Can Earthquakes Be Predicted?

Earthquake prediction is one of the most challenging problems in all of science. Despite decades of research, scientists still cannot reliably predict the exact time, location, and magnitude of a future earthquake. Here's why — and what we can do.

Why Is Precise Prediction So Difficult?

Earthquakes occur deep underground along fault systems that stretch for hundreds of kilometres. The complexity of rock mechanics, fluid pressures, and stress accumulation makes it impossible — with current understanding — to know exactly when and where rock will rupture. Dozens of proposed precursors (animal behaviour, radon gas emissions, ground deformation, electromagnetic signals) have been studied, but none have proven reliable enough for routine prediction.

Earthquake Forecasting vs. Prediction

Scientists distinguish between prediction and forecasting:

• Prediction would mean specifying that an earthquake of a given magnitude will occur in a specific location at a specific time — within a narrow window. This is not currently possible.
• Forecasting means estimating the probability of an earthquake over a longer period (decades) based on fault history, geological data, and statistical models. This is well-established science and informs building codes and emergency planning.

For example, the USGS can say with high confidence that there is a roughly 60% probability of an M6.7+ earthquake in the San Francisco Bay Area within the next 30 years. But they cannot say it will happen on a specific date.

Earthquake Early Warning Systems

While prediction is elusive, early warning systems are increasingly effective. These systems detect the first seismic waves from an earthquake (P-waves) and send automated alerts to nearby areas before the more destructive S-waves and surface waves arrive. The warning time is typically between a few seconds and about a minute, depending on distance from the epicentre.

Operational systems include:

• ShakeAlert (Western United States) — alerts sent to millions of smartphones via FEMA's Wireless Emergency Alerts
• J-Alert (Japan) — one of the world's most advanced systems, sending warnings via TV, radio, and phones
• SED ShakeMap (Switzerland/Europe)
• Systems in Mexico, China, Taiwan, and South Korea

Even a few seconds of warning is enough time to drop and take cover, pull a car to the side of the road, halt high-speed trains, or pause surgical procedures.

The Future of Earthquake Science

Machine learning and AI are being applied to seismic datasets to identify patterns that human analysts might miss. Dense sensor networks, satellite-based ground deformation monitoring (InSAR), and GPS data are all improving our understanding of where stress is building. Probabilistic seismic hazard analysis is becoming ever more precise — even if the dream of day-accurate prediction remains beyond current science.