Reflection seismic profiles show that drift sediments occur in the depth range of 1200–2800 m on the continental margin off Southeast China. These drift deposits were generated by upward flow near the Dongsha Islands of the North Pacific Deep Water (NPDW), which enters the South China Sea via the Bashi Channel (sill depth >2500 m). This flow results in deposition of slope sediments resuspended off East and South Taiwan on the slope southeast of the Dongsha Islands. At 1200 m water depth, this upward flow probably terminates and the NPDW mixes with the surface water. Sediment ages from ODP site 1144 southeast of the Dongsha Islands show that this process started at least 1.1 m.y. ago. Indications for mass wasting in the area of upslope water flow document the simultaneous occurrence of current- and gravity-controlled sedimentation processes. However, bottom current-related drift sedimentation dominates.
 The South China Sea (SCS) is a semi-enclosed sea bounded to the west by Vietnam, to the north by southeastern China and Taiwan, to the east by the Philippine arc and to the south by Malaysia and Indonesia (Figure 1). Connections with the East China Sea, the Pacific Ocean, the Sulu Sea, the Java Sea, and the Indian Ocean exist via several oceanic straits of which the Luzon Strait is the deepest (>2500 m). Shallow circulation in the SCS is dominated by the Southeast Asia monsoon system [Wyrtki, 1961]. In winter, the monsoon is directed southwestward, while in summer it reverses to northeastward. The resulting wind stress curl controls the shallow circulation [e.g., Wyrtki, 1961; Chao et al., 1996; Qu, 2000]. Moreover, circulation in the northern SCS is significantly affected by the flow of the Kuroshio Current through the Luzon Strait [e.g., Metzger and Hurlburt, 1996, 2001; Yang et al., 2002].
 Our study area is located south of Hong Kong on the continental margin of Southeast China centred surround the Dongsha Islands (Figure 1). Here, three cruises with the German research vessel Sonne were carried out in the years 1987 (SO-50B), 1990 (SO-72A) and 1994 (SO-95), and over 6,600 km of multi-channel reflection seismic data and high-resolution 3.5 kHz echograms were collected. During cruise SO-95, the seismic equipment used consisted of a GECO-PRAKLA mini-streamer (active length 100 m) and three GECO-PRAKLA air guns (total volume 5 l). The seismic data were recorded digitally and later processed with a commercial software package.
 The continental margin of the northern SCS is characterised by a complex tectonic history dominated by the collision of Taiwan with the Chinese continent, resulting in uplift of the Dongsha region at the Mio-Pliocene boundary and in the lower Middle Pleistocene [Lüdmann and Wong, 1999]. The present depositional environment is controlled largely by strong bottom currents and by the Pearl River that reached the shelfedge west of the Dongsha Islands during sea-level lowstands [Lüdmann et al., 2001]. Although shallow circulation in the SCS has been repeatedly studied [e.g., Metzger and Hurlburt, 1996; Shaw and Chao, 1994; Yang et al., 2002], information on the deep circulation pattern is still sparse. Upwelling of deep water is known only off Vietnam (especially in summer) and the coast of Luzon [e.g., Chao et al., 1996; Wyrtki, 1961]. Generally, four major water masses can be distinguished in the SCS: (1) surface water (0–300 m); (2) intermediate water (300–1000 m); (3) upper deep water (1000–2500 m); and (4) lower deep water (>2500 m) [Wyrtki, 1961].
 The occurrence of drift sediments has been reported from different continental margin settings [e.g., Locker and Laine, 1992; Stoker et al., 1998; Wong et al., 2003]. The study of these deposits provides insight into the action of bottom currents which can erode and transport vast amounts of sediment along the margin. In this paper, we present evidence for intrusion of the NPDW into the SCS through the Luzon Strait as manifested by the occurrence of drift deposits on the continental margin of Southeast China.
 A seismo-stratigraphic interpretation of our seismic reflection profiles provides clear evidence for drift sedimentation at the northern continental margin of the SCS. Seismic profile 20, which runs parallel to the continental margin, crosses a drift body on the continental slope off the Dongsha Islands (Figure 2). This mounded drift occurs at water depths between 1200–2800 m in the northeastern part of a depression-like structure and is accompanied by three moats (channel-like structures), to which bottom currents are confined. The seismofacies of the drift is characterised by a succession of wavy reflectors of high-to-medium amplitude in its upper part and medium-to-low amplitude in the lower part, where the reflectors intercalate with transparent layers. The continuity of the reflectors is generally high, although it decreases with depth. The mounded, sigmoidal reflector pattern and the lack of a second levee clearly differentiate the observed drift from a turbidity channel-levee system. The drift consist generally of two depositional (seismic) units separated by an erosional unconformity. The stacking pattern of the layers in these units suggests an upslope progradation. The units have internal discontinuities (Figure 2), which typically arise from erosive episodes of increased bottom current intensity, although they could also be a result of marked changes in grain size or composition accompanying small fluctuations in the current regime, sediment supply or drastic changes in the water chemistry [Faugères et al., 1999]. The termination of aggradation of sediment layers in a depositional unit is accompanied by a significant shift of the moat from NE to SW (Figure 2), leading to a southwestward younging of the units. The southwestern part of the depression is dominated by erosional truncations or scarps; here, reflectors crop out directly at the seafloor. Locally, sediment slides may overlie the drift.
 Another drift body crossed by two of our seismic lines is located southeast of the Dongsha Islands (Figures 3a and 3b). Profile 10, a dip profile, shows a lenticular drift body with a length of over 23 km and a convex-upward geometry (Figure 3a). Three moats marked by channel-like incisions are located on the northwestern flank of the drift. The same drift body is crossed by the margin-parallel profile 20 (Figure 3b). Its seismic facies is identical with that of the drift body to the northeast (Figure 2). It consists of three depositional units marked by a distinct downslope migration of their moats (Figures 3a and 3b). Associated with this migration is a significant lithological change at 252 mbsf at ODP site 1144 from iron sulphide-rich clay to siliceous biogenic clay devoid of iron sulphide [Shipboard Scientific Party, 2000]. It occurs at the base of depositional unit 2 at 264 ± 11 mbsf (average p-wave velocity 1700 ± 50 m/s) (Figure 3b). The moats are restricted to the northeastern part of the depression, while erosional truncations dominate in the southwest (Figure 3b) analogous to the drift body to the northeast (Figure 2). In places, mass wasting deposits such as debris flows may be intercalated (Figure 3b). The sediment drifts are oriented WNW-ESE oblique to the trend of the continental margin (Figure 4), and downslope migration of the drift is perpendicular to the direction of current flow.
 The drift sediments observed on the continental margin of Southeast China are restricted to two limited areas southeast of the Dongsha Islands (Figure 4). The lithostratigraphy of ODP site 1144 shows that the siliciclastic fraction of the sediments consists mainly of clay and silt [Boulay et al., 2003]. The drift nature of the deposits is indicated by high sedimentation rates, the geometry and morphology of the drift and its surficial and internal structures [Shipboard Scientific Party, 2000].
3. Discussions and Conclusion
 We interpret the lenticular depositional bodies with their mounded external geometry mapped near the Dongsha Islands as giant elongated drifts [Faugères et al., 1999]. Bottom currents confined to moats are presumably responsible for the transport and construction of these drift bodies. These currents flow to the WNW upslope (see orientation of moats, Figure 4) because the moats are located on the NE flank of the NW-SE striking depressions. If the currents flow downslope in the ESE direction, it would have been deflected to the right by the Coriolis force, so that the moats would be on the SW depression flank. The overall downslope migration of the moats is difficult to explain. We speculate that this may be due to an abrupt decrease in current speed, so that drift sedimentation starts lower on the slope and the net Coriolis deflection to the northeast is less, resulting in a southwestward moat migration downslope.
 The only source for a current approaching from the east is the Luzon Strait (Figure 4). Here, the Bashi Channel with a sill depth of around 2500 m provides a possible pathway for deep water masses to ventilate into the SCS. This is corroborated by the suggestion that the Luzon Strait transport has vertically a sandwich-like structure with an inflow from the Pacific in the upper and deeper layers and an outflow from the SCS in the intermediate layer [e.g., Chao et al., 1996; Yuan, 2002; Qu and Lindstrom, 2004]. However, the water mass in question is probably not fed by the Kuroshio Current that flows northward along the eastern Philippines and enters the SCS through the southern Luzon Strait, because otherwise considerable sinking (>1000 m) and vertical mixing would be necessary. In fact, a CTD transect off northeastern Luzon points to the existence of a southward flowing current (namely the Luzon Undercurrent) beneath the Kuroshio at water depths of 500 to >2500 m [Qu et al., 1997]. Thus, we conclude that intrusion of deep waters into the SCS must take place through the Bashi Channel. However, because of the NNE-SSW orientation of this channel, water inflow can only occur from the north (Figure 4). Direct current measurements at the centre of the Bashi Channel in water depths of about 2500 m show an inflow of cold Pacific deep waters with a volume transport of 1.2 Sv (Figure 4) [Liu and Liu, 1988]. Liu and Liu  suggested a largely tidal origin for this current, which is strong enough (about 14 cm/s) to erode unconsolidated mud and to transport clay and silt in suspension [Nichols, 1998, Figure 4.5]. Although the current speed within the narrow channel is higher than that outside the channel, it still appears likely that this current is responsible for the deposition of the clayey and silty drift deposits south and southeast of the Dongsha Islands.
 Our hypothesis of inflow of the NPDW is consistent with results from tracer experiments over the course of a year and their interpretation as indicators for upwelling of a Pacific water mass to a depth of about 900 m southwest of Taiwan [Chao et al., 1996]. Recently, Qu and Lindstrom  postulated an inflow of oxygen-rich Pacific deep water around the sill depth of the Luzon Strait. Additional supporting evidence comes from provenance analyses of rare earth abundances in sediment samples from ODP site 1144 [Shao et al., 2001], which suggest Taiwan as the source area rather than the Pearl River or the Philippines (Figure 4). They speculated that sediments were transported from Taiwan via the Taiwan Strait onto the slope off the Dongsha Islands (Figure 4). However, this scenario is consistent neither with the WNW-ESE trend of the moats nor with the prevailing direction of surface currents in the Taiwan Strait which flow year-round northward, although in winter there is an additional southward flow along the east coast of China, resulting in a zero to slightly northward net flow [Chen, 2003; Liang et al., 2003].
 We postulate that the southward flowing NPDW resuspends slope sediments off eastern and southern Taiwan and transports the fine fraction through the Bashi Channel into the SCS. After passing through the narrow elongated channel, the pathway widens drastically and the water mass is literally “injected” into the SCS Basin. Here, the current possibly sinks and is deflected to the west by the Coriolis effect, following the seafloor topography. At the continental slope of Southeast China, it is pushed upslope, where it slows down and deposits its load at water depths of 1200–2800 m. By 1200 m depth, the kinetic energy is probably reduced enough for mixing with surface waters to occur. The surface waters are monsoon-controlled and flow parallel to the continental margin. In our study area, they result in erosion and sediment reworking at water depths of 200–1000 m around the Dongsha Islands (Figure 4). Lüdmann et al.  demonstrated that in a large area surrounding the Dongsha Islands, Holocene deposits are absent and that fields of sediment waves occur on the upper slope. Our seismic profiles suggest that in addition to upslope sediment transport, downslope processes may also occur. The erosional truncations observed on profile 20 (Figures 2 and 3b) are probably due to mass wasting processes such as slumping and sliding.
 We conclude that deposition on the continental slope southeast of the Dongsha Islands is mainly controlled by upward flow of the NPDW, with monsoon-induced circulation playing a secondary role. That NPDW intrusion is independent of sea-level fluctuations accounts for the overall extremely high sedimentation rates observed near ODP site 1144 (core 17940: about 700 m/m.y. in the Holocene and around 400 m/m.y. during glacial periods [Wang et al., 1999]). Since the ODP drill hole has not reached the lower boundary of the drift body (Figure 3b), upward flow of the NPDW must have lasted at least 1.1 m.y. [Boulay et al., 2003]. Further studies are necessary to determine the distribution and detailed evolution of the drift sediments as well as the complex water circulation pattern and sedimentation processes in the northern SCS.
 We gratefully acknowledge the unfailing help of the captain, officers and crew of the R/V Sonne during the three South China Sea cruises reported here. Our thanks are due to Prof. Shao Lei (Tongji University, Shanghai) for fruitful discussions and to two anonymous reviewers for their constructive criticisms which provided essential help in improving our manuscript. This work was funded by the German Federal Ministry of Education, Science, Research and Technology (Projects MFG0052, 03G0072A and 03G0095A).