Orbitally forced ice sheet fluctuations during the Marinoan Snowball Earth glaciation

Abstract

Two global glaciations occurred during the Neoproterozoic. Snowball Earth theory posits that these were terminated after millions of years of frigidity when initial warming from rising atmospheric CO2concentrations was amplified by the reduction of ice cover and hence a reduction in planetary albedo. This scenario implies that most of the geological record of ice cover was deposited in a brief period of melt-back. However, deposits in low palaeo-latitudes show evidence of glacial-interglacial cycles. Here we analyse the sedimentology and oxygen and sulphur isotopic signatures of Marinoan Snowball glaciation deposits from Svalbard, in the Norwegian High Arctic. The deposits preserve a record of oscillations in glacier extent and hydrologic conditions under uniformly high atmospheric CO2concentrations. We use simulations from a coupled three-dimensional ice sheet and atmospheric general circulation model to show that such oscillations can be explained by orbital forcing in the late stages of a Snowball glaciation. The simulations suggest that while atmospheric CO2concentrations were rising, but not yet at the threshold required for complete melt-back, the ice sheets would have been sensitive to orbital forcing. We conclude that a similar dynamic can potentially explain the complex successions observed at other localities.

Full text

LETTERS
PUBLISHED ONLINE: 24 AUGUST 2015 | DOI: 10.1038/NGEO2502
Orbitally forced ice sheet fluctuations during the
Marinoan Snowball Earth glaciation
Douglas I. Benn1,2*, Guillaume Le Hir3, Huiming Bao4, Yannick Donnadieu5, Christophe Dumas5,
Edward J. Fleming1,6†, Michael J. Hambrey7, Emily A. McMillan6, Michael S. Petronis8,
Gilles Ramstein5, Carl T. E. Stevenson6, Peter M. Wynn9and Ian J. Fairchild6
Two global glaciations occurred during the Neoproterozoic.
Snowball Earth theory posits that these were terminated
after millions of years of frigidity when initial warming from
rising atmospheric CO2concentrations was amplified by the
reduction of ice cover and hence a reduction in planetary
albedo1,2. This scenario implies that most of the geological
record of ice cover was deposited in a brief period of
melt-back3. However, deposits in low palaeo-latitudes show
evidence of glacial–interglacial cycles4–6. Here we analyse the
sedimentology and oxygen and sulphur isotopic signatures
of Marinoan Snowball glaciation deposits from Svalbard, in
the Norwegian High Arctic. The deposits preserve a record
of oscillations in glacier extent and hydrologic conditions
under uniformly high atmospheric CO2concentrations. We
use simulations from a coupled three-dimensional ice sheet
and atmospheric general circulation model to show that such
oscillations can be explained by orbital forcing in the late stages
of a Snowball glaciation. The simulations suggest that while
atmospheric CO2concentrations were rising, but not yet at
the threshold required for complete melt-back, the ice sheets
would have been sensitive to orbital forcing. We conclude
that a similar dynamic can potentially explain the complex
successions observed at other localities.
The Wilsonbreen Formation in northeast Svalbard contains a
detailed record of environmental change during the Marinoan, the
second of the major Cryogenian glaciations (650–635 Ma; refs 7,8).
At this time, Svalbard was located in the Tropics on the eastern
side of Rodinia9,10. The <180 m thick Wilsonbreen Formation was
deposited within a long-lived intracratonic sedimentary basin11. It
is subdivided into three members (W1, W2 and W3) based on
the relative abundance of diamictite and carbonate beds7,8 (Fig. 1
and Supplementary Figs 1 and 2). The occurrence throughout the
succession of lacustrine sediments containing both precipitated car-
bonate and ice-rafted detritus, and intermittent evaporative carbon-
ates and fluvial deposits, indicates that the basin remained isolated
from the sea, consistent with eustatic sea-level fall of several hundred
metres and limited local isostatic depression (Supplementary Infor-
mation; ref. 12). This makes it ideal for investigating environmental
change within a Neoproterozoic panglaciation, as it provides direct
evidence of subaerial environments and climatic conditions.
We made detailed sedimentary logs at ten known and new
localities extending over 60 km of strike (Fig. 1 and Supplemen-
tary Fig. 1; see Methods). Seven sediment facies associations were
identified, recording distinct depositional environments that varied
in spatial extent through time (Supplementary Fig. 3 and Supple-
mentary Information). These are: FA1: subglacial, recording di-
rect presence of glacier ice; FA2: fluvial channels; FA3: dolomitic
floodplain, recording episodic flooding, evaporation and microbial
communities; FA4: carbonate lake margin, including evidence of
wave action; FA5: carbonate lacustrine, including annual rhythmites
and intermittent ice-rafted debris; FA6: glacilacustrine, consisting
of ice-proximal grounding-line fans (FA6-G) and ice-distal rainout
deposits (FA6-D); and FA7: periglacial, recording cold, non-glacial
conditions. Further descriptions are provided in the Supplementary
Information. The vertical and horizontal distribution of these facies
associations (Fig. 1) allows the sequence of environmental changes
to be reconstructed in detail:
(1) The base of the Wilsonbreen Formation is a well-marked
periglacially weathered horizon with thin wind-blown sands
(Supplementary Fig. 4a,b). This surface records very limited
sediment cycling in cold, arid conditions.
(2) At all localities, the weathering horizon is overlain by fluvial
channel facies (FA2) and mudstones, marking the appearance
of flowing water in the basin and implying positive air temper-
atures for at least part of the time (Supplementary Fig. 5a).
(3) Glacilacustrine deposits (FA6-D) record flooding of the basin
and delivery of sediment by ice rafting (Supplementary
Fig. 4c,d). Far-travelled clasts are common, indicating transport
by a large, continental ice sheet.
(4) Warm-based, active ice advanced into the basin, indicated by
traction tills and glacitectonic shearing (FA1; Supplementary
Fig. 4e–g). (1–4 make up Member W1).
(5) Ice retreat is recorded by a second periglacial weathering
surface (FA7). This is overlain by fluvial channel, flood-
plain, lake-margin and carbonate lacustrine sediments of
W2 (FA2-5; Supplementary Fig. 5), recording a shifting mo-
saic of playa lakes and ephemeral streams. Lakes and river
channels supported microbial communities. Millimetre-scale
carbonate-siliciclastic rhythmites indicate seasonal cycles of
1Department of Geology, The University Centre in Svalbard (UNIS), N-9171 Longyearbyen, Norway. 2School of Geography and Geosciences, University of
St Andrews, St Andrews KY16 8YA, UK. 3Institut de Physique du Globe de Paris, 75238 Paris, France. 4Department of Geology and Geophysics, E235
Howe-Russell Complex, Louisiana State University, Baton Rouge, Louisiana 70803, USA. 5Laboratoire des Sciences du Climat et de l’Environnement,
CNRS-CEA, 91190 Gif-sur-Yvette, France. 6School of Geography, Earth and Environmental Sciences, University of Birmingham, Birmingham B15 2TT, UK.
7Institute of Geography and Earth Sciences, Aberystwyth University, Aberystwyth SY23 3DB, UK. 8Natural Resource Management, Environmental
Geology, New Mexico Highlands University, Las Vegas, New Mexico 87701, USA. 9Lancaster Environment Centre, University of Lancaster, Lancaster
LA1 4YQ, UK. †Present address: CASP, West Building, 181A Huntingdon Road, Cambridge CB3 0DH, UK. *e-mail: Doug.Benn@unis.no
NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience 1
© 2015 Macmillan Publishers Limited. All rights reserved© 2015 Macmillan Publishers Limited. All rights reserved© 2015 Macmillan Publishers Limited. All rights reserved© 2015 Macmillan Publishers Limited. All rights reserved

LETTERS NATURE GEOSCIENCE DOI: 10.1038/NGEO2502
010
180
150
100
50
m
0
km
20 30 40 50 60
D1
W3
W2
W1
E4
DRA
DIT
REIN
KLO
McD
ORM
PIN
AND
BAC
SLA
Marine carbonate FA1: subglacial (till) FA2: fluvial
FA6: (proximal)
FA7: periglacial
FA6: glacilacustrine (distal)FA1: (glacitectonite)
FA2-5: (carbonate) fluvial and lacustrine
Figure 1 | Sedimentary architecture and palaeoenvironments of the
Wilsonbreen Formation. Regional correlation of facies associations and
members W1, W2 and W3 across northeast Svalbard. From north to south,
study locations are: DRA, Dracoisen; DIT, Ditlovtoppen; AND, East
Andromedafjellet; REIN, Reinsryggen (informal name); KLO, Klofjellet;
McD, MacDonaldryggen; BAC, Backlundtoppen–Kvitfjellet ridge; PIN,
Pinnsvinryggen (informal name); SLA, Slangen and ORM, Ormen.
photosynthesis. The environment seems to have been closely
similar to that of the present-day McMurdo Dry Valleys in
Antarctica, although with less extreme seasonality owing to its
low latitude13.
(6) Water levels and glacier extent underwent a series of oscil-
lations, recorded by switches between glacilacustrine diamic-
tite (FA6-D) and fluvial, lacustrine and lake-margin sediments
(FA2-5) in W2. Sedimentation rates inferred from annual
rhythmites in W2 suggest that each retreat phase may have
lasted ∼104years.
(7) A second major ice advance marks the base of W3, with
widespread deposition of subglacial tills and glacitectonism of
underlying sediments. Basal tills are absent from the northern-
most locality, but close proximity of glacier ice is recorded by
grounding-line fans (FA6-G; Supplementary Fig. 4h,i).
(8) Ice retreated while the basin remained flooded and glaci-
genic sediment continued to be delivered to the lake by
ice rafting. Thin laminated carbonates (FA5) in W3 indi-
cate periods of reduced glacigenic sedimentation, indicative
of minor climatic fluctuations over timescales of ∼103years
(Supplementary Fig. 5g).
(9) A sharp contact with overlying laminated ‘cap’ carbonate
(Supplementary Fig. 2) records the transition to post-glacial
conditions. At some localities, basal conglomerates provide ev-
idence of subaerial exposure, followed by marine transgression.
The cap carbonate closely resembles basal Ediacaran carbonates
elsewhere, and marks global deglaciation, eustatic sea-level rise
and connection of the basin to the sea1,12,14.
Environmental and atmospheric conditions during deposition
of W2 and W3 can be further elucidated by isotopic data from
carbonate-associated sulphate in lacustrine limestones (Fig. 2
and Supplementary Fig. 6). These exhibit negative to extremely
negative 117O values, with consistent linear co-variation with δ34S,
indicating mixing of pre-glacial sulphate and isotopically light
sulphate formed in a CO2-enriched atmosphere15,16. The observed
0
−1.6
−1.4
−1.2
−1.0
−0.8
−0.6
−0.4
−0.2
0.0
y = 0.106x − 3.37
Heavy
endmember
Towards light
endmember
r2 = 0.59
W2 limestones, existing data
W2 limestones, new data
W3 limestones, all new data
51015202530
CAS δ34S (% )
%
CAS Δ17O (% )
%
Figure 2 | Co-variation of 117O and δ34S from carbonate-associated
sulphate in W2 and W3. ‘Existing data’ (ref. 16) and new data define a
mixing line between pre-glacial sulphate (top) and an isotopically light
sulphate formed by oxidation of pyrite including incorporation of a
light-117O signature from a CO2-enriched atmosphere. Data from W2 and
W3 lie on closely similar trend lines, indicating no detectable change in
pCO2between deposition of the two members.
values could reflect non-unique combinations of pCO2,pO2, O2
residence time and other factors, but a box model17 indicates pCO2
was most likely ∼10 to 100 mbar (1 mbar =1,000 ppmv).
These values are far too high to permit formation of low-
latitude ice sheets in the Neoproterozoic, but they are consistent
with a late-stage Snowball Earth. For an ice-free Neoproterozoic
Earth, model studies indicate mean terrestrial temperatures in the
range 30–50 ◦C for pCO2=10 to 100 mbar (ref. 18). Formation of
low-latitude ice sheets requires much lower pCO2, on the order of
0.1–1 mbar (refs 2,19,20). Once formed, however, ice sheets can
persist despite rising CO2from volcanic outgassing, as a result of
a high planetary albedo. This hysteresis in the relationship between
pCO2and planetary temperature is a key element of Snowball Earth
theory. It implies that W2 and W3 were deposited relatively late in
the Marinoan, after volcanic outgassing had raised pCO2from 0.1 or
1 mbar to 10 or 100 mbar. Modelled silicate weathering and volcanic
outgassing rates indicate that this would require 106to 107years21.
The consistent co-variation of 117O and δ34S in lacustrine
limestones in both W2 and W3 suggests no detectable rise in
atmospheric pCO2, as this would alter the slope of the mixing
line (Fig. 2). This implies that the glacier oscillations recorded
in W2 and W3 occurred during a relatively short time interval
(<105years21) towards the end of the Marinoan. In turn, this implies
that the remainder of the Wilsonbreen Formation (including the
basal weathering horizon) represents many millions of years, during
which pCO2built up from the low values necessary for inception
of low-latitude glaciation to those indicated by the geochemical
evidence. The weathering horizon provides direct evidence of cold,
arid conditions during this interval, before the appearance of fluvial
and glacilacustrine sediments in the basin.
The evidence for ice sheet advance/retreat cycles at low latitudes
in a CO2-enriched atmosphere motivated a series of numerical
simulations to test the hypothesis that these cycles were linked
to Milankovitch orbital variations. We employed asynchronous
coupling of a three-dimensional ice sheet model and an atmospheric
general circulation model using the continental configuration of
ref. 22. We first ran simulations with a modern orbital configuration
to examine ice sheet behaviour through a large range of pCO2
values from 0.1 to 100 mbar (ref. 23; Supplementary Figs 7–10).
Consistently with previous results2,20, at low pCO2(0.1 mbar), global
ice volume reaches 170 ×106km3, but substantial tropical land
2NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience
© 2015 Macmillan Publishers Limited. All rights reserved© 2015 Macmillan Publishers Limited. All rights reserved© 2015 Macmillan Publishers Limited. All rights reserved© 2015 Macmillan Publishers Limited. All rights reserved

NATURE GEOSCIENCE DOI: 10.1038/NGEO2502 LETTERS
Surface NH
Surface SH
Land ice surface
(×1013 m2)
0
−1,000
−600
−200
1,000
600
200
1,000
2,000
3,000
4,000
Ice thickness (CSO, 30 kyr of simulation)Ice thickness (WSO, 20 kyr of simulation)
Ice thickness (WSO minus CSO)
0°
20° N
20° S
40° N
a
40° S
60° S
0°
20° N
20° S
40° N
40° S
60° S
0°
20° N
20° S
40° N
40° S
60° S
2.5
3.0
5.5
6.0
0 10,000 20,000
Time (yr)
30,000 40,000
[CSO][WSO] [WSO][CSO]
m
m
Latitude
Latitude
Latitude
b
cd
0.50
0.50
0.50
Figure 3 | Modelled ice sheet oscillations in response to orbital forcing. a,b, Shaded contours show land ice thickness obtained with 20 mbar of carbon
dioxide in response to changes of orbital forcing (WSO (a)) and CSO (b), warm/cold summer orbit for the northern hemisphere) over the course of two
precession cycles (40kyr of simulation). In the light brown continental areas without ice, the white line is used to represent the old ice sheet extension
(WSO case). The Svalbard area is indicated by a red circle. c, Ice thickness variation in 10 kyr (WSO case after 20 kyr minus CSO case after 30kyr of
simulation). In a–cthe continental outline is shown by the 0.5m elevation contour. d, Surface of hemisphere covered by ice (m2) through time ([WSO] and
[CSO] indicate which orbital configuration is used).
areas remain ice free as a result of sublimation exceeding snowfall
(Supplementary Fig. 10a). Ice volume remains relatively constant
for pCO2=0.1 to 20 mbar (Supplementary Fig. 10b), owing to an
increase in accumulation that compensates for higher ablation rates
(Supplementary Fig. 13). In contrast, above 20 mbar, ice extent in the
eastern Tropics significantly decreases (Supplementary Fig. 10c). At
pCO2=100 mbar, most of the continental ice cover disappears, except
for remnants over mountain ranges (Supplementary Fig. 10d).
To test the sensitivity of the tropical ice sheets to Milankovitch
forcing, experiments with changing orbital parameters were initial-
ized using the steady-state ice sheets for pCO2=20 mbar. Although
obliquity has been invoked as a possible cause of Neoproterozoic
glaciations24, this mechanism remains problematical and cannot
account for significant climatic oscillations at low latitudes25,26 . We
therefore focused on precession as a possible driver, and used two
opposite orbital configurations, favouring cold and warm summers,
respectively, over the northern tropics (CSO: cold summer orbit
and WSO: warm summer orbit; Supplementary Fig. 14). Switch-
ing between these configurations causes tropical ice sheets to ad-
vance/retreat over several hundred kilometres in 10 kyr (Supple-
mentary Movie 1), with strong asymmetry between hemispheres
(Fig. 3). Shifting from WSO to CSO causes ice retreat in the South-
ern Hemisphere and ice sheet expansion in the Northern Hemi-
sphere (Supplementary Fig. 14c,d). Significant ice volume changes
occur between 30◦N and S, but are less apparent at higher lati-
tudes. This reflects higher ablation rates in the warmer low lati-
tudes (Supplementary Fig. 14e,h), and higher ice sheet sensitivity
to shifting patterns of melt. Larger greenhouse forcing at the end of
the Snowball event implies increasing ice sheet sensitivity to subtle
insolation changes. Given a strong diurnal cycle23, our simulations
also predict a significant number of days above 0◦C in the tropics
(Supplementary Fig. 15), consistent with geological evidence for
ice rafting, liquid water in lakes and rivers, and photosynthetic
microbial communities.
Our results show that geological evidence for glacial–interglacial
cycles5–7 is consistent with an enriched Snowball Earth theory. Ter-
mination of the Marinoan panglaciation was not a simple switch
from icehouse to greenhouse states, but was characterized by a
climate transition during which glacial cycles could be forced
by Milankovitch orbital variations. The geochemical evidence
presented here implies that at least the upper 60–70% of the
Wilsonbreen Formation was deposited in ∼105years, on the as-
sumption that a trend in pCO2would be evident over longer
timescales21. Rates of CO2build-up, however, may have slowed in
the later stages of Snowball Earth owing to silicate weathering of
exposed land surfaces, so it is possible that the oscillatory phase was
more prolonged.
Initiation of low-latitude glaciation in the Neoproterozoic re-
quires low pCO2(0.1–1 mbar; refs 2,19,20), implying that the os-
cillatory phase was preceded by a prolonged colder period (∼106
to 107years) during which pCO2increased gradually as a result of
volcanic outgassing21. This timescale is in agreement with recent
dates indicating the Marinoan lasted ∼15 million years27. The basal
weathering horizon is consistent with a period of low temperatures
and limited hydrologic cycle before the oscillatory phase2,19 .
Further work is needed to refine the upper and lower limits
of pCO2conducive to climate and ice sheet oscillations in Snow-
ball Earth. Factors not included in the present model, such as
supraglacial dust or areas of ice-free tropical ocean28–30 , can be
expected to make the Earth system more sensitive to orbital forcing.
While many details remain to be investigated, our overall conclu-
sions remain robust.
The Neoproterozoic Snowball Earth was nuanced, varied and
rich. We anticipate that detailed studies of the rock record in other
parts of the world, in conjunction with numerical modelling studies,
will continue to yield insight into the temporal and regional diversity
of this pivotal period in Earth history.
Methods
Methods and any associated references are available in the online
version of the paper.
Received 17 February 2015; accepted 8 July 2015;
published online 24 August 2015
NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience 3
© 2015 Macmillan Publishers Limited. All rights reserved© 2015 Macmillan Publishers Limited. All rights reserved© 2015 Macmillan Publishers Limited. All rights reserved© 2015 Macmillan Publishers Limited. All rights reserved

LETTERS NATURE GEOSCIENCE DOI: 10.1038/NGEO2502
References
1. Hoffman, P. F. & Schrag, D. P. The Snowball Earth hypothesis: Testing the limits
of global change. Terra Nova 14, 129–155 (2002).
2. Donnadieu, Y., Goddéris, Y. & Le Hir, G. Treatise on Geochemistry 2nd edn,
Vol. 6, 217–229 (Elsevier, 2014).
3. Hoffman, P. F. Strange bedfellows: Glacial diamictite and cap carbonate
from the Marinoan (635 Ma) glaciation in Namibia. Sedimentology 58,
57–119 (2011).
4. Allen, P. A. & Etienne, J. L. Sedimentary challenge to Snowball Earth. Nature
Geosci. 1, 817–825 (2008).
5. Rieu, R., Allen, P. A., Plötze, M. & Pettke, T. Climatic cycles during a
Neoproterozoic ‘‘snowball’’ glacial epoch. Geology 35, 299–302 (2007).
6. Le Heron, D. P., Busfield, M. E. & Kamona, F. An interglacial on snowball
Earth? Dynamic ice behaviour revealed in the Chuos Formation, Namibia.
Sedimentology 60, 411–427 (2013).
7. Fairchild, I. J. & Hambrey, M. J. Vendian basin evolution in East Greenland and
NE Svalbard. Precambrian Res. 73, 217–233 (1995).
8. Halverson, G. P. in Neoproterozoic Geobiology and Paleobiology (eds Xiao, S. &
Kaufman, A. J.) 231–271 (Springer, 2006).
9. Li, X.-X., Evans, D. A. & Halverson, G. P. Neoproterozoic glaciations in a
revised global palaeogeography from the breakup of Rodinia to the assembly of
Gondwanaland. Sedim. Geol. 294, 219–232 (2013).
10. Petronis, M. S. et al. in AGU Fall Meeting 2013 Abstract #GP41A-1107
(American Geophysical Union, 2013).
11. Harland, W. B. The Geology of Svalbard (Geological Society, 1997).
12. Creveling, J. R. & Mitrovica, J. X. The sea-level fingerprint of a Snowball Earth
deglaciation. Earth Planet. Sci. Lett. 399, 74–85 (2014).
13. Lyons, W. B. et al. The McMurdo Dry Valleys long-term ecological research
program: New understanding of the biogeochemistry of the Dry Valley lakes:
A review. Polar Geogr. 25, 202–217 (2001).
14. Hoffman, P. et al. Are basal Ediacaran (635 Ma) basal ‘cap dolostones’
diachronous? Earth Planet. Sci. Lett. 258, 114–131 (2007).
15. Fairchild, I. J., Hambrey, M. J., Spiro, B. & Jefferson, T. H. Late Proterozoic
glacial carbonates in northeast Spitsbergen: New insights into the
carbonate-tillite association. Geol. Mag. 126, 469–490 (1989).
16. Bao, H., Fairchild, I. J., Wynn, P. M. & Spötl, C. Stretching the envelope of past
surface environments: Neoproterozoic glacial lakes from Svalbard. Science 323,
119–122 (2009).
17. Cao, X. & Bao, H. Dynamic model constraints on oxygen-17 depletion in
atmospheric O2after a snowball Earth. Proc. Natl Acad. Sci. USA 110,
14546–14550 (2013).
18. Le Hir, G. et al. The snowball Earth aftermath: Exploring the limits of
continental weathering processes. Earth Planet. Sci. Lett. 277,
453–463 (2009).
19. Pierrehumbert, R., Abbot, D. S., Voigt, A. & Koll, D. Climate of the
Neoproterozoic. Ann. Rev. Earth Planet. Sci. 39, 417–460 (2011).
20. Pollard, D. & Kasting, J. F. in The Extreme Proterozoic: Geology, Geochemistry,
and Climate (eds Jenkins, G. S., McMenamin, M. A. S., McKay, C. P. & Sohl, L.)
91–105 (Geophysical Monograph Series 146, American Geophysical
Union, 2004).
21. Le Hir, G., Ramstein, G., Donnadiieu, Y. & Goddéris, Y. Scenario for the
evolution of atmospheric pCO2during a snowball Earth. Geology 36,
47–50 (2008).
22. Hoffman, P. F. & Li, Z. X. A palaeogeographic context for Neoproterozoic
glaciation. Palaeogeogr. Palaeoclimatol. Palaeoecol. 277, 158–172 (2009).
23. Pierrehumbert, R. T. Climate dynamics of a hard Snowball Earth. J. Geophys.
Res. 110, D01111 (2005).
24. Spiegl, T. C., Paeth, H. & Frimmel, H. E. Evaluating key parameters for
the initiation of a Neoproterozoic Snowball Earth with a single Earth
System Model of intermediate complexity. Earth Planet. Sci. Lett. 415,
100–110 (2015).
25. Donnadieu, Y., Ramstein, G., Fluteau, F., Besse, J. & Meert, J. Is high obliquity a
plausible cause for Neoproterozoic glaciations? Geophys. Res. Lett. 29,
2127 (2002).
26. Paillard, D. Quaternary glaciations: From observations to theories. Quat. Sci.
Rev. 107, 11–24 (2015).
27. Rooney, A. D. et al. A Cryogenican chronology: Two long-lasting synchronous
Neoproterozoic glaciations. Geology 43, 459–462 (2015).
28. Abbot, D. S. & Pierrehumbert, R. T. Mudball: Surface dust and Snowball Earth
deglaciation. J. Geophys. Res. 115, D03104 (2010).
29. Abbot, D. S., Voigt, S. & Koll, D. The Jormungand global climate state and
implications for Neoproterozoic glaciations. J. Geophys. Res. 116,
D18103 (2011).
30. Rose, B. E. J. Stable ‘‘Waterbelt’’ climates controlled by tropical ocean heat
transport: A nonlinear coupled climate mechanism of relevance to Snowball
Earth. J. Geophys. Res. 120, 1404–1423 (2015).
Acknowledgements
This work was supported by the NERC-funded project GR3/ NE/H004963/1 Glacial
Activity in Neoproterozoic Svalbard (GAINS). Logistical support was provided by the
University Centre in Svalbard. This work was granted access to the HPC resources of
CCRT under allocation 2014-017013 made by GENCI (Grand Equipement National de
Calcul Intensif). We also thank D. Paillard and P. Hoffman for stimulating discussions
and valuable insights.
Author contributions
Field data were collected and analysed by I.J.F., D.I.B., E.J.F., M.J.H., E.A.McM., M.S.P.,
P.M.W. and C.T.E.S. Geochemical analyses were conducted by H.B. and P.M.W. Model
experiments were designed and conducted by G.L.H., Y.D., C.D. and G.R. The
manuscript and figures were drafted by D.I.B., I.J.F. and G.L.H., with contributions from
the other authors.
Additional information
Supplementary information is available in the online version of the paper. Reprints and
permissions information is available online at www.nature.com/reprints.
Correspondence and requests for materials should be addressed to D.I.B.
Competing financial interests
The authors declare no competing financial interests.
4NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience
© 2015 Macmillan Publishers Limited. All rights reserved© 2015 Macmillan Publishers Limited. All rights reserved© 2015 Macmillan Publishers Limited. All rights reserved© 2015 Macmillan Publishers Limited. All rights reserved

NATURE GEOSCIENCE DOI: 10.1038/NGEO2502 LETTERS
Methods
Sedimentology. Lithofacies were classified based on grain size, internal
sedimentary structures and deformation structures, and bounding surfaces.
Detailed stratigraphic logs were made in the field, supplemented by drawings and
photographs of key features. Samples were taken for polishing and thin sectioning,
to allow detailed examination of microstructures in the laboratory. In addition,
data were collected on clast lithology, shape, surface features and fabric.
Diamictites of the Wilsonbreen Formation are commonly very friable, allowing
included clasts to be removed intact from the surrounding matrix, allowing
measurement of both clast morphology and orientation, using methods developed
for unlithified sediments. Clast morphology (shape, roundness and surface texture)
was measured for samples of 50 clasts to determine transport pathways. Clast fabric
analysis was performed by measuring a-axis orientations of samples of 50 clasts
with a compass-clinometer, and data were summarized using the eigenvalue or
orientation tensor method. Oriented samples for measurement of anisotropy of
magnetic susceptibility (AMS) were collected using a combination of field-drilling
and block sampling. AMS was measured using an AGICO KLY-3 Kappabridge
operating at 875 Hz with a 300 A m−1applied field at the University of Birmingham
and an AGICO MFK-1A Kappabridge operating at 976Hz with a 200 A m−1
applied field at New Mexico Highlands University.
Geochemistry. Laboratory procedures for extracting, purifying and measuring the
triple oxygen (δ18O and 117O) and sulphur (δ34 S) isotope composition of
carbonate associated sulphate (CAS) in bulk carbonates are detailed in ref. 16.
Briefly, fresh carbonate-bearing rock chips were crushed into fine grains and
powders using mortar and pestle. Rinsing the fines with 18 Mwater revealed little
water-leachable sulphate in any of the Wilsonbreen carbonates. Subsequently,
about 10 to 30 g carbonates were slowly digested in 1–3 M HCl solutions. The
solution was then centrifuged, filtered through a 0.2µm filter, and acidified before
saturated BaCl2droplets were added. BaSO4precipitates were collected after >12 h
and purified using the DDARP method (see Supplementary Information). The
purified BaSO4was then analysed for three different isotope parameters: 117O, by
converting to O2using a CO2-laser fluorination method; δ18O, by converting to CO
through a Thermal Conversion Elemental Analyzer (TCEA) at 1,450◦C; and δ34 S,
by converting to SO2by combustion in tin capsules in the presence of V2O5
through an Elementar Pyrocube elemental analyser at 1,050 ◦C. The 117O was run
in dual-inlet mode, whereas the δ18O and δ34 S were run in continuous-flow mode.
Both the 117O and δ18 O were run on a MAT 253 at Louisiana State University,
whereas the δ34S was determined on an Isoprime 100 continuous-flow mass
spectrometer at the University of Lancaster, UK. The 117O was calculated as
117O≡δ017 O−0.52×δ018 O in which δ0≡1,000ln (Rsample/Rstandard ) and Ris the
molar ratio of 18O/16 O or 17O/16O. All δvalues are in Vienna Standard Mean Ocean
Water VSMOW and Vienna Canyon Diablo Troilite (VCDT) for sulphate oxygen
and sulphur, respectively. The analytical standard deviation (1σ) for replicate
analysis associated with the 117O, δ18O and δ34 S are ±0.05h,±0.5hand ±0.2h,
respectively. As the CAS is heterogeneous in hand specimens, the standard
deviation is for laboratory procedures. δ34S values were corrected against VCDT
using within-run analyses of international standard NBS-127 (assuming δ34S values
of +21.1h). Within-run standard replication (1 s.d.) was <0.3h. All geochemical
data are included in Supplementary Table 1.
Numerical modelling. Model runs were conducted with a coupled
atmospheric general circulation model (LMDz) and ice sheet model (GRISLI:
GRenoble Ice Shelf and Land Ice model). LMDz (spatial resolution 4◦in latitude
×5◦in longitude with 38 vertical levels) was run with prescribed continental ice
to climatic equilibrium. GRISLI has a 40km grid size and is driven with
downscaled climatic fields of surface air temperature, precipitation and
evaporation. To capture ice sheet–climate feedbacks, LMDz is rerun using the new
ice sheet distribution and topography. This procedure was repeated each 10 kyr to
investigate orbital forcing.
Surface mass balance (accumulation minus sublimation and melting) was
computed from monthly mean temperature, precipitation and evaporation rate.
Melt rate is calculated using the positive-degree-day method.
No sea ice dynamics treatment is specified, the sea ice cover is prescribed and a
thickness of 10 m is imposed. Ice albedo is fixed at 0.6, whereas snow albedo varies
from 0.9 from 0.55 as a function of the zenith and ageing process. Land ice/snow
free surface has the characteristic of a bare soil (rocky regolith) with an
albedo of 0.3.
Code availability. Code for the GCM LMDz can be accessed at:
http://lmdz.lmd.jussieu.fr. Code for the ISM GRISLI (GRenoble Ice Shelf and Land
Ice model) is not available.
Further details of the methods and modelling procedures are provided in the
Supplementary Information in the online version of the paper.
NATURE GEOSCIENCE |www.nature.com/naturegeoscience
© 2015 Macmillan Publishers Limited. All rights reserved© 2015 Macmillan Publishers Limited. All rights reserved© 2015 Macmillan Publishers Limited. All rights reserved© 2015 Macmillan Publishers Limited. All rights reserved