Chemical weathering beneath ice sheets may explain why some snowball Earths lasted so long

Between about 720 and 635 million years ago, Earth froze over at least twice. Ice stretched from the poles to the equator, oceans crusted over, and the planet became a white marble drifting through space. These snowball Earth events are among the most extreme climate episodes in geological history. But they raise a puzzle that has nagged geologists for decades: the first one, the Sturtian glaciation, lasted roughly four to fifteen times longer than the second, the Marinoan. Why?

A study from the Earth-Life Science Institute (ELSI) at the Institute of Science Tokyo offers a surprising answer. The ice itself may have been fighting against its own melting.

The standard escape route and its problem

The textbook explanation for how Earth escapes a snowball state goes like this. Under normal conditions, volcanic CO2 emissions are balanced by chemical weathering of rocks on land, which consumes CO2. When ice covers the continents, weathering shuts down because there is no liquid water to drive it. Volcanoes keep erupting, CO2 accumulates in the atmosphere, and eventually the greenhouse effect becomes strong enough to melt the ice.

This model is clean and logical. It also predicts that both snowball events should have lasted roughly similar lengths of time, since volcanic emission rates and continental configurations were broadly comparable. The fact that the Sturtian glaciation persisted for tens of millions of years longer than the Marinoan has been difficult to reconcile with this framework.

Water under the ice

Lead author Shintaro Kadoya and co-author Mohit Melwani Daswani developed numerical models of what happens at the base of kilometer-thick continental ice sheets. The key insight is that geothermal heat from Earth's interior, combined with the insulating effect of overlying ice, can generate meltwater at the glacier base even when surface temperatures are far below freezing.

This is not theoretical speculation. Modern ice sheets in Antarctica and Greenland have subglacial lakes and drainage networks. During a snowball Earth, with ice potentially several kilometers thick, basal melting would have been widespread.

That meltwater does not just sit there. It flows through crushed rock produced by glacial erosion, the grinding action of ice moving over bedrock. This contact between water and freshly exposed mineral surfaces creates conditions for chemical weathering, the same CO2-consuming reactions that normally happen on exposed land surfaces under rain and flowing rivers.

CO2 consumption that could rival volcanic output

The models track how dissolved elements, secondary minerals, and fluid chemistry evolve as subglacial meltwater interacts with crushed rock. The central finding is that under plausible snowball Earth conditions, the rate of CO2 consumption by subglacial weathering could approach the rate of volcanic CO2 emissions.

If that seems like a small detail, it is not. In the standard model, zero weathering means CO2 accumulates steadily and deglaciation proceeds on a predictable timeline. But if weathering beneath the ice consumes CO2 nearly as fast as volcanoes emit it, the net accumulation slows to a crawl. The greenhouse escape valve still works, but it takes far longer to build up enough pressure.

The efficiency of this subglacial weathering depends on two factors: how much meltwater is available and how much fresh rock glacial erosion delivers. The balance between these two determines the chemical output. Even modest variations in meltwater supply or erosion rate could shift the system from one where CO2 accumulates quickly to one where it accumulates barely at all.

Two glaciations, different plumbing

This sensitivity provides a potential explanation for the duration difference between the Sturtian and Marinoan glaciations. If subglacial hydrology and erosion rates differed between the two events, perhaps due to differences in continental positions, ice sheet dynamics, or bedrock geology, the rate of CO2 consumption beneath the ice would have differed too. A more efficient subglacial weathering system during the Sturtian could have delayed deglaciation by tens of millions of years.

The models also suggest that subglacial meltwater flowing into the oceans could have delivered nutrients like phosphorus, with potential implications for marine biological productivity once the ice finally retreated. Subglacial environments, in other words, were not frozen wastelands. They were active chemical reactors.

Model assumptions and open questions

The study relies on numerical simulations, not direct measurements of 700-million-year-old subglacial chemistry. The models require assumptions about ice sheet thickness, geothermal heat flow, erosion rates, and rock composition that carry significant uncertainty. The actual conditions beneath Neoproterozoic ice sheets are, by nature, impossible to observe directly.

The models also treat subglacial weathering in a relatively simplified way, focusing on steady-state chemical interactions. Real subglacial environments involve complex, intermittent water flow, variable rock exposure, and biological activity that could either enhance or inhibit weathering rates. These complexities are not fully captured.

The geological evidence that supports subglacial weathering during snowball Earth, including the presence of minerals like dolomite in glacial-era sediments, is suggestive but not conclusive. Alternative explanations for these mineral deposits exist, and the debate is ongoing.

What the study does convincingly is challenge the assumption that continental weathering simply stops during global glaciation. If even a fraction of the modeled CO2 consumption occurred beneath Neoproterozoic ice sheets, it would represent a significant and previously overlooked feedback in Earth's climate system.