Snowball Earth: Testing the Limits of Climate Change

Paul Hoffman (Department of Earth & Planetary Sciences,

Harvard University, Cambridge, Massachusetts, USA)

Dept. of Geol. Sci., Ohio State University, Columbus, Ohio, USA

15 November 2001


Joe Kirschvink of Caltech came up with the Snowball Earth hypothesis in 1992.

Namibia - where most of the original evidence came from.  New places from which evidence comes include Spitsbergen, nw Canada (Mackenzie Mts.), Mauritania, Australia.

The Pleistocene ice advances reach about the same maxima and then snap back each time - are they the biggest glaciations ever?  The Pleistocene glaciations were the largest during the entire Phanerozoic.  There were dozens of ice advances in the last 2-3 m.y., not 2 or 3 advances.


Fragmentary evidence for glacial events are known for the Paleoproterozoic and there is well known and better evidence of glaciation in the Neoproterozoic.  Both appear to have had 100-200 m.y. durations, and both had multiple glacial events.  There was a 1.5 b.y. ice-age gap between them.

The Neoproterozoic glaciations seem to have occurred from ~780-580 m.y., although the interval could be as short as 100 m.y.  At least two glaciations occurred then, possibly four or five in the Neoproterozoic interval.


Douglas Mawson was the first to suggest the possibility of superglaciations.  Mawson is best known for his 1907-1920 work in Antarctica.  In a 15 July 1948 talk by Mawson, he proposed that the late Precambrian ice age was a global ice age, and the disappearance of the ice age resulted in the explosion of fossils we see at the beginning of the Paleozoic.  See J. Proc. Roy. Soc. NSW 82: 150-174 (1949).


Mackenzie Mountains - 200 meters worth of diamictite is sandwiched between carbonates (with sharp contacts).  Carbonate rocks are usually low-latitude & warm-water deposits.  Ice sheets were running across pure carbonate shelves.  Another general feature of the Neoproterozoic glaciation is the abruptness of the end of the interval and the immediate deposition of overlying carbonates, with no evidence of erosion in between.  This is a universally observed, sudden transition from glaciation to warm water carbonate deposition.  The overlying carbonate rocks are called cap carbonates.  Cap carbonates are usually thick successions, representing deepwater facies (10s to 100s of meters of water).  The depth was created by the glacial event & got rapidly filled in by carbonates.


These Neoproterozoic glacial deposits are widespread - they are known in North America, South America, Africa, Arabia, Europe, Asia, Australia, Greenland, Scandinavia, China, Siberia, central Asia (not Antarctica yet).


W. Brian Harland - interested in determining paleolatitudes of these Neoproterozoic deposits.  See Geol. Rundschau 54: 45-61 (1964) and Scientific American 211: 28-36 (1964).  He found that many of these Neoproterozoic glacial deposits formed at high latitudes.  The thought then was: normal remnant magnetism in rocks was believed to be primary if it hadn’t been heated beyond the Curie Point.  Now, though, we know remagnetization can be done by low-temperature chemical magnetization.

Can do a fold test to see this.  Can also do a reversal test.  Do both for determination of 1˚ vs. 2˚ magnetization for the rocks you’re interested in.


Flinders Ranges, Australia - the Elatina Formation (glacial interval) in the Neoproterozoic has been determined to be within 10˚ of the paleoequator.  The Elatina laminations are tidal - the best tidal rhythmites in the world - allows for accurate calculation of the Earth-Moon distance, to within 100 km.  The Elatina rocks are marine, since the laminations are tidal.  Glaciers thus descended to sea level at very low latitudes.  The magnetization of the Elatina Formation was found to be 1˚ - with sedimentation.


David Evans (2000) - American Journal of Science 300: 347-433.  He has a histogram of occurrences of different glacial units formed at or close to sea level at different latitudes.  Lots occur <10˚ from the equator.  There are some from 40˚-60˚ from the equator.  There are none >60˚ from the equator.  This is similar to the histogram of latitudinal trends of evaporation/precipitation - lots near the equator, less at mid-latitudes.  The expectation of glacial deposits near the poles is not seen here.


Kirschvink (a biogeologist and paleomag. guy at Caltech) tested the Elatina Formation and agreed that ice did go to the equator in the Neoproterozoic.  Though a fanciful idea, this is a well-known concept to climate modelers and climate physicists.

Heat absorbed = Heat emitted

πR2Es[1-a] = 4πR2[fsTs4]

where R = radius of the Earth

            Es = solar irradiance

            a = planetary albedo

            f = effective infrared transmission factor (greenhouse effect)

            s = Stefan-Boltzman constant

            Ts = surface temperature


Runaway ice albedo - ice reflects solar radiation and further cooling results (positive feedback).  Effect on a spherical Earth - the strength of the feedback increases as ice reaches further from the pole.

Instability in the climate system does occur in this ice-albedo feedback - there is a point where the runaway ice albedo positive feedback becomes unstoppable - ice would freeze over even the equator, resulting in an ice-covered world.  The instability point is where ice reaches ~30˚ from the equator.

See Caldeira & Kasting (1992) - Nature 359: 226.

Ikeda & Tajika (1999) - Geophys. Res. Letters 267: 349.


Snowball Freeze Scenario

1) large polar sea - ice caps - most continents are on the equator in the Neoproterozoic.  A little over 0˚ C glacial mean temperature - bright whitish “land” where there would otherwise be water.  Sea level drops - more true land is exposed, and more albedo.  Equatorial continental distribution makes for a colder Earth.


2) runaway ice-albedo feedback - ice on upland areas - tropical ocean totally frozen or not?  Some open water in the tropics?  Planetary albedo goes up to 0.5 or 0.6.  Have 1-1.5 km thick ice sheets overall.

Climate physicists never believed that the Earth did ever experience this.  This was considered to be a terminal condition.  To escape from this, solar flux has to be raised 25%.  Obviously, we did escape.


Iron formations are associated with the Neoproterozoic glacial events.  This is strange, since BIFs were largely gone by then.  All Neoproterozoic ironstones are within glacial intervals.


The escape of the Snowball Earth condition was plate tectonics.  See geologic carbon cycle by Walker et al. (1981).  In a Snowball Earth, plate tectonics continues - get a slow increase in CO2 in the atmosphere (from volcanoes), without the usual CO2 sinks (no photosynthesis on a Snowball Earth and no silicate weathering on the land).  So, there is a continual source of CO2 (volcanism) and no sinks, resulting in rising CO2 levels after the onset of Snowball Earth.  Need to reach high CO2 levels in the air to overcome a 0.6 planetary albedo.  The amount you need is 120,000 ppm CO2 (12% CO2) in the air.  These levels would be reached in ~4 m.y. (rough estimate).


There is an escape to the Snowball Earth - CO2 rises slowly, to where the melting point is reached in the tropics, which becomes an area that resists perennial ice growth.  Once melting starts, reverse albedo kick in - dark water + high CO2 results in faster and faster melting.  Under these conditions, global sea ice 300-400 meters thick would disappear in 100-1000 years.  Ice disappears far faster than CO2 is consumed.  This results in a transient ultragreenhouse effect.  The Earth reaches the ice-free branch and surface temperatures skyrocket.  Estimated surface sea temperatures at the tropics were 40-50˚ C (120-130˚ F).


Intense evaporation of sea water - a strong hydrologic cycle kicks in - rain is pouring down on the landscape after being glacially altered.  The landscape becomes an intense chemical reaction factory - silicate weathering levels are high, and there is a big flush of alkalinity into the oceans.  So, there is an abrupt beginning to Snowball Earth and a gradual amelioration, but reaching a point were the tropical oceans begin to open, resulting in a sudden transient ultragreenhouse effect.  Then CO2 is consumed by silicate weathering on the newly-exposed land, and CO2 gradually return to steady state levels.


Kirschvink presented this in 1989, and published it as a chapter in the Proterozoic Biosphere book in 1991.  Kirschvink suggested some tests of this hypothesis to see if the Snowball Earth hypothesis was right:

1) Neoproterozoic glaciations should by synchronous.  It has been tough getting geochronologic dating on these deposits.  Even naysayers agree that these glacial deposits are correlatable, though, based on coincidence with isotope curves.  These deposits are global events.  Carbon isotopes, though ambiguous, are robust recorders of events - it is hard to change isotopes, and anomalies at this time are very large.  Secular variations in carbon isotopes (in carbonates and in organic matter) are due to burial flux.  Volcanic carbon input = -5 to -6 d13C VPDB (‰).  When carbon is fixed by organism, the organic matter is depleted ~3% in carbon-13, compared with dissolved carbon in water.  The more you bury organic matter, then carbonate d13C (= dissolved carbon in water) gets higher.  Get heavier residual seawater occurring when organic matter burial rates increase.  Actually, not when organic matter burial rates increase, but when you get an increase in the ratio of Corg/Ctotal burial flux.  dcarb varies ~3 per mil in the Phanerozoic.  The Proterozoic values vary 5-10 per mil.  The Cretaceous-Tertiary event saw only a 2 per mil change.  These unusual events correlate.  All agree that these isotope changes (dramatic changes at that) are correlatable - at the worst, they are not strongly diachronous.  Reliable records, though, depend on sound understanding of the stratigraphy.  Hoffman & company spent many years doing Neoproterozoic stratigraphy in the Otavi Group of Namibia.


2) Expect to see a sudden lithologic change associated with the onset and end of Snowball Earth.  Kirschvink didn’t originally know about cap carbonates.  Having glacial deposits sharply but conformably overlain by warm-water carbonates - this is a predictable consequence of the Snowball Earth Hypothesis, not a paradox.  Get a tremendous alkalinity flux into the oceans after glaciation - this drives precipitation of CaCO3 - cap carbonates have to be there, according to the model.  Only silicate weathering brings CO2 down to steady state levels.  Cap carbonates have dolomite at the base and limestone above that.  The limestone was originally crystals of aragonite growing on the seafloor.  We know this based on primary fabric details that are preserved.  This indicates high rates of CaCO3 precipitation due to supersaturation of carbonate in the water.  Anything that sticks up off the seafloor is in higher alkalinity, and will have aragonite growth.  Rapid deposition of cap carbonates is thy key to understanding the carbon isotope curve changes.  A -5 to -6 per mil value in carbon isotopes is close to mantle values (= volcanically derived CO2).  What does this mean?  A dead ocean & little to no organic production/photosynthesis?  It seems so.  So, oceans get carbon isotope values of -5 to -6 per mil.  However, the lowest values are not at the base of the cap carbonates.  Was organic productivity fine (“normal”)?  There may have been lots of productivity but organic matter burial rates are low because carbonate precipitation rates are so high.  Low carbon isotope values in cap carbonates are due to the overwhelming of organic matter burial by carbonate production.  Atmospheric source gets carbon-12 rich because HCO3- conversion from CO2 includes an 8 per mil shift in preference for carbon-13.  Maielberg cap carbonates - the base of the succession has a carbon isotope value of -2 d13C.


3) Expect little air-sea gas exchange in a Snowball Earth.  Since there are lots of O2 sinks in the ocean (Fe in vent areas, for example), oceans will get anoxic.  In the absence of molecular O2, Fe2+ gets widely distributed in the ocean water - this should explain the presence of iron formations.  Lots of Fe in solution in anoxic water.  Ferric oxide formation requires an absence of sulfide (derived from sulfate as a weathering product from the land, reduced by bacteria to sulfide - this can’t be happening in order to get ferric oxide formation & ironstone formation).  So, can get lots of ferric oxide formation.  The last BIFs are at 1.8 by.  But, after a 1.2 by-long hiatus, get new BIFs associated with Neoproterozoic glacial deposits.  Not all Neoproterozoic glacial deposits have associated iron formations, though.  Within the cap carbonates of the Mackenzie Mountains and in Australia, the upper dolomite part and the lower limestone part of the cap carbonate have a 40-50 cm thick layer of 1˚ barite at the dolomite-limestone boundary.  It looks like stromatolites.  It is primary!  Barite is an incredibly insoluble mineral.  Need to boil barite in HF for a week to dissolve it.  It always occurs between the dolomite and the limestone.  Why is it there?  Can’t transport lots of Ba into modern oceans, due to the presence of sulfate.  If there is little or no sulfate in the oceans, can get an oxic/anoxic chemocline.  If that chemocline stabilizes, below the chemocline will be Ba-rich waters.  Sulfate ions are poisoners for dolomite formation.  Below the chemocline, barite waters lacked sulfate and above the chemocline, the water had sulfate.  Dolomite will have faithful carbon isotope values.  People assume all dolomite is secondary, but here it is primary - get fractionation that is not usually accounted for by isotope stratigraphers.


Neoproterozoic enigmas - worldwide distribution of glacial deposits, low-latitude ice lines at sea level, Fe-formations with ice-rafted debris, post-glacial cap carbonates, and multicellular animal evolution within 20 my of the end of the Snowball Earth.


Why didn’t these Snowball Earth events occur all the time?  Need an equatorial distribution of continents - this hasn’t happened yet in the Phanerozoic.  Equatorially-distributed continents inevitably results in a colder Earth.  Normally, land in the tropics has higher weathering rates, resulting in CO2 levels dropping and temperatures dropping, which is a check on weathering rates.  In addition to albedo effects, equatorially distributed continents results in extensive polar ice caps.  Then, the atmosphere will be drier (less latent heat in it), and atmospheric energy is reduced, and circulation is less effective in transporting heat from low latitudes to high latitudes.  The tropics will be warmer but the poles will be colder, and will be more susceptible to albedo runaway.


What triggers individual glaciation events?  Can see a 10 per mil drop in carbon isotopes before the onset of glacial deposits.  This indicates a changing input of carbon, likely from generation of methane (from seafloor clathrates) - methane has a significant greenhouse forcing effect.  The carbon cycle responds to this - CO2 levels go down (because it’s warmer and weathering rates go up, which is a CO2 sink).  If you get an interruption in methane supply, the methane already in the air oxidizes to CO2 (a weaker greenhouse gas).  All of this is from an unusual Neoproterozoic paleogeography which allows a Snowball Earth to happen.


Darwin: bad hypotheses can’t explain lots of facts.


Weaknesses of the Snowball Earth hypothesis - the results/conditions are so far outside what we know, so we have to figure out things based on first principles only.


Lesson: Juicy stuff remains to be found, even after 200 years of geologic investigations.



Home page