A Model Where Warming Makes East Antarctica Snowier, and Harder to Melt
Warm air carries more water, so a warmer world should snow harder on Antarctica. Climate scientists have known that for a long time and have mostly set it aside, because over the next century or two the melting, the calving and the ice flowing out to sea are expected to outrun the extra snow. Stretch the same question across a hundred thousand years and, in one new set of simulations, the balance comes out the other way.
Those simulations are the work of Kyung-Sook Yun and Axel Timmermann, at the Center for Climate Physics of the Institute for Basic Science and at Pusan National University in Busan, South Korea. Their study was published Oct. 30, 2026, in Nature Climate Change. Models of this kind usually treat the great ice sheets as scenery: fixed in place, or pushed around by a climate that never feels them back. Yun and Timmermann wired the two halves together, so the ice could reshape the air above it and the air could reshape the ice. Then they held carbon dioxide steady at a series of raised levels and let the pair run for 100,000 years of ice-model time, with the Earth system model run in short bursts between long stretches of ice. It is peer-reviewed, openly available and a model result rather than a measurement.
The two Antarctic ice sheets then come apart differently. Above 2 °C of warming over pre-industrial levels, together with the ocean warming at depth that comes with it, the West Antarctic ice sheet disintegrates on a millennial scale: thousands of years rather than decades, but it goes. Greenland's ice follows the temperature down steadily. East Antarctica does neither. The paper's phrase for it is "a much greater resilience", which in these runs persists at 10 °C of warming and more. Resilience is not survival. The grounding line, where ice starts to float, retreats through the marine basins around East Antarctica's rim and then slows sharply once it reaches rock above sea level. The Antarctic ice sheet as a whole still gives up enough ice in these runs to raise the sea by about 6 meters.
The extra snow outpaces the extra melt
In the simulations a warmer Southern Hemisphere pushes the westerly winds further south and leaves the air over Antarctica moister and cloudier. Total cloud cover over the polar regions rises by 20–40%. More moisture arriving means more snowfall; more cloud means less sunlight reaching the surface to melt it. Over East Antarctica those two curves separate. Snowfall there rises by 11.5% for every degree of local warming, while melting, calving and ice flowing out rise by 8.7%. Over Greenland and West Antarctica the gap runs the other way, and both lose ice.
How much of that mattered is testable, and they tested it. When the ice model is run on its own with the two changes removed, snowfall and sunlight pinned at pre-industrial values, the whole Antarctic ice sheet loses about 70% of its volume under 12 times pre-industrial carbon dioxide. With them in, it loses about 25%. That second number is the one worth carrying away, and it belongs to the entire continent rather than to East Antarctica alone. Of the two effects, the snow does more work than the sunlight. Every experiment is available as downloadable files on the authors' data server.
The authors' own tests point the other way
Yun and Timmermann are careful about what this does and does not establish, and the careful parts are the most useful in the paper. Their standalone ice-sheet runs do not collapse East Antarctica either, which means the link to the atmosphere may not be the whole explanation; how the ice model itself is configured may contribute too. Every robustness test they report moves the answer toward more ice loss, not less. A steeper fall of temperature with altitude takes more off the Antarctic ice sheet. Adding meltwater-driven cracking and the collapse of tall ice cliffs costs more ice again. Stronger melting underneath the floating shelves loses it faster. And there is one model pair here, one run per carbon dioxide level, and no ensemble of runs to put a spread around the answer.
A published coupled model disagrees outright. A 2025 study in Geophysical Research Letters joined a simpler Earth system model, with its clouds prescribed rather than simulated, to a different ice-sheet model, and got a darker, less stable East Antarctica. Yun and Timmermann name that result and argue against it: the southward wind shift and the extra cloud in their own run keep the snow cover bright and the surface high. The older uncoupled work sits on the same side. Simulations that drove an ice-sheet model with simplified climate forcing put the complete melting of East Antarctica at 6–10 °C, a band this coupled run passes through without it. A 2026 map of tipping risks across Antarctica's ice basins in the same journal is in the paper's own reference list, and most of the recent work on Antarctic tipping points has pointed the other way.
Greenland's number is the easiest one to over-read
The same runs put Greenland's threshold at 6 °C above pre-industrial. That is this model's figure, and it sits well above the published ones: standalone ice-sheet models give 0.8–3.2 °C, and the coupled study these authors compare themselves against came out lower too. The paper notes the pattern itself, that thresholds inferred from coupled models tend to run higher than earlier offline estimates, probably because a uniform climate forcing misses regional detail. Nobody should read that as Greenland's safe margin.
One ingredient does have support from outside the model. East Antarctic snowfall really is unusually sensitive to temperature: a 2023 study in The Cryosphere put the plateau at more than 10% more precipitation per degree of warming, which is where this simulation's own value sits. That corroborates the mechanism's first step. It says nothing about the 100,000-year result.
And these are not forecasts. The high-warming cases hold carbon dioxide fixed at concentrations nobody projects, for longer than modern humans have existed, in order to ask what state the ice would eventually settle into. The finding underneath is narrower than reassurance, and more interesting: when an ice sheet is allowed to change the weather that falls on it, the answer a model gives about Antarctica changes too.