When Was the Last Time the San Andreas Fault Went Off?

The answer depends on which section of the fault you mean, because the San Andreas is not one uniform crack but a roughly 1,200-kilometer system that behaves differently along its length. The most recent large earthquake on the northern segment was the famous 1906 San Francisco event. The south-central section last ruptured in the 1857 Fort Tejon earthquake. And the southernmost stretch, running through the Coachella Valley, has not produced a major quake since about 1690. That last figure is the one that keeps seismologists up at night, because it means more than 300 years of accumulated stress on a segment that historically breaks every couple of centuries.

Three Segments, Three Different Timelines

The San Andreas fault runs from the Salton Sea in the south to Cape Mendocino in the north, but it does not rupture as a single unit. Geologists generally divide it into at least three major sections, each with its own earthquake history and recurrence pattern. Thinking of the fault as one entity that either “goes off” or doesn’t is one of the most common misconceptions about it.

The northern segment last broke on April 18, 1906, producing the magnitude 7.9 San Francisco earthquake. That rupture stretched roughly 470 kilometers. Triangulation surveys showed that slip along the fault reached as high as 8.6 meters near Shelter Cove and 7.5 meters at Tomales Bay, tapering to about 2.7 meters near Loma Prieta at the southern end.¹ Seismological analysis of 1906 identified two main patches of intense energy release separated by about 110 kilometers: one near the Golden Gate and a larger one between Point Reyes and Fort Ross.²

The south-central segment produced the magnitude 7.9 Fort Tejon earthquake on January 9, 1857. Its surface rupture ran about 350 kilometers. Detailed measurements from high-resolution topographic data show the 1857 event had an overall average displacement below 3.5 meters, with higher slip of 4 to 6 meters along the northwestern half of the rupture.³ Along the Carrizo Plain, a stretch that long loomed large in forecasting models, new measurements put the average 1857 slip at about 5.3 meters, higher than earlier estimates and reshaping assumptions about how that segment behaves.⁴

The southernmost segment, running through the Coachella Valley past Palm Springs, has not produced a large earthquake during any period of historical record. Paleoseismic trenching places the most recent event around 1690, more than 300 years ago.⁵ Geodetic measurements show the crust on either side of this quiet segment is accumulating strain at a high rate, and geological evidence indicates very large earthquakes have struck every two to three centuries.⁶

Why the Southernmost Section Worries Seismologists Most

Of the three major segments, the Coachella Valley stretch has the clearest case for being overdue. Studies at the Thousand Palms Oasis site on the Mission Creek strand found an average repeat time of about 215 years between surface-rupturing earthquakes, yet more than 325 years have passed since the last one.⁷ That gap has widened with every passing year and now exceeds any known interval in the paleoseismic record for that site.

Satellite radar measurements reinforce the concern. High-resolution data from synthetic aperture radar show that the southern San Andreas is splitting strain roughly evenly with the parallel San Jacinto fault, and the observed strain rates confirm the southern San Andreas may be nearing the end of its quiet period between large earthquakes.⁸ InSAR and GPS velocities in the area agree to within a couple of millimeters per year, giving scientists confidence that their strain maps are accurate.⁹

“Overdue” in earthquake science is a tricky word. Faults don’t run on a fixed clock the way a bus schedule does. But the combination of a long quiet period, a short historical average, and direct measurements of accumulating strain makes this segment the consensus top concern along the entire San Andreas system.

What Wrightwood Tells Us About Recurrence

One of the richest earthquake records anywhere on the San Andreas comes from Wrightwood, a small community in the San Gabriel Mountains northeast of Los Angeles. A structural depression there has been filling with sediment for thousands of years, trapping evidence of past earthquakes in its layered deposits. Fault scarps, colluvial wedges, and broken sediment layers record large events in 1857, 1812, and approximately 1700, 1610, and 1470.¹⁰

The full record at Wrightwood extends back about 6,000 years and captures at least 30 prehistoric earthquakes. Over the past 1,500 years the record appears complete, with a mean recurrence interval of about 105 years, though individual intervals range widely from 31 to 165 years. Average slip per event is about 3.2 meters, but that too spans a wide range from less than a meter to 7 meters.¹¹ What stands out from this long record is how irregular the pattern is. Earthquakes cluster during some centuries and thin out during others. The slip in one earthquake does a poor job of predicting when the next will strike, and the sequence is only slightly more ordered than random. The strongest pattern the researchers found is that earthquakes tend to bunch together after periods when strain has been building faster than it is released across multiple cycles.

That finding matters for anyone expecting the San Andreas to behave like a pendulum swinging back and forth at a predictable tempo. The Wrightwood data suggest the fault is more like a dripping faucet where the drops sometimes come rapidly and sometimes pause for an uncomfortably long time.

The Forgotten 1812 Earthquake

Between the well-known 1857 and 1906 events sits another large earthquake that often gets overlooked: the estimated magnitude 7.5 quake of December 8, 1812. This one is peculiar because the damage it caused and the paleoseismic evidence do not fit neatly on any single fault strand. Dynamic rupture modeling has shown that the 1812 event is best explained by a rupture that began on the San Jacinto fault and jumped onto the San Andreas fault during the same earthquake.¹² Old coast redwood and other trees near Wrightwood captured the event in their growth rings, recording disruption in the fall or winter of 1812 to 1813.¹³

The possibility that ruptures can jump between the San Andreas and the San Jacinto is important for hazard estimates, because the two faults run roughly parallel through the most populated parts of southern California. Paleoseismic dating suggests that joint ruptures bridging the northern San Jacinto and the San Andreas have occurred more than once, though only a couple of ruptures appear to have broken through the complicated San Gorgonio Pass region that separates them.¹⁴ The lesson: future San Andreas earthquakes in southern California might not stay on the San Andreas alone.

An Ancient Lake That May Have Triggered Earthquakes

One of the more surprising recent findings involves Lake Cahuilla, a vast body of water that repeatedly filled and dried up in the Salton Trough over the past millennium as the Colorado River changed course. New geologic and paleoseismic data show that the past six major earthquakes on the southernmost San Andreas probably occurred during times when the lake was full.¹⁵ A lake the size of Cahuilla (comparable to the Salton Sea but far deeper) would have pressed enormous weight onto the crust. Modeling shows that this load increased stress on the fault by several hundred kilopascals and more than doubled the rate at which stress built up, likely enough to push the fault over the edge.

The flip side is also interesting. Lake Cahuilla has been dry for centuries. Researchers have tested whether the current drought in the Salton Trough could explain why the southernmost segment has been quiet for so long. Earthquake-cycle modeling finds that the pore-pressure reduction needed to delay a rupture from the expected sub-200-year interval to more than 300 years is unrealistically large.¹⁶ In other words, the absence of the lake may be slowing things down somewhat, but it cannot be the whole explanation for the long gap.

The Creeping Section and Why Part of the Fault Never “Goes Off”

Between the locked northern and south-central segments lies a roughly 150-kilometer stretch through central California, near the town of Parkfield, where the San Andreas fault moves almost continuously without producing major earthquakes. This creeping section releases strain through slow, steady sliding and frequent small earthquakes rather than through rare, catastrophic ones.

Drilling into the fault at the San Andreas Fault Observatory at Depth (SAFOD) near Parkfield revealed why. Core samples from about 2.7 kilometers down showed that the actively creeping strands have extremely low friction, with a coefficient of friction around 0.1, far below what rock typically exhibits. The culprit is a magnesium-rich clay mineral, saponite, one of the weakest known minerals in this class.¹⁷ Surrounding wall rock and other fault strands at the same depth behave very differently, exhibiting higher friction and the kind of stick-slip behavior associated with earthquakes.¹⁸ The creeping section and the locked sections of the fault also differ in the statistical patterns of their small earthquakes, with locked segments showing distinctly different fractal clustering of quake locations compared to creeping ones.¹⁹

The creeping section matters to the “when did it last go off” question because it effectively divides the San Andreas into independent northern and southern halves. A rupture on one side is unlikely to propagate through the creep zone to trigger the other. That is why the 1906 quake did not extend south past central California, and why 1857 did not propagate north.

Offshore Evidence and the Northern Segment’s Rhythm

On land, paleoseismic records for the northern San Andreas are sparser than for the south. But the ocean floor off northern California has preserved a complementary archive. Researchers collected dozens of piston and gravity cores from submarine channels draining the continental margin and found a series of turbidite beds, layers of disturbed sediment triggered by strong shaking. The late Holocene turbidite record passes multiple tests for having been triggered by earthquakes rather than storms or other causes, and it can be correlated from site to site along the coast.²⁰

During the last roughly 2,800 years, 15 turbidites have been identified offshore, including one attributed to the 1906 earthquake. That works out to an average repeat time of about 200 years, close to the onshore estimate of about 230 years for the northern segment. At least eight of the ten most recent events appear to have ruptured the full 320-kilometer distance from the Mendocino Triple Junction to near San Francisco. Onshore paleoseismic data at several sites place the penultimate northern event in the mid-1600s, and a third event around 1300.²¹ So the northern segment’s most recent rupture in 1906 came after a gap of roughly 250 to 300 years from its predecessor, slightly longer than average.

Could a Nearby Quake Set Off the San Andreas?

Earthquakes interact through the crust, transferring stress to neighboring faults. The 2019 Ridgecrest earthquake sequence in the Mojave Desert prompted researchers to calculate whether stress changes had nudged the San Andreas closer to failure. On its own, the Ridgecrest event had a modest effect. But modeling showed that if a follow-on rupture had occurred on a nearby fault and reached within about 45 kilometers of the San Andreas, it would have raised the probability of a Mojave-section San Andreas rupture by a factor of roughly 150. Even without that worst case, the estimated chance of a large San Andreas earthquake in the 12 months after Ridgecrest was about 1.15 percent, or three to five times the background probability.²²

That might sound small, but for context, background annual probabilities for any individual fault section are typically well below one percent. Regional earthquakes that shift stress onto the San Andreas are a recognized ingredient in the overall hazard picture, even if no single nearby event makes a San Andreas rupture likely in absolute terms.

What a Future Southern Rupture Might Look Like

California’s emergency planners have spent years modeling a scenario called ShakeOut, in which a magnitude 7.8 earthquake ruptures the southern San Andreas from the Salton Sea northwest toward Los Angeles. Simulations of this scenario use spontaneous rupture models to estimate ground shaking across the region.²³ Early versions predicted very strong long-period shaking in the Los Angeles Basin, driven by seismic waves channeling through a corridor of interconnected sedimentary basins.²⁴ Later research incorporating updated subsurface geology and surface topography has refined those estimates, though the broad picture of severe shaking in the LA area remains.²⁵

Beyond shaking, the ShakeOut scenario highlighted infrastructure vulnerabilities that many people do not think about. The major aqueducts that carry water into southern California from the north and east all cross the San Andreas fault. Simulations of the water system’s response to a magnitude 7.8 event found that aqueducts would sustain significant damage, with restoration of water flow estimated to take between 4 and 18 months.²⁶ A region of nearly 20 million people could face months of severely constrained water supply.

Using Tree Rings to Fill the Gaps

Between written records (which only go back to the mid-1700s in California) and paleoseismic trenches (which are expensive and site-specific), dendrochronology offers a third way to date past San Andreas earthquakes. Trees growing along or near the fault record earthquake effects in their annual rings: tilted trunks produce reaction wood, severed roots slow growth, and co-seismic changes in groundwater can alter ring width for years afterward.

The approach has already confirmed the 1812 earthquake in tree rings at Wrightwood. Researchers are now extending the method to coast redwoods along the northern San Andreas, where the trees’ exceptional lifespans could potentially resolve late-Holocene earthquakes year by year. Progress has been limited by the difficulty of crossdating rings in these long-lived species and the scarcity of suitable old-growth specimens, but the technique holds promise for building a more continuous timeline.²⁷ Researchers have also used USGS data from other tree species to identify fault movements that predate the earliest historical accounts.²⁸ With traditional trenching studies expensive and spatially limited, trees offer a way to fill in gaps in both time and geography along the fault.

Parkfield’s repeating microearthquakes, meanwhile, have provided a testing ground for whether any part of the San Andreas is temporally predictable at all. Retrospective forecasting experiments on small repeating quakes at Parkfield demonstrated genuine predictability in their timing, suggesting that at least some fault behaviors follow quantifiable patterns even if large earthquakes remain stubbornly hard to forecast.²⁹ Whether those patterns scale up to the big ruptures that people care about most remains an open question, and the irregular Wrightwood record suggests the answer is: only loosely, at best.