Millennial-Scale Dynamics of the East Asian Winter Monsoon During the Last 200,000 Years T Garidel-Thoron, L Beaufort, Bk Linsley, S Dannenmann

Millennial-scale dynamics of the East Asian winter monsoon during the last 200,000 years T Garidel-Thoron, L Beaufort, Bk Linsley, S Dannenmann

To cite this version:

T Garidel-Thoron, L Beaufort, Bk Linsley, S Dannenmann. Millennial-scale dynamics of the East Asian winter monsoon during the last 200,000 years. Paleoceanography, American Geophysical Union, 2001, 16 (5), pp.491-502. 10.1029/2000PA000557. hal-01460389

HAL Id: hal-01460389 https://hal.archives-ouvertes.fr/hal-01460389 Submitted on 8 Feb 2021

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. PALEOCEANOGRAPHY, VOL. 16, NO. 5, PAGES 491 - 502, OCTOBER 2001

Millennial-scale dynamics of the East Asian winter monsoon during the last 200,000 years

Thibault de Garidel-Thoron and Luc Beaufort CentreEuropben de Rechercheet d'Enseignementen Gbosciencesde l'Environnement(CEREGE), Aix-en-Provence, France

BraddockK. Linsley and StefanieDannenmann Departmentof Earthand Atmospheric Sciences, University at Albany,State University of New York,Albany, New York, USA

Abstract. The primary productivitydynamics of the last 200,000 years in the Sulu Sea was reconstructedusing the abundanceof the coccolithophoreFlorisphaera profunda in the IMAGES MD97-2141 core. We find that primary productivitywas enhancedduring glacial periods, which we suggestis due to a strongerEast Asian winter monsoon.During the last 80 kyr, eight significantincreases in primary productivity(PP) in the Sulu Sea are similar to East Asian winter monsoonchanges recorded in Chineseloess. The PP maxima are not linked with Heinrich events (HE) in the North Atlantic, althoughfour PP peaks are synchronouswith HE. The PP oscillationshave frequenciesnear those of the Dansgaard-Oeschgercycles in NorthernHemisphere ice recordsand indicatea teleconnectionof the East Asian winter monsoonwith Greenlandclimate. In this Sulu Searecord the EastAsian winter monsoonoscillates with periodicitiesof •6, 4.2-3.4, 2.3, and 1.5 kyr. In particular,the 1.5 kyr cycleexhibits a strongand pervasive signal from stage6 to the Holocene without any ice volumemodulation. This stationaritysuggests that the 1.5 kyr cycle is not driven by somehigh-latitude forcing.

1. Introduction tions which appear synchronouswith some Heinrich events [Linsley, 1996]. The couplingbetween the atmosphereand the oceanis funda- For the Dansgaard-Oeschger oscillations recorded in ice mental to climate dynamics over seasonalto millennial time- cores, similar changesare documentedin sea surfacetemper- scales.Variations in this couplinginfluence the oceanicbiosphere ature (SST) reconstructions,percentage of carbonate,magnetic and particularlythe phytoplanktonof the upper oceanlayer. In susceptibility,planktonic foraminifera oxygen isotopiccomposi- low-latitude areas the phytoplanktonicactivity quantified by tion, and foraminiferal assemblagerecords from the North primaryproductivity (PP) is correlatedto the wind stressdynam- Atlantic Ocean (for a review, see Corto'o et al. [2000]). The ics on the sea surface[Nair et al., 1989]. Previousstudies have loess-paleosolrecord in China exhibitscomparable shifts [Chen concentratedon long-termchanges in the PP [e.g., Beaufortet al., et al., 1997]. In the Santa Barbara basin a bioturbation index 1997; Mix, 1989], but little is known about the millennial-scale records nearly all stadials and interstadialsdescribed in the dynamicsof PP. Greenland record [Behl and Kennet, 1996]. At low latitudes, At present,we know that during the last glacial stage our organic carbon changesin the Arabian Sea appearto correlate climate systemwent throughrapid changesthat are well docu- with D-O cycles [Schulzet al., 1998]. mented in the North Atlantic area. Two major types of abrupt To constrainthe past millennialvariations of low-latitudepale- changeshave been described.Heinrich events are documented oproductivity,we investigate,at high-resolution,variations in the massive iceberg dischargesof ice-rafted debris to the North coccolithassemblages in IMAGES giant piston core MD97-2141 Atlantic deep-sea sedimentsoccurring with a periodicity of locatedin the Sulu Sea (Figure 1). •6-7 kyr [Bond et al., 1992; Heinrich, 1988]. These events The Sulu Sea is located between the Asian continent and the are followed by consecutiveabrupt warmings on the Greenland "Western Pacific Warm Pool" (WPWP), where annual SST is continent.Dansgaard-Oeschger (D-O) climatic oscillationswith above29øC [Yah et al., 1992]. The climate of the Sulu Sea is periods of 1000-3000 years have been describedin the Green- strongly influenced by the East Asian monsoon. The East land ice cores records [Dansgaard et al., 1993]. These oscilla- Asian monsoon results from the different potential heating tions correspondto •15øC air temperatureshifts between a betweenthe WPWP and the Asian continent.During the boreal stadial (cold phase) and an interstadial(warm phase) [douzel, winter the main heating source is located in the ocean. The 1999]. latent heat release associatedwith intense convectiveprecipita- The Heinrich eventsare correlativewith major changesin the tion fuels the meridionalcirculation. Tropical convectionin the climate dynamicsrecorded worldwide, from the North Atlantic to western equatorial Pacific is connected to the descending the Antarctic.In the Asian region the East Asian winter monsoon branch over the Siberianregion, forming a strong local Hadley strengthenedduring Heinrich events[Porter and Zhisheng,1995; cell in the East Asian region [Zhang et al., 1997]. The East Xiao et al., 1999]. Planktonicforaminifera assemblages in South Asian winter monsoon winds in the Sulu Sea result from the China Sea recorded these abrupt changes [Chen and Huang, merging of the northerly East Asian monsoonwith the Pacific 1998; Wanget al., 1999]. In the Sulu Sea the oxygen isotopes trade winds over the South China Sea [McGregor and (6280)of planktonicforaminifera show important rapid oscilla- Nieuwolt, 1998]. East Asian winter monsoon bursts during Januaryto March (Figure 2) can induce blooms of coccolitho- Copyright2001 by the AmericanGeophysical Union. phorids [Wiesneret al., 1996]. The PP rises correlativelyto Paper number2000PA000557. the wind stressstrengthening because of the strongermixing 0883-8305/01/2000PA000557512.00 of the upper ocean(Figure 2) [Nair et al., 1989]. Thus coccolith

491 492 DE GARIDEL-THORON ET AL.: DYNAMICS OF THE EAST ASIAN WINTER MONSOON

ASIA

2O N ••.. • PhilippinesSea }' .• '•--,MD972141 •,•-' .• ••% •/•x• OdP769 10N[• , 0'• Sulu.Sea 4; ? c2• /• c• , PcificOcean ,• • • CelebesSea

O-

10S I I I 90 E 100 E 110 E 120 E 130 E

Figure 1. Map of SoutheastAsia andof the marginalseas of thewestern Pacific. Note locationof coreMD97-2141 in the Sulu Sea.

assemblagesrecord information on both paleoproductivity The 95% confidenceinterval for %Fp varies between+2 and changesand also on East Asian winter monsoon variations in +6% dependingon the percentageof Fp [Pattersonand Fishbein, the Sulu Sea. 1989].

3.1. Age Model 2. Material The age model of the core MD97-2141 was developedby The 36 rn giantpiston core IMAGES MD97-2141 (08ø47'N, Dannenmannet at. [1998] and Oppoet at. [1998]. It was obtained 121ø17'E,3633 m depth)was retrieved during the IPHIS-IMAGES using28 acceleratormass spectrometer (AMS) 14Cages on III cruiseof the R/V Marion Dufresnein May 1997.This position Globigerinoidesruber and G. sacculiferand by comparisonof is locatedin thevicinity of the OceanDrilling Program (ODP) Site theplanktonic foraminifera 15180 curve (on Gtobigerinoides tuber 769 [Linstey,1996]. The core is locatedon the Cagayanridge, and G. saccutifertests) with the SPECMAPstack [Imbrie et at., which protectsthe site from downslopeprocesses, and abovethe 1984] (Table 1). The radiocarbondates were convertedto calendar presentlysocline depth (•3800 m), allowingfor goodpreservation agesusing (1) a correctionof 400 years,according to the age of carbonates[Linstey et at., 1985; Miao et at., 1994]. The reservoirof carbonin the ocean[Bard, 1988], and then (2) the sedimentsare composedpredominantly of well-preservednanno- CALIB3 calibrationsoftware [Stuiver and Reimer, 1993] to take fossil-foraminifera oozes. into accountpast atmospheric changes in cosmogenicproduction. All the agesdiscussed here are calendarages B.P. The carbon reservoirage in this work is assumedto be invariantwith time. We 3. Methods alsoassumed that the carbonreservoir age changes in the tropical For coccolithcounting, the corewas sampledevery 2 cm in the upperoceanic layer where G. tuber and G. saccutiferlive are not upper 6 m of the core, allowingfor a resolutionof •70 years, varyingmore than •200 years[Duptessy et at., 1991]. The last and every 3-4 cm in the lower 30 m for a resolutionof •200- appearancedatum of G. tuber pink at 1593 cm fits with the 500 years.A smearslide was preparedfor each sample,and at TerminationII [Thompsonet at., 1979],strengthening the validity least 300 coccolithswere countedfor each slide (mean of 357 of the age model(Figure 3). Two ninth-orderpolynomial regres- coccoliths)on a Zeiss Axioscopat a 1000x resolution.Percen- sionswere usedfrom the coretop to 400 cm and from 440 to 920 tages of Ftorisphaeraprofunda (Fp) were computedusing the cm on the 14Cages and SPECMAP tie pointsto smooththe following equation: sedimentationrate for the last60 kyr. This smoothingis indispen- sablefor the spectralanalyses to avoidspurious peaks linked with %Fp= 100(numberFp)/(totalcoccoliths) sedimentationrates changes. DE GARIDEL-THORON ET AL.' DYNAMICS OF THE EAST ASIAN WINTER MONSOON 493

summer winter 3.2.2. Singular-spectrum analysis. Singular-spectrum analysis(SSA) is designedto extract the informationcontained monsoon monsoon in a short, nonstationary,and noisy signal [Vautard and Ghil, 18 7 1989]. This method is based on the computing of empirical orthogonalfunctions in the time domain. SSA can give insight into the dynamics of the underlying system that generatesthe signal.Using data-adaptivefilters which are not perioddependant, SSA allows the separationof the noise from the trend and the deterministicoscillations of the signal. • 14 6 3.3. FIorisphaera profunda: A Paleoproductivity Marker The coccolithophoddae(Prymnesiophycaea) are phytoplank- tonic organismsthat live in the oceanicphotic zone. They are very sensitive to variations in light and nutrient availability, .0 12 5.5' which are both depthdependent. The lower photiczone is darker and richerin nutrientsthan the upperphotic zone. In the tropical ocean the lower photic zone coccolithophoridaecommunity is dominatedby Florisphaeraprofunda associatedwith Gladioli- ••_10 5 thusfiabellatus and Algirosphaerarobusta, while most of the other specieslive in the upper photic zone [Okada and Honjo, 1973]. :LY'__ .p.p. (gC.m_2.m.1, •/'•4.5 When the nutricline is shallow,the upper photic zone species dominatethe coccolithcommunity, whereas when the nutriclineis deeper,the relativeproportion of lowerphotic community is more 6 t, •,•,•,•,•,•,1,•,•,•,•,-----wind speed (m.s-1) i,• 4 important.The depth of the nutdcline in the low-latitudeopen oceanis mainly drivenby wind intensity.When winds are strong, may jul sep nov jan mar may the upper layers are well mixed, and the nutrientsare upwelled into the upperphotic zone. Inversely, when wind stressdecreases, month the mixing is less effective, and the photic zone is depletedin nutrients.This depthrelationship between coccolithophorid com- munitieswas successfullyused by Molfino and Mcintyre [1990] Figure 2. Wind strengthin Sulu Sea at 10øN (Comprehensive to monitorchanges in the nutriclinedepth and has been calibrated Ocean-AtmosphereData Set (COADS) Atlas) and estimated to primary productivityby Beaufort et al. [1997] for paleopro- monthly primary productionat the same location from Antoine ductivity reconstructions. and Morel [ 1996]. The relationshipbetween the Florisphaeraprofunda ratio and primaryproduction has alreadybeen quantified by Beaufortet al. [1997] with the following equation:

The averagesedimentation rate is •10.5 cm kyr-•, with y: 617 -[2791og(x+ 3)], maximaduring glacial stage 2 of 34 cm kyr-•. For example, thesedimentation rate during the stage 3 is •30 cmkyr -•, which wherey is theyearly PP (g C m-2 yr-•) andx is thepercent Fp. allows a 70 year resolution(2 cm sampling).A hiatus occurs This equationis basedon Indian Oceanlow-latitude core tops.We between 30 and 22 ka (Figure 3). The sedimentationrates are assumethat the coccolithophodds'assemblage distributions are coherentwith other records;that is, they are increasedduring homogenousin the intertropicalzone. The variationsof Fp are in glacial periods and lower during interglacialstages [e.g., Chen agreementwith other paleoproductivityproxies [Beaufortet al., and Huang, 1998]. Reduced benthic mixing due to dysaerobic 1997]. conditionsin the Sulu Sea lessensthe bioturbationsmoothing effect [Kuehl et al., 1993]. The sharptransition in the coccolith record at 14.55 ka, occurswithin 2 cm, anotherargument for a 4. Results weak bioturbation effect. Thepresent annual PP in theSulu Sea is 148g C m-2 yr -• 3.2. Signal analysis [Antoineand Morel, 1996], which is close to the averagerecon- 2 1 3.2.1. Spectral analysis. To extract the significant perio- structedPP over the last 200 kyr of 135 g C m- yr- . The PP dicities contained in the PP signal, we performed spectral (data are accessible at http://www. cerege.fr and at ftp:// analysisusing different algorithms(Blackman-Tukey, maximum ftp.noaa.ngdc.gov/paleo)oscillates during the last200 kyr between 2 1 entropy, and multitaper methods) provided in the package 81 and 223 g C m- yr- (Figure 3). Analysedes [Paillard et al., 1996]. The comparisonof these different methodsallows the discriminationof spuriousresults 4.1. Glacial-Interglacial Variations due to biases of a particular method. We present only the On a glacial-interglacialtimescale, PP increasesduring glacial multitaper method (MTM) results here. The MTM is able to periodsand decreasesduring interglacials(Figure 3). PP is mod- detect low-amplitudeoscillations in relatively short time series eratelycorrelated (r2 = 0.49)with the ice volume curve (SPEC- with a high degree of statisticalsignificance [Thomson, 1982]. MAP stackof Imbrie et al. [1984]). This is shownby the spectral The statisticalsignificance reported in this work is computed analysisof the paleoproductivityrecord, which containsorbital using a Fisher test (F test). This F test is performed on the frequencypeaks of Milankovitchtheory (i.e., 1/100 kyr, 1/41 kyr, amplitudeto analyzethe harmonicoscillations assuming that the and •1/20 kyr) (Figure4) andconfirmed by cross-spectralanalysis signal containsperiodic and separatedcomponents [Iqou et aL, betweenthe PP and the SPECMAP stack(not shown).This orbital 1997]. forcingexplains about half of the varianceof the PP record.Half of 494 DE GARIDEL-THORON ET AL.: DYNAMICS OF THE EAST ASIAN WINTER MONSOON

Table 1. RadiocarbonAges and Tie PointsFrom the SPECMAPStack Used in MD97-2141 Chronologya Depthin Core,cm AMS •4CAge Errorb CalendarAge, years Species/AgeModel Accession Number 1 4,560 +50 4,798 G. tuber (white) OS- 16971 10.5 4,210 +40 4,286 G. tuber (white) OS- 16926 14 4,740 +40 4,962 G. tuber (white) OS- 16410 29 4,700 +55 4,873 G. tuber (white) OS-16972 59 6,020 +40 6,416 G. sacculifer(w/out sac) OS-16411 73 6,810 +55 7,274 G. sacculifer(w/out sac) OS-16973 85 6,830 +90 7,295 G. sacculifer(w/out sac) OS- 18401 94 8,850 +55 9,455 G. sacculifer(w/out sac) OS-16974 99 10,700 +90 12,152 G. sacculifer(w/out sac) OS-16975 120 93,80 + 160 10,001 G. sacculifer(w/out sac) OS-16977 150 10,200 +80 11,045 G. tuber (white) OS-16978 158 10,250 +120 11,153 G. sacculifer(w/out sac) OS-16980 162 10,750 +50 12,228 G. sacculifer(w/out sac) OS-16979 205.5 11,750 +130 13,258 G. sacculifer(w/out sac) OS-17238 212 12,350 +65 13,931 G. tuber (white) OS-16412 226.5 13,000 +95 14,796 G. tuber (white) OS-16970 244 14,100 +70 16,422 G. tuber (white) OS-16413 269 14,750 +70 17,200 G. tuber (white) OS-16361 282 15,100 +90 17,591 G. sacculifer(w/out sac) OS-17913 339 17,150 +140 19,749 G. tuber (white) OS-17914 368 17,650 +85 20,430c G. tuber (white) OS-22672 400 18,850 +140 21,847c G. sacculifer(w/out sac) OS-17916 440 28,000 +130 32,101c G. sacculifer(w/out sac) OS-17882 487 30,900 +260 35,146c G. tuber (white) OS-17912 506.5 33,000 +310 37,290c G. sacculifer(w/out sac) OS-17917 543 33,600 +590 37,893c G. sacculifer(w/out sac) OS-17911 553 34,300 +390 38,790c G. sacculifer(w/out sac) OS-17915 594 36,900 +460 41,134c G. sacculifer(w/out sac) OS-17918 920 59,000 SPECMAP 1150 71,000 SPECMAP 1190 80,000 SPECMAP 1400 99,000 SPECMAP 1450 109,000 SPECMAP 1540 115,000 SPECMAP 1650 131,000 SPECMAP 1790 146,000 SPECMAP 1860 151,000 SPECMAP 2000 171,000 SPECMAP 2200 182,000 SPECMAP aSeeDannenmann et al. [1998] and Oppoet al. [1998]. bError is given in 1 cCalendar ages have been calculated using a 400 yearreservoir correction and applying the Stuiver and Braziunas [ 1993]calibration curve for samples youngerthan 20,000 calendaryear in age and a U/Th calibrationcurve for the samplesolder than 20,000 calendaryears [Bard et al., 1993].

the varianceof the PP time seriesremains to be exploredin the Project (GRIP) and Greenland Ice Sheet Project 2 (GISP2) sub-Milankovitch timescale. isotopic records also display an abrupt warming at •14.5 ka. This abruptwarming seemsto be at least hemispheric[Bard et al., 1997]. The sea level rise (meltwater pulse (MWP) Ia of 4.2. Sub-Milankovitch Dynamics Fairbanks [1989]) was invoked by Pelejero et al. [1999c] to 4.2.1. Bolling/Allerod and the Younger Dryas event. The account for the thermal and ten'igenousinput changesin the last deglaciationis marked by the Younger Dryas event, which South China Sea. However, the abruptnessof the PP change interruptsthe global warming trend in the NorthernHemisphere. (<200 years) is not easily attributableto the flooding of the The YoungerDryas seemsto be at leasthemispheric in extentand Sundaland.An abrupt drop in the East Asian winter monsoon has been alreadydescribed in the Sulu Sea [Kudrasset al., 1991; intensity can explain this PP change in a more plausible Linsley and Thunell, 1990]. The Younger Dryas event is also manner. presentin theMD97-2141 15•80 record between 13.5 and 11.5 ka The interpretationof the YoungerDryas event in the Sulu Sea (Figure 3). The paleoproductivityrecord inferred from Fp is is still controversial.Two main hypotheseswere summarizedby characterizedby an abrupt decreaseat 14.55 ka (from 170 to Andersonand Thunell [1993]: either it reflectsa coolingevent or •125 g C m-2 yr-1) toa plateauuntil 11.5 ka, followed by another it is dueto a changein the 15•80of seawater.Modem analog decreaseto 100g C m-2 yr-•. technique reconstructionof SST with planktonic foraminifera We interpretthis PP recordto indicatea sharpdecrease of the doesnot recordenough SST differenceduring the YoungerDryas East Asian winter monsoonstrength after 14.55 kyr. A similar to account for the observedisotopic shift [Thunell and Miao, abrupt transition at 14.5 ka is recorded in the alkenone-sea 1996]. So the enrichmentis probablydue to changesin seawater surfacetemperature (SST) record of the meridionalSouth China isotopic compositionduring the Younger Dryas. The origin of Sea as a 1.5øCstep [Pelejero et al., 1999a].Greenland Ice Core this changeis still being debatedas an oceanicsource [Duplessy DE GARIDEL-THORON ET AL.: DYNAMICS OF THE EAST ASIAN WINTER MONSOON 495

Age (cal. yrs BP) 0 50000 100000 150000 200000 ' • 2000 a • • 1000 "• 0 , -3 '-- b ...... ' .... ' ' ' I L.A.DGlobigerinoides - _ •,• v ruberpink ' -2 :• -2,,5

...... ' . 0

-1

•.•-1,5 _

E 200

• 160 o

'• 120 o

'- 80 I I ...... ]•;,,l l]•j I I I I 0 50000 100000 150000 200000 Age (cal. yrs BP) Figure3. (a)Position of acceleratormass spectrometry (AMS) dates on planktonic foraminifera converted in calendarages (small triangles) and the SPECMAP tie points used in the stratigraphy (solid circles). Note the hiatus between400 and 440 cm (shaded area). (b) MD97-21410:80 planktonicforaminifera results at 10cm intervals (shadedline) compared with the SPECMAP stacked deep-ocean record (thick solid line). (c) MD97-2141 primary productivity(PP) as reconstructed from the coccolithophorid record.

et aL, 1991]or an atmosphericvariation [Anderson and Thunell, PP variations in the Sulu Sea we smoothedthe PP record with a 1993]. 2 kyr averagemoving window and resampled ata 100year time The PP plateauduring the Allerodand the YoungerDryas is step(Figure 5). This2 kyrwindow keeps only the variance at the similarto the warmingrecorded by alkenonesrecords from the millennialand longer timescales and attenuates the century-scale cores 17961 and 17964 in the South China Sea (SCS) [Pelejero variations.During the last 80 kyr the•PPrecord reveals eight et al., 1999a,1999b]. The YoungerDryas-Allerod SST differ- eventsnumbered PP1-PP8, some of wIiichmay be synchronous ence in the SCS is 0.4øC. This correspondsto an increaseof with North Atlantic environmentalchanges (Figure 5). The first •0.1%o of the foraminiferal15:80, 20% of the 0.5%0change one(PP1) at •8.3 kyr agois in phase,with the earlyto middle observedin Sulu Sea sediments[Linsley, 1996]. Our PP record Holocenetransition period [Alley et al., 1997].PP2 matches with strengthensthe argument that East Asian winter monsoon dynam- HeinrichEvent 1 (HE1). PP3 occursduring the Last Glacial ics duringYounger Dryas were not significantlydifferent from Maximum. PP4 seemsto correlatewith HE4. PP5 and PP6 do not the Allerod becauseof the small measuredchange in the PP matchwith any major HeinrichEvents, and HE5 is being record betweenthe Allerod and the YoungerDryas. Thus the intercalated between these two PP increases. However, PP5, at changein planktonicforaminifera 15:80 composition in the Sulu •44 ka is correlated to an increase of the winter monsoon Sea betweenthe Allerod and the YoungerDryas seemsto be strengthrecorded in paleoloessfrom China, dated between 43.3 mainlydue to an isotopicvariation in oceanwater linked to and 45.2 kyr BP (PL7 of Chenet al. [1997]).PP7 couldbe salinity,after a majorchange in atmosphericprocesses 14.5 kyr relatedto HE6. PP8 doesnot matchany HeinrichEvent. Chen et ago, duringthe Bolling. al. [1997]also describe two eventsat 59.2-66.2ka (PL9)and 4.2.2. Millennial-scale PP events during MIS 3. The 68.6-71.2 ka (PL10),which may correlatewith PP7 and PP8 Heinrich events in the North Atlantic have been correlated to givenage model uncertainties. In conclusion, all PP maximain short increasesof the East Asian winter monsoon dynamics the Sulu Sea can be correlatedwith Chinese 1oessevents and [Chenet al., 1997;Porter and Zhisheng,1995]. In orderto notablyevents PP5 and PP8 may correspondonly with the comparethe millennialvariations in theNorth Atlantic and the Chineseloess record. Therefore we concludethat these regional 496 DE GARIDEL-THORON ET AL.: DYNAMICS OF THE EAST ASIAN WINTER MONSOON

I i i i i 1 106 i 2,510 7 i i i

111 ka i I i i i i i i i i i I 39 ka 8 10 5 x --.

6 105 m

4 105

21 ka

i i 2 105

--.

o I o 105 104

period (ka)

Figure 4. Spectralanalysis ofthe MD97-2141 PP recordbetween 4.1 and200 ka. Threepeaks appear in the spectral analysiscomputed using the Blackman-Tukeyalgorithm (solid line) and the maximum entropyalgorithm (dotted line), which correspondto orbital frequenciesof the Milankovitchtheory. events are indicative of significantchanges in the East Asian investigatedbecause of limited synchronizationbetween the ice winter monsoondynamics. PP1, PP2, PP4, and PP7 match clearly core records and the marine records inherent to differences in with North Atlantic Heinrich Events. In stage 3, uncertaintiesin establishedchronologies. the age modelthat cannotbe smallerthan +2 kyr may explainthe A strengtheningof the East Asian winter monsoonduring the relative discrepancybetween PP5 or PP6 and HE5 but cannot glacialstages is compatiblewith the conclusionsof severalstudies accountfor the occurrenceof supplementarypeaks (PP5 or PP6 alreadycarried out in this area [Chen and Huang, 1998; Wanget and PP8). Our data indicatethat the dynamicsof the East Asian al., 1999]. It correspondsto a strengtheningof the Hadley cell winter monsoon is not directly linked with the major iceberg between the Western Pacific Warm Pool Low and the Siberian dischargesin the North Atlantic as statedby Porter and Zhisheng High. This mode of atmosphericcirculation is compatiblewith last [1995]. The East Asian winter monsoonexhibits higher-frequency glacial stage simulations[Kutzbach et al., 1993]. The Siberian dynamicsthan that of the main HE. However, it remains to be Highs are directly connectedto the Ferrel cells, which influence determinedif a commondynamic is presentbetween the high- the Greenland and Asian climates. This teleconnection should and the low-latituderecords that is not linked with major icebergs operatevia the coupling of these two cells (Ferrel and Hadley) discharges.We next compare the Dansgaard-Oeschgercycles during the winter. recorded in the Greenland ice core records with our PP record In conclusion,the lack of a systematiccorrelation between the from the Sulu Sea. Sulu SeaPP and HE indicatesthat icebergdischarges in the North Toevaluate Dansgaard-Oeschger scalevariability, weper- Atlantic and the East Asian winter monsoon follow different formedSSA on the Sulu Sea PP and on the GRIP •580 records dynamics.However, the correlationbetween the Greenlandclimate to examinethe relationshipsbetween Greenland climate and East PC2 and the East Asian winter monsoon PC2 indicates that similar Asian winter monsoondynamics. For this analysis,we resampled millennial-scaleclimate variability affectsboth the Greenlandand the two recordsat 200 year time intervalsafter interpolation.We the Western Pacific climates. computed SSA with the Vautard-Ghil autocovarianceestimator embeddedin 20 dimensionscorresponding to 2000 years. The 4.3. Analysis in the Frequency Domain (Suborbital first principal component(PC) in the two recordsdescribes the Frequencies) long-term trend, and the secondPC showsmillennial dynamics Threetime sliceswere definedto explorehigh-frequency cycles (Figure 6 and Table 2). There is a good agreementbetween the in the PP record.The first one, from 22 to 4.1 ka, spansthe whole two PC2 recordsof millennial dynamicsin PP and temperaturein deglaciation.The secondone, from 60 to 30 ka, includesmost of Greenlandduring the last 70 kyr. PP increasesin the Sulu Sea stage3. The third one, from 160 to 130 ka, correspondsto the end when temperaturesin Greenlanddecrease. Cross-spectral analysis of stage6. The PP recordswere derrendedin eachsegment using a indicateshigh coherencybetween the PC2 of PP and the PC2 of ninth-orderpolynomial for the deglaciation,a fourth-orderpoly- GRIP for the •6 and 3.5 kyr frequencybands with phasesof nomialfor stage6, and a linearderrend for stage3 (Figure7c). The •160 ø and of 140ø, which is indicativeof the oppositephase three recordsshow variancepeaks at periodsof •6, 3.5, and 2.4 describedabove. This phaserelationship is well constrainedfor kyr and between1780 and 1200 yearswith a mean at 1500 years the deglaciationby •4C ages.The phasecannot be further (Figure 7e-7g). The spectralpeaks could be related to the age DE GARIDEL-THORONET AL.: DYNAMICS OF THE EAST ASIAN WINTER MONSOON 497 180t PP1 PP2PP3 PP4 PP5PP6 PP7 PP8 t180 •'E160 / !•:'• 160 •/.. ' õ -40 I I -40 I ! a_. 120 120 ., .., ',.ViSit lOOb & .4 'V' lOO 83i00tD'2 80 80 0 10 20 30 40 50 60 70 80 Age (kyrs BP) Figure5. MD97-2141primary productivity versus calendar age (solid line) smoothed with a 2 kyrmoving average window.Between 22and 30 calendar kyrB.P., ahiatus exists inthe sediment record identified with •4C dates (shaded area).The major peaks in PP are numbered from 1 to8. The Younger Dryas and timing of Heinrich Events inthe NorthAtlantic (shaded area) are shown intoward bottom. The ages for these events are from Bard [1998] for the YoungerDryas, Thouveny etal. [2000] for North Atlantic Heinrich Events 1-4, and Chapman andShackleton [1998] for the chronologyof H5 and H6. model.However, changing the age model from polynomial fits to 5. Discussion:High-Frequency Cycles linearfits does not change the periods of thepeaks > 100years, and the relativetiming of thepeaks is kept.Thus the spectralpeaks Becausethe frequency of •6 kyr-• is closeto theHeinrich seem to be robust featuresof the signal with this current age Events'frequency band, we filtered the PP record in thisfrequency model. bandusing a Gaussian filter (bandwidth of0.1666 kyr -• + 0.03) 4.3.1. The •1.5 kyr cycle. Forthe time slice from 4.1 to 22 (Figure8a). The resulting PP series clearly shows an amplification ka themost significant peak occurs at 1.38kyr (significanceof duringthe glacial periods, with minima during interglacials (i.e., 93%).In the 30-60 ka timeslice, two peaksoccur at 1.54kyr Holoceneand stage 5:75 - 130ka), and indicates that this frequency (98.5%)and 1.2 kyr (99.8%).In thelast time slice the peak is at seemsto be ice volumeforced. The amplificationcould depend on 1.48ka (99%).The strongest periods are all around1.4-1.5 kyr, thehigh-latitude ice volume available for ice rafting events. andthe occurrencein threedifferent time-slices indicates probably A 3.7 kyr cyclehas already been reported by Pestiauxet al. a commonorigin and an almoststationary signal across different [1988]in theIndian Ocean. A similarperiodicity of 3.3 kyr was climatic conditions. alsoreported by Sirockoet al. [ 1996]in a productionrecord from 4.3.2. The m2.4kyr cycle. In the4.1-22 ka timeslice, two the Arabian Sea. These authorsattributed this frequencyband to minorpeaks occur (2.58 and 2.17 kyr), but they are not significant. combination tones of orbitalforcing on monsoondynamics. In the In the 30-60 ka timeslice the prominent peak occurs at 2.32 kyr SuluSea this period is significantduring the MIS3. This period is (significanceof 79.8%). In the130-160 ka time slice a significantnot very stable in thePP record, oscillating between 3.62 kyr for peakis at 2.43 kyr (96%).The period of 2.4 kyr seemsalso stage3 and3.36 kyr for stage 6. Theidentification ofthis period in stationary,but its strength varies through time. It is moremarked severalpaleomonsoon records seems to indicatethat this period is duringthe glacialstages and may correspondto ice volume significantin Asian monsoon system dynamics. The hypothesis of forcing. combinationtones related to precessionaland to obliquityfrequen- 4.3.3. The •4.2-3.3 kyr cycle. In the time slicefrom 22 to cies,which forces the climatesystem like a nonlinearoscillator 4.1 ka thepeak is at •4.2 (significanceof 97%). The 30-60 ka [Pestiauxet al., 1988],could explain the variations of thespectral timeslice peaks at 3.6 kyr (97%), and in the130-160 ka time slice peakscontained in this frequencyband. Even if a nonlinear thesignificant peak is at3.36 kyr (86%). The strength of thispeak climaticoscillator predicts only periods >5 kyr [LeTreut and Ghil, is strongin all thetime slices, but the period does not seem to be 1983],the monsoonsystem by its amplificationof insolation stationary. forcingcould probably generate cycles in thisfrequency band 4.3.4. The •6 kyr cycle. In thetime slice between 4.1 and22 [Pestiauxet al., 1988]. ka thepeak at •8 kyr is notsignificant because of thedominant deglaciationsignal overprints onthe low-frequency band. In the30- 5.1. The 2.4 kyr Cycle 60ka time slice a peak occurs at 5.7 kyr (significance of 89.9%), and The PP exhibitsa strong2.4 kyr cycle,especially during the between130 and160 ka the significantpeak is at 6.7 kyr (92.1%) lasttwo glacial stages (MIS 2-3 and6). Thisperiod has already with anotherconcentration of varianceat •5.06 kyr (90.5%). This been described in oceanic sediments. Pestiaux et al. [1988] frequencyismost pronounced during the glacials periods. observea 2.3 kyr periodin theIndian Ocean.hydrography, which 498 DE GARIDEL-THORON ET AL.: DYNAMICS OF THE EAST ASIAN WINTER MONSOON

Age (kyrs BP)

10 20 30 40 50 60 70

80 _ _ ß .... ! .... ß .... ß .... ß .... ß .... ß - -I

E 100 PP (D

• 12o ._ > ._

-o 140 o

-35 •" 160 E n 180 ,, ., ., }, , _/,, ) , ,

-40

8180GRIP - 80 -45 PC2 PP - 60 PC2 GRIP

o - 40

ß• - 20

o 0

m 20 .E_ n 40

60 ß 10 2o 3o 4o

Age (kyrs BP) Figure6. (top)MD97-2141 PP (smoothed; solid line) record versus the 0•80 recordof theGRIP ice core (dotted line).(bottom) PC2 of thePP record (solid line) and PC2 of the15180 record of GRIP(dotted line) as reconstructed by singularspectrum analysis (SSA), which show similar oscillations.Note that both scalesin the bottompanel are invertedcompared to the top panel. was interpretedas a combinationtone of the precessionaland 1993], which depends on the solar flux (Hallstattzeit cycle obliquity cycles, representingan internal but nonlinearresponse [Damon and dirikowic, 1992]) and on the oceanic-atmosphere of the monsoonsystem to solarforcing. This cycle is alsopresent exchangeof •4C.A 2.2 kyr periodicitywas also reported in a in the atmospheric•4C excessrecord [Stuiver and Braziunas, monsoon record from the Oman margin [Naidu and Malgrem,

Table 2. PercentVariance Described by the PrincipalComponents of the Singular-SpectrumAnalysis and Mean Periodsof the PCs PP in Sulu Sea •80 GRIP Described Cumulative Mean Period, Described Cumulative Mean Period, Variance,% Variance,% years Variance,% Variance,% years PC 1 82.52 82.52 >8000 67.64 67.64 >8000 PC2 6.76 89.29 6200 13.56 81.21 5500 PC3 2.98 92.28 3690 8.4 89.61 3500 PC4 1.67 93.95 2400 3.45 93.06 2760 PC5 1.15 95.10 1870 1.76 94.83 1870 PCs 6-20 4.9 100 ß ß ß 6.07 100 ß ß ß DE GARIDEL-THORON ET AL.: DYNAMICS OF THE EAST ASIAN WINTER MONSOON 499

Age (kyrs BP)

0 50 100 150 200

25O i

200

150

ß E. 100 o

o 22O ø o

180

'•- 140

140

100 121 D 100 4 12 20 40 50 60 140 150 160 age (kyrs) age (kyrs) age (kyrs)

i i i i i i i i i i i i i i i i i ! i i i i I 8• oø42ooo ø oø oø I oo •o •o noll,o•I I'I P•o,o o o , o.to I] 1

' ', I,', I I 1 0.9

' , , 2 ' • [ , ',, ,•, , , ",,•,,, ,•1 ','• 0•""' ' ' '' ' " ',/, • ,-, • ,•, ,,,,,,,•,' t' 0.7 10000 2000 10000 2000 10( •00 2000 period (a) period (a) period (a)

Figure 7. (a) MD97-2141 primaryproductivity (0-200 ka). Three intervalsare enlargedin Figures7a-7c. (b) From 4 to 22 calendarkyr B.P. A polynomialof third order(thin solidline) was appliedto detrendthe primaryproductivity signal(thick solid line) beforethe multitapermethod (MTM) spectralanalysis. (c) From 35 to 60 ka. PP (thick solid line) is represented,a linear detrendwas performedbefore spectralanalysis. (d) From 130 to 160 ka. A fifth-order polynomial(thin solid line) of the PP (thick solid line) was also subtractedfrom the PP recordbefore the spectral analysis.(e-g) MTM spectraof the three time windows in Figures 7b, 7c, and 7d, respectively.Line amplitudes (dashedline) and F test (estimateof confidence;solid line) are plottedversus period. The shadedareas represent significantfrequency bands.

1995], which was attributed to interactions between oceanic The low expressionof this cycle in the Sulu Sea during the last circulationchanges and atmospheric ]4C changesat a 2.3 kyr terminationis perhapsdue to the strong overprint of the 1.5 kyr period. This cycle is not expressedin TerminationI of the Sulu cycle on the PP record. This cycle seems to be stationary Sea PP record. However, it is significantfor the glacials stages. throughoutthis record. 500 DE GARIDEL-THORON ET AL.' DYNAMICS OF THE EAST ASIAN WINTER MONSOON

5.2. Pseudo 1.5 kyr Cyclicity Age (kyrs BP)

A 1.47 kyr period was first describedby Dansgaard et al. o 20 40 60 80 lOO 12o 14o 16o [1984] in the Camp century 6•80 ice core record. In theSummit ice lO core(GISP2) the 6•80 andthe polar component of a principal componentanalysis computed on variouschemical markers in the ice both show periods•,,1.5 kyr [Mayewsla'et al., 1997]. A 1.47 '- 5 kyr period was also describedby Bond et al. [1997] in North Atlantic deep-seacores during the Holoceneand the deglaciation. In Alaska, a lake recordcontains climatic variations with a 1.5 kyr cycle [Campbell et al., 1998]. These recordsprovide growing evidencefor cyclicityof •,,1.5 kyr in many high-latituderecords. This cycle was also recentlylinked to fluctuationsin continental -5 ice massduring periods of loweredsea level (-45 m belowpresent level) by a simplemodel of ice dynamics[Schulz et al., 1999]. At lower latitudes the southwest Indian monsoon also shows clear -lO I , , , I 11 Ii :: : I, , ,I , ,, I ,, , I, , ,I 15 1.45 kyr cycles in a paleoproductivityrecord spanningthe last deglaciation[Sirocko et al., 1996]. Sirockoet al. speculatethat this 10 relationshipcould be due to subprecessionalforcing. In the Sulu Sea the primaryproductivity shows clear peaks in a broader1.5 kyr band(Figure 7). The peak in PP is at 1.38 kyr for the last 22 kyr. For stage2-3 the major peak is at 1.37 kyr. That •,,1.5 kyr cyclicity is also presentduring stage6. During the last deglaciationthe cycles are well defined. This pseudoperiodicity -5 appearssignificant in the threetime slices,suggesting that it is a pervasive feature of the East Asian winter monsoondynamics. -10 Our low-latituderecord clearly doesnot record an amplification with increasein the polar continentalice masses(i.e., the sea -15 level does not modulate the envelope of the 1.5 kyr cycle). 0 20 40 60 80 100 120 140 160 During the last 160 kyr, eight maxima in the 1.5 kyr envelope Age (kyrs BP) were counted,which correspondroughly to the precessionperiod (160/8 = 20 kyr) (Figure 8b). The East Asian monsoon1.5 kyr Figure 8. FilteredPP with a Gaussianfilter centeredon (a) 6000 cycle thuscould be the climaticexpression of a combinationtone yearsand (b) 1500 years.The envelopeof the 1500 yearsfiltered of the orbital insolation frequenciesas already expressedby signalexhibits a •,,20,000 yearsmodulation. Pestiaux et al. [1988] and Sirocko et al. [1996]. The pervasive occurrenceof this cycle during both glacial and interglacial periods in the subtropicsseems to indicate that if a 1.5 kyr climateoscillator exists, a commonorigin betweenhigh and low latitudes is expected for this cyclicity. We suggest that its with peaks of the East Asian monsoonrecorded in the Chinese presencein the Sulu Sea is not forced by high latitudes, as loess.In the timescaleclose of the pacingof Dansgaard-Oeschger indicatedby its presenceduring stages1 and 5 in the Sulu Sea cycles,singular-spectrum analysis reveals that the PP recordin the and absencein the intervals in the GRIP record [Stuiver and Sulu Sea can be correlated with the Greenland climate. Therefore a Braziunas, 1993]. Further data are necessaryto improve the commondynamic, which is not forcedby icebergsdischarges in definition of the frequencybands and the geographicalextent the North Atlantic, is present in both high- and low-latitude of theserapid climate cycles. paleoclimaticrecords. 4. Four dominant frequencieswere isolated in the high- frequencypacing of the East Asian winter monsoon.They occur 6. Conclusions (1) at •6 kyr, (2) in a 3.3-4.2 kyr frequencyband of unknown origin, (3) in a 2.4 kyr frequencyband probablyresulting from We reconstructedprimary productivity (PP) fromcoccoliths in the the couplingbetween solar flux and oceanicprocesses, and (4) MD97-2141 core, locatedin the Sulu Sea duringthe last 200 kyr. in a 1.5 kyr band. This •1.5 kyr pseudoperiodicityis a 1. We find that PP increasedduring the glacial stages,whereas pervasivefeature of the East Asian winter monsoonduring the the interglacialsare timesof lower PP.We attributethis PP change marinestages 1, 3, 5, and 6. The origin of this climatic cyclicity to a strengtheningof the East Asian winter monsoonin the Sulu remainsunknown, but we suggestthat its presenceat this site in Sea duringglacial stages. the low-latitude western Pacific is not forced from the high 2. The PP record indicatesthat an abrupt decreasein the East latitudes. Asian winter monsoonoccurred 14.55 calendarkyr B.P., followed by a •,,2 kyr plateauduring both Allerod and YoungerDryas. It showsthat changes inthe •80 recordfrom the Sulu Sea during the Acknowledgments. The supportof French MENRT, TAAF, CNRS/ Younger Dryas are most likely linked with sea surfacesalinities INSU, and IFRTP to the Marion-DuJ?esneand the IMAGES Program changes. was necessaryto performthis work. T.G. was supportedby a predoctoral 3. The PP recordexhibits eight rapid oscillationsduring the last grant of the French Ministbre de la Recherche.Financial supportfrom 70 kyr. Only four PP eventsare synchronouswith HeinrichEvents. INSU grant to TG. and L.B. is acknowledged.B.K.L. acknowledges This implies that the East Asian winter monsoon and North grant OCE 9710156 and technical assistantStephen Howe. We are indebted to Delia Oppo, who contributed to the collection of the Atlantic iceberg dischargesfollow different dynamics,and that AMS14Cages at theWHOI-NOSAMS AMS facilityand •180 data with HE are not systematicallyforcing increasesin the East Asian the NSF grant OCE-9710097, and the technicalhelp of SusanTrimarchi wintermonsoon. However, these PP oscillationsappear to correlate and Luping Zou. DE GARIDEL-THORON ET AL.: DYNAMICS OF THE EAST ASIAN WINTER MONSOON 501

References Alley, R. B., P. A. Mayewski, T. Sowers,M. Eos Trans. AGU, 79(45), Fall Meet. Suppl., Climatology,339 pp., JohnWiley, New York, Stuiver,K. C. Taylor, and P. U. Clark, Holo- F454, 1998. 1998. cene climatic instability:A prominent,wide- Dansgaard,W., S. J. Johnsen,H. B. Clausen,N. Miao, Q., R. Thunell, and D. M. Anderson,Gla- spread event 8200 yr ago, Geology, 25(6), Dahl-Jensen,N. Gundestrup,and C. U. Ham- cial-Holocene carbonate dissolution and sea sur- 483-486, 1997. mer, North Atlantic climatic oscillations re- facetemperatures in the SouthChina and Sulu Anderson,D. M., and R. C. Thunell, The oxy- vealed by deep Greenland ice cores, in Seas,Paleoceanography, 9(2), 269- 290, 1994. gen-isotopecomposition of tropicalocean sur- Climate Processesand Climate Sensitivity, Mix, A., Influence of productivityvariations on face water during the last deglaciation,•uat. Geophys.Monogr. Set., vol. 29, edited by J. long-term atmospheric CO2, Nature, 337, Sci. Rev., 12, 456-473, 1993. E. Hansen and T. Takahashi,pp. 288-298, 541-544, 1989. Antoine, D., and A. Morel, Oceanicprimary pro- AGU, Washington,D.C., 1984. Molfino, B., and A. Mcintyre, Precessionalfor- duction,2, Estimationat global scalefrom sa- Dansgaard,W., et al., Evidencefor generalin- cing of nutriclinedynamics in the equatorial tellite (coastalzone color scanner) chlorophyll, stability of past climate from a 250 kyr ice- Atlantic, Science,249, 766- 769, 1990. Global Biogeochem.Cycles, 1 O, 57- 70, 1996. core record, Nature, 364, 218-220, 1993. Naidu, P. D., and B. A. Malgrem, A 2,200 years Bard, E., Correlationof acceleratormass spectro- Duplessy,J.-C., E. Bard,M. Arnold,N.J. Shack- periodicityin the Asianmonsoon system, Geo- metry•4C ages measured in planktonic fora- leton,J. Duprat,and L. Labeyrie,How fastdid phys.Res. Lett., 22(17), 2361- 2364, 1995. minifera: Paleoceanographicimplications, the ocean-atmospheresystem run during the Nair, R. R., V. Ittekkot, S. J. Manganini, V. Paleoceanography,3(6), 635-645, 1988. last deglaciation?,Earth Planet. Sci. Lett., Ramaswamy,B. Haake, E. T. Degens, B. N. Bard, E., Geochemicaland geophysicalimplica- 103, 27-40, 1991. Desai, and S. Honjo, Increasedparticle flux to tions of the radiocarboncalibration, Geochim. Fairbanks,R. G., A 17000 year glacio-eustatic the deep ocean related to monsoons,Nature, Cosmochim.Acta, 62(12), 2025-2038, 1998. sea-levelrecord: Influence of glacial melting 338, 749- 751, 1989. Bard, E., M. Arnold, R. G. Fairbanks, and B. rates on the Younger Dryas event and deep- Okada, H., and S. Honjo, The distribution of Hamelin,e3øTh-e34U and •4C ages obtained oceancirculation, Nature, 342, 637- 642, 1989. oceaniccoccolithophorids in the Pacific,Deep by massspectrometry on corals,in Calibration Heinrich,H., Origin and consequencesof cyclic Sea Res., 20, 355-374, 1973. 1993, editedby M. Stuiver,A. Long, and R. S. ice rafting in the northeastAtlantic duringthe Oppo,D. W., B. K. Linsley,S. Dannenmann,and Krea, pp. 191-200, Dcp. of Geosci.,Univ. of past 130 000 years, Quat. Res., 29, 142-152, R. D. Norris, Centennial-to-millennial scale Ariz., Tucson, 1993. 1988. variability recorded in Holocene sediments Bard, E., F. Rostek,and C. Sonzogni,Interhemi- Imbrie, J., J. D. Hays, D. G. Martinson,A. Mcin- from the Sulu and SouthChina Seas,Eos Trans. sphericsynchrony of the last deglaciationin- tyre, A. C. Mix, J. J. Morley, N. G. Pisias,W. AGU, 79(45), Fall Meet. Suppl.,F454, 1998. ferred from alkenone palaeothermometry, L. Prell, and N.J. Shackleton,The orbital the- Paillard,D., L. Labeyrie,and P. Yiou, Macintosh Nature, 385, 707- 710, 1997. ory of Pleistoceneclimate: Support from a re- performs time-series analysis, Eos Trans. Beaufort, L., Y. Lancelot, P. Camberlin, O. visedchronology ofthe marine •80 record,in AGU, 77(39), 379, 1996. Cayre, E. Vincent,F. Bassinot,and L. Labeyr- Milankovitch and Climate, edited by A. L. Patterson, R. T., and E. Fishbein, Re-examina- ie, Insolationcycles as a major controlof equa- Berger et al., pp. 269-305, D. Reidel, Nor- tion of the statistical methods used to deter- torial Indian Ocean primary production, well, Mass., 1984. mine the number of point countsneeded for Science, 278, 1451-1454, 1997. Jouzel, J., Calibrating the isotopicpaleotherm- micropaleontologicalquantitative research, d. Behl, R., andJ.P. Kennet,Briefinterstadial events ometer, Science,286, 910- 911, 1999. Paleontol., 63(2), 245-248, 1989. in the SantaBarbara basin, NE Pacific,during Kudrass, H. R., H. Erlenkeuser,R. Vollbrecht, Pelejero,C., J. O. Grimalt, S. Heilig, M. Kie- the past60 kyr, Nature,379, 243-246, 1996. and W. Weiss, Global nature of the Younger nast,and L. Wang,High-resolution u3k?tem - Bond,G., et al., Evidencefor massivedischarges Dryas coolingevent inferred from oxygeniso- peraturereconstructions in the South China of icebergsinto the North Atlantic Oceandur- tope data from Sulu sea cores, Nature, 349, Sea over the past220 kyr, Paleoceanography, ing the last glacial period, Nature, 360, 245- 406- 409, 1991. 14(2), 224-231, 1999a. 249, 1992. Kuehl, S. A., T. G. Fuglseth,and R. C. Thunell, Pelejero,C., J. O. Grimalt, M. Samthein,L. Wang, Bond, G., W. Showers,M. Chezebiet,R. Lotti, P. Sedimentmixing andaccumulation rates in the and J. A. Flores,Molecular biomarkerrecord of Almasi, P. deMenocal,P. Priore, H. Cullen, I. Sulu and the South China Seas:Implications sea-surfacetemperature and climaticchange in Hajdas,and G. Bonani,A pervasivemillennial for organicpreservation in deep-seasediments, the South China Sea during the last 140,000 scale cycle in North-Atlantic Holocene and Mar. Geol., 111, 15- 35, 1993. years,Mar. Geol., 156, 109-121, 1999b. glacial climates, Nature, 278, 1257-1266, Kutzbach,J. E., P.J. Guetter,P. J. Behling,and R. Pelejero,C., M. Kienast,L. Wang,and J. O. Gri- 1997. Selin, Simulatedclimatic changes:Results of malt, The floodingof Sundalandduring the last Campbell,I. D., C. Campbell,M. J. Apps,N. W. the COHMAP climate-modelexperiments, in deglaciation:Imprints in hemipelagic sedi- Rutter, and A. B. G. Bush, Late Holocene ca. Global Climates Since the Last Glacial Maxi- ments from the southern South China Sea, 1500 yr climaticperiodicities and their impli- mum,edited by H. E. Wright,Jr. et al., pp. 24- Earth Planet. Sci. Lett., 171, 661 - 671, 1999c. cations,Geology, 26(5), 471-473, 1998. 93, Univ. of Minn. Press,Minneapolis, 1993. Pestiaux,P., I. Van Der Mersch, A. Berger,and Chapman,M. R., and N.J. Shackleton,Millen- Le Treut, H., and M. Ghil, Orbital forcing,cli- J.-C. Duplessy,Paleoclimatic variability at fre- nial-scale fluctuations in North Atlantic heat matic interactionsand glaciationscycles, d. quenciesranging from 1 cycle per 10,000 flux duringthe last 150,000 years,Earth Pla- Geophys.Res., 88, 5167-5190, 1983. years to 1 cycle per 1000 years--Evidence net. Sci. Lett., 159, 57-70, 1998. Linsley,B. K., Oxygen-isotoperecord of sea-le- for non-linearbehaviour of the climatesystem, Chen, F. H., J. Bloemendal, J. M. Wang, J. Li, vel and climate variations in the Sulu Sea over Clim. Change, 12, 9-37, 1988. and L. Oldfield, High-resolutionmulti-proxy the past 150,000years, Nature, 380, 234-237, Porter, S. C., and A. Zhisheng,Correlation be- climate records from Chinese loess: Evidence 1996. tween climate events in the North Atlantic and for rapid climaticchanges over the last 75 kyr, Linsley,B. K., and R. C. Thunell,The recordof China duringthe last glaciation,Nature, 3 75, Palaeogeogr. Palaeoclimatol. Palaeoecol., deglaciationin the Sulu Sea:Evidence for the 305-308, 1995. 130, 323-335, 1997. YoungerDryas event in the tropical western Schulz, H., U. von Rad, and H. Erlenkeuser, Chen, M. T., and C. Y. Huang, Ice-volume for- Pacific, Paleoceanography,5(6), 1025-1039, Correlation between Arabian Sea and Green- cing of the winter monsoonclimate in the 1990. land climate oscillationsof the past 110,000 South China Sea, Paleoceanography,13(6), Linsley,B. K., R. C. Thunell,C. Morgan, and D. years,Nature, 393, 54-57, 1998. 622-633, 1998. Williams, Oxygen minimumexpansion in the Schulz,M., W. H. Berger, M. Sarnthein,and P. Cortijo, E., L. Labeyrie, M. Elliot, E. Balbon, Sulu Sea, western equatorialPacific, during M. Grootes,Amplitude variations of 1470-year and N. Tisnerat,Rapid climaticvariability of the last glacial low stand of sea level, Mar. climate oscillations during the last 100,000 the North Atlantic Oceanand globalclimate: A Micrøpaleøntøl" 9, 395-418, 1985. yearslinked to fluctuationsof continentalice focus of the IMAGES program, Quat. Sci. Mayewski, P. A., L. D. Meeker, M. S. Twickler, mass, Geophys. Res. Lett., 26(22), 3385- Rev., 19, 2000. S. Whitlow, Q. Yang, W. B. Lyons, and M. 3388, 1999. Damon, P. E., and J. L. Jirikowic, The sun as a Prentice,Major featuresand forcing of high- Sirocko,F., D. Garbe-Sch--nberg,A. Mcintyre, low-frequencyharmonic oscillator, Radiocar- latitudeNorthern Hemisphere atmospheric cir- and B. Molfino, Teleconnectionsbetween the bon, 34(2), 199-205, 1992. culationusing a 110,000-yearlong glaciogeo- subtropicalmonsoons and high-latitude cli- Dannenmann,S., B. K. Linsley,D. W. Oppo,and chemical series, d. Geophys. Res., 102, mates during the last deglaciation,Science, R. D. Norris, Centennial resolution isotopic 26,345-26,366, 1997. 2 72, 526- 529, 1996. record from the Sulu Sea: 10 kyr to 60 kyr, McGregor, G. R., and S. Nieuwolt, Tropical Stuiver, M., and T. F. Braziunas, Sun, ocean, 502 DE GARIDEL-THORON ET AL.: DYNAMICS OF THE EAST ASIAN WINTER MONSOON

climateand atmospheric 14CO2' An evaluation to paleoclimatictime series,Physica D, 35, Jansen,and P. A. Mayewski, Paleoclimatic of causaland spectralrelationships, Holocene, 395-424, 1989. variability inferred from the spectralanalysis 3(4), 289-305, 1993. Wang, L., M. Sarnthein,H. Erlenkeuser,J. Gri- of Greenland and Antartic ice-core data, J. Stuiver, M., and P. J. Reimer, Calib Rev. 3, malt, P. Grootes,S. Heilig, E. Ivanova,M. Kie- Geophys.Res., 102, 26, 441-26, 454, 1997. Radiocarbon, 35, 215-230, 1993. nast,C. Pelejero,and U. Pflaumann,East Asian Zhang, Y., East Asian winter monsoon:Results Thompson,P. R., A. W. H. Bb, J.-C. Duplessy, monsoonduring the latePleistocene: high-reso- from eight AMIP models,Clim. Dyn., 13(11), and N. C. Shackleton,Disappearance of pink- lution sediment records from the South China 797- 820, 1997. pigmentedGlobigerinoides tuber at 120,000 Sea, Mar. Geol., 156, 245-284, 1999. yr BP in the Indian and Pacific Oceans,Nat- Wiesner,M. G., L. Zheng, H. K. Wong, Y. Wang, ure, 280, 554-558, 1979. and W. Chen, Fluxes of particulatematter in Thomson,D. J., Spectrumestimation and harmo- the South China Sea, in Particle Flux in the L. Beaufort and T. de Garidel-Thoron, Centre nic analysis, IEEE Proc., 70, 1055, 1982. Ocean, editedby J. Ittekkot et al., pp. 293- Europbende Rechercheet d'Enseignementen Thouveny,N., E. Moreno,D. Delanghe,L. Can- 312, JohnWiley, New York, 1996. G•osciences de l'Environnement, Europ61e don, Y. Lancelot, and N.J. Shackleton,Rock- Xiao, J. L., Z. S. An, T S. Liu, Y. Inouchi, H. mbditerranben de l'Arbois, BP 80, Aix-en- magneticdetection of distalice rafted debries: Kumai, S. Yoshikawa, and Y. Kondo, East Provence cedex 4, 13545 France. (garidel• Clue for the identificationof Heinrich layers Asian monsoon variation during the last cerege.fr) on the Portuguesemargin, Earth Planet. Sci. 130,000 years:Evidence from the LoessPla- S. Dannenmannand B. K. Linsley, Depart- Lett., 180, 61-75, 2000. teau of centralChina and Lake Biwa of Japan, ment of Earth and Atmospheric Sciences, Thunell, R. C., and Q. C. Miao, Sea-surfacetem- Quat. Sci. Rev., 18, 147-157, 1999. University at Albany, State University of New peratureof the westernequatorial Pacific dur- Yan, X.-H., C.-R. Ho, Q. Zheng, and V. Klemas, York, Albany, NY 12222, USA. ing the YoungerDryas, Quat. Res.,46, 72-77, Temperatureand size variabilitiesof the Wes- 1996. tern Pacific Warm Pool, Science, 258, 1643- (ReceivedJune 26, 2000; Vautard,R., and M. Ghil, Singularspectrum ana- 1645, 1992. revisedApril 26, 2001; lysis in nonlineardynamics, with applications Yiou, P., K. Fuhrer, L. D. Meeker, J. Jouzel, E. acceptedMay 25, 2001.)