A capsizing iceberg grinding against the glacier that calved it radiates long-period ground motion strong enough to travel thousands of kilometres [10]. It radiates little high-frequency energy, and high frequencies are what detection and location routines key on for ordinary earthquakes, volcanoes and nuclear explosions [11]. So the signal was arriving at stations that were switched on and working, and the software was not built to see it. That is also why the whole class of event was only identified in the northern hemisphere about twenty years ago, long after other seismic sources became routine catalogue entries [12].
Magnitude then decides who is close enough to listen. Greenland's glacial earthquakes are large, the biggest comparable to North Korean nuclear tests, which is why a worldwide network logs them without special effort [13]. If the Antarctic equivalents are much smaller, distance defeats that network [6], and the remedy is proximity rather than better algorithms: the study used stations on the continent itself, where earlier searches had relied on the global array [5]. The result is 362 events across a 2010 to 2023 window, about 26 a year on average [3], with 245 of them at the marine end of Thwaites, or roughly 68 percent [1], and the remaining 117 forming the Pine Island cluster [2]. Thwaites alone averages about 17.5 events a year over the record, so the 2018 to 2020 peak [7] ran above that rate [4].
The timing is the paper's most interesting point, though the evidence behind it is thin. That peak overlaps an acceleration of the ice tongue toward the sea that satellites independently confirmed [7], and the author writes that ocean conditions could have driven the speed-up, with the effect not yet well understood [8]. The more useful evidence is negative: the Thwaites events do not follow the annual warm-air cycle that sets the late-summer rhythm in Greenland [9]. Ruling out the obvious atmospheric explanation is worth more than the coincidence itself, because a three-year overlap inside a fourteen-year record does not establish order of operations. Nothing in this account tells you whether faster flow produced more capsizing icebergs or whether the loss of those icebergs let the flow accelerate.
The missing piece here is the denominator. There is no count of how many icebergs calved over the period, so there is no detection rate, and this write-up is the researcher's own account of the study, published by The Conversation in August 2026 [17], without event magnitudes to anchor the comparison with Greenland's much larger quakes [13]. An early-warning instrument needs a calibrated baseline, and one fourteen-year series at one glacier is the beginning of a baseline rather than the possession of one.
My read: this is a credible cheap monitor of calving activity, cheap because the sensors and the recordings already exist and the marginal cost is analysis, and it is not yet a collapse forecast. It earns attention anyway because of what sits behind it. Thwaites is judged capable of falling apart quickly, and a complete collapse would put 3 metres on global sea level [15]. A monitor with modest confidence is still worth running against a number like that.
