What if earthquakes could be predicted?

An earthquake is when the ground shakes.

What if we knew the day of an earthquake ahead of time? Let's imagine it together.

An earthquake happens when two blocks of Earth's rock suddenly slip past each other along a crack called a fault, and the ground shakes. Today, nobody can say exactly when one will come.

What if we could? In this thought experiment, you can follow three paths: the warning systems people already use, why prediction is so hard, and how buildings are made safer. Then you can imagine what a true prediction might change.

USGS is direct about it: neither USGS nor any other scientists have ever predicted a major earthquake, and it does not expect to know how in the foreseeable future.

This thought experiment imagines that a real prediction became possible. To reason about it fairly, we first need to separate prediction from three tools that already exist and are often confused with it: long-term probabilities, short-term aftershock forecasts and earthquake early warning. Each outcome of the imagined prediction is an idea, not a forecast.

Scientists cannot say when an earthquake will come.

But special machines can notice shaking very fast. They send a quick warning to phones. People then drop, cover and hold on.

USGS says a real earthquake prediction has to name three things: the date and time, the place and the size. Nobody can do that today.

Instead, USGS works out probabilities: the chance that a big earthquake happens in an area over a number of years. It also runs ShakeAlert, a warning system on the West Coast of the United States. ShakeAlert notices an earthquake that has already started and can send an alert seconds before the strong shaking arrives, so people can drop, cover and hold on.

USGS distinguishes four things. Probabilities give long-term chances, mostly from the average rate of past earthquakes, with odds that might range from 1-in-30 to 1-in-300. Forecasts apply to shorter windows, usually aftershocks, whose sequences tend to follow the same pattern. Early warning is issued after an earthquake starts. A prediction must give date and time, location and magnitude, which is not yet possible.

ShakeAlert detects an earthquake already under way, estimates its location, magnitude and shaking, and lets partners deliver alerts or trigger automatic actions such as slowing trains and closing valves.

Let's pretend we knew the day of an earthquake.

Families might get ready early. Schools might practice. These are only ideas.

Now imagine scientists could name the day, the place and the size of an earthquake, long before it came. That is pretend: USGS says nobody knows how.

In that pretend world, families might pack supplies and fix heavy shelves to the wall ahead of time. Trains might slow down early, and people might stay away from places where things could fall. But even then, buildings and bridges might still need to be strong, because the shaking itself might not change at all. These are ideas, reasoned from the tools people use today.

If a reliable prediction of time, place and magnitude existed, the seconds of warning ShakeAlert gives might stretch to days. Actions that today must be automatic, such as slowing trains or closing valves, might be planned calmly in advance.

Yet a prediction might not reduce the shaking. Strong structures might still matter most. Today USGS focuses its efforts on long-term mitigation and the safety of structures rather than short-term prediction. A false prediction might also cause harm of its own. All of these are reasoned ideas, not predictions.

Sometimes small signs come before an earthquake. But most of the time, the same signs come and no earthquake follows.

Many earthquakes have no signs at all.

People have looked for clues before earthquakes, such as swarms of small quakes, more radon gas in local water or animals acting oddly. USGS says these possible clues often happen without any earthquake afterward, and most earthquakes have no clues before them at all.

Animals are a good example. People have told stories of strange animal behavior before earthquakes for a very long time, but USGS says scientists have not found behavior that is consistent and reliable, or a way it could work.

USGS lists commonly claimed precursors: swarms of small earthquakes, rising radon in local water, unusual animal behavior, a series of growing moderate events, or a moderate event that might be a foreshock. Most of these frequently occur without being followed by an earthquake, and most earthquakes have no precursory events whatsoever.

Claims of prediction also fail USGS's three-part test: they lack a scientific basis, leave out the date, place or size, or are so broad that some earthquake will always fit. On animals, USGS notes that consistent and reliable behavior before earthquakes, and a mechanism explaining it, still eludes scientists.

Some buildings sit on special soft pads. The pads help the building shake less.

At school, children practice drop, cover and hold on.

Engineers use earthquake knowledge to build stronger. USGS hazard maps are the basis for the earthquake rules in building codes. Some buildings rest on base isolators, flexible pads that partly separate them from the ground. In a 1994 earthquake in Los Angeles, USC University Hospital stood on 149 isolators, and the hospital and its contents had no damage.

People help too. In the Great ShakeOut drill, families, schools and businesses practice drop, cover and hold on. USGS also advises securing heavy furniture and tall shelves at home.

USGS's National Seismic Hazard Model, updated for all 50 states in 2023, is the basis for seismic provisions in building codes and for insurance risk models. Recordings from instrumented buildings help engineers design new structures and strengthen old ones.

Base isolation partly decouples a building from ground shaking. In the magnitude 6.7 Northridge earthquake of 1994, the eight-story USC University Hospital, on 149 isolators, came through with no damage: the isolators cut the motion fed into its base by about two-thirds, and peak shaking at the roof was only about 40 percent of that on the ground about 200 feet away.