Late Quaternary Deep-Water Circulation in the South China Sea

A suite of four new benthic foraminiferal ( Cibicidoides wuellerstorfi) stable carbon isotope records produced from South China Sea (SCS) sedi­ ment cores ranging in water depth from 1515 and 3766 m are used here to evaluate deep-water circulation in the SCS over the last complete glacial­ interglacial cycle (i.e., the last 200 kyrs). A common pattern of benthic (i13C variations is observed between cores, with pronounced o 13C deple­ tions recorded during glacial periods and o 13C enrichments during interglacials. The overall similarity of o 13C values between sites indicates a homogeneous water mass below 1500 m in the SCS basin through time, except for the interval between 120 and 65 ka. Temporal changes in venti­ lation and exchange between the SCS and western Pacific are indicated by comparison with records from the Ontong Java Plateau (OJP). The pres­ ence of Pacific Intermediate Water can be inferred based on the bathymet­ ric gradient of o13C between SCS and OJP (Llo13C). The Llo13C patterns do not appear to be correlated with glacial-interglacial cycles. However, the Llo180 is greater in warm intervals (late MIS 4 and early MIS 3; MIS 1) than during cold intervals. (


INTRODUCTION
The distribution of o13C of LC02 in the surface ocean is largely controlled by biological activity and air-sea exchange, whereas seawater o 13C below the permanent thermocline is largely a function of deep circulation and the effects of remineralization of sinking organic material (Kroopnick 1980(Kroopnick , 1985)).Forarniniferal o13C measurements made on benthic taxa were the first applied to paleo-reconstructions, specifically to infer past variability in deep water circulation patterns.This is because the processes that control deepwater o 13C are less 322 TAO, Vol. 14, No. 3, September 2003 complicated than those at the surface.In recent decades, a large number of investigations have utilized o 13C records of benthic foraminifera as deepwater paleoceanographic tracers (e.g., Broecker 1982;Shackleton et al. 1983;Duplessy et al. 1988;Oppo and Fairbanks 1987;Oppo et al. 1990;Mix et al. 1991;Mackensen et al. 1993;Bicker! and Wefer 1999).In most of these studies, the benthic foraminifer species Cibicidoides wuellerstorfi and related taxa have been measured since they appear to calcify their tests with respect to o 13C composition close to that of the ambient bottom-water LC0 2 (e.g., Zahn et al. 1986).Subsequently, the "Mackensen effect" was proposed to explain the additional depletion of o 13C that can occur as a result of high surface productivity and the additional input of 12C from the decomposition of organic matter on the sea floor (Mackensen et al. 1993).
Deep-water circulation is closely related to climate change (e.g., Mix et al. 1991;Oppo et al. 1997;Mcintyre et al. 1999).A longstanding paleoceanographic question is whether, dur ing glacial periods, the North Pacific may have been a source of deep water (e.g., Berger 1987;Curry et al. 1988;Duplessy et al. 1988;Ohkouchi et al. 1994;Jiau et al. 1997;Keigwin 1998) As a marginal basin along the western edge of the Pacific, the South China Sea (SCS) is separated from the western Pacific and Indian Oceans by a series of shallow sills that range in water depth from 36 to 2200 m.The SCS is characterized by sedimentation rates than are an order of magnitude higher than found in most of the Pacific and carbonate sediments are widespread.The SCS thus offers a special opportunity for reconstructing deepwater paleoceanography in the Pacific.However, only a few studies to date have addressed glacial deep circulation in the SCS based on benthic foraminifer assemblages (e.g., Jiau and Wang 1997; Jian et al. 1999) and the isotopic compositions of benthic foraminifera (e.g., Winn et al. 1992;Wang et al. 1999).The present study focuses on benthic o13C records from different water depths in the SCS basin and on their comparison with the western equatorial Pacific record to infer past variations in the ventilation of deep water.

MATERIALS AND METHODS
A total of four benthic stable isotope records were generated for this study.The locations of cores that were used are indicated on the bathymetric map (Fig. 1), and information regard ing the geographic coordinates, water depth and core length are listed in Table 1.Benthic foraminifers of the species Cibicidoides wuellerstorfi were analyzed both at the Leibniz Labo ratory of Kiel University and at the University of California, Davis.Results are presented relative to the standard PDB-scale and are reported with a precision better than 0.07 %0 for o 180 and 0.06 %0 for o 13C.

Distribution of 8180 and o"C in the Water Column
Before beginning the downcore study, the isotopic compositions of C. wuellerstorfi from   ........ .: a                       -0.5 0.0 0.5 1.0 1.5 2.0  0  : �:179 50 Recent coretops were compared with the distribution of o 180 and o 13C in the water column (Fig. 2).The C. wuellerstorfi coretop data were taken from Grothmann (1996) while the water column data are based on our own measurements.Bemis et al. (1998) concluded that Cibicidoides precipitates its shell close to oxygen isotopic equilibrium with seawater and that the expected equilibrium values can be estimated by: T(°C) = 16.5 (±0.The o 13C compositions of C. wuellerstorfi from all coretops are close to the measured o13C values of ambient seawater 2,C02 (open circles on Fig. 2; Lin et al. 1999).Interestingly, the coretop o 13C values are closer to the direct measurements of 2,C02 than to the calculated values of calcite o13C in equilibrium (dotted diamonds on Fig. 2).The calculated values of calcite o13C are based on Grossman (1998)  Coretop values of benthic o 180 from the four SCS cores appear to reflect bottom water temperature; the deeper the core is situated, the heavier the o 180 that is recorded (Fig. 3).For example, there is a systematic 0.2-0.4%0 shift between Cores 17954 and 17950 throughout the two records.Furthermore, the range of o 180 variation in Core 17950 is very similar to that of ODP Site 806B, indicating a similarity of the watermasses at these two locations.While there is no significant observed difference in o 180 amplitude between the last glacial maxi mum (LGM) and Holocene for all individual cores, the o 180 gradient (Lio 180 1 _0) between the intermediate (ODP 806B) and deep water (S05-37 and -29 KL) cores, however, is quite dif ferent through the glacial cycles.The Ll o 180 1 _0 is generally reduced in cold intervals relative to warm periods, i.e., it is <0.2 %0 during the LGM in contrast to >0.5 9/oo in the Holocene glacials (wann periods) and lighter 8 "C values during glacials (cool periods; Fig. 4).However, the magnitude of the 8 "C depletion during glacials decreases progressively from MIS 6 to MIS2.The amplitude of glacial-interglacial 813C fluctuations from deeper cores (i.e., >2000 m; Cores S050-37 and -29 KL; -0.7 %0) are generally larger than those from shallower cores, which display changes more in line with the global average 8 "C shift of 0.46 '.Yoo (Curry et al. 1988).The hydrographic gradient reflected by the L\8180 1 _ 0 contrasts discussed above was not found in the 81'C records.Instead, the Ll813C gradient between the intermediate (ODP Site 806B) and deep water (S050-37 and -29 KL) cores are more or less consistent (0.2-0.3 %0) throughout the glacial cycles.One interesting feature is the pronounced but intermittent 8 13C depletion exhibited in the deeper sites (>2000 m; Cores S050-37 and -29 KL) during the Last Interglacial (MIS 5), especially the conspicuous 8 "C low near the end of MIS 5 that seems to be contemporaneous between cores.In addition, the two highly 8 "C-depleted intervals around 65-60 ka and 40 ka recorded in Core 17950 are also seen in the records of the other cores, though with much weaker signals (Fig. 4 ).The 8 "C event around 65-60 ka has also been documented at one other core from the southern margin of the basin (Core 17961; Wang et al. 1999) as an abrupt and pronounced 0.5 '.Yoo drop.Unfortunately, due to the relatively short core lengths of S050-37 and -29KL, a complete deep water 813C record from the SCS for MISS was not possible in this study.

Sill Depth in the Luzon Strait
As a marginal basin surrounded by shallow sills that are mostly <50 m in water depth, deep water in the SCS is largely derived from the western Pacific through the deep passage at Luzon Strait between Taiwan and Luzon.The depth of the deepest channel along the Luzon Strait, however, has been variously reported as shallow as 1900 m (Broecker et al. 1986) to as deep as 2500-2600 m (J ian and Wang 1997;Wang 1999).Hence, the depth of the controlling sill adopted in this study is critical to any further discussion since we are comparing our SCS data with a record from the western Pacific that comes from a depth of 2520 m (ODP Site 806B; Bicker! and Wefer 1999).Hydrographic observations were used as the primary basis for our determination of sill depth for the following discussion.Water properties in the SCS, including potential temperature and salinity, remain fairly constant below 2000 m and are very similar to those in the western Pacific (Gong et al. 1992).A sill depth of 2000 to 2200 m was therefore taken as the critical threshold separating the SCS from the western Pacific in this study.Thus, both the 8180 and 813C signals from cores deeper than 2200 m (i.e., S050-37 and -29KL) in the SCS are expected to be analogous to those from the western Pacific.ODP Hole 806B from the OJP hence serves as an appropriate reference for comparison and for the following discussion.

Deep-water Circulation in the South China Sea
The benthic 8 180 and 813C records from the SCS are best considered in the context of data from the western Pacific.Figure 5 illustrates the covariance patterns of benthic 8180 and 813C values from the suite of cores.The upper panel contains data from the cores shallower Hui-Ling Lin to 8180.This shift can be ascribed to a temperature effect resulting from the shallower water depth of Core 17954.In contrast, there is no clear offset in the 8 13C data between cores; indeed, the similar range of 8 13C values among the three cores implies a close relationship between the SCS and the western Pacific in terms of ventilation in the mid-depth range of the controlling sills.Deep water renewal in the SCS has been reported to be rapid based on radio carbon measurements (Broecker et al. 1986(Broecker et al. , 1988)), with a residence time estimated between 40 and 120 years according to calculations of influxes and basin volume (Gong et al. 1992).
Even though deep water of the SCS below the sill depth of the Luzon Strait is provided by the western Pacific, significant deviations in 8 180 -8 13C between the deep SCS and the OJP are clearly revealed in the lower panel of Fig. 5.Both the 8 180 and 8 13C of benthic forarnini fers in the deep part of the SCS (37 and 29 KL) are enriched relative to those from the OJP.
The SCS enrichment of 8180 (-0.4 o/oo) relative to the OJP indicates slightly cooler deep water in the SCS during glacial time (about l.9°C, according to the equation of Bemis et al. 1998), not counting for possible salinity effects.A 1.1 °C cooling of deep water has been suggested as an upper limit in the western equatorial Pacific (Birchfield 1987;Bicker! et al. 1993).The additional glacial cooling in the SCS compared with the open ocean is possibly due to the enclosed-basin hydrographic setting induced by lowered sea level during the glacial (Wang 1999).Wang et al. (1999) have inferred that an extreme oxygen-minimum layer developed near the southern margin of the SCS during glacials based on low epibenthic 8 13C signals, leading to an estuarine stratification of the surface water.However, data from this study do not seem to support this hypothesis.A different Ll813C 1 _ 0 pattern between glacial periods would be ex pected if the deepwater circulation mode changed substantially in response to climate.The large negative 813C excursions observed in the SCS between 70 and 60 ka (late MIS 4 and early MIS 3 for deep cores) relative to ODP Hole 806B were not repeated in the SCS during the LGM.It therefore seems more likely that these depleted events can be attributed to the "Mackensen effect" as a result of episodes of increased surface productivity (Mackensen et al. 1995).On the other hand, some evidence for different deepwater masses in the SCS can per haps be found in the different L\8180 1 _ 0 observed between glacials and interglacials.
. Today, deep water formation in the North Pacific is suppressed by the low salinity of the surface waters.Unfortunately, relevant research into potential sources of North Pacific Deep Water is restricted by poor carbonate preservation in most of the Pacific Ocean.

+
Fig. 2. Vertical profiles of i5180 -i513C in the northern South China Sea shown in comparison to the C.wuellerstorfi coretop data ofGrothmann (1996).

�
2) -4.80 (±0.16)* ( &-0w ).The estimated o 180 plotted on Fig. 2 were calculated based on the above equation and using field data for temperature and o 180 (i.e., 0w in equation; open squares on Fig. 2) from the northeast SCS (unpublished data; Station E in Lin et al. 1999).Overall, the coretop data are in good agreement with the calculated o 180 equilibrium values, supporting the basis for the following discussion.
, as modified fromEpstein et al. (1953): o13C = ( o 13C of 2,C02) + [10.51-(2980ffw+273)].The o 13C of C. wuellerstorfi from coretops fall well within the range of o 13C in the water column at depths below 1500 m, except for the data point enclosed in parentheses which comes from a location near the mouth of the Pearl River.The similarity between coretop o 13C of C. wuellerstoifi and o 13C i:co 2 below 1500 m water depth endorses the feasibility of using C. wuellerstorfi o 13C for the downcore records.
3.2 o 180 and o 13C StratigraphyDowncore variations in the o 180 and o 13C compositions of C. wuellerstorfi for the four study cores are plotted in Figs.3 and 4in ascending order of water depth, together with the record from ODP Hole 806B for reference.ODP Hole 806B was drilled on the Ontong Java Plateau (OJP) in the western equatorial Pacific (0°19.1 'N, 159°21.TE) and is currently bathed by Pacific Deep Water(Craig et al. 1981 ).The chronology for individual cores is based on their o 180 stratigraphy and the vertical and horizontal axes are scaled the same for easy comparison.Cores S050-37 and -29 KL did not penetrate marine isotope stage 5e (MIS 5e;Winn et al. 1992) and hence do not span the complete last glacial-interglacial cycle.

(Fig. 3 .Fig. 4 .
Fig. 3. Time-series of the stable oxygen ( 8 180) isotope composition of C.wuellerstorfi for the four SCS cores and for ODP Hole 806B plotted in order of coring water depth.Superimposed on the individual SCS core records is the record of ODP Hole 806B for comparison.

Fig. 5 .
Fig. 5. Crossplots of benthic o 180 and o 13C values from the set of cores to examine covariance relationships.The upper panel includes cores from depth shallower than ODP Hole 806B (17954 and 17950) while the lower panel shows cores deeper (37 and 29 KL) than ODP Hole 806B.In both panels, the data from ODP Hole 806B are shown for comparison.

Table 1 . Location, water depth, and core recovery for cores discussed in this study
Winn et al. (1992) (1992)