arrow_back
In Depth
In Depth · 9 min read

Faults and Stress: Why Earthquakes Happen

An earthquake is the sudden release of stress that a fault has stored for decades or centuries. Here is how that stress builds, how faults give way, and how one rupture can set up the next.

Elastic Rebound Theory and the 1906 San Francisco Earthquake

On 18 April 1906 the San Andreas Fault ruptured along about 477 kilometers of northern California, in an earthquake now estimated at magnitude 7.9 (USGS n.d.b). Research by Gladys Hansen and Emmet Condon put the number of people who died directly or indirectly at more than 3,000, several times the figure of about 700 quoted for decades (USGS n.d.c).

The earthquake also produced the first working explanation of why faults fail. Harry Fielding Reid, a geologist at Johns Hopkins University, examined how the ground surface had moved during the earthquake and concluded that the rupture must have involved an elastic rebound of previously stored elastic stress (USGS n.d.a). Fences, roads and other straight lines that crossed the fault had been offset; a fence photographed by G. K. Gilbert near Bolinas was displaced about 8.5 feet along the fault trace (USGS n.d.a).

Reid set out the argument in the second volume of the state commission’s report (Reid 1910). Rock on either side of a fault is carried along by slow, steady motion, but the fault itself stays stuck. The crust bends like a spring, storing elastic strain, until the stress exceeds the strength of the fault. The fault then slips suddenly, the bent rock snaps back toward an unstrained shape, and the stored energy leaves as seismic waves.

The theory has a direct consequence. Once a fault has slipped, the same slow motion begins to load it again, so earthquakes on a given fault are expected to recur. Most of the modern science of faults and stress elaborates on that cycle of loading and release.

Types of Faults: Normal, Reverse, Thrust and Strike-Slip

The U.S. Geological Survey defines a fault as a fracture or zone of fractures between two blocks of rock that allows the blocks to move relative to each other (USGS n.d.d). Faults are classified by the angle of the fault surface, called the dip, and by the direction of slip along it (USGS n.d.d).

On a normal fault, the block above the inclined fault surface moves down relative to the block below. On a reverse fault the upper block moves up and over the lower one, and a reverse fault with a shallow dip is called a thrust fault. On a strike-slip fault the blocks slide sideways past each other, and faults that combine sideways and vertical motion are called oblique-slip faults (USGS n.d.d). A strike-slip fault is described as right-lateral or left-lateral according to which way the far side appears to move when viewed from across the fault. The San Andreas and North Anatolian faults are right-lateral; the East Anatolian Fault is left-lateral.

Each type reflects a different arrangement of stress. Ernest Masson Anderson (1905) observed that because the ground surface carries no shear stress, one of the three principal stresses in the shallow crust is close to vertical. If the largest compressive stress is vertical, as where the crust is being stretched, normal faults form. If the smallest is vertical, as where the crust is being squeezed, reverse and thrust faults form. If the intermediate stress is vertical, strike-slip faults form. With typical rock friction, the theory predicts normal faults dipping at about 60 degrees, thrusts at about 30 degrees and near-vertical strike-slip faults.

Stick-Slip Friction: How Faults Store and Release Energy

Elastic rebound describes the loading. Friction explains why the release is sudden rather than gradual. In laboratory experiments on rock, Brace and Byerlee (1966) found that surfaces under high pressure did not slide smoothly but stuck, slipped abruptly and stuck again. They proposed that shallow earthquakes may be this stick-slip motion on faults, and noted that the stress released in each slip event was only a small part of the stress the rock was carrying (Brace and Byerlee 1966).

Later work turned this observation into laws of rock friction. Scholz (1998) summarized how the outcome depends on whether friction rises or falls as sliding speeds up. Where friction strengthens with speed, any slip slows itself down and the fault slides stably. Where friction weakens with speed, slip can accelerate into an unstable rupture, which is an earthquake. Scholz argued that a single constitutive law of this kind accounts for a broad range of earthquake behavior, including how much of a fault’s motion is released in earthquakes and how much passes without them (Scholz 1998).

The distinction matters at the scale of whole faults. A fault patch with velocity-weakening friction can lock and later fail violently. A patch with velocity-strengthening friction tends to creep.

Locked Versus Creeping Faults on the San Andreas

Most of the San Andreas Fault is locked. Between large earthquakes its two sides stay stuck at depth while the plates continue to move, and strain accumulates in the surrounding rock as Reid described. The section that broke in 1906 behaves this way.

Between San Juan Bautista and Parkfield in central California, the fault behaves differently. There the two sides slide past each other more or less continuously in a process called aseismic creep, which moves the ground without producing large earthquakes. Titus et al. (2006) resurveyed seven alignment arrays, rows of survey markers set across the fault, and obtained average minimum slip rates of 21 to 26 millimeters per year over 33 to 36 years. The results indicate approximately steady creep along the central creeping segment for at least 35 years (Titus et al. 2006).

Creep is visible at the surface. Where a creeping fault crosses a town, curbs, sidewalks and walls are bent or offset a little more each year. The city of Hollister, near the northern end of the creeping section, is a well-known example, although the fault that runs through its streets is the Calaveras Fault, a branch of the San Andreas system, rather than the San Andreas itself.

A creeping section does not store strain in the way a locked one does, and it can also limit the spread of ruptures. After the magnitude 6.0 Parkfield earthquake of 2004, which struck at the southern end of the creeping section, GPS stations showed little slip during or after the earthquake along the central creeping segment (Titus et al. 2006). Creep does not make a fault harmless, since creeping faults can still host moderate earthquakes, but it changes how and where stress builds.

Stress Transfer and Coulomb Stress Change

When a fault slips, the stress it was carrying does not disappear. Part of it is redistributed to the surrounding crust, pushing some nearby faults closer to failure and moving others further away. Seismologists measure this with the Coulomb failure criterion, which combines the change in shear stress that drives a fault to slip with the change in the clamping stress that holds it shut (King et al. 1994). Stress changes are commonly given in bars; one bar is 0.1 megapascals, roughly atmospheric pressure at sea level.

The changes are small compared with the total stress in the crust, but they matter for faults that are already close to failure. King et al. (1994) found that aftershocks were abundant where the Coulomb stress on favorably oriented faults rose by more than about half a bar, and sparse where it fell by a similar amount. Regions where stress dropped are called stress shadows. The same study calculated that the 1992 Landers and Big Bear earthquakes in southern California raised stress on the San Bernardino section of the San Andreas Fault by 2 to 6 bars, enough to bring forward the next large earthquake there by about a decade (King et al. 1994).

Stein (1999) reviewed the evidence and concluded that when stress changes are translated into changes in probability, seismic hazard is strongly influenced by interactions between earthquakes. A fault’s chance of rupturing therefore depends on its own loading and on what its neighbors have recently done.

The North Anatolian Fault: An Earthquake Sequence Since 1939

The North Anatolian Fault, a right-lateral strike-slip fault that runs across northern Türkiye, offers the clearest example of earthquakes marching along a fault. The sequence began on 26 December 1939 with a magnitude 7.8 earthquake near Erzincan in the east, which killed 32,700 people (USGS n.d.e).

Over the following decades, large earthquakes broke successive stretches of the fault, mostly progressing westward. Stein et al. (1997) studied ten earthquakes of magnitude 6.7 or larger that together ruptured 1,000 kilometers of the fault between 1939 and 1992. They calculated that nine of the ten ruptures had been brought closer to failure by the shocks before them, typically by 1 to 10 bars, equivalent to 3 to 30 years of ordinary tectonic loading. The stress changes produced an average threefold gain in the probability of an earthquake during the decade after each event (Stein et al. 1997).

The authors used the same reasoning to look ahead. For the 30 years from 1996 they estimated a 12 percent chance of a magnitude 6.7 or larger earthquake near the port city of Izmit (Stein et al. 1997). On 17 August 1999, an earthquake of magnitude 7.4 in early estimates (USGS 2000), and 7.6 in the USGS report of the event (USGS 1999), struck that stretch of the fault. Government estimates put the losses at 17,127 deaths and 43,953 injuries, and more than 250,000 people were displaced (USGS 2000). The authors had not predicted the earthquake; they had identified the segment as more likely than most to fail.

The 1999 rupture moved the stress concentration further west, toward the faults beneath the Sea of Marmara near Istanbul. Taking the Izmit earthquake into account, Parsons et al. (2000) calculated a 62 ± 15 percent probability of strong shaking in Istanbul during the next 30 years and 32 ± 12 percent during the next decade. The calculation is a probability, not a prediction, and the faults beneath the Sea of Marmara remain a central concern for hazard planning in the region.

The 2023 Türkiye–Syria Earthquakes on the East Anatolian Fault

The East Anatolian Fault forms the southeastern boundary of the Anatolian block and, together with the North Anatolian Fault, accommodates the westward movement of Türkiye toward the Aegean Sea (USGS n.d.f). On 6 February 2023 it was the source of two large earthquakes nine hours apart.

The magnitude 7.8 Pazarcık earthquake ruptured nearly 354 kilometers of the East Anatolian Fault. The second, a magnitude 7.5 earthquake near Elbistan, ruptured roughly 161 kilometers of the separate Çardak Fault to the north (USGS n.d.g). USGS reported that the death toll in Türkiye and Syria exceeded 57,000 by the end of March 2023 (USGS n.d.g).

The two earthquakes show stress transfer at work, and its limits. Coulomb calculations indicate that the first earthquake raised stress by about 1 to 4 bars on the section of the Çardak Fault where the second rupture began, while lowering it by about 3 to 10 bars on the part of that fault running parallel to the East Anatolian Fault (Toda and Stein 2024). A stress increase of this size is small compared with the stress released in an earthquake, but it can be enough to trigger a fault that is already near failure (Stein et al. 2023).

The same calculations left open questions. Some sections of the East Anatolian Fault that received increased stress did not rupture, and part of the Çardak Fault that had not been stressed joined the magnitude 7.5 rupture. The second earthquake was also not preceded by foreshocks, despite the large and sudden loading (Stein et al. 2023). Stress transfer explains much of where earthquakes cluster, but it does not yet say which loaded fault will fail, or when.

Frequently asked questions

What is elastic rebound theory?
Elastic rebound theory holds that rock on either side of a locked fault slowly bends as the crust moves, storing elastic strain. When the stress exceeds the fault’s strength, the fault slips and the rock snaps back, releasing the energy as an earthquake. Harry Fielding Reid proposed it after studying ground movement in the 1906 San Francisco earthquake.
What are the three main types of faults?
Normal faults, where the upper block moves down, form where the crust is stretched. Reverse and thrust faults, where the upper block moves up, form where it is compressed. Strike-slip faults, where the blocks slide sideways past each other, include the San Andreas and the North Anatolian faults.
What is a creeping fault?
A creeping fault slides slowly and more or less continuously instead of staying locked and failing in large earthquakes. The San Andreas Fault creeps between San Juan Bautista and Parkfield in central California, where surveys show steady slip of roughly 21 to 26 millimeters per year for more than three decades.
Can one earthquake trigger another?
Yes. An earthquake redistributes stress, raising it on some nearby faults and lowering it on others. Faults already close to failure can be pushed over the edge, which is why aftershocks cluster where Coulomb stress has increased and large earthquakes sometimes follow one another along a fault.
Why did earthquakes move west along the North Anatolian Fault?
Between 1939 and 1992, ten earthquakes of magnitude 6.7 or larger ruptured about 1,000 kilometers of the fault, and nine of them struck where earlier shocks had raised the stress. The 1999 Izmit earthquake continued the sequence westward, increasing the estimated hazard for Istanbul.
What fault caused the 2023 Türkiye–Syria earthquake?
The magnitude 7.8 earthquake of 6 February 2023 ruptured nearly 354 kilometers of the East Anatolian Fault. A magnitude 7.5 earthquake about nine hours later ruptured roughly 161 kilometers of the nearby Çardak Fault, which the first earthquake had partly loaded with additional stress.

References

  1. Anderson, E. M. 1905. “The Dynamics of Faulting.” Transactions of the Edinburgh Geological Society 8 (3): 387–402. doi.org/10.1144/transed.8.3.387.
  2. Brace, W. F., and J. D. Byerlee. 1966. “Stick-Slip as a Mechanism for Earthquakes.” Science 153 (3739): 990–92. doi.org/10.1126/science.153.3739.990.
  3. King, Geoffrey C. P., Ross S. Stein, and Jian Lin. 1994. “Static Stress Changes and the Triggering of Earthquakes.” Bulletin of the Seismological Society of America 84 (3): 935–53. doi.org/10.1785/BSSA0840030935.
  4. Parsons, Tom, Shinji Toda, Ross S. Stein, Aykut Barka, and James H. Dieterich. 2000. “Heightened Odds of Large Earthquakes Near Istanbul: An Interaction-Based Probability Calculation.” Science 288 (5466): 661–65. doi.org/10.1126/science.288.5466.661.
  5. Reid, Harry Fielding. 1910. The Mechanics of the Earthquake. Vol. 2 of The California Earthquake of April 18, 1906: Report of the State Earthquake Investigation Commission. Publication 87. Washington, DC: Carnegie Institution of Washington.
  6. Scholz, Christopher H. 1998. “Earthquakes and Friction Laws.” Nature 391 (6662): 37–42. doi.org/10.1038/34097.
  7. Stein, Ross S. 1999. “The Role of Stress Transfer in Earthquake Occurrence.” Nature 402 (6762): 605–9. doi.org/10.1038/45144.
  8. Stein, Ross S., Aykut A. Barka, and James H. Dieterich. 1997. “Progressive Failure on the North Anatolian Fault since 1939 by Earthquake Stress Triggering.” Geophysical Journal International 128 (3): 594–604. doi.org/10.1111/j.1365-246X.1997.tb05321.x.
  9. Stein, Ross S., Shinji Toda, Ali Deger Özbakir, Volkan Sevilgen, Hector Gonzalez-Huizar, Gabriel Lotto, and Serkan Sevilgen. 2023. “Interactions, Stress Changes, Mysteries, and Partial Forecasts of the 2023 Kahramanmaraş, Türkiye, Earthquakes.” Temblor. doi.org/10.32858/temblor.299.
  10. Titus, Sarah J., Charles DeMets, and Basil Tikoff. 2006. “Thirty-Five-Year Creep Rates for the Creeping Segment of the San Andreas Fault and the Effects of the 2004 Parkfield Earthquake: Constraints from Alignment Arrays, Continuous Global Positioning System, and Creepmeters.” Bulletin of the Seismological Society of America 96 (4B): S250–68. doi.org/10.1785/0120050811.
  11. Toda, Shinji, and Ross S. Stein. 2024. “The Role of Stress Transfer in Rupture Nucleation and Inhibition in the 2023 Kahramanmaraş, Türkiye, Sequence, and a One-Year Earthquake Forecast.” Seismological Research Letters 95 (2A): 596–606. doi.org/10.1785/0220230252.
  12. USGS (U.S. Geological Survey). 1999. “Magnitude 7.6 Turkey, 1999 August 17 00:01:39 UTC.” National Earthquake Information Center preliminary report. Archived web page. webarchive.library.unt.edu/eot2008/20090510165706/http:/neic.usgs.gov/neis/eq_depot/1999/eq_990817.
  13. USGS (U.S. Geological Survey). 2000. Implications for Earthquake Risk Reduction in the United States from the Kocaeli, Turkey, Earthquake of August 17, 1999. Circular 1193. doi.org/10.3133/cir1193.
  14. USGS (U.S. Geological Survey). n.d.a. “Reid’s Elastic Rebound Theory.” The Great 1906 San Francisco Earthquake. Accessed November 16, 2026. earthquake.usgs.gov/earthquakes/events/1906calif/18april/reid.php.
  15. USGS (U.S. Geological Survey). n.d.b. “How Long Was the 1906 Rupture?” The Great 1906 San Francisco Earthquake. Accessed November 16, 2026. earthquake.usgs.gov/earthquakes/events/1906calif/18april/howlong.php.
  16. USGS (U.S. Geological Survey). n.d.c. “Casualties and Damage after the 1906 Earthquake.” The Great 1906 San Francisco Earthquake. Accessed November 16, 2026. earthquake.usgs.gov/earthquakes/events/1906calif/18april/casualties.php.
  17. USGS (U.S. Geological Survey). n.d.d. “What Is a Fault and What Are the Different Types?” Frequently Asked Questions. Accessed November 16, 2026. www.usgs.gov/faqs/what-fault-and-what-are-different-types.
  18. USGS (U.S. Geological Survey). n.d.e. “Historic Earthquakes: Erzincan, Turkey, 1939 December 26.” Archived web page. Accessed November 16, 2026. webarchive.library.unt.edu/web/20090509190328/http://earthquake.usgs.gov/regional/world/events/1939_12_26.php.
  19. USGS (U.S. Geological Survey). n.d.f. “M 7.8 – Pazarcik Earthquake, Kahramanmaras Earthquake Sequence.” Earthquake Hazards Program event page. Accessed November 16, 2026. earthquake.usgs.gov/earthquakes/eventpage/us6000jllz.
  20. USGS (U.S. Geological Survey). n.d.g. “Frequently Asked Questions about 2023 Earthquakes in Türkiye.” Earthquake Hazards Program. Accessed November 16, 2026. www.usgs.gov/programs/earthquake-hazards/science/frequently-asked-questions-about-2023-earthquakes-turkiye.
Next in In DepthEarth’s Interior, Seismic Waves and How Earthquakes Are MeasuredAll In Depth articles
sisma
The Pulse of a Living Planet
Loading earthquakes…