In 2003, off the coast of Hokkaido, two sensors on the seafloor were sitting in exactly the right place. They belonged to a Japanese cable network, and they happened to lie in the patch of bottom that the earthquake bent. They measure pressure so finely that they can pick up fractions of a millimetre of water standing above them. When the ground broke, they wrote everything down at once.
What they wrote was a mess. Mikhail Nosov, an oceanologist who studies tsunami waves, got the record and had to pull it apart. The seabed was moving, so the water level over the sensors was changing. The tsunami itself was starting to run. And something else was shaking the readings.
Nosov's group filtered that something out first, and the quake's work showed up cleanly: the seabed had risen by 30 to 50 centimetres. A rise like that is the classic start of a tsunami, since the lifted floor pushes the water above it up. In 2011 at Tohoku the displaced volume was 100 cubic kilometres. Why bother with the leftover shaking, then? Because it came from a place nobody had ever recorded from. The closer sensors sit to a tsunami's birthplace, the faster they react and the better they describe what is happening down there.
The leftover signal was the water itself, trembling. The whole layer, kilometres deep, behaves like jelly, Nosov says. Sound waves run through it, and it shivers because it can be squeezed.
Acoustic waves run through it, and it starts to shake like this, because it is compressible.
В нем бегают акустические волны, он так начинает колебаться из-за того, что он сжимаемый.
— Mikhail Nosov, Osnova, 30:58
That sounds wrong. Water is the textbook thing that does not squeeze. Nosov's answer is that everything squeezes and the only question is how hard you have to push: water just needs "the right crowbar". He measures stiffness by multiplying a substance's density by the speed of sound in it. Air has about one kilogram per cubic metre and sound moves through it at 340 metres per second, which gives a few hundred. Water has a thousand and sound moves at 1,500, which gives 1.5 million. Water is stiffer than air by several orders of magnitude, but it still gives a little. The seabed beneath it is stiffer again, by roughly another factor of ten.
A quake is the crowbar. It heaves a slab of floor under a column of water, and the column rings.
The sound is low, about one hertz, far below what an ear can pick up. It does not last. The energy runs down into the seabed, bends into the rock and is gone. So the evidence is brief: a short wobble on a pressure record, easy to mistake for noise.
Others have since worked on the same physics. A 2025 paper in Geoscientific Model Development by Usama Kadri describes a warning technology built on the acoustic signals that appear with a tsunami because the water layer is compressed. A September 2025 arXiv paper by Ravindra Pethiyagoda describes pressure waves bouncing between surface and floor and then travelling off as acoustic-gravity waves.
The wobble also fit a mathematical model Nosov had been developing years before and was looking for a way to apply. The Tokachi-oki record was, as he tells it, the first ever made inside the source of a tsunami. Before 2003 the first minutes of a tsunami were something people could only compute.
