kim cobb's borneo stalagmite talk - agu 2015

Kim Cobb (Georgia Tech) @coralsncaves

Stacy Carolin (Oxford) Sang Chen (Caltech) David Lund (U. Connecticut) Nele Meckler (U. Bergen) Jess F. Adkins (Caltech) Sharon Hoffmann (UNC-Wilmington) Julien Emile-Geay (U. Southern California) Andrew A. Tuen (U. Sans Malaysia) Alison Pritchard, Brian Clark, Syria Lejau, Jenny Malang (Gunung Mulu National Park) Jean Lynch-Stieglitz (Georgia Tech) Jessica Moerman (Johns Hopkins)

Western tropical Pacific hydrology as inferred from northern Borneo speleothems: interannual to orbital-scale variability

OK

Varied Response of Western PacificHydrology to Climate Forcings overthe Last Glacial PeriodStacy A. Carolin,1* Kim M. Cobb,1 Jess F. Adkins,2 Brian Clark,3 Jessica L. Conroy,1 Syria Lejau,3

Jenny Malang,3 Andrew A. Tuen4

Atmospheric deep convection in the west Pacific plays a key role in the global heat and moisturebudgets, yet its response to orbital and abrupt climate change events is poorly resolved. Here,we present four absolutely dated, overlapping stalagmite oxygen isotopic records from northern Borneothat span most of the last glacial cycle. The records suggest that northern Borneo’s hydroclimateshifted in phase with precessional forcing but was only weakly affected by glacial-interglacial changesin global climate boundary conditions. Regional convection likely decreased during Heinrich events,but other Northern Hemisphere abrupt climate change events are notably absent. The new recordssuggest that the deep tropical Pacific hydroclimate variability may have played an important role inshaping the global response to the largest abrupt climate change events.

The response of the tropical Pacific tochanges in Earth’s climate system remainshighly uncertain. Themost recent glacial-

interglacial cycle encompasses several preces-sional cycles; changes in ice volume, sea level,global temperature, and atmospheric partial pres-sure of CO2; and millennial-scale climate events,thus providing insights into the tropical Pacificresponse to a variety of climate forcings. Chinesestalagmites show that East Asian monsoonstrength closely tracks precessional insolationforcing over several glacial-interglacial cycles andexhibits prominent millennial-scale variability(1, 2). The timing and structure of these abruptclimate changes are nearly identical tomillennial-scale events recorded in the Greenland ice cores[Dansgaard-Oeschger (D/O) events] (3) and insediment records that document ice-rafted debrisacross the North Atlantic (Heinrich events) (4, 5).A Borneo stalagmite record spanning the past27,000 years provides a markedly different viewof hydrology in the western tropical Pacific, withthe Heinrich 1 excursion and spring-fall preces-sional insolation forcing explaining much of thevariability (6). At its most basic, this finding il-lustrates the complexity of regional responses tovarious climate forcings, especially at sites lo-cated far from the North Atlantic, and demands amore exhaustive tropical Pacific hydrologic recordencompassing a full glacial-interglacial cycle.

Here, we present four overlapping stalagmiteoxygen isotopic (d18O) records fromGunungBudaand GunungMulu national parks, located in north-ern Borneo (4°N, 115°E) (fig. S1), that togetherspan most of the last glacial cycle. The researchsite is located near the center of the west Pacificwarm pool (WPWP), where changes in sea surfacetemperatures and sea-level pressure have consid-

erable impacts on large-scale atmospheric circu-lation and global hydrology (7). Using multiplestalagmites from different caves, we distinguishshared climate-related features from cave-specificsignals in the overlapping d18O records.

The four stalagmite records span portions ofthe last glacial cycle with many intervals of over-lap, based on U-series dates (Fig. 1). Stalagmiteswere recovered fromSecret Cave atGunungMulu[SC02, 37 to 94 thousand years before the present(ky B.P.), and SC03, 32 to 100 ky B.P.] and fromBukit Assam (BA02, 15 to 46 ky B.P.) and SnailShell Cave (SCH02, 31 to 73 ky B.P.) at GunungBuda, 25 km from Gunung Mulu (fig. S2). Thedeglacial andHolocene d18O records from stalag-mite SCH02 were presented in (6). Eighty-sixnew U/Th dates measured across the four stalag-mites fall in stratigraphic order within 2s errors(8). Large uncertainties in the 230Th/232Th ratio ofthe contaminant phases translate into large un-certainties associated with the correction for de-trital thorium contamination. Fourteen isochronsmeasured across stalagmites from three separatecaves give initial 230Th/232Th atomic ratios of56 T 11 (2s) for Bukit Assam Cave, 59 T 13 (2s)for Snail Shell Cave, and 111 T 41 (2s) × 10−6 forSecret Cave (8), which fall within the range ofpreviously published values from our site (6).Absolute age errors for each U/Th date were cal-culated with a Monte Carlo approach that com-bined multiple sources of error. The resultingdating errors average T200, T250, T400, andT500 years (2s) for BA02, SCH02, SC02, andSC03, respectively. Age models were initiallyconstructed by linearly interpolating betweeneach date and were refined by aligning five majormillennial-scale d18O excursions visible acrossall four records within age error (8). The fact thatboth chronologies fall nearly completely withinthe StalAge (9) algorithm’s 95% confidence in-terval (figs. S3 to S6) adds credibility to our as-signed chronologies and associated error estimates.With our 1-mm sampling interval, the temporalresolution of the associated d18O records aver-ages 60 years per sample for the faster-growing

stalagmites BA02 and SCH02 and 200 years persample for the slower-growing stalagmites SC02and SC03. During the 50– to 38–ky B.P. interval,SC02 and SC03 were sampled at 0.5-mm toachieve a resolution of ~100 years per sample.Ultraslow growth intervals (<10 mm/year for thefaster-growing stalagmites and <3 mm/year forthe slower growing stalagmites) may representunresolved hiatuses and, as such, were excludedfrom the resulting paleoclimate reconstructions(8), following (6).

The stalagmite d18O records provide recon-structions of rainfall d18O variability at the re-search site, which, in turn, tracks the strength ofregional convective activity (10). Consistent withthe tropical amount effect (11, 12), rainfall d18Ovariations measured at the site from 2006 to 2011are significantly anticorrelated with regional pre-cipitation amount and closely track the El Niño–Southern Oscillation on monthly time scales (10).A weak semi-annual seasonal cycle in rainfalld18O is characterized by relative minima in Juneto July and November to January and relativemaxima in February to April and August to Oc-tober. Such a pattern suggests that the twice-yearly passage of the Intertropical ConvergenceZone (ITCZ) over the site is associatedwith shiftsin the moisture sources and/or trajectories thatdrive the observed seasonal fractionations (10).Dripwater d18O values match rainfall d18O valuesaveraged over the preceding 2 to 6 months (13),suggesting a short residence time of dripwaterd18O relative to our centennial-scale sampling ofstalagmite d18O. Time series of Buda and Mulustalagmite d18O are highly reproducible (6, 14),strongly supporting their interpretation as rainfalld18O reconstructions and, by extension, as recordsof past regional convective activity.

The overlapping Borneo stalagmite d18Orecords show orbital-scale variability related toprecessional insolation forcing and glacial-interglacial (G-I) changes (Fig. 2). The similarityof our stalagmite d18O time series to indices ofG-I variability greatly diminishes after removingthe mean d18O of seawater due to changes in icevolume (8, 15) from the Borneo d18O records(Fig. 2 and fig. S7). After this correction, LastGlacial Maximum (LGM) d18O values are nearlyidentical to d18O values at ~85 ky B.P., despite thepresence of substantially larger ice sheets, coolerregional temperatures (16, 17), and a completelyexposed Sunda Shelf during the LGM. In partic-ular, Sunda Shelf emergence has been implicatedin shaping glacial western tropical Pacific hydro-climate in previous studies (6, 18, 19). However,we find little correspondence between Borneostalagmite d18O and an index of Sunda Shelf arealextent over the entire glacial cycle (fig. S7). Forexample, Borneo stalagmite d18O variations in the70– to 90–ky B.P. interval bear little resemblanceto reconstructed sea-level changes, especially from~71 to 76 ky B.P. (20), when a large drop in sealevel almost doubled the size of the exposed shelf(fig. S7). As such, the new Borneo d18O recordssuggest that the cumulative influence of G-I

1School of Earth and Atmospheric Sciences, Georgia Institute ofTechnology, Atlanta, GA 30332, USA. 2Division of Geologicaland Planetary Sciences, California Institute of Technology,Pasadena, CA91125,USA. 3GunungMuluNational Park, Sarawak,Malaysia. 4Institute of Biodiversity and Environmental Con-servation, Universiti Malaysia Sarawak, Sarawak, Malaysia.

*Corresponding author. E-mail: [email protected]

28 JUNE 2013 VOL 340 SCIENCE www.sciencemag.org1564

REPORTS

Varied Response of Western PacificHydrology to Climate Forcings overthe Last Glacial PeriodStacy A. Carolin,1* Kim M. Cobb,1 Jess F. Adkins,2 Brian Clark,3 Jessica L. Conroy,1 Syria Lejau,3

Jenny Malang,3 Andrew A. Tuen4

Atmospheric deep convection in the west Pacific plays a key role in the global heat and moisturebudgets, yet its response to orbital and abrupt climate change events is poorly resolved. Here,we present four absolutely dated, overlapping stalagmite oxygen isotopic records from northern Borneothat span most of the last glacial cycle. The records suggest that northern Borneo’s hydroclimateshifted in phase with precessional forcing but was only weakly affected by glacial-interglacial changesin global climate boundary conditions. Regional convection likely decreased during Heinrich events,but other Northern Hemisphere abrupt climate change events are notably absent. The new recordssuggest that the deep tropical Pacific hydroclimate variability may have played an important role inshaping the global response to the largest abrupt climate change events.

The response of the tropical Pacific tochanges in Earth’s climate system remainshighly uncertain. Themost recent glacial-

interglacial cycle encompasses several preces-sional cycles; changes in ice volume, sea level,global temperature, and atmospheric partial pres-sure of CO2; and millennial-scale climate events,thus providing insights into the tropical Pacificresponse to a variety of climate forcings. Chinesestalagmites show that East Asian monsoonstrength closely tracks precessional insolationforcing over several glacial-interglacial cycles andexhibits prominent millennial-scale variability(1, 2). The timing and structure of these abruptclimate changes are nearly identical tomillennial-scale events recorded in the Greenland ice cores[Dansgaard-Oeschger (D/O) events] (3) and insediment records that document ice-rafted debrisacross the North Atlantic (Heinrich events) (4, 5).A Borneo stalagmite record spanning the past27,000 years provides a markedly different viewof hydrology in the western tropical Pacific, withthe Heinrich 1 excursion and spring-fall preces-sional insolation forcing explaining much of thevariability (6). At its most basic, this finding il-lustrates the complexity of regional responses tovarious climate forcings, especially at sites lo-cated far from the North Atlantic, and demands amore exhaustive tropical Pacific hydrologic recordencompassing a full glacial-interglacial cycle.

Here, we present four overlapping stalagmiteoxygen isotopic (d18O) records fromGunungBudaand GunungMulu national parks, located in north-ern Borneo (4°N, 115°E) (fig. S1), that togetherspan most of the last glacial cycle. The researchsite is located near the center of the west Pacificwarm pool (WPWP), where changes in sea surfacetemperatures and sea-level pressure have consid-

erable impacts on large-scale atmospheric circu-lation and global hydrology (7). Using multiplestalagmites from different caves, we distinguishshared climate-related features from cave-specificsignals in the overlapping d18O records.

The four stalagmite records span portions ofthe last glacial cycle with many intervals of over-lap, based on U-series dates (Fig. 1). Stalagmiteswere recovered fromSecret Cave atGunungMulu[SC02, 37 to 94 thousand years before the present(ky B.P.), and SC03, 32 to 100 ky B.P.] and fromBukit Assam (BA02, 15 to 46 ky B.P.) and SnailShell Cave (SCH02, 31 to 73 ky B.P.) at GunungBuda, 25 km from Gunung Mulu (fig. S2). Thedeglacial andHolocene d18O records from stalag-mite SCH02 were presented in (6). Eighty-sixnew U/Th dates measured across the four stalag-mites fall in stratigraphic order within 2s errors(8). Large uncertainties in the 230Th/232Th ratio ofthe contaminant phases translate into large un-certainties associated with the correction for de-trital thorium contamination. Fourteen isochronsmeasured across stalagmites from three separatecaves give initial 230Th/232Th atomic ratios of56 T 11 (2s) for Bukit Assam Cave, 59 T 13 (2s)for Snail Shell Cave, and 111 T 41 (2s) × 10−6 forSecret Cave (8), which fall within the range ofpreviously published values from our site (6).Absolute age errors for each U/Th date were cal-culated with a Monte Carlo approach that com-bined multiple sources of error. The resultingdating errors average T200, T250, T400, andT500 years (2s) for BA02, SCH02, SC02, andSC03, respectively. Age models were initiallyconstructed by linearly interpolating betweeneach date and were refined by aligning five majormillennial-scale d18O excursions visible acrossall four records within age error (8). The fact thatboth chronologies fall nearly completely withinthe StalAge (9) algorithm’s 95% confidence in-terval (figs. S3 to S6) adds credibility to our as-signed chronologies and associated error estimates.With our 1-mm sampling interval, the temporalresolution of the associated d18O records aver-ages 60 years per sample for the faster-growing

stalagmites BA02 and SCH02 and 200 years persample for the slower-growing stalagmites SC02and SC03. During the 50– to 38–ky B.P. interval,SC02 and SC03 were sampled at 0.5-mm toachieve a resolution of ~100 years per sample.Ultraslow growth intervals (<10 mm/year for thefaster-growing stalagmites and <3 mm/year forthe slower growing stalagmites) may representunresolved hiatuses and, as such, were excludedfrom the resulting paleoclimate reconstructions(8), following (6).

The stalagmite d18O records provide recon-structions of rainfall d18O variability at the re-search site, which, in turn, tracks the strength ofregional convective activity (10). Consistent withthe tropical amount effect (11, 12), rainfall d18Ovariations measured at the site from 2006 to 2011are significantly anticorrelated with regional pre-cipitation amount and closely track the El Niño–Southern Oscillation on monthly time scales (10).A weak semi-annual seasonal cycle in rainfalld18O is characterized by relative minima in Juneto July and November to January and relativemaxima in February to April and August to Oc-tober. Such a pattern suggests that the twice-yearly passage of the Intertropical ConvergenceZone (ITCZ) over the site is associatedwith shiftsin the moisture sources and/or trajectories thatdrive the observed seasonal fractionations (10).Dripwater d18O values match rainfall d18O valuesaveraged over the preceding 2 to 6 months (13),suggesting a short residence time of dripwaterd18O relative to our centennial-scale sampling ofstalagmite d18O. Time series of Buda and Mulustalagmite d18O are highly reproducible (6, 14),strongly supporting their interpretation as rainfalld18O reconstructions and, by extension, as recordsof past regional convective activity.

The overlapping Borneo stalagmite d18Orecords show orbital-scale variability related toprecessional insolation forcing and glacial-interglacial (G-I) changes (Fig. 2). The similarityof our stalagmite d18O time series to indices ofG-I variability greatly diminishes after removingthe mean d18O of seawater due to changes in icevolume (8, 15) from the Borneo d18O records(Fig. 2 and fig. S7). After this correction, LastGlacial Maximum (LGM) d18O values are nearlyidentical to d18O values at ~85 ky B.P., despite thepresence of substantially larger ice sheets, coolerregional temperatures (16, 17), and a completelyexposed Sunda Shelf during the LGM. In partic-ular, Sunda Shelf emergence has been implicatedin shaping glacial western tropical Pacific hydro-climate in previous studies (6, 18, 19). However,we find little correspondence between Borneostalagmite d18O and an index of Sunda Shelf arealextent over the entire glacial cycle (fig. S7). Forexample, Borneo stalagmite d18O variations in the70– to 90–ky B.P. interval bear little resemblanceto reconstructed sea-level changes, especially from~71 to 76 ky B.P. (20), when a large drop in sealevel almost doubled the size of the exposed shelf(fig. S7). As such, the new Borneo d18O recordssuggest that the cumulative influence of G-I

1School of Earth and Atmospheric Sciences, Georgia Institute ofTechnology, Atlanta, GA 30332, USA. 2Division of Geologicaland Planetary Sciences, California Institute of Technology,Pasadena, CA91125,USA. 3GunungMuluNational Park, Sarawak,Malaysia. 4Institute of Biodiversity and Environmental Con-servation, Universiti Malaysia Sarawak, Sarawak, Malaysia.

*Corresponding author. E-mail: [email protected]

28 JUNE 2013 VOL 340 SCIENCE www.sciencemag.org1564

REPORTS

24. J. J. Warren, T. A. Tronic, J. M. Mayer, Chem. Rev. 110,6961 (2010).

25. M. K. Mahanthappa, K.-W. Huang, A. P. Cole,R. M. Waymouth, Chem. Commun. (Camb.) 2002, 502(2002).

26. M. K. Mahanthappa, A. P. Cole, R. M. Waymouth,Organometallics 23, 836 (2004).

27. K. Schröder et al., Organometallics 27, 1859 (2008).28. J. M. Mayer, Acc. Chem. Res. 44, 36 (2011).29. J. Jin, M. Newcomb, J. Org. Chem. 73, 7901 (2008).30. S. Morrison, Electrochemistry at Semiconductor and

Oxidized Metal Electrodes (Plenum, New York, 1980).31. B. Enright, G. Redmond, D. Fitzmaurice, J. Phys. Chem.

98, 6195 (1994).32. F. Marken, A. S. Bhambra, D.-H. Kim, R. J. Mortimer,

S. J. Stott, Electrochem. Commun. 6, 1153 (2004).33. M. Zukalova, M. Kalbáč, L. Kavan, I. Exnar, M. Graetzel,

Chem. Mater. 17, 1248 (2005).

34. A. Minguzzi, F.-R. F. Fan, A. Vertova, S. Rondinini,A. J. Bard, Chem. Sci. 3, 217 (2012).

35. Y. Ren, L. J. Hardwick, P. G. Bruce, Angew. Chem. Int. Ed.49, 2570 (2010).

36. C. Venkataraman, A. V. Soudackov, S. Hammes-Schiffer,J. Phys. Chem. C 114, 487 (2010).

37. H. Petek, J. Zhao, Chem. Rev. 110, 7082 (2010).38. C. Costentin, Chem. Rev. 108, 2145 (2008).

Acknowledgments: Materials and methods and additionalfigures are presented in the supplementary materials onScience Online. We are grateful for financial support fromthe University of Washington, the American Chemical SocietyPetroleum Research Fund (51178-ND3), and the U.S. NSF(CHE-1151726). This research is also supported in part bythe Department of Energy Office of Science GraduateFellowship Program, made possible in part by the AmericanRecovery and Reinvestment Act of 2009, administered by

Oak Ridge Institute for Science and Education–Oak RidgeAssociated Universities under contract no. DE-AC05-06OR23100,and in part by the NSF Center for Enabling New Technologiesthrough Catalysis. We especially thank D. Gamelin, membersof the Gamelin laboratory, and L. Maibaum for very valuablediscussions.

Supplementary Materialswww.sciencemag.org/cgi/content/full/336/6086/1298/DC1Materials and MethodsSupplementary TextFigs. S1 to S14Table S1References (39–54)

7 February 2012; accepted 10 April 201210.1126/science.1220234

Interglacial Hydroclimate in theTropical West Pacific Through theLate PleistoceneA. N. Meckler,1,2* M. O. Clarkson,3 K. M. Cobb,4 H. Sodemann,5 J. F. Adkins1

Records of atmospheric carbon dioxide concentration (PCO2) and Antarctic temperature haverevealed an intriguing change in the magnitude of interglacial warmth and PCO2 at around430,000 years ago (430 ka), but the global climate repercussions of this change remainelusive. Here, we present a stalagmite-based reconstruction of tropical West Pacific hydroclimatefrom 570 to 210 ka. The results suggest similar regional precipitation amounts across the fourinterglacials contained in the record, implying that tropical hydroclimate was insensitive tointerglacial differences in PCO2 and high-latitude temperature. In contrast, during glacialterminations, drying in the tropical West Pacific accompanied cooling events in northern highlatitudes. Therefore, the tropical convective heat engine can either stabilize or amplify globalclimate change, depending on the nature of the climate forcing.

The study of past interglacial climates pro-vides valuable insight into natural climatevariability under conditions roughly sim-

ilar to those of the present day, but with differ-ences in atmospheric carbon dioxide concentration(PCO2) (1) and astronomic forcing. Especially in-triguing is a step-like increase in interglacial Ant-arctic temperature (2), global ice melting (3), andPCO2 (4, 5) that took place at roughly 430,000years ago (430 ka) and is referred to as theMid-Brunhes Event (MBE). To understand theunderlying cause of this shift toward warmerinterglacials, it is crucial to establish the globalsignature of interglacial climates during that time,but sufficiently long records, especially from lowlatitudes and terrestrial archives, are sparse. Be-cause PCO2 and ice cover affect the planet’s ra-diative balance, a global change in interglacial

climate at the MBEmight be expected. Althoughmany marine sea surface temperature (SST) re-constructions from around the globe show qual-itatively similar changes to those observed inAntarctica (6), the few available lower-latitudeSST reconstructions (7–9) from this time perioddo not show the same consistent increase in post-MBE interglacial temperatures. Given that manymodels simulate strong dynamical links betweenhigh-latitude and tropical climate [e.g., (10)], moretropical records are required to test and improveour understanding of low-latitude climate sensi-tivity to a range of different forcing factors.

Here, we present stalagmite records fromnorthern Borneo (4°N, 115°E; fig. S1) that spanseveral glacial-to-interglacial cycles and includethe MBE. We reconstruct Warm Pool hydrocli-mate from the oxygen isotope (d18O) variabilityrecorded in three stalagmites with overlappinggrowth periods. Borneo lies in the West PacificWarm Pool, which is an important heat and mois-ture source for higher latitudes (11). Althoughthere is little seasonality inmodern rainfall amountat our site, the isotopic composition of precipita-tion varies seasonally by up to 2 to 3‰ (12), withthe most enriched d18O values occurring in Feb-ruary and March, when the intertropical conver-

gence zone (ITCZ) is furthest south of our site.However, on interannual time scales, the El Niño–Southern Oscillation (ENSO) phenomenon dom-inates both precipitation amount and d18O, withEl Niño events resulting in large-scale drying andrainfall d18O that is roughly 5 to 6‰ higher thanaverage (12). Two recent studies of water iso-topes in the same area conclude that regional(much more than local) rainfall amount is themain driver for rainfall d18O variability on in-traseasonal and longer time scales today (13) andduring modeled millennial-scale events in thepast (14).

We analyzed stalagmites from three differentcaves in GunungMulu andGunung Buda Nation-al Parks, each about 5 to 10 km apart. Thestalagmites were dated by 238U-234U-230Th mea-surements (15) and cover the time period from570 (T20) ka to 210 (T10) ka [from marine iso-tope stage (MIS) 13 to the beginning of MIS7],with some gaps due to hiatuses. Sample GC08covers the whole time period, whereas samplesSqueeze1 and WR5 yield shorter records thatoverlap GC08. In comparison to other cave sys-tems, dating of these stalagmites is challengingdue to their unusually low 234U/238U ratios and/orlow U concentrations (fig. S4 and tables S1 toS3). The old age of the samples exacerbates thesedating issues. Furthermore, there is evidence thatslight open-system exchange of U has affectedsample GC08, requiring us to model the ages inthis stalagmite by assuming a constant initial234U/238U ratio (15). Varying detrital 230Th con-tamination cannot explain the age reversals inthis stalagmite, owing to the short half-life of230Th compared to the age of our samples (15).The error assigned to the assumed constant initial234U/238U ratio leads to increased age uncertaintyin the GC08 chronology, with individual age er-rors ranging from T9200 to T13,500 years (2s;table S1), limiting our ability to compare themillennial-scale timing of climate change in Bor-neo with other records. The age constraints aresufficient, however, to identify each interglacialstage, which is the main focus of this study. Insample WR5, open-system alteration might alsohave occurred but cannot be corrected for (15).The d18O record from this stalagmite therefore

1Geological and Planetary Science Department, California In-stitute of Technology, Pasadena, CA, USA. 2Geological Institute,ETH Zurich, Switzerland. 3School of Geosciences, University ofEdinburgh, Edinburgh EH9 3JW, UK. 4School of Earth andAtmospheric Sciences, Georgia Institute of Technology, Atlanta,GA 30332, USA. 5Institute for Atmosphere and Climate, ETHZurich, Switzerland.

*To whom correspondence should be addressed. E-mail:[email protected]

www.sciencemag.org SCIENCE VOL 336 8 JUNE 2012 1301

REPORTS

on

Augu

st 2

, 201

2w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

24. J. J. Warren, T. A. Tronic, J. M. Mayer, Chem. Rev. 110,6961 (2010).

25. M. K. Mahanthappa, K.-W. Huang, A. P. Cole,R. M. Waymouth, Chem. Commun. (Camb.) 2002, 502(2002).

26. M. K. Mahanthappa, A. P. Cole, R. M. Waymouth,Organometallics 23, 836 (2004).

27. K. Schröder et al., Organometallics 27, 1859 (2008).28. J. M. Mayer, Acc. Chem. Res. 44, 36 (2011).29. J. Jin, M. Newcomb, J. Org. Chem. 73, 7901 (2008).30. S. Morrison, Electrochemistry at Semiconductor and

Oxidized Metal Electrodes (Plenum, New York, 1980).31. B. Enright, G. Redmond, D. Fitzmaurice, J. Phys. Chem.

98, 6195 (1994).32. F. Marken, A. S. Bhambra, D.-H. Kim, R. J. Mortimer,

S. J. Stott, Electrochem. Commun. 6, 1153 (2004).33. M. Zukalova, M. Kalbáč, L. Kavan, I. Exnar, M. Graetzel,

Chem. Mater. 17, 1248 (2005).

34. A. Minguzzi, F.-R. F. Fan, A. Vertova, S. Rondinini,A. J. Bard, Chem. Sci. 3, 217 (2012).

35. Y. Ren, L. J. Hardwick, P. G. Bruce, Angew. Chem. Int. Ed.49, 2570 (2010).

36. C. Venkataraman, A. V. Soudackov, S. Hammes-Schiffer,J. Phys. Chem. C 114, 487 (2010).

37. H. Petek, J. Zhao, Chem. Rev. 110, 7082 (2010).38. C. Costentin, Chem. Rev. 108, 2145 (2008).

Acknowledgments: Materials and methods and additionalfigures are presented in the supplementary materials onScience Online. We are grateful for financial support fromthe University of Washington, the American Chemical SocietyPetroleum Research Fund (51178-ND3), and the U.S. NSF(CHE-1151726). This research is also supported in part bythe Department of Energy Office of Science GraduateFellowship Program, made possible in part by the AmericanRecovery and Reinvestment Act of 2009, administered by

Oak Ridge Institute for Science and Education–Oak RidgeAssociated Universities under contract no. DE-AC05-06OR23100,and in part by the NSF Center for Enabling New Technologiesthrough Catalysis. We especially thank D. Gamelin, membersof the Gamelin laboratory, and L. Maibaum for very valuablediscussions.

Supplementary Materialswww.sciencemag.org/cgi/content/full/336/6086/1298/DC1Materials and MethodsSupplementary TextFigs. S1 to S14Table S1References (39–54)

7 February 2012; accepted 10 April 201210.1126/science.1220234

Interglacial Hydroclimate in theTropical West Pacific Through theLate PleistoceneA. N. Meckler,1,2* M. O. Clarkson,3 K. M. Cobb,4 H. Sodemann,5 J. F. Adkins1

Records of atmospheric carbon dioxide concentration (PCO2) and Antarctic temperature haverevealed an intriguing change in the magnitude of interglacial warmth and PCO2 at around430,000 years ago (430 ka), but the global climate repercussions of this change remainelusive. Here, we present a stalagmite-based reconstruction of tropical West Pacific hydroclimatefrom 570 to 210 ka. The results suggest similar regional precipitation amounts across the fourinterglacials contained in the record, implying that tropical hydroclimate was insensitive tointerglacial differences in PCO2 and high-latitude temperature. In contrast, during glacialterminations, drying in the tropical West Pacific accompanied cooling events in northern highlatitudes. Therefore, the tropical convective heat engine can either stabilize or amplify globalclimate change, depending on the nature of the climate forcing.

The study of past interglacial climates pro-vides valuable insight into natural climatevariability under conditions roughly sim-

ilar to those of the present day, but with differ-ences in atmospheric carbon dioxide concentration(PCO2) (1) and astronomic forcing. Especially in-triguing is a step-like increase in interglacial Ant-arctic temperature (2), global ice melting (3), andPCO2 (4, 5) that took place at roughly 430,000years ago (430 ka) and is referred to as theMid-Brunhes Event (MBE). To understand theunderlying cause of this shift toward warmerinterglacials, it is crucial to establish the globalsignature of interglacial climates during that time,but sufficiently long records, especially from lowlatitudes and terrestrial archives, are sparse. Be-cause PCO2 and ice cover affect the planet’s ra-diative balance, a global change in interglacial

climate at the MBEmight be expected. Althoughmany marine sea surface temperature (SST) re-constructions from around the globe show qual-itatively similar changes to those observed inAntarctica (6), the few available lower-latitudeSST reconstructions (7–9) from this time perioddo not show the same consistent increase in post-MBE interglacial temperatures. Given that manymodels simulate strong dynamical links betweenhigh-latitude and tropical climate [e.g., (10)], moretropical records are required to test and improveour understanding of low-latitude climate sensi-tivity to a range of different forcing factors.

Here, we present stalagmite records fromnorthern Borneo (4°N, 115°E; fig. S1) that spanseveral glacial-to-interglacial cycles and includethe MBE. We reconstruct Warm Pool hydrocli-mate from the oxygen isotope (d18O) variabilityrecorded in three stalagmites with overlappinggrowth periods. Borneo lies in the West PacificWarm Pool, which is an important heat and mois-ture source for higher latitudes (11). Althoughthere is little seasonality inmodern rainfall amountat our site, the isotopic composition of precipita-tion varies seasonally by up to 2 to 3‰ (12), withthe most enriched d18O values occurring in Feb-ruary and March, when the intertropical conver-

gence zone (ITCZ) is furthest south of our site.However, on interannual time scales, the El Niño–Southern Oscillation (ENSO) phenomenon dom-inates both precipitation amount and d18O, withEl Niño events resulting in large-scale drying andrainfall d18O that is roughly 5 to 6‰ higher thanaverage (12). Two recent studies of water iso-topes in the same area conclude that regional(much more than local) rainfall amount is themain driver for rainfall d18O variability on in-traseasonal and longer time scales today (13) andduring modeled millennial-scale events in thepast (14).

We analyzed stalagmites from three differentcaves in GunungMulu andGunung Buda Nation-al Parks, each about 5 to 10 km apart. Thestalagmites were dated by 238U-234U-230Th mea-surements (15) and cover the time period from570 (T20) ka to 210 (T10) ka [from marine iso-tope stage (MIS) 13 to the beginning of MIS7],with some gaps due to hiatuses. Sample GC08covers the whole time period, whereas samplesSqueeze1 and WR5 yield shorter records thatoverlap GC08. In comparison to other cave sys-tems, dating of these stalagmites is challengingdue to their unusually low 234U/238U ratios and/orlow U concentrations (fig. S4 and tables S1 toS3). The old age of the samples exacerbates thesedating issues. Furthermore, there is evidence thatslight open-system exchange of U has affectedsample GC08, requiring us to model the ages inthis stalagmite by assuming a constant initial234U/238U ratio (15). Varying detrital 230Th con-tamination cannot explain the age reversals inthis stalagmite, owing to the short half-life of230Th compared to the age of our samples (15).The error assigned to the assumed constant initial234U/238U ratio leads to increased age uncertaintyin the GC08 chronology, with individual age er-rors ranging from T9200 to T13,500 years (2s;table S1), limiting our ability to compare themillennial-scale timing of climate change in Bor-neo with other records. The age constraints aresufficient, however, to identify each interglacialstage, which is the main focus of this study. Insample WR5, open-system alteration might alsohave occurred but cannot be corrected for (15).The d18O record from this stalagmite therefore

1Geological and Planetary Science Department, California In-stitute of Technology, Pasadena, CA, USA. 2Geological Institute,ETH Zurich, Switzerland. 3School of Geosciences, University ofEdinburgh, Edinburgh EH9 3JW, UK. 4School of Earth andAtmospheric Sciences, Georgia Institute of Technology, Atlanta,GA 30332, USA. 5Institute for Atmosphere and Climate, ETHZurich, Switzerland.

*To whom correspondence should be addressed. E-mail:[email protected]

www.sciencemag.org SCIENCE VOL 336 8 JUNE 2012 1301

REPORTS

on

Augu

st 2

, 201

2ww

w.sc

ienc

emag

.org

Down

load

ed fr

om

LETTERS

Millennial-scale trends in west Pacific warm poolhydrology since the Last Glacial MaximumJudson W. Partin1, Kim M. Cobb1, Jess F. Adkins2, Brian Clark3 & Diego P. Fernandez2

Models and palaeoclimate data suggest that the tropical Pacificclimate system plays a key part in the mechanisms underlyingorbital-scale and abrupt climate change1–7. Atmospheric convec-tion over the western tropical Pacific is a major source of heat andmoisture to extratropical regions, and may therefore influencethe global climate response to a variety of forcing factors. Theresponse of tropical Pacific convection to changes in globalclimate boundary conditions, abrupt climate changes and radi-ative forcing remains uncertain, however. Here we present threeabsolutely dated oxygen isotope records from stalagmites innorthern Borneo that reflect changes in west Pacific warm poolhydrology over the past 27,000 years. Our results suggest thatconvection over the western tropical Pacific weakened 18,000–20,000 years ago, as tropical Pacific2,5,6,8 and Antarctic9 tempera-tures began to rise during the early stages of deglaciation.Convective activity, as inferred from oxygen isotopes, reached aminimum during Heinrich event 1 (ref. 10), when the Atlanticmeridional overturning circulation was weak11, pointing to feed-backs between the strength of the overturning circulation andtropical Pacific hydrology. There is no evidence of the YoungerDryas event12 in the stalagmite records, however, suggesting thatdifferent mechanisms operated during these two abrupt deglacialclimate events. During the Holocene epoch, convective activityappears to track changes in spring and autumn insolation, high-lighting the sensitivity of tropical Pacific convection to externalradiative forcing. Together, these findings demonstrate that thetropical Pacific hydrological cycle is sensitive to high-latitudeclimate processes in both hemispheres, as well as to external radi-ative forcing, and that it may have a central role in abrupt climatechange events.

Numerous palaeoclimatic studies have focused on changes in thetropical Pacific zonal sea surface temperature (SST) gradient andassociated shifts in convection2,4,5, extending an El Nino/SouthernOscillation (ENSO) framework to interpretations of millennial-scaleclimate variability. Recently, the potential for meridional changesin tropical Pacific temperatures has emerged as an importantmechanism in shaping hydrological responses to abrupt climatechange7,13–15. Indeed, new modelling and palaeoclimate results sug-gest that during Heinrich event 1 (H1), when a near-collapse of theAtlantic meridional overturning circulation11 cooled most of theNorthern Hemisphere, the Intertropical Convergence Zone (ITCZ)shifted southwards7,13–15. However, the extent to which tropicalPacific climate feedbacks were involved in deglacial abrupt climateevents such as H1, the Antarctic Cold Reversal9 and subsequentYounger Dryas12 remains unclear. A combination of zonal and/ormeridional changes in the distribution of tropical convection prob-ably played a key part in shaping the global climate response toabrupt climate changes during the deglaciation. Uncovering the

mechanisms and feedbacks that govern the response of the tropicalPacific climate system to both internal and external forcings requireswell-dated, high-resolution records from climatic centres of action.

Here we present three absolutely dated stalagmite oxygen isotopic(d18O) records from northern Borneo that document changes inwestern tropical Pacific atmospheric circulation and hydrology overthe past 27,000 yr. Tropical stalagmite d18O records are particularlywell-suited to the investigation of centennial-to-millennial hydro-logical change because tropical rainfall d18O is inversely correlatedto precipitation amount and because U–Th dating provides excellentchronological control.

The research site is located in Gunung Buda National Park (4uN,114u E) in the northwestern corner of Malaysian Borneo (seeMethods for detailed site description). The ITCZ lies above northernBorneo year-round, delivering 5m of rainfall with little seasonality(Supplementary Fig. 1). ENSO exerts a dominant control on north-ern Borneo precipitation, with anomalies as large as 650% duringENSO extremes (Fig. 1). Interannual (2–7 yr) changes in rainfallaccount for ,20% of total precipitation variance in northernBorneo and are highly correlated to the Southern Oscillation Index(R520.80). Seasonal cycles in rainwater d18O (210% duringboreal autumn and 24% during boreal spring) probably reflectnorthern versus southern moisture trajectories16. ENSO-relatedinterannual rainfall d18O variability is consistent with the ‘amounteffect’17,18, whereby periods of increased precipitation are character-ized by lighter rainfall d18O (ref. 16).

Several tests confirm that the Gunung Buda stalagmites formedunder oxygen isotopic equilibrium, allowing carbonate d18O changes

1School of Earth and Atmospheric Sciences, Georgia Institute of Technology, Atlanta, Georgia 30332, USA. 2Division of Geological and Planetary Sciences, California Institute ofTechnology, Pasadena, California 91125, USA. 3Gunung Mulu National Park, Sarawak, Malaysia.

16° N

8° N

0°

8° S

16° S

100° E

–30 –20 –10 0Precipitation anomaly (cm month–1)

10 20 30

120° E 140° E

Figure 1 | Map of the west Pacific December-January-Februaryprecipitation anomaly during the 1997–98 El Nino event. Data are fromref. 31. A white triangle marks the approximate location of the research site.

Vol 449 |27 September 2007 |doi:10.1038/nature06164

452Nature ©2007 Publishing Group

to be interpreted as rainwater d18O changes. First, the correlationbetween oxygen and carbon isotopic variability is low (R, 0.05)after linearly detrending for glacial–interglacial isotopic changes.Second, all three samples pass the ‘Hendy test’19, exhibiting d18Ovariations of less than ,0.5% across the axial portion of a singlegrowth layer (see Methods and Supplementary Figs 9 and 10).Third, the equilibrium calcite d18O value equals that measured formodern stalagmite calcite, given measured present-day rainwaterd18O of26.8% and 26 uC cave temperatures16. Last, the high degreeof millennial-scale reproducibility of d18O in the three stalagmitesfrom caves located,5 km apart strongly suggests that the stalagmited18O records regional climate changes associated with rainfall d18Ovariability (Fig. 2). Poor sub-millennial reproducibility between thethree records can be attributed to dating uncertainties and/or site-specific stalagmite d18O variability. It is important to note that tem-perature changes could only account for ,0.7% of stalagmite d18Ovariability over the past 27,000 yr, given that warmpool temperatureswere no more than 3.5 uC colder during the Last Glacial Maximum(LGM)2,5,6,8.

A total of 75 U–Th dates and 11 isochrons provide excellent chro-nological control for the three stalagmite d18O records (see Methodsand Supplementary Information). The age model for each recordconsists of 24–26 U–Th dates (Fig. 2) that were corrected for detritalthorium using stalagmite-specific detrital 230Th/232Th values calcu-lated using isochrons. Age errors of up to 2% (2s) represent a com-bination of analytical uncertainty and uncertainty in the detrital232Th correction. Slow growth rates and/or unresolved hiatuses rep-resent the largest sources of chronological uncertainty, so we limitour climatic interpretations to portions of the records with growthrates higher than 10 mmyr21.

LGM stalagmite d18O values (averaged over the period 19–23 kyrago) are 1.36 0.3% heavier than modern values, and reflect a com-bination of global ice volume (11%)20, LGM cooling of ,2–3.5 uCin the western tropical Pacific2,5,6,8 (10.4% to 10.7%), and poorlyconstrained changes in LGM regional seawater d18O (20.5% to10.5%)2,5,6,21. Thus, potentially small changes in northern Borneo

rainfall d18O during the LGM are difficult to resolve. Possible con-trols on northern Borneo rainfall d18O over our whole 27-kyr recordinclude: (1) changes in the tropical Pacific zonal SST gradient; (2)changes in the location and/or intensity of the ITCZ; and (3) changesin eustatic sea level, which determine the size of the emergent SundaShelf. Exposure of the Sunda Shelf undoubtedly altered atmosphericcirculation in the west Pacific by increasing continentality andlengthening moisture trajectories, both of which deplete rainfalld18O (refs 17, 18). A complete understanding of warm pool hydro-logical changes during the LGM and deglaciation must account forthe effect of the Sunda Shelf on the tropical Pacific coupled system.

Stalagmite d18O values begin a protracted trend towards morepositive values 18–20 kyr ago, as tropical Pacific SST2,5,6,8 andAntarctic9 temperatures began to rise during the early deglaciation.This stalagmite d18O excursion cannot be attributed to temperatureor local seawater d18O changes, as recorded in nearby marine sedi-ments5,6,8,21, and is therefore interpreted as a positive rainfall d18Oanomaly (dry conditions based on the amount effect). The inferredtrend towards drier conditions in northern Borneo culminates inmaximum stalagmite d18O values 16.36 0.3 kyr ago, coincident withthe timing of a d18Omaximum in aChinese stalagmite22 attributed toH1 (Fig. 3). Conservative age error bars for specific features of theBorneo stalagmite d18O records take into account uncertaintiesassociated with stalagmite d18O reproducibility, U-series datingerrors, and the potential for nonlinear growth rates. Dry conditionsin northern Borneo during H1 are consistent with model results15

indicating a southward shift of the ITCZ in conjunction with a wea-kened Atlantic meridional overturning circulation inferred fromproxy data11. Together with evidence for dry conditions in southeast

–10

δ18 O

(‰)–9

–8

–7

–10

δ18 O

(‰) –9

–8

–7

0 5 10 15Age (kyr ago)

20 25 30

–10

δ18 O

(‰) –9

–8

–7

Figure 2 | Three absolutely dated stalagmite d18O records from northernBorneo. U–Th dates for each record are plotted in corresponding coloursalong the top and bottom; error bars represent 2s analytical uncertainty plusuncertainty in the detrital thorium correction. The average d18O temporalresolution is 72, 56 and 60 yr per sample for SCH02, SSC01 and BA04,respectively, but varies from 1 to 100 yr per sample depending on growthrate. Data depicted in grey represent slow-growing (,10mmyr21) portionsof the records, and are considered untrustworthy for climatic interpretationowing to poor chronological control (up to 61 kyr). Sample BA04 grew inBukit Assam cave, while SSC01 and SCH02 grew,20m apart in Snail Shellcave, roughly 6 km from Bukit Assam (see Methods for more detailed caveand sample descriptions).

a

b

c

d

e

f

–34

–9

–44

–46

–48

–50

–40–20020

–52

Inso

latio

ndi

ffere

nce

(W m

–2)

–8

–7

–6

–5

δ18 O

(‰)

δ18 O

(‰)

δ18 O

(‰)

δ18 O

(‰)

SS

T (°

C)

–36

–38

–40

–42

–44

–46

–10

–9

–8

–7

30

0 5 10 15Age (kyr ago)

20 25 30

29282726

Figure 3 | Comparison of the Borneo stalagmite d18O records with otherpalaeoclimate records. Records plotted are as follows. a, Greenland(NGRIP) ice core d18O (ref. 12). b, Hulu/Dongge caves stalagmite d18Orecords (ref. 22 and R. L. Edwards, personal communication). c, Borneostalagmite d18O records (SCH02, blue; SSC01, red; BA04, green). Slow-growing (,10 mmyr21) portions of the records are excluded. Note thatBA04 d18O values have been shifted by 10.4%. d, EPICA Dronning MaudLand ice core d18O (ref. 9). e, Sediment core reconstruction of SST from theSulu Sea6. f, March minus September insolation at the Equator.

NATURE |Vol 449 |27 September 2007 LETTERS

453Nature ©2007 Publishing Group

h"p://www.esrl.noaa.gov/psd/map/clim/olr.shtml

Borneo daily rainfall δ18O

large day-to-day range (20‰) dominated by intraseasonal and interannual variability

Moerman et al., 2013

Borneo daily rainfall δ18O

little to no seasonality in rainfall δ18O

Moerman et al., 2013

observational confirmation of numerical modeling results invok-ing post-condensation evaporative enrichment as a primary con-trol on rainfall isotopes during individual precipitation events (Leeand Fung, 2008; Risi et al., 2008a).

At the diurnal timescale, several lines of evidence suggest thatisotopic fractionation associated with the “local” amount effect isdwarfed by δ18O variations that reflect regional convective activity.For one, our results indicate that post-condensation evaporation offalling raindrops, which amounts to fractionations of ∼1–2‰within the March 7, 2010 convective event, is relatively smallcompared to day-to-day rainfall δ18O variations of ∼2–10‰.Indeed, previous observational and GCM-based studies also findthat local post-condensation evaporation has relatively littleimpact on rainfall δ18O values on daily and longer timescales inwet tropical regions (Kurita et al., 2009; Breitenbach et al., 2010;Field et al., 2010). Secondly, the weak correlation between dailyrainfall δ18O and daily Mulu precipitation amount further suggeststhat local rainout has relatively little impact on cumulative dailyMulu rainfall δ18O values. Similarly weak correlations betweendaily rainfall δ18O and local precipitation amount are also observedat other tropical and subtropical sites (e.g. Yamanaka et al., 2004;Risi et al., 2008b, Kurita et al., 2009, Breitenbach et al., 2010;Vimeux et al., 2011).

The weak daily rainfall δ18O–precipitation relationship can beattributed to the time- and space-integrative properties of rainfallδ18O, illustrated by the high correlation between daily rainfall δ18Oand regional precipitation amount averaged over the precedingweek (Fig. 4D). The 5–8-day vapor residence time that we observeat Mulu suggests that isotopic signals persist within the atmo-spheric vapor above Mulu for several days. These findings areconsistent with those of studies at other tropical sites, whichexplain the time-integrative behavior of rainfall δ18O via arepeated atmospheric vapor recycling process, in which low-level vapor depleted by earlier convective activity is fed intosuccessive convective systems (Risi et al., 2008a, 2008b; Vimeuxet al., 2011). In this way, the isotopic composition of the recycledwater vapor – and that of the resultant rainfall – reflects the

cumulative intensity of the previous days' convective activity, thuscausing daily rainfall δ18O to be poorly correlated with dailyprecipitation amount. Taken together, these results suggest thatdiurnal rainfall δ18O variability at our site is more dependent onthe isotopic composition of the water vapor from which itcondenses than the local fractionation processes that occur duringindividual rainfall events. The space-integrative property of dailyMulu rainfall δ18O, illustrated by the strong correlations betweenMulu rainfall δ18O and regional precipitation amount in Fig. 4Cand D, suggests that Mulu vapor δ18O is also significantly influ-enced by regional-scale convective processes. This is furthersupported by Fig. 5, which shows that correlations between dailygridded TRMM precipitation amount and daily Mulu rainfall δ18Oare far more regionally extensive than they are for daily Muluprecipitation amount.

The large multi-day rainfall δ18O depletion events observed atMulu typically coincide with the passage of mesoscale convectivesystems associated with the MJO (Fig. 7). We infer that these large-scale convective systems deliver anomalously depleted watervapor δ18O to Mulu, thus resulting in depleted rainfall δ18O.Indeed, Kurita et al. (2011) and Berkelhammer et al. (2012) observe

60E 80E 100E120E140E160E 180

Sep2008

Oct2008

Nov2008

Dec2008

Jan2009

Feb2009

Mar2009

Apr2009

May2009

OLR

(W/m

2 )

180

190

200

210

220

230

240

250

260

270

280

Fig. 7. Time-longitude map of daily NOAA interpolated OLR filtered with a 30–96day bandpass filter and averaged over 0–51N for the time window spanning 8/26/2008–5/23/2009. Open circles mark the timing of intraseasonal rainfall δ18Odepletion events that occur during the time window. White dashed arrowshighlight examples of eastward-propagating enhanced convection.

-12

-10

-8

-6

-4

δ18O

(%o)

Mulu precipitation

25

20

15

10

5

0

Prec

ip (m

m/d

ay)

Prec

ip (m

m/d

ay)

TRMM

OLR

Jan

Feb Mar AprMay Ju

n Jul

Aug Sep OctNov Dec

Jan

Feb Mar Apr May Jun

Jul

Aug Sep Oct Nov Dec

25

20

15

10

5

0

180

200

220

240

OLR

(W/m

2 )

Mulu δ18O

Fig. 8. Long-term monthly mean Mulu rainfall δ18O (red), local Mulu precipitationamount (blue), satellite-measured TRMM 3B43 precipitation amount (black), andNOAA interpolated OLR (green) from July 2006 to May 2011. TRMM and OLR arespatial averages integrated over a 2.51!2.51 gridbox centered about 51N, 1151E.Also plotted are versions of each long-term monthly mean with ENSO-relatedvariability removed (gray). Error bars represent the 1s spread of each month'smean value. Note that axes for δ18O and OLR are inverted. (For interpretation of thereferences to color in this figure legend, the reader is referred to the web version ofthis article.)

J.W. Moerman et al. / Earth and Planetary Science Letters ∎ (∎∎∎∎) ∎∎∎–∎∎∎8

Please cite this article as: Moerman, J.W., et al., Diurnal to interannual rainfall δ18O variations in northern Borneo driven byregional hydrology. Earth and Planetary Science Letters (2013), http://dx.doi.org/10.1016/j.epsl.2013.03.014i

δ18O to integrate regional convective activity across space and timemakes it a much better indicator of large-scale hydrology thanlocal precipitation amount, which contains much more noise.Therefore, while precipitation variability from nearby sites maybe poorly correlated (Dayem et al., 2010), rainfall δ18O variabilityfrom nearby sites may indeed be highly correlated due to theintegrative properties of rainfall δ18O. This explains why stalagmiteδ18O records from caves that are hundreds of kilometers apartreflect the same regional-scale hydroclimate influences (e.g. Yuanet al., 2004; Wang et al., 2001). Likewise, well-reproduced cavestalagmite records from a single site can be interpreted as robustindicators of regional-scale hydroclimate variability (e.g. Partinet al., 2007).

When considering how to relate rainfall δ18O variability tostalagmite calcite δ18O variability, one must keep in mind thatseveral processes mediate the cloud-to-calcite transformation ofrainfall δ18O signals. First, the infiltration of rainwaters through thekarst environment inevitably leads to some measure of signal

attenuation from mixing, as well as signal delay depending onwater residence times in the karst. In comparing several Muludripwater δ18O timeseries with rainfall δ18O timeseries, Cobb et al.(2007) argue for dripwater residence times of 2–3 months, basedon the preservation of a weak seasonal cycle in dripwater δ18O.However, owing to logistical difficulties in collecting timeseries ofdripwater δ18O from slow to ultra-slow drips that typically formstalagmites, residence time estimates for the most relevant dripsremain poorly constrained. A new 5-yr-long timeseries of drip-water δ18O from both fast and slow Mulu drips will help quantifygroundwater transit times across a broad range of drip environ-ments (Moerman et al., 2012). A complementary approachinvolves pursuing a “calibration” of annual to sub-annual stalag-mite δ18O with climatic timeseries over the 20th century. Thelatter approach requires unusually fast-growing stalagmites andsmall chronological errors afforded by either annual layer countingand/or many U/Th dates (e.g. Frappier et al., 2002, 2007; Fleitmannet al., 2004; Treble et al., 2005) and, as such, are relatively rare.

60E 120E 180 120W 60W

40N

20

0

20S

40S

Correlation (R)

Correlation between OLR and Mulu 18O

0 0.2 0.4 0.6 -0.6 -0.4 -0.2

60E 120E 180 120W 60W

40N

20N

0

20S

40S

Correlation (R)

Correlation between OLR and Mulu precip

0 0.2 0.4 0.6 -0.6 -0.4 -0.2

60E 120E 180 120W 60W

40N

20N

0

20S

40S

Correlation (R)

Correlation between TRMM and Mulu 18O

0 0.2 0.4 0.6 -0.6 -0.4 -0.2

60E 120E 180 120W 60W

40N

20N

0

20S

40S

Correlation (R)

Correlation between TRMM and Mulu precip

0 0.2 0.4 0.6 -0.6 -0.4 -0.2

A B

C D

x

x x

x

Fig. 10. Correlation maps of monthly mean Mulu rainfall δ18O and local Mulu precipitation amount with monthly TRMM precipitation product 3B43 and monthly NOAAinterpolated OLR for July 2006–May 2011 constructed using ordinary least squares regression. (A) Correlation between monthly mean Mulu rainfall δ18O and gridded TRMMprecipitation. (B) Same as (A) but with gridded OLR. (C) Correlation between monthly mean Mulu precipitation amount and gridded TRMM precipitation. D) Same as (C) butwith gridded OLR. Black contour lines indicate 95% significance regions as determined by the student's t-test using effective degrees of freedom (Bretherton et al., 1999). Thewhite ‘x’ in each panel marks the location of our study site at Gunung Mulu National Park (4°N, 114°E). Color scales for panels (B) and (C) are inverted. (For interpretation ofthe references to color in this figure legend, the reader is referred to the web version of this article.)

J.W. Moerman et al. / Earth and Planetary Science Letters ∎ (∎∎∎∎) ∎∎∎–∎∎∎10

Please cite this article as: Moerman, J.W., et al., Diurnal to interannual rainfall δ18O variations in northern Borneo driven byregional hydrology. Earth and Planetary Science Letters (2013), http://dx.doi.org/10.1016/j.epsl.2013.03.014i

Moerman et al,. 2013

Figure 1. Comparison of Mulu dripwater δ18O to rainfall δ18O and local and regional climate variables. (a) Monthly NINO4SST anomalies [Reynolds et al., 2002]. Red and blue bars along the upper x axis represent weak-to-moderate El Niño andmoderate-to-strong La Niña events as defined by the Oceanic Niño Index (http://www.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ensoyears.shtml). (b) Nonamount-weighted daily Mulu rainfall δ18O (gray circles) shownwith a 2 month running average (black line), and nonamount-weighted (dotted line) and amount-weighted (dashed line)means of the entire time series. (c) Mulu dripwater δ18O from drips WF (orange), WS (blue), and L2 (green) plotted withthe composite mean of the three dripwater δ18O time series (black line) and mean amount-weighted rainfall δ18O fromFigure 1b (dashed line). Box-and-whisker plots represent the median (bold bar), 25–75% quartile range (red box), andmaximum and minimum δ18O values (whiskers) for system-wide field expedition surveys of stalagmite-forming dripwatersconducted in August 2008 (n = 36), February–March 2010 (n = 54), and October–November 2012 (n = 135). (d) Drip ratesfrom drip sites WF (orange), WS (blue), and L2 (green). Note that WS drip rate has been shifted by +30 dpm and L2 by+20 dpm. (e) Rain gauge measurements of daily Mulu precipitation amount (gray bars) plotted with a 2 month runningaverage (black line). Y axes in Figures 1a–1c are inverted.

Geophysical Research Letters 10.1002/2014GL061696

MOERMAN ET AL. ©2014. American Geophysical Union. All Rights Reserved. 4

Figure 1. Comparison of Mulu dripwater δ18O to rainfall δ18O and local and regional climate variables. (a) Monthly NINO4SST anomalies [Reynolds et al., 2002]. Red and blue bars along the upper x axis represent weak-to-moderate El Niño andmoderate-to-strong La Niña events as defined by the Oceanic Niño Index (http://www.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ensoyears.shtml). (b) Nonamount-weighted daily Mulu rainfall δ18O (gray circles) shownwith a 2 month running average (black line), and nonamount-weighted (dotted line) and amount-weighted (dashed line)means of the entire time series. (c) Mulu dripwater δ18O from drips WF (orange), WS (blue), and L2 (green) plotted withthe composite mean of the three dripwater δ18O time series (black line) and mean amount-weighted rainfall δ18O fromFigure 1b (dashed line). Box-and-whisker plots represent the median (bold bar), 25–75% quartile range (red box), andmaximum and minimum δ18O values (whiskers) for system-wide field expedition surveys of stalagmite-forming dripwatersconducted in August 2008 (n = 36), February–March 2010 (n = 54), and October–November 2012 (n = 135). (d) Drip ratesfrom drip sites WF (orange), WS (blue), and L2 (green). Note that WS drip rate has been shifted by +30 dpm and L2 by+20 dpm. (e) Rain gauge measurements of daily Mulu precipitation amount (gray bars) plotted with a 2 month runningaverage (black line). Y axes in Figures 1a–1c are inverted.



daily rain δ18O


bi-weekly dripwater δ18O

While the autogenic recharge modelreproduces the timing of L2 dripwaterδ18Ominima andmaxima, it overestimatesthe amplitude of the drip’s δ18O variations(Figure 2). Amount-weighted rainfall δ18Oaveraged over the previous 42weeks(~10months) best reflects the timing ofdripwater δ18O maxima and minimaobserved in L2 (R=0.84), but the predictedvariations are roughly 1‰ higher thanobserved (Figure 2). This suggests thatthe flow pathway to this drip site is morecomplicated than that feeding WF andWS. L2’s amplitude attenuation suggestsa likely contribution from a second, well-mixed reservoir, which we model using abivariate mixing model,

XM ¼ XA 1" f Bð Þ þ XBf B (1)

where XA is the isotopic composition ofReservoir A,XB is the isotopic compositionof Reservoir B, and fB is the mixingparameter. Modeled dripwater δ18Osimulated by the autogenic rechargemodel with an ~10 month residencetime represents Reservoir A. In theinterest of simplicity, we set the isotopiccomposition of Reservoir B as amount-weighted Mulu rainfall δ18O averagedacross the study period ("8.6‰). AtfB = 0.6, the residuals between themodeled and observed L2 time seriesare minimized. With this modelconfiguration, the muted variability in L2is reproduced particularly well during

the 2009–2012 ENSO cycle (Figure 2), suggesting that two distinct reservoirs likely feed L2—one with aresidence time of ~10months that is recharged autogenically (Reservoir A) and a second with a significantlylonger residence time (Reservoir B). However, prior to 2009, the above parameters overdamp the variability,resulting in a poor fit between the modeled and observed time series. Increasing the Reservoir A flux from40% to 80% (i.e., fB = 0.2) during late 2008/early 2009 greatly improves model-data fit, suggesting anabrupt switch in flow to L2 during this period (Figure 2). Indeed, Reservoir A may represent 100% of flowin December 2008. Throughout 2008, modeled dripwater δ18O is ~0.5‰ more negative than observed L2values, indicating that either (i) the isotopic composition of Reservoir B varies over the study period and/or(ii) that our simple two-reservoir model fails to capture the complexity of flow feeding L2. Longer time seriesof rainwater and dripwater δ18O are required to differentiate between these scenarios.

The isotopic spread of hundreds of stalagmite-forming drips collected during field expeditions falls withinthe variability of the dripwater δ18O time series (Figure 1c), indicating that the three time series drips arebroadly representative of system-wide shifts in isotopic composition. The relatively small spread in both the2008 and 2012 expedition data sets (~0.3‰ (1σ)) suggests that residence times for stalagmite-forming dripsin Mulu are long enough to homogenize large intraseasonal shifts in rainfall δ18O of ~10‰ [Moerman et al.,2013], establishing a lower bound for residence times of stalagmite-forming drips of approximately 1–2months. Likewise, residence times for many stalagmite-forming drips appear to be less than 2 years, as only25% of the 2010 expedition drip distribution overlaps that of the 2008 expedition (Figure 1c). This remainstrue even when a subset of the slowest dripping samples (<2 dpm) is considered.

Figure 2. Observed dripwater δ18O (circles) for dripsWF (orange),WS (blue),and L2 (green) compared with best fit modeled dripwater δ18O (solid lines)using amount-weighted Mulu rainfall δ18O as input. Bold black linesrepresent dripwater simulated by the autogenic recharge model with aresidence time of ~3months for WF, ~4months for WS, and ~10monthsfor L2. Bivariate mixing model results for drip L2 modeled as 40% ofReservoir A with an ~10 month residence time and 60% of Reservoir Breflecting mean amount-weighted rainfall δ18O are plotted as a gray line.During the October 2008 to June 2009 interval, the red line reflectsmodeled dripwater δ18O assuming contributions of 20% from Reservoir Aand 80% from Reservoir B. Y axes are inverted in all panels.



drips reflect running average of rainfall �18O over previous 3-4 months


While the autogenic recharge modelreproduces the timing of L2 dripwaterδ18Ominima andmaxima, it overestimatesthe amplitude of the drip’s δ18O variations(Figure 2). Amount-weighted rainfall δ18Oaveraged over the previous 42weeks(~10months) best reflects the timing ofdripwater δ18O maxima and minimaobserved in L2 (R=0.84), but the predictedvariations are roughly 1‰ higher thanobserved (Figure 2). This suggests thatthe flow pathway to this drip site is morecomplicated than that feeding WF andWS. L2’s amplitude attenuation suggestsa likely contribution from a second, well-mixed reservoir, which we model using abivariate mixing model,

XM ¼ XA 1" f Bð Þ þ XBf B (1)

where XA is the isotopic composition ofReservoir A,XB is the isotopic compositionof Reservoir B, and fB is the mixingparameter. Modeled dripwater δ18Osimulated by the autogenic rechargemodel with an ~10 month residencetime represents Reservoir A. In theinterest of simplicity, we set the isotopiccomposition of Reservoir B as amount-weighted Mulu rainfall δ18O averagedacross the study period ("8.6‰). AtfB = 0.6, the residuals between themodeled and observed L2 time seriesare minimized. With this modelconfiguration, the muted variability in L2is reproduced particularly well during

the 2009–2012 ENSO cycle (Figure 2), suggesting that two distinct reservoirs likely feed L2—one with aresidence time of ~10months that is recharged autogenically (Reservoir A) and a second with a significantlylonger residence time (Reservoir B). However, prior to 2009, the above parameters overdamp the variability,resulting in a poor fit between the modeled and observed time series. Increasing the Reservoir A flux from40% to 80% (i.e., fB = 0.2) during late 2008/early 2009 greatly improves model-data fit, suggesting anabrupt switch in flow to L2 during this period (Figure 2). Indeed, Reservoir A may represent 100% of flowin December 2008. Throughout 2008, modeled dripwater δ18O is ~0.5‰ more negative than observed L2values, indicating that either (i) the isotopic composition of Reservoir B varies over the study period and/or(ii) that our simple two-reservoir model fails to capture the complexity of flow feeding L2. Longer time seriesof rainwater and dripwater δ18O are required to differentiate between these scenarios.

The isotopic spread of hundreds of stalagmite-forming drips collected during field expeditions falls withinthe variability of the dripwater δ18O time series (Figure 1c), indicating that the three time series drips arebroadly representative of system-wide shifts in isotopic composition. The relatively small spread in both the2008 and 2012 expedition data sets (~0.3‰ (1σ)) suggests that residence times for stalagmite-forming dripsin Mulu are long enough to homogenize large intraseasonal shifts in rainfall δ18O of ~10‰ [Moerman et al.,2013], establishing a lower bound for residence times of stalagmite-forming drips of approximately 1–2months. Likewise, residence times for many stalagmite-forming drips appear to be less than 2 years, as only25% of the 2010 expedition drip distribution overlaps that of the 2008 expedition (Figure 1c). This remainstrue even when a subset of the slowest dripping samples (<2 dpm) is considered.

Figure 2. Observed dripwater δ18O (circles) for dripsWF (orange),WS (blue),and L2 (green) compared with best fit modeled dripwater δ18O (solid lines)using amount-weighted Mulu rainfall δ18O as input. Bold black linesrepresent dripwater simulated by the autogenic recharge model with aresidence time of ~3months for WF, ~4months for WS, and ~10monthsfor L2. Bivariate mixing model results for drip L2 modeled as 40% ofReservoir A with an ~10 month residence time and 60% of Reservoir Breflecting mean amount-weighted rainfall δ18O are plotted as a gray line.During the October 2008 to June 2009 interval, the red line reflectsmodeled dripwater δ18O assuming contributions of 20% from Reservoir Aand 80% from Reservoir B. Y axes are inverted in all panels.



for 1 of 3 drips, two reservoirs needed; infers changes in residence times

Carolin et al,. submitted

newest data extend reconstruction through Termination 2

more convective activity

Mulu δ18O stack a work in progress see Partin et al., 2007 (0-30kybp) Meckler et al., 2012 (200-550kybp) Carolin et al., 2013 (30-100kybp)


fa

ll in

sola

tion

0°

(W/m

2 )


strong precessional signal tied to boreal fall insolation


fa

ll in

sola

tion

0°

(W/m

2 )


strong precessional signal tied to boreal fall insolation hard to reconcile with lack of modern-day seasonality


fa

ll in

sola

tion

0°

(W/m

2 )


Do data reflect boreal fall insolation control on ENSO?


Age (yr BP)0 2000 4000 6000 8000 10000 12000

δ18

-9.5

-9.0

-8.5

-8.0

-7.5

-7.0

-6.5

-6.0

25°N

JJA

inso

latio

n (W

/m2 )

450

455

460

465

470

475

480

485

δ18

-10.5

-10.0

-9.5

-9.0

-8.5

-8.0

-7.5

-7.0

-6.5

Equ

ator

SO

N in

sola

tion

(W/m

2)

405

410

415

420

425

430

435

440

445

Figure 3��D��6SHOHRWKHP�į18O records from Borneo. The growth axis of BA03 was sampled HYHU\��PP�IRU�ORZ�UHVROXWLRQ�į18O analysis (orange line). Also shown are U-Th dates for BA03 (orange triangles) used to construct the age model (Hoffmann et al., in prep). Thick black OLQHV�RQ�WKH�RUDQJH�FXUYH�PDUN�WKH�ZLQGRZV�RI�KLJK�UHVROXWLRQ�į18O analysis shown in Figure 4. į18O records of stalagmite SCH02 (red), BA04 (blue) and SSC01 (green) from the same region are also plotted with corresponding U-Th dates (Partin et al., 2007). Also plotted is the September-October-November (SON) insolation on the equator (black, Berger & Loutre, 1991). �E��6SHOHRWKHP�į18O record from Dongge Cave, China (grey, Dykoski et al., 2005), plotted with corresponding U-Th ages and June-July-August (JJA) insolation at 25° N (black, Berger & Loutre, 1991).

(a)

(b)Age (yr BP)

0 2000 4000 6000 8000 10000 12000

δ18

-9.5

-9.0

-8.5

-8.0

-7.5

-7.0

-6.5

-6.0

25°N

JJA

inso

latio

n (W

/m2 )

450

455

460

465

470

475

480

485

δ18

-10.5

-10.0

-9.5

-9.0

-8.5

-8.0

-7.5

-7.0

-6.5

Equ

ator

SO

N in

sola

tion

(W/m

2)

405

410

415

420

425

430

435

440

445

Figure 3��D��6SHOHRWKHP�į18O records from Borneo. The growth axis of BA03 was sampled HYHU\��PP�IRU�ORZ�UHVROXWLRQ�į18O analysis (orange line). Also shown are U-Th dates for BA03 (orange triangles) used to construct the age model (Hoffmann et al., in prep). Thick black OLQHV�RQ�WKH�RUDQJH�FXUYH�PDUN�WKH�ZLQGRZV�RI�KLJK�UHVROXWLRQ�į18O analysis shown in Figure 4. į18O records of stalagmite SCH02 (red), BA04 (blue) and SSC01 (green) from the same region are also plotted with corresponding U-Th dates (Partin et al., 2007). Also plotted is the September-October-November (SON) insolation on the equator (black, Berger & Loutre, 1991). �E��6SHOHRWKHP�į18O record from Dongge Cave, China (grey, Dykoski et al., 2005), plotted with corresponding U-Th ages and June-July-August (JJA) insolation at 25° N (black, Berger & Loutre, 1991).

(a)

(b)

Stalagmite BA03 70-150μm/yr 30-60 μm sampling = 2-5 samples/yr

Age (yr BP)2390 2410 2430 2450 2470 2490 2510 2530 2550 2570 2590

δ18-10.0

-9.8-9.6-9.4-9.2-9.0-8.8-8.6-8.4

Age (yr BP)3240 3260 3280 3300 3320 3340 3360 3380

Age (yr BP)5170 5190 5210

Age (yr BP)5640 5660 5680 5700

Age (yr BP)6590 6610 6630 6650 6670 6690 6710 6730

Age (yr BP)8140 8160 8180 8200 8220 8240 8260 8280 8300

Age (yr BP)2390 2410 2430 2450 2470 2490 2510 2530 2550 2570 2590

2-7y

r filte

red δ18

-0.20-0.15-0.10-0.05

00.050.100.150.20

Age (yr BP)3240 3260 3280 3300 3320 3340 3360 3380

Age (yr BP)5170 5190 5210

Age (yr BP)5640 5660 5680 5700

Age (yr BP)6590 6610 6630 6650 6670 6690 6710 6730

Age (yr BP)8140 8160 8180 8200 8220 8240 8260 8280 8300

Figure 4 High-resolution BA03 δ18O time series and 2-7 year bandpass filtered series Top row Blue crosses mark the δ18O values of individual growth bands milled at sub-annual resolution (3032 total), plotted with a two-year running mean (red line) based on a boxcar averaging algorithm. The black lines represent 95% confidence intervals of the boxcar running mean. Bottom row Bandpass (2-7 yr) filtered δ18O series of the six time series interpolated to 0.5-yr resolution. The black dashed lines mark the ±0.1‰ range. Age (yr BP)

2390 2410 2430 2450 2470 2490 2510 2530 2550 2570 2590

δ18

-10.0-9.8-9.6-9.4-9.2-9.0-8.8-8.6-8.4

Age (yr BP)3240 3260 3280 3300 3320 3340 3360 3380

Age (yr BP)5170 5190 5210

Age (yr BP)5640 5660 5680 5700

Age (yr BP)6590 6610 6630 6650 6670 6690 6710 6730

Age (yr BP)8140 8160 8180 8200 8220 8240 8260 8280 8300

Age (yr BP)2390 2410 2430 2450 2470 2490 2510 2530 2550 2570 2590

2-7y

r filte

red δ18

-0.20-0.15-0.10-0.05

00.050.100.150.20

Age (yr BP)3240 3260 3280 3300 3320 3340 3360 3380

Age (yr BP)5170 5190 5210

Age (yr BP)5640 5660 5680 5700

Age (yr BP)6590 6610 6630 6650 6670 6690 6710 6730

Age (yr BP)8140 8160 8180 8200 8220 8240 8260 8280 8300

Figure 4 High-resolution BA03 δ18O time series and 2-7 year bandpass filtered series Top row Blue crosses mark the δ18O values of individual growth bands milled at sub-annual resolution (3032 total), plotted with a two-year running mean (red line) based on a boxcar averaging algorithm. The black lines represent 95% confidence intervals of the boxcar running mean. Bottom row Bandpass (2-7 yr) filtered δ18O series of the six time series interpolated to 0.5-yr resolution. The black dashed lines mark the ±0.1‰ range.

Age (yr BP)2390 2410 2430 2450 2470 2490 2510 2530 2550 2570 2590

δ18

-10.0-9.8-9.6-9.4-9.2-9.0-8.8-8.6-8.4

Age (yr BP)3240 3260 3280 3300 3320 3340 3360 3380

Age (yr BP)5170 5190 5210

Age (yr BP)5640 5660 5680 5700

Age (yr BP)6590 6610 6630 6650 6670 6690 6710 6730

Age (yr BP)8140 8160 8180 8200 8220 8240 8260 8280 8300

Age (yr BP)2390 2410 2430 2450 2470 2490 2510 2530 2550 2570 2590

2-7y

r filte

red δ18

-0.20-0.15-0.10-0.05

00.050.100.150.20

Age (yr BP)3240 3260 3280 3300 3320 3340 3360 3380

Age (yr BP)5170 5190 5210

Age (yr BP)5640 5660 5680 5700

Age (yr BP)6590 6610 6630 6650 6670 6690 6710 6730

Age (yr BP)8140 8160 8180 8200 8220 8240 8260 8280 8300

Figure 4 High-resolution BA03 δ18O time series and 2-7 year bandpass filtered series Top row Blue crosses mark the δ18O values of individual growth bands milled at sub-annual resolution (3032 total), plotted with a two-year running mean (red line) based on a boxcar averaging algorithm. The black lines represent 95% confidence intervals of the boxcar running mean. Bottom row Bandpass (2-7 yr) filtered δ18O series of the six time series interpolated to 0.5-yr resolution. The black dashed lines mark the ±0.1‰ range.

2.5kybp 3.3kybp 5.2kybp

5.7kybp 6.7kybp

8.2kybp

20yrs

sub-annually resolved stalagmite δ18O records, 50-200yrs long

Chen et al,. submitted

Cha

nge

in c

oral

δ18

O s

tdev

of 2

-7yr

ban

d(%

)

-80

-60

-40

-20

0

20

Cha

nge

in P

eru

mol

lusk

∆T

varia

nce

from

mod

ern

(%)

-80

-60

-40

-20

0

20

Age (yr BP)0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000

Red

inte

nsity

std

ev o

f 2-7

yr b

and

0

5

10

15

20

25

30

35

El J

unco

San

d Pe

rcen

tage

(%)

0

5

10

15

20

25

30

35

BA03

δ18

0.02

0.03

0.04

0.05

0.06

0.07

Tota

l Var

ianc

e of

For

amin

ifera

l δ18

O

0.1

0.15

0.2

0.25

0.3

0.35 Figure 7 Comparison of Holocene ENSO variability estimates from BA03 with foraminiferal, coral, molluscan and lake sediment records. (a) Estimates of ENSO variability from stalagmite BA03 (squares) and foraminifera (circles, Koutavas & Joanides, 2012). %$�� į18O estimates are based on the standard deviation of the 2-7 yr band in overlapping 30-yr windows (5-yr step). )RUDPLQLIHUDO� į18O variance is FDOFXODWHG� XVLQJ� WKH� į18O of single tests in each 1-cm stratum (representing a deposition interval of approximately 80 years), and plotted with the original published age model (Koutavas & Joanides, 2012). �E�� &RUDO� į18O standard deviations are calculated in the same way as BA03 and plotted as percent differences from modern corals (1968-1998) for each site. Coral time series shorter than 30 years are plotted with filled circles (Cobb et al., 2013; McGregor et al., 2013). Also plotted is ENSO-driven seasonal SST range variance on Peruvian coast calculated from PRQWKO\� PROOXVFDQ� VKHOO� į18O measurements (pink, Carre et al., 2014). Variance is plotted as percent difference from modern (0-60 yr BP). The vertical bars PDUN� �ı� XQFHUWDLQW\� RI� WKH�estimates, and the horizontal bars mark the time range of shells for the variance estimate. (c) Lake records from Lake Pallcacocha in Ecuador (blue,

Moy et al., 2002) and El Junco in the Galapagas Islands (red, Conroy et al., 2008). Lake Pallcacocha record is based on reflective intensity of red light of the lake deposits. The variance of red intensity within overlapping 30-yr windows (5-yr step) is shown. Lake El Junco record is based on percentage of sand in the sediment grains. The low ENSO activity in early Holocene (6-10 kyr BP) in the lake records is distinct from the interannual records in the (a) and (b).

(a)

(b)

(c)

BA03

V21-30

PNG

Northern LineIslands

Peru Molluscs

Lake Pallcacocha

El Junco

Cha

nge

in c

oral

δ18

O s

tdev

of 2

-7yr

ban

d(%

)-80

-60

-40

-20

0

20

Cha

nge

in P

eru

mol

lusk

∆T

varia

nce

from

mod

ern

(%)

-80

-60

-40

-20

0

20

Age (yr BP)0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000

Red

inte

nsity

std

ev o

f 2-7

yr b

and

0

5

10

15

20

25

30

35

El J

unco

San

d Pe

rcen

tage

(%)

0

5

10

15

20

25

30

35

BA03

δ18

0.02

0.03

0.04

0.05

0.06

0.07

Tota

l Var

ianc

e of

For

amin

ifera

l δ18

O

0.1

0.15

0.2

0.25

0.3



(a)

(b)

(c)

BA03

V21-30

PNG


Peru Molluscs

Lake Pallcacocha

El Junco

pronounced minimum in ENSO variance ~5kybp similar to foram-based estimates . . . (Koutavas & Joanides, 2012) and mollusk-based ENSO reconstructions (Carre et al., 2013)


stronger ENSO

Cha

nge

in c

oral

δ18

O s

tdev

of 2

-7yr

ban

d(%

)

-80

-60

-40

-20

0

20

Cha

nge

in P

eru

mol

lusk

∆T

varia

nce

from

mod

ern

(%)

-80

-60

-40

-20

0

20

Age (yr BP)0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000

Red

inte

nsity

std

ev o

f 2-7

yr b

and

0

5

10

15

20

25

30

35

El J

unco

San

d Pe

rcen

tage

(%)

0

5

10

15

20

25

30

35

BA03

δ18

0.02

0.03

0.04

0.05

0.06

0.07

Tota

l Var

ianc

e of

For

amin

ifera

l δ18

O

0.1

0.15

0.2

0.25

0.3



(a)

(b)

(c)

BA03

V21-30

PNG


Peru Molluscs

Lake Pallcacocha

El Junco

Cha

nge

in c

oral

δ18

O s

tdev

of 2

-7yr

ban

d(%

)-80

-60

-40

-20

0

20

Cha

nge

in P

eru

mol

lusk

∆T

varia

nce

from

mod

ern

(%)

-80

-60

-40

-20

0

20

Age (yr BP)0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000

Red

inte

nsity

std

ev o

f 2-7

yr b

and

0

5

10

15

20

25

30

35

El J

unco

San

d Pe

rcen

tage

(%)

0

5

10

15

20

25

30

35

BA03

δ18

0.02

0.03

0.04

0.05

0.06

0.07

Tota

l Var

ianc

e of

For

amin

ifera

l δ18

O

0.1

0.15

0.2

0.25

0.3



(a)

(b)

(c)

BA03

V21-30

PNG


Peru Molluscs

Lake Pallcacocha

El Junco

01000200030004000500060007000 −80

−60

−40

−20

0

20

Chan

ge in

std

ev o

f ENS

O (%

)

Year (BP)

Reduced ENSO

1983-2003 ENSO Variance

20th century ENSO range>82 yrs

60-82 yrs

40-60 yrs

20-40 yrs<20 yrs

Fanning

Christmas

Palmyra


time

. . . and fossil coral ENSO reconstructions (Cobb et al., 2013; Grothe et al., in prep)

stronger ENSO

stronger ENSO

Several lines of evidence suggest that Holocene ENSO variance reached a minimum ~4-5kybp, suggesting that ENSO may be sensitive to boreal fall insolation forcing.

Several lines of evidence suggest that Holocene ENSO variance reached a minimum ~4-5kybp, suggesting that ENSO may be sensitive to boreal fall insolation forcing. Consistent with 550ky of boreal fall precessional sensitivity in low-resolution Borneo stalagmite δ18O.

Several lines of evidence suggest that Holocene ENSO variance reached a minimum ~4-5kybp, suggesting that ENSO may be sensitive to boreal fall insolation forcing. Consistent with 550ky of boreal fall precessional sensitivity in low-resolution Borneo stalagmite δ18O. But via what dynamical mechanisms?

GCM simulations of ENSO show sensitivity to summer/winter forcing à Holocene-long trend

e.g.CCSM4Liuetal.,2014

−0.5 −0.4 −0.3 −0.2 −0.1 0 0.1 0.2 0.3 0.4 0.50

1000

2000

3000

4000

5000

6000

7000

mean=0.138mean=0

x95%=0.106 x95%=0.036

x95%=0.052 x95%=0.088

x95%=0.024 x95%=0.116

Null TRACE

Linear regression coefficient (oC/10000yr)N

um

ber

of o

ccu

ren

ces

Nino3.4 pseudocoral PDFsets:100000 win:30yrs

nsample=50

nsample=200

nsample=1000

−60

−40

−20

0

% c

han

ge

of T

RA

CE

ENSO

ben

chm

ark

last

1ka

20 18 16 14 12 10 8 6 4 2 0

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

0.55

Age (ka)

ENSO

var

iab

ility

(1.5

−7

yr)

a

b

0 2 4 6 8 10 12 14 16 18 20

-3

-2

-1

0

1

2

3

4

5

6

7

8Proxy Observations

Scal

ed E

NSO

var

ianc

e

Scaled Seasonality

TLS fit, o = +0.1±0.1

TLS fit, 95% CI

0 5 10 15 20 25 30 35 40 45 50

1

1.2

1.4

1.6

1.8

2

2.2

2.4

2.6

2.8

3General Circulation Models

Scal

ed E

NSO

var

ianc

e

Scaled seasonality

HadGEM2 PI

GISS PI

KCM PI

CCSM4 PI

MIROC PI

MPI PI

CNRM PI

IPSL PI

CSIRO PI

TLS fit, m

= -0.012±0.0042

TLS fit, 95% CI

ref. 13, 14

this study

ref. 11, 12

ref. 18, 19

ref. 15, 16, 20, 21

ref. 17

ref. 22, 23

ref. 24, 25

Emile-Geay et al., Nat. Geo. 3 days ago

frequency entrainment: stronger seasonal cycle = reduced ENSO (e.g. Chang et al., 1994; Jin et al., 1996)

Emile-Geay et al., Nat. Geo. 3 days ago

but proxy records do not support this

0 2 4 6 8 10 12 14 16 18 20

-3

-2

-1

0

1

2

3

4

5

6

7

8Proxy Observations

Scal

ed E

NSO

var

ianc

e

Scaled Seasonality

TLS fit, o = +0.1±0.1

TLS fit, 95% CI

0 5 10 15 20 25 30 35 40 45 50

1

1.2

1.4

1.6

1.8

2

2.2

2.4

2.6

2.8

3General Circulation Models

Scal

ed E

NSO

var

ianc

e

Scaled seasonality

HadGEM2 PI

GISS PI

KCM PI

CCSM4 PI

MIROC PI

MPI PI

CNRM PI

IPSL PI

CSIRO PI

TLS fit, m

= -0.012±0.0042

TLS fit, 95% CI

ref. 13, 14

this study

ref. 11, 12

ref. 18, 19

ref. 15, 16, 20, 21

ref. 17

ref. 22, 23

ref. 24, 25

need to revisit relationship between annual cycle and ENSO variability

CLEMENT ET AL' ORBITAL CONTROLS 449

April

• . "• •;.,•- z., .,.•. •.,•- ,•. •:. ,.?- •-- 'i•

Uniform heating of the tropical Pacific

Atmospheric heating Atmospheric heating

Winds

Coupled interactions -..'-,', amplify on an intemnnual time.ale

SST

Figure 7. Schematic representation of the mechanism by which the tropical Pacific coupled ocean- atmosphere system responds to Milankovitch forcing. We consider (top) a uniform heating of the tropical Pacific ocean (cross-hatching represents a heating or warm SST anomaly). The seasonal position (April or October) of the Intertropical Convergence Zone (ITCZ)- the region of mean wind fidd convergence, high precipitation, or atmospheric heating, is represented by the shaded region. In April the total effect of the warm SST anomaly and the background wind field convergence gives an atmospheric heating that is uniform across the equator and hence will not be amphfied by zonally asymmetric coupled interactions. On the other hand, in October, when the ITCZ is north of the equator in the eastern Pacific, the combined effect of the SST anomaly and the background wind field convergence gives a zonally asymmetric atmospheric heating that drives easterly winds. The coupled system amplifies this signal on an interannual timescale. The result is a (bottom) La Nin5-1ike SST anomaly (shaded region).

be understood about the connection between climate and the structure of the equatorial therinocline.

The phase of the mean seasonal cycle, essential to the model's response to precessional forcing, is also specified in the background state. The mean change in NINO3 is related to the sign of the forcing in late summer/fall when the divergence field is most zonally asymmetric and the growth of perturbations is the largest. If the changes in seasonal cycle substantially went beyond what the model can simulate, it is not clear how the precessional forcing would affect the ENSO variability since it originates in the nonlinear interactions between the seasonal cycle and the wave dynamics [Zebiak and Cane, 1987; Tziperman et al., 1994; 1997; Chang et al., 1994]. However, recent investigations of the factors

controlling the phase of the seasonal cycle in the eastern equatorial Pacific generally point to the continental configuration and heating over land [Philander et al., 1996]. DeWitt and Schneider, submitted manuscript, 1999] performed a coupled GCM experiment changing the phase of the precessional cycle opposite to today's. While the seasonal cycle in the eastern equatorial Pa- cific is slightly shifted relative to today, the phasing is approximately the same with warming in the first half of the year and cooling in the latter half. Thus, it is possible that the phase of the seasonal cycle remained unchanged during the late Quaternary when the continental configuration was approximately the same as in the present day.

A more serious caveat concerns the simplicity of the

cold tongue seasonality defined by spring/fall conditions

Clement et al., 1999

recent resurgence of ENSO-annual cycle research: seasonally-dependent O-A feedbacks (e.g. DiNezio and Deser, 2014) state-dependent westerly wind bursts (e.g. Aaron Levine, AGU 2015; Kug et al., 2008, others) impacts of forcing on ENSO growth rates in fall (e.g. Karamperidou et al., 2015) annual cycle-ENSO combination modes (e.g. Stuecker et al., 2013)

Conclusions Borneo stalagmites provide evidence for boreal fall insolation control of ENSO:

- low resolution across 500ky - high resolution across 10ky

Stalagmite evidence corroborated by other ENSO-resolving proxies. Collectively, proxy data provide clear target for modeling investigations of annual cycle-ENSO interactions.