Can earthquakes be predicted?
No. Nobody can say where an earthquake will be, when it will happen, and how big it will be. Not USGS, not any university, not any app, and not this site. USGS states plainly that it does not predict earthquakes and has never predicted a major one.
That is the whole answer. The rest of this page explains what is possible, because the honest alternatives are genuinely useful and are often confused with prediction.
Prediction, forecasting and early warning are three different things
| Which | What it claims | Does it exist? |
|---|---|---|
| Prediction | This fault will produce a magnitude 7 next Tuesday | No. Not achievable with any current science. |
| Forecasting | Over the next 30 years there is an X% chance of a magnitude 6.7 or larger in this region | Yes, and it is the basis of building codes and insurance pricing. |
| Early warning | Shaking has started 60 km away and will reach you in 15 seconds | Yes — ShakeAlert. It detects an earthquake already in progress. |
Forecasting is why the National Seismic Hazard Model exists and why a house in Los Angeles is built to a different standard than one in Ohio. It says nothing about which day.
Why "overdue" is not a real idea
You will read that a fault is "overdue" for a large earthquake. It is a misunderstanding of what an average recurrence interval is. If a fault segment has produced large earthquakes an average of every 150 years, that average is drawn from intervals that varied enormously — and elapsed time does not accumulate pressure on a schedule. An earthquake does not become more likely tomorrow because one has not happened for a long time in the way a bus becomes more likely the longer you wait.
QuakeIndex does not use the word, and it is on our banned phrase list — a page that used it would fail our own build.
Foreshocks look useful and are not
Some large earthquakes are preceded by smaller ones nearby. The problem is that small earthquakes happen constantly, and there is no property of a small earthquake that identifies it as a foreshock at the time. The label is applied retrospectively, after a larger event follows. So "there has been a swarm, is a big one coming?" has no answer, and anyone giving you one is not reading the data.
What our activity score is, and is not
The QuakeIndex score compares the last seven days at a place against that place's own ten-year record. If it reads "elevated", that means more energy has been released here recently than usual. Past tense, and nothing more. It is not a statement about next week, and we designed it not to be mistaken for one: it is always shown alongside the normal range it was measured against, and we decline to publish a number at all where the record is too sparse for "normal" to mean anything.
How to recognise a claim that is not science
- It gives a specific date, or a narrow window.
- It cites planetary alignments, tides, animal behaviour or cloud shapes.
- It is unfalsifiable — the window is wide enough that something will happen in it.
- It is not published anywhere that other scientists could check it.
- It is selling something.
The productive thing to do with the energy a prediction claim generates is preparation, which works regardless of timing: a kit, a retrofit if your house needs one, early warning on your phone, and knowing what to do when shaking starts.
Sources
Last checked: