A new study led by Woods Hole Oceanographic Institution (WHOI), published in Science on or around October 15, 2026, provides the most comprehensive observational analysis of ice cliffs at the margins of Earth’s ice sheets (primarily Antarctica and Greenland).

Researchers, including lead author Caroline Needell (MIT- WHOI Joint Program doctoral student), Catherine Walker (WHOI), and Jeremy Bassis (University of Michigan), analyzed 15,330 individual satellite observations (mainly from NASA’s ICESat- 2 laser altimetry, plus some earlier airborne data).

These measured cliff heights, ice thickness, water depth, crevassing, calving rates, and how fronts evolved over time (often in ~3- month intervals over multi- year periods).

Marine Ice- Cliff Instability (MICI)

MICI is a long- debated theory proposing that ice cliffs taller than roughly 100 m above the waterline can become structurally unstable under their own weight. Collapse would expose even taller inland ice, potentially trigger rapid, self- sustaining retreat and substantially accelerate sea- level rise contributions from ice sheets. It has featured in some of the highest sea- level projections but has never been directly observed in a runaway form, leading to debate over its inclusion in models.

Key Findings

• Tall ice cliffs (>~90- 100 m) are rare. Most are much shorter (e.g., 95% under ~52 m in Antarctica and ~77 m in Greenland); the tallest observed reached a bit over ~125 m. Only a small number of glaciers had cliffs near the theoretical critical height.⁠
• When tall cliffs do retreat and calve, they do so faster than existing parameterizations assume, sometimes more than three times the rate predicted by formulas used in high-end projections for a ~100 m cliff. Models may therefore underestimate short- term calving rates from tall cliffs.⁠
• However, the rapid retreat is typically short- lived rather than runaway. Glaciers often thin, retreat into regions with less favorable conditions (e.g., varying bed topography that reduces effective height, or other stabilizing factors), and re- stabilize as shorter, more stable cliffs. No observed cases showed sustained unstable retreat.⁠
• Other factors beyond pure height matter for stability, including bed topography, ice thinning rates, water depth (buoyancy effects), and mélange (the mix of ice debris at the front).

Implications

The observations confirm that some mechanical processes described by MICI occur in nature, but they also show limiting factors that can halt rapid retreat. This provides an empirical global dataset to better constrain ice- sheet models, helping reduce uncertainty in sea- level rise projections. Accurate modeling will need to incorporate variables like bed topography and thinning rates, not just cliff height.⁠

The work was supported by the National Science Foundation, NASA, the U.S. Department of Energy, the DOMINOS project (part of the International Thwaites Glacier Collaboration), and others.

Full paper: “Empirical constraints on ice-cliff calving” (DOI: 10.1126/science.aee0658).

Official WHOI release: whoi.edu/press-room/news-release/ice-survey/.

Marine ice- cliff instability (MICI) is a hypothesized process in which tall, near- vertical ice cliffs at the edges of marine- terminating glaciers or ice sheets become structurally unstable and collapse under their own weight, potentially triggering rapid, self- sustaining retreat of the ice sheet.

Basic mechanism

Ice sheets in Greenland and Antarctica often end in the ocean either as floating ice shelves or as grounded ice cliffs. When a protective ice shelf collapses or thins away (due to warming oceans or atmosphere), it can expose a steep cliff of grounded ice.

Theory and modeling indicate that once the cliff height above the waterline exceeds a critical threshold, commonly estimated around 80- 100 meters (roughly the height of the Statue of Liberty), the stresses in the ice approach or exceed the material strength of the ice.

This can cause the cliff to fail through brittle fracture, shear, or slumping. The collapse produces icebergs and exposes a new, often taller cliff farther inland (especially in regions where the bed deepens inland, called a retrograde slope). If successive failures continually expose taller ice, the process can become a positive feedback loop of accelerating retreat.

Key factors that influence whether this occurs include:

• Cliff height above water
• Water depth (deeper water provides more buoyant support)
• Ice thickness gradients and bed topography
• Pre- existing damage or crevassing in the ice
• Presence of proglacial mélange (a mix of icebergs and sea ice that can buttress the cliff)
• Rates of dynamic thinning (the ice flowing and stretching and thinning as it approaches the front)

Why it matters for sea- level rise

MICI has been included in some ice- sheet models that produce the highest projections of future sea- level rise. In particular, early parameterizations suggested it could contribute to multi- meter rises over centuries by enabling rapid collapse of vulnerable regions such as the West Antarctic Ice Sheet (which holds enough ice to raise global sea level by about 3- 5 meters if fully lost). The Intergovernmental Panel on Climate Change has treated projections involving MICI with low confidence because of large uncertainties.

Controversies and observational constraints

The idea remains debated for several reasons:

• True runaway MICI has never been directly observed in the modern era; it is largely inferred from theory, limited observations of tall cliffs, high- fidelity fracture models, and interpretations of past sea-level high stands.
• Tall ice cliffs (near or above ~100 m) are rare today. Comprehensive satellite surveys (e.g., of more than 15,000 cliffs using ICESat- 2 and airborne laser altimetry) show that most cliffs are far shorter (means around 30 m; 95th percentiles typically under 50- 80 m depending on the ice sheet).

A 2026 study in Science (Needell, Walker, and Bassis) provided the largest observational dataset to date. It found that the tallest cliffs do calve (break off) faster than many existing model parameterizations assume, sometimes more than three times as fast for ~100 m cliffs. However, the rapid retreat phases were short-lived. Dynamic thinning and retreat into areas with smaller ice-thickness gradients caused the cliffs to stabilize rather than continue runaway collapse. The work indicates that models may underestimate short-term calving rates from tall cliffs, but that instability depends sensitively on factors beyond height alone (especially thickness gradients and topography).

In short, MICI describes a plausible mechanical pathway for accelerated ice loss after ice- shelf disintegration, but real- world observations so far show that stabilizing processes often interrupt the feedback. Improved models will need to incorporate these additional controls to narrow the range of future sea-level projections.

Empirical constraints on ice cliff calving

Editor’s summary

Sections of the ice cliffs at the margins of Earth’s ice sheets constantly collapse into the adjacent sea in a process called calving. Could such calving turn into a runaway retreat of the ice sheet edge if it proceeds into thick enough parts of the ice sheet past the limits of ice cliff stability? Needell et al. measured the profiles of more than 15,000 ice cliffs in Greenland and Antarctica and examined the characteristics that promote calving. Using a subset of suitable glaciers, they found that the tallest cliffs calved more rapidly than current parameterizations assume, but also that dynamic thinning and margin retreat into regions with low ice thickness gradients prevented runaway losses. —Jesse Smith

Abstract

When calving cliffs at the margins of Earth’s ice sheets exceed a critical height, runaway retreat may be initiated in a process called the marine ice cliff instability. However, sufficiently tall ice cliffs have rarely been observed. To address this gap, we measured the characteristics of 15,330 ice cliffs across Earth’s ice sheet margins using laser altimetry. We show that cliff height is highly correlated with mélange thickness, water depth, and crevassing. We found that tall cliffs calve more rapidly than parameterizations assume, but dynamic thinning and retreat into regions with low ice thickness gradients prevent runaway retreat. These observations suggest that model projections may underestimate subannual rates of cliff calving, but that the potential for instability will depend sensitively on additional quantities, not cliff height alone.

Key points from the study (drawing on the published abstract/editor summary and closely matching preprint)

• Researchers compiled ~15,000 and more laser altimetry observations (primarily NASA ICESat- 2, 2018- 2023, plus earlier Operation IceBridge airborne data) of ice- cliff heights around the Antarctic and Greenland ice sheets.
• Cliff- height distributions are right- skewed. Means are ~31- 33 m. The 95th percentile is only ~52 m in Antarctica and ~77 m in Greenland. Maximum observed heights are ~124- 128 m. Cliffs near or above the commonly cited ~100 m theoretical critical height (*H*crit) are rare.
• Cliff height correlates positively with water depth and proglacial mélange thickness and shows a relationship with a proxy for pre- existing ice damage and crevassing (taller cliffs tend not to be the most damaged, suggesting highly stressed, damaged ice fails before reaching extreme heights). No strong link to atmospheric or sea- surface temperature was found.
• Focusing on the tallest cliffs (>90 m; 16 glaciers identified), four with good repeat sampling that retreated during the study period were analyzed in detail (Fleming, Commandant Charcot, and Frost glaciers in Antarctica; Tuttulikassaap Sermia in Greenland).
• Calving rates increase with cliff height and exceed existing parameterizations (e.g., the DeConto & Pollard formulation used in high -end sea- level projections) by up to a factor of ~3 for ~100 m cliffs.
• Rapid retreat was observed but was short- lived. Dynamic thinning combined with retreat into regions of low ice- thickness gradient suppressed further cliff growth and prevented runaway (self- sustaining) retreat. No cases of sustained marine ice-cliff instability (MICI) were observed.

Implications: Ice- sheet models that rely solely on cliff height for calving parameterizations may underestimate short- term (sub-annual) retreat rates from tall cliffs. However, the potential for true MICI-style instability depends sensitively on additional factors, especially ice -thickness gradients, bed topography, dynamic thinning, and mélange, rather than height alone. The new observational dataset provides empirical constraints to improve these parameterizations and reduce uncertainty in sea- level rise projections.

The full text is behind the Science paywall (or institutional access). A preprint version with highly similar content is available on Research Square (DOI: 10.21203/rs.3.rs-7133627/v1). Supplementary materials (methods, additional figures, data) are linked from the Science page, and replication data have been deposited (e.g., Dryad).

Journal: Science, 394, 311–315 (2026)

DOI: 10.1126/science.aee0658

Provided: Woods Hole Oceanographic Institution

Authors: Caroline Needell, Catherine C. Walker, Jeremy N. Bassis