Algorithm Built to Study Mars Finds 73 Unknown Calderas on the Ocean Floor

A team of researchers from Université Paris-Saclay has shown that a tool built to scan the surface of Mars works just as well several kilometers beneath the ocean's surface. Using a machine-learning algorithm, the researchers identified 73 previously unknown volcanic calderas hidden beneath the seafloor, more than tripling the number of documented structures of this kind on Earth.

From Mars to the Ocean Floor

The starting point was an algorithm originally designed to automatically detect impact craters on the surface of Mars, a task in which artificial intelligence has for years helped planetary scientists analyze satellite imagery. Verolino's team applied the same shape-recognition logic to bathymetric maps, digital models of the seafloor's terrain built from sonar and satellite data.

In its first pass, the algorithm flagged more than 87,000 locations that could correspond to the crater-like depressions typical of volcanic calderas. That was clearly too many to count as a genuine discovery, so the researchers moved on to painstaking verification, combining automated analysis with manual assessment of each location's shape, depth, and geological context.

Three Times More Structures Than Before

After that screening, 78 of the most likely calderas remained on the list. Comparing them with existing scientific literature showed that as many as 73 had never been described before. That means the global count of documented underwater calderas jumped from fewer than 30 to more than a hundred within a single study, a result rarely seen in marine geology.

The newly discovered structures aren't clustered in one place on the map. Eight sit on mid-ocean ridges, where tectonic plates are pulling apart; nine lie in volcanic arcs linked to subduction zones; and the largest group, as many as 61 calderas, sits within tectonic plates, mostly along long chains of underwater volcanic mountains known as seamounts.

Why It Matters for Risk Assessment

Underwater volcanic calderas aren't just a geological curiosity. Their presence matters directly for assessing the risks tied to submarine eruptions, which can trigger tsunamis threatening coastlines thousands of kilometers away. The researchers point to the 2022 eruption of the Hunga Tonga-Hunga Ha'apai caldera, which generated tsunami waves and atmospheric disturbances felt around the world, as an example of how severe the consequences of such events can be.

Our data fill observational gaps and provide a repeatable, targeted framework for characterizing underwater volcanoes - research team led by Andrea Verolino, Université Paris-Saclay

The study's authors also flagged seven calderas they consider especially important targets for further observation, given their depth, location, and morphological features suggesting potential activity. These are the sites set to be examined first by future research expeditions equipped with sonar and remotely operated underwater vehicles.

What's Next for the Method

The significance of this discovery extends beyond the list of new objects itself. It shows that algorithms trained to recognize patterns on one planet can be effectively transferred to entirely different data, as long as the structure of the problem, in this case detecting circular depressions in terrain, remains similar. For oceanography, that opens a chance to fill in the blank spots on the seafloor map faster, since the ocean floor is still less well understood than the surface of Mars or the Moon.

For coastal residents and institutions handling natural disaster risk management, the practical takeaway is that the list of potential sources of submarine eruptions and tsunamis is considerably longer than believed until recently. That could shape future research priorities and plans for seismic and volcanic monitoring in regions not previously treated as especially at risk.

Sources: Artificial intelligence discovers 73 new calderas beneath the oceans (benchmark.pl), Surprising discovery, 73 previously unknown volcanic calderas in the oceans (tech.wp.pl)