Possibly by a lowering of atmospheric greenhouse gases to near-present levels through tectonically-mediated rock weathering, when the Sun was considerably dimmer than present.

Evidence points to a reduction in so-called "greenhouse" gases in the atmosphere, principally CO₂ (carbon dioxide) and CH₄ (methane). This would have made the global climate colder, creating larger areas of ice and snow. Ice and snow reflect more Solar radiation than does bare ground or liquid water, which creates a "positive feedback". If the earth ever became approximately half-covered by ice or snow, the feedback would become self-sustaining and glacial ice would rapidly spread to the equator. Climate physicists (notably Mikhail Budyko in Leningrad, USSR and William Sellers in Tucson, Arizona, USA) accidentally discovered the runaway ice-albedo feedback in the 1960's as a result of calculations (using simple radiative energy-balance models) concerning the stability of Arctic sea-ice. The physicists did not believe a runaway feedback had ever actually occurred. They were unaware that a geologist, Brian Harland at Cambridge University in the UK, was marshalling geological evidence for global glaciation. (Harland was equally unaware that physicists had an explanation for his observations.) The physicists assumed that life could not have survived a snowball earth and they knew of no way the planet could have escaped from the icy grip. The subsequent discoveries of deep-sea and hydrothermal-vent biotas and Antarctic psychrophiles (cold-loving organisms) removed the first objection, and in 1992 Joe Kirschvink (read: How did the snowball earths terminate?) postulated a natural escape mechanism involving CO₂.

What caused the lowering of greenhouse gas concentration in the first place? The scenarios for CO₂ and CH₄ are quite different, and each has been advocated as an agent of snowball earth. First CO₂. On geological time scales, the ocean and atmosphere are in equilibrium with respect to CO₂ and can be treated as a single reservoir. CO₂ is supplied to this reservoir by volcanic and metamorphic emanations, and is removed as sediment in the form of CaCO₃ (limestone) and organic matter (roughly CH₂O). The atmospheric CO₂ forms carbonic acid rain, which is neutralized (protons are consumed) by silicate rock "weathering" (conversion to soil). The resulting solutes include Ca₂⁺ and HCO₃^- (bicarbonate) ions that rivers carry to the ocean, where CaCO₃ is precipitated by calcifying organisms and organic matter by primary producers like cyanobacteria and algae. The entire process is often simply referred to as "silicate weathering", because that is the rate-limiting step. Silicate weathering rate is sensitive to climate, faster where hot and wet, slower where cold and dry.

During the Cryogenian Period, encompassing the Sturtian and Marinoan snowball earths, there was a rare preponderance of continents in the tropics, where it is hot and wet. Therefore, the global rate of silicate weathering was high. As a result, CO₂ concentrations fell and the global climate cooled because there was less "greenhouse" warming. Global cooling lowered the silicate weathering rate, ultimately stabilizing the climate system at a new colder state. Two additional phenomena are known to have occurred at the same time which further contributed to high silicate weathering rates and therefore a cold climate. The first was the breakup of a pre-Pangean supercontinent named Rodinia, which began ~830 Ma and continued for nearly 200 million years. A supercontinent is the assembly of almost all continents into a single mass. Silicate weathering rates are low when a supercontinent exists, because most land area is far from the ocean and therefore very dry. When a supercontinent breaks up into small fragments, formerly arid regions become wetter and weathering rates increase accordingly. The second phenomenon was the massive eruption of basalt lava ("flood" basalt) at 723 Ma in Arctic Canada, which was then very close to the equator. Basalt lava weathers rapidly and is a rich source of Ca₂₊ ions. The combined effects of tropical continents, supercontinent breakup, and equatorial flood basalt emplacement are sufficient to cause a snowball earth in model simulations published in 2004 by Yannick Donnadieu, Yves Goddéris and coworkers.

Scenarios involving CH₄ are simpler but more ad hoc. CH₄ is supplied to the atmosphere by microbes (methanogens) that live in poorly-drained soils (e.g., tropical wetlands) and in organic-rich sediments below the sea-floor. It is removed by oxidation (combination with O₂ or its derivitives). Molecule for molecule, CH₄ is ~30 times more powerful as a greenhouse gas than CO₂, but it is very unstable in our present O₂-rich atmosphere where its "residence" time is ~20,000 times less than CO₂. O₂ levels in the early atmosphere were extremely low (less than 1% PAL, present atmospheric level) and CH₄ levels were presumably far higher than present, creating a strong greenhouse effect. The early CH₄ greenhouse helped offset the lower Solar luminosity, which has increased by an average of ~6% per billion years since the origin of the Solar System. When O₂ levels rose, CH₄ levels fell accordingly, causing a loss of greenhouse warming. If the loss was rapid (less than 1 million years), the cooling could not be stabilized by silicate weathering feedback, which would cause a slow compensatory rise in CO₂. If the previous CH₄ greenhouse forcing was large, a snowball earth would result from its rapid destruction.

The most likely cause of a sudden rise in O₂ is the evolution of oxygenic photosynthesis (a biological revolution), but preliminary evidence form fossil organic molecules suggests that oxygenic photosynthesis was in existence half a billion years before the Makganyene snowball earth. Different geochemical data-sets suggest that the rise of O₂ above 1% PAL occurred sometime between ~2.4 and 2.2 Ga (billion years ago). Resolving the detailed trajectory of O₂ rise and its relation to the Makganyene snowball earth is an active area of current research.

Various astronomical theories for triggering snowball earths have been proposed. In 2005, Alex Pavlov and associates suggested that snowball earths occurred when the Solar System encountered giant molecular clouds in the spiral arms of our galaxy. An isotopic test of their hypothesis has yet to bear fruit.