The events sit 60 to 90 miles beneath David Glacier, far from any plate boundary, and they were hiding in twenty years of archived seismic data the field had largely ignored.
For decades, Antarctica has been treated as the planet's quietest continent, a frozen interior where the tectonic plates beneath the ice barely twitch. A new analysis overturns that picture. Researchers at the University of Alabama and their collaborators have used a machine-learning model to recover more than 500 previously unrecognized earthquakes from roughly twenty years of archived Antarctic seismic data, many of them sitting 60 to 90 miles beneath David Glacier, in the middle of a tectonic plate where the standard textbook model says they should not exist.
Published in the journal Science and reported May 28 by Live Science, the catalog comes from running a machine-learning reanalysis over two archival seismic datasets collected between 2001 and 2004 and again from 2012 to 2015, drawn from 49 ground instruments spread across the continent. Those instruments were deployed for other purposes, and the intermediate-depth events buried in their recordings had been classified as background noise. A model trained to look for faint earthquake signatures pulled them out.
David Glacier stretches roughly 1,100 kilometers, or about 700 miles, and bridges the East and West Antarctic ice sheets, draining close to 4 percent of the East Antarctic Ice Sheet's total ice into the Ross Sea. Beneath it, the new catalog finds more than 500 events at depths between 100 and 150 kilometers, with magnitudes ranging from 1.6 to 3.5. The lead author, Long Ho of the University of Alabama, told Live Science that the events are too small to threaten the ice sheet or the surrounding ecosystem, a point echoed by independent glaciologist Richard Alley of Penn State, who was not involved in the study.
The deeper claim is geological. Intermediate-depth earthquakes, those deeper than about 80 kilometers, are normally observed only at subduction zones, places where one tectonic plate dives beneath another. The new catalog places them in the middle of a plate, hundreds of kilometers from any active boundary. Ho and his coauthors propose a mechanism: warm, buoyant upper-mantle material extending under the edges of David Glacier from below, pushing up on the cold, rigid East Antarctic crust and upper mantle. That flexure concentrates stress at depth, and the boundary between cold East Antarctic rock and warmer West Antarctic rock creates an abrupt change in tectonic strength. The result is a zone where the plate interior can break.
Alley framed the result in plainer terms, telling Live Science that the long-held view that Antarctica largely lacks earthquakes was less a fact about the continent and more a fact about the listening tools. The new model suggests similar intraplate events may be hiding in archived seismic data from other continental interiors, and that the role of such events in plate-tectonic theory may need to be re-evaluated.
The paper stops short of claiming a hazard link. The detected events are small, the proposed mechanism is the authors' interpretation rather than established consensus, and an independent seismologist's read on the flexure model would strengthen the case. What the catalog does establish is a method. Machine-learning reanalysis of legacy seismic data is a reusable playbook for stress, ice, and tectonic questions the existing 49-station network was never designed to answer, and the first pass has already redrawn the map of where intermediate-depth earthquakes can occur.