Hektoria Glacier retreated eight kilometers in sixty days. The retreat rate was an order of magnitude faster than any previously recorded for a grounded glacier. Six glacial earthquakes punctuated the collapse — seismic evidence that the ice was tearing from bedrock, not merely floating away from an already-detached shelf. The glacier lost nearly half its length. Then it stopped, not because conditions improved, but because it ran out of flat floor.
The mechanism is geometric. Hektoria sat on an ice plain — a broad, flat expanse of soft sediment below sea level, with almost no topographic variation. The glacier rested lightly on this surface, grounded at multiple points across the plain rather than at a single clean grounding line. When warming reduced the sea ice that buttressed the glacier front, the usual calving began: tabular icebergs breaking off in orderly succession. But as the front retreated across the flat plain, the calving regime changed entirely.
The ice plain had no ridge, no rocky protrusion, no topographic feature that could arrest the retreat. On a typical glacier bed — irregular, ridged, carved by flowing ice — retreat is self-limiting. The grounding line migrates inland until it reaches a pinning point where the bedrock rises enough to prevent further flotation. The glacier stabilizes. The feedback loop is negative: retreat creates conditions that slow retreat.
On the flat floor, the feedback is positive. As the front thins and retreats, more of the glacier lifts off the smooth bed and begins floating. Floating ice is exposed to tidal flexing, which opens crevasses from the bottom — a direction that grounded ice rarely fractures from. These bottom-up crevasses propagate upward until they meet crevasses opening downward from the surface. When the two sets of fractures connect, the entire ice column is severed. The severed blocks topple. The front retreats further across the featureless plain, exposing more ice to the same mechanism. The process is self-accelerating: retreat enables more retreat because there is nothing in the geometry to stop it.
The result is a new calving mechanism the researchers call ice plain calving. It produces retreat rates that match paleoglacial evidence from 15,000 to 19,000 years ago — rates previously attributed to exceptional ice-age conditions — from modern warming. The explanation is not that the forcing is exceptional. The explanation is that the geometry is permissive.
Hektoria is small — roughly the size of Philadelphia. Its collapse does not significantly affect global sea levels. The importance lies in the mechanism's portability. Ice plains have been identified beneath Thwaites Glacier, Pine Island Glacier, Whillans Ice Stream, portions of the Ross and Filchner-Ronne ice shelves, and unnamed glaciers in Greenland. All are larger than Hektoria by orders of magnitude. All sit on flat submarine beds that lack the topographic features required to arrest buoyancy-driven retreat.
The standard question about glacier collapse is: how much warming does it take? The flat floor answer is different. The question is not how hard you push. The question is whether there is anything to push against. A glacier on a ridged bed can absorb substantial warming because the geometry provides natural arrest points. A glacier on a flat bed cannot absorb even modest warming once the front retreats past the last pinning point, because the geometry provides no arrest at all.
Current ice sheet models assume tabular calving — the orderly shedding of icebergs at the glacier front. This assumption is geometrically valid for irregular beds. It is invalid for ice plains. The six-fold flow acceleration and forty-fold thinning increase observed at Hektoria are not anomalies in the models. They are consequences of a mechanism the models do not contain. The flat floor does not appear in the projections because the projections assume a floor that can hold.