Contents lists available atScienceDirect
Earth and Planetary Science Letters
www.elsevier.com/locate/epsl
Astronomically paced changes in deep-water circulation in the western North Atlantic during the middle Eocene
Maximilian Vahlenkamp
a,∗, Igor Niezgodzki
b,c, David De Vleeschouwer
a, Torsten Bickert
a, Dustin Harper
d, Sandra Kirtland Turner
e, Gerrit Lohmann
b, Philip Sexton
f, James Zachos
d, Heiko Pälike
aaMARUM- CenterforMarineEnvironmentalSciences,UniversityofBremen,LeobenerStr.8,28359Bremen,Germany bAlfredWegenerInstitute- HelmholtzCentreforPolarandMarineResearch,Bussestr.24,27570Bremerhaven,Germany
cINGPAN- InstituteofGeologicalSciences,PolishAcademyofSciences,ResearchCenterinKraków,BiogeosystemModellingLaboratory,Kraków,Poland dDepartmentofEarth&PlanetaryScience,UniversityofCalifornia,SantaCruz,CA95064,USA
eDepartmentofEarthSciences,UniversityofCalifornia,Riverside,CA92521,USA
fSchoolofEnvironment,Earth&EcosystemSciences,TheOpenUniversity,MiltonKeynesMK76AA,UK
a r t i c l e i n f o a b s t ra c t
Articlehistory:
Received19July2017
Receivedinrevisedform5December2017 Accepted6December2017
Availableonlinexxxx Editor:M.Frank
Keywords:
NADW AMOC
astronomicalforcing middleEocene IODP
NorthAtlanticDeep Water(NADW)currentlyredistributesheatandsaltbetweenEarth’soceanbasins, andplaysavitalroleintheocean-atmosphereCO2exchange.Despiteitscrucialroleintoday’sclimate system,vigorousdebateremainsastowhendeep-waterformationintheNorthAtlanticstarted.Here,we presentdatasetsfromcarbonate-richmiddleEocenesedimentsfromtheNewfoundlandRidge,revealing a unique archive of paleoceanographic change from the progressively cooling climateof the middle Eocene.Well-definedlithologicalternationsbetweencalcareousoozeandclay-richintervalsoccuratthe
∼41-kyrbeatofaxialobliquity.Hence,weidentifyobliquityasthedriverofmiddleEocene(43.5–46 Ma) NorthernComponentWater(NCW,thepredecessorofmodernNADW)variability.High-resolutionbenthic foraminiferalδ18O andδ13C suggestthatobliquityminimacorrespondtocold,nutrient-depleted,western NorthAtlanticdeepwaters.We thuslink strongerNCWformationwithobliquity minima.Incontrast, during obliquity maxima, Deep WesternBoundary Currents wereweaker and warmer, while abyssal nutrientsweremore abundant.Theseaspects reflect amore sluggishNCWformation. Thisobliquity- pacedpaleoceanographicregimeisinexcellentagreementwithresultsfromanEarthsystemmodel,in whichobliquityminimaconfigurationsenhanceNCWformation.
©2017TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/4.0/).
1. Introduction
Modern North Atlantic Deep Water (NADW) production ac- countsfor ∼40 to 50% of AtlanticMeridional Overturning Circu- lation (AMOC) (Broecker, 1998). As an important part of global thermohalinecirculation,theAMOC helpsregulateglobalclimate inthree primaryways:i)throughthezonalandlatitudinalredis- tribution of heat, salt, and nutrients (Broecker and Peng, 1982), ii)via thecarbon cycle, by AMOC’sdominant role inmoderating oceanicCO2 uptake(Zickfeldetal., 2008) and iii)by itseffecton atmosphericcirculationthrough modulation ofglobalseasurface temperatures(SST)(Mulitzaetal.,2008).Accordingly,variationsin AMOCintensitycan causelarge-scaleperturbations to globaland regionalclimates.
*
Correspondingauthor.E-mailaddress:mvahlenkamp@marum.de(M. Vahlenkamp).
Oceancirculationhasevolvedinresponsetochangesinpaleo- geographic configurations over time. During the early Paleocene, theoceanicconnectionbetweentheGreenland-NorwegianSeaand theNorthAtlanticwas notyetestablishedandtheAtlanticwasa much narrower elongated basin with extended adjacent shallow shelf areas(Scoteseetal., 1988). Overturningprincipallyoccurred in theSouthern Ocean during thelate Paleocene to early Eocene (PakandMiller,1992; Thomasetal.,2003),withpossiblecontribu- tionofdeep-watersourcesintheNorthPacific(Thomas,2004) and warmsalinedeepwateroriginatingintheTethys(ScherandMar- tin, 2004). Extremely highdeep ocean temperatures(up to 12◦C higher than modern) (Cramer et al., 2011; Sexton et al., 2006a;
Zachos et al., 2001) and elevated atmospheric CO2 (Anagnostou et al., 2016) were associated with decreased latitudinal temper- ature gradients (Tripati and Elderfield, 2005) and an enhanced hydrological cycle (Barron et al., 1989) that freshened the sur- face ocean at high latitudes. During the early Eocene, rifting in the Greenland-Norwegian Sea (Mosar et al., 2002) created the https://doi.org/10.1016/j.epsl.2017.12.016
0012-821X/©2017TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/4.0/).
the late Eocene to early Oligocene when the Drake andTasman Passages began to open (Cramer et al., 2009; Katz et al., 2011;
Scher and Martin, 2008). Davies etal. (2001) proposed an onset ofNCW at35Mabasedontheidentificationanddatingoftheat thistimeoldestknownNorthAtlanticDriftSediments.Ndisotopes fromthe SouthAtlantic andthe Southern Oceanhave later been usedtoverifythisage(ViaandThomas,2006).However,theSouth Atlanticwaslessradiogenic(ascharacteristicofNCW)duringthe middleEocenethanduringtheOligocene(ScherandMartin,2008;
ViaandThomas,2006).Hohbeinetal. (2012)pushedbackthedate oftheonsetofNCWbasedontheonsetofsedimentdriftdeposits within a restricted sedimentary basin at the Greenland–Scotland Ridge, a key gateway for modern NADW outflow into the North Atlantic. This age close to the early-middle Eocene boundary is supported by the onsetof the Newfoundland Drifts(Boyle etal., 2017) andwinnowingatBlakeNoseasindicatedbythedeposition offoraminiferalsandswhichhavebeendepositedacrosstheearly–
middle Eocene boundary disconformity (Norris et al., 2001). The proposed timing offirstNCW atthe early-middleEocenebound- arycoincides withtheinvigorationofbottomcurrentsinferredby large scale erosion in the North Atlantic (Berggren andHollister, 1974), theonset oftheCenozoicglobaldeep-water coolingtrend (Zachosetal.,2001), majorchangesindeep-seacirculationasev- ident through changes in the global inter-basinal δ13C gradient (Sexton et al., 2006a) warming of the Atlantic relative to Pacific bottomwaters(Cramer etal.,2009) andenhancedglobalproduc- tivity(Nielsenetal.,2009).
Most attempts to characterize the oceanic response to astro- nomicalforcing undergreenhouseconditionsduring theCenozoic havefocusedontheearlyEocenegreenhouse,beforetheonsetof NCW formation (e.g. Lunt et al., 2011; Sloan and Huber, 2001).
Here, we particularly focuson the response to changes in obliq- uity(i.e.thetiltofEarth’srotationalaxis)aftertheonsetofNCW (Boyleetal.,2017; Hohbeinetal.,2012).
Theeffectofobliquityislargestathighlatitudes,wherethecli- materesponsetoobliquityforcingisenhancedbyvariousfeedback mechanisms(Mantsis etal., 2011). Severalmechanisms including atmosphericandoceancirculationareknowntotransferhighlat- itudeinsolationforcingintomidorlowlatitudinalclimatesignals (LiuandHerbert,2004).DuringthePleistocene,overturninginthe NorthAtlanticwashamperedbyfreshwaterreleasefromicesheet collapseduringintervalswithunusuallywarmsummerscausedby maxima in obliquity (Sigman et al., 2007). Furthermore, in pre- PleistocenetimeswhenNorthernHemisphereicesheetswereab- sent orsmaller,cooler globaltemperaturesare usuallyassociated withminimain obliquity (De Vleeschouwer etal., 2017; Hays et al.,1976) facilitatingseaiceformationandthusstimulatingNCW formation.ThisisconsistentwithstrongNADW formationduring earlyPliocene(4.7–4.2 Ma)obliquityminima(Billupsetal.,1997).
Intheabsenceofcontinentalicesheets,obliquitycanregulatethe
Fig. 1.PaleoclimatesettingofthemiddleEocene(45Ma).a)Globaltopographyand bathymetryusedformiddleEocenemodelsimulationsinthisstudy(modifiedafter Luntetal.,2016).b)ThemiddleEoceneischaracterizedbythelong-termcooling afterthewarmestclimatesoftheCenozoic,asindicatedbyaglobalcompilationof Cenozoicδ18Obenthicrecords(Crameretal.,2009).
strength ofthethermohalinecirculationviaits influenceontem- peratureanddensitygradients,thehydrologicalcycle,theextentof seaice, andpossiblyvia thecarboncycle(Kuhlbrodt etal., 2009;
Rahmstorf,1995).
SedimentsatSiteU1410,drilledduringIODPExpedition342at theNewfoundlandRidge(Norrisetal.,2014) provideauniqueop- portunity to study middle Eocene deep ocean circulation in the western North Atlantic on orbital timescales. Here, we use the X-ray fluorescence (XRF) derived ratio of Ca/Fein bulk sediment and benthic foraminiferal stable carbon and oxygen isotopes to constrain the intensity of the Deep Western Boundary Currents aswell as deep-water nutrient availabilityand paleotemperature.
Together, these allow us to infer orbital-scale variations in deep ocean circulationinthewestern NorthAtlanticduringan interval ofmiddleEocenecooling.Weestablishthedeep-waterresponseto astronomical parameters and complement our observations with simulationsusingthecoupledEarthsystemmodelCOSMOS.
2. Materialandmethods
Site U1410 (41◦19.6993N, 49◦10.1847W; ∼3387.5 m water depth) (Fig. 1) was drilled during Integrated Ocean Drilling Pro- gram (IODP)Expedition342–PaleogeneNewfoundlandSediment Drifts (Norris et al., 2014). Middle Eocene sedimentation at this siteoccurred ata paleodepth of∼2950 m at∼50 Ma(Norris et al., 2012; TucholkeandVogt,1979).TherecoveredmiddleEocene sedimentsholdarecordofcarbonate-rich,cyclicsequences.More- over,relativelyhighsedimentationrates(2–4 cm/kyr)(Norrisetal., 2014) characterizethesesediments,allowingforahigh-resolution reconstructionofoceanographicandclimaticchange.
2.1. XRFdata
X-rayfluorescencemeasurementsinthestudiedintervalswere carriedoutontheAvaatechXRFCoreScanneratScrippsInstitution ofOceanographyGeological Collections,U.C.SanDiego.Elemental
intensitieswerecollectedevery2cm(ca.1kyr/sample)down-core over a 1.2cm2 area with a down-coreslit size of10 mm using generatorsettingsof10kVandacurrentof0.1mA.Thesplitsur- faceofthearchivehalfwas coveredwitha4 μm thickSPEXCerti PrepUltralene1foilto avoidcontamination oftheXRF measure- mentunit anddesiccation ofthesediment.The samplingtimeat thesplitcoresurfacewas20seconds.
2.2.Isotopesamplingstrategyandchronology
Forbenthic foraminifer stableisotopes, we sampled the sedi- mentaryrecordofSiteU1410at4cmintervals(ca.2kyr/sample).
Somesampleswere takenfromsectionsthatarenotincorporated inthe latest version ofthe splice, asthe initial shipboard splice was revisedbasedon onshore-acquiredXRF data(Supplementary Tables1and2).WealignedallcoresofSite U1410onacentime- terscaletothe depthscale ofthisrevisedsplice(Supplementary Fig. 1andSupplementaryTable3).Allisotopedataareplacedona common depth scale, and are directly comparable to other geo- chemical measurements presented here, though not consistently obtainedfromthesamehole.
Weupdated theshipboardagemodel forSite U1410 byusing theGTS2012AstronomicAgeModel(Gradstein etal., 2012) ages ofchronboundaries:42.351MaforC19r/20n(131.76mcdinHole U1410A);43.505MaforC20n/C20r(163.86mcdinHoleU1410A);
45.942MaforC20r/C21n(218.40mcdinHoleU1410A);47.837Ma forC21n/C21r (252.53mcdinHoleU1410A) (Norris etal.,2014).
Inbetweenthesechronboundaries,weuselinearinterpolationto transferthe XRF andstableisotope depth-series intotime-series.
BenthicstableisotopedataaregiveninSupplementaryTable4.
2.3.Analyticalmethodsforstableisotopesofbenthicforaminifera
Sedimentsampleswere soaked insodiummetaphosphate(ac- cordingtoa cleaning protocolforEoceneExp. 342samples; Hull etal., 2017) andwashedthrougha150 μmsieve.Basedonavail- ability, 5–15 well preserved, ‘glassy’ (sensu Sexton et al., 2006b) specimens of the benthic foraminifera Nuttallidestruempyi with- outanyvisible infillingswere picked fromthe>150 μmfraction inordertogeneratestablecarbonandoxygenisotopedata.These
‘glassy’foraminiferacontrastsharplywiththe‘frosty’preservation (Sextonetal.,2006b) thatistypicaloffossilforaminiferafromthe vast majority of deep-sea sedimentary sequences. Samples were measured on a Finnigan MAT 251 gas isotope ratio mass spec- trometerconnectedtoa KielIII automated carbonatepreparation deviceat MARUM,Bremen.Data are reportedinthe usual delta- notationversus V-PDB.Theinstrumentwas calibratedagainst the in-house standard (ground Solnhofen limestone), which in turn was calibrated against the NBS-19 standard reference material.
Over themeasurement period the standard deviations of thein- housestandardwere0.03h forδ13C and0.07hforδ18O.Benthic foraminiferalcarbonandoxygenisotopesfromNuttallidestruempyi were corrected to Cibicidoides usingwell established interspecies correctionfactors(Katz etal., 2003).Weplotanddiscussall ben- thicforaminiferalisotopedatacorrectedtoCibicidoides.
2.4.Time-seriesanalysis
Time-series analysis was carried out using the REDFIT algo- rithm(Schulzand Mudelsee,2002) forunevenly spaced time se- riesasimplementedin thePASTsoftware(Hammeretal., 2001).
The analyses were performedin thetime-domain with a Welch- windowandwithoutoversamplingorsegmentation.Priortospec- tral analyses a Lowess smoother (smoothing factor = 0.66 equal to∼1.7Myr) was applied toall data seriesto remove long-term trends. Bandpass filters isolate and extract the components of
signals associated with a specific range of frequencies. We em- ployedband-passfilterstoassessthebehaviour ofaspecificrange of frequencies in a studied signal using the R astrochron pack- age (Meyers, 2014). The periodicity range for bandpass filtering of obliquity cycles is between 29 and 55 kyr. We chose such a relativelybroadrangeinordertocapturethemajorobliquitycom- ponents at41kyr,54kyr and29kyr(Laskar etal.,2004),which is crucial for a correct representation of amplitude modulation patterns. We obtained the amplitude envelope of the obliquity- filteredsignal by theapplicationoftheHilbertTransform, asim- plementedinastrochron.Alowpassfilterat130kyrwasappliedto theobliquity amplitudemodulationtocapturethemainobliquity amplitude modulationperiods.Thefrequencyrangeforthe direct 173-kyrfilterisbetween130and220kyr.Phaseestimatesandco- herences werecalculatedwiththe Blackman-Tukeymethod using theAnalyseriessoftware(Paillardetal.,1996).
2.5. Earthsystemmodel
Eocene climate simulations were performed using the Earth system modelCOSMOS,in the coupledatmosphere-ocean config- uration with prescribed vegetation. The model configuration in- cludestheatmospherecomponentECHAM5atT31/L19resolution, i.e. a horizontal resolution of ∼3.75◦ with nineteen vertical lay- ers(Roeckneretal.,2006).TheMaxPlanckInstitute OceanModel (MPI-OM) runs in a GR30/L40 configuration, i.e. has an average horizontal resolution of 3◦ × 1.8◦ with 40 uneven vertical lay- ers(Marslandetal.,2003).MPI-OM includesthedynamicsofsea iceformulatedusingviscous-plasticrheology(HiblerIII,1979).Our versionofCOSMOShasbeenextensivelyusedandvalidatedinthe context ofglacial (Werner etal., 2016; Zhang etal., 2014), inter- glacial (Lohmann et al., 2013; Pfeiffer and Lohmann, 2016), and Miocene(Knorretal.,2011; KnorrandLohmann,2014) climates.
Inordertomodeldeep-watercirculationatandaftertheearly- middleEoceneboundary, we employedan Ypresian (56–47.8 Ma) paleogeographycompiledbyGETECH(Luntetal.,2016) andmod- ified the paleogeography in the North Atlantic to account for the tectonic changes at the early-middle Eocene boundary. We havemodified theoriginal paleogeography by closing theGibral- tar Straitsfollowing the reconstruction by Stampfli et al. (2002).
AdeepconnectionbetweentheGreenland-NorwegianSeaandthe NorthAtlanticviathedeepestpartoftheGreenland-ScotlandRift was establishedby atleast∼48.5Ma(Hohbeinetal., 2012).The ArcticOcean wasan isolated,poorly ventilated basin,comparable to the modern Black Sea (Stein et al., 2006) and theFram Strait onlyconnectedtheGreenland-NorwegianSeaandtheArcticfrom the early Mioceneonwards (Jakobssonetal., 2007). Forthis rea- son, we employ a paleogeography witha closed Fram Strait and aGreenland-NorwegianSeagatewayof500 mwaterdepthinour obliquityminimum/maximumexperiments.
Based on this paleogeography, the boundary conditions to run the modelinclude theprescription ofvegetation distribution (Sewalletal.,2007),theset-upofthehydrologicaldischargemodel (HagemannandDümenil, 1997),orography-relatedparametersfor the gravity wave drag parameterization (Lott and Miller, 1997), glaciermask,the concentrationofgreenhousegases intheatmo- sphere,andorbitalparameters.Thesolarconstantwasreducedby 0.6% compared to present-day andis equalto 1358.8 W/m2.Ec- centricityandprecession weresettopresentdayvalues.TheCO2 levelwas setto3xpre-industriallevel(840ppm)(Anagnostou et al., 2016), whileother greenhouse gases were set to presentday values.Weprescribednoicesheetsineitherhemisphere.
We set up two experiments, with different obliquity angles, keeping eccentricityand precession constant. In the first experi- ment(OBLMIN),obliquityissettoaminimumvalueof22.1◦,while inthe second (OBLMAX) to amaximum value of24.5◦. Both sim-
Fig. 2.Obliquity-pacedpaleoceanographicproxyrecordsfromIODPSiteU1410.(a)XRF-derivedCa/Feindicatetheonsetofdriftsedimentationat244mcd.(b–d)Ca/Fe, δ13Cbenthicandδ18ObenthicrecordsofchronC20rinthetimedomain.WeconsiderC20rboundaryagesof43.51and45.94Maandapplylinearinterpolationinbetweenthese tie-points,sotoconvertdepthintotime.BlueandRedlinesontheisotopiccurvesrepresentthe6-kyrrunningaverages.(Forinterpretationofthereferencestocolourin thisfigurelegend,thereaderisreferredtothewebversionofthisarticle.)
ulations were integrated for 7000 model years to reach steady state. The analysisis basedon a climatology(long-term average) calculatedfromthelast100yearsofsimulations.Werunanaddi- tionalminimumobliquityexperimentwithanadjustedGreenland- ScotlandRidgesilldepth of200 mtotest thesensitivityofNCW formationtothisbathymetricfeature.Thesimulationwasrestarted fromexperimentOBLMINatamodelyear6200andwasfurtherrun for800yearstothemodelyear7000.
3. Results
3.1. Paleoceanographicproxiesandtheirastronomicalinterpretation
Cyclic drift deposits replace continuous pelagic sediment at 244 mcdwithinthe deeperthirdofchronC21n.The investigated sedimentsection (∼158–219 mcd)inchronC20rcontains middle Eocene sediment drift deposits composed of light grayish green nannofossil clay interlayered with nannofossil ooze. These well- developed alternations between nannofossil clay and nannofossil ooze reflect the primary lithological cycles in carbonate content.
Carbonatecontentvariesbetween∼30wt%inthe darkernanno- fossilclayand80wt%inthenannofossilooze(Boyleetal.,2017;
Norris et al., 2014). The alternation between both lithologies is wellcapturedby theXRF-derivedratiobetweencalciumandiron (Ca/Fe),withhighCa/Feratiosinthenannofossiloozeandlowra- tiosin theclay-rich intervals. Theonset ofthedrift sedimentsis markedby a sharpdecrease inthemean Ca/Feratioat244mcd (Fig. 2a).
3.1.1. XRF-derivedCa/Fe
The Ca/Fe ratioin chron C20r is characterized by low ampli- tude cyclesoflessthan10counts/countsintheolderpartofthe chron(46–44.1 Ma)andhighamplitudeofupto80counts/counts in its younger part (44.1–43.5 Ma). The power spectrum of the log(Ca/Fe)time-seriessuggestobliquityasthemaindriverofCa/Fe variance, asit exhibitsa doublespectral peak exceeding the99%
confidence level (CL) at obliquity-related periods of 29–55 kyrs (between0.018–0.034cycles/kyr;Fig. 3a).
Another strongspectral peak emerges ata period of18.7 kyr (0.053 cycles/kyr), very closeto the frequency where one would expecttheimprintofthe∼20-kyrprecessioncycle.However,this spectral peak ismore likely to be the first harmonicof the fun- damental frequencyoftheobliquitycycles.Indeed,thelog(Ca/Fe) cycles are characterized by a non-sinusoidal periodic waveform (SupplementaryFig. 2a). UsingFourierexpansion,suchasignal is representedbythesumofaninfiniteseriesofall-integerharmonic frequencies. This means that a power spectrum of such a signal willhaveaharmonicpeakatallmultiplesofthefundamentalfre- quency, inthiscaseobliquity (SupplementaryFig. 2b). Hence, we interpret the observed peak at 18.7 kyr asthe first harmonic of obliquity,rather thanasthe imprintofprecession. Apparent am- plitude modulations of obliquity with a frequency of ∼173 kyr result from the resonance or “beat” between the (present-day) 41-kyr and53-kyrobliquity cycles,forwhichthe frequencies are determined by the precession constant andthe frequency of the ascending node precession of respectively the Earth–Moon body (p+s3)andSaturn(p+s6)(Laskaretal.,2004).
Fig. 3.Spectral andphaseanalysisofproxyrecords.Redfitpowerspectraofthe a)Ca/Fe,b)δ13Cbenthic,andc)δ18Obenthic.Mainperiodsarehighlighted.d)Phase wheelillustratingthephaserelationsbetweenbenthicforaminiferalstableisotopes andCa/Feattheobliquityband(41kyr).Inthephasewheelrepresentation,vec- torsinthe12o’clockpositionareinphasewithmaximumCa/Fe,andphaselags increaseintheclockwisedirection(forexample,3o’clockrepresentsa90◦lagrela- tivetoCa/Fe,6o’clockrepresentsanantiphaseresponse,and9o’clockrepresentsa 90◦lead).Vectorlength(fromcirclecentretomiddleofarc)representscoherence, andtheassociatedarcdenotesthephasewithin2σ phaseerror.Circlesmark99%
(dashed-dotted),95%(dashed)and90%(dotted)coherence.
3.1.2. Benthicforaminiferalcarbonisotopes
Ourbenthicforaminiferalisotoperecordconsistsof1185mea- surementsbetween43.35–46 Ma.Benthiccarbonisotopesinchron C20rrangebetween−0.04hand0.91h.Peaksexceedingthe95%
confidencelevelinthepowerspectrumofδ13C appearatfrequen- ciesrelated to shorteccentricity (∼0.086 cycles/kyr)and preces- sion(0.058cycles/kyr), while apeak closeto the95% confidence leveloccursatthefrequencyofobliquity (43kyr)(Fig. 3b).Other significant peaks at frequencies of 1.29 Myr (0.00078cycles/kyr) and173 kyr(0.0058cycles/kyr)are atthe timescaleof thelong- termresponseofthecarboncycletothe1.2Myrandthe173-kyr obliquityamplitudemodulation,whileanothersignificantpeakap- pearsat a frequency of 800kyr. The carbon isotope record thus seems to be characterized by short and long-term orbital vari- abilitysuperimposed on a low amplitude increase of0.02h/Myr throughoutchronC20r.
3.1.3. Benthicforaminiferaloxygenisotopes
Benthic foraminiferal oxygen isotope values range between
−0.51h and 0.62h and follow the previously-observed gen- eralcooling trendinbenthic foraminiferaloxygen stableisotopes throughoutthemiddleEocene(Crameretal.,2009).Inourrecord, thisincrease amounts to0.11h/Myr throughoutchronC20r. Sig- nificant peaks in the power spectrum of benthic δ18O occur at 42 kyr(0.024cycles/kyr)andat∼330kyr(0.003cycles/kyr),close to the periods of obliquity (41 kyr) andpossibly related to long eccentricity(405kyr)(Fig. 3c).
3.1.4. Proxyphaserelationship
Thefitbetweenobservedfrequenciesandexpectedorbitalfre- quencies in all three proxy records is remarkablegiven the fact that the agemodel is basedon linear interpolation betweenthe chronboundariesratherthanonastronomicaltuningandthuspre- ventscircularreasoningbyintroducingpowerintoexpectedorbital
frequencies. The records of benthic foraminiferal δ13C and δ18O areinphase(i.e.positivelycorrelated).Minimainδ13C correspond to minima in δ18O and vice versa. On the other hand, benthic foraminiferal isotopes are in antiphase(i.e. negatively correlated) with log (Ca/Fe). Minima in δ13C and δ18O lead maxima in log (Ca/Fe)by ∼4.5kyrand∼2.5kyr, andviceversa. Thecoherence betweenδ13C andCa/Feexceedsthe 99%CLandthus indicatesa verystablephasedifferencebetweenthetwoproxyrecords.
3.2. Earthsystemmodel
Figs. 4aand4d displayglobalSSTs,assimulated bythe Earth system model, under minimum and maximum obliquity condi- tions,respectively.Onaverage,SSTsare1.5◦Ccoolerduringobliq- uity minima relative to obliquity maxima (Fig. 4g). The aver- agedeep-watertemperaturedifferencebetweenbothastronomical configurations is 0.2◦C, with the obliquity minimum simulation beingcoolerthantheobliquitymaximumsimulation.Theamount ofrelativecoolinggraduallyincreasestowardsthepoles;theArctic is characterized by a 5.9◦C difference betweenboth simulations.
The latitudinal SST gradient increases by 5.5◦C during obliquity minima compared to obliquity maxima, a feature seen in other models(Mantsisetal.,2011).
Deep-watercoolingofupto1.3◦Coccursduringobliquitymin- imaintheArcticOcean,theNorthAtlanticandtheNordicSea.The modelpredictsthatSiteU1410wasexposedto∼1.2◦Ccoolerbot- tom watersduring middle Eoceneobliquity minima comparedto obliquitymaxima(Fig. 5).
Average global sea surface salinities (SSS) are 0.4 psu higher during obliquity minima compared to obliquity maxima. Differ- encesaremostpronounced intheTethys,wheresurfacesalinities increase by more than 2 psu during obliquity minima (Figs. 4b, 4e and4h). At thesame time, SSSdecreases by ∼0.5psuin the LabradorSea,whileSSSintheGreenland-NorwegianSeaisbarely affectedbyvariationsinobliquity.
The simulated deep-water formation is strongly seasonal and occurs primarily in the Weddell Sea and RossSea during boreal summer and in the Eastern North Atlantic and the Greenland- Norwegian Seaduringboreal winter.Mixed-layerdepthis deeper during obliquity minima in boreal winter in the Greenland- Norwegian Sea(Figs.4c,4f and4i).Deep Western BoundaryCur- rentsatthepaleodepthofSiteU1410(∼2950 m)arefasterduring obliquityminima,whileduringobliquitymaximastrongercurrents occurinintermediatewaterdepth∼1500 m.
Our simulations exhibit deep-water formation and a Deep Western BoundaryCurrent intheNorthAtlantic inallthree sim- ulations (Fig. 6), indicated also in the global meridional ocean circulation.Enhancedoverturning withdepthdown to4200 m is detected for minimum obliquity, whereas a reduced overturning withdepthsdown to3500 mis detectedformaximumobliquity, andforthemodelsetupwithminimumobliquityandassuminga 200 msilldepthoftheGreenland-ScotlandRidge(Fig. 6).
The simulation with the shallower Greenland-Scotland Ridge silldepthshowsthattheexactgeometryofNCWandSCWissus- ceptibletochangesintheoverflowcapacityoutoftheGreenland- NorwegianSea.Thedeep-waterformationisweakerandshallower in the shallow sill simulation compared to simulations with a 500 mdeepGreenland-Scotland Ridge, whiletheSouthern Ocean componentisstrengthened.
4. Discussion
In thissection, we will use theresults fromour paleoclimate simulation experiments to illustrate the mechanism involved in the invigoration of NCW formation in response to orbital forc- ing. It should be noted, however, that the exact amplitude of
Fig. 4.BorealwinterpaleoceanographyassimulatedbytheCOSMOSmodel.SeasonallyaveragedSeaSurfaceTemperature(SST),SeaSurfaceSalinity(SSS)andmixedlayer depthforborealwinterunder(a–c)obliquityminimumconfiguration,andunder(d–f)obliquitymaximumconfiguration.(g–i)SeasonaldifferencesinSST,SSSandmixed layerdepthbetweentheobliquityminimumandobliquitymaximumsimulations.
Fig. 5.Annualdifferenceintemperatureat2785 mwaterdepthbetweentheobliq- uityminimumandtheobliquitymaximumsimulation.Thedepthof2785 mcorre- spondstotheapproximatepaleodepthofSiteU1410duringthemiddleEocene.
astronomically-forcedchangedependsondifferentboundary con- ditions (e.g. paleobathymetry, atmospheric CO2 level, vegetation distribution,continentalicevolume).
4.1. PaleoceanographicchangeintheWesternNorthAtlantic
Proxyrecords atSite U1410in theWestern NorthAtlanticoff Newfoundlandreveal astrongimprint ofobliquity.However, dur- ingthe middleEocene,Site U1410was locatedatapaleolatitude of∼41◦N,where83%ofthevariabilityinincomingsummerinso- lation(June21)isascribedtoprecessionandonly17%toobliquity (relative amplitude of obliquity compared to precession at 41◦N over the past 1 Ma from Laskar et al., 2004). In that respect, it isremarkablethatthelog(Ca/Fe)andstableisotopeproxyrecords
exhibitanastronomicalimprintthatisdominatedbyobliquity.The most likely explanation for this observation is an oceanographic teleconnectionbetweenthepaleolocation ofSite U1410andhigh latitudes,whereobliquityhasitsmaximuminfluenceoninsolation variability.
A mechanismto enable such an oceanographicteleconnection istheinvigorationofDeepWesternBoundaryCurrentsinresponse totheonsetofNCWformation,whichhaspreviouslybeenlinked tothedepositionofcontouritedriftsedimentsattheearly-middle Eocene boundary across the North Atlantic (Boyle et al., 2017;
Hohbeinetal.,2012).AtSite U1410,thischangeinbottomwater circulationismarkedbythesharpdecreaseinCa/Feat∼47 Maco- incidingwitha lithologychange frombiogenic pelagicsediments tomuddydriftsediments.Severallinesofevidenceimplythatthe lithologicalvariabilityincarbonatecontent isrelatedtotheinput of clay: (1)Middle Eocene carbonate deposition between ∼3300 and∼4300 mpaleo-waterdepthatIODPExpedition342sitesin- dicatea deepNorthAtlanticcarbonatecompensationdepth(CCD) atthistime(Norrisetal.,2014),withtheCCDlyingwellbelowthe paleodepthofSiteU1410.Theexcellentpreservationofcalcareous microfossilsatSite U1410 furtherstrengthenstheseobservations.
(2) An order-of-magnitude increase in bulk sediment mass accu- mulation rates, due primarily to an increase in the terrigenous component (Boyle et al., 2017). (3) Increases in the sedimenta- tion rate at Site U1410 with the onset of drift sediments have the same proportional magnitude asdecreases inCaCO3 content at thesame time (Norris et al., 2014). These argumentstogether excludechangesinproductivityastheprimary driveroftheCa/Fe signal. In contouritedeposits, winnowing andsortingof particles ofdifferentgrain-sizeandspecificweightshouldhaveasignificant influencethegrainsizeandelementalcomposition.Atflowspeeds below 10–15 cm s−1 common for the deposition of muddy con- tourites, sediment composition is controlled by selective deposi- tion,whereaswinnowinganderosiononlyplayaroleatfastercur-
Fig. 6.Deepwesternboundarycurrentvelocityandglobalmeridionaloverturningcirculation.Annuallyaveragedcurrentvelocitiesduring(a)theobliquityminimumand(b) theobliquitymaximumsimulation.(c)AnnuallyaveragedcurrentvelocitiesunderaminimumobliquityconfigurationandforaGreenland-ScotlandRidge200 msilldepth (500 minnominalsimulationsinpanels(a–b)).Positivevalues(red)indicateclockwiseoverturningcells;negativevalues(blue)indicatecounter-clockwiseoverturningcells.
Thecontourspacingis4Sverdrup.(Forinterpretationofthereferencestocolourinthisfigurelegend,thereaderisreferredtothewebversionofthisarticle.)
Fig. 7.The173-kyrcycleintheamplitudemodulationofobliquity.Theamplitudeenvelope(lightbrown)oftheobliquityband-passfilteredsignal(brown)generallyexhibits anin-phaserelationwiththedirect173-kyrband-passfilteredsignal,asextractedfromthe(a)Ca/Fe(black/grey),(b)δ13Cbenthic(blue),(c)δ18Obenthic(red).Phasewheel (d)illustratingthephaserelationsbetweenbenthicforaminiferalstableoxygenisotopes(red)andCa/Fe(black)attheobliquityamplitudemodulationband(173kyr).Inthe phasewheelrepresentation,vectorsinthe12o’clockpositionareinphasewithmaximumCa/Fe,andphaselagsincreaseintheclockwisedirection(forexample,3o’clock representsa90◦lagrelativetoCa/Fe,6o’clockrepresentsanantiphaseresponse,and9o’clockrepresentsa90◦lead).Vectorlength(fromcirclecentretomiddleofarc) representscoherence,andtheassociatedarcdenotesthephasewithin2σ phaseerror.(Forinterpretationofthereferencestocolourinthisfigurelegend,thereaderis referredtothewebversionofthisarticle.)
rentvelocities(McCaveandHall,2006).Theincreaseinterrigenous accumulation andthe drop in the ratio of Ca/Fe at the onset of thedriftsindicatesthatcurrentvelocitiesweresufficienttotrans- port clay from sources along the northeastCanadian continental margin southward to the topographically isolated Newfoundland ridgesbutinsufficientforwinnowing(Boyleetal.,2017).Forthis reason, we attribute the terrigenous component of the sediment reflected by the ratio of Ca/Feas a proxy for the Deep Western BoundaryCurrentvelocity.Consequently,lowCa/Fecorrespondsto strongDeepWesternBoundaryCurrentsandviceversa.
Benthic foraminiferal carbonisotope ratios are primarily con- trolled by deep-water nutrient availability. Modern NADW is a young deep-water mass formed by convection of surface waters predominantly sourced from low latitude areas marked by low nutrient andhigh chlorophyll where intense phytoplankton pro- ductivity effectively strips mostof the available nutrients out of surface waters. As such, NADW is relatively nutrient-depleted.
Plankton preferentially extracts 12C from surface waters. Young deep-water masses (i.e. close to their source region) forming in regions supplied by surface waters sourcedfrom low latitudear- eas marked by low nutrients and high chlorophyll are thus en- riched in 13C and can be distinguished by their relatively high δ13C values (Kroopnick, 1985). Alternatively, the changes in ben- thic foraminiferalcarbon isotopes could reflect globalchanges in δ13CDIC due tothe growth anddecay ofthe terrigenous and ge- ologicalcarbonreservoirs (e.g. terrestrialbiosphere, gashydrates, volcanic outgassing, organic matter accumulation) on obliquity timescales. The phase relation between our proxies (Fig. 3d)re- veals that the invigoration of Deep Western Boundary Currents (Ca/Fe minima) corresponds to a greater influence of nutrient- depletedbottom watersand/or thegrowthof 13C-depleted reser- voirs (δ13C maxima) and cold bottom water temperatures (δ18O maxima)atSiteU1410 onobliquitytimescales.Thestrongcoher- encebetweenCa/Feandδ13C (Fig. 3)supportsa commonmecha- nismdrivingbothproxies,relatedtoNCWvariabilityinthewest- ernNorthAtlanticattherhythmofobliquity.
WeproposethatepisodesofstrongNCWformationcorrespond to obliquity minima based on the following arguments: (1) The phase relationship between obliquity amplitude modulation pat- ternsandadirect173-kyrfilterexhibita positivephase relation- ship forCa/Fe (Fig. 7) andnegative phase relationships forδ18O.
Thesephaserelationships indicatethatincreasedCa/Feratioscor- respondtohighvariabilityinCa/Fe.Thus,we canlinkhighCa/Fe toobliquitymaxima.Inthebenthicforaminiferalδ18O record,the direct 173-kyr filter andthe obliquity amplitude modulation are in antiphase, linking high obliquity-variability in δ18O (during a 173-kyrobliquitymaximum)withlow δ18O.Inother words,min- imainbenthicforaminiferalstableoxygenisotopescorrespond to maxima in obliquity. However, the phase analysis between the 173-kyr δ13C filter andthe δ13C obliquityenvelope didnot yield resultsthatwerestatisticallysignificant.Nevertheless,weinferthe sameobliquityphaserelationshipforδ13C asforδ18O,asbothare in phasewitheach other (andinantiphasewithCa/Fe).(2) δ18O maxima inthe deep oceanare generally linked toglobal cooling in response to obliquity minima (De Vleeschouwer et al., 2017;
Haysetal.,1976).(3)OursimulationsinCOSMOS showthesame proportional cooling andinvigoration ofDeep Western Boundary Currentsduring obliquityminimathatweseeinourdata,provid- ing independentsupportforthephasing ofourdatawithrespect toaxialobliquity.
Theamplitudeofbenthicforaminiferaloxygenisotopevariabil- ityatthe41kyrrhythmofobliquityrangesbetween0.2and0.4h. UsingtheequationfromBemisetal. (1998),thisvariabilitycorre- spondstodeep-watertemperaturechangesbetween0.8and1.6◦C over the course ofa middle Eoceneobliquity cycle. Thesevalues areinexcellentagreementwiththeresultsfromourclimatesim- ulations, which exhibit a 1.2◦C difference betweenthe obliquity minimum andobliquity maximumsimulation inthe deepnorth- westernAtlantic(Fig. 5).
In accordance with our paleoceanographic proxy data, our oceancirculationsimulationsindicatethatDeepWesternBoundary Currents were invigorated during obliquity minima, transporting moreclaytotheNewfoundlandRidgethanduring obliquitymax- ima (Fig. 6). These simulations also predict a larger amount of nutrient-depleted NCW reachedthe Newfoundland region during obliquityminima(Fig. 6a, b).
4.2. SourceandformationofmiddleEocenenortherncomponentwater
The modernAMOCischaracterizedbyoverturning cellsinthe SouthernOcean(AntarcticBottomWater)andintheNorthAtlantic (NADW).ThetwomajorcomponentsofmodernNADWareNordic
Fig. 8.Globalbenthicforaminiferalisotopecomposite(black)plottedagainstδ13C (blue)andδ18O (red)fromsiteU1410.U1410N.truempyidataarecorrectedtoCibicidoides usingcorrectionfactorsforδ13C (Nut+0.34h)andδ18O ((Nut+0.10)/0.89)fromKatzetal. (2003).(Forinterpretationofthereferencestocolourinthisfigurelegend,the readerisreferredtothewebversionofthisarticle.)
SeaOverflow WaterandLabradorSeaWater.NordicSeaOverflow Wateris the densest partof NADW andis formed through con- vectionofhighlysalinesurfacewaterintheGreenland-Norwegian Sea induced by cooling. Nordic Sea Overflow Water is fed back into the North Atlantic by overflow over the Greenland-Scotland Ridge and forms the lower part of the Deep Western Boundary Current.LabradorSeaWater,ontheotherhand,islessdensethan NordicSeaOverflowWaterandformstheupperpartoftheWest- ernBoundaryCurrentafterconvectionintheLabradorSea(Hansen andØsterhus,2000).Ourproxydatadonotallowustodistinguish betweendeepwaters convectedin theGreenland-Norwegian Sea versus Labrador Seaduring the middle Eocene.Yet, several lines ofevidencesuggesttheGreenland-NorwegianSeaasthe(primary) sourceofNCWinthemiddleEocene:First,thepaleodepthofSite U1410 is ∼3000 m, much deeper than the convection depth of modernLabradorSeaWater(Pickartetal.,2002).Second,theonset of drift sedimentation at the Judd Falls Drift on the Greenland- Scotland Ridge suggestsdeep-water formation in the Greenland- NorwegianSeaintheearlymiddleEocene(Hohbeinetal.,2012).
At the same time, the residual depth of the Iceland plume de- creasedbyupto400 m(Parnell-Turneretal.,2014).Astheresid- ualdepthoftheIcelandPlumewasrobustlylinkedtothepercent- ageofNCWforthelast10Ma(Pooreetal.,2006),thelowresidual depthduring the middleEocenecould have enabledfirstconsid- erable NCW overflow over the GSR. Third, our modelling results showtheEasternNorthAtlanticandtheGreenland-NorwegianSea astheprimaryoverturninglocationsduringborealsummer,while nodeep-waterformationoccursintheLabradorSea.
AlateronsetofLabradorSeaWaterinthelateEocenetoearly Oligocene is in agreement with previous studies (Borrelli et al., 2014; Kaminskiet al., 1989). Based on the available paleoceano- graphicevidence,weproposetheonsetofNCWintheGreenland- Norwegian Sea atthe early-middle Eocene boundary and a later initiationofdeep-waterformationintheLabradorSea.
4.3. TheroleofNorthernComponentWaterintheevolutionofglobal climate
Inter-basinal benthic δ13C and δ18O gradients are commonly usedtodistinguishbetweendifferentwatermasses. Crameretal.
(2009) used a global compilation of benthic isotope records to assess Cenozoicpaleocenaographic change, whereas Sexton et al.
(2006a) focused on the Eocene. These authors show that Paleo- gene δ13C signatures of deep waters were very similar between thedifferentoceanbasins,particularlybetweentheNorthAtlantic andtheSouthAtlantic.TheU1410benthicδ13C valuesduring the middle Eocene are on the lower range of the global composite and inthe same range with previous values from the North At- lantic.MiddleEoceneNorthAtlantic δ13C values areintermediate between, butvery similar to thoseof theSouth Atlanticand the SouthernOcean.Accordingly,interbasinalbenthicδ13C was lowto non-existent, until Pacific benthicδ13C startedto become lighter fromabout14Ma onwards,whereas theSouthernOceandemon- stratesatrend towardslighter valuesfromabout8Ma.Together, thissuggeststhatinter-basinalδ13C gradientsmaynot havebeen asusefulasawatermasstracerasinthemodernoceanandhave been modulated by additional mechanisms such as the link be- tween climate and the biological productivity. This is supported bythecorrespondencebetweenδ13C andδ18O duringthemiddle Eocene.
Benthic foraminiferal δ18O gradients have been used to track theonsetofthemodernbimodaloceancirculationbasedonther- maldifferentiationbetweentheoceanbasins.Studiesbasedonthis proxyfor thermaldifferentiationbetweenthe North Atlanticand theSouthern OceansuggesttheonsetofNCWat38.5Ma, yetdo notprecludeanearliercontributionfromGSRoverflow(Borrelliet al.,2014; Langtonetal.,2016).However, thisdateisunsupported byevidencefromotherproxiessuchastheonsetofNorthAtlantic drift sedimentation and Nd isotopes. Just as for the carbon iso- topes,ourbenthicforaminiferaloxygenisotopesareintherangeof existingglobalbenthicforaminiferalisotopevaluesforthemiddle
Oligoceneboundary.
4.4. ResponseofNCWproductiontoastronomicalforcing
NCW formation in the Greenland-Norwegian Sea is driven by the density gradient between the surface ocean and underlying waterlayers.Thisdensitygradientrespondssensitivelytochanges in the salt balance andtemperatures at the North Atlantic con- vection sites.Wepropose acombinationoffivedifferentclimatic feedbackmechanisms,respondingtoastronomicalforcing,explain- ing thestrongNCW signal during obliquity minimainsediments fromtheNewfoundlandRidge.
Coolerglobaltemperatures duringobliquity minima,amplified intheNorth Atlantic,resultin enhancedoceanicheat losstothe atmosphereintheNorthAtlantic.Thestrongcoolingincreasesthe densityof thesaline surface watersflowing northfromthe sub- tropics.Thisprocesscouldincreasesurfacewaterdensityuntilthat waterstartstosink.
Moreover,duringobliquityminima,comparativelyweakinsola- tionathighlatitudescoolsthewater(densityincreases),whilethe stronger thannormal insolationat low latitudes induces warmer water(densitydecreases).Theresultingenhancedmeridionalden- sity gradient favours astronger overturning (Butler etal., 2016).
Interestingly,theoccurrenceoficerafteddebrisindicatestheiniti- ationofglaciationaroundtheArcticandtheGreenland-Norwegian SeaaroundthesametimeastheinitiationofNCWat∼47Maand
∼44Ma,respectively(Stickleyetal.,2009; Tripatietal.,2008).Be- causelower obliquitydecreases annual-averageinsolationathigh latitudes,seaicegrowthshouldpeakduringlocalizedcoldperiods coevalwithobliquityminima.Sea-iceformationremovesfreshwa- ter from the ocean, leaving behind enough salt to increase the salinity of the upper ∼25 m of the watercolumn by 1 psu per meterofseaicethickness(Maykut,1985).
A decline in seasonality owing to low obliquity weakens the hydrological cycle in the Northern Hemisphere and preferen- tiallydecreasestotalannualprecipitationacrossthehighlatitudes (Lawrenceetal.,2003).Theresultingreductionoffreshwaterinflux into the Greenland-Norwegian Seaand the North Atlantic would increasethedensityofsurfacewaters,therebystrengtheningther- mohalinecirculation.
TheobservedchangesinNorthAtlanticcirculationinourproxy datacouldresultfromchanges inatmosphericCO2 respondingto variations inobliquity.Inthe geologicalrecord andclimatemod- els,elevateddeep-seatemperatures(DeVleeschouweretal.,2017;
Hays et al., 1976) and higher atmospheric CO2 concentrations (Schefferetal.,2006) correspondtoobliquitymaxima(Lüthietal., 2008).Globalwarming andelevatedatmosphericCO2 canweaken the thermohaline circulation by surface water warming and sur- facewaterfreshening (Rahmstorf, 2006). Consequently,increasing atmospheric CO2 concentration during obliquity maxima hasthe
5. Conclusions
High-resolutionbenthicforaminiferalstableoxygenandcarbon isotopes from contourite deposits in the western North Atlantic (NewfoundlandDrift, IODPExpedition342,SiteU1410)combined with geochemical data from XRF reflect bottom water nutrient- content, temperatureandDeep Western Boundary Currentveloc- ity. Records from the Newfoundland Drifts indicate variations in NCW formation and associated Deep Western Boundary Current strength onorbital timescales,dominated byobliquity.Combined with globalcirculation model experiments the results show that enhancedoverturningisassociatedwithastrongcoolingofsurface waters intheGreenland-Norwegian Seaduring obliquity minima, while NCW formation is weaker with relatively warm tempera- tures in the Greenland-Norwegian Sea during obliquity maxima.
Ourmodelling results indicatethat middleEocene NCWwas pri- marily formed in the Greenland Norwegian Sea and the eastern NorthAtlantic.
Acknowledgements
M.V.,D.D.V.andH.P.arefundedbyEuropean ResearchCouncil ConsolidatorGrantEarthSequencing(grantagreement617462).I.N.
isfundedbytheNationalCenterofScience,Poland(DEC-2012/07/
N/ST10/03419)andtheDAAD (grant57130104).I.N.gratefullyac- knowledgesG. Knorr (AWI)andJ.Tyszka (INGPAN) fortheirsci- entific support. G.L. is funded by the Helmholtz society through the PACES and REKLIM programs. This research used data pro- vided by the Integrated Ocean Drilling Program (IODP). IODP is sponsored by theUS NationalScience Foundation (NSF)andpar- ticipatingcountriesunderthemanagementofJointOceanographic Institutions (JOI), Inc. We thank H. Kuhnertand theteam of the stableisotopelaboratoryatMARUM,Bremen,forthesupportwith stableisotopemeasurementsandtheIODPBremencorerepository (BCR) forlogistical support. We thank all participating scientists, technical staff,andcrew ofIODP Expedition342for theacquisi- tionandshipboardanalysisofcoresusedinthisstudy.
Appendix A. Supplementarymaterial
Supplementarymaterialrelatedtothisarticlecanbefoundon- lineathttps://doi.org/10.1016/j.epsl.2017.12.016.
References
Anagnostou,E.,John,E.H.,Edgar,K.M.,Foster,G.L.,Ridgwell,A.,Inglis,G.N.,Pancost, R.D., Lunt,D.J.,Pearson,P.N., 2016.Changingatmospheric CO2 concentration wastheprimarydriverofearlyCenozoicclimate.Nature 533,380–384.
Barron,E.J.,Hay,W.W.,Thompson,S.,1989.Thehydrologiccycle:amajorvariable duringEarthhistory.Palaeogeogr.Palaeoclimatol.Palaeoecol. 75,157–174.
