Mars: Gullies Carved by Dry Ice, Not Water
A new analysis of data from two orbiters shows that gullies in the Sisyphi Cavi region of Mars are shaped not by water or geysers but by late-season melting of carbon dioxide ice.

A team of scientists from Paris-Saclay University has shown that the gullies that shift from season to season on the surface of Mars are shaped neither by liquid water nor by carbon dioxide geysers, but by a late stage of ice sublimation that triggers avalanches of dry ice and dust down a crater’s slope. The finding concerns Sisyphi Cavi, a system of deep polar pits near the south polar cap, and appears in a preprint posted to arXiv on September 3, 2026, under the number 2609.03885. The paper’s lead author is Apolin Leklef of the Institute for Space Astrophysics at Paris-Saclay University, with ten authors listed in total.
Martian gullies are geologically young features that resemble water-carved gullies on Earth, and for decades that resemblance has fueled debate over what actually drives them. Two mechanisms have most often been proposed, periodic ice melting that would briefly release liquid water, and the so-called Kieffer geyser mechanism, in which gas pressure building beneath a translucent layer of carbon dioxide ice breaches the crust and blasts jets of gas and dark material onto the surface. The new study is the first to combine ice composition with the seasonal timing of these events at one and the same active site, rather than examining them separately.
The analysis draws on data from the OMEGA spectrometer aboard the European orbiter Mars Express and the CRISM spectrometer aboard the American Mars Reconnaissance Orbiter. The site sits at 68 degrees south latitude and 1 degree east longitude, outside the geographic band where Martian gullies typically form, which makes it a useful test of how far existing explanations actually hold up beyond the usual latitudes. The results are still preliminary, drawn from a preprint that has not yet gone through peer review at a scientific journal.
Why water and geysers are ruled out
The team’s first task was to check whether liquid water was involved at all. Water’s absorption band, which shows up at a wavelength of 1.5 micrometers, was completely absent during the spring thaw, when the layer of frozen carbon dioxide that blankets the surface over winter disappears. Water ice was not detected as a standalone surface deposit, though the team does not rule out its presence in trace amounts, mixed in as a minor component within the carbon dioxide ice layer itself. Salts and clays, materials that normally require water to form, were found, but only on the high plateaus and crater rims, not near the gullies themselves.
Geysers looked like the more plausible candidate, but the timing did not line up. The dark spots that in images betray gas venting from geysers appear in late winter and early spring, peaking around the spring equinox. Changes in the gullies, however, only begin in mid to late spring, well after the geysers have already quieted down. That gap in timing, the authors argue, does not support the idea that geysers are what directly shapes the gullies. Another thread the team followed was the physical state of the carbon dioxide ice itself, which stays translucent through most of the season instead of turning opaque, which would have hinted at a different process at work.
What actually moves the slope
Since neither water nor geysers fit the data, the authors propose a third mechanism, one close to how a puck floats on a cushion of air on an air hockey table. In the final stage of carbon dioxide ice sublimation, the gas generated underneath the ice slab can act as a kind of lubricant, rather than venting through a narrow crack the way it does in a geyser. A previously stationary mass of dry ice and dust then slides down the slope and turns into a fast-moving flow that carves channels and leaves fan-shaped deposits at the bottom, the same shape visible in images of active gullies. The team’s conclusion is that the changes seen in the Sisyphi Cavi gullies are better explained by the liquefaction, or avalanche, of carbon dioxide ice in the last stages of its sublimation, rather than by liquid water or a classic geyser.
Limits of the findings
The paper is a preprint and has not yet been reviewed by other planetary geologists, so the results should be treated as preliminary. The analysis covers a single site, Sisyphi Cavi, which is itself an outlier since it lies outside the latitudes where gullies typically form, so the conclusion cannot automatically be extended to every gully on Mars, including those at latitudes where ice melting has previously been floated as a possibility. The data also do not prove that water played no role at all during some earlier geological epoch on Mars, they only show that there is no trace of liquid water in the current seasonal activity at this particular site. What was observed is also a matter of timing, or the lack of it, between phenomena tracked from orbit, not a direct measurement of the mechanism itself on the ground, so what is presented here is above all the best explanation that currently fits the available data, not final proof.



