The Cryogenian
Neoproterozoic Snowball Scenarios

"Snowball" Scenarios of the Cryogenian


There is now good evidence that at least three Proterozoic ice ages culminated in glaciation extending to very low latitudes; possibly to the equator. Joseph Kirschvink (1992) coined the term 'snowball Earth' to describe these events, suggesting that if ice had also covered the oceans at these times, then the Earth might have resembled a highly reflective snowball (p. 52).

Evidence in support of the snowball hypothesis has been accumulating since the 1960s, although the two landmark papers on the subject, by Kirschvink and Hoffman et al., did not appear until the 1990s. Although these are powerful theories which potentially explain a number of disparate and otherwise problematic phenomena, none of the several versions of the snowball hypothesis yet provides a complete explanation of all field observations.


“In 1891 the Norwegian geologist Hans Henrik Reusch found an ancient deposit he interpreted as being a glacial moraine. The deposit, now believed to be a tillite, lay atop a striated rock surface beside Varanger Fjord in northern Norway. Both the tillite and the rock surface are demonstrably Pre-Cambrian” (Harland & Rudwick, 1964). A few years later, around 1900, Sir Douglas Mawson recognised the glaciogenic nature of the Sturt Tillite, a few kilometres south of Adelaide in South Australia. Since then, Late Proterozoic glaciogenic sequences have become known from almost all of the major cratonic areas, including North America, the Gondwana continents, and the Baltic Platform.

Although the earliest snowball event or events may have occurred as early as 2,300 to 2,200 Ma (Kirschvink, 2002), our reconstruction of these ancient times is still not clear. Far stronger evidence supports recognition of three Neoproterozoic events: Sturtian, Marinoan and Varanger. Corsetti & Kaufman (1999); Rice et al. (2003). Some publications (e.g. Narbonne, 1998: fig. 2) suggest the possibility of multiple individual glaciations in the Sturtian, usually by means of cryptic indicator symbols on time charts; it is difficult to guess how to interpret such typography.

Evidence for Glaciation at Low Paleolatitudes

Whereas the Permian and Quaternary glacial deposits formed at relatively high latitudes, those of the Proterozoic are believed to have formed much closer to the equator. In 1992, Joseph Kirschvink (1992) coined the expression “snowball earth” to evoke his conjectured appearance of a fully glaciated planet.

Arguments in support of this contention have been repeatedly advanced since the 1960s (e.g. Harland & Rudwick, 1964; Harland, 1964; Hambrey & Harland, 1981; Kirschvink, 1992; Hoffman et al., 1998), prompted by observations that some of the diamictites contain an unexpected abundance of carbonate debris, presumably derived from nearby carbonate platforms. Additionally, many of these units are bounded above and below by thick carbonate sequences which today are known to form only at tropical latitudes, within about 33° of the equator. Ziegler et al. (1984); Kirschvink (1992). Other anomalies include dropstones and varves in the carbonates, evaporites, and anomalous iron enrichment (iron-rich mudstones and even some BIFs). The iron deposits should have been able to form only if the contemporary Proterozoic oceans contained little or no dissolved oxygen, but by that time the atmosphere is believed to have had nearly the same composition as it has today.

Early paleomagnetic data presented in support of low-latitude interpretations were famously suspect, particularly in terms of constraining the time at which remanent magnetisation was acquired. (Two important exceptions to this were the reports of Embleton & Williams, 1986 and Sumner et al. 1987, for the varved sediments of the Elatina Formation of South Australia.)  However, subsequent studies have confirmed the equatorial placement of Rodinia. Kirschvink concludes that, “During the uppermost Marinoan glaciation in Australia, it now seems clear that these extensive, sea-level deposits (including varves and dropstones) were formed by widespread continental glaciers which were within a few degrees of the equator. The data are difficult to interpret in any fashion other than that of a widespread, equatorial glaciation” (Kirschvink 1992, p. 51).

“The distribution of Infra-Cambrian water-deposited tillites is almost worldwide. Whether they are considered according to the present position of the continents or according to a possible Pre-Cambrian arrangement, … it is difficult to confine them … to a restricted portion of the globe. There are two alternative hypotheses to account for this fact. One states that the ice was widespread at all latitudes” (Harland & Rudwick 1964, p. 33).

Diamictite: massive, structureless unsorted rock. The sort of mess a really big glacier would leave.

Drop stone: stones carried along by glaciers and dropped when the bottom of the glacier melts over water.

Varve: A rhythmic sequence of sediments deposited in annual cycles in glacial lakes. Light-colored, coarse summer grains are deposited by rapid melting of the glacier. The summer layers grade upward to layers of finer, dark winter grains of clay minerals or organic material that are deposited slowly from suspension in quiet water while streams and lakes are icebound. Oilfield Glossary- Term 'varve'

Marinoan: Early Ediacaran (Vendian), about 580 Mya. This was the most recent of the putative three snowball events. The others occurred in the "Varanger" (~600 Mya, Cryogenian - Ediacaran boundary) and the Sturtian (~800 Mya, Early Cryogenian).

Initial Scepticism

Initial reluctance to accept that glaciations had extended to low latitudes was largely sustained by two theoretical difficulties:

Joseph Kirschvink (1992) appears to have been first to offer both.

Possible Causal Mechanisms for Low-Latitude Glaciation

Early converts to the hypothesis offered a variety of causal explanations.

One suggestion, proposed in Williams 1975, was that the obliquity of the Earth's orbit may have been greater than its present value of around 20°. Were it to exceed about 54°, the Sun would heat the poles more than the equator. Glaciers might then form in the equatorial regions. This proposal, however, poses more difficulties than it resolves: Whereas the physical basis for the Milanković-scale changes (a few degrees with a period of a few tens of thousands of years) is fairly well understood, no mechanism has yet been proposed that would lead to the much larger oscillations required by the Williams hypothesis. Moreover, detailed studies of both modern and ancient heliotropic stromatolites (Vanyo & Awramik, 1982; Vanyo & Awramik, 1985; Awramik & Vanyo, 1986; and Vanyo et al., 1986) argue convincingly that the obliquity at 800 Ma was in the range of the present values. Kirschvink, 1992: 51).

“[Another] possibility to consider is that the Neoproterozoic sun was weaker by approximately 6 percent, making the earth more susceptible to a global freeze. The slow warming of our sun as it ages might explain why no snowball event has occurred since that time. But convincing geologic evidence suggests that no such glaciations occurred in the billion or so years before the Neoproterozoic, when the sun was even cooler”  Hoffman & Schrag (2000: 75).

Paleogeographic reconstructions for this time indicate that the bulk of the continental land mass probably lay in middle to low latitudes during the late Precambrian, a paleogeography which has not recurred subsequently.  Kirschvink suggests, “[i]n a qualitative sense, this could have had a fundamental impact on global climate, as most of the solar energy adsorbed by the earth today is trapped in the tropical oceans (in contrast to the continents which are relatively good reflectors) and in high latitude oceans which often have fog or other cloud cover. Furthermore, if extensive areas of shallow, epicontinental seas were within the tropics, a slight drop in sea level would convert large areas of energy-absorbing oceanic surface to highly reflective land surface, perhaps enhancing the glacial tendency.”  Kirschvink (1992: 51-52).

When a significant proportion of the continental landmass lies near the poles, as it does today, carbon dioxide in the atmosphere remains in high enough concentrations to keep the planet warm. When global temperatures drop enough that glaciers cover the high-latitude continents, as they do in Antarctica and Greenland, the ice sheets prevent chemical erosion of the rocks beneath the ice. Moreover, ice-cover is inhospitable to most plant life, so photosynthesis is also inhibited. With the principal carbon sinks suppressed, the carbon dioxide in the atmosphere rises to a level high enough to fend off the advancing ice sheets, maintaining an equilibrium. If all the continents cluster in the tropics, on the other hand, they would remain ice-free even as the earth grew colder and approached the critical threshold for a runaway freeze. The carbon dioxide ‘safety switch’ would fail because carbon burial continues unchecked. (Partly after Hoffman & Schrag 2000: 75).

Note, however, that the onset of glaciation may have begun with the break-up of Rodinia, some 750 Ma (Walker, 2003: 241), an inconsistency which requires some accommodation from current hypotheses.

Mechanisms Permitting Thawing

Having found a plausible mechanism to permit low-latitude glaciation, there remains the opposite problem: although the snowball events appear to have lasted a very long time, obviously they did end, and more quickly than continental rafting away from the equator alone would permit.

Kirschvink appears to have been first to suggest that the reversal of ice house conditions could be effected, “through the gradual buildup of the greenhouse gas, CO2, contributed to the air through volcanic emissions. The presence of ice on the continents and pack ice on the oceans would inhibit both silicate weathering and photosynthesis, which are the two major sinks for CO2 at present. Hence, this would be a rather unstable situation with the potential for fluctuating rapidly between the ‘ice house’ and ‘greenhouse’ states.”  Kirschvink (1992: 52), the onset of the latter possibly occurring in as little as a few hundred years. Hoffman & Schrag (2000: 68).

An interesting constraint on this hypothesis is provided by relatively high the freezing point of CO2: about -78° C.  If the polar areas were consistently colder than this, they would form an additional CO2 sink.

“With this greenhouse scenario in mind, climate modelers Kenneth Caldeira of Lawrence Livermore National Laboratory and James F. Kasting of Pennsylvania State University estimated in 1992 [Caldeira & Kasting (1992)] that overcoming the runaway freeze would require roughly 350 times the present-day concentration of carbon dioxide [~0.12 bar]. Assuming volcanoes of the Neoproterozoic belched out gases at the same rate as they do today, the planet would have remained locked in ice for up to tens of millions of years before enough carbon dioxide could accumulate to begin melting the sea ice. A snowball earth would be not only the most severe conceivable ice age, it would be the most prolonged.”  Hoffman & Schrag (2000: 72).

Phenomena Relating to the Global Snowball Model

The test of any historical model lies in its consistency with observation – past and future – which is a form of prediction. The global snowball model is potentially very powerful, providing a unifying explanation for a number of diverse phenomena. Further, several implications may themselves be testable.

Banded Iron formations "occur in Proterozoic rocks, ranging in age from 1.8 to 2.5 billion years old. They are composed of alternating layers of iron-rich material (commonly magnetite) and silica (chert). Each layer is relatively thin, varying in thickness from a millimeter or so up to several centimeters. Here is one theory as to how they might have formed: It is theorized that the Earth's primitive atmosphere had little or no free oxygen. In addition, Proterozoic rocks exposed at the surface had a high level of iron, which was released at the surface by weathering. Since there wasn't any oxygen to combine with it at the surface ... the iron entered the ocean as iron ions. At the same time, primitive photosynthetic blue/green algae were beginning to proliferate in the near surface waters. As the algae would produce O2 as a waste product of photosynthesis, the free oxygen would combine with the iron ions to form magnetite (Fe3O4), an iron oxide. This cleansed the algae's environment. As the biomass expanded beyond the capacity for the available iron to neutralize the waste O2 the oxygen content of the sea water rose to toxic levels. This eventually resulted in large-scale extinction of the algae population, and led to the accumulation of an iron-poor layer of silica on the sea floor. As time passed and algae populations re-established themselves, a new iron-rich layer began to accumulate. Unfortunately, the algae were of relatively low intelligence and were unable to learn from their past excesses (this was also before the EPA), so they would again proliferate beyond the capacity of the iron ions to clean up their waste products, and the cycle would repeat. This went on for approximately 800,000,000 years!"  GeoMan's Banded Iron Page.

Snowball Events

Although there is some limited evidence for a profound global ice age at about 2,300 to 2,200 Ma (Kirschvink, 2002), the well-documented events are all Neoproterozoic.

~710 to 680 Ma: The Sturtian Glaciation(s)

Rice et al. (2003) concludes that glacial deposits corresponding to the earliest (Sturtian) glaciation are absent in Norway, Svalbard, eastern Greenland, Scotland and Death Valley. “However, cap-carbonates to this glaciation can be recognized in many sequences, based on the isotopic and sedimentological characteristics of the Sturtian cap-carbonates in Namibia (Rasthof), NW Canada (Rapitan), and South Australia (Sturt). In all these cap-carbonates, δ13C rises sharply from –4‰ to +5‰ in relatively organic-rich sediments. Probable Sturtian cap-carbonates, without underlying diamictites, include the lower Russøya Member from Svalbard and the lower Beck Springs Formation from Death Valley.”  Rice et al. (2003).

605 to 585 Ma: The Varanger-Marinoan Ice Ages

The Marinoan glaciation is the most widespread and most easily recognised of the snowball events. Unlike the earlier Sturtian glaciation, the Marinoan was presaged by a large (up to 15‰) though gradual decline in δ13C. Unequivocal Marinoan deposits include the Ghaub (northern Namibia), Elatina (South Australia), and Ice Brook (north-western Canada) formations, all of which are the higher of two diamictites.

Marinoan glacial deposits are overlaid by a distinctive transgressive, laminated cap-dolostone, which variably contains isopachous cements, accretionary oscillation megaripples, tubestones, and peloids. The cap-dolostone is bounded above by a flooding surface that corresponds to an increase in the fraction of siliciclastic sediments and, commonly, a shift to from dolomite to calcite. In some successions, seafloor barite and aragonite cements occur at this transition.

Throughout the cap-dolostone, δ13C remains consistently in the range -2 to -4‰.

Applying these unique isotopic and sedimentological boundary conditions as correlation tools, Rice et al. (2003) concluded that the diamictite pairs Petrovbreen + Wilsonbreen (northeast Svalbard), Ulvesø + Storeelv (eastern Greenland), and Surprise + Wildrose (Death Valley) are jointly Marinoan in age. These criteria also indicate that the thick Port Askaig (750m) and Smalfjord (420m) diamictites in Scotland and Norway, respectively, are Marinoan. These correlations are important because, in both cases, the Marinoan diamictite is the lower of two glacial horizons. Thus, it is concluded that the upper diamictites in Norway (Mortenses) and Scotland (Loch na Cille) correspond to a third glaciation: the Varangian.

Varangian glacial deposits are not widespread, but overlie and appear to be related to the largest Neoproterozoic negative δ13C anomaly (-8‰). This shows up globally between Marinoan and Ediacaran-aged strata (e.g. the Wonoka Formation in South Australia and the Huqf Group in Oman).

(After Rice et al. 2003.)

Biological Consequences


It has been suggested that the time of the Varanger-Marinoan glaciations, which lasted from approximately 605 to 585 Ma (Martin et al., 2000), was an interval of widespread extinction, a contention based mainly on carbon isotopic profiles, which “display strong negative as well as positive excursions. Negative excursions are specifically associated with the major ice ages that mark immediately pre-Ediacaran time. Much research is currently focused on this unusual coupling of climate and biogeochemistry, and both paleoceanographic models and clustered phytoplankton extinctions suggest that these ice ages had a severe impact on the biota – potentially applying brakes to early animal evolution.”  Knoll & Carroll (1999: 2135).

Acritarchs are sometimes supposed to have been major victims of a mass extinction, around 610 Ma, perhaps associated with the glaciations, when some estimates suggest that up to 70% of taxa went extinct. Interestingly though, the late Gonzalo Vidal, perhaps the most widely quoted and respected of researchers into Precambrian acritarchs, while acknowledging the very low diversity of acritarchs reported from this interval, also points out the scarcity of rocks likely to yield good acritarch assemblages, and stops short of any causal speculation. Vidal (1981).

When considering a possible mass extinction at this time, it must be remembered that the Twitya fauna, Aspidella terranovica and Nimbia occlusa, or at least forms which left behind indistinguishable fossils, passed through the Varanger-Marinoan.

Effects on Metazoan Evolution

The Sturtian snowball period is shortly succeeded by the earliest unambiguous record of metazoan animals and, after an additional 170 Ma and two more low-latitude glaciations, by the appearance of shelly Cambrian faunas. Thus, it was an interesting stage in the evolution of multicellular animals, posing not only the question of how early life survived under such environmental stress (Hyde et al., 2000) but whether the glaciations actually acted to shape metazoan evolution in some way, as first proposed by Martin Rudwick in the 1960s. See, e.g. Harland & Rudwick (1964: 36): “[A]t the end of the ice age, the improvement in climate and the rise of the sea level would have re-created a variety of favourable but biologically empty environments, in which the opportunity would exist for radical evolutionary changes to take place.”

This argument is rather compelling: “Explosive” radiations following mass extinction events are well-documented from the Phanerozoic so it is tempting to extend the snowball earth speculation to suggest that these evolutionary changes were actually driven by the glaciations – “the periodic removal of all life from higher latitudes would create a series of post-glacial sweepstakes, perhaps allowing novel forms to establish themselves, free from the competition of a preexisting biota.”  Kirschvink (1992: 52).

With their usual flair for the dramatic, Hoffman & Schrag (2000: 74) observe (not quite accurately): “Eukaryotes … had emerged more than one billion years earlier, but the most complex organisms that had evolved when the first Neoproterozoic glaciation hit were filamentous algae and unicellular protozoa. It has always been a mystery why it took so long for these primitive organisms to diversify into the 11 animal body plans that show up suddenly in the fossil record during the Cambrian explosion. … A series of global freeze-fry events would have imposed an environmental filter on the evolution of life. All extant eukaryotes would thus stem from the survivors of the Neoproterozoic calamity.”

Some evidence for the extent of eukaryotic extinctions may be evident in the universal tree of life. Hoffman & Schrag (2000) propose that eukaryotic lineages may have been ‘pruned’ during the snowball earth episodes – a concept seemingly akin to Gould’s (1989) ‘decimation by lottery’ – which is certainly plausible. However, their supporting contention that universal trees “depict the eukaryotes’ phylogeny as a delayed radiation crowning a long, unbranched stem” (p. 74) is neither strong evidence for pruning per se, nor is the claim consistent with any published molecular biology as far as I am aware. My own understanding of the literature is that this model of eukaryote phylogeny is unique to Hoffman & Schrag, and contradicted by the overwhelming bulk of published research.

More plausibly, Hoffman & Schrag suggest that, in the face of varying environmental stress, many organisms respond with wholesale genetic alterations. Severe stress encourages a great degree of genetic change in a short time, because organisms that can most quickly alter their genes will have the most opportunities to acquire traits that will help them adapt and proliferate.

Widely separated refuge communities could accumulate genetic diversity over millions of years. When two groups that start off the same are isolated from each other long enough under different conditions, chances are that independent mutation will produce new species. Repopulations occurring after each glaciation would come about under unusual and rapidly changing selective pressures quite different from those preceding the glaciation; such conditions would also favour the emergence of new life-forms. After Hoffman & Schrag (2000: 75).

However, their view does find an echo in Ernst Mayr’s otherwise inexplicable comment that “diversity of the early eukaryotes seemingly remained rather low for the period from 1,700 to 900 million years ago, but then rose rapidly to experience a veritable explosion of protistan microfossils during the Cambrian.”  Mayr (2001: 48-50).


Founder communities must have survived the snowball events, perhaps in a variety of habitats. Psychrophilic (cold-loving) representatives are today known from among the cyanobacteria, dinoflagellates, and some algae, which can live in snow and on the surfaces of rock particles in floating sea-ice. Less cold-tolerant organisms may have held out in locations where geothermal action preserved warm micro-climates – some perhaps from around deep-sea fumaroles, though photoautotrophs must clearly have ‘over-wintered’ elsewhere. Hoffman & Schrag (2000: 74) make the reasonable point that the steep and variable temperature and chemical gradients endemic to ephemeral hot springs would preselect for survival in the runaway greenhouse conditions which they postulate to succeed the snowball events.

However, we may not yet need to invoke the image of a few last bastions of life huddled around some deep-sea vent. It is not clear what fraction of the equatorial oceans in deep water would form pack ice, as these zones would still absorb large amounts of the incident solar radiation, perhaps enough to prevent ice formation. Hence, we might expect to find some warm tropical “puddles” in the sea of ice, shifting slightly from north to south with the seasons. In turn, this should produce extreme climatic shifts in some local areas. Kirschvink (1992: 52). A variant of this supposition finds support from the results of computer simulations with a coupled climate/ice-sheet model, reported in Hyde et al. (2000): “To simulate a snowball Earth, we use only a reduction in the solar constant compared to present-day conditions and we keep atmospheric CO2 concentrations near present levels. We find rapid transitions into and out of full glaciation that are consistent with the geological evidence. When we combine these results with a general circulation model, some of the simulations result in an equatorial belt of open water that may have provided a refugium for multicellular animals.”

“Although the extent of glaciation remains uncertain, if the protostome-deuterostome divergence occurred before these world-wide glaciations, they are likely to have imposed a severe ecological constraint on the forms that could have survived. Runnegar (2000) argued that conditions even within the refugia would have allowed survival only of small, simply constructed, pelagic bilaterian stem group forms (such as proposed for the remote ancestors of the Bilateria by Davidson et al., 1995). Evolution of adult body plans in the bilaterian stem group would have had to await the more favorable late Neoproterozoic environments.”  Erwin & Davidson (2002: 3024).

Also see Runnegar (2000); Peterson & Davidson (2000).

Conclusion: Current Status of the Snowball Hypotheses

Overall, the various snowball earth hypotheses have potential to explain diverse observations of the Proterozoic geological record: synchronous low latitude diamictites associated with carbonate deposits, carbon isotope excursions, banded iron formations, and so on. Nevertheless, I feel Hoffman & Schrag (2000: 74) somewhat overstate their case with the claim that the “strength of the hypothesis is that it simultaneously explains all these salient features, none of which had satisfactory independent explanations.”  As Kirschvink (2002: table 1) makes clear, neither of the major variants – ‘hard’ snowball or slushball – is presently able to explain all observations, and “a more complex scenario may be closer to the actual truth than any of the discrete models” proposed to date.

The link with evolutionary phenomena, though tantalising, is at this time a less well-developed speculation.

Chris Clowes 0309xx

Related Topics

Further Reading

Related Pages

Other Web Sites

contact us

page uploaded 15 May 2002
checked ATW051009, edited RFVS111203
© Chris Clowes 2003