Why Earthquakes Cluster Along Faults

If you map every earthquake recorded in California over the past century and overlay it on the state's geology, you see something striking: the dots do not scatter evenly. They trace lines. The San Andreas, the Hayward, the Garlock, the Calaveras — each fault appears as a bright seam of seismicity, while the rock between them sits quiet. Earthquakes, it turns out, are creatures of habit. They prefer to happen where earthquakes have already happened.

The first reason is mechanical. Intact rock is strong. To break a fresh slab of granite under crustal conditions, you need to overcome enormous cohesive forces — the chemical bonds that hold mineral grains together. A pre-existing fault, by contrast, is not intact rock but a zone of crushed, ground-up material called fault gouge, often lubricated by water and clay minerals. Sliding on such a surface requires only that friction be overcome, and friction in a mature fault zone can be a small fraction of the strength of unbroken rock. Once a crack exists, the cheapest way for the crust to relieve stress is to use it again. This is why faults persist for millions of years: each rupture leaves the surface a little weaker, biasing the next failure toward the same plane.

The second reason is that earthquakes talk to each other. When one patch of fault slips, it does not simply release stress; it redistributes it. The rock on either side of the rupture relaxes, but the rock at the rupture's tips is squeezed harder than before. This phenomenon, called stress transfer, means that a magnitude-7 event on one segment can load adjacent segments closer to their own failure threshold. Sometimes the loaded segment ruptures within seconds, producing a cascading mainshock. Sometimes it waits years. The 1992 Landers earthquake in the Mojave, for instance, is thought to have advanced the clock on several nearby faults by decades. Aftershock sequences are the most visible expression of this dialogue: the thousands of small tremors that follow a large quake are not random rattles but the crust's way of catching up with the new stress field.

A third factor is geometry. Plate motion is steady on long timescales — the Pacific Plate slides past the North American Plate at roughly 50 millimeters per year — but that motion has to be accommodated somewhere. Once a fault system has organized itself to carry that displacement, it would take a major reorganization of the entire crust to abandon it for a fresh one. The path of least resistance, again, is the path already cut. Faults thus inherit their importance: a strand that took up slip a million years ago is likely still taking up slip today, and the strain budget of the region keeps loading it.

This clustering has practical consequences. It is why seismic hazard maps are not uniform smears of risk but sharply patterned, with the highest probabilities pressed against known fault traces. It is also why paleoseismology — the practice of trenching across faults to read the layered evidence of past ruptures — works at all. If earthquakes were scattered randomly across the crust, the recurrence intervals of individual faults would be meaningless. Because they are not, a trench across the San Andreas can yield a calendar of prehistoric quakes stretching back thousands of years.

There is a temptation to read all this as a tidy story: weak faults, stress transfer, inherited geometry, repeat. But the picture has ragged edges. Some large earthquakes occur on faults nobody had mapped, in places the clustering rule did not predict. The 1994 Northridge event ruptured a blind thrust beneath Los Angeles that left no surface trace. The crust contains more flaws than we can see, and our maps of where earthquakes belong are always provisional. Clustering is the dominant pattern, not the whole pattern. The faults we know about will probably host the next big one — but the crust reserves the right to surprise us.