Early Last Interglacial ocean warming drove

Early Last Interglacial ocean warming drove substantial ice mass loss from Antarctica

Authors Chris S.M. Turney, Christopher J. Fogwill, Nicholas R. Golledge, Nicholas P. McKay, Erik van Sebille, Richard T. Jones, David Etheridge, Mauro Rubino, David P. Thornton, Siwan M. Davies, Christopher Bronk Ramsey, Zoë Thomas, Michael I. Bird, Niels C. Munksgaard, Mika Kohno, John Woodward, Kate Winter, Laura S. Weyrich, Camilla M. Rootes, Helen Millman, Paul G. Albert, Andres Rivera, Tas van Ommen, Mark Curran, Andrew Moy, Stefan Rahmstorf, Kenji Kawamura, Claus-Dieter Hillenbrand, Michael E. Weber, Christina J. Manning, Jennifer Young and Alan Cooper

THIS IS A NON-PEER REVIEWED PREPRINT SUBMISSION TO EARTHARXIV MANUSCRIPT SUBMITTED TO PNAS

Early Last Interglacial ocean warming drove substantial ice mass loss from Antarctica 1

Title: Early Last Interglacial ocean warming drove substantial ice

2

mass loss from Antarctica

3 4

Classification: Physical Sciences

5

(Earth, Atmospheric, and Planetary Sciences)

6

Authors and Affiliations

7

Chris S.M. Turney1,2,3*, Christopher J. Fogwill1,2,4, Nicholas R. Golledge5,6, Nicholas

8

P. McKay7, Erik van Sebille2,8,9, Richard T. Jones10†, David Etheridge11, Mauro

9

Rubino4, David P. Thornton11, Siwan M. Davies12, Christopher Bronk Ramsey13, Zoë

10

Thomas1,2,3, Michael I. Bird14,15, Niels C. Munksgaard15,16, Mika Kohno17, John

11

Woodward18, Kate Winter18, Laura S. Weyrich19,20, Camilla M. Rootes21, Helen

12

Millman2, Paul G. Albert13, Andres Rivera22, Tas van Ommen23,24, Mark Curran23,24,

13

Andrew Moy23,24, Stefan Rahmstorf25,26, Kenji Kawamura27,28,29 Claus-Dieter

14

Hillenbrand30, Michael E. Weber31, Christina J. Manning32, Jennifer Young19,20 and

15

Alan Cooper19,20

16

1. Palaeontology, Geobiology and Earth Archives Research Centre, School of

17

Biological, Earth and Environmental Sciences, University of New South Wales,

18

Australia

19

2. Climate Change Research Centre, School of Biological, Earth and Environmental

20

Sciences, University of New South Wales, Australia

21

3. ARC Centre of Excellence in Australian Biodiversity and Heritage, School of

22

Biological, Earth and Environmental Sciences, University of New South Wales,

23

Australia

24

4. School of Geography, Geology and the Environment, Keele University, ST5 5BG,

25

UK

1


5. Antarctic Research Centre, Victoria University of Wellington, Wellington 6140,

27

New Zealand

28

6. GNS Science, Avalon, Lower Hutt, New Zealand

29

7. School of Earth and Sustainability, Northern Arizona University, Flagstaff,

30

Arizona 86011, USA

31

8. Grantham Institute & Department of Physics, Imperial College London, London,

32

UK

33

9. Institute for Marine and Atmospheric Research Utrecht, Utrecht University,

34

Utrecht, Netherlands

35

10. Department of Geography, Exeter University, Devon, EX4 4RJ, UK

36

11. Climate Science Centre, CSIRO Ocean and Atmosphere, Aspendale, Victoria,

37

3195 Australia

38

12. Department of Geography, Swansea University, Swansea, UK

39

13. Research Laboratory for Archaeology and the History of Art, University of

40

Oxford, Dyson Perrins Building, South Parks Road, Oxford, OX1 3QY, UK

41

14. Centre for Tropical Environmental and Sustainability Science, College of

42

Science and Engineering, James Cook University, Cairns, Australia

43

15. ARC Centre of Excellence in Australian Biodiversity and Heritage, James Cook

44

University, Cairns, Australia

45

16. Research Institute for the Environment and Livelihoods, Charles Darwin

46

University, Australia

47

17. Department of Geochemistry, Geoscience Center, University of Göttingen,

48

Goldschmidtstr. 1, 37077 Göttingen, Germany

49

18. Department of Geography and Environmental Sciences, Faculty of Engineering

50

and Environment, Northumbria University, Newcastle upon Tyne, NE1 8ST, UK

51

19. Australian Centre for Ancient DNA, University of Adelaide, Australia

2


20. ARC Centre of Excellence in Australian Biodiversity and Heritage, University of

53

Adelaide, Australia

54

21. Department of Geography, University of Sheffield, UK

55

22. Glaciology and Climate Change Laboratory, Centro de Estudios Cientficos,

56

Valdivia, Arturo Prat 514, Chile

57

23. Department of the Environment and Energy, Australian Antarctic Division, 203

58

Channel Highway, Kingston, Tasmania 7050, Australia

59

24. Antarctic Climate & Ecosystems Cooperative Research Centre, University of

60

Tasmania, Private Bag 80, Hobart, Tasmania 7001, Australia

61

25. Potsdam Institute for Climate Impact Research (PIK), PO Box 60 12 03,

62

D-14412 Potsdam, Germany.

63

26. University of Potsdam, Institute of Physics and Astronomy, Germany.

64

27. National Institute of Polar Research, Research Organizations of Information and

65

Systems, 10-3 Midori-cho, Tachikawa, Tokyo 190-8518, Japan

66

28. Department of Polar Science, Graduate University for Advanced Studies

67

(SOKENDAI), 10-3 Midori-cho, Tachikawa, Tokyo 190-8518, Japan

68

29. Institute of Biogeosciences, Japan Agency for Marine-Earth Science

69

and Technology, 2-15 Natsushima-cho, Yokosuka 237-0061, Japan

70

30. British Antarctic Survey, High Cross, Madingley Road, Cambridge CB3 0ET,

71

UK

72

31. Steinmann-Institute, University of Bonn, Germany.

73

32. Department of Earth Sciences, Royal Holloway University of London, TW20

74

OEX, UK

75

†. Deceased.

76

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

3


The future response of the Antarctic ice sheets to rising temperatures remains highly

78

uncertain (1, 2). An excellent period for assessing the sensitivity of Antarctica to warming

79

is the Last Interglacial (LIG; 129-116 kyr) (3-5), which experienced warmer polar

80

temperatures (6, 7) and higher global mean sea level (GMSL +6 to 9 m) (3, 8) relative to

81

present day. LIG sea level cannot be fully explained by Greenland Ice Sheet melt (4˚C) ocean

132

warming required for the loss of the WAIS, exceeding paleoclimate estimates (10,

133

22, 27, 28).

134 135

Here we report a new high-resolution record of environmental change and ice flow

136

dynamics from the Patriot Hills Blue Ice Area (BIA), exposed in Horseshoe Valley

137

(Ellsworth Mountains; see Methods) (Fig. 1). Importantly, Horseshoe Valley is a

138

locally-sourced compound glacier system (i.e. with negligible inflow) that is

139

buttressed by, but ultimately coalesces with, the Institute Ice Stream via the

140

Horseshoe Valley Trough, making the area sensitive to dynamic ice-sheet changes

141

across the broader WSE (13). Due to strong prevailing katabatic airflow, an

142

extensive BIA (more than 1150 m across) has formed to the leeward side of the

143

Patriot Hills, where ancient ice is drawn up from depth within Horseshoe Valley

144

(Fig. 1). Regional airborne and detailed local ground-penetrating radar (GPR)

145

surveys show a remarkably coherent series of dipping (24-45°) layers, broken by two

146

discontinuities, which represent isochrons across the Patriot Hills BIA, extending

147

thousands of meters into Horseshoe Valley. The BIA transect spans the time intervals

148

0-80 kyr and 130-134 kyr, constrained by analysis of trace gases and geochemically

149

identified volcanic layers exposed across the transect which have been Bayesian age

6


modelled against the recently compiled continuous 156 kyr global greenhouse gas

151

time series (CO2, CH4, and N2O) (29) on the AICC2012 age scale (30) (see

152

Methods). The record is located 50 km inland from the modern grounding line of the

153

Filchner-Ronne Ice Shelf in the Weddell Sea Embayment (WSE) (31) and close to

154

the Rutford Ice Stream, one of the largest methane hydrate reserves identified in

155

Antarctica (total organic carbon estimated to be 21,000 petagrams, or 21x1018 g)

156

(14). Today precipitation at the site is delivered via storms originating from the

157

South Atlantic or Weddell Sea. Crucially, the Ellsworth Mountains also lie in a

158

sector of the continent that is highly responsive to isostatic rebound under a scenario

159

of substantial WAIS mass loss, potentially preserving ice from around the time of the

160

LIG in small valley glaciers and higher ground areas (32).

161 162

Results

163

The isotopic series of dD across the Patriot Hills BIA exhibits a coherent record of

164

relatively low values between 18 and 80 kyr, consistent with a glacial-age sequence

165

(Fig. 2). Below these layers and at the periphery of zones of higher ice flow (31) we

166

find an older unit of ice exposed at the surface expressed by a step change to

167

enriched (interglacial) isotopic values (Fig 2 and S5), implying proximal warmer

168

conditions and reduced sea ice extent (33). Importantly, we identify a distinct tephra

169

horizon near the boundary of this older unit of ice which, based on major and trace

170

element geochemical fingerprinting (Figs 3 and S9), is correlated to a volcanic ash

171

from the penultimate deglaciation (Termination II) referred to as Tephra B in marine

172

sediments on the West Antarctic continental margin (34) and identified at 1785.14 m

173

depth in the Dome Fuji ice core where it is dated to 130.7±1.8 kyr (AICC2012

174

timescale) (30, 33, 34). The start of the oldest section of the sequence is dated here to

7


134.1±2.2 kyr, consistent with modelling studies, airborne radio-echo sounding lines,

176

and GPR profiles, which imply older ice exists at depth in the Ellsworth Mountains

177

(31, 32) (Fig. 1).

178 179

The combined tephra and trace gas analyses suggest a ~50 kyr hiatus after

180

Termination II (130.1±1.8 kyr). Radio-echo sounding surveys across the WSE have

181

identified a large subglacial basin comprising landforms reflecting restricted,

182

dynamic, marine-proximal alpine glaciation, with hanging tributary valleys feeding

183

an overdeepened Ellsworth Trough (35). The extensive nature of the subglacial

184

features implies substantial and repeated mass loss of the marine sections of the

185

WAIS, with the ice margin some 200 km inland of present day (35). However, the

186

timing of most recent retreat is currently unknown. Whilst previous surface exposure

187

dating in the region has suggested that the WAIS contribution to global sea level rise

188

during warmer periods was limited to 3.3 m above present, relatively short-duration

189

interglacial periods may have resulted in near-complete deglaciation (36),

190

contributing to the sub-glacial landscape identified across the WSE during recent

191

surveys (35). Previous work has interpreted erosional features D1 and D2 in the

192

Patriot Hills BIA to be a consequence of extensive ice surface lowering in Horseshoe

193

Valley (up to ~500 m since the Last Glacial Maximum, 21 kyr) and more exposure of

194

katabatic-enhancing nunataks, resulting in increased wind scour (13, 37). Whilst this

195

scenario may explain unconformity D0, previous work has demonstrated Horseshoe

196

Valley and the wider WSE to be highly sensitive to periods of rapid ice stream

197

advance or retreat in the last glacial cycle and Holocene with dramatic reductions in

198

surface elevation (13, 37-39). Furthermore, glaciological investigations assessing the

199

impact of ice shelf loss on glaciers along the Antarctic Peninsula provides important

8


insights, albeit on a smaller scale. The 2002 Larsen B ice shelf collapse led to many

201

of the tributary glaciers abruptly changing from a convex to a concave profile (40),

202

with relic ice left isolated on the upper flanks of the valleys (41). Both scenarios are

203

consistent with extensive grounding line retreat across the inner shelf of the Weddell

204

Sea and associated substantial ice loss across the wider WSE (31).

205 206

The ice at Patriot Hills therefore appears to preserve a record of glacier flow in

207

Horseshoe Valley up to the moment when the Filcher-Ronne Ice Shelf collapsed,

208

after which the sequence likely remained isolated due to regional ice flow

209

reconfiguration for multiple millennia; a situation that persisted until the ice surface

210

had risen sufficiently to enable the regional ice flow to recover sometime during late

211

MIS 5. The presence of a discrete older ice unit along the flanks of the Ellsworth

212

Mountains (31) (Figs 1 and S2) and the subsequent inferred highly variable climate

213

and/or sea ice extent across the wider WSE (Figs S5 and S11) implies the

214

preservation of ice from Marine Isotope Stage (MIS) 6/5 (Termination II) and 5/4

215

transitions in Horseshoe Valley. Our data provide the first direct evidence for a

216

WAIS collapse during the LIG and supports previous suggestions that the southern

217

polar region was a major driver of high global sea level (4, 11, 25), most likely at the

218

onset of the interglacial (34, 42, 43).

219 220

Discussion

221

What could be the cause of this ice loss? Recent work has proposed that the iceberg-

222

rafted Heinrich 11 event between 135 and 130 kyr (during Termination II) may have

223

significantly reduced North Atlantic Deep Water (NADW) formation and shut down

9


the Atlantic Meridional Overturning Circulation (AMOC) (12), resulting in a net heat

225

transport to the Southern Hemisphere (the bipolar seesaw pattern of northern cooling

226

and southern warming) (11) (Fig. 4). Under this scenario, surface cooling during

227

Heinrich 11 increased the northern latitudinal temperature gradient and caused a

228

southward migration of the Intertropical Convergence Zone and mid-latitude

229

Southern Hemisphere westerly airflow (9, 44). In the Southern Ocean, the associated

230

northward Ekman transport of cool surface waters (something akin to today; Fig. 1)

231

would have been compensated by increased delivery of relatively warm and nutrient-

232

rich deep Circumpolar Deep Water onto the Antarctic continental shelf (9, 11, 34,

233

44), leading to enhanced thermal erosion of ice at exposed grounding lines (2, 11).

234

The precise correlation between the Patriot Hills ice and West Antarctic marine

235

records (34) afforded by the Termination II tephra demonstrates for the first time that

236

the warming recorded in the BIA is coincident with a major, well-documented peak

237

in marine temperatures and productivity around the Antarctic continent and in the

238

Southern Ocean (34, 44) (Fig. 2). Recent modelling results suggest that increased

239

heat transport beneath the ice shelves can drive extensive grounding-line retreat,

240

triggering substantial drawdown of Antarctic ice sheets (9, 15, 27), and projected to

241

increase in the WSE during the twenty-first century (17). The subsequent delivery of

242

large volumes of associated freshwater into the Southern Ocean during the LIG

243

would have reduced Antarctic Bottom Water (AABW) production, resulting in

244

increased deepwater formation in the North Atlantic (11, 45) (Fig. 4).

245 246

With Southern Ocean warming and concurrent ice-sheet retreat, the large methane

247

reservoirs in Antarctic sedimentary basins (e.g. Rutford Ice Stream) would have

248

become vulnerable to release (14), contributing to elevated atmospheric levels

10


through the LIG (6, 29) (Fig. 2). High-latitude open water and sea ice are rich in

250

microbial communities, components of which may be collected by passing storms

251

and delivered onto the ice sheet (e.g. prokaryotes, DNA), offering insights into

252

offshore environmental processes (46, 47). To investigate environmental changes

253

prior to and after the ice-sheet reconfiguration recorded in the Patriot Hills BIA, we

254

applied a strict ancient DNA methodology and sequencing to provide the first

255

description of ancient microbial species locked within the ice (Methods). Methane-

256

utilizing microorganisms were found in three samples along the Patriot Hills transect

257

and were absent from other samples and laboratory controls. The most striking

258

feature of the Patriot Hills BIA genetic record was detected immediately prior to

259

inferred ice loss, where Methyloversatilis microbes dominated the detectable

260

microbial diversity (~130 kyr) (Fig. 2 and Fig. S13). Methyloversatilis was only

261

found in high abundance in this sample (with trace amounts identified at ~22 kyr).

262

Crucially, Methyloversatilis species are facultative methylotrophs (48), consistent

263

with elevated levels of CH4 in the water column near the end of Termination II.

264 265

The inferred substantial mass loss across the WSE implies a major role for ocean

266

warming during Termination II and the LIG. The most comprehensive published

267

high-latitude (≥40˚S) network of quantified sea surface temperature (SST) estimates

268

suggests an early LIG (~130 kyr) warming of 1.6±0.9˚C relative to present day (22,

269

28), providing an upper limit on the sensitivity of Antarctic ice sheets to ocean

270

temperatures. To investigate ice-sheet dynamics around the Patriot Hills and across

271

Antarctica in response to a range of ocean warming scenarios (1˚-3˚C), we applied

272

the Parallel Ice Sheet Model v.0.6.3 (Fig. 5) (15). The pattern of circum-Antarctic

273

ocean warming during this time period is not well established so we assume a

11


spatially-uniform warming pattern relative to present day temperatures. Our model

275

time series illustrates that the majority of ice loss takes place within the first two

276

millennia, depending on the magnitude of the forcing (Fig. 5 and Table 1). This

277

corresponds to the time period of inferred loss of marine-based sectors of the ice

278

sheet (Fig. 2), primarily in West Antarctica. In contrast to some whole-continent

279

models, our simulations do not include mechanisms by which a grounded ice cliff

280

may collapse (10), a process that produces considerably faster and greater ice margin

281

retreat than reported here.

282 283

For the 2˚C warmer than present day scenario, our model predicts a contribution to

284

GMSL rise of 3.2 m in the first millennium of forcing (Fig. 5). The loss of the

285

Filchner-Ronne Ice Shelf within 200 years of warming triggers a non-linear response

286

by removing the buttressing force that stabilises grounded ice across large parts of

287

the WSE and the EAIS (most notably the Recovery Basin) (Fig. 6), consistent with

288

other modelling studies (9, 26, 27). Ongoing slower ice loss subsequently occurs

289

around the margins of East Antarctica, producing a sustained contribution to sea-

290

level rise.

291 292

Previous work has suggested that the Ellsworth Mountains would have experienced a

293

relatively large positive isostatic adjustment (~500 m) accompanying the loss of the

294

WAIS (32). To investigate how an evolving ice-sheet geometry would manifest

295

across the wider region, we extracted local ice surface and bed elevations for the

296

WSE from the model simulation that uses a 2°C ocean warming with no atmospheric

297

warming. Fig. 6 illustrates the sequence of events that take place as the ice sheet

12


evolves. First, loss of the Filchner-Ronne Ice Shelf in the Weddell Sea triggers a

299

non-linear response, removing the buttressing force that stabilises grounded ice

300

across large parts of the WSE and the EAIS (most notably the Recovery Basin). The

301

loss of back-stress allows for an acceleration of grounded ice and a rapid but short-

302

lived thinning episode (32). At the Patriot Hills, bedrock uplift of c. 30 m over this

303

0.2 kyr period is outpaced by a surface lowering of c. 75 m, implying a net ice-sheet

304

thinning of around 105 m. Subsequently, regional-scale isostatic uplift elevates both

305

the bed topography (c. 250 m) and ice-sheet surface (c. 350 m) relative to the initial

306

configuration. The difference between these two values reflects positive net mass

307

balance of the ice sheet here (~0.055 m/year). After around 2.5 kyr, renewed

308

dynamic thinning of the ice sheet in the Patriot Hills leads to a rapid thinning and

309

lowering of the ice sheet surface, at a rate exceeding regional-scale bedrock

310

subsidence (120 m over 0.4 kyr, or 0.3 m/year, compared to c. 70 m over 3.2 kyr, or

311

0.022 m/year respectively) (Fig. 6). The Patriot Hills record is consistent with the

312

loss of grounded ice early in the LIG (32) as a consequence of regional ice dynamic

313

changes and isostatically-driven isolation of Horseshoe Valley from sustained ocean

314

forcing. For the 1˚ and 3˚ warming scenarios, similar spatial losses are modelled,

315

with GMSL rises of 1.8 and 4.0 m for the first millennium, respectively (Table 1).

316

Atmospheric warming of the magnitude suggested in Antarctic cores (>4˚C) (24, 25,

317

49-51) adds an additional meter of equivalent global sea level within the first

318

millennium (Fig. S15). Whilst some modelling studies have argued the loss of the

319

Filchner–Ronne Ice Shelf does not display a strong marine ice sheet instability

320

feedback (52), our results suggest otherwise. The ice sheet modelling outputs support

321

our interpretation of the Patriot Hill BIA record, demonstrating the substantial ice-

322

sheet flow reconfiguration and resultant isostatic response that would occur under

323

such a scenario (Fig. 6).

13

Early Last Interglacial ocean warming drove substantial ice mass loss from Antarctica 324 325

The evidence for substantial mass loss from Antarctica in the early LIG has

326

important implications for the future (3, 42). Our field-based reconstruction and

327

modelling results support a growing body of evidence that the Antarctic ice sheets

328

are highly sensitive to ocean temperatures. Driven by enhanced basal melt through

329

increased heat transport into cavities beneath the ice shelves (2, 15), this process is

330

projected to increase with a weakening AMOC during the twenty-first century (1, 16-

331

18), and one which may lead to other positive feedbacks such as destabilization of

332

methane hydrate reserves (14).

333 334

Acknowledgements

335

CSMT, CJF, MIB, AC and NRG are supported by their respective Australian

336

Research Council (ARC) and Royal Society of New Zealand fellowships. Fieldwork

337

was undertaken under ARC Linkage Project (LP120200724), supported by Linkage

338

Partner Antarctic Logistics and Expeditions (ALE). JW and KW undertook GPR

339

survey of the Patriot Hills record through NERC support with logistical field support

340

from the British Antarctic Survey. SMD acknowledges financial support from Coleg

341

Cymraeg Cenedlaethol and the European Research Council. KK was supported by

342

JSPS KAKENHI. We thank Dr Chris Hayward and Dr Gwydion Jones for electron

343

microprobe assistance, Kathryn Lacey and Gareth James for help with preparing the

344

tephra samples, Drs Nelia Dunbar, Nels Iverson and Andrei Kurbatov for discussions

345

on the tephra correlations, CSIRO GASLAB personnel for support of gas analysis,

346

Prof. Bill Sturges and Dr Sam Allin of the Centre for Ocean and Atmospheric

347

Sciences (University of East Anglia, UK) for performing the sulfur hexafluoride

348

analyses, Levke Caesar (Potsdam Institute for Climate Impact Research) for

14


preparing the recent trend in SSTs in Fig. 1, Vicki Taylor (British Ocean Sediment

350

Core Research Facility, Southampton) for assistance with marine core sampling, and

351

Dr Emilie Capron (British Antarctic Survey) for advice on reconstructing early

352

southern Last Interglacial temperatures. CSIRO’s contribution was supported in part

353

by the Australian Climate Change Science Program (ACCSP), an Australian

354

Government Initiative.

355 356

Author contributions

357

CSMT, CJF, NRG and RTJ conceived the research; CT, CJF, NRG, AC, LW, JY,

358

SMD, PGA, CJM, RTJ, DE, MR, DPT, CBR, ZT, MB, NCM, MK, JW and KW

359

designed the methods and performed the analysis; CT wrote the paper with input

360

from all authors. There are no competing interests.

15


361 362

Fig. 1 Location and age profile of Patriot Hills blue ice area. (a) Location of

363

Antarctic ice and marine records discussed in this study and austral spring-summer

364

(October-March) sea surface temperature trends (over the period 1981-2010;

365

HadISST data). (b) Trace gas (circles), tephra (triangles) and boundary (square) age

366

solutions for surface ice along transect B-B’ relative to an arbitrary datum along the

367

transect (displayed in panel d). Dashed lines denote unconformities D0-D2 at their

16


surface expression. (c) Basal topography of the Ellsworth Subglacial Highlands

369

(West Antarctica) with the locations of airborne radio-echo sounding transect A-A’

370

(displayed in panel e) and Rutford Ice Stream (IS) (31). (d) The location of Patriot

371

Hills in Horseshoe Valley (LIMA background image) with the BIA climate line

372

(marked by transect B-B’), dominant ice flow direction and distance to grounding

373

line. The Horseshoe Valley, Independence and Ellsworth troughs are given by the

374

initials HV, IT and ET respectively. (e) Airborne radio-echo sounding cross section

375

of ice within Horseshoe Valley, Independence and Ellsworth troughs (modified from

376

ref. (31)). Digitization highlights basal topography (brown), lower basal ice unit

377

(grey) and upper ice unit (red) as well as internal stratigraphic features (black for

378

observed, dashed for inferred, and purple for best estimate).

17


379 380

Fig. 2. Climate, ocean circulation and sea level changes over the past 140 kyr. (a)

381

d18O record from the North Greenland (NGRIP) ice core (55, 56). (b) Bermuda Rise

382

231

383

ratio of 0.093 marking sluggish/absent AMOC (12). Selected North Atlantic

384

Heinrich (H) events and reduced AMOC shown. (c) Biogenic opal flux from ODP

385

Site 1094 (53.2˚S) as a measure of wind-driven upwelling in the Southern Ocean

386

(44). (d) Comparison between the recently compiled global atmospheric methane

Pa/230Th data (reversed axis; 1s uncertainty) with dashed line denoting production

18


time series (red line; 2s envelope) (29) with the methane record from the West

388

Antarctic Patriot Hills (black circles with 1s uncertainty; open circles mark

389

anomalously high concentration data excluded from age model). (e) Patriot Hills dD

390

record with mean Holocene values; envelope 1s. Grey shading denotes the timing of

391

the surface elevation change across the Weddell Sea Embayment as indicated by the

392

hiatus in the Patriot Hills sequence and inferred substantial Antarctic ice mass loss,

393

consistent with the reported divergence of the isotopic signal observed between the

394

horizontal Mount Moulton ice core record from the WAIS and East Antarctic ice

395

cores (24, 25, 33), and peak global sea level (8). Triangles denote the presence of

396

geochemically-identified tephra layers in the Patriot Hills record. Pie-chart

397

representation of percentage methane-utilising bacteria in 16S rRNA samples from

398

Patriot Hills; crosses denote absence of these bacteria (Methods). (f) Temporal

399

changes in ocean productivity with peak productivity (PP; green shading) during

400

interglacials and subsequent enhanced content of calcareous microfossils in Antarctic

401

continental margin sediments (red shading) (34). Dashed black line shows position of

402

tephra identified in the Patriot Hills (-340 m), Dome Fuji (1785.14 m) and Tephra B

403

in marine sediments from the West Antarctic continental margin. (g) East Antarctic

404

Dome Fuji d18O record (30, 33). (h) Reconstructed relative sea level curve with 2s

405

envelope (8). Yellow shading highlights the timing of iceberg-rafted Heinrich debris

406

event 11 (H11), when large amounts of iceberg-rafted debris were deposited in the

407

North Atlantic (11) and the 231Pa/230Th ratio on Bermuda Rise shifted towards the

408

production ratio of 0.093, representative of sluggish or absent AMOC (12). Circled

19


numbers 1 and 2 denote enhanced upwelling-induced warming in the Southern

410

Ocean and Antarctic ice mass loss, respectively.

411 412

Fig. 3. Average trace element concentrations of Patriot Hills tephra at -340 m and

413

Tephra B from marine sediment cores PC108 (4.65 m depth) and PC111 (6.86 m

414

depth) (34) normalised to Primitive Mantle (53) (a). Biplots show comparison

415

between selected trace element concentrations of the tephra in the different

416

sequences. Error bars on plots show 2s of replicate analyses of MPI-DING

417

StHs6/80-G (54), but errors are typically smaller than the data symbols (b-e).

20

Early Last Interglacial ocean warming drove substantial ice mass loss from Antarctica a Onset Heinrich 11 AIS

ITCZ CH4

0

AAIW

1

D e pth (k m)

Heinrich 11 icebergs

Sea ice

NAIW

2 AABW

3

AABW

4 5 60° S

30° S

b Late Heinrich 11 0

30° N

60° N

ITCZ

AIS CH4 AAIW

1

D e pth (k m)

0°

NADW

2 3

AABW

4 5 60° S

30° S

c Early Last Interglacial

0°

30° N

60° N

ITCZ

AIS 0 AAIW

D e pth (k m)

1 2 3

NADW AABW

4 5 60° S

30° S

0°

30° N

60° N

Latitude

418 419

Fig. 4. Ocean-atmospheric interactions during Termination II and the Last

420

Interglacial. Panels show changing Atlantic Meridional Overturning Circulation

421

(AMOC) in response to iceberg discharge (a,b) in the North Atlantic (Heinrich event

422

11) during Termination II and (c) from the Antarctic Ice Sheets (AIS) during the Last

423

Interglacial, with inferred shifts in atmospheric circulation including mid-latitude

424

Southern Hemisphere westerly (crossed circle) airflow and Intertropical

21


Convergence Zone (ITCZ) (9, 11, 44, 45). Vertical arrows denote CH4 and heat flux

426

associated with Antarctic coastal easterly (dot in circle) and westerly (crossed circle)

427

airflow (2, 14). AABW, AAIW, NAIW and NADW define Antarctic Bottom Water,

428

Antarctic Intermediate Water, North Atlantic Intermediate Water and North Atlantic

429

Deep Water, respectively.

22


Dome Fuji Patriot Hills

Mt Moulton

Taylor Glacier

430 431

Fig. 5. Modelled Antarctic ice-sheet evolution under Last Interglacial forcing. (a)

432

Sea-level equivalent mass loss for ice-sheet simulations forced by a range of air and

433

ocean temperature anomalies relative to present day. ‘dT’ and ‘dOT’ describes

434

atmospheric and ocean temperature anomalies respectively. (b-d) shows Antarctic

435

Ice Sheet extent and elevation with 2˚C warmer ocean temperatures over time

436

intervals of 1, 2 and 5 kyr, respectively (with no atmospheric warming); equivalent

437

sea-level contribution is given in the bottom left corner of each panel. Locations of

438

Patriot Hills (Ellsworth Mountains, West Antarctic Ice Sheet) and ice core records

439

discussed in this study are shown in panel b. Inset box in (b) outlines region shown in

440

Fig. 6.

23

Early Last Interglacial ocean warming drove substantial ice mass loss from Antarctica Ice−shelf collapse Regional uplift Dynamic thinning

a 600

b

0 yr

c

200 yr

d

400 yr

e

800 yr

f

1200 yr

g

3000 yr

Bed elevation (m)

1300 400

Surface elevation (m)

1400 500

1200 300

0

441

1000 2000 3000 4000 5000

Years

442

Fig. 6. Bed (black line) and surface (blue) elevation changes at Patriot Hills

443

(Ellsworth Mountains, West Antarctic Ice Sheet) in response to 2˚C warmer ocean

444

temperatures over a time interval of 5 kyr (with no atmospheric warming) (a). Bed

445

(black line) and surface (blue) elevation changes vs time, with phases of the

446

prevalence of particular processes, such as ice-shelf collapse (mint shaded), regional

447

uplift (grey shaded) and dynamic thinning (light-brown shaded), highlighted. (b-g)

448

Selected time slices corresponding to dashed lines in panel a showing ice shelf extent

449

and ice sheet elevation in the Weddell Sea Embayment (WSE) over the first 3 kyr.

450

Location of Patriot Hills is marked by the red square; grey shaded areas are ice-shelf

451

covered, whilst white areas are free of both grounded and floating glacial ice.

24

Early Last Interglacial ocean warming drove substantial ice mass loss from Antarctica 1000 yrs

2000 yrs

5000 yrs

10,000 yrs

1˚C SST warming 0˚C air

1.81

3.82

4.98

5.25

2˚C air

2.13

4.82

5.84

6.04

4˚C air

2.47

5.84

6.89

7.28

0˚C air

3.20

5.32

5.86

6.81

2˚C air

3.61

5.97

6.95

7.69

4˚C air

4.19

6.83

8.37

9.82

0˚C air

4.01

5.77

6.58

8.22

2˚C air

4.66

6.25

7.46

9.16

4˚C air

5.11

7.17

9.01

10.92

2˚C SST warming

3˚C SST warming

452

Table 1 Sea-level equivalent mass loss (meters) for Antarctic ice-sheet simulations

453

forced over 10,000 years by range of annual air and ocean temperature anomalies

454

relative to present day.

25


METHODS

456

Patriot Hills

457

Site description and geomorphological context. The Patriot Hills BIA (Horseshoe

458

Valley, Ellsworth Mountains; 80°18′S, 81°21′W) is a slow flowing (3 m depth to

515

minimise modern air contamination and/or alteration (60). The samples were double

516

bagged and sealed in the field, and transported frozen to CSIRO’s ICELAB facility

517

in Melbourne for the extraction and measurement of trace gases using a modified dry

518

extraction ‘cheese grater’ and cryogenic trapping technique (61, 62). The trapped air

519

samples were analysed by gas chromatography (GC) and the trace gas concentrations

520

are reported against the calibration scales maintained by CSIRO GASLAB (63).

521

Where sufficient material was available, duplicates were analysed.

522

The presence of visible tephra layers (volcanic ash horizons) provides

523

additional chronological control for the Patriot Hills BIA. Here we report two new

524

tephras from Patriot Hills at 10 m and -340 m, both observed as ~4 cm units of

525

dispersed shards (Fig. S7). Shards were extracted by centrifugation of the melted ice

526

samples and put onto a glass slide for electron microprobe analysis. The slides were

527

ground and polished using silica carbide paper and decreasing grades of diamond

528

suspension to expose fresh sections of glass. Single-grain analyses of ten oxides were

529

performed on a Cameca SX-100 electron microprobe at the Tephrochronology

530

Analytical Unit, University of Edinburgh. See SI for operating conditions (64).

531

Geochemical results are provided in Table S1. The shards from 10 m are bimodal,

28


with a basanitic and trachytic composition (Fig. S8). The shards from -340 m are

533

trachytic in composition and exhibit a tightly-clustered population (Fig. S9). Both

534

were compared to published tephras from across Antarctica (34, 65-75). The 10 m

535

tephra has the closest match to be the basanite Tephra C from the WAIS Divide at

536

3149.12 m (Similarity Coefficient or SC = 0.98), equivalent to 44.9±0.3 kyr (75).

537

The -340 m tephra revealed the closest match to a tephra layer in the Dome Fuji ice

538

core at 1785.14 m depth (SC = 0.966; equivalent to 130.7±1.8 kyr on the AICC2012

539

timescale (30, 68, 76); data previously unpublished).

540

A widespread tephra found in marine sedimentary records on the West

541

Antarctic continental margin (Tephra B) has been proposed to correlate to the tephra

542

at Dome Fuji 1785.14 m but the correlation has until now remained only tentative in

543

the absence of any reported geochemistry from the latter (34). Here we find the

544

major oxides from Tephra B have a close match to Patriot Hills -340 m (SC = 0.948),

545

consistent with this interpretation. To test this correlation, we undertook trace

546

element analysis of the glass shards from Patriot Hills at -340 m. Unfortunately, the

547

Dome Fuji shards were too thin for analysis. However, we were able to undertake

548

trace element analyses on Tephra B samples from two marine sediment cores from

549

the West Antarctic continental margin: PC108 (4.65 m depth) and PC111 (6.86 m

550

depth) (34). Trace element analysis of volcanic glass shards were performed using an

551

Agilent 8900 triple quadrupole ICP-MS (ICP-QQQ) coupled to a Resonetics 193nm

552

ArF excimer laser-ablation in the Department of Earth Sciences, Royal Holloway,

553

University of London. See SI for operating conditions (77). Accuracies of LA-ICP-

554

MS analyses of ATHO-G and reference StHs6/80-G MPI-DING (54) glass were

555

typically ≤ 5%. Identical trace element glass chemistries (Fig. 2, Table S2) strongly

556

support the correlation of Patriot Hills -340 m tephra horizon and the marine West

557

Antarctic Tephra B (34) which is in turn correlated to Dome Fuji 1785.14 m (33, 34,

29


68, 76), and probably originates from the Marie Byrd Land volcanic province (West

559

Antarctica) (34). The recognition of a widespread tephra horizon across a large sector

560

of the Antarctic at the very onset of the LIG provides a time-parallel marker horizon

561

crucial for future studies investigating Antarctic ice-sheet mass loss.

562

To develop an age model we undertook Bayesian age modelling using a

563

Poisson process deposition model (P_sequence) in the software package OxCal

564

v.4.2.4 (https://c14.arch.ox.ac.uk/) (Tables S3 and S4) (78, 79). Using Bayes

565

theorem, the algorithms employed sample possible solutions with a probability that is

566

the product of the prior and likelihood probabilities (80, 81). ‘Calibration curves’

567

with 20 year resolution were developed for the three trace gas species using the 156

568

kyr time series (29). Taking into account the deposition model, the reported ages of

569

the tephra layers, and the common age solutions offered by the trace gas

570

measurements, the posterior probability densities quantify the most probable age

571

distributions. The available constraints suggest the 1156-m long Patriot Hills BIA

572

transect spans time intervals from ~134.2 to ~1.3 kyr comprising four key zones: 4 (-

573

362 to -339 m, equivalent to 134.2±2.2 to 130.1±1.8 kyr), 3 (-326 to 240 m,

574

equivalent to 80±6.1 to 22.7±2.8 kyr), 2 (240 to 360 m, equivalent to 22.7±2.8 to

575

10.3±0.4 kyr) and 1 (360 to 800 m, 10.3±0.4 to 1.3±0.6 kyr). The Agreement Index

576

for the Patriot Hills age model was 101.6% (Aoverall=71.2%), exceeding the

577

recommended rejection Agreement Index threshold of 60% (79) (Methods).

578

Regardless of the relatively large uncertainty associated with the oldest section of ice

579

(zone 4), the identification of the 130.7±1.8 kyr (AICC2012 timescale) Tephra

580

B/Dome Fuji 1785.14 m (30, 33, 34) within Patriot Hills at -340 m unambiguously

581

demonstrates the presence of Termination II-age ice.

582

30


Isotopes. δD and δ18O isotopic measurements were performed between 1 and 3 m

584

resolution at James Cook University (JCU) using Diffusion Sampling - Cavity Ring-

585

down Spectrometry (DS-CRDS) (International Atomic Energy WICO Lab ID.

586

16139) (82). This system continuously converts liquid water into water vapour for

587

real-time stable isotope analysis by laser spectroscopy (Picarro L2120-i, Sunnyvale,

588

CA, USA). See SI for operating conditions. To ensure reproducibility, a subset of

589

samples was rerun at UNSW ICELAB for δD and δ18O using a Los Gatos Research

590

Liquid Water Isotope Analyser 24d (International Atomic Energy WICO Lab ID.

591

16117). Reported overall analytical precision on long term ice core standards is

592