Every tsunami warning system has the same structural weakness.
The instruments sit in the water, so the wave has to physically arrive at one before anybody learns anything from it.
Deep ocean buoys help, and they are expensive, sparse and still in the path.
Seismometers get around that by reading the earthquake instead, but an earthquake is only a clue about a tsunami rather than a measurement of one.
Which leaves an obvious gap. Something that watches the wave itself, without waiting for it and without sitting where it is going.
The answer turned out to be above the water.
A wave in the sea writes itself into the sky
A tsunami in open ocean lifts and drops the entire surface over an enormous area, slowly and by only a few inches.
That motion pushes on the air resting on top of it, and the pressure disturbance does not stay near the surface. It climbs.
As it rises the air around it thins, and because the energy has to go somewhere the disturbance grows, so a few inches of sea becomes a large atmospheric wave by the time it reaches the ionized layer a couple of hundred miles up.
Up there the air is electrically charged, and satellite navigation signals passing through it are slowed by exactly how many free electrons lie along the path.
Move that layer around and the delay changes, which a ground receiver sees as its position solution wobbling in a very particular pattern.
The sky becomes the instrument.
Nothing new had to be built
The unusual part of this is that the hardware was already there and was installed for something else entirely.
Geodetic receivers sit on concrete pillars all over the world, logging satellite signals continuously for surveying, plate motion studies and reference frames.
The system reads more than 350 of them around the Pacific rim, and each one can see disturbances out to roughly 745 miles from where it stands.
No ship had to be sent anywhere, no cable laid, no mooring maintained in deep water where maintenance costs more than the instrument.
The signals are free, the receivers are somebody else’s, and the processing is the only new thing in the chain.
It is an instrument made of existing infrastructure.
Eight minutes up, thirty two minutes ahead
The earthquake struck at 23:24 and 52 seconds UTC on July 29 of 2025, off the Kamchatka coast, at a depth of about 22 miles.
It came in at magnitude 8.8, which ties it with two others as the sixth strongest ever recorded on a seismometer and makes it the largest since 2011.
The first atmospheric disturbances showed up in the data 8 minutes after the seafloor moved, and the approaching wave was flagged 32 minutes before it made landfall in Hawaii.
The peer reviewed write up of the event is slightly more conservative, putting the detection at 30 minutes before the wave reached the coast.
What arrived was an amplitude of about 5.7 feet at one Maui harbor and 4.9 feet at Hilo.
The tsunami itself was running at over 500 mph.
What the signal does not tell you
The honest limit is that this measures a disturbance in the atmosphere rather than a height of water.
It says something is coming and roughly when, which is genuinely valuable, but converting the ionospheric wobble into an expected run up at a particular beach is not solved.
Coverage is also uneven by construction, since the receivers stand on land and large stretches of ocean have no land within range, in the way that a gap in the telemetry or a slow process nobody watched leaves a blind stretch.
The output still needs a trained analyst to read, which the developers say plainly rather than selling it as a finished product.
The formal description of what it did during this event runs to a single detection case, not a track record.
One event is one event.
Thirty two minutes is not a warning by itself
None of this replaces anything, and the people building it are the first to say so.
A tsunami warning is already a chain of seismometers, pressure recorders on the seafloor, coastal gauges and models, and each link covers the weaknesses of the others.
What this adds is a reading taken from a place the water has not reached, early in the sequence, when every other source is still either guessing from the quake or waiting.
The developers describe the value in minutes rather than in certainty, and say those minutes are what could make a difference to a community deciding whether to move.
The Pacific tsunami announced itself twice.
The sky heard it first.
Read the whole thing?
Get the week's signal, not the noise
Our sharpest reporting on energy, climate and nature — free, once a week.
