Why do earthquakes happen?

Why do earthquakes happen?

If Earth's tectonic plates are always moving, why does the ground not slide smoothly all the time? Plate motion is roughly as fast as fingernails grow — a few centimetres a year, and steady. Nothing about that description predicts a jolt. So what could make slow, continuous motion accumulate and then arrive all at once?

Along a fault, rough masses of rock press against one another. Plate motion continues, but friction can hold part of the fault in place. Here is the step that matters: the plates do not stop just because the fault has. The motion has to go somewhere, so it goes into deforming the rock on either side. The crust bends elastically, and stress climbs.

Follow that and the abruptness explains itself. Elastic deformation stores energy the way a bent spring does, and friction can only hold so much. Once the stress on the fault exceeds what friction can resist, the locked patch lets go, the bent rock springs back toward its unstrained shape, and the stored energy leaves in seconds. That is Harry Fielding Reid's elastic rebound theory, worked out from survey lines across the San Andreas fault that had visibly bent before the 1906 San Francisco earthquake and snapped straight through it. The name is worth knowing because it makes the right prediction: the ground had already moved before the earthquake, slowly, and the earthquake was the catching-up.

The released energy travels through Earth as seismic waves, and it travels in more than one form. Compressional P waves move fastest — roughly six kilometres a second in the crust — and shake the ground back and forth along their direction of travel. Slower shear S waves follow, moving the ground sideways, and they are usually the ones that do damage. Two consequences fall out of that gap in speed. A seismometer can turn the delay between the two arrivals into a distance from the source, which is how a network of stations locates a quake it never felt. And because the fast wave is the harmless one, early-warning systems can detect the P wave and issue an alert before the S wave arrives — seconds of notice near the epicentre, tens of seconds further away. It is not prediction; it is a message outrunning a slower messenger.

The rupture begins at the focus below ground; the epicenter is the point directly above it at the surface. The amount of damage, however, also depends on depth, distance, soil, building design, and the duration and frequencies of shaking. This is why magnitude and intensity are separate ideas. Magnitude is one number describing the source. Intensity describes the shaking experienced at a particular place, and one earthquake has as many intensities as it has places. A moderate quake in soft sediment beneath unreinforced buildings can do more harm than a larger one under bedrock.

Most earthquakes are not caused by plates colliding at that instant. They are the sudden release of stress accumulated over time, which is why a fault can be silent for centuries and none of that silence is reassuring.

Magnitude is the number most often misread. The scale is logarithmic, so each whole step is ten times the ground-motion amplitude on a seismogram — but energy scales faster than amplitude, by a factor of about 31.6 per unit, since energy goes as ten to the power of one-and-a-half times the magnitude. A magnitude 7 therefore releases roughly a thousand times the energy of a magnitude 5, not a hundred. This is also why the folk belief that small earthquakes bleed off stress and prevent large ones fails arithmetically as well as physically: it would take on the order of a thousand magnitude 5 events to match one magnitude 7, and they would have to occur on the right patch of the right fault.

Prediction and forecasting are also different things, and only one of them works. Forecasting — the probability of a quake of some size in some region over decades — is genuinely useful and underwrites building codes. Prediction, naming a time and place, has failed every serious attempt. The Parkfield experiment is the honest illustration: a stretch of the San Andreas with unusually regular moderate quakes was instrumented, and in 1985 the U.S. Geological Survey publicly forecast the next one before

1. It arrived in 2004, eleven years late, and when it came it showed no clear precursors even

though it happened inside the densest instrument array ever assembled for the purpose.

One more honest gap. Most earthquakes do occur at plate boundaries, and the model above is built from them. But some happen deep inside a plate, far from any boundary — the New Madrid sequence of 1811-12 shook the central United States hard — and why stress concentrates there, on what appear to be old healed faults, is much less well understood than the boundary case.

• Why aftershocks follow a major earthquake.
• How seismometers locate an earthquake.
• Why soft soils can amplify shaking.

Every term the collection defines is gathered in the glossary.