Why Earthquakes Cannot Be Predicted (and What Scientists Do Instead)

Direct answer: nobody can predict earthquakes. A prediction would require specifying the place, time and magnitude narrowly enough to act on, and no reliable method exists. The USGS states plainly that it has never predicted a major earthquake and does not expect to. What scientists can do is estimate long-term probabilities, issue warnings in the seconds after a rupture starts, and map where shaking will be worst.

Why prediction fails

An earthquake is the sudden failure of rock that has been slowly strained for decades or centuries. The problem is that the strain is invisible, the fault is kilometres underground, and the moment of failure appears to be extremely sensitive to conditions we cannot measure.

There is also a deeper problem: evidence suggests a small rupture and a great one begin the same way. Whether a slip that starts on a small patch of fault stops after a metre or cascades into a thousand-kilometre M9 seems to be decided as it propagates. If that is right, then even perfect knowledge of the starting conditions would not tell you the final size.

The precursors that did not work out

A century of research has tested candidate warning signs. Radon gas in groundwater, changes in electrical resistivity, unusual animal behaviour, water-level shifts in wells, foreshock patterns, and "earthquake lights" have all been proposed. The pattern is consistent: signals show up before some earthquakes, fail to show up before others, and show up plenty of times when no earthquake follows. A precursor that produces false alarms is not usable — evacuating a city is enormously costly and cannot be repeated on a false signal without destroying public trust.

Foreshocks are the most tantalising case. Roughly a proportion of large earthquakes are preceded by smaller ones — but the overwhelming majority of small earthquakes are not foreshocks of anything, and there is no known way to tell the difference in advance.

The famous near-miss

China's evacuation of Haicheng before an M7.3 earthquake in 1975 is often cited as a successful prediction. It followed an unusual foreshock swarm and local decisions to evacuate. The following year the Tangshan earthquake struck with no comparable warning and killed a catastrophic number of people. One success followed immediately by one catastrophic failure is the honest summary of the state of the art.

What scientists do instead

1. Long-term probabilistic forecasting

Statements like "a 60 percent chance of an M6.7 or larger in this region in the next 30 years" are genuine science. They are built from fault slip rates, palaeoseismology of past ruptures, and GPS measurements of crustal strain. They cannot tell you what to do on Tuesday. They can tell a government what to write into the building code and where to strengthen schools and hospitals.

2. Aftershock forecasting

This is the one place where short-term forecasting works, because aftershocks obey reliable statistical laws — their rate decays predictably with time, and their magnitudes follow a known distribution. Agencies routinely publish credible statements such as the probability of an M5 or larger in the coming week. See how aftershocks work.

3. Earthquake early warning

Early warning is not prediction. It detects an earthquake that has already started and races an alert ahead of the damaging waves, buying seconds to tens of seconds. Japan, Mexico, Taiwan and the US West Coast run such systems. See how earthquake early warning works.

4. Hazard mapping and engineering

This is where the lives are actually saved. Seismic hazard maps show expected ground motion, soil amplification and liquefaction potential. Building codes translate those into design requirements. The reason the same magnitude produces vastly different death tolls in different countries is almost entirely construction, not geology.

How to spot a fake prediction

Prediction claims circulate after every large earthquake, usually on social media. The tells are consistent:

• Planetary alignments, the Moon, or "supermoons" as the mechanism. Tidal forces are real but tiny, and no robust link to large earthquakes has been demonstrated.
• A vague window — "a big one in the next two weeks somewhere in the Pacific" — which is guaranteed to be right and means nothing.
• Claimed predictions revealed only after the event.
• An individual or private company rather than a national seismological agency.
• Anything asking for payment or clicks.

Legitimate agencies — USGS, EMSC, JMA and national geological surveys — never issue day-specific predictions, because they cannot.

What this means for you

If you live in a seismic zone, treat earthquakes as a certainty on a long timescale and an unknown on a short one. Secure heavy furniture and water heaters, know Drop, Cover and Hold On, keep a 72-hour kit, and know your tsunami evacuation route if you are coastal. Preparation is the only strategy that works when timing is unknowable.

Track live seismic activity on the earthquake monitoring page.

Frequently asked questions

Can any country predict earthquakes?

No. No national agency anywhere issues day-specific earthquake predictions, because no reliable method exists. Japan, the United States, China and others fund extensive research; none of it has produced usable short-term prediction.

What is the difference between an earthquake prediction and a forecast?

A prediction names a specific place, time and magnitude. A forecast gives a probability over a long window — for example a 60 percent chance of a damaging earthquake in a region within 30 years. Only forecasting is scientifically supported.

Do animals sense earthquakes before they happen?

There are many anecdotes and no reproducible evidence of useful lead time. Animals do react to the fast P-wave, which arrives seconds before the damaging shaking — that is detection, not prediction.

Can the Moon or planetary alignments trigger earthquakes?

Tidal stresses are real but very small compared with the strain stored on a fault. No robust link between lunar phase, planetary alignment and large earthquakes has been demonstrated.

If earthquakes cannot be predicted, what should I do?

Prepare on the assumption that timing is unknowable: secure heavy furniture, practise Drop, Cover and Hold On, keep a 72-hour kit, know your building's weaknesses, and learn your tsunami evacuation route if you live on a coast.

Sources

• USGS — Can you predict earthquakes?
• USGS — Earthquake Hazards Program
• USGS — Aftershock Forecasts
• Global Earthquake Model Foundation — Global Seismic Hazard Map
• EMSC — European-Mediterranean Seismological Centre

Found an error? See our corrections policy.