The East Asian winter monsoon variability in response to precession during the past 150 000 yr

Abstract. The response of the East Asian winter monsoon variability to orbital forcing is still unclear, and hypotheses are controversial. We present a 150 000 yr record of sea surface temperature difference (ΔSST) between the South China Sea and other Western Pacific Warm Pool regions as a proxy for the intensity of the Asian winter monsoon, because the winter cooling of the South China Sea is caused by the cooling of surface water at the northern margin and the southward advection of cooled water due to winter monsoon winds. The ΔSST showed dominant precession cycles during the past 150 000 yr. The ΔSST varies at precessional band and supports the hypothesis that monsoon is regulated by insolation changes at low-latitudes (Kutzbach, 1981), but contradicts previous suggestions based on marine and loess records that eccentricity controls variability on glacial–interglacial timescales. Maximum winter monsoon intensity corresponds to the May perihelion at precessional band, which is not fully consistent with the Kutzbach model of maximum winter monsoon at the June perihelion. Variation in the East Asian winter monsoon was anti-phased with the Indian summer monsoon, suggesting a linkage of dynamics between these two monsoon systems on an orbital timescale.


Introduction
The Asian monsoon is the largest monsoon system on Earth (e.g.Wang et al., 2003).It consists of the East Asian and Indian monsoons and is dynamically linked with the Australian and African monsoons.While the Indian summer monsoon winds are stronger than the East Asian summer monsoon winds, the East Asian winter monsoon winds are much stronger than the Indian winter monsoon winds (Wang et al., 2003).In boreal winter, the winter monsoon occurs over East Asia and the adjacent marginal seas and is directly linked to the Australian summer monsoon through crossequatorial flows over the South China Sea (SCS) (Wang et al., 2003).
Long-term changes in the Asian monsoon are an important topic of paleoclimatology.The Asian monsoon responds to insolation changes at low latitudes on theoretical ground, which is regulated by precession, and hence it has been assumed to respond to precessional forcing (Kutzbach, 1981).According to this hypothesis, the summer monsoon is maximized when the Northern Hemisphere summer insolation is maximal on precessional cycle.However, the periodicity and phase of the monsoon variability in proxy records were inconsistent.Clemens and Prell (2003) argued that the Indian summer monsoon varied at obliquity and precession bands and is strongest at the November perihelion on precession cycles.In contrast, based on marine records, it was suggested that monsoon intensity is regulated by glacial conditions (e.g.Huang et al., 1997a, b;Jian et al., 2001;Steinke et al., 2010).In the last glacial period, the summer monsoon was weaker, and the winter monsoon was stronger.A paleoproductivity record from the Sulu Sea suggested that the East Asian winter monsoon was stronger in glacial periods (de Garidel-Thoron et al., 2001).Chinese loess records have suggested that the East Asian winter monsoon was stronger in glacials than in interglacials, and the variability shows strong eccentricity cycles (Ding et al., 1995;Xiao et al., 1995).Thus far different proxies and archives provided different conclusions on the periodicity and the phase of the Asian monsoon variability.
The South China Sea is a part of the Western Pacific Warm Pool (WPWP) region, but the winter sea surface temperature in the SCS is significantly lower than that in other regions of the WPWP.An oceanographic study demonstrated that this winter cooling is caused by the winter monsoon: winter monsoon northerly winds cool the surface water in the northern margin and advect the cooled water southward (Liu et al., 2004).Using this relationship, winter monsoon variations can be reconstructed by generating sea surface temperature (SST) records for the SCS.Shintani et al. (2008) discussed winter monsoon variation over the past 23 000 yr.The temperature difference between the northern SCS and the Sulu Sea suggests that the winter monsoon was stronger in the last deglaciation, in particular, during the Younger Dryas period, and that the intensity in the last glacial maximum (LGM) was the same as that during the middle Holocene.Tian et al. (2010) compared Mg/Ca-derived SSTs between the northern and southern SCS and showed that the SST gradient was higher in Heinrich 1 and the Younger Dryas periods.Huang et al. (2011) reported that SST gradient at the western and eastern margins of the southern SCS was higher in Heinrich 2, Heinrich 1 and Younger Dryas periods, suggesting enhanced East Asian winter monsoon activity in these periods.This strategy is useful for understanding the Asian winter monsoon variation, but a longer record is necessary to understand the response of the Asian winter monsoon to orbital forcing.
Many paleoceanographic studies of SST have been conducted in the SCS.Wang and Wang (1990) and Wang et al. (1995) reconstructed summer and winter SSTs in the SCS based on foraminifer assemblages, and found that during the LGM, the SCS experienced larger seasonal SST differences and a steeper latitudinal SST gradient than it does presently.Intense glacial cooling in the northern SCS was reported based on foraminifer-and alkenone-based SST records (Huang et al., 1997a, b;Chen and Huang, 1998;Pelejero et al., 1999;Chen et al., 2003).Kienast et al. (2001) found a millennium-scale temperature variation that mimics Greenland ice core records.Oppo and Sun (2005) and Zhao et al. (2006) reported millennium-scale temperature variations in the northern and southern SCS, respectively, for the last two glacial-interglacial cycles.The glacial-interglacial contrast of SST in the SCS have been attributed to either the inflow of cold water from the North Pacific (e.g.Wang and Wang, 1990;Wang et al., 1995) or changes in the winter monsoon intensity (e.g.Huang et al., 1997a, b).Disagreement regarding estimated paleotemperatures in the SCS was reported among different proxies such as alkenone U K' 37 , the foraminiferal Mg / Ca ratio, and the transfer function.This is potentially attributable to differences in the season and depth that each proxy reflects (e.g.Steinke et al., 2008).Alternative approaches such as TEX 86 are useful for better understanding paleotemperature changes in the SCS (Shintani et al., 2011).
TEX 86 is a recently developed paleotemperature proxy (Schouten et al., 2002) based on glycerol dialkyl glycerol tetraethers (GDGTs).The TEX 86 paleothermometer has the advantage that it does not appear to be influenced by salinity (Wuchter et al., 2004), and it is more sensitive to temperature changes in tropical waters (Kim et al., 2010) compared with the U K' 37 method (Pelejero and Grimalt, 1997).TEX 86 has been used for paleotemperature estimations in the northern SCS (Shintani et al., 2011;Li et al., 2013).Shintani et al. (2011) assumed that TEX 86 reflects SST weighted towards warmer seasons, but Li et al. (2013) subsequently assumed that TEX 86 reflects subsurface temperature, based on the results of a surface sediment study (Jia et al., 2012).In the shallow water region of the SCS, Zhang et al. (2013) suggested that TEX 86 records temperature during cooler season.
Here, we present a 150 000 yr record of TEX H 86 -derived temperatures from southern SCS core MD97-2151 offshore Vietnam.We calculated SST difference between the SCS and the central WPWP to understand the East Asian winter monsoon variability and its underlying mechanisms during the past 150 000 yr.

Oceanographic settings
The SCS is a marginal sea of the North Pacific with seven connections through the Taiwan Strait to the East China Sea (sill depth ∼ 70 m), the Bashi Strait to the North Pacific (sill depth ∼ 2500 m), the Mindoro and Balabac Straits to the Sulu Sea (sill depths ∼ 450 and ∼ 100 m, respectively), the Malacca Strait to the Indian Ocean (sill depth ∼ 30 m), and the Gaspar and Karimata Straits (∼ 40-50 m) to the Java Sea (Fig. 1; Wyrtki, 1961).The surface circulation in the SCS is driven by large-scale, seasonally reversed monsoon winds (Wyrtki, 1961;Hu et al., 2000).In the boreal summer, southwesterly winds drive an inflow of Indian Ocean water through the Sunda Shelf and a clockwise surface circulation in the SCS.In the boreal winter, northeasterly winds drive an inflow of North Pacific water through the Bashi Strait, an inflow of East China Sea water through the Taiwan Strait, and a counterclockwise surface circulation in the SCS.The southern SCS is part of the WPWP region, but the winter surface temperature is significantly lower than that in the central WPWP.A recent oceanographic study demonstrated that  (Liu et al., 2004) and (lower) the locations of cores MD97-2151 and MD97-2140 and ODP Site 806 and the annual mean sea surface temperatures in the entire western Pacific warm pool (Locarnini et al., 2010).
this winter cooling is caused by the winter monsoon.Winter monsoon northerly winds cool the surface water in the northern margin of the SCS and advect the cooled water southward (Liu et al., 2004).The SST in the study region is governed by the winter wind strength and the effect of cooling reaches 2 • C in the study area (Huang et al., 2011).Summer monsoon southerly winds develop an anticyclonic eddy and cool the surface water by the advection of coastal cool water and upwelling in the study area, and the effect of cooling is maximum 1 • C (Xie et al., 2003).The present-day SST at the study site shows a seasonal variation between 26.1 • C in January and 29.6 • C in May, with a mean annual value of 28.0 • C (Fig. 2; Locarnini et al., 2010).

Samples and age-depth model
During the IMAGES 1997 Marion Dufresne cruise, a giant piston core (MD97-2151; 26.72 m long) was collected from a water depth of 1598 m on the southwestern slope of the SCS at 8 • 43.73 N; 109 • 52.17 E (Fig. 1).The sediments consist of olive to dark gray silty clay with nannofossils, foraminifera, and diatoms (Chen et al., 1998).Locarnini et al., 2010).Months "J" to "D" means January to December.
An age model in calendar years (Fig. 3; Table 1) was created by oxygen isotope stratigraphy (Lisiecki and Raimo, 2005) of benthic foraminifera Cibicidoides wuellerstorfi (Zhao et al., 2006), and by the AMS 14 C ages of 13 samples of the planktonic foraminifera Globigerinoides sacculifer.The 14 C age at 699.5 cm was not used for the age-depth model.The calendar age was converted from the AMS 14 C age using the CALIB6.1.0program and marine09 data set (Reimer et al., 2009) with a 400 yr global reservoir correction.Larger carbon reservoir ages have been suggested in the SCS for the last glacial (e.g.Wang et al., 1999;Lin et al., 2013).It is, however, not easy to turn the knowledge to improve the age-depth model in the SCS core.We thus used the constant local reservoir effect of R = 0.The deviation of carbon reservoir age throughout several thousand years does not change significantly the entire variation of SST during the past 150 000 yr.The Young Toba Tuff appeared at 1556 cm in core MD97-2151, and its age was assigned 75.1 ka in this age-depth model, which is consistent with recent dating results (75.0 ka; Mark et al., 2013).
A total of 178 samples were collected every 15 cm on average (equivalent to approximately 840 yr intervals) down to a depth of 26.7 m (150 ka).

Alkenone analysis
Gas chromatography (GC) of F3 (alkenones) was conducted using a Hewlett Packard 5890 series II gas chromatograph with on-column injection and electronic pressure control systems, and a flame ionization detector.Samples were dissolved in hexane.Helium was used as the carrier gas at 30 cm s −1 .A Chrompack CP-Sil5CB column was used (60 m × 0.25 mm i.d.; film thickness, 0.25 µm).The oven temperature was programmed to rise from 70 to 290 The alkenone unsaturation index U K' 37 was calculated from the concentrations of di-and tri-unsaturated C 37 alken-2ones ([C 37 : 2 MK] and [C 37 : 3 MK], respectively) using the following expression+ (Prahl et al., 1988): Temperature was calculated according to the equation where T is temperature [ • C] based on core-top calibration in the SCS (Pelejero and Grimalt, 1997); analytical accuracy (standard deviation in a replicate analysis) was 0.24 • C in our laboratory.

Glycerol dialkyl glycerol tetraether (GDGT) analysis
Glycerol dialkyl glycerol tetraethers (GDGTs) were analyzed following Yamamoto and Polyak (2009).An aliquot of F4 was dissolved in hexane-2-propanol (99 : 1) and filtered.Glycerol dialkyl glycerol tetraethers (GDGTs) were analyzed using high performance liquid chromatographymass spectrometry (HPLC-MS) with an Agilent 1100 HPLC system connected to a Bruker Daltonics micrOTOF-HS timeof-flight mass spectrometer.Separation was conducted using a Prevail Cyano column (2.1 × 150 mm, 3 µm; Alltech) and maintained at 30 • C following the method of Hopmans et al. (2000) and Schouten et al. (2007).Conditions were: flow rate 0.2 mL min −1 , isocratic with 99 % hexane and 1 % 2propanol for the first 5 min followed by a linear gradient to 1.8 % 2-propanol over 45 min.Detection was achieved using atmospheric pressure, positive ion chemical ionizationmass spectrometry (APCI-MS).The spectrometer was run in full scan mode (m/z 500-1500).Compounds were identified by comparing mass spectra and retention times with those of GDGT standards (formed from the main phospholipids of Thermoplasma acidophilum via acid hydrolysis) and those in the literature (Hopmans et al., 2000).Quantification was achieved by integrating the summed peak areas in the (M+H) + and the isotopic (M+H+1) + ion traces.TEX H 86 was calculated from the concentrations of GDGT-1, GDGT-2, GDGT-3 and a regioisomer of crenarchaeol using the following expression (Schouten et al., 2002;Kim et TEX H 86 is defined as the logarithmic function of TEX 86 and yields the best correlation with SST when the data from polar and subpolar oceans are removed (Kim et al., 2010).SST was calculated according to the following equation based on a global core-top calibration (Kim et al., 2010): where SST = sea surface temperature [ • C]; analytical accuracy was 0.45 • C in our laboratory.The temperature at 30-125 m was calculated according to the following equation based on a local core-top calibration in the SCS (Jia et al., 2012): where T 30−125 m is the temperature at 30-125 m [ • C]; analytical accuracy was 0.36 • C in our laboratory.

Spectral analysis
Spectral analysis was performed using the Blackman-Tukey and Cross-Blackman-Tukey methods provided in the Analyseries software package (Paillard et al., 1996).

Branched and isoprenoid tetraether (BIT) index
The BIT index, a proxy for soil versus marine organic matter input to sediments (Hopmans et al., 2004), varied between 0.02 and 0.54, with an average of 0.14 (Fig. 4).The BIT index was higher in marine isotope stage (MIS) 6, MIS 4, and MIS 2 than in MIS 5, MIS 3, and MIS 1.The variation is consistent with the variation of δ 18 O of benthic foraminifera (Fig. 4), suggesting that the mouth of the Mekong River came closer to the study site due to marine regression, resulting in an increase in soil organic matter input in the periods of low sea level stand.

TEX H
86 Weijers et al. (2006) noted that samples with high BIT (> 0.4) may cause anomalously high TEX H 86 -derived temperatures.This was not a concern for most of the samples used in the present study.The samples showing BIT values higher than 0.4 were omitted for paleotemperature calculations in the profile in Fig. 4.
Applying a global core-top calibration (Kim et al., 2010), the TEX H 86 -derived temperature varied between 22.7 • C and 29.4 • C, with an average of 26.1 • C (Fig. 4).The TEX H 86 -derived temperature was ∼ 26  (Kim et al., 2010).A correlation analysis indicated that TEX H 86 correlated better with the mean annual temperature at 30-125 m (r = 0.89) than with the temperature in the 0-30 m mixed layer (r = 0.69).This suggests that TEX H 86 reflects a deeper and cooler subsurface temperature, rather than the surface or mixed-layer temperature in the SCS.Applying the SCS local core-top calibration (Jia et al., 2012), the TEX H 86 -derived temperature at 30-125 m varied between 18.0 • C and 23.4 • C, with an average of 20.7 • C (Fig. 4).The TEX H 86 -derived temperature was ∼ 20.5 in MIS 4 to 2, and 22-23 • C in MIS 1 (Fig. 4).The core-top temperature was 22.3 • C, which agrees with the mean annual temperature at ∼ 90 m (Locarnini et al., 2010).
The TEX H 86 -derived temperature showed a suborbital-scale variation with an amplitude of > 1 • C (Fig. 4).Remarkable cooling occurred at 135 ka in MIS 6, 81 ka in MIS 5d, and 13 ka in the last deglaciation.

U K' 37
The U K 37 in this core was reported by Zhao et al. (2006) and was also analyzed at lower resolution in the present study (Fig. 4).Column chromatography was used to separate alkenones in this study, but not in Zhao et al. (2006).Despite the difference in preparation steps, both records showed an almost identical pattern (Fig. 4).One major difference was that the large cold excursion during 140-135 ka in the high-resolution record was not revealed in the new lower-resolution record.Applying a South China Sea coretop calibration (Pelejero and Grimalt, 1997), the U K' 37 -derived temperature obtained by Zhao et al. (2006) varied between 23.5 • C and 28.9 • C with an average of 26.2 • C (Fig. 4).The U K 37 -derived temperature was ∼ 26 to 2, and 27-28 • C in MIS 1 (Fig. 4).

Season and depth reflected by TEX H 86 and U K' 37
As indicated from satellite and shipboard observations, the chlorophyll concentrations did not display large seasonal variations, with only a moderate maximum in the winter season in the study area (Feldman et al., 1989;Liu et al., 2002).These observations suggested that the production of GDGTs does not change seasonally, and that TEX H 86 reflected the mean annual temperature at the study site.Although the pro- duction depth of GDGTs and the depth at which TEX H 86 indicates the temperature are still not clear, the investigation of surface sediments in the SCS suggested that TEX H 86 reflects a deeper and cooler subsurface temperature, rather than the surface or mixed layer temperature (Jia et al., 2012).In the present study, we assumed that TEX H 86 reflected a subsurface mean annual temperature at 30-125 m.
The core-top U K' 37 -derived temperature obtained by the equations of Pelejero and Grimalt (1997) based on core-top calibration in the SCS and Prahl et al. (1988) based on calibration by a culture experiment were 27.9 • C and 27.0 • C, respectively.The former agrees with the mean annual SST (28.0 • C; Locarnini et al., 2010), but the latter is 0.9 • C lower than the mean annual SST at the study site.The chlorophyll concentrations do not display large seasonal variations, with only a moderate maximum in the winter season in the study area (Feldman et al., 1989;Liu et al., 2002), suggesting that alkenones are produced throughout the year, but was more weighted to the winter season.The U K' 37 -derived temperature by Prahl's culture-based calibration thus reflects winterweighted SST, whereas that by the SCS core-top calibration reflects the mean annual SST.Because alkenone production is more weighted in winter, the variation of U K' 37 -derived temperature reflects more sensitively SST changes during winter.
Variation in the temperature at 30-125 m using TEX H 86 (T 30−125 m ) was nearly parallel to the variation in the SST obtained from the study core using U K' 37 (Zhao et al., 2006 and this study) (Fig. 4).The difference between the U K' 37derived SST and TEX H 86 -derived temperature at 30-125 m ( T alkenone-GDGT as defined in Jia et al., 2012) varied between 3.2 and 8.7 • C with an average of 5.6 • C (Fig. 5).The T alkenone-GDGT was generally around 6 • C from 150 to 16 ka, dropped to 5 • C between 16 and 3 ka and rose to 6 • C after 3 ka (Fig. 5).Jia et al. (2012) suggested that T alkenone-GDGT is an index of the depth of the thermocline; higher T alkenone-GDGT indicates a shallow thermocline.If this is true, the lower T alkenone-GDGT in 16-3 ka reflects the thermocline deepening in this period.Although there were some changes, T alkenone-GDGT was nearly constant, suggesting that the thermocline structure of the SCS was largely stable in most periods during the past 150 000 yr.
In the present study, assuming that the depth profile of the water temperature was constant, the TEX H 86 -derived SST in the MD97-2151 core (SST*) was obtained according to the following equation by adding the average value of T alkenone-GDGT (5.6 • C) during the last 150 000 yr to the T 30−125 m (Fig. 6b): The SST* is a SST calculated based on the assumption of a constant gradient of the water temperature for the use of SST calculation in the next section.The SST* variation is similar to U K' 37 -derived SST variation (Fig. 6b), suggesting that this assumption is reasonable.

Response of SCS SSTs to orbital forcing
In the central WPWP region (ODP Site 806 in the Ontong Java Plateau (Lea et al., 2000) and MD97-2140 at the northern margin of New Guinea Island (de Garidel-Thoron et al., 2005), Globigerinoides ruber Mg/Ca-derived temperatures showed a pattern that was broadly similar to that of the atmospheric CO 2 concentration recorded in Antarctic ice cores (Fig. 6b; Kawamura et al., 2007).
SST change at the study site was delayed behind that in the central WPWP region and changes in the atmospheric CO 2 concentration (Fig. 6b).Periodic cooling events occurred every 23 kyr and were superimposed on the variation seen in the other regions of the WPWP.To characterize the SST variations specific to the SCS, the SST at the study site was subtracted from the SSTs at ODP Site 806 (0 • 19 N; 159 • 22 E, 2520 m, Lea et al., 2000) and MD97-2140 (2 • 02 N; 141 • 46 E, 2547 m, de Garidel-Thoron et al., 2005).The SST records at ODP Site 806 and MD97-2140 were reassessed in the present study by tuning the oxygen isotope profile of G. ruber to that of Lisiecki and Raymo (2005) (Fig. 6a).The SST difference (SST WPWP -SST SCS ) was defined as SST.Four different SSTs (TEX H 86 -derived SST* ODP806-MD972151 , TEX H 86 -derived SST* MD972140-MD972151 , U K' 37 -derived SST ODP806-MD972151 , U K' 37 -derived SST MD972140-MD972151 ) showed a similar variation (Fig. 6c) and were formed as a stacked SST record (Fig. 6d).The stacked SST is designed to minimize the effect of spatial heterogeneity and the errors that arise from independent age-depth models and proxies.The stacked SST shows a regular pattern, with cooler events every 23 kyr (Fig. 6d) and the best correlation with the May insolation at the equator.
The glacial-interglacial SST variation in the SCS was often interpreted by changes in the sea level and the resultant basin geometry (e.g.Pelejero et al., 1999;Zhao et al., 2006).Since the SST drop was also seen in the other WPWP region in the glacial periods, the SST is used to distinguish  (Zhao et al., 2006) and of planktonic foraminifera (PF) in ODP 806 (Lea et al., 2000) and core MD97-2140(de Garidel-Thoron et al., 2005).(B) TEX H 86 -derived SSTs* and U K' 37 -derived SSTs in core MD97-2151, Mg/Ca-derived SSTs in core MD97-2140 (de Garidel-Thoron et al., 2005), and ODP 806 (Lea et al., 2000), and CO 2 concentrations from Dome Fuji (Kawamura et al., 2007).(C) Temperature differences ( SSTs) between the central Western Pacific Warm Pool region (MD97-2140 and ODP 806) and Site MD97-2151 in the southern South China Sea.(D) The stacked SST.The insolation on May 13 at the equator is shown for comparison.Temperatures were calculated from U K' 37 using an equation based on the core-top calibration (Pelejero and Grimalt, 1997) the local cooling of the SCS from the regional cooling of the entire WPWP area.The variation of SST was not synchronous with the variation of sea level change shown by the oxygen isotopes of benthic foraminifera, and the maximum local cooling of the SCS (maximum SST) appeared several thousand years after the minimum sea level (Fig. 6).The local cooling of the SCS was thus independent of sea level change and the resultant basin isolation.
Oxygen isotope and Mg/Ca studies for G. ruber in a core from the northern WPWP region indicate that the calcification depth of G. ruber is ∼ 60 m in the surface mixed layer (Sagawa et al., 2012).The temperature in the surface mixed layer shows little variation on seasonal and interannual timescales in the WPWP (Data from Locarnini et al., 2010).It is thus likely that the G. ruber Mg/Ca in ODP 806 and MD97-2140 cores reflected the mean annual SSTs.In core MD97-2151, TEX H 86 and U K' 37 reflected subsurface and surface mean annual temperatures, respectively, as discussed in the previous section.Thus, it is unlikely that a possible bias of individual paleotemperature proxy records affects the variation of the stacked SST.
The winter cooling of the SCS surface water is caused by the winter monsoon (Liu et al., 2004).Winter monsoon northerly winds cool the surface water in the northern margin of the SCS and advect the cooled water southward (Liu et al., 2004).The SST in the study region is mainly governed by the winter wind strength (Huang et al., 2011), though it is also affected by the advection of coastal cool water and upwelling induced by the summer monsoon winds to some extent (Xie et al., 2003).These modern observations suggest that SST is an index of the Asian winter monsoon activity with a higher SST corresponding to a stronger winter monsoon.According to this interpretation, the East Asian winter monsoon varied in response to precessional forcing and was maximized at the May perihelion.
This variation is consistent with previously published shorter records of the East Asian winter monsoon based on SCS SSTs (Shintani et al., 2008;Huang et al., 2011) that showed an intensification of the East Asian winter monsoon during the last deglaciation and its gradual weakening during the Holocene.The variation is also consistent with a diatom record from the Huguang Maar Lake in south China that showed an intensification of the East Asian winter monsoon during the last deglaciation and the early Holocene (Wang et al., 2012).Our result indicates that intensification of the Asian winter monsoon also occurred in the penultimate deglaciation (glacial termination II), and the variation has responded to a precessional cycle.
Cross-spectral analysis between SST and the δ 18 O of benthic foraminifera in a period of 9 to 149 ka indicates that the periodicity at 23-25 kyr is dominant and highly coherent at this frequency (Fig. 7).The SST maximum lags the precession maximum (the Northern Hemisphere summer insolation minimum) by 138  The bandwidth is 0.021.We performed spectral analyses using the Blackman-Tukey and Cross-Blackman-Tukey methods provided in the Analyseries software package (Paillard et al., 1996).
The strong precession signal seen in SST supports the hypothesis that monsoon is regulated by insolation changes at low-latitudes (Kutzbach, 1981).The obliquity and eccentricity signals in SST are much weaker than the precession signal (Fig. 7).This is in contrast with the previous marine records of the East Asian winter monsoon that has a similar variation to that of continental ice volume (e.g. de Garidel-Thoron et al., 2001).The glacial-interglacial pattern in proxy records presumably reflects the variation of the entire WPWP region rather than local variation sensitive to the East Asian monsoon activity.The approach of SST successfully extracted the signal of the East Asian winter monsoon from the climate variation in this region.

Phase relationships with the Indian and East Asian summer monsoons
On the 23 kyr cycles, the maximum phase of the winter monsoon (maximum SST) corresponds to the May perihelion (Fig. 8).This is nearly anti-phased with the variation in the Indian summer monsoon.The phase of the Indian summer monsoon stack was reconstructed using marine upwelling records from the Arabian Sea, and the maximum phase of the summer monsoon corresponded to the October to November perihelion (Fig. 8; Clemens and Prell, 2003;Caley et al., 2011).Caley et al. (2011) assigned upwelling and nonupwelling assemblages of planktonic foraminifera to the summer and winter monsoon assemblages, respectively, in a core from the Arabian Sea.Based on the record of these assemblages, they suggested that the Indian winter monsoon was anti-phase with the summer monsoon stack and maximal at the May perihelion.Because the relative abundance of the winter monsoon assemblage was affected mainly by the abundance of the summer monsoon assemblage (Caley the Indian summer monsoon stack maximum (Clemens and Prell, 2003), the Indian summer and winter monsoon maximum (Caley et al., 2011), the East Asian summer monsoon maximum (Wang et al., 2008), and the maximum of the summer precipitation in central Japan (Igarashi and Oba, 2006).et al., 2011), the phase of the winter monsoon assemblage maxima must be considered with caution.Despite this, the phase coincided with the maximum phase of the East Asian winter monsoon proposed in this study.This correspondence supports the perspective of Caley et al. (2011) that the Indian summer and winter monsoons were in anti-phase at the precession band.It also suggests that the Asian winter monsoon system behaved synchronously in the East Asian and Indian regions.
The phase relationship between the East Asian winter and summer monsoons is more complex.On the 23 kyr cycle, the variation in SST was out-of-phase by ∼ 80 • with changes in the East Asian summer monsoon intensity suggested by the stalagmite oxygen isotope record from the Hulu and Sanbao caves in central China (Fig. 8; Wang et al., 2001Wang et al., , 2008)).However it was anti-phased with changes in the summer monsoon precipitation in central Japan, as demonstrated by a pollen record from MD01-2421 (Fig. 8; Igarashi and Oba, 2006;Yamamoto, 2009).If the Chinese stalagmite records indicate the phase of the East Asian summer monsoon variability (Wang et al., 2001(Wang et al., , 2008)), the relationship between the summer and winter monsoons was neither in-phase nor antiphase, suggesting that different mechanisms should drive summer and winter monsoons.On the other hand, if the summer precipitation record in central Japan indicates the phase of the East Asian summer monsoon variability (Igarashi and Oba, 2006;Clemens et al., 2010), the phase relationship between the summer and winter monsoons was anti-phase, suggesting that common mechanisms should drive summer and winter monsoons.Since the phase of the East Asian summer monsoon is still debated (Clemens et al., 2010), the phase relationship between summer and winter monsoons remains inconclusive.
In the Asian monsoon system, the Indian summer monsoon and the East Asian winter monsoon are more powerful than the East Asian summer monsoon and the Indian winter monsoon, respectively (Wang et al., 2003).Consequently, regardless of the uncertain phase of the East Asian summer monsoon, the antiphase variation of the East Asian winter monsoon and the Indian summer monsoon is a dominant phenomenon in the entire Asian monsoon system.

Response of the East Asian winter monsoon to interhemispheric heat imbalance
On the 23 kyr cycles, the maximum phase of the East Asian winter monsoon (maximum SST) corresponds to the May perihelion (Fig. 8).However, this is not consistent with the Kutzbach model that the winter monsoon is maximum at the June perihelion.This inconsistency suggests that the East Asian winter monsoon did not simply respond to insolation changes on precession cycle.Two possible interpretations are proposed for the finding that the East Asian winter monsoon was maximum at the May perihelion rather than at the ice volume maxima (the March perihelion) or the insolation minimum in boreal winter (the June perihelion).First, the May perihelion is between the ice volume maximum and the insolation minimum in boreal winter.The combined effect of cooling by ice volume forcing and the decrease in sensible heating owing to the weakest winter insolation could intensify the Siberian High and result in a maximum in the East Asian winter monsoon at the May perihelion.Second, the Indian summer monsoon is minimum at the May perihelion, which was assumed to be caused by the combined effect of the minimum latent heat export from the southern Indian Ocean at the June perihelion and minimum sensible heating of the Asian plateau at the March perihelion (Fig. 8; Clemens and Prell, 2003).The weakest Indian summer monsoon at the May perihelion decreased the heat transfer from the Southern Hemisphere to the Northern Hemisphere and cooled Siberia in winter, causing a stronger Siberian High.It is, however, not clear how much the Indian summer monsoon regulated the interhemispheric heat balance.In both cases, seasonal heat imbalance between the Northern and Southern Hemispheres determined the intensity of the cross-equatorial flows of the East Asian monsoon.
The Asian monsoon system transfers energy and water vapor across the equator.At the November perihelion, the Indian summer monsoon was strongest and transferred more heat from the Southern Hemisphere to the Northern Hemisphere, resulting in warming in the Northern Hemisphere and cooling in the Southern Hemisphere.In contrast, at the May perihelion, the monsoon system transferred more energy to the Southern Hemisphere, resulting in cooling in the Northern Hemisphere and warming in the Southern Hemisphere.
The role of the cross-equatorial monsoon like the East Asian winter monsoon in regulating interhemispheric heat balance on an orbital timescale is still unclear.Further studies on the phase of the cross-equatorial monsoons will shed light on changes in interhemispheric heat balance on an orbital timescale.

Conclusions
The sea surface temperature difference ( SST) between the South China Sea and other Western Pacific Warm Pool regions reflected the intensity of the Asian winter monsoon.The SST shows strong variability at precession cycle during the past 150 000 yr.The strong precession signal supports the hypothesis that monsoon is regulated by insolation changes at low-latitudes (Kutzbach, 1981), but contradicts previous suggestions based on marine and loess records that eccentricity controls variability on glacial-interglacial timescales.
At the precession cycle, the maximum winter monsoon intensity shown by the SST corresponds to the May perihelion, which is not fully consistent with the Kutzbach model of maximum winter monsoon at the June perihelion.There are two possible interpretations; (1) the combined effect of maximum cooling by ice volume forcing at the March perihelion and the decrease in sensible heating owing to the weakest winter insolation at the June perihelion could intensify the Siberian High and result in a maximum in the East Asian winter monsoon at the May perihelion, and (2) the weakest Indian summer monsoon at the May perihelion decreased the heat transfer from the Southern Hemisphere to the Northern Hemisphere and cooled Siberia in winter, causing a stronger Siberian High.
The East Asian winter monsoon was anti-phased with the Indian summer monsoon and the summer monsoon precipitation in central Japan.This antiphase response suggests a linkage of dynamics between summer and winter monsoons in the Asian monsoon system.Since the cross-equatorial flows can transfer energy and water vapor across the equator, better understanding of the East Asian winter monsoon is a key issue to unravel changes in interhemispheric energy balance on an orbital timescale.

Fig. 1 .
Fig. 1.Map showing (upper) the location of core MD97-2151 and the winter sea surface temperatures in the South China Sea (Liu et al., 2004) and (lower) the locations of cores MD97-2151 and MD97-2140 and ODP Site 806 and the annual mean sea surface temperatures in the entire western Pacific warm pool (Locarnini et al., 2010).

Fig. 2 .
Fig. 2. Depth and seasonal variation of water temperatures at the study site (compiled data fromLocarnini et al., 2010).Months "J" to "D" means January to December.

Fig. 3 .
Fig.3.Age-depth model of core MD97-2151(Zhao et al., 2006 and  this study).The age at 699.5 cm (open circle) was not used for the age-depth model.Black and blue sold circles indicate 14 C based calendar ages and the tie points of oxygen isotope stratigraphy, respectively.

Fig. 8 .
Fig. 8. Precession wheel showing the phases of stacked SST maximum (with errors of the estimate shown by the dashed lines), the Indian summer monsoon stack maximum(Clemens and Prell, 2003), the Indian summer and winter monsoon maximum(Caley et al., 2011), the East Asian summer monsoon maximum(Wang et al., 2008), and the maximum of the summer precipitation in central Japan(Igarashi and Oba, 2006).

Table 1 .
Control points for the age model for core MD97-2151.
Locarnini et al., 2010)o 5a, ∼ 25 • C in MIS 4 to 2, and 28-29 • C in MIS 1 (Fig.4).The core-top temperature was 28.1 • C, which agrees with the mean annual SST (28.0 • C;Locarnini et al., 2010).Recently, Jia et al. (2012) correlated core-top TEX H 86 and water column temperature at 40 different sites in the SCS and found that the SSTs in the SCS were overestimated by a global core-top calibration • C in MIS 6, ∼ 29.5 • C in MIS 5e,