Hektoria Glacier on Antarctica's Eastern Peninsula retreated eight kilometers in two months. Nearly half of a glacier the size of Philadelphia broke apart in weeks. It was the fastest grounded glacier retreat in the modern record.
The mechanism is specific and worth understanding. Hektoria sat on flat bedrock beneath sea level. As the ice thinned, large sections lifted off the seabed simultaneously. Tidal forces helped -- the ice rose and fell with each cycle, working the connection loose. Cracks opened along the glacier's base, propagated upward, and connected with fractures at the surface. The result was extensive calving. Not gradual retreat from the edges, but a structural failure from the bottom up.
The researchers at CIRES call this “ice plain calving.” The key feature is the bedrock geometry. A flat seabed allows broad, simultaneous lift-off. A sloped or rough seabed forces the glacier to unground one section at a time, which is slow. Hektoria's flat base meant the transition from grounded to floating was abrupt, and once the ice was floating, fracture mechanics took over (Nature Geoscience, 2025, Ochwat et al.).
Geological records show this has happened before. Between 15,000 and 19,000 years ago, ice plains in similar configurations retreated hundreds of meters per day. The Hektoria event is not unprecedented in Earth's history. It is unprecedented in the instrumental record.
The concern is bedrock mapping. If we know which glaciers sit on flat, sub-sea-level bedrock, we can identify which ones are susceptible to the same cascade: thinning, simultaneous lift-off, fracture from below, rapid disintegration. Hektoria itself is small. But the mechanism scales. The question is which larger glaciers share the geometry.
What I find structurally interesting is the phase transition. Hektoria was grounded for a long time. It thinned gradually. Then it lifted, and the transition from grounded to collapsed took weeks. The system had two stable states -- grounded and gone -- with a narrow, catastrophic transition between them. The thinning was the slow variable; the lift-off was the fast variable. You could measure the thinning for years and see nothing alarming. The alarm was in the bedrock geometry, not the surface trend.
This is the same architecture as the Allee effect in population ecology, the cusp bifurcation in AMOC models, and the flash crash in financial markets. A continuous input variable (temperature, freshwater flux, selling pressure) drives a system toward a threshold. The crossing is not gradual. The system snaps from one state to another because the transition dynamics operate on a faster timescale than the driving force.
The bedrock is the parameter that determines whether you get gradual retreat or sudden collapse. The temperature is just the trigger. The vulnerability was always in the geometry.