Why Earthquakes Happen (and Why We Can't Predict Them)

It was 5:12 in the morning on April 18, 1906, and most of San Francisco was still asleep. Beneath Sacramento Street, and along roughly 477 kilometers of the San Andreas Fault, two enormous slabs of the Earth's crust that had been locked together for decades finally let go. The Pacific plate lurched northwest past the North American plate in a matter of seconds, the ground heaved, and a magnitude 7.9 earthquake rolled through the city. Within minutes the first fires were climbing through the wreckage toward the bay, and over the next three days those fires would do far more damage than the shaking itself. By the time the smoke cleared, much of the city was gone.

What happened that morning was not random violence from an angry planet. It was the sudden discharge of energy that had been quietly accumulating in the rocks for a human lifetime, released in the only way the physics allows. Understanding how that works explains both why earthquakes are inevitable along certain lines on the map and why, despite more than a century of careful study, we still cannot tell you the day one will strike. This is the story of what an earthquake actually is, how we measure it, and why the prediction we wish for remains out of reach.

The Slow Squeeze and the Sudden Snap

Strip away the drama and an earthquake is a remarkably simple mechanical event. It is the rapid release of elastic strain stored in rocks on either side of a fault. A fault is a fracture in the crust where two blocks of rock can move relative to each other, and the key word in the definition is "elastic." Rock, despite feeling utterly rigid when you stand on it, behaves a little like a steel spring under sustained pressure. It bends.

The crucial misunderstanding to clear away is the image of plates gliding smoothly past one another. They do not. The two sides of a fault catch on friction and lock together, sometimes for hundreds of years, but the forces driving plate motion never stop pushing, so while the fault surface stays stuck, the rock on either side slowly deforms, storing energy exactly as a bent ruler does. Eventually the accumulated stress exceeds the strength of the friction holding the fault shut. The rock snaps back toward its unstressed shape, the two blocks jerk past each other in seconds, and all that stored elastic energy radiates away as the shaking we feel.

This idea, that faults load slowly and rupture suddenly, is called elastic rebound theory, and it was worked out in the immediate aftermath of 1906 by the geophysicist Henry Fielding Reid. He noticed that survey lines crossing the San Andreas had been visibly bent in the years before the quake and then violently straightened during it, as if a drawn bow had been released. More than a century later, elastic rebound remains the foundation of how geologists understand fault behavior, and it delivers an uncomfortable corollary: any locked fault is by definition busy storing the energy for its next earthquake right now.

The San Andreas and the Three Ways Rock Can Break

The fault that destroyed San Francisco is the textbook example of one of three basic fault types, and it is worth meeting all three because they map neatly onto the architecture of plate tectonics. The San Andreas is a near-vertical fracture marking the boundary between the Pacific plate and the North American plate, running about 1,200 kilometers from the Salton Sea in the south to Cape Mendocino in the north, passing close enough to cities like Daly City that the boundary between two major plates runs almost through people's backyards.

Faults are classified by how the two blocks move relative to each other. The San Andreas is a strike-slip fault, meaning the blocks slide horizontally past one another with little vertical motion, like two hands rubbing palm to palm. The second type is the normal fault, which forms where the crust is being pulled apart and stretched, allowing one block (the hanging wall) to drop down relative to the other. The third is the reverse fault, or its low-angle cousin the thrust fault, which forms where the crust is being squeezed, driving the hanging wall up and over the other block.

These three styles are not arbitrary. They correspond to the three things plate boundaries can do. Where plates slide past each other, you get strike-slip faulting; where they pull apart, you get normal faulting; where they collide and one is forced beneath the other, you get reverse and thrust faulting. The way the ground breaks in any given earthquake is therefore a direct readout of the larger forces shaping that stretch of the planet, which is why a geologist can often tell you the local plate-tectonic story just from the motion recorded on a single fault.

Putting a Number on the Shaking

For most of human history, earthquakes could only be described by their effects, which made one event impossible to compare meaningfully with another. That changed in 1935, when Charles Richter, working with Beno Gutenberg at Caltech, published the local magnitude scale to put southern California's earthquakes on a single number derived from the size of the wiggles recorded on a standard seismograph.

The scale's defining feature is that it is logarithmic, and this is the single most misunderstood fact about earthquakes. Each whole-number step up the scale represents roughly a tenfold increase in the amplitude of ground motion and, because energy scales even more steeply, about 32 times more energy released. So a magnitude 7 is not "twice as big" as a magnitude 6 in any ordinary sense. It shakes the ground about ten times as hard and releases something like 32 times the energy. Climb two steps, from magnitude 6 to magnitude 8, and you are talking about roughly a thousandfold jump in energy.

The Richter scale served well for moderate earthquakes but had a fatal flaw at the top end. It saturates above about magnitude 7, meaning the very largest earthquakes all came out with similar numbers even when one was vastly more destructive than another, because the waves Richter measured stop growing in proportion to the true size of giant ruptures. In 1979, Thomas Hanks and Hiroo Kanamori introduced the moment magnitude scale to fix this. Instead of reading a single wave amplitude, it is calculated from the physical ingredients of the rupture itself: the fault area that slipped, how far it slipped, and the rigidity of the rock. This gives the largest events the size they genuinely deserve, and it is the scale quoted whenever a major earthquake makes the news, even though people still loosely call it "the Richter scale."

From the Hidden Focus to the Triangulated Map

Every earthquake begins at a point in the crust where the rupture first breaks loose, called the focus, or hypocenter. The spot on the surface directly above it, the place people think of as the earthquake's location, is the epicenter. From the focus, energy radiates outward in several families of seismic waves that travel at different speeds, and that difference in speed turns out to be the trick that lets us find earthquakes at all.

The fastest are the P-waves, or primary waves, which are compressional, pushing and pulling the rock in the direction they travel, much like sound through the ground. Behind them come the slower S-waves, or secondary waves, which shear the rock side to side and cannot pass through liquid. Last and usually most destructive are the surface waves, which travel along the ground itself and do much of the shaking that topples buildings. A single seismograph trace shows this sequence laid out in time: P-wave first, then S-wave, then the long roll of surface waves.

The gap between the P-wave and S-wave arrivals is the key. Because the two waves travel at known, different speeds, the size of that time gap tells a seismologist how far away the focus was, the same way you estimate a lightning strike's distance from the delay between flash and thunder. A single station gives you the distance but not the direction, so you need at least three. Each station draws a circle of possible distances around itself, and the one place where all three circles meet is the epicenter. This triangulation, performed today across global networks that share their traces in near real time, pins down an earthquake anywhere on the planet within minutes.

When the Seafloor Lifts and the Ocean Answers

The largest earthquakes Earth produces, and the deadliest by a wide margin, occur on a special kind of fault hidden beneath the sea. At subduction zones, where one tectonic plate dives beneath another, the contact between the descending oceanic plate and the overriding plate forms a gently sloping surface called the megathrust. Because this interface can lock over an enormous area, it stores a colossal amount of strain before it fails, and when it does, it produces the biggest numbers on the magnitude scale.

The 1960 Valdivia earthquake off the coast of Chile reached magnitude 9.5, the largest ever instrumentally recorded, and the 2011 Tohoku earthquake off northeastern Japan reached magnitude 9.0 or 9.1. What makes megathrust events uniquely dangerous is not only their size but their geometry. When the fault ruptures, it thrusts a vast area of the seafloor upward by several meters, and that sudden displacement shoves the entire column of water above it. The result is a tsunami, a series of long-wavelength waves that can cross an ocean and devastate coastlines thousands of kilometers from the epicenter. The water that swept inland in Japan in 2011 was the direct consequence of the seabed jumping during those few minutes of rupture.

Why the Honest Answer Is "We Don't Know When"

Given everything we now understand, the obvious question is why we cannot simply predict the next one. We know where the dangerous faults are, and we know they load slowly and rupture suddenly. We can even estimate, from past earthquakes and the slow accumulation of strain, the long-term probability that a given fault will rupture within the coming decades. That kind of forecasting, expressed as odds over a span of years, is genuinely useful and underpins building codes and insurance.

What we cannot do is predict the day, the hour, the magnitude. The reason is woven into the physics described earlier. A fault stays locked until the stress just barely exceeds the friction holding it shut, and that threshold depends on details we cannot measure, the exact roughness of the fault surface kilometers underground, the precise distribution of stress, the pressure of fluids in tiny cracks. A tiny, unmeasurable difference at the critical moment decides whether a small patch slips harmlessly or whether the rupture cascades into a great earthquake. Decades of searching for reliable precursor signals, foreshocks, ground tilt, changes in well water, animal behavior, have produced nothing that works consistently, because the same small signals appear constantly without any earthquake following. The intellectually honest position, held across the scientific community, is that short-term earthquake prediction is not currently possible, and may never be. The best protection is not foretelling the future but building for the rupture we know is coming.

Key Takeaways

An earthquake is the rapid release of elastic strain that accumulates as the rock on either side of a locked fault slowly deforms under relentless plate motion, until friction gives way and the blocks snap past each other, an idea first formalized as elastic rebound theory after the 1906 San Francisco quake. Faults come in three types, strike-slip, normal, and reverse, that correspond directly to the three things plate boundaries do (slide past, pull apart, or collide), with the San Andreas as the classic strike-slip example. Magnitude is logarithmic, so each step up means about ten times the ground motion and roughly 32 times the energy, which is why the 1935 Richter scale was eventually replaced for large events by the 1979 moment magnitude scale built from fault area, slip, and rock rigidity. Earthquakes begin at a focus below an epicenter, send out P-, S-, and surface waves whose differing speeds let three stations triangulate the location, and reach their most catastrophic form at subduction megathrusts, where ruptures like Valdivia 1960 and Tohoku 2011 lift the seafloor and launch ocean-crossing tsunamis. We can forecast long-term probabilities, but the exact moment of rupture depends on inaccessible underground conditions and a vanishingly sensitive triggering threshold, which is why reliable short-term prediction remains beyond us.