Snowball Earth Hypothesis

The hypothesis

The snowball Earth hypothesis (SEH) suggests that the Earth experienced surface temperatures so low that virtually its entire surface was covered by glaciers and/or thick sea ice periodically during its early history. Such a condition has been hypothesized for parts of the Neoproterozoic Era from about 750 million years ago (Ma) to about 600 Ma (Figure S21). A similar frozen state has been proposed during the early part of the Paleoproterozoic Era at about 2,300 Ma. There is little evidence of glaciation in the long intervening period but recently Williams (2005) presented evidence for glaciation in the Kimberley region of Western Australia at around 1,800 Ma.

Figure S21

Distribution of glaciations throughout geologic time. Note the large gap between the Paleo- and Neoproterozoic glaciations. Question mark indicates uncertainty concerning the number of Neoproterozoic glacial episodes.

The evidence

Support for this hypothesis comes primarily from field evidence, including the widespread preservation of diamictites, which are conglomerates containing large rock fragments separated by abundant fine grained matrix materials in which they appear to “float.” Included rock fragments commonly have scratched or striated surfaces. Rock fragments with planar surfaces are said to be “faceted.” The presence of many faceted and striated clasts in a diamictite suggests a glacial origin. Another glacial indicator is the preservation of large scale striated or furrowed rock surfaces that are thought to have formed by grinding action as rock-charged glaciers moved over older rocks or penecontemporaneously deposited sediments. Glacial activity is also indicated by the presence of finely bedded (laminated) sedimentary rocks that contain isolated large rock fragments. These fragments commonly show evidence of vertical emplacement in the form of downward penetration of the clast into enclosing sedimentary layers or as “splash up” structures – upfolding of the finely laminated sedimentary rocks on the sides of the fragment. Such rocks are called “dropstones” and are thought to have been released from melting icebergs.

These kinds of evidence have been found in Neoproterozoic rocks on all of Earth’s continents. Rocks with similar glaciogenic characteristics have also been described from older (∼2,300 Ma) successions on several continents. The older glacial deposits are best known from North America, where they occur in the Huronian Supergroup on the north shore of Lake Huron in Ontario, Canada. Similar rocks occur in Michigan and bordering states, in SE Wyoming, on the west side of Hudson Bay and in the Chibougamau area of Quebec. Paleoproterozoic glaciogenic rocks also occur in NW Europe, South Africa and Western Australia. Although preserved on four continents, evidence of Paleoproterozoic glaciation is much less widespread than for the younger Neoproterozoic occurrences. It is not known whether this reflects an originally restricted distribution or whether much of the older rock record has been lost as a result of subsequent erosion.

Discovery of the wide distribution of Neoproterozoic glaciogenic rocks led Mawson (1949) and Harland (1964) to suggest that the surface of the entire planet may have been frozen during parts of the Neoproterozoic. This concept was crystallized in the phrase “snowball Earth” by Kirschvink (1992), who proposed a genetic relationship among Neoproterozoic glaciogenic rocks, banded iron-formations (BIFs), and the snowball Earth state. Banded iron formations are thought to be chemically precipitated sediments that typically consist of finely interbedded (or “banded”) layers of siliceous material (chert and jasper) and iron oxides and other iron-rich minerals. Banded iron formations are common in Late Archean to Palaeoproterozoic successions (2,500 Ma to ∼1,800 Ma) but are rare-to-absent from most sedimentary successions deposited in the next billion years. Deposition of the Paleoproterozoic BIF has been attributed to the onset of the oxygenation of Earth’s atmosphere, beginning at about 2,300 Ma. Extensive bodies of younger Fe-rich sedimentary rocks are unlikely because, following oxygenation of the Earth’s atmosphere, large amounts of Fe could not have accumulated in the world’s oceans, since Fe (as Fe ^{+ 3}) is virtually insoluble under oxidizing conditions. Alternatively, the deep oceans were not oxygenated during the period from about 2.3 Ga to 800 Ma but rather were rich in sulfide, which would have reacted with any dissolved iron to precipitate iron sulfides (Anbar and Knoll, 2002). Both scenarios predict a dearth of Fe-oxide-rich chemical sedimentary rocks. The fact that some widespread BIFs, such as those in the Hamersley Basin of Western Australia, appear to pre-date the oxygenation of the atmosphere as indicated by various parameters in the Huronian Supergroup, suggests that deposition of these iron formations was also controlled by other (tectonic?) factors. Kirschvink (1992), however, linked the Neoproterozoic BIFs to the snowball Earth hypothesis. He considered that isolation of the atmosphere (oxygenated) and oceans by ice would have led to the buildup of dissolved Fe in the oceans as the result of on-going hydrothermal activity at subaqueous mid-ocean ridges and other volcanic centers. Once oxygen was re-introduced to the oceans at the termination of the world-encircling glaciation, iron would have precipitated.

Another line of evidence, proposed by Hoffman et al. (1998), involves low δ ¹³C values from carbonate rocks both below and above Neoproterozoic glaciogenic diamictites in Namibia. They attributed these unusually low δ ¹³C values to the near-cessation of photosynthetic activity in the world oceans, due to the inferred widespread climatic deterioration and development of world-encircling ice sheets and sea ice. Carbonate rocks precipitated from the oceans inherit many of their stable isotopic characteristics from sea water. Many photosynthetic micro-organisms preferentially secrete the lighter isotope of carbon (¹²C) so that during periods of high biological productivity the oceans are correspondingly enriched in the heavier isotope (¹³C). Thus, δ ¹³C values of carbonates formed in such organic-rich periods would be relatively high, and conversely these values would be low if photosynthetic activity significantly decreased, as might be expected during a global glaciation. Because little evidence exists to support a significant decrease in photosynthetic activity in the Neoproterozoic, Hoffman and Schrag (2002) subsequently suggested that the low δ ¹³C values commonly found in post-glacial carbonate rocks (so-called “cap carbonates”) are due to rapid introduction of large amounts of CO₂ to the oceans from the atmosphere, rather than due to a biological catastrophe. The ultimate source of this CO₂ may have been dissociation of oceanic methane hydrate during deglaciation, followed by oxidation to CO₂ (Jiang et al., 2003).

Low latitude glaciation

The angle of inclination of the Earth’s magnetic field to the planet’s surface varies systematically with latitude. The field is steeply inclined in polar regions whereas it is nearly parallel to the Earth’s surface at low latitudes. Thus, if the magnetic field direction is “frozen” into sedimentary rocks at or close to the time of their deposition then the paleolatitude can readily be determined. There are several reports of low (tropical to subtropical) paleolatitudes from both Neoproterozoic and Paleoproterozoic glaciogenic rocks. In some cases, it has been shown that these low latitude glacial deposits were laid down close to sea level. This is very important as some glaciers occur in tropical latitudes (e.g., on Mt. Kiliminjaro) at present but they can exist only above the snowline – at very high altitudes. The widespread nature of low paleolatitude magnetic signals from Neoproterozoic glacial deposits in SE Australia has been demonstrated. This condition was probably long-lived, for it has been shown that it spanned times of magnetic reversals (periods when the Earth’s magnetic field spontaneously reverses direction, so that the north pole becomes the south). Similar low paleolatitudes are also claimed for some Paleoproterozoic glaciogenic rocks in North America and South Africa. These indications of ice at sea level in tropical latitudes have been considered to provide support for the idea of global glaciation – the snowball Earth hypothesis. Even before such paleomagnetic evidence was widely available, Williams (1975) suggested that the unusual rock associations in Neoproterozoic sedimentary successions might be explained as a result of differences in the tilt angle of the Earth’s spin axis. He proposed that there was a significant increase in the obliquity of the Earth’s ecliptic in the Precambrian. He noted that, with a high obliquity (greater than 54°), tropical regions would receive less insolation than polar regions in a given year. If the Earth underwent cooling, the effect would have been greater at low latitudes so that glaciation would have occurred preferentially near the equator. Such a high obliquity would also tend to accentuate seasonal differences throughout the globe and might provide an explanation for evidence of seasonality in some glaciogenic rocks that apparently formed at low latitude (see following section).

Evidence of seasonality at low paleolatitudes

At present, strong seasonality is not observed at low latitudes. In both Paleo- and Neoproterozoic glaciogenic successions, however, there is evidence of seasonality in sedimentary rocks that, according to paleomagnetic results, formed in tropical regions. This is manifested by the preservation of laminated sedimentary rocks that closely resemble varves in younger (Pleistocene) glacial sediments formed at high-to-intermediate latitudes. These sediments are organized into couplets, each of which comprises a relatively coarse and fine layer. The fine laminae are thought to represent cold winter conditions when little water or sediment entered the depositional basin and fine material in suspension settled from the water column. The coarse layers formed during summer months when there was abundant sediment-charged meltwater. Many Pleistocene varves contain isolated dropstones. Such clasts are also common in Proterozoic laminated sequences and lend credence to their interpretation as annual deposits formed by significant seasonal temperature changes.

Evidence of strong seasonality in Proterozoic glacial deposits is also provided by large structures that, in plan view, are polygonal and in sectional view take the form of downward-tapering wedges. These structures are typically filled with sand that exhibits laminations sub-parallel to the edges of the structure. These structures have been described from a number of Neoproterozoic glacial successions, including some that formed at low paleolatitudes. They are interpreted as fossil ice-wedge structures, which are typical of permafrost regions today. They form as a result of thermal contraction and expansion cycles and indicate strong seasonal fluctuations in surface temperature over long periods of time. Although attempts have been made to explain these structures as a result of other phenomena (such as rapid advance-retreat cycles of surging glaciers), the modern and Pleistocene permafrost examples provide the best analog.

Problems with the snowball Earth hypothesis

Lack of precise age control

The ages of Proterozoic glaciations are largely unknown. There is stratigraphic evidence (i.e., one glaciogenic formation occurring on top of another) of several glacial episodes in some areas in both the Paleo- and Neoproterozoic, but the number of glaciations is still under debate. Contemporaneous glaciation of huge areas of the planet is a sine qua non for the snowball Earth hypothesis but it has yet to be demonstrated. Part of the problem lies in the fact that most sedimentary rocks (including glaciogenic deposits) cannot be dated using currently available geochronological techniques although some dates are available from contemporaneous volcanic rocks.

Inconsistent geochemical trends

Many diamictite-bearing glaciogenic successions are overlain by carbonate rocks (commonly dolostones). According to Hoffman et al., (1998), the cap carbonates are thought to be the products of extreme alkalization of the oceans brought about by reaction between the abundant CO₂ (that caused destruction of the global ice cover) and fine grained material (rock flour) produced by the grinding of rocks in the glaciers. If this theory is correct then any sediment formed from the residue of the putative extreme alteration should carry chemical evidence of very strong weathering, followed by an upward decrease in weathering intensity as the CO₂ content of the atmosphere diminished. A method of quantifying the degree of weathering of sediments and sedimentary rocks, known as the Chemical Index of Alteration (CIA) was used by Nesbitt and Young, (1982) to study post-glacial sedimentary rocks of the Paleoproterozoic Gowganda Formation in Ontario. This study showed the opposite trend to that predicted by the SEH, i.e., upward-increasing CIA values following the end of glaciation. Similar studies have not been carried out on post-glacial siliciclastic sediments of Neoproterozoic age, but in many cases there could be problems in interpreting such data because of the high proportion of recycled materials (older sedimentary rocks that have been through a previous weathering cycle) and carbonate rocks (which affect the CIA).

Great stratigraphic thicknesses of Proterozoic glaciogenic deposits

Under the snowball Earth hypothesis, it is difficult to envisage accumulation of a thick succession of glaciogenic sedimentary rocks, especially at low latitudes. For glaciation to reach low latitudes, the SEH predicts a totally frozen condition, preceded by rapid onset and followed by abrupt disappearance of glacial ice (Figure S22). On an Earth whose surface was completely frozen, the hydrologic cycle would virtually stop, so that the great ice masses and large-scale ice movements required for production of thick sedimentary successions are unlikely. Many glaciogenic successions, in both the Paleo- and Neoproterozoic, are several km in thickness. It has been suggested by proponents of the SEH that ablation of ice at low paleolatitudes might provide sufficient hydrologic activity but without access to a huge reservoir of water (such as the world oceans). The production of great volumes of glacial sediments remains problematic. It is further hypothesized that much of the continental lithosphere was situated at low paleolatitudes so that in tropical regions, where most ablation would have occurred, there should only have been a thin ice cover and therefore limited ablation (as opposed to having the entire ocean as a moisture source). However, a low paleolatitudinal distribution of continents may not have existed for all snowball Earth glaciations.

Figure S22

Proposed relationships between geologic time and surface temperature during a snowball Earth period (modified from Hoffman, 2000). Note the rapid growth and destruction of ice sheets.

Rapid onset and demise of snowball glaciations

Onset of the “snowball Earth” condition has been explained as the result of a “runaway albedo” feedback. If glaciation initiates at the poles and gradually descends to lower latitudes, a threshold occurs beyond which there is rapid lowering of surface temperatures and the entire surface of the planet becomes frozen, due to a positive feedback mechanism. Ice reflects solar energy more efficiently than land or water; therefore, as more of the Earth’s surface becomes ice-covered, it becomes progressively cooler until ice spreads rapidly to cover even low latitude regions of the planet (Hoffman, 2000). Once the entire planetary surface is frozen, the problem becomes one of finding a mechanism to release it from its frozen state. It was proposed (Kirschvink, 1992; Hoffman et al., 1998) that the Earth’s surface was re-warmed through buildup of atmospheric CO₂ from ongoing volcanic eruptions. This would have been accentuated by the absence of normal CO₂ drawdown from rock weathering, which would have ceased during the snowball state because of the widespread ice cover and low temperatures. Thus, the SEH invokes sudden (a few thousand years) onset and demise of glaciations, separated by much longer (several millions of years) glacial periods (Figure S22).

In the Paleoproterozoic Huronian Supergroup, however, there is sedimentological and geochemical evidence that the formations both below and above widespread glaciogenic rocks of the Gowganda Formation testify to a very gradual deterioration of climate before the glaciation and slow amelioration at the end. Likewise, in the Neoproterozoic of the Adelaide geosyncline in South Australia, Sturtian glaciogenic successions are overlain by thick (up to 1 km) accumulations of mudstone with abundant dropstones (Young and Gostin, 1989). Such occurrences testify to a long-lived, gradual amelioration of climate, with abundant floating glacial ice, in contradiction to the rapid amelioration implicit in the SEH. A less-than-catastrophic end to glaciation is also supported by recent evidence for multiple geomagnetic reversals within a cap carbonate, over a span of at least hundreds of thousands of years.

Origin of banded iron-formations associated with Proterozoic glaciogenic deposits

As noted above, the association of banded iron formations (BIFs) with some Neoproterozoic glaciogenic deposits led Kirschvink (1992) and others to propose that there was a genetic link with the snowball Earth condition, with BIFs depositing once deglaciation began. This relationship has also been proposed for the Paleoproterozoic but in the best-known basins, such as the Huronian of Ontario, glaciogenic rocks and Superior-type BIF are separated by about 300 Ma. Likewise, in the Hamersley basin of Western Australia, glacial deposits formed after deposition of major BIF, and not before, as predicted by the SEH. In contrast, there is a much more intimate association between BIF and glacial deposits in the Neoproterozoic. For example, ice-rafted dropstones are present in some laminated iron-formations, clearly attesting to the contemporaneity of glaciation and deposition of iron-formation. This relationship was attributed by some to precipitation of Fe concentrated in the world’s oceans as a result of isolation of the atmosphere from the oceans during a snowball Earth period. There are, however, problems with this interpretation. The distribution of Neoproterozoic BIF is much more sporadic than that of the glaciogenic rocks. If deposition of the BIF were the result of global processes in the oceans, they should be much more common. In some cases (e.g., in the Rapitan Group of the northern Cordilleran region of North America), huge deposits of BIF occur below rather than above the glaciogenic diamictites. This is significant because the theory predicts that Fe should be released after the destruction of the oceans’ ice cover. Many of the Neoproterozoic BIF occurrences formed in rift basins. This is shown by the presence of strong facies and thickness changes (Yeo, 1981; Young and Gostin, 1989) in the glaciogenic successions and the association, in some basins, with volcanic rocks. Furthermore, the geochemistry of the BIF supports a hydrothermal origin, which is compatible with the glaciated rift hypothesis for their origin (Yeo, 1981). This theory proposes that the iron was produced by hydrothermal circulation of seawater in restricted Red Sea-type rift basins. Iron-charged brines were moved and mixed with “normal” seawater and meltwater as a result of overturn caused by water movements related to release of meltwaters from glaciers that debouched into the basin from surrounding rift shoulders. Experimental studies show that such dilution and oxidation could lead to precipitation of dissolved metals such as Fe. This mechanism provides explanations for the hydrothermal nature of the BIF, their relatively restricted occurrence, fault-related facies and thickness changes and the intimate relationship with glaciogenic deposits.

Conclusions

Glaciogenic rocks are widely preserved in supracrustal successions of early Paleoproterozoic and late Neoproterozoic age and sparse in the long (∼1,500 Ma) intervening interval of geologic time. This evidence suggests that the Earth went through significant climatic perturbations near to the beginning and end of the Proterozoic Eon. These perturbations in the Earth’s surficial environments may have provided the impetus for significant evolutionary changes in the biosphere, especially near the end of the Proterozoic Eon when the first animals emerged. Whether the Earth’s surface achieved a totally frozen condition (the snowball Earth hypothesis) during the great Proterozoic ice ages remains a possibility but currently available evidence is by no means conclusive.

Cross-references

Albedo Feedbacks

Atmospheric Evolution, Earth

Banded Iron Formations and the Early Atmosphere

Carbon Isotopes, Stable

Carbonates, Cool Water

Diamicton

Glacial Sediments

Glaciations, Pre-Quaternary

Ice-Rafted Debris (IRD)

Obliquity

Periglacial Geomorphology

Proterozoic Climates

Varved Sediments

Weathering and Climate