Earthquakes occur near the San Andreas Fault because two massive slabs of Earth’s crust are grinding past each other, and the friction between them causes rock to lock up, bend under stress, and eventually snap. The Pacific Plate slides north-northwestward past the North American Plate at roughly 36 millimeters per year, about the speed your fingernails grow. That motion never stops, but the fault itself doesn’t move smoothly. Stress builds for decades or centuries, then releases in seconds as an earthquake.
Two Plates Sliding Past Each Other
The San Andreas Fault is a transform plate boundary, meaning the two plates involved move sideways relative to each other rather than colliding head-on or pulling apart. The Pacific Plate, carrying a thin sliver of western California, grinds past the North American Plate along a zone stretching about 1,200 kilometers (800 miles) from the Salton Sea in the south to Cape Mendocino in the north. It is the longest fault in California and one of the longest in North America.
At this type of boundary, crust is neither created nor destroyed. Instead, blocks of rock are torn apart across a broad zone of shearing. That grinding action produces shallow earthquakes and large sideways displacement of rock over time. Entire landscapes have been reshaped by it: Point Reyes, the Channel Islands, and Pinnacles National Park all owe their positions to millions of years of this lateral movement.
How Stress Builds and Releases
The plates move constantly, but friction along the fault surface keeps most sections locked in place. The rock on either side of the fault bends and deforms as the plates push past each other, storing enormous elastic energy the way a rubber band stretches when you pull it. This process can continue for decades or even centuries.
Eventually, the accumulated strain overcomes the friction holding the fault together. The rock snaps back to its unstrained shape in a sudden rupture, releasing energy as seismic waves. This concept, known as elastic rebound, was first proposed after the 1906 San Francisco earthquake and remains the core explanation for why earthquakes happen along faults like the San Andreas. The stronger the friction and the longer the fault stays locked, the more energy builds up and the larger the eventual earthquake.
Locked Segments vs. Creeping Segments
Not every part of the San Andreas Fault behaves the same way, and this is key to understanding where the biggest earthquakes strike. The fault has three distinct behavioral zones.
The northern section, running roughly from San Juan Bautista up through the San Francisco Bay Area, is currently locked. It has produced no detectable movement and very few earthquakes since the magnitude 7.9 San Francisco earthquake of 1906. The southern section, from about Cholame down past Los Angeles, is also locked. It last ruptured in the magnitude 7.9 Fort Tejon earthquake of 1857. Both of these locked segments are accumulating stress right now.
Between them sits a creeping section where the fault slips continuously without building up large amounts of strain. This stretch, running from San Juan Bautista to Parkfield, produces frequent small earthquakes (mostly magnitude 5 and below) but no large ones. The town of Parkfield itself sits in a transition zone between the creeping and locked behavior, making it one of the most intensely studied earthquake sites in the world.
The locked segments are the dangerous ones. Because they don’t release strain gradually, they store energy over long periods and eventually produce powerful earthquakes. The creeping section acts more like a pressure valve, letting stress escape in small, mostly harmless increments.
How Often Large Earthquakes Strike
Geological trenching along the southern San Andreas near Wrightwood, about 70 kilometers northeast of Los Angeles, has revealed evidence of five large earthquakes over the past five centuries: in roughly 1470, 1610, 1700, 1812, and 1857. That works out to an average recurrence interval of about 100 years. The southern section has now gone over 165 years without a major rupture, well beyond that historical average.
The northern section tells a similar story. The 1906 San Francisco earthquake ruptured a 350-kilometer (220-mile) segment of the fault and was felt as far away as Las Vegas. More than a century has passed since then, and geodetic measurements show that section remains locked and loading with stress. The 1989 Loma Prieta earthquake (magnitude 6.9) occurred on a nearby segment but did not relieve the strain on the main locked portion.
How Scientists Track Fault Movement
Because the San Andreas produces earthquakes on human timescales, it is one of the most heavily monitored faults on Earth. Networks of GPS stations and strainmeters across western California continuously measure how the ground deforms as the plates move. Even tiny shifts of a millimeter or less can be detected, allowing scientists to map exactly where strain is building and how fast.
At Parkfield, the USGS operates the San Andreas Fault Observatory at Depth, which monitors small to moderate earthquakes and measures rock deformation deep underground throughout the earthquake cycle. These instruments track the transition between the creeping and locked zones in real time, providing data that helps refine estimates of when and where the next large rupture is most likely.
Why the Risk Remains High
The fundamental reason earthquakes keep happening along the San Andreas is simple: the plates never stop moving. Every year, another 36 millimeters of motion is added to the strain budget on locked segments. That energy has to go somewhere eventually, and the only way it can be released on a locked fault is through an earthquake. The longer a segment stays quiet, the more strain it accumulates and the larger the potential rupture becomes.
Southern California’s locked section is of particular concern because it runs near densely populated areas including Los Angeles and sits well past its average recurrence interval. The physics of the fault guarantee that large earthquakes will continue to occur here. The only uncertainty is timing.