Earthquake Early Warning Systems

Earthquake early warning is often described as though it were prediction. It is not. It is a race, and the reason it can be won is that the messenger is faster than the message.

The physics that makes it possible

Two facts combine.

First, seismic waves travel at different speeds. P-waves move through the crust at roughly 6 km per second and are comparatively weak. S-waves and surface waves follow at roughly 3.5 km per second and carry most of the destructive energy. Over 100 km, that difference opens a gap of roughly 12 seconds.

Second, electronic signals travel at close to the speed of light. A message from a sensor near the epicentre to a city 200 km away arrives essentially instantly compared with the seismic waves crossing the same ground.

So: instrument the fault zones densely, detect the P-wave the moment it arrives at the nearest stations, characterise the earthquake in a few seconds, and transmit an alert that overtakes the damaging waves.

The warning is never long. It is enough to matter.

The blind zone

The unavoidable limitation is that the system cannot warn the area closest to the epicentre — precisely where shaking is strongest.

Detection takes finite time: waves must reach at least a few stations, algorithms must associate and characterise them, and the alert must be issued. That processing occupies several seconds, during which the S-waves are already propagating outwards. The region they have covered by the time the alert goes out receives no useful warning at all.

This blind zone typically has a radius of a few tens of kilometres, depending on station density and processing speed. Denser networks shrink it but cannot eliminate it.

The result is an inherent trade-off: warning time increases with distance from the epicentre, but shaking intensity decreases with distance. The places that would benefit most get the least notice.

What can be done with seconds

The value of a short warning lies in automation. Machines act far faster than people.

Rail systems brake automatically. Japan's Shinkansen network has an established record of decelerating trains before shaking arrives, and derailment at high speed is among the most consequential outcomes early warning prevents.

Lifts stop at the nearest floor and open their doors, preventing entrapment — one of the more common problems after urban earthquakes.

Industrial processes shut down: valves close on gas and chemical lines, sensitive manufacturing pauses, reactors and refineries enter safe states.

Hospitals pause surgery. Even a few seconds' notice lets a surgeon withdraw an instrument.

Utilities isolate sections of gas networks, reducing post-earthquake fire risk, historically one of the largest secondary hazards.

Emergency services open firehouse doors before power fails or frames distort.

People drop, cover and hold on — and for those a moderate distance from the epicentre, that alone reduces injuries substantially.

Systems in operation

Japan operates the most mature system. The Japan Meteorological Agency's nationwide network delivers alerts through television, radio, mobile phones and public address systems. It performed as designed during the 2011 Tōhoku earthquake, providing many seconds of warning to Tokyo, though the event also exposed limitations in rapidly estimating magnitude for very large ruptures.

Mexico's SASMEX was the first public system, developed after the devastating 1985 Mexico City earthquake. Because the subduction zone that threatens Mexico City lies several hundred kilometres away on the coast, the capital can receive up to a minute of warning — an unusually favourable geometry.

Taiwan, South Korea, China, Israel, Turkey, Costa Rica, Chile and India operate systems at varying scales.

ShakeAlert covers California, Oregon and Washington, delivering alerts through mobile applications, Wireless Emergency Alerts and automated industrial systems.

Why magnitude estimation is hard

The subtlest technical challenge is that a large earthquake takes time to happen.

A magnitude 9 rupture may take several minutes to propagate along its entire fault. In the first few seconds, its P-wave looks much like that of a magnitude 6. An early-warning system must estimate final magnitude from the beginning of a rupture that has not finished — and there is genuine scientific debate about how much information the initial seconds actually contain about the eventual size.

Systems handle this by issuing an initial estimate and updating it as more data arrives. That creates a real operational tension: alert early and risk under- or over-estimating, or wait for accuracy and lose warning time.

Both error types carry costs. Underestimating leaves people unprepared. Overestimating produces false alarms that erode trust and can cause injury during unnecessary evacuations.

What to do when you get an alert

Treat it as the start of shaking, not as advance notice to go somewhere.

Drop, Cover and Hold On immediately. Do not attempt to leave a building or run outside — you have seconds, and moving during shaking is how people get hurt.

If driving, slow down gradually and pull over away from overpasses and bridges.

If near the coast, be ready to move to high ground once shaking stops, since strong shaking may indicate tsunami potential.

And accept that alerts will sometimes arrive simultaneously with the shaking, or after it. Physics guarantees this near the epicentre. The system is not broken when that happens; it is operating at its limit.