Is the San Andreas “locked, loaded, and ready to go”?

Send me Earthquake Insights

Volkan and I presented and exhibited Temblor at the National Earthquake Conference in Long Beach last week. Prof. Thomas Jordan, USC University Professor, William M. Keck Foundation Chair in Geological Sciences, and Director of the Southern California Earthquake Center (SCEC), gave the keynote address. Tom has not only led SCEC through fifteen years of sustained growth and achievement, but he’s also launched countless initiatives critical to earthquake science, such as the Uniform California Earthquake Rupture Forecasts (UCERF), and the international Collaboratory for Scientific Earthquake Predictability (CSEP), a rigorous independent protocol for testing earthquake forecasts and prediction hypotheses.

In his speech, Tom argued that to understand the full range and likelihood of future earthquakes and their associated shaking, we must make thousands if not millions of 3D simulations. To do this we need to use the next generation of super-computers—because the current generation is too slow! The shaking can be dramatically amplified in sedimentary basins and when seismic waves bounce off deep layers, features absent or muted in current methods. This matters, because these probabilistic hazard assessments form the basis for building construction codes, mandatory retrofit ordinances, and quake insurance premiums. The recent Uniform California Earthquake Rupture Forecast Ver. 3 (Field et al., 2014) makes some strides in this direction. And coming on strong are earthquake simulators such as RSQsim (Dieterich and Richards-Dinger, 2010) that generate thousands of ruptures from a set of physical laws rather than assumed slip and rupture propagation. Equally important are CyberShake models (Graves et al., 2011) of individual scenario earthquakes with realistic basins and layers.

But what really caught the attention of the mediaand the public—was just one slide

Tom closed by making the argument that the San Andreas is, in his words, “locked, loaded, and ready to go.” That got our attention. And he made this case by showing one slide. Here it is, photographed by the LA Times and included in a Times article by Rong-Gong Lin II that quickly went viral.


Believe it or not, Tom was not suggesting there is a gun pointed at our heads. ’Locked’ in seismic parlance means a fault is not freely slipping; ‘loaded’ means that sufficient stress has been reached to overcome the friction that keeps it locked. Tom argued that the San Andreas system accommodates 50 mm/yr (2 in/yr) of plate motion, and so with about 5 m (16 ft) of average slip in great quakes, the fault should produce about one such event a century. Despite that, the time since the last great quake (“open intervals” in the slide) along the 1,000 km-long (600 mi) fault are all longer, and one is three times longer. This is what he means by “ready to go.” Of course, a Mw=7.7 San Andreas event did strike a little over a century ago in 1906, but Tom seemed to be arguing that we should get one quake per century along every section, or at least on the San Andreas.

Could it be this simple?

Now, if things were so obvious, we wouldn’t need supercomputers to forecast quakes. In a sense, Tom’s wake-up call contradicted—or at least short-circuited—the case he so eloquently made in the body of his talk for building a vast inventory of plausible quakes in order to divine the future. But putting that aside, is he right about the San Andreas being ready to go?

Because many misaligned, discontinuous, and bent faults accommodate the broad North America-Pacific plate boundary, the slip rate of the San Andreas is generally about half of the plate rate. Where the San Andreas is isolated and parallel to the plate motion, its slip rate is about 2/3 the plate rate, or 34 mm/yr, but where there are nearby parallel faults, such as the Hayward fault in the Bay Area or the San Jacinto in SoCal, its rate drops to about 1/3 the plate rate, or 17 mm/yr. This means that the time needed to store enough stress to trigger the next quake should not—and perhaps cannot—be uniform. So, here’s how things look to me:

The San Andreas (blue) is only the most prominent element of the 350 km (200 mi) wide plate boundary. Because ruptures do not repeat—either in their slip or their inter-event time—it’s essential to emphasize that these assessments are crude. Further, the uncertainties shown here reflect only the variation in slip rate along the fault. The rates are from Parsons et al. (2014), the 1857 and 1906 average slip are from Sieh (1978) and Song et al. (2008) respectively. The 1812 slip is a model by Lozos (2016), and the 1690 slip is simply a default estimate.

So, how about ‘locked, generally loaded, with some sections perhaps ready to go’

When I repeat Tom’s assessment in the accompanying map and table, I get a more nuanced answer. Even though the time since the last great quake along the southernmost San Andreas is longest, the slip rate there is lowest, and so this section may or may not have accumulated sufficient stress to rupture. And if it were ready to go, why didn’t it rupture in 2010, when the surface waves of the Mw=7.2 El Major-Cucapah quake just across the Mexican border enveloped and jostled that section? The strongest case can be made for a large quake overlapping the site of the Great 1857 Mw=7.8 Ft. Teton quake, largely because of the uniformly high San Andreas slip rate there. But this section undergoes a 40° bend (near the ‘1857’ in the map), which means that the stresses cannot be everywhere optimally aligned for failure: it is “locked” not just by friction but by geometry.

A reality check from Turkey

Sometimes simplicity is a tantalizing mirage, so it’s useful to look at the San Andreas’ twin sister in Turkey: the North Anatolian fault. Both right-lateral faults have about the same slip rate, length, straightness, and range of quake sizes; they both even have a creeping section near their midpoint. But the masterful work of Nicolas Ambraseys, who devoured contemporary historical accounts along the spice and trade routes of Anatolia to glean the record of great quakes (Nick could read 14 languages!) affords us a much longer look than we have of the San Andreas.

The idea that the duration of the open interval can foretell what will happen next loses its luster on the North Anatolian fault because it’s inter-event times, as well as the quake sizes and locations, are so variable. If this 50% variability applied to the San Andreas, no sections could be fairly described as ‘overdue’ today. Tom did not use this term, but others have. We should, then, reserve ‘overdue’ for an open interval more than twice the expected inter-event time.

This figure of North Anatolian fault quakes is from Stein et al. (1997), updated for the 1999 Mw=7.6 Izmit quake, with the white arrows giving the direction of cascading quakes. Even though 1939-1999 saw nearly the entire 1,000 km long fault rupture in a largely western falling-domino sequence, the earlier record is quite different. When we examined the inter-event times (the time between quakes at each point along the fault), we found it to be 450±220 years. Not only was the variation great—50% of the time between quakes—but the propagation direction was also variable.

However, another San Andreas look-alike, the Alpine Fault in New Zealand, has a record of more regular earthquakes, with an inter-event variability of 33% for the past 24 prehistoric quakes (Berryman et al., 2012). But the Alpine fault is straighter and more isolated than the San Andreas and North Anatolian faults, and so earthquakes on adjacent faults do not add or subtract stress from it. And even though the 31 mm/yr slip rate on the southern Alpine Fault is similar to the San Andreas, the mean inter-event time on the Alpine is longer than any of the San Andreas’ open intervals: 330 years. So, while it’s fascinating that there is a ‘metronome fault’ out there, the Alpine is probably not a good guidepost for the San Andreas.

If Tom’s slide is too simple, and mine is too equivocal, what’s the right answer?

I believe the best available answer is furnished by the latest California rupture model, UCERF3. Rather than looking only at the four San Andreas events, the team created hundreds of thousands of physically plausible ruptures on all 2,000 or so known faults. They found that the mean time between Mw≥7.7 shocks in California is about 106 years (they report an annual frequency of 9.4 x 10^-3 in Table 13 of Field et al., 2014; Mw=7.7 is about the size of the 1906 quake; 1857 was probably a Mw=7.8, and 1812 was probably Mw=7.5). In fact, this 106-year interval might even be the origin of Tom’s ‘once per century’ expectation since he is a UCERF3 author.

But these large events need not strike on the San Andreas, let alone on specific San Andreas sections, and there are a dozen faults capable of firing off quakes of this size in the state. While the probability is higher on the San Andreas than off, in 1872 we had a Mw=7.5-7.7 on the Owen’s Valley fault (Beanland and Clark, 1994). In the 200 years of historic records, the state has experienced up to three Mw≥7.7 events, in southern (1857) and eastern (1872), and northern (1906) California. This rate is consistent with, or perhaps even a little higher than, the long-term model average.

So, what’s the message

While the southern San Andreas is a likely candidate for the next great quake, ‘overdue’ would be over-reach, and there are many other fault sections that could rupture. But since the mean time between Mw≥7.7 California shocks is about 106 years, and we are 110 years downstream from the last one, we should all be prepared—even if we cannot be forewarned.

Ross Stein (, Temblor

You can check your home’s seismic risk at Temblor

References cited:

Sarah Beanland and Malcolm M. Clark (1994), The Owens Valley fault zone, eastern California, and surface faulting associated with the 1872 earthquake, U.S. Geol. Surv. Bulletin 1982, 29 p.

Kelvin R. Berryman, Ursula A. Cochran, Kate J. Clark, Glenn P. Biasi, Robert M. Langridge, Pilar Villamor (2012), Major Earthquakes Occur Regularly on an Isolated Plate Boundary Fault, Science, 336, 1690-1693, DOI: 10.1126/science.1218959

James H. Dietrich and Keith Richards-Dinger (2010), Earthquake recurrence in simulated fault systems, Pure Appl. Geophysics, 167, 1087-1104, DOI: 10.1007/s00024-010-0094-0.

Edward H. (Ned) Field, R. J. Arrowsmith, G. P. Biasi, P. Bird, T. E. Dawson, K. R., Felzer, D. D. Jackson, J. M. Johnson, T. H. Jordan, C. Madden, et al.(2014). Uniform California earthquake rupture forecast, version 3 (UCERF3)—The time-independent model, Bull. Seismol. Soc. Am. 104, 1122–1180, doi: 10.1785/0120130164.

Robert Graves, Thomas H. Jordan, Scott Callaghan, Ewa Deelman, Edward Field, Gideon Juve, Carl Kesselman, Philip Maechling, Gaurang Mehta, Kevin Milner, David Okaya, Patrick Small, Karan Vahi (2011), CyberShake: A Physics-Based Seismic Hazard Model for Southern California, Pure Appl. Geophysics, 168, 367-381, DOI: 10.1007/s00024-010-0161-6.

Julian C. Lozos (2016), A case for historical joint rupture of the San Andreas and San Jacinto faults, Science Advances, 2, doi: 10.1126/sciadv.1500621.

Tom Parsons, K. M. Johnson, P. Bird, J.M. Bormann, T.E. Dawson, E.H. Field, W.C. Hammond, T.A. Herring, R. McCarey, Z.-K. Shen, W.R. Thatcher, R.J. Weldon II, and Y. Zeng, Appendix C—Deformation models for UCERF3, USGS Open-File Rep. 2013–1165, 66 pp.

Seok Goo Song, Gregory C. Beroza and Paul Segall (2008), A Unified Source Model for the 1906 San Francisco Earthquake, Bull. Seismol. Soc. Amer., 98, 823-831, doi: 10.1785/0120060402

Kerry E. Sieh (1978), Slip along the San Andreas fault associated with the great 1857 earthquake, Bull. Seismol. Soc. Am., 68, 1421-1448.

Ross S. Stein, Aykut A. Barka, and James H. Dieterich (1997), Progressive failure on the North Anatolian fault since 1939 by earthquake stress triggering, Geophys. J. Int., 128, 594-604, 1997, 10.1111/j.1365-246X.1997.tb05321.x