Geologically speaking, East Antarctica is supposed to be one of the quietest places on Earth — a vast, frozen interior sitting firmly within a single tectonic plate, far from any boundary where earthquakes typically form. Nothing there should be shaking.
Yet a joint US-Spanish research team, scanning data from dozens of seismic stations scattered across the continent, has identified more than 500 earthquakes rumbling deep beneath the ice — clustered between 60 and 90 miles underground, in rock that every standard model of plate tectonics says should be inert. They found them only with the help of AI.
A continent that wasn’t supposed to shake
Antarctica has long been treated as seismically dead. No major tectonic plate boundaries cross it, and the standard model of plate tectonics holds that plate interiors experience little to no deformation. Geologists expected silence — and for decades, that is largely what they recorded.
The new study upended that assumption. Researchers identified 510 intermediate-depth earthquakes, or IDEs, clustered between 100 and 150 kilometers below the surface. Local magnitudes ranged from 1.6 to 3.5 — modest by any measure, but scientifically significant given where they occurred.
Intraplate earthquakes are already difficult to explain. IDEs within intraplate settings are harder still, because the high temperatures and pressures in the upper mantle do not normally permit the brittle rock failure that produces earthquakes. Finding them beneath East Antarctica forced researchers to ask uncomfortable questions about what “stable” geology actually means — and whether the category has ever been as reliable as assumed.
How AI helped hear the silence break
Detecting hundreds of faint earthquakes buried under kilometers of Antarctic ice is not something human analysts could manage by hand at any useful scale. The team turned to a deep learning AI detection package, enhanced by a technique called transfer learning — a method that lets a model apply patterns learned in one context to a new, related problem.
Data came from 49 seismic monitoring stations spread across East Antarctica. The AI separated earthquake signals from background noise by analyzing two distinct wave types: faster p-waves, which travel through any material, and slower s-waves, which travel only through solid rock. Comparing the arrival times and characteristics of both allowed the system to pinpoint rock-fracture events and map their locations with precision.
Transfer learning proved especially valuable here. Antarctica’s seismic environment is unique, and training a model from scratch would have required far more local data than currently exists. Drawing on models trained elsewhere let researchers work with what they had — and still detect 510 events that would otherwise have gone entirely unnoticed.
What could be causing them
The earthquakes cluster near a significant geological boundary: the meeting point of the thick, cold East Antarctica lithospheric slab and the thinner, hotter West Antarctica slab. That contrast in density and rigidity creates what researchers describe as a steep lithospheric strength gradient, where stress concentrations naturally build.
Two competing forces may be driving the fractures — hot mantle material pushing upward from below, and the enormous weight of glaciers pressing downward from above. The key mechanism involves softer, warmer rock heating and bending the rigid, brittle crust from below. That bending generates stress that eventually breaks something. Not a dramatic process, but deep underground and across geological time, it accumulates into something measurable.
A mystery within the mystery
The proposed mechanisms offer a plausible explanation for why earthquakes occur so deep beneath this region. What they do not explain is why the activity clusters specifically beneath David Glacier.
Similar lithospheric boundaries run along other stretches of the Transantarctic Mountains, and those areas appear quiet. Something local — some additional factor the researchers have not yet identified — seems to be concentrating seismic activity in this one location. The study acknowledges this gap directly rather than working around it. Antarctica is not alone in presenting this kind of puzzle; intraplate seismic activity has also been documented in Afghanistan, Morocco, and Romania, places that similarly lack obvious tectonic explanations. Antarctica now joins that list of geological anomalies awaiting fuller answers.
What this means for earthquake science worldwide
The implications reach well beyond Antarctica. The AI detection method used here could be deployed in other supposedly quiet regions around the world, scanning for hidden intraplate IDEs that current monitoring has missed entirely.
The researchers suggest these deep earthquakes may be far more widespread than the existing record indicates. Better detection tools, more finely tuned to subtle signals, could reveal a global picture of intraplate seismicity that looks very different from what scientists currently assume. Published in the journal Science, the study calls for a fundamental rethink of what “seismically stable” actually means — and the next step will likely involve denser monitoring networks, both in Antarctica and elsewhere, to determine whether David Glacier is a rare outlier or an early sign of discoveries still to come.