№ 174 · earth science

A glacier flows because solid ice creeps

A glacier is solid ice, yet it moves downhill, typically by tens to hundreds of metres a year. It does this in two ways. The ice itself slowly deforms under its own weight, the way cold honey or warm wax would if you waited long enough. And where the bottom of the glacier is wet, the whole mass can also slide over the ground beneath it. This page is mostly about the first, because ice deforms in a lopsided way: a little extra push makes it flow a lot faster.

Ice is a solid that creeps

An ice crystal is built from flat layers of water molecules stacked one on another. Ralf Greve's review of how ice flows on Earth and other worlds describes how those layers can glide over each other when the crystal is sheared, like the cards in a deck. Real crystals contain defects called dislocations that make this gliding much easier, so the process is known as dislocation creep. Glacier ice is made of countless such crystals, millimetres to centimetres across, and together they behave like a very stiff, very slow liquid.

Put a steady load on a block of glacier ice and it does not just squash a little and stop. After a short settling-in period it keeps deforming at a steady rate for as long as the load stays on. That rate grows steeply with the load. In the 1950s John Glen measured this in the laboratory, and the rule that came out of his experiments, now called Glen's flow law, says that the rate of deformation goes as the stress raised to a power n. Robert Law and colleagues point out that Glen's own figure was 3.2, and glacier models usually round it to 3. With n = 3, doubling the stress makes the ice deform eight times as fast.

Temperature matters too. Warmer ice is softer, and Greve gives the standard constants that set how much softer for ice at −10 °C and colder. Just below the melting point ice softens faster still. Maria Zeitz and colleagues give the larger value that ice models usually use for that warm range. The figure below joins the two.

Predict first

The ice at the bottom of a glacier carries the weight of everything above it. Where in the glacier does the ice move fastest?

Check your answer

At the surface. Steven Earle's Physical Geology explains that the lower ice does most of the deforming, while the upper ice is carried along on top of it and so moves fastest. The base moves more slowly than the surface, and the edges of a valley glacier, held back by friction with the valley sides, move more slowly than the middle.

Why the top moves fastest

Picture the glacier as a thick slab lying on a slope. At any depth, the ice above is trying to slide downhill, and the ice below has to hold it back. The deeper you go, the more ice there is above, so the shearing stress grows in step with depth. At the bed it is greatest. Since the ice deforms as the cube of the stress, the deforming is heavily concentrated near the bottom. In a slab like the one in the figure, the bottom quarter of the ice does about two-thirds of all the shearing, and the top quarter hardly shears at all.

Now think of it as a deck of cards on a tilted table. Each card slides a little over the one beneath it. The top card ends up moving farthest, because it carries every slip below it added together. That is why the surface of a glacier moves fastest even though it deforms least. Earle's textbook treats the top 50 to 100 metres as effectively rigid ice riding on the flowing ice below. In Glen's law no ice is perfectly rigid, but at a tenth of the depth the stress is a tenth of that at the bed, and the rate of shearing is a thousandth. That stiff upper layer cannot bend where the flow below speeds up, so it cracks, and Earle explains that this is where crevasses open.

Because the stress at the bed depends on both the thickness and the slope, the speed is extremely sensitive to each. In the figure, doubling the thickness from 150 m to 300 m makes the surface move 16 times as fast, since a thicker slab has both more stress at the bed and more layers adding up their slips. Doubling the slope from 3° to 6° makes it move almost 8 times as fast. Cooling the ice from 0 °C to −10 °C slows it about ten times.

Interactive A slab of glacier ice resting on a slope, seen from the side, with a pipe pushed straight down through it to the bed. Set how thick the ice is, how steep its surface is, how cold the ice is and how fast it slides on its bed, which it can only do when the ice is at 0 °C. The arrows show how far each depth moves in a year, worked out from Glen's flow law. Press Watch a year to see the pipe bend. Every control works with Tab, Enter and the arrow keys.

ice thickness200 m
surface slope5.0°
ice temperature0 °C
sliding at the bed0 m/yr

Surface speed 60 m a year

flow → sideways motion drawn 1.0 × the depth scale surface slope 5.0° 0 200 m no sliding at the bed bedrock pipe after 0 days
200 m of ice on a 5.0° slope.

Sliding on the bed

Earle explains that if the base of a glacier is at the melting point, a film of water usually separates the ice from the rock or sediment beneath, and the ice can slide. If the base is below freezing, the ice is frozen to the ground and all of its motion comes from deforming. Thick ice is a good insulator, so heat from Earth's interior can keep the bottom of a thick glacier at the melting point, while a thin snout may freeze to its bed. Earle describes how, at the front of the Athabasca Glacier in Alberta, ice coming from behind has been pushed up and over the stuck front edge.

How much of the motion is sliding varies enormously. Law and colleagues, in a review of glacier sliding, note that in borehole measurements and simulations over rough beds, slip at the base has ranged from 5% to 95% of the surface speed. They also describe the old explanations of how ice gets past bumps in the bed: it can melt on the upstream side, where the pressure is higher, and refreeze downstream; or it can deform faster around the obstacle. The melting trick only matters for bumps smaller than a few centimetres. Their review argues there is still no agreed way to turn all this into a single rule for computer models, and it notes that the rules in use can make large differences to how much ice a model loses.

Is the power really 3?

The n = 3 in Glen's law is a convention, and several recent studies push on it. In 2022 Joanna D. Millstein, Brent M. Minchew and Samuel S. Pegler used satellite measurements of how Antarctic ice shelves stretch and found n = 4.1 ± 0.4 in their fast-flowing parts. In 2024 Meghana Ranganathan and Minchew argued that the mechanism with n = 4, dislocation creep, likely dominates in all the fast-flowing areas of Antarctica. In 2025 Sheng Fan and colleagues pooled 70 years of laboratory experiments and concluded that, for ice deformed a great deal, n should be 4, or 3.5 if temperatures above −5 °C are included.

Ice right at its melting point may be different again. Also in 2025, Collin Schohn and colleagues sheared watery ice in a large ring-shaped machine at Iowa State University and found it behaved with n close to 1, over the ranges of water content and stress they expect near glacier beds. And at Argentière Glacier in the French Alps, Juan-Pedro Roldán-Blasco and colleagues, measuring how a borehole bent over a year, found the shape of the bending was explained by the ice getting softer with depth as it held more water, and that it said little about n. So the shape of the figure's curve is right in kind, but exactly how steeply ice responds to stress is still being measured.

Flowing while shrinking

A glacier flows downhill even while it is retreating. Earle explains that the front of a glacier is always melting or breaking off as icebergs. If the ice arrives faster than the front melts, the front advances. If it arrives more slowly, the front pulls back, though the ice itself is still moving forward. Whether a glacier grows or shrinks is set by how much snow it gains and how much ice it loses each year. Flow only moves the ice from where it is gained to where it is lost.

Right now glaciers are losing far more than they gain. The Glacier Mass Balance Intercomparison Exercise, a combined estimate published in 2025, found that the world's glaciers, not counting the Greenland and Antarctic ice sheets, lost 273 ± 16 billion tonnes of ice a year from 2000 to 2023, and the rate rose by 36% from the first half of that period to the second. The World Glacier Monitoring Service reported in April 2026 that glaciers lost 408 ± 132 billion tonnes in the 2025 hydrological year, and that six of the highest mass-loss years on record came in the past seven years.

In short

Glacier ice is a solid that creeps: its crystals glide along their layers, and the rate of deformation rises roughly as the cube of the stress. The stress grows with depth, so most of the deforming happens near the bed, and the surface, carried on all the layers below, moves fastest. Where the bed is wet, the glacier slides as well, by anything from a small to a large share of its motion. Thickness, slope and temperature all change the speed steeply. How steeply ice responds to stress is still argued, with recent work favouring a power nearer 4 for cold ice deforming fast and nearer 1 for watery ice at the melting point. Flow carries ice downhill whether a glacier is growing or shrinking, and today the world's glaciers are shrinking fast.

Where this comes from

  1. Fluid dynamics of planetary ices linked only, not reproduced
    Ralf Greve · arXiv:0903.3773 · 2009
    arxiv.org/abs/0903.3773
  2. What is glacier sliding reuse permitted with attribution
    Robert Law, David Chandler, Phillip Voigt, Ivan Utkin, Andreas Born · arXiv:2407.13577 · 2024
    arxiv.org/abs/2407.13577
  3. Physical Geology, 2nd edition, 16.2 How Glaciers Work linked only, not reproduced
    Steven Earle · 2019
    geo.libretexts.org/Bookshelves/Geology/Physical_Geology_(Earle)/16%3A_Glaciation/16.02%3A_How_Glaciers_Work
  4. Sensitivity of ice loss to uncertainty in flow law parameters in an idealized one-dimensional geometry (The Cryosphere 14, 3537-3550) linked only, not reproduced
    Maria Zeitz, Anders Levermann & Ricarda Winkelmann · 2020
    doi.org/10.5194/tc-14-3537-2020
  5. Ice viscosity is more sensitive to stress than commonly assumed (Communications Earth & Environment 3, 57) linked only, not reproduced
    Joanna D. Millstein, Brent M. Minchew & Samuel S. Pegler · 2022
    doi.org/10.1038/s43247-022-00385-x
  6. A modified viscous flow law for natural glacier ice: Scaling from laboratories to ice sheets (PNAS) linked only, not reproduced
    Meghana Ranganathan & Brent Minchew · 2024
    pmc.ncbi.nlm.nih.gov/articles/PMC11164037/
  7. Flow laws for ice constrained by 70 years of laboratory experiments (Nature Geoscience 18, 296-304) linked only, not reproduced
    Sheng Fan, Ting Wang, David J. Prior, Thomas Breithaupt, Travis F. Hager & David Wallis · 2025
    pmc.ncbi.nlm.nih.gov/articles/PMC11981940/
  8. Linear-viscous flow of temperate ice (Science 387, 182-185) linked only, not reproduced
    Collin M. Schohn, Neal R. Iverson, Lucas K. Zoet, Jacob R. Fowler & Natasha Morgan-Witts · 2025
    doi.org/10.1126/science.adp7708
  9. Creep enhancement and sliding in a temperate, hard-bedded alpine glacier (The Cryosphere 19, 267-282) linked only, not reproduced
    Juan-Pedro Roldán-Blasco, Adrien Gilbert, Luc Piard, Florent Gimbert, Christian Vincent, Olivier Gagliardini, Anuar Togaibekov, Andrea Walpersdorf & Nathan Maier · 2025
    doi.org/10.5194/tc-19-267-2025
  10. Community estimate of global glacier mass changes from 2000 to 2023 (Nature 639, 382-388) linked only, not reproduced
    The GlaMBIE Team · 2025
    doi.org/10.1038/s41586-024-08545-z
  11. Global glacier mass change in 2025 (Nature Reviews Earth & Environment 7, 213-215) linked only, not reproduced
    The WGMS Network (Zemp, Welty, Nussbaumer et al.) · 2026
    doi.org/10.1038/s43017-026-00777-z