occur alternately in the Luzon Strait. In the upper layer (<500 m), the strongest westward flow, exceeding 25 cm s−1, occurs at a depth of about 100 m in the northern part of the strait, while the strongest eastward flow, exceeding 7 cm s−1, occurs at depth of about 100 m around 20°15′N. It is obvious that the flow is predominantly westward in the upper layer. Westward flow reaching about 7 cm s−1 also occurs in the southern part of the strait in the intermediate layer (between 500 and 1500 m), while eastward flow reaching10 cm s−1 is observed in the north at depths around 1000 m. In contrast with the upper layer, the eastward flow is dominant in the intermediate layer. Westward flow is widespread in the deep layer (>1500 m), with its maximum speed exceeding 7 cm s−1, while eastward flow is primarily confined to the northern part of the strait. The abyssal westward flow is forced by a baroclinic pressure gradient between the Pacific and the SCS [Qu et al., 2006].
 It is noteworthy that the vertical structure of subinertial flow may vary with time. Given the fact that most observations used for this study were conducted in the first half of October, around the transition season of monsoon, we feel that they likely provide reasonable representation of the annual mean condition. However, this needs to be investigated by further research.
3.1.2. Geostrophic Flow
 Applying the same technique described in sections 2.2 a and b to isopycnal displacements ξ(z) and assuming a reference level at 1500 db, we also obtain the geostrophic flow through the Luzon Strait, where isopycnal displacements ξ(z) were fit solely to the oscillatory components and defined in such a way that their time-mean is zero. The LADCP measurements at 1500 db were then used to obtain the absolute velocity through the Strait. The results indicate that the amplitudes of semidiurnal tides vary from 10 to 50 m, diurnal tides from 5 to 35 m and near-inertial motions from 5 to 20 m. The rms isopycnal displacements are 13–20 m. The uncertainties of the subinertial profiles are ±3 m. Compared with the subinertial flow shown in Figure 2, the westward geostrophic flow is more pronounced in the upper layer (Figure 3), especially in the southern part of the strait. The flow has two cores: one at about 19°N and the other around 21°30′N, both exceeding 20 cm s−1 in magnitude. In the intermediate layer, the geostrophic flow is westward in the south, primarily south of 20°20′N, and eastward in the north. The eastward flow has a maximum velocity of about 7 cm s−1, weaker by about 3 cm s−1 than the subinertial flow. In the deep layer, the flow is predominantly westward and its maximum velocity reaches 4 cm s−1. The eastward flow is observed mostly in the region between 20°20′N and 21°30′N, with a maximum velocity 7 cm s−1. In general, the geostrophic flow shows a pattern similar to the subinertial flow, and in a quantitative sense the discrepancies are likely due to errors in both deriving geostrophic calculation and LADCP measurement.
3.2. Volume Transport
 The volume transports from subinertial flow and geostrophic flow are shown in Table 1. The uncertainty of transport in each individual layer is about 1 ∼ 2 Sv, and the uncertainty in the whole depth range is about 3 Sv. The net volume transport of the subinertial flow through the Luzon Strait is 9 Sv (westward) in the upper layer (<500 m) and 5 Sv (eastward) in the intermediate layer (500–1500 m). Water enters the SCS again in the deep layer (>1500 m) with a net volume transport of 2 Sv. This result compares favorably with earlier observations [Wang, 1986; Liu and Liu, 1988; Qu et al., 2006] and confirms the hypothesis that the LST has a sandwiched vertical structure. The total net volume transport of subinertial flow through the Luzon Strait is 6 Sv, of which about one fifth is due to barotropic flow.
Table 1. Volume Transports at Different Depth Ranges of the Luzon Straita
500 to 1500 m
Units are Sv = 106 m3 s−1. Positive values indicate eastward transport.
Subinertial flow transport
Geostrophic flow transport
 The geostrophic flow shows the same vertical structure as the subinertial flow (Table 1), with a net volume transport of 10 Sv (westward), 5 Sv (eastward), and 3 Sv (westward) in the upper, intermediate, and deep layers, respectively. The total volume transport of geostrophic flow through the Luzon Strait is 8 Sv, which is 2 Sv larger than that of subinertial flow measured by LADCP.
 To gain an overview of water mass distribution, we also divide the volume transport into three density layers: an upper layer (<26.4 Kg m−3), an intermediate layer (26.4 to 27.4 Kg m−3), and a deep layer (>27.4 Kg m−3 ) (Figure 4). Here, the intermediate layer represents the North Pacific Intermediate Waters. Again, the vertical distribution clearly shows a sandwiched structure (Table 2), despite some quantitative discrepancies in some layers.
Table 2. Same as Table 1 Except at Different Potential Density Ranges
<26.4 Kg m−3
26.4 to 27.4 Kg m−3
>27.4 Kg m−3
Subinertial flow transport
Geostrophic flow transport
 The subinertial flow, geostrophic flow, and the volume transport through the Luzon Strait were examined using recently collected velocity and hydrographic data. Analysis of these data provides a high resolution picture of the subinertial flow and an estimate of volume transport through the Luzon Strait. The results clearly show a westward flow in the upper layer (<500 m), an eastward flow in the intermediate layer (500–1500 m), and a westward flow again in the deeper layer (>1500 m), confirming the earlier hypothesis that the Luzon Strait transport has a sandwiched vertical structure. The deepwater overflow through the Luzon Strait is 2 Sv.
 This study shows that the effect of non-steady flows, such as tidal currents and near-inertial motions, can be effectively removed from high-resolution observations, allowing us to obtain more reliable subinertial flow fields and volume transports. Some estimates obtained from this study may serve as a reference for future studies of the SCS.
 The authors would like to thank Martin Visbeck for kind help in processing the LADCP data, to Yanwei Zhang, Lifen Yang and Jianjun Kang for useful discussions, and Captain Liujia Jiang and the crew of R/VDongfanghong No.2 for enthusiastic support of the project. The authors wish to acknowledge the support from Key International S&T Cooperation Projects of China (project 2004DFB02700), National Natural Science Foundation of China (project 40552002), and National Basic Research Program of China (project 2005CB422303). T. Qu was supported by the National Aeronautics and Space Administration through grant NAG5-12756 and by Japan Agency for Marine-Earth Science and Technology through its sponsorship of the International Pacific Research Center (contribution IPRC-400). Two anonymous reviewers provided helpful suggestions that improved the manuscript.