© 2006 Nature Publishing Group

Subtropical Arctic Ocean temperatures during thePalaeocene/Eocene thermal maximumAppy Sluijs1*, Stefan Schouten2*, Mark Pagani3, Martijn Woltering2, Henk Brinkhuis1,Jaap S. Sinninghe Damste2,4, Gerald R. Dickens5, Matthew Huber6, Gert-Jan Reichart4, Ruediger Stein7,Jens Matthiessen7, Lucas J. Lourens4, Nikolai Pedentchouk3, Jan Backman8, Kathryn Moran9

& the Expedition 302 Scientists†

The Palaeocene/Eocene thermal maximum, ,55 million years ago,was a brief period of widespread, extreme climatic warming1–3,that was associated with massive atmospheric greenhouse gasinput4. Although aspects of the resulting environmental changesare well documented at low latitudes, no data were available toquantify simultaneous changes in the Arctic region. Here weidentify the Palaeocene/Eocene thermal maximum in a marinesedimentary sequence obtained during the Arctic CoringExpedition5. We show that sea surface temperatures near theNorth Pole increased from ,18 8C to over 23 8C during thisevent. Such warm values imply the absence of ice and thus excludethe influence of ice-albedo feedbacks on this Arctic warming. Atthe same time, sea level rose while anoxic and euxinic conditionsdeveloped in the ocean’s bottom waters and photic zone, respect-ively. Increasing temperature and sea level match expectationsbased on palaeoclimate model simulations6, but the absolute polartemperatures that we derive before, during and after the event aremore than 10 8C warmer than those model-predicted. Thissuggests that higher-than-modern greenhouse gas concentrationsmust have operated in conjunction with other feedback mecha-nisms—perhaps polar stratospheric clouds7 or hurricane-inducedocean mixing8—to amplify early Palaeogene polar temperatures.

Stable carbon isotope (d13C) records of carbonate and organiccarbon from numerous sites show a prominent negative carbonisotope excursion across the Palaeocene/Eocene thermal maximum(PETM)2,9. The carbon isotope excursion is expressed as a .2.5‰drop in d13C, which signifies an input of at least 1.5 £ 1018 g of13C-depleted carbon, somewhat analogous in magnitude and com-position to current and expected fossil fuel emissions. The PETMcaptures ,200 kyr (ref. 10) and is associated with profound environ-mental changes that are well-documented at low- to mid-latitudes(,608), including a 4–8 8C temperature rise of surface and deepocean waters1–3 and major terrestrial and marine biotic changes11–13.Terrestrial mammal turnovers are consistent with mass migrationsacross Arctic regions resulting from high-latitude warming14, but noArctic data have existed to evaluate this hypothesis.

The Integrated Ocean Drilling Program Expedition (IODP) 302(or the Arctic Coring Expedition) recently recovered a Palaeogenemarine sedimentary record from Hole 4A (,878 52.00 0 N;1368 10.64

0E; 1,288 m water depth), on the Lomonosov ridge in

the central Arctic Ocean5. This ridge represents a fragment ofcontinental crust that rifted from the Eurasian shelf margin at highlatitudes (.858; Fig. 1) during the latest part of the Palaeocene epochand subsided to present depths after the Palaeocene. Upper Palaeoceneand Lower Eocene sediments between approximately 406 and 263 mcomposite depth below sea floor (m.c.d.) at Hole 4A consist oforganic-rich (,2% total organic carbon (TOC) by mass on average)siliciclastic claystone5. Shipboard observations showed that thisinterval is barren of calcareous and siliceous microfossils butyields rich assemblages of palynomorphs, notably organic-walleddinoflagellate cysts (dinocysts) and terrestrial pollen and spores5.

The PETM was identified from the top of Core 32X to within Core

LETTERS

Figure 1 | Location of IODP Hole 302-4A within the palaeogeographicreconstruction of the Arctic Basin at late Palaeocene–early Eocenetimes. The figure is modified from ref. 24. NP, North Pole.

1Palaeoecology, Institute of Environmental Biology, Utrecht University, Laboratory of Palaeobotany and Palynology, Budapestlaan 4, 3584 CD Utrecht, The Netherlands. 2RoyalNetherlands Institute for Sea Research (NIOZ), Department of Marine Biogeochemistry and Toxicology, PO Box 59, 1790 AB, Den Burg, Texel, The Netherlands. 3Department ofGeology and Geophysics, Yale University, PO Box 208109, New Haven, Connecticut 06520, USA. 4Department of Earth Sciences, Utrecht University, Budapestlaan 4, 3584 CDUtrecht, The Netherlands. 5Department of Earth Sciences, Rice University, 6100 Main Street, Houston, Texas 77005, USA. 6Earth and Atmospheric Sciences Department and thePurdue Climate Change Research Center, Purdue University, 550 Stadium Mall Drive, West Lafayette, Indiana 47906, USA. 7Alfred-Wegener-Institute for Polar and MarineResearch, Columbusstrasse, 27568 Bremerhaven, Germany. 8Department of Geology and Geochemistry, Stockholm University, Stockholm, SE-106 91, Sweden. 9University ofRhode Island, Bay Campus, Narragansett, Rhode Island 02882, USA.*These authors contributed equally to this work.†A list of authors and affiliations appears at the end of the paper.

Vol 441|1 June 2006|doi:10.1038/nature04668

610

© 2006 Nature Publishing Group

29X (,387–378.5 m.c.d.) by the occurrence of the dinocyst speciesApectodinium augustum, which is diagnostic of the PETM15 (Fig. 2;Supplementary Fig. S–1a). The lower bound is somewhat problem-atic, though, because the upper 50 cm of Core 32X has been disturbedby drilling and various proxies suggests that the sediment from thisinterval represents a mixture of uppermost Palaeocene and PETMmaterial16. Moreover, only 55 cm of section was recovered of thecritical Core 31X, which has an uncertain stratigraphic positionrelative to Cores 30X and 32X (see error bars in Fig. 2 andSupplementary Information). Stable carbon isotopes of bulk organiccarbon (d13CTOC) show a prominent ,6‰ drop between the top ofCore 32X (388 m.c.d.) and 31X (,386 m.c.d.), apart from one valuefrom the disturbed zone, followed by a gradual recovery throughCores 30X and 29X to ,378.5 m.c.d. (Fig. 2). The d13CTOC pattern isgenerally reproduced in the carbon isotope record of the C27 and C29

n-alkanes, which are biomarkers derived from the leaf waxes ofterrestrial higher plants16. Despite the core gaps, the magnitude andshape of the d13CTOC excursion resembles other shallow marinePETM sections, such as Doel in Belgium17, confidently correlatingthis interval to the Palaeocene/Eocene boundary event.

Before the PETM, Apectodinium was a subtropical dinoflagellaterestricted to low latitudes11,15. Thus, the sudden influx of Apectodi-nium spp. dinocysts across the PETM at Hole 4A (Fig. 2) suggests asubstantial rise in Arctic sea surface temperature (SST) to subtropicalor tropical levels. Angiosperm pollen becomes more abundant at theexpense of spores and gymnosperm pollen (Fig. 2), suggesting anexpanded growing season. The lack of calcareous microfossils pro-hibits the use of the common techniques for quantifying past SSTs.Instead, we employed the newly developed palaeothermometerTEX86

0 (see Supplementary Information), which is based on thedistribution of crenarchaeotal membrane lipids18. This distributionis independent of surface water parameters such as nutrient avail-ability or salinity18,19, and shows a highly significant linear correlationwith present-day mean annual SST in the range of 10 to 28 8C(Supplementary Fig. S–2B). Because the export of crenarchaeotallipids to the sea floor predominantly occurs during the season with

highest phytoplankton productivity, which in the Arctic Ocean issummer, our TEX86

0 record is probably skewed towards summertemperatures (see also Supplementary Information).

Arctic SSTs rose from,18 8C in the latest Palaeocene, to over 23 8Cduring the PETM, and subsequently decreased to ,17 8C by the endof the event (Fig. 2). Background SSTs from the latest Palaeocene andearly Eocene are generally consistent with the few other proxy dataestimates from Arctic locations with late Cretaceous and earlyPalaeogene strata20–22. The significantly lower terrestrial temperatureestimates from Ellesmere Island at 738N palaeolatitude23 are derivedfrom upper Lower Eocene strata and are similar to TEX86-derivedSSTs in the Arctic Ocean for that time period24; these estimates arethus not in disagreement with our data. Maximum SSTs coincidewith minimum d13C values during the PETM, while the coolingtrend mirrors the recovery pattern in d13C and a decrease ofangiosperm pollen.

Several lines of evidence (Fig. 2) suggest that the location of Hole4A was proximal to the coast and strongly influenced by fluvial inputsin the latest Palaeocene. For example, palynomorph assemblages fromupper Palaeocene strata are dominated by terrestrial spores and pollen(,90%). Those samples with sufficient dinocysts yield abundantSenegalinium spp. and Cerodinium spp. (Supplementary Fig. S–1),which probably come from dinoflagellates that tolerated low surface-water salinities24 and required nutrient-rich conditions25. Sedimentsfrom this interval also contain abundant amorphous organic matter,presumably of terrestrial origin. Moreover, values of the Branchedand Isoprenoid Tetraether (BIT) index—a measure for the amount ofriver-derived terrestrial organic matter relative to marine organicmatter26—are relatively high.

In contrast to uppermost Palaeocene sediments, palynomorphassemblages from the PETM interval are characterized by abundantdinocysts (60%) and substantially lower BIT indices (Fig. 2), indicat-ing a relative decrease of riverine-derived organic carbon. Also, theincrease in the Rock Eval hydrogen index suggests a relative increase inaquatic versus terrestrial organic matter (Supplementary Fig. S–3A).We attribute the enhanced influence of marine conditions during the

Figure 2 | Core recovery and palynological and geochemical results acrossthe PETM of IODP Hole 302-4A. Core 31X was plotted 100 cm lower thanm.c.d.5 for illustration purposes. Error bars connected to Core 31X in therecovery column indicate the uncertainty of its stratigraphic position (seeSupplementary Information). Orange bars indicate intervals affected bydrilling disturbance. Stable carbon isotopes are expressed relative to the

PeeDee Belemnite standard. Low-salinity-tolerant dinocysts compriseSenegalinium spp., Cerodinium spp., and Polysphaeridium spp.,while Membranosphaera spp., Spiniferites ramosus complex, andAreoligera-Glaphyrocysta cpx. represent the typical normal marine species25

(Supplementary Fig. S–1). Arrows and A. aug indicate the first and lastoccurrences of dinocyst Apectodinium augustum.

NATURE|Vol 441|1 June 2006 LETTERS

611

© 2006 Nature Publishing Group

PETM to a sea level rise, an interpretation consistent with evidencefrom other neritic locations27. The gradual return to more terrestrialinfluence later in the event probably results from subsequentregression. Despite the transgression, low-salinity-tolerant dinocystsremain dominant (Fig. 2), showing that brackish surface waterspersisted during the PETM. If the earliest Palaeogene greenhouseworld was continental-ice-free, a mechanism for the sea-level risemay comprise the ,5 m thermal expansion of sea water expectedfrom a 5–8 8C (refs 2, 3) increase in deep ocean temperatures.

The occurrence of laminated sediments from the onset of thePETM (although hardly visible in Core 31X due to its disturbedstate) up to 382.5 m.c.d., and the absence of benthic foraminiferallinings (Supplementary Fig. S–3A) suggest that bottom watersbecame anoxic during the PETM. Within the laminated interval,derivatives of the characteristic pigment isorenieratene are recordedin concentrations of up to 2 mg g21 sediment; they are below thedetection limit outside this interval (Fig. 2; Supplementary Fig. S–3).These compounds are derived from the brown strain of photosyn-thetic green sulphur bacteria, which requires euxinic (anoxic andsulphidic) conditions to thrive28. Accordingly, at the PETM photiczone euxinia developed at the drill site coincident with bottom wateranoxia, which gradually disappeared during the recovery of SST andd13C excursion (Fig. 2).

We can exclude selective preservation as a mechanism to explainthe marked changes in organic biomarkers and palynomorph assem-blages that occur in coincidence with water column anoxia. First, thepreservation of organic matter is also excellent outside the laminatedinterval (.2% TOC on average; Supplementary Fig. S–3A) andsecond, most of our proxies compare the relative abundance ofstructurally similar organic compounds that are equally susceptibleto oxidation (see Supplementary Information).

The euxinic conditions were potentially caused by several factors.For example, increased freshwater input, greater nutrient load andwarmer temperatures would all conspire to reduce dissolved O2 inthe water column. However, given the shallow water depth of the site,an important factor was probably intense stratification due to theinfluence of a brackish surface-water lid. Although several mecha-nisms could drive such stratification, given that low-salinity-tolerantdinocysts remain dominant despite the more distal position of thesite, and given the data presented in a companion paper16, thesimplest explanation is that decreased mixing resulted fromincreased SSTs and enhanced fluvial runoff, with the latter alsosupplying extra nutrients to increase production and saturate photiczone respiration. The termination of euxinic conditions coincideswith increasing surface salinities16 (Fig. 2) and cooling, suggesting anincrease of mixing with more-saline deeper waters.

Even if we assume that our TEX860 temperatures represent sum-

mer values (see Supplementary Information), palaeoclimate modelssimulating the early Palaeogene world with 2,000 p.p.m.v. of CO2 inthe atmosphere6 underestimate Arctic Ocean summer SSTs by at least15 8C for the PETM and 10 8C for the surrounding late Palaeoceneand early Eocene. This discrepancy may be even larger because theinitial part of the PETM, and potentially also the strata formed undermaximum temperatures, were possibly not recovered (Fig. 2). On theother hand, the magnitude of the carbon isotope excursion iscomparable to previous studies and peak PETM temperatures laggedthe onset of the carbon isotope excursion by ,40 kyr (ref. 1),indicating that optimum Arctic SSTs are probably covered in ourrecord.

The models consistently predict pole-to-Equator temperaturegradients of ,30 8C (ref. 29). Such gradients represent significantoverestimates because they would imply unrealistically warm tropicalSSTs considering our polar temperatures. The high polar tempera-tures and reduced pole-to-Equator temperature gradients cannot beexplained by invoking even greater greenhouse gas concentrationsbecause this would elevate tropical SSTs, which in existing modelpredictions already match or exceed those determined from proxy

records at low-latitude locations6. Also, ocean heat transport isunlikely to be the cause because this requires a threefold increase,which cannot be simulated in the current generation of fully coupledocean–atmosphere climate models29. Similarly, atmospheric generalcirculation models do not support strong enough positive feedbacksin atmospheric heat transport30. Consequently, we surmise thatphysical processes that are not yet incorporated in the modelsoperated in conjunction with high greenhouse gas concentrationsto enhance polar warmth and reduce the pole-to-Equator tempera-ture gradient during the early Palaeogene. These processes couldinclude high-latitude warming and tropical cooling through theenhancement of polar stratospheric clouds7, and hurricane-inducedocean mixing8, respectively.

With latest Palaeocene SSTs of 18 8C it is not likely that ice waspresent in the Arctic. This implies that the PETM at Hole 4A allows usto examine the Arctic environment and the nature of polar ampli-fication during a time of massive greenhouse gas emissions andextreme global warming in the absence of ice-albedo feedbacks.Interestingly, polar amplification of temperature rise at the PETMappears to have been minor (Fig. 2; refs 1–3), suggesting that thestrengthening of the mechanism that caused above early Palaeogenepolar temperature amplification was small at the PETM. Ourextremely warm polar temperatures indicate that, despite muchrecent progress, feedbacks responsible for early Palaeogene mid- tohigh-latitude warmth remain poorly understood and are notimplemented in existing climate models.

METHODSPalynology. Sediments were oven-dried at 60 8C. To ,2 g of sediment, a knownamount of Lycopodium spores were added, after which the sample was treatedwith 30% HCl and twice with 30% HF for carbonate and silicate removal,respectively. After sieving over a 15-mm nylon mesh sieve, residues were analysedat 500 £ magnification until a minimum of 200 dinocysts had been recorded persample. Absolute quantitative numbers were calculated using the relativenumber of Lycopodium.Organic geochemistry. Powdered and freeze-dried sediments were analysed for%TOC and d13CTOC with a Fison NA 1500 CNS analyser, connected to aFinnigan Delta Plus mass spectrometer. Analytical precision and accuracywere determined by replicate analyses and by comparison with internationaland in-house standards, and were better than 0.1% and 0.1‰ for %TOC andd13CTOC, respectively.

Powdered and freeze-dried sediments (1–3 g dry mass) were extracted withdichloromethane (DCM)/methanol (2:1) by using the Dionex acceleratedsolvent extraction technique. The extracts were separated by Al2O3 columnchromatography using hexane/DCM (9:1) and DCM/methanol (1:1) to yield theapolar and polar fractions, respectively. The apolar fractions were analysed forisorenieratene derivatives by gas chromatography and gas chromatography/massspectrometry, while the polar fractions were analysed for tetraether lipids andused to calculate TEX86

0 (see Supplementary Information; reproducibility waswithin ,1 8C) and BIT (see ref. 26 for method description) indices.

Received 7 July 2005; accepted 21 February 2006.

1. Zachos, J. C. et al. A transient rise in tropical sea surface temperature during thePaleocene-Eocene thermal maximum. Science 302, 1151–-1154 (2003).

2. Kennett, J. P. & Stott, L. D. Abrupt deep-sea warming, palaeoceanographicchanges and benthic extinctions at the end of the Palaeocene. Nature 353,225–-229 (1991).

3. Tripati, A. & Elderfield, H. Deep-sea temperature and circulation changes atthe Paleocene-Eocene thermal maximum. Science 308, 1894–-1898 (2005).

4. Dickens, G. R., O’Neil, J. R., Rea, D. K. & Owen, R. M. Dissociation of oceanicmethane hydrate as a cause of the carbon isotope excursion at the end of thePaleocene. Paleoceanography 10, 965–-971 (1995).

5. Backman, J., Moran, K., McInroy, D. B., Mayer, L. A. & the Expedition 302Scientists. Arctic Coring Expedition (ACEX). Proc. ODP 302 I/doi:10.2204/iodp.proc.302.2006 (Integrated Ocean Drilling Program ManagementInternational, College Station, Texas, 2006).

6. Shellito, C. J., Sloan, L. C. & Huber, M. Climate model sensitivity toatmospheric CO2 levels in the Early-Middle Paleogene. Palaeogeogr.Palaeoclimatol. Palaeoecol. 193, 113–-123 (2003).

7. Sloan, L. C. & Pollard, D. Polar stratospheric clouds: A high latitude warmingmechanism in an ancient greenhouse world. Geophys. Res. Lett. 25, 3517–-3520(1998).

LETTERS NATURE|Vol 441|1 June 2006

612

© 2006 Nature Publishing Group

8. Emanuel, K., DesAutels, C., Holloway, C. & Korty, R. Environmental Control ofTropical Cyclone Intensity. J. Atmos. Sci. 61, 843–-858 (2004).

9. Koch, P. L., Zachos, J. C. & Gingerich, P. D. Correlation between isotoperecords in marine and continental carbon reservoirs near the Palaeocene/Eocene boundary. Nature 358, 319–-322 (1992).

10. Rohl, U., Bralower, T. J., Norris, G. & Wefer, G. A new chronology for the latePaleocene thermal maximum and its environmental implications. Geology 28,927–-930 (2000).

11. Crouch, E. M. et al. Global dinoflagellate event associated with the latePaleocene thermal maximum. Geology 29, 315–-318 (2001).

12. Thomas, E. & Shackleton, N. J. in Correlation of the Early Paleogene inNorthwestern Europe (eds Knox, R. W. O. B., Corfield, R. M. & Dunay, R. E.)401–-441 (Geological Society of London Special Publication 101, 1996).

13. Wing, S. L. in Late Paleocene–-early Eocene Climatic and Biotic Events in theMarine and Terrestrial Records (eds Aubry, M.-P., Lucas, S. G. & Berggren, W. A.)380–-400 (Columbia Univ. Press, New York, 1998).

14. Bowen, G. J. et al. Mammalian dispersal at the Paleocene/Eocene boundary.Science 295, 2062–-2065 (2002).

15. Bujak, J. P. & Brinkhuis, H. in Late Paleocene–-early Eocene Climatic and BioticEvents in the Marine and Terrestrial Records (eds Aubry, M.-P., Lucas, S. G. &Berggren, W. A.) 277–-295 (Columbia Univ. Press, New York, 1998).

16. Pagani, M. et al. Arctic’s hydrology during global warming at thePalaeocene/Eocene thermal maximum. Nature (submitted).

17. Steurbaut, E. et al. in Causes and Consequences of Globally Warm Climates in theEarly Paleogene (eds Wing, S. L., Gingerich, P., Schmitz, B. & Thomas, E.)291–-317 (Geological Society of America Special Paper 369, Boulder, Colorado,2003).

18. Schouten, S., Hopmans, E. C., Schefuß, E. & Sinninghe Damste, J. S.Distributional variations in marine crenarchaeotal membrane lipids: a new toolfor reconstructing ancient sea water temperatures? Earth Planet. Sci. Lett. 204,265–-274 (2002).

19. Wuchter, C., Schouten, S., Coolen, M. J. L. & Sinninghe Damste, J. S.Temperature-dependent variation in the distribution of tetraether membranelipids of marine Crenarchaeota: Implications for TEX86 paleothermometry.Paleoceanography 19, doi:10.1029/2004PA001041 (2004).

20. Markwick, P. J. Fossil crocodilians as indicators of Late Cretaceous andCenozoic climates: Implications for using palaeontological data inreconstructing palaeoclimate. Palaeogeogr. Palaeoclimatol. Palaeoecol. 137,205–-271 (1998).

21. Tripati, A., Zachos, J., Marincovich, L. Jr & Bice, K. Late Paleocene Arcticcoastal climate inferred from molluscan stable and radiogenic isotope ratios.Palaeogeogr. Palaeoclimatol. Palaeoecol. 170, 101–-113 (2001).

22. Jenkyns, H. C., Forster, A., Schouten, S. & Sinninghe Damste, J. S. Hightemperatures in the Late Cretaceous Arctic Ocean. Nature 432, 888–-892(2004).

23. Fricke, H. C. & Wing, S. L. Oxygen isotope and paleobotanical estimates of

temperature and d18O-latitude gradients over North America during the earlyEocene. Am. J. Sci. 304, 612–-635 (2004).

24. Brinkhuis, H. et al. Episodic fresh surface waters in the Eocene Arctic Ocean.Nature doi:10.1038/nature04692 (this issue).

25. Sluijs, A., Pross, J. & Brinkhuis, H. From greenhouse to icehouse; organic-walleddinoflagellate cysts as paleoenvironmental indicators in the Paleogene. EarthSci. Rev. 68, 281–-315 (2005).

26. Hopmans, E. C. et al. A novel proxy for terrestrial organic matter in sedimentsbased on branched and isoprenoid tetraether lipids. Earth Planet. Sci. Lett. 224,107–-116 (2004).

27. Speijer, R. P. & Morsi, A.-M. M. Ostracode turnover and sea-level changesassociated with the Paleocene-Eocene thermal maximum. Geology 30, 23–-26(2002).

28. Sinninghe-Damste, J. S., Wakeham, S. G., Kohnen, M. E. L., Hayes, J. M. & DeLeeuw, J. W. A 6,000-year sedimentary molecular record of chemoclineexcursions in the Black Sea. Nature 362, 827–-829 (1993).

29. Huber, M., Sloan, L. C. & Shellito, C. J. in Causes and Consequences of GloballyWarm Climates in the Early Palaeogene (eds Wing, S. L., Gingerich, P. D.,Schmitz, B. & Thomas, E.) 25–-47 (Geological Society of America Special Paper369, Boulder, Colorado, 2003).

30. Caballero, R. & Langen, P. L. The dynamic range of poleward energy transportin an atmospheric general circulation model. Geophys. Res. Lett. 32, doi:10.1029/2004GL021581 (2005).

Supplementary Information is linked to the online version of the paper atwww.nature.com/nature.

Acknowledgements A.S. thanks the Utrecht Biogeology Centre for funding. H.B.thanks the Netherlands Organization for Scientific Research, and UtrechtUniversity for enabling participation in the ACEX expedition. M.H. thanks thePurdue Climate Change Research Center, ITaP and the Purdue ResearchFoundation for their continued support. This research used samples and dataprovided by the IODP. We thank L. Bik, J. van Tongeren, N. Welters and A. vanDijk for technical support, and C. E. Stickley for discussions.

Author Contributions A.S. and H.B. carried out the palynology, A.S. & G.-J.R. thed13CTOC, S.S., M.W. and J.S.S.D. the TEX86

0 , BIT and isorenieratane analyses. R.S.generated the hydrogen index data. J.B. and K.M. were the co-chiefs of theACEX expedition. N.P., J.M and the Expedition 302 Scientists were involved ingenerating shipboard and shore-based ACEX data. A.S., S.S., M.P., H.B., J.S.S.D.,G.R.D., M.H. and L.J.L. contributed to interpreting the data and writing the paper.

Author Information Reprints and permissions information is available atnpg.nature.com/reprintsandpermissions. The authors declare no competingfinancial interests. Correspondence and requests for materials should beaddressed to A.S. (a.sluijs@bio.uu.nl) or S.S. (schouten@nioz.nl)

Expedition 302 Scientists (those not listed above): Steve Clemens1, Thomas Cronin2, Frederique Eynaud3, Jerome Gattacceca4,Martin Jakobsson5, Ric Jordan6, Michael Kaminski7, John King8, Nalan Koc9, Nahysa C. Martinez10, David McInroy11, Theodore C. Moore Jr12,Matthew O’Regan8, Jonaotaro Onodera13, Heiko Palike14, Brice Rea15, Domenico Rio16, Tatsuhiko Sakamoto17, David C. Smith8, Kristen E.K. St John18, Itsuki Suto19, Noritoshi Suzuki20, Kozo Takahashi13, Mahito Watanabe21 & Masanobu Yamamoto22

1Geological Sciences, Brown University, 324 Brook Street, PO Box 1846, Providence, Rhode Island 02912-1846, USA. 2US Geological Survey, Eastern Earth Surface ProcessesTeam, 926A USGS National Center, Reston, Virginia 20192, USA. 3Department de Geologie et Oceanographie, Universite Bordeaux 1, Avenue des facultes, c/o Bernei Housen,33405 Talence Cedex, France. 4Department of Geophysics, CEREGE (CNRS)/University of Aix-Marseille 3, BP 80, 13545 Aix-en-Provence Cedex 4, France. 5Department ofGeology and Geochemistry, Stockholm University, 106 91 Stockholm, Sweden. 6Department of Earth and Environmental Sciences, Faculty of Science, Yamagata University, 1-4-12Kojirakawa-machi, Yamagata 990-8560, Japan. 7Department of Earth Sciences, University College London, Gower Street, London WC1E 6BT, UK. 8Graduate School ofOceanography, University of Rhode Island, Narragansett Bay Campus, South Ferry Road, Narragansett, Rhode Island 02882, USA. 9Norwegian Polar Institute, Polar EnvironmentalCenter, N-9296 Tromsø, Norway. 10Department of Earth Sciences, Boston University, 685 Commonwealth Avenue, Boston, Massachusetts 02215, USA. 11British GeologicalSurvey, Murchison House, West Mains Road, Edinburgh, EH9 3LA, UK. 12Geological Sciences, University of Michigan, Ann Arbor, Michigan 48109-1063, USA. 13Department ofEarth and Planetary Sciences, Graduate School of Sciences, Kyushu University, Hakozaki 6-10-1, Higashi-ku, Fukuoka 812-8581, Japan. 14School of Ocean and Earth Science,University of Southampton, Southampton Oceanography Center, European Way, Southampton SO14 3ZH, UK. 15Department of Geography and Environment, School ofGeosciences, University of Aberdeen, Elphinstone Road, Aberdeen AB24 3UF, UK. 16Department of Geology, Paleontology and Geophysics, University of Padova, Via Giotto 1I-35137 Padova, Italy. 17Institute for Research on Earth Evolution (IFREE), Japan Agency for Marine-Earth Science and Technology (JAMSTEC), Natsushima-cho 2-15, Yokosuka237-0061, Japan. 18Department of Geology and Environmental Science, MSC 7703, James Madison University, Harrisonburg, Virginia 22807, USA. 19Institute of Life andEnvironmental Science, University of Tsukuba, Tennoudai 1-1-1, Tsukuba, Ibaraki 305-8572, Japan. 20Institute of Geology and Paleontology, Graduate School of Science, TohokuUniversity, Aramaki, Aoba, Aoba-ku, Sendai City 980-8578, Japan. 21Institute of Geoscience, National Institute of Advanced Industrial Science, and Technology (GeologicalSurvey of Japan) AIST Tsukuba Central 7, Higashi-1-1-1, Tsukuba, Ibaraki 305-8567, Japan. 22Graduate School of Environmental Earth Science, Hokkaido University, Kita-10,Nishi-5, Kita-ku, Sapporo 060-0810, Japan

NATURE|Vol 441|1 June 2006 LETTERS

613

1

Supplementary Information

Subtropical Arctic Ocean temperatures during the Palaeocene-Eocene thermal maximum. Appy Sluijs, Stefan Schouten, Mark Pagani, Martijn Woltering, Henk Brinkhuis, Jaap S. Sinninghe Damsté, Gerald R. Dickens, Matthew Huber, Gert-Jan Reichart, Ruediger Stein, Jens Matthiessen, Lucas J. Lourens, Nikolai Pedentchouk, Jan Backman, Kathryn Moran and the Expedition Scientists

Relative stratigraphic position of Core 31X

The stratigraphic position of the 55 cm recovered section of Core 302-4A-31X relative to Cores 30X and 32X is uncertain (Fig. 2). It may be located anywhere in the interval between 384.54 and 388 meters composite depth (mcd). For illustration purposes, we have placed this core 1m lower than indicated in ref. 1.

Description and calibration of TEX86’

Originally TEX86 values were calculated as described in ref. 2 and converted to sea surface temperature (SST). However, the glycerol dialkyl glycerol tetraether (GDGT) lipid containing three cyclopentane rings (GDGT-3), almost always the least abundant GDGT lipid in marine sediments, was unusually high in Core 302-4A-32X (Fig. S-2A). Since terrestrial organic matter also contains isoprenoid GDGT lipids with cyclopentane rings3 (and Weijers, J Schouten, S., Sinninghe Damsté, J.S., unpublished results), it is likely that the high terrestrial input in this section has disturbed the aquatic signal. The high abundance of GDGT-3 resulted in unusually high TEX86 values in Core 32X. To circumvent this problem, TEX86’ was devised which has the same definition as TEX86 except that isomer GDGT-3 was removed from the denominator2. The TEX86’ was determined for 104 marine surface sediments and found to correlate very well with annual mean SST: TEX86’ = 0.016 x SST + 0.20 with R2 = 0.93 (Fig S-2B). This equation was used to convert TEX’86 into SST. Sediments from Cores 30X and 31X show a normal marine tetraether lipid distribution, i.e., showing a very minor peak at GDGT-3, resulting in the TEX86’ values only slightly different from TEX86 values (Fig SI-2C). Interestingly, the top part of Core 30X, which shows a large terrestrial influence (Fig. 2), shows again a relatively large offset between TEX86’ and TEX86, evidencing a larger contribution of the GDGT-3. This is consistent with a terrestrially-derived contribution to the GDGT-3 peak.

2

Which temperature does TEX86’ indicate in the Arctic Ocean?

TEX86 and TEX86’ are calibrated by core top analysis to mean annual mean SST. This empirical relation is not the same as a causal relationship. Several studies have shown that the cell number of crenarchaeota in the water column strongly depends on the season4. In most studies a negative correlation has been observed between the cell abundance of crenarchaeota and phytoplankton, likely because they compete for the nutrient ammonia. As crenarchaeota are chemoautotrophic organisms5 and thus not directly depend on light it is likely that crenarchaeota predominantly thrived during times of low cell abundances of phytoplankton and low light intensities, i.e. in the Arctic winter. However, for the crenarchaeotal signal to reach the sediment floor a significant sedimentation flux is needed. In the present day ocean significant organic carbon fluxes are observed during periods of high phytoplanktonic productivity because an active food web leads to grazing and feacal pellet packaging6,7. In agreement with this we recently recorded the highest fluxes of crenarchaeotal lipids in the Arabian Sea coinciding with the periods of high productivity, despite that their relative abundance was lower in these periods (Wuchter C., Schouten S., Wakeham, S.G. and Sinninghe Damsté, J.S., unpublished results). In the Eocene Arctic Ocean the bulk of the signal will be derived from those periods with highest biological productivity, which at this high latitude must be the summer months. Hence we refined our interpretation of the TEX86’ estimates by suggesting that its signal in the Eocen Arctic Ocean, although calibrated to globally-derived annual mean SSTs, may record the, on average, higher summer SSTs due to the flux-dependency of the signal.

Preservation of the organic matter used in this study

Water column anoxia, in this paper indicated by the laminated sediments and the presense of isoreneratene derivatives, will lead to an improved preservation of certain organic compounds8. Below, we discuss the preservation of the different types of organic matter used in this paper and conclude that selective preservation did not affect our proxy records.

The presence of isorenieratene derivatives has been frequently used to detect anoxia extending up into the photic zone of the water column during the Phanerozoic9,10, including late Pliocene sapropel formation in the eastern Mediterranean11. These compounds are derived from photosynthetic green sulfur bacteria which require both light and free sulfide (i.e. euxinic conditions) to thrive. In the particular, these organisms are found in lakes and marine settings where the water column is euxinic such as the present day Black Sea12. Accordingly, the presence of isorenieratene

3

derivatives in ACEX Cores 30X and 31X indicates that euxinic conditions prevailed in the photic zone in the Arctic ocean during the PETM. These compounds preserve quite well, even when the water column is re-oxygenated and sediments are inhabited by benthic organisms13. Therefore we exclude selective preservation as possible bias on the presence/absence pattern of these compounds. Moreover, well preserved palynomorphs and substantial organic matter concentrations (~2% TOC) outside the laminated interval with isorenieratene derivatives, indicates that no severe organic matter degradation occurred. In this respect, we also exclude a bias on the terrestrial vs. marine palynomorph ratio, since it has recently been shown that preferential preservation of terrestrial palynomorphs only occurs in well-oxygenated bottom water settings14.

The terrestrial and marine compounds used in the BIT index represent structurally very similar compounds15. It has been shown that the relative distribution of the different isomers of glycerol dialkyl glycerol tetraethers (GDGTs) is independent of the oxygen concentration in the water column16. Thus, the absolute amounts of GDGTs may have increased within the laminated interval, but their relative distribution, i.e. the BIT index, has remained unaffected.

Bottom water anoxia at the PETM

Organic linings of benthic foraminifera are common through the latest Palaeocene and earliest Eocene except during the photic zone euxinia, indicated by isorenieratane (Fig. S-3A). Sediments from the latter interval are laminated (Fig. S-3B), which implies that no bioturbation occurred after deposition of the sediments and that bottom waters were anoxic. Despite a large scatter % TOC is on average ~1.3% higher during the PETM compared to the latest Palaeocene. Unfortunately, due to the core recovery problems and potential changes in siliciclastic sediment supply related to the transgression there is relatively poor grip on sediment accumulation rates across the studied interval. However, some of the enhanced % TOC may be due to increased nutrient supply by rivers, resulting in higher phytoplankton production and, under anoxic bottom water conditions, high organic matter accumulation.

4

1. Backman, J., Moran, K., McInroy, D. B., Mayer, L. A. & Expedition-Scientists. Proceedings of the Integrated Ocean Drilling Program, 302 (Integrated Ocean Drilling Program Management International, Inc., College Station TX, in the press).

2. Schouten, S., Hopmans, E. C., Schefuß, E. & Sinninghe Damsté, J. S. Distributional variations in marine crenarchaeotal membrane lipids: a new tool for reconstructing ancient sea water temperatures? Earth and Planetary Science Letters 204, 265-274 (2002).

3. Weijers, J. W. H., Schouten, S., van der Linden, M., van Geel, B. & Sinninghe Damsté, J. S. Water table related variations in the abundance of intact archaeal membrane lipids in a Swedish peat bog. FEMS Microbiol. Lett. 239, 51-56 (2004).

4. Murray, A. E. et al. A time series assessment of planktonic archaeal variability in the Santa Barbara Channel. Aquatic Microbial Ecology 20, 129-145 (1999).

5. Wuchter, C., Schouten, S., Boschker, H. T. S. & Sinninghe Damste, J. S. Bicarbonate uptake by marine Crenarchaeota. FEMS Microbiology Letters 219, 203-207 (2003).

6. Wakeham, S. G. & Canuel, E. A. Organic geochemistry of particulate matter in the eastern tropical North Pacific Ocean: Implications for particle dynamics. Journal of Marine Research 46, 183-213 (1988).

7. Wakeham, S. G. & Lee, C. in Organic Geochemistry (eds. Engel, M. & Macko, S.) 145-169 (Plenum Press, New York, 1993).

8. Sinninghe Damste, J. S., Rijpstra, W. I. C. & Reichart, G.-j. The influence of oxic degradation on the sedimentary biomarker record II. Evidence from Arabian Sea sediments. Geochimica et Cosmochimica Acta 66, 2737-2754 (2002).

9. Koopmans, M. P. et al. Diagenetic and catagenetic products of isorenieratene: Molecular indicators for photic zone anoxia. Geochimica et Cosmochimica Acta 60, 4467-4496 (1996).

10. Sinninghe Damste, J. S. & Koster, J. A euxinic southern North Atlantic Ocean during the Cenomanian/Turonian oceanic anoxic event. Earth and Planetary Science Letters 158, 165-173 (1998).

11. Passier, H. F. et al. Sulphidic Mediterranean surface waters during Pliocene sapropel formation. Nature 397, 146-149 (1999).

12. Sinninghe-Damsté, J. S., Wakeham, S. G., Kohnen, M. E. L., Hayes, J. M. & Leeuw, J. W. d. A 6,000-year sedimentary molecular record of chemocline excursions in the Black Sea. Nature 362, 827 - 829 (1993).

13. Kenig, F., Hudson, J., D., Sinninghe Damste, J. S. & Popp, B. N. Intermittent euxinia: Reconciliation of a Jurassic black shale with its biofacies. Geology 32, 421-424 (2004).

14. Reichart, G.-J. & Brinkhuis, H. Late Quaternary Protoperidinium cysts as indicators of paleoproductivity in the northern Arabian Sea. Marine Micropaleontology 49, 303-370 (2003).

15. Hopmans, E. C. et al. A novel proxy for terrestrial organic matter in sediments based on branched and isoprenoid tetraether lipids. Earth and Planetary Science Letters 224, 107-116 (2004).

16. Schouten, S., Hopmans, E. C. & Sinninghe Damsté, J. S. The effect of maturity and depositional redox conditions on archaeal tetraether lipid palaeothermometry. Organic Geochemistry 35, 567-571 (2004).

A

Figure S-2. TEX86' calibration A) HPLC/MS base peak chromatograms showing the chrenarchaeotal membrane lipid distribution of samples from Cores 31X and 32X. B) Calibration of TEX86' to mean annual sea surface temperature. C) Comparison between TEX86' and TEX86 across the PETM of Site 302-4A.

TEX86' = 0.016T + 0.20R2 = 0.93

0.1

0.2

0.4

0.5

0.6

0.7

0.8

0 10 15 20 25 30Mean annual SST (ºC)

TE

X86

'

B C

17 19 2115 23 25

Core

380

382

384

386

390

De

pth

(m

cd)

378

32X

31X

30X

29X

TEX86'(°C)

TEX86

(°C)

32X-1, 101-102389.01 mcd

31X-CC, 23-25385.74 mcd

GDGT-0

GDGT-1

GDGT-2

GDGT-3

Crenarchaeolisomer

Crenarchaeol

2 40

% TOCIsorenieratane(µg/g)

1 20

Figure S-3. Additional indicators for water column euxinia and terrestrial influence. A) Isorenieratane and foraminifer lining distribution through the latest Palaeocene - earliest Eocene. Laminated interval indicated by stripes. Laminae are unclear in Core 31X due to drilling disturbance. The hydrogen index shows a substantial increase during the PETM, consistent with increased auquatic versus terrestrial organic matter. Despite significante scatter, average % Total Organic Carbon (TOC) increases by ~1% at the PETM. B) Core picture (interval 302-4A-30X-3, 123-129 cm) showing laminations.

BA 11 22 33 440

foraminiferal linings(x10.000/g)

?

Hydrogen Index(mgHC/gC)

100 200 3000

δ13CTOC

(‰ vs PDB)

-32 -28 -26 -24

380

382

384

386

390

De

pth

(m

cd)

378

Hole 4ACore

-30

32X

31X

30X

29X

Exp

editi

on

Site

Hol

e

Cor

e

type

sec

t (c

m)

b (c

m)

dep

th (m

cd)

Ape

ctod

iniu

m s

pp.

Are

olig

era/

Gla

phyr

ocys

ta c

px

Cal

igod

iniu

m a

cera

s

Cer

odin

ium

/Def

land

rea

cpx

Cor

dosp

haer

idiu

m c

px

Sen

egal

iniu

m s

pp.

Spi

nife

rites

cpx

Mem

bran

osph

aera

cpx

Oth

ers

n di

nocy

sts

% A

pect

odin

ium

spp

.

% A

reol

iger

a/G

laph

yroc

ysta

cpx

% C

alig

odin

ium

ace

ras

% C

erod

iniu

m/D

efla

ndre

a c

px

% C

ordo

spha

erid

ium

cpx

% S

eneg

alin

ium

spp

.

% S

pini

ferit

es c

px

% M

embr

anos

phae

ra c

px

% P

olys

phae

ridiu

m c

px

% O

ther

s

% lo

w s

alin

ity to

lera

nt d

inoc

ysts

Fora

min

ifer l

inin

gs p

er g

ram

Ang

iosp

erm

s

Spo

res

Gym

nosp

erm

s

n te

rr. P

alm

orph

s

% A

ngio

sper

ms

% S

pore

s

% G

ymno

sper

ms

num

ber o

f mar

ine

paly

nom

orph

s

% m

arin

e pa

lyno

mor

phs

% te

rrest

rial p

alyn

omor

phs

302 4 A 29 X 1 41 43 378.21 0 50 0 4 0 63 31 44 0 17 209 0 24 0 2 0 30 15 21 0 8 36 0 141 175 327 643 22 27 51 274 30 70302 4 A 29 X 1 61 63 378.41 129 52 94 275 47 19 34 12 4 96302 4 A 29 X 1 81 83 378.61 1 56 0 25 0 58 10 37 1 19 207 0 27 0 12 0 28 5 18 0 9 45 0 137 176 165 478 29 37 35 244 34 66302 4 A 29 X 1 101 103 378.81 0 1 0 17 0 94 57 33 1 15 218 0 0 0 8 0 43 26 15 0 7 55 6.E+03 282 206 215 703 40 29 31 316 31 69302 4 A 29 X CC 1 3 379.01 0 2 4 35 0 92 23 36 1 8 201 0 1 2 17 0 46 11 18 0 4 66 2.E+04 503 300 1527 2330 22 13 66 263 10 90

302 4 A 30 X 1 1 3 380.31 77 101 94 272 28 37 35 67 20 80302 4 A 30 X 1 21 23 380.51 89 127 57 273 33 47 21 32 10 90302 4 A 30 X 1 41 43 380.71 91 124 114 329 28 38 35 27 7 93302 4 A 30 X 1 61 63 380.91 100 113 98 311 32 36 32 11 3 97302 4 A 30 X 1 81 83 381.11 102 78 106 286 36 27 37 7 2 98302 4 A 30 X 1 101 103 381.31 0 0 2 10 0 71 57 49 0 10 200 0 0 1 5 0 36 29 24 0 5 43 1.E+03 159 176 100 435 37 40 23 219 12 88302 4 A 30 X 1 121 123 381.51 16 2 1 10 2 29 20 35 1 5 122 13 2 1 8 2 23 16 29 1 4 40 5.E+04 411 326 576 1313 31 25 44 201 13 87302 4 A 30 X 1 141 143 381.71 13 0 1 29 0 27 47 80 0 18 215 6 0 0 13 0 13 22 37 0 8 30 3.E+04 267 362 491 1120 24 32 44 253 18 82302 4 A 30 X 2 1 3 381.82 7 0 2 18 0 13 29 124 1 8 201 3 0 1 9 0 7 14 62 1 4 17 8.E+03 300 344 421 1065 28 32 40 230 18 82302 4 A 30 X 2 21 23 382.02 5 1 0 25 0 40 41 93 0 5 210 2 0 0 12 0 19 20 44 0 2 33 2.E+04 345 796 296 1437 24 55 21 261 15 85302 4 A 30 X 2 41 43 382.22 5 3 16 45 0 15 50 59 14 11 218 2 1 7 21 0 7 23 27 6 5 37 2.E+04 116 67 106 289 40 23 37 255 47 53302 4 A 30 X 2 61 63 382.42 4 1 8 15 1 27 58 82 7 12 215 2 0 4 7 0 13 27 38 3 6 25 1.E+04 101 70 131 302 33 23 43 291 49 51302 4 A 30 X 2 81 83 382.62 2 3 10 72 0 35 50 21 0 14 207 1 2 5 35 0 17 24 10 0 7 58 2.E+04 244 60 83 387 63 16 21 216 32 68302 4 A 30 X 2 101 103 382.82 6 4 1 121 5 50 27 3 0 5 222 3 2 0 55 2 23 12 1 0 2 81 0 70 20 47 137 51 15 34 256 65 35302 4 A 30 X 2 121 123 383.02 15 6 0 97 1 44 26 5 1 8 203 7 3 0 48 0 22 13 2 0 4 79 0 122 25 56 203 60 12 28 238 54 46302 4 A 30 X 2 141 143 383.22 17 0 0 73 0 35 3 81 0 2 211 8 0 0 35 0 17 1 38 0 1 56 0 129 22 17 168 77 13 10 231 58 42302 4 A 30 X 3 1 3 383.34 18 1 1 138 2 37 9 31 6 4 247 7 0 0 56 1 15 4 13 2 2 80 0 68 14 25 107 64 13 23 264 71 29302 4 A 30 X 3 21 23 383.54 30 0 0 132 1 19 4 49 2 2 239 13 0 0 55 0 8 2 21 1 1 74 0 74 8 13 95 78 8 14 269 74 26302 4 A 30 X 3 41 43 383.74 20 1 1 151 1 22 3 17 1 6 223 9 0 0 68 0 10 1 8 0 3 88 0 13 3 6 22 59 14 27 232 91 9302 4 A 30 X 3 61 63 383.94 52 0 0 51 0 54 3 55 0 10 225 23 0 0 23 0 24 1 24 0 4 64 0 92 12 3 107 86 11 3 251 70 30302 4 A 30 X 3 81 83 384.14 48 2 0 104 1 52 1 15 0 5 228 21 1 0 46 0 23 0 7 0 2 89 0 72 19 21 112 64 17 19 250 69 31302 4 A 30 X 3 101 103 384.34 26 2 0 83 0 98 6 10 0 5 230 11 1 0 36 0 43 3 4 0 2 91 0 51 9 8 68 75 13 12 244 78 22

302 4 A 31 X CC 10 12 385.80 49 3 0 143 1 30 13 16 0 2 257 19 1 0 56 0 12 5 6 0 1 84 0 71 15 97 183 39 8 53 290 61 39302 4 A 31 X CC 23 25 385.93 29 0 0 140 0 43 1 61 4 5 283 10 0 0 49 0 15 0 22 1 2 75 0 81 4 9 94 86 4 10 297 76 24302 4 A 31 X CC 40 42 386.10 47 1 0 156 0 29 2 11 0 1 247 19 0 0 63 0 12 1 4 0 0 93 0 40 0 4 44 91 0 9 256 85 15

302 4 A 32 X 1 1 3 388.01 24 5 2 92 0 7 10 0 0 5 145 17 3 2 63 0 5 7 0 0 3 85 0 40 34 150 224 18 15 67 175 44 56302 4 A 32 X 1 21 23 388.21 6 0 0 93 0 11 19 0 2 0 130 4 0 0 71 0 9 14 0 1 0 85 0 54 66 165 285 19 23 58 142 33 67302 4 A 32 X 1 41 43 388.41 82 0 0 43 0 4 7 0 2 0 137 60 0 0 31 0 3 5 0 1 0 87 0 3 9 4 16 20 60 20 144 89 11302 4 A 32 X 1 61 63 388.61 181 45 95 321 56 14 30 19 6 94302 4 A 32 X 1 81 83 388.81 0 0 1 84 0 71 19 30 0 10 215 0 0 0 39 0 33 9 14 0 5 76 1.E+04 96 87 78 261 37 33 30 348 57 43302 4 A 32 X 1 101 103 389.01 34 23 20 77 44 30 26 7 8 92302 4 A 32 X 1 121 123 389.21 43 29 19 91 47 32 21 2 2 98302 4 A 32 X 1 141 143 389.41 42 22 20 84 50 26 24 8 9 91302 4 A 32 X 2 1 3 389.51 49 27 32 108 45 25 30 20 16 84302 4 A 32 X 2 21 23 389.71 55 33 24 112 49 29 21 17 13 87302 4 A 32 X 2 41 43 389.91 0 0 0 55 0 105 27 13 1 5 206 0 0 0 27 0 51 13 6 0 2 80 7.E+03 260 154 228 642 40 24 36 295 31 69302 4 A 32 X 2 61 63 390.11 66 21 23 110 60 19 21 3 3 97302 4 A 32 X 2 81 83 390.31 0 0 0 71 0 65 44 20 0 4 204 0 0 0 35 0 32 22 10 0 2 68 2.E+04 323 233 324 880 37 26 37 340 28 72302 4 A 32 X 2 101 103 390.51 56 35 23 114 49 31 20 9 7 93302 4 A 32 X 2 121 123 390.71 22 25 50 97 23 26 52 22 18 82

Pol

ysph

aerid

ium

cpx

T able S -1. S ummarized palynological results across the P E T M of IODP Hole 4A. C ore 31X has been shifted down 1 meter relative to ref. S I-1 (see S upplementary Information text). G reen bars indicate intervals affected by drilling disturbance. Low salinity tolerant dinocyst percentages are normalised to the sum of all dinocysts minus the species with unknown salinity tolerance. T he latter group is 10% on average.