Late Miocene global cooling and the rise of modern ecosystems

Timothy D.Herbert1*,Kira https://www.360docs.net/doc/308376175.html,wrence2,Alexandrina Tzanova1,Laura Cleaveland Peterson3,

Rocio Caballero-Gill1and Christopher S.Kelly1

During the late Miocene epoch,about seven million years ago,large areas of the continents experienced drying,enhanced seasonality,and a restructuring of terrestrial plant and animal communities.These changes are seen throughout the subtropics,but have typically been attributed to regional tectonic forcing.Here we present a set of globally distributed sea surface temperature records spanning the past12million years based on the alkenone unsaturation method.We?nd that a sustained late Miocene cooling occurred synchronously in both hemispheres,and culminated with ocean temperatures dipping to near-modern values between about7and5.4million years ago.The period of maximum cooling coincides with evidence for transient glaciations in the Northern Hemisphere and with a steepening of the pole-to-equator temperature gradient,as well.We thus infer that late Miocene aridity and terrestrial ecosystem changes occurred in a global context of increasing meridional temperature gradients.We conclude that a global forcing mechanism,such as the previously hypothesized decline in atmospheric CO2levels between eight and six million years ago,is required to explain the late Miocene changes in temperature,climate and ecosystems.

E normous changes to terrestrial environments and ecosystems

occurred in the late Miocene.The Sahara Desert became established at this time,with the oldest known Aeolian deposits dating to7Myr ago(Ma;ref.1).This event was part of a larger drying of the landscape in the subtropics1,2,a trend further substantiated by the global radiation of succulent plant lineages in the late Miocene3. Perhaps the most spectacular landscape change came from the late Miocene expansion of C4grasslands(that is,plants using the C4photosynthetic pathway)at the expense of C3-dominated ecosystems in large areas of the tropics and subtropics4–https://www.360docs.net/doc/308376175.html,rge-scale shifts in vegetation and landscape in the late Miocene coincided with major turnovers in terrestrial fauna favouring browsers feeding on grasses and shrubs7,https://www.360docs.net/doc/308376175.html,stly,the base of the hominin lineage appears to lie at about7Ma(ref.9).

What is perplexing is that,to date,this environmental upheaval on land occurred without substantial evidence for late Miocene temperature change.Our single best marine index of the global climate state,the marine oxygen isotope record derived from benthic foraminifera,records an isotopic enrichment and growth of Antarctic ice volume at～13.9Ma and cyclic glacial-interglacial anomalies that intensify in the latest Pliocene and mid Pleistocene10–12.However,the late Mioceneδ18O record is singularly devoid of a strong trend that would suggest sustained climatic change during this time.Likewise,current proxy reconstructions of past atmospheric CO2levels show no significant shifts in greenhouse gas levels in late Miocene time13.

Given the importance of the ocean in storing heat and in determining the levels of atmospheric water vapour,it is clearly desirable to supplement the view of late Miocene climate derived from benthicδ18O measurements,which are sensitive only to two aspects of the climate system:the inventory of isotopically depleted continental ice,and the temperatures during the coldest times in the year of the small areas of the polar ocean that form the world’s

densest waters and fill the deep sea.However,remarkably few

estimates exist documenting the evolution of the marine surface

temperature field during the mid to late Miocene(Supplementary

Table1).Here,we provide new insights into late Miocene climate

evolution by providing the first globally distributed set of17marine

sea surface temperature(SST)estimates based on the organic U k

37 proxy,combining previously published works and nine new records,

including sites in all major ocean basins and both tropical and

mid-to high-latitude regions(Fig.1and Supplementary Table1).

Thanks to the wide geographic range of alkenone-producing

haptophyte algae,the U k

SST proxy provides an internally

consistent measurement of past marine surface temperatures on a

nearly global basis.Although variations in seasonal production and

depth habit of alkenone producers undoubtedly occur,a large body

of modern measurements supports the use of the U k

SST proxy as a

close approximation to mean annual SST14.We also note that while

U k

estimates become scattered in polar waters below～5?C and

in the warm pool regions of the ocean above28?C(ref.14),these

regions do not dominate our SST reconstructions.

Our reconstructions reveal substantial and coherent ocean

cooling in the terminal(Tortonian and Messinian stages)Miocene

(Figs2and3b).It is important to note that temperature changes

reflect both long-term climate change and changes over shorter

(orbital scale)time,the last of which are poorly resolved by

the present sampling.Converting the absolute values of inferred

SST to anomalies relative to present mean annual SST at the

site location(Fig.2e–h)helps to clarify major patterns.Because

plate motions become significant on this timescale,we referenced

anomalies to the palaeo-location of the sites;in regions of rapid

plate motion and strong temperature gradients such as the eastern

tropical Pacific,this correction makes a substantial di?erence

1Department of Earth,Environmental and Planetary Sciences,Brown University,Providence,Rhode Island02912,USA.2Department of Geology and Environmental Geosciences,Lafayette College,Easton,Pennsylvania18042,USA.3Environmental Studies,Luther College,700College Drive,Decorah, Iowa52101,USA.*e-mail:timothy_herbert@https://www.360docs.net/doc/308376175.html,

Temperature (°C)

Longitude

L a t i t u d e

60° E

30° E

0°

30° W

60° W 90° W 150° W 180°150° E

90° E 90°E

120° W 120° E Figure 1|Map of the Deep Sea Drilling Program,the Ocean Drilling Program and Integrated Ocean Drilling Program site locations used in this study.Black stars indicate sites with time series generated solely in association with this study.Black dots indicate site with some new data in addition to some previously published data.Grey dots indicate sites with data solely generated from previous work.For site information,modern SST and citations for previously published datasets,see Methods.

to the calculated anomalies (see Methods and Supplementary Information).Warmer than modern temperatures plot positively,while colder temperatures correspond to time periods when surface temperatures at the sites were colder than today;these occur during prior glacial periods.High-latitude amplification of anomalies and rough hemispheric symmetry are both evident (Figs 2and 3b and Supplementary Figs 1and 2).Warm temperatures at high latitudes,particularly in the North Atlantic,with temperatures up to 17?C warmer relative to modern conditions (Fig.2)typify the late Serravalian–early Tortonian (12–11Ma)interval of the Miocene.Weak latitudinal temperature gradients were presumably a relict of an even more equable world that began to change after the ‘refrigeration’of the Antarctic at 13.9Ma.Note that,because the alkenone unsaturation index occasionally approaches its limit of 1.0at some tropical and subtropical sites prior to 8Ma (Supplementary Table 1),our data set may underestimate tropical temperatures prior to 8Ma and overestimate equator-pole temperature gradients before that time.Modest cooling then gave way to accelerated cooling of mid-to high-latitude regions in the Messinian (～7.2–5.3Ma)that brought SSTs in these regions to approximately their modern values (Fig.2e,f,h).In contrast,late Miocene cooling is expressed only moderately in the tropics (alkenone estimates from the eastern tropical Pacific and Arabian Sea shown in Figs 2g and 3b are consistent with long-term late Miocene cooling indicated by the TEX86proxy in the Pacific Warm Pool region 15).Within the uncertainties of dating,temperatures then synchronously rebounded in the early Pliocene,with most of the Pliocene represented by temperatures warmer than modern.

The long-term,large-magnitude (～6?C)interhemispheric cool-ing evident in our SST records prior to widespread evidence for Northern Hemisphere glaciation supports previous work arguing for very di?erent global temperature thresholds for northern and

southern polar glaciation 16.Extensive East Antarctic glaciation at 13.9Ma (ref.17)apparently coexisted with rather equable SST gradi-ents,especially in the Northern Hemisphere.Substantial global SST cooling was required to bring the climate system to a state close to Holocene levels in the Messinian stage of the Miocene (Fig.2).The lack of sustained enrichment in the marine benthic δ18O record (Fig.3a)indicates that the late Miocene cooling did not lead to a large permanent increase in continental ice volume.An implication would be that the polar regions,or at least the parts of the polar oceans that formed deep water,were already very cold prior to 8Ma;the cooling we detect at tropical and intermediate latitudes therefore did not register in the benthic δ18O proxy.It appears that during the Messinian,Northern Hemisphere temperatures were frequently poised near,but not below the threshold for permanent continental ice formation,the East Antarctic ice sheet having long since reached a volume not dissimilar to today.Amplified cooling in our Northern Hemisphere high-latitude SST records could reflect the development of annually persistent Arctic sea ice,which may have acted as a potent positive feedback,perhaps also in conjunction with the albedo feedback of decreased forest area in mid-to high-latitudes 18.Polar sea ice expansion would not leave an imprint on the δ18O of sea water,and thus would be invisible in the isotopic proxy.Alternatively,given the signal-to-noise ratio and the multiple controls on the δ18O proxy,oxygen isotope data do not preclude the existence of 5–20m of sea level equivalent (equal to one to four times the present Greenland ice cap)stored in a Messinian Northern Hemisphere cryosphere and/or West Antarctic ice sheet 19.

Late Miocene conditions indeed seem to have allowed for ephemeral build ups of the continental cryosphere beyond the East Antarctic ice sheet.The cold interval from circa 7–5.4Ma that we reconstruct coincides very closely in time with previously enigmatic

Absolute temperature

ODP 887 N. Paci?c (this study)

ODP 907 Norwegian Sea (+ data from this study)ODP 982 N. Atlantic (+ data this study)ODP 883/884 N. Paci?c (this study)ODP 1021 CA Margin ODP 1208 N. Paci?c

Mediterranean (+ data from this study)ODP 1010 CA Margin ODP U1338 EEP

ODP 846 EEP (+ data from this study) ODP 850 EEP ODP 1241 EEP ODP 722 Arabian Sea

ODP 1088 S. Atlantic (this study)

ODP 1125 S. Paci?c (+ data from this study)ODP 1085 Beguela Margin

DSDP 594 S. Paci?c (this study)Temperature change

S e a s u r f a c e t e m p e r a t u r e (°C )

S e a s u r f a c e t e m p e r a t u r e c h a n g e (°C )

Age (Ma)

b c

d e

2426

2830

?4?202

4610

?505

15Figure 2|Temperature evolution over the past 12Myr for the sites in Fig.1.a –d ,Alkenone temperature records generated in this study or as previously published;all datasets use the calibration of ref.23.e –h ,SST changes as di erences relative to the modern mean annual sea surface temperature at the site.Data are binned by latitude:Northern Hemisphere (>50?N);Northern Hemisphere (30?–50?N);Southern Hemisphere (30?–50?S);and Tropics.Grey highlighted interval is late Miocene cooling (LMC)common to all sites.Site information and citations for all datasets are available in Supplementary Table 1;see Methods for SST backtracking procedure.

evidence of late Miocene glaciations of southeast Greenland 17,southeastern Alaska 20and South America 21,and in the Southern Hemisphere with pulses of ice-rafted detritus o?Wilkes Land and Adelie Land 22and,perhaps the formation of an ice sheet on West Antarctica 23.The coincidence of the first ice rafting in the North Atlantic 17and the North Pacific 20suggests that ‘overshoots’in cooling may have allowed for transient Northern Hemisphere glaciations during an approximately 2-Myr window of time.Multiple episodes of Messinian stage isotope enrichments (TG22-TG6events 19)in fact nest in the interval of the most pronounced cooling between 6and 5.4Ma identified in Fig.2.It is important to note,however,that the late Miocene cooling did not lead monotonically to the great Northern Hemisphere ice ages of the Pleistocene.Instead,much of Pliocene time apparently represents a modest reversal in the late Neogene cooling trend (Figs 2and 3b).This global temperature reversal may explain evidence for Pliocene warming in the Arctic 24and Antarctic 23.

The late Miocene SST pattern revealed here provides context for the rapid aridification and ecosystem shifts observed in the terrestrial record at this time 1–3,5–9.Prior to roughly 8Ma,dramatically reduced meridional temperature gradients would have been accompanied by expanded and weakened Hadley cells,an overall intensified hydrological cycle,and a contraction of hyper-arid zones in comparison to the modern climate state 25.On land,these di?erences are manifested as a greatly enlarged area of tropical and subtropical forest in the Tortonian (11.6–7.25Ma)as compared to modern conditions 26.Areas occupied by deserts today were inhabited by shrublands,grasslands,savannahs and woodlands 26.However,with the progressive late Miocene increase in equator–pole temperature gradients,the Hadley cells would have contracted and strengthened,focusing and intensifying the evaporative capabilities of the descending limbs of these cells.The resultant increase in aridity in subtropical regions is likely to have contributed to the pronounced floral and faunal changes that occurred in these regions during the late Miocene 2–9.

Large-scale cooling and aridification in the late Miocene would have enlisted a number of positive feedbacks.In addition to the possible Arctic and Antarctic sea ice feedbacks,replacement of forest biomes by grasslands,shrublands and deserts would have sig-nificantly increased continental albedo 18.Drier conditions on land also probably promoted cooling through a dustier atmosphere 27,although the contribution of this e?ect to global cooling is di?cult to constrain.Crucially,the ocean cooling itself may have played an essential role in amplifying the initial forcing of temperature changes,through the temperature-solubility e?ect on CO 2,the link between colder SST and diminished atmospheric water vapour content,and through temperature-mediated changes in stratification and wind-driven deep ocean mixing at high latitudes.The ultimate explanation of the changes revealed by declining ocean temperatures must account for the strong high-latitude

%δ18O (% P D B )

%δ13C (% P D B )

%δ13C (% )

p C O 2 (p p m )

Age (Ma)

S S T a n o m a l y (°C )

Figure 3|Late Miocene to present climate and carbon cycle changes.a ,Benthic δ18O composite 11.PDB,PeeDee Belemnite.b ,SST anomalies for each region in Fig.2e–h computed by binning anomaly estimates in 250-kyr windows (50%overlap)at each site and averaging the anomaly estimates.c ,δ13C (50-point running mean)from marine carbonates 15.d ,Soil δ13C data from Pakistan,an indicator of C 3versus C 4plant contribution to soil carbon 34.e ,Synthesis of published p CO 2proxy data obtained through a variety of di erent methodologies (references cited in Methods).Within each panel are relevant qualitative climate or carbon cycle events that have been reported in the literature (see Methods).Grey highlighted interval is late Miocene cooling (LMC)identi?ed in Fig.2.

amplification of the cooling and the observation that both hemispheres cooled in tandem.At 10Ma,we reconstruct a global SST anomaly of about +6?C compared to the present day

(see Methods).The global air temperature anomaly would have been enhanced by feedbacks over land and areas covered by sea ice by a factor of approximately 1.8over the SST anomaly 28.However,the large magnitude of cooling detected in the Southern (oceanic)Hemisphere rules out terrestrial feedbacks as the dominant source of di?erences between temperatures at 10Ma and today.The late Miocene cooling we detect is actually unfavourable to the expansion of C 4vegetation 4and requires o?setting competitive advantages such as tradeo?s of carbon fixation to photorespiration and water loss,or an unlikely scenario of enhanced summer temperatures (preferred by C 4plants)despite a generally cooler climate.We therefore view our SST information as strong support for the hypothesis advanced by refs 4,5to explain the rise of C 4plants in an environment of declining CO 2.The link we propose between late Miocene reduction in CO 2levels and increases in aridity satisfies two ecological constraints—the C 4pathway confers benefits both in carbon assimilation under low CO 2levels and in water retention 4.In the marine realm,carbon cycle indicators also suggest major late Miocene perturbations to the global carbon cycle.Approximately coincident with Antarctic refrigeration (13.9Ma),carbon isotope gradients,an index of ocean carbon sequestration,increased markedly between the Atlantic and Pacific basins 12.Subsequently,concurrent with the first significant appearance of ice in Greenland and South America (around 7Ma),a stepped decrease (～0.5–1 )occurs in benthic carbon isotope records 29(Fig.3c).Palaeoproductivity indicators suggest a widespread late Miocene—early Pliocene marine biogenic bloom 27.Diester-Haas et al.27explained this late Miocene carbon cycle change by suggesting that a drawdown of atmospheric CO 2,in conjunction with changes in atmospheric circulation patterns,caused a cooling of climate,an increase in seasonality and aridity,and related changes to C 3versus C 4vegetation.They argue that,as a result,land surfaces were then covered by less dense grassland/woodland vegetation,triggering greater erosion of δ13C-depleted organic soil material which increased nutrient loads in rivers,and in turn led to increased marine productivity.This carbon eventually became sequestered into the dissolved carbon reservoir of the deep ocean.

Although late Miocene uplift of the Himalayan mountain belt may have a?ected the southeast Asian monsoon and therefore regional vegetation 30,and the closure of the Paratethys seaway may have aided in the drying of the Saharan region 31,the changes observed on land were part of a much larger global temperature decrease that cannot be explained by the atmospheric e?ects of regional tectonic forcing.The positive feedbacks between late Miocene cooling,enhanced equator–pole temperature gradients and aridity identified above would provide amplification,but not initiation of the https://www.360docs.net/doc/308376175.html,Riviere and colleagues 32argue that changing ocean gateways between the late Miocene and Pliocene may have caused the thermocline to shoal past the depth critical for coupling it to SST,allowing the climate system to become sensitive to CO 2forcing.However,this hypothesis o?ers no clear explanation for the broad pattern of ocean cooling,the observed increase in terrestrial aridity and ecosystem changes,or the observed changes in the global carbon cycle during the late Miocene.

We propose that the late Miocene atmosphere descended to a critical level of CO 2,from above 500to below 350ppm,a range consistent with the threshold associated with conditions that favour C 4over C 3plants at tropical and subtropical temperatures 4and with the reduction in the greenhouse e?ect needed to explain the large global cooling we reconstruct.Messinian CO 2levels may have sunk episodically to the threshold of 280ppm necessary for Northern Hemisphere glaciation 16.We note that a recent study 33based on a physiological interpretation of marine coccolithophorid algae vital-e?ects is consistent with a late Miocene decline in CO 2,and in contrast to the results of other CO 2proxy reconstructions (Fig.3e).Thus,photosynthetic organisms in both terrestrial and

marine realms may have both been forced to adjust to a late Miocene

decline in p CO

2levels33.

Regional di?erences in the timing and magnitude of terrestrial and marine carbon cycle changes are not precluded by a scenario of dominant CO2forcing of late Miocene climatic and ecosystem

change.We note that while p CO

2seems to be the primary driver of the

viability of the C4pathway,a variety of additional factors—including temperature,aridity,light intensity,growth season temperature and warm-season rainfall4,6—may influence the competitiveness of di?erent photosynthetic strategies.A number of these factors would be controlled by regional climatic influences.Furthermore,our temperature data indicate that the pace of cooling and aridification was halted and perhaps reversed for much of the Pliocene,and therefore we would expect that biotic and geochemical proxies should reveal significant complexity in the timing and magnitude of change following the late Miocene excursion in the climate system.

The late Miocene(Tortonian through Messinian)now stands out as the time in the Neogene with the largest sustained(for example,>orbital scale)shift from an equable climate toward the modern world of strong equator–pole temperature gradients in both hemispheres.Given the significant uncertainties associated

with existing p CO

2reconstructions for the late Miocene13,we submit

that the SST pattern revealed by our synthesis provides strong circumstantial evidence supporting the hypothesis4,5that global changes in terrestrial vegetation were a direct consequence of a decline in a late Miocene atmospheric CO2.

Methods

Methods,including statements of data availability and any associated accession codes and references,are available in the online version of this paper.

Received21January2016;accepted17August2016;

published online26September2016

References

1.Schuster,M.et al.The age of the Sahara desert.Science311,821(2006).

2.Dupont,L.M.,Rommerskirchen,F.,Mollenhauer,G.&Schefu?,E.Miocene to

Pliocene changes in South African hydrology and vegetation in relation to the expansion of C4plants.Earth Planet.Sci.Lett.375,408–417(2013).

3.Arakaki,M.et al.Contemporaneous and recent radiations of the world’s major

succulent plant lineages.Proc.Natl https://www.360docs.net/doc/308376175.html,A108,8379–8384(2011).

4.Ehleringer,J.R.,Cerling,T.E.&Helliker,B.R.C4photosynthesis,atmospheric

CO2and climate.Oecologia112,285–299(1997).

5.Cerling,T.E.et al.Global vegetation change through the Miocene/Pliocene

boundary.Nature389,153–158(1997).

6.Str?mberg,C.A.Evolution of grasses and grassland ecosystems.Annu.Rev.

Earth Planet.Sci.39,517–544(2011).

7.Wang,Y.,Cerling,T.E.&MacFadden,B.J.Fossile horses and carbon isotopes:

new evidence for Cenozoic dietary,habitat,and ecosystem changes in North America.Palaeogeogr.Palaeoclimatol.Palaeoecol.107,269–279(1994).

8.Badgley,C.et al.Ecological changes in Miocene mammalian record show

impact of prolonged climatic forcing.Proc.Natl https://www.360docs.net/doc/308376175.html,A105,

12145–12149(2008).

9.Brunet,M.Two new Mio-Pliocene Chadian hominids enlighten Charles

Darwin’s1871prediction.Phil.Trans.R.Soc.B365,3315–3321(2010).

10.Lewis,A.R.et al.Mid-Miocene cooling and the extinction of tundra in

continental Antarctica.Proc.Natl https://www.360docs.net/doc/308376175.html,A105,10676–10680(2008). 11.Zachos,J.,Pagani,M.,Sloan,L.C.,Thomas,E.&Billups,K.Trends,

rhythms,and aberrations in global climate65Ma to present.Science292,

686–693(2001).

12.Cramer,B.,Toggweiler,J.R.,Wright,J.D.,Katz,M.E.&Miller,K.G.Ocean

overturning since the late Cretaceous:inferences from a new benthic

foraminiferal isotope compliation.Paleoceanography24,PA4216(2009).

13.Beerling,D.J.&Royer,D.L.Convergent Cenozoic CO2history.Nat.Geosci.4,

418–420(2011).

14.Müller,P.J.,Kirst,G.,Ruhland,G.,von Storch,I.&Rosell-Melé,A.Calibration

of the alkenone paleotemperature index U k

37on core-tops from the eastern

South Atlantic and the global ocean(60?N–60?S).Geochem.Cosmochim.Acta 62,1757–1772(1998).15.Zhang,Y.G.,Pagani,M.&Liu,Z.A12-million-year temperature history of the

tropical Pacific Ocean.Science344,84–87(2014).

16.DeConto,R.M.et al.Thresholds for Cenozoic bipolar glaciation.Nature455,

652–656(2008).

https://www.360docs.net/doc/308376175.html,rsen,H.C.et al.Seven million years of glaciation in Greenland.Science264,

952–955(1994).

18.Bradshaw,C.D.et al.The relative roles of CO2and palaeogeography in

determining late Miocene climate:results from a terrestrial model-data

comparison.Clim.Past8,715–786(2012).

19.Hodell,D.A.,Curtis,J.H.,Sierro,F.J.&Raymo,M.E.Correlation of late

Miocene to early Pliocene sequences between the Mediterranean and North Atlantic.Paleoceanography16,164–178(2001).

20.Krissek,L.in Proceeding of the Ocean Drilling Program,Scientific Results(eds

Rea,D.K.,Basov,L.A.,Scholl,D.W.&Allan,J.F.)179–194(Ocean Drilling Program,1992).

21.Mercer,J.H.&Sutter,https://www.360docs.net/doc/308376175.html,te Miocene-earliest Pliocene glaciation in

southern Argentina:implications for global ice-sheet history.Palaeogeogr.

Palaeoclimatol.Palaeoecol.38,185–206(1982).

22.Williams,T.et al.Evidence for iceberg armadas from east Antarctica in the

Southern Ocean during the late Miocene and early Pliocene.Earth Planet.Sci.

Lett.290,351–361(2010).

23.Kennett,J.P.&Barker,https://www.360docs.net/doc/308376175.html,test Cretaceous to Cenozoic climate and

oceanographic developments in the Weddell Sea,Antarctica:an ocean-drilling perspective.Proc.Ocean Drilling Program,Scientific Results937–960(Ocean Drilling Program,1990).

24.Thiede,J.et https://www.360docs.net/doc/308376175.html,te Cenozoic history of the polar North Atlantic:results from

ocean drilling.Quat.Sci.Rev.17,185–208(1998).

25.Brierley,C.M.et al.Greatly expanded tropical warm pool and weakened

Hadley circulation in the early Pliocene.Science323,1714–1718(2009).

26.Pound,M.J.et al.A Tortonian(late Miocene11.61–7.25Ma)global

vegetation reconstruction.Palaeogeogr.Palaeoclimatol.Palaeoecol.300,

29–45(2011).

27.Diester-Haass,L.,Billiups,K.&Emeis,https://www.360docs.net/doc/308376175.html,te Miocene carbon isotope

records and marine biological productivity:was there a(dusty)link?

Paleoceanography21,PA4216(2006).

28.Lunt,D.J.et al.Earth system sensitivity inferred from Pliocene modelling and

data.Nat.Geosci.3,60–64(2009).

29.Hodell,D.A.&Venz-Curtis,https://www.360docs.net/doc/308376175.html,te Neogene history of deepwater

ventilation in the Southern Ocean.Geochem.Geophys.Geosyst.7,

Q09001(2006).

30.Quade,J.&Cerling,T.E.Expansion of C4grasses in the late Miocene of

Northern Pakistan:evidence from stable isotopes in paleosols.Palaeogeogr.

Palaeoclimatol.Palaeoecol.115,91–116(1995).

31.Zhang,Z.et al.Aridification of the Sahara desert caused by Tethys Sea

shrinkage during the late Miocene.Nature513,401–404(2014).

https://www.360docs.net/doc/308376175.html,Riviere,J.P.et https://www.360docs.net/doc/308376175.html,te Miocene decoupling of oceanic warmth and

atmospheric carbon dioxide forcing.Nature486,97–100(2012).

33.Bolton,C.T.&Stoll,https://www.360docs.net/doc/308376175.html,te Miocene threshold response of marine algae to

carbon dioxide limitation.Nature500,558–562(2013).

34.Huang,Y.,Clemens,S.C.,Liu,W.,Wang,Y.&Prell,https://www.360docs.net/doc/308376175.html,rge-scale

hydrological change drove the late Miocene C4plant expansion in the

Himalayan foreland and Arabian Peninsula.Geology35,531–534(2007). Acknowledgements

Portions of this work were funded by the Petroleum Research Fund of the American Chemical Society,by the Evolving Earth Foundation,and by NSF grants to T.D.H.,K.T.L. and L.C.P.(0623487,1304366,1459280,and1545859).We thank D.Muller,J.Cannon and M.Seton for help with GPlates backtracking software,A.Alpert and A.Martin for laboratory analyses,C.Riihimaki for assistance in creating Fig.1and J.-E.Lee,

S.C.Clemens,C.Janis and C.Bolton for comments on earlier drafts.We gratefully acknowledge assistance by the curators for the IODP

Author contributions

T.D.H.and K.T.L.conceived the idea of a global synthesis of late Miocene SST estimates based on the alkenone method and wrote the majority of the text.K.T.L.,L.C.P.,C.S.K., R.C.-G.and A.T.provided primary data based on their laboratory work,and contributed to the analysis and the text.K.T.L.produced the figures.

Additional information

Supplementary information is available in the online version of the paper.Reprints and permissions information is available online at https://www.360docs.net/doc/308376175.html,/reprints. Correspondence and requests for materials should be addressed to T.D.H. Competing?nancial interests

The authors declare no competing financial interests.

Methods

Alkenone U k’

37-SST estimates in https://www.360docs.net/doc/308376175.html,rmation on location,modern annual

SST,sample resolution and identification of primary sources of U k

37data15,31,34–46

other than those generated for this study are reported for each site in Supplementary Table1.Supplementary Table2provides primary sample

information,along with U k

37,estimated SST,and reconstructed age,for each site.

We adopt the calibration of ref.14to mean annual SST,as it is based on a large

global data set of modern sediment U k

37values.Although we and other workers

recognize that the correlation of the unsaturation ratios measured in sediments to mean annual temperatures is an approximation to a more complex reality of synthesis and export of the biomarkers from the photic zone,and remineralization in surficial sediments,recent reanalyses of even more comprehensive alkenone data sets rea?rm the consistent correlation of the index to mean annual SST47.

The bulk of the U k

37analyses were generated at Brown University or at Lafayette

College using identical analytical procedures.Briefly,sediment samples of1–5g dry weight were extracted via a Dionex Accelerated Solvent Extractor and prepared for gas chromatography using toluene as the solvent for sample injection.Gas chromatographic analyses were carried out on Agilent5890or6890gas chromatographs(GCs),with a temperature ramp at25?C to240?C,followed by ramping at1.5?C to320?C.Gas chromatographic quality control was maintained by daily analysis of a laboratory standard alkenone mixture,and by running the same sample at the beginning and end of each GC run to determine sample drift due to changing chromatographic conditions.Analysis of both lab standard and

sample replication indicated a long-term replication of U k

37values equivalent to

0.1–0.2?C.With the exception of a handful of sites(ODP883,884,907and the Mediterranean),chromatograms were free from interferences.At the sites listed immediately above,lipid extracts were eluted through silica gel flash columns to allow for better GC analyses.Although there is some potential for interlaboratory

di?erences between investigators,we adopt U k

37values previously published as

listed in Supplementary Table1with one exception.When we merged our own data set from ODP722(ref.35)with Pliocene-Miocene data from ref.10an interlaboratory o?set became evident.We adjusted the data of ref.34upwards by 0.85?C,based on the di?erence between40samples analysed by both laboratories. We believe that this o?set comes from ref.34using a much faster temperature ramp in GC analysis,which tends to enhance co-elution of a C36:2fatty acid methyl ester(FAME)compound with the C37:3alkenone,thereby introducing a

‘cold’bias to chromatography.Primary(depth,U k

37,and inferred SST)as well as

derived(estimated age,SST anomaly relative to modern day,and SST anomaly corrected for plate motion)data are contained as Supplementary Table2with conversions to SST following refs14,48.

We also note a concern that the alkenone unstaturation index becomes insensitive to SST at very warm ocean temperatures(～28.3?C with the calibration of ref.48;29?C with the Muller et al.14calibration).To minimize this limitation,we excluded potential alkenone records from warm pool regions in our synthesis.We also report the number of data points at each site that exceed an unsaturation index of0.99in Supplementary Table1,along with the number of samples prior to8Ma with an unsaturation index of0.99or higher.Nearly all cases of very high unsaturation index occur prior to late Miocene cooling,consistent with our reconstruction of a very much warmer world before roughly8Ma.Our estimates are therefore probably biased to underestimate the tropical cooling,although the pattern we derive is consistent with long-term cooling reconstructed in warmer areas of the equatorial Pacific by the TEX86method15.However,it is also

evident that U k

37values near saturation are very rare in our data sets and

do not fundamentally bias our interpretations of global cooling in late

Miocene time.

Age models.Data collected with reference to nominal core depth(metres below sea floor,mbsf)were converted to estimated ages,along with an assessment of the uncertainty of assigned ages.In some cases,splicing between o?set holes allows for the development of a composite depth section that accounts for gaps or duplications in the coring process(metres composite depth,mcd).Data tables list primary sampling information,along with the corresponding mbsf,and,in some cases,mcd.We used the Bayesian age model program Bacon49

(https://www.360docs.net/doc/308376175.html,/blaauw/bacon.html)to fit age–depth curves to uncertainties associated with each age control point.Age–depth files used as input to Bacon are provided,along with the final Bacon parameters chosen for the age models reported here.The age–depth files(Supplementary Table3)identify the nature of the age control points used.Uncertainties in age fits that we report are the 2-sigma(95%confidence)limits returned by the program.

We used the following hierarchy of stratigraphic control to construct age models.Wherever possible,we used orbital-resolution age information that tunes stratigraphic records to the timing of earth orbital variations.In most cases such resolution is limited to Plio-Pleistocene timescales,but in a few cases(ODP Sites 982and1085and the Mediterranean),orbital resolution has been achieved in the Miocene section as well(refs19,42,43).We assign age uncertainties for orbital stratigraphies of±5kyr for the time interval0–2Ma,±10kyr to the time interval 2–3Ma,and±20kyr for the time interval3–12Ma.Next in preference comes the age information derived from magnetic polarity reversals captured in palaeomagnetic investigations of the study sites.The ages of these boundaries have been very precisely determined,and provide globally synchronous time reference points.We assign an age uncertainty of5kyr for the time interval0–3Ma,±20kyr for the polarity sequence between3and6Ma,and±40kyr for polarity boundaries older than6Ma.Unfortunately,only a handful of sites studied yielded reliable palaeomagnetic information in their Miocene and Pliocene sections(portions of Site883,continuously for Sites884,887,907,and Mediterranean sections).

Age models from the many sites rely on biostratigraphy.The resolution and accuracy of biostratigraphical datums varies widely,depending both on the sampling density of palaeontological studies,and on the synchroneity of microfossil events.We used biostratigraphic estimates derived from calcareous microfossils(planktonic foraminifera and nannofossils)exclusively,with the exception of sites in the North Pacific(ODP883,884,and887)dominated by siliceous microfossils.Wherever possible,we aligned the calibration of biostratigraphic datums in the primary literature with the current astronomically calibrated timescale50.We relied on literature evidence that nannofossil events as a rule are better synchronized to palaeomagnetic and astronchronological age control,so that we use smaller uncertainties in nannofossil age constraints than for planktonic foraminifera.A file listing site-by-site age control information where we report new data or have modified previous age models of U k

data sets from their published values is provided in Supplementary Table3.

Site backtracking for SST anomaly calculation.We calculated SST anomalies at each site with respect to the modern annual SST reported by the World Ocean Atlas (1998)database at the location of the site51,backtracked for plate motion (Supplementary Table4).We used the GPlates software(https://www.360docs.net/doc/308376175.html,)to calculate palaeolatitude and palaeolongitude in0.5Myr intervals for each location. Rotations were calculated using ref.52;coastlines were taken from ref.53.In the case of longitude,absolute values mean little—it is the distance to coastlines,where gradients in SST develop,that may matter.For California margin(Sites1010and 1021),eastern tropical Pacific(Sites846,850,1241,U1338),Benguela region(ODP 1085)and Somalia(ODP722)we referenced longitude to the location of a point on the coast of California,South America,Southern Africa,and the Somali Peninsula, respectively,of the same palaeolatitude and rotated the sites with respect to the relevant fixed coastline.The palaeolongitude thus approximates the location of the site in relation to present-day onshore–o?shore SST gradients,with a latitude correction.In the case of ODP1125,which lies near a frontal boundary that is believed to be anchored to a topographic anomaly on the sea bed,we account only for the northward movement of the site from the late Miocene to present,and fix the reference palaeolongitude at the present day.Sites722and1085have shifted slightly polewards/equatorwards,respectively,toward the present day.However, because they lie in regions where the present-day isotherms almost perfectly parallel the coastline,we have not implemented a correction for palaeolatitude. The Mediterranean sites lie in a region where tectonic movements have changed the shape of the basin extensively over time.The sites have drifted northwards

by a little over2?latitude since15Ma,and we applied a correction over time based on modern average Northern Hemisphere SST gradients over this

latitude range.

We used a MatLab program(provided as Supplementary Information)to derive the interpolated value of SST at each pair of palaeolatitude/palaeolongitude values from the WOA database.A cubic spline interpolation was used—except near coastal boundaries,where linear interpolation was applied.The plate tectonic trajectory thus defined a moving reference SST framework at each site that we converted to a smooth function over time for calculating the SST anomaly from the present day backward in https://www.360docs.net/doc/308376175.html,rmation on coordinates used for each site over time and the interpolated modern SST are given as Supplementary Table3 (separate file).

Stacking regionalδ18O events in Fig.3a.Continuous isotopic curve from ref.11, growth in global ice volume following ref.54,glaciation of SE Greenland from ref.17,glaciation in South America from ref.21.

Stacking regional SST Anomalies in Fig.3b.We recognize that each regional stack derives from a small number(3–6)of sites,thus we use this stacking approach to qualitatively illustrate the broad scale features these datasets reveal rather than to draw quantitative conclusions.At each site,we binned data over250kyr,with50% overlap,and calculated the mean of the backtracked SST anomaly(calibration of ref.14),the number of data points in the bin,and the standard deviation of the anomalies around their mean.For each region shown in Fig.3,we then averaged the mean anomaly at each time step between the sites(Supplementary Table5).We tested the sensitivity of the main features of the synthesis time series in Fig.3b by an alternative approach in which we interpolate all datasets to an even spacing of 70kyr,which is the average resolution of all16datasets.Our analysis(not shown) revealed that fundamental features of the stacked anomalies datasets remain similar to the time-binning approach,although the time-binning method yields smoother time series because it averages rather than interpolates data.

Marine carbon isotope events in Fig.3c.Continuous isotopic curves derive from a 50-point running mean of data provided by ref.15;timing of the late

Miocene–early Pliocene ‘biogenic bloom’is based on ref.27and the inferred global marine productivity increase was reported by ref.55.

Marine soil carbonate events in Fig.3d.δ13C in pedogenic carbonates from Pakistan based on information provided by ref.30.Timing of aridification of

eastern Asia from ref.56and references therein,from northern Africa 1,57,southern Africa from ref.2,diversification of succulent plants from ref.3,and the global rise of C 4-dominated ecosystems from ref.58and references therein.

Sources for palaeo-CO 2estimates in Fig.3e.CO 2proxies representing middle Miocene–Pleistocene time are represented in Fig.3;we attempted to use the most recent data sets and/or interpretations for each method.The source for alkenone-based palaeo-CO 2estimates are (brown diamonds)59and (orange

squares)60;for the boron-based approach (purple circles)61,62,(light purple triangles and blue diamonds)63,(light blue triangles and dark blue inverted triangles)47;for the palaeosol δ13C method 13(red triangles)and an estimate based on stomatal density from ref.13(green circles).

A 10Ma global SST reconstruction.We chose 10Ma as representative of global SST prior to late Miocene accelerated cooling,and prior to the major terrestrial land-surface and ecosystem changes reported in the literature.SST estimates

(corrected for plate motion)within a 10Ma ±0.25Ma time span were averaged at each site.To approximate the surface area of the Earth represented by each

location,we plot the SST anomaly as a function of the sine of latitude.Note that we do not have sites poleward of 68?N and 46?S,so the choice of ocean temperatures poleward of these constraints is rather uncertain.We took as a reasonable estimate a mean annual temperature of 0?C for the polar oceans at 10Ma,which is a warming of 1to 1.5?C compared to the present.We also note that our values of the tropical ocean temperature anomalies at 10Ma are probably underestimates,as values of U k 37>0.99become more frequent in this time interval.We applied two fits,the first a continuous sixth-order polynomial with the polar ocean constraint included in the fit,and a second third-order polynomial fit only to the actual SST data,with a linear fit from the poleward sites to the 0?C imposed polar ocean temperature.Both fits were then integrated globally to yield an estimate of the average ocean SST anomaly at 10Ma.The fit of the continuous function gave a warming of ～6?C,while the piecewise fit gave a warming of ～5.9?C.These estimates represent an approximately two-to fourfold increase in global temperature (relative to pre-industrial)in comparison to the mid-Pliocene ocean warming 25,51.

Supplementary Table 1reports information relevant to site location,sampling

resolution and references to sources of original U k

37measurements.Supplementary Table 2reports U k

37data for each site represented in Fig.2,primary sampling information,age determination along with estimated age

uncertainty,conversion of to SST U k

37according to two reference calibrations,and SST anomalies both with respect to unfixed geographic location and after correcting for plate motion over time.

Supplementary Table 3provides age–depth control points used at Sites 594,722,846,982,1085,1088,1125and 1208,the nature of these control points,and the parameters input to the Bacon age modelling program 49for each site.

Supplementary Table 4reports site location as a function of age after plate tectonic backtracking.

Supplementary Table 5reports stacked SST as a function of age shown in Fig.3.Supplementary Fig.1displays a sixth-order polynomial fit to all SST data at 10Ma plus polar ocean imposed temperature anomalies (circles represent alkenone-derived SST anomalies with reference to the present day and squares represent the imposed polar ocean temperature constraint).

Supplementary Fig.2presents an alternative third-order polynomial fit to

actual SST data at 10Ma with a piecewise linear fit to polar ocean constraints in the high northern and southern latitudes.

Code availability.The Matlab code used to interpolate modern SST from the SST grid taken from the Levitus World Ocean Atlas database is available in the supporting information files.

Data availability.We declare that the data that support the findings of this study are available in the supplementary files as described below.

References

35.Herbert,T.D.,Cleaveland Peterson,L.,Lawrence,K.T.&Liu,Z.Tropical ocean

temperature over the past 3.5million years.Science 328,1530–1534(2010).36.Liu,Z.&Herbert,T.D.High-latitude influence on the eastern equatorial

Pacific climate in the early Pleistocene epoch.Nature 427,720–723(2004).

https://www.360docs.net/doc/308376175.html,wrence,K.T.,Liu,Z.&Herbert,T.D.Evolution of the eastern tropical

Pacific through Plio-Pleistocene glaciation.Science 312,79–83(2006).

https://www.360docs.net/doc/308376175.html,wrence,K.T.,Herbert,T.D.,Brown,C.M.,Raymo,M.E.&Haywood,A.M.

High-amplitude variations in North Atlantic sea surface temperature during the early Pliocene warm period.Paleoceanography 24,PA2218(2009).39.Rommerskirchen,F.,Condon,T.,Mollenhauer,G.,Dupont,L.M.

&Schefuss,E.Miocene to Pliocene development of surface and subsurface temperatures in the Benguela Current system.Paleoceanography 26,PA3216(2011).

40.Seki,O.et al .Paleoceanographic changes in the Eastern Equatorial Pacific over

the last 10Myr.Paleoceanography 27,PA3224(2012).

41.Rousselle,G.,Beltran,C.,Sicre,M.A.,Ra?,I.&De Rafelis,M.Sea-surface

condition changes in the Equatorial Pacific during the Mio-Pliocene as inferred from coccolith geochemistry.Earth Planet.Sci.Lett.361,412–421(2013).42.Tzanova,A.,Herbert,T.D.&Cleaveland Peterson,L.C.Cooling

Mediterranean Sea Surface Temperatures during the late Miocene provide a climate context for evolutionary transitions in Africa and Eurasian flora and fauna.Earth Planet.Sci.Lett.419,71–80(2015).

43.Herbert,T.D.,Peterson,L.C.&Ng,G.Evolution of Mediterranean sea surface

temperatures 3.5-1.5Ma:regional and hemispheric influences.Earth Planet.Sci.Lett.409,307–318(2015).

44.Emeis,K.-C.et al .Temperature and salinity variations of Mediterranean Sea

surface waters over the last 16,000years from records of planktonic stable oxygen isotopes and alkenone unsaturation ratios.Palaeogeogr.Palaeoclimatol.Palaeoecol.158,259–280(2000).

45.Emeis,K.-C.et al .Eastern Mediterranean surface water temperatures and δ18O

composition during deposition of sapropels in the late Quaternary.Paleoceanography 18,1005(2003).

46.Fedorov,A.V .,Burls,N.,Lawrence,K.&Peterson,L.Tightly linked ocean

zonal and meridional temperature gradients over geological timescales.Nat.Geosci.8,975–980(2015).

47.Rosell-Melé,A.&Prahl,F.G.Seasonality of U K

37temperature estimates as

inferred from sediment trap data.Quat.Sci.Rev.72,128–136(2013).48.Prahl,F.G.,Muehlhausen,L.A.&Zahnle,D.L.Further evaluation of

long-chain alkenones as indicators of paleoceanographic conditions.Geochim.Cosmochim.Acta 52,2303–2310(1988).

49.Blaauw,M.&Christen,J.A.Flexible paleoclimate age-depth models using an

autoregressive gamma process.Bayesian Anal.6,457–474(2011).

50.Gradstein,F.M.,Ogg,G.&Schmitz,M.The Geologic Time Scale 20122-Volume

Set (Elsevier,2012).

51.NODC_WOA98(NOAA/OAR/ESRL PSD,1998);https://www.360docs.net/doc/308376175.html,/psd 52.Seton,M.et al .Global continental and ocean basin reconstructions since 200

Ma.Earth Sci.Rev.113,212–270(2012).

53.Wessel,P .&Smith,W .H.F.A global,self-consistent,hierarchical,

high-resolution shoreline database.J.Geophys.Res.101,8741–8743(1996).54.Lear,C.H.,Rosenthal,Y.&Wright,J.D.The closing of a seaway:ocean

water masses and global climate change.Earth Planet.Sci.Lett.210,425–436(2003).

55.Rea,D.K.,Snoeckx,H.&Joseph,https://www.360docs.net/doc/308376175.html,te Cenozoic eolian deposition in the

North Pacific:Asian drying,Tibetan uplift,and cooling of the Northern Hemisphere.Paleoceanography 13,215–224(1998).

56.Molnar,P .Mio-Pliocene growth of the Tibetan Plateau and evolution of East

Asian climate.Palaeontol.Electron.8,1–23(2005).

57.Ruddiman,W .F.et al .Late Miocene to Pleistocene evolution of climate in

Africa and the low-latitude Atlantic:overview of Leg 108results.in Proceeding of the Ocean Drilling Program,Scientific Results (eds Ruddiman,W .F.,Sarnthein,M.,Baldaufand,J.G.&Heath,R.)463–482(Ocean Drilling Program,1989).

58.Tipple,B.J.&Pagani,M.The early origins of terrestrial C 4photosynthesis.

Annu.Rev.Earth Planet.Sci.35,435–461(2007).

59.Seki,O.et al .Alkenone and boron-based Pliocene p CO 2records.Earth Planet.

Sci.Lett.292,201–211(2010).

60.Zhang,Y.G.,Pagani,M.,Liu,Z.,Bohaty,S.&DeConto,R.A 40-million-year

history of atmospheric CO 2.Phil.Trans.R.Soc.A 371,20130146(2013).61.Bartoli,G.,Hoenisch,B.&Zeebe,R.E.Atmospheric CO 2decline during the

Pliocene intensification of Northern Hemisphere Glaciation.Paleoceanography 26,PA4213(2011).

62.Martinez-Boti,M.A.et al .Plio-Pleistocene climate sensitivity evaluated using

high-resolution CO 2records.Nature 518(2015).

63.Foster,G.L.,Lear,C.H.&Rae,J.W .B.The evolution of p CO 2,ice volume

and climate during the middle Miocene.Earth Planet.Sci.Lett.341–344,243–254(2012).