From the Android cell phone network that alerted users in Venezuela to the public systems in Mexico and Japan, technology can’t predict earthquakes: it just tries to buy us a few seconds after the earthquake starts — and turn those seconds into protection.
Those who hit northern Venezuela on June 24 once again exposed an old frustration: science calculates eclipses centuries in advance and anticipates hurricanes days, but it continues unable to tell when the Earth will shake.
The seismologists’ explanation is straightforward — it is possible to know where and how often large earthquakes occur, but not the day nor the hour. And this limitation is not born from a lack of technology: it is in the very nature of the phenomenon.
Still, according to , a few seconds before the shock was felt in several areas, millions of Android users received a warning on their cell phones. It wasn’t a prediction — it was real-time detection.
In a country with no official seismic early warning system, it was a Google system, based on the sensors in millions of Android phones, that gave some people the few seconds that in some cases made the difference.
To understand the apparent contradiction, it is necessary to separate three things that are often confused:
- One is predict: saying when, where and with what magnitude an earthquake will occur, within useful margins — something that cannot be done today.
- Another is estimate the risk: Calculate the probability of a major shake in a region over decades — this is done, and it is what underpins earthquake-resistant building codes.
- The third is alert: detecting an earthquake that has already started and warning before the most destructive waves arrive — and it works.
Forecasting is the impossible link. The rupture that causes a large earthquake occurs several kilometers deep, far from any instrument that measures it directly, and behaves like a chaotic system: a small fracture can either lock or propagate and trigger a devastating shock — and there is no way to know, right away, which of the two paths it will follow.
Over the last few decades, they have been tested supposed precursor signs — such as radon emanations in the soil, changes in groundwater, electromagnetic disturbances, so-called ““. None have proven reliable and repeatable.
As for premonitory tremors, they are only identified as such after the big earthquake: the overwhelming majority of small tremors are not followed by anything.
A confusion between forecasting and risk communication has already had serious consequences. Three years after the L’Aquila earthquake in 2009which killed more than 300 people, seven members of the Italian Major Risks Commission (six scientists and a Civil Protection officer) were convicted of involuntary manslaughter.
Contrary to what has often been repeated, these experts were not tried for not having predicted the earthquake, but for the way they communicated the risk to the population, after a series of small tremors were described as not representing imminent danger.
The convictions of the six scientists would eventually be and the acquittal was confirmed by the Italian supreme court. All that remained was the partial condemnation of Bernardo de Bernardinisthe head of Civil Protection — but the case served as a warning about the dangers of convey false certainties.
If it cannot be predicted, it can be detected
Early warning systems take advantage of an uneven race between different types of .
When an earthquake starts, it first releases P waves, fast (about six km/s) but not very destructive, they arethen the S and surface wavesslower (3 to 4 km/s) and responsible for most of the damage.
How do the Digital data travels much faster than any seismic wave, a sensor close to the epicenter can detect P waves and trigger an alert that reaches the most distant areas before the strongest impact of the tremor is felt, explains .
There are a few seconds of advancement, sometimes tens of seconds — enough to stop trains, immobilize elevators or seek shelter.
O Mexico was a pioneer in public seismic early warning systems: their system has been operating since the early 1990s and could give Mexico City up to 60 seconds of warningwith sirens and radio alerts.
O Japan, with hundreds of stationsincluding sensors on the seabed, sends warnings directly to cell phones; In the United States, it covers the west coast and stops the subway automatically.
Portugal has seismic surveillance and information in almost real time through IPMA, but not through a mass public system that warns the population seconds before the earthquake arrives. For , there is an alert system operated by IPMA and integrated into .
A Venezuela did not have an official early warning system of this type. Situated between the Caribbean and South American plates, the country is seismically active, but its national network of monitoring stations is limited., and it does not have an official warning system: the surveillance body monitors and reports, but does not issue advance warnings.
What worked was Google’s alert system for Android, which turns the accelerometers of more than two billion cell phones into miniseismographs: when a sufficient number of devices in the same area simultaneously detect a seismic pattern, a central server confirms the event and triggers the warning.
According to one published in Science in 2025, the system has already detected more than 18 thousand earthquakes since 2021 and showed that cell phones can complement conventional seismic networks, especially in regions without official early warning systems.
The system is not infallibleand was criticized for underperformance in the 2023 Turkey and Syria earthquakes, when it underestimated the magnitude and did not send maximum level alerts to most at-risk users. But, in the absence of an alternativewas what gave many in Caracas a few seconds to .
Until science learn to read the Earth before it moves“predicting” an earthquake continues to come down to a single certainty: that .
