Was a famous earthquake precursor a phantom?

By David Jacobson, M.Sc. and Ross Stein, Ph.D., Temblor

Check your hazard rank

The 1999 Izmit earthquake killed 17,000 people and left half a million homeless.


In 1999, the M=7.6 Izmit earthquake killed approximately 17,000 people and left half a million homeless. In a study published in 2011, Michel Bouchon (Institut des Sciences de la Terre de l’Université de Grenoble) and his colleagues discovered a burst of foreshocks at or near the epicenter of the mainshock during 44 min before the mainshock, and inferred that these quakes were accompanied by slow slip. They argued that the slow slip could be a diagnostic precursor detectable for other quakes by GPS receivers. Now, a new study by William Ellsworth (Stanford University) and Fatih Bulut (Kandilli Observatory and Earthquake Research Institute, Boğaziçi University, İstanbul) used seismic techniques unavailable in 2011 to re-examine the sequence. They conclude while the foreshocks are real, the slow slip is not.


What is a foreshock?

How to read the Izmit foreshock could have great implications and benefits for future earthquake prediction, that evanescent but elusive seismological goal.

This figure, simplified from Ellsworth and Bulut (2018), shows the location of the 1999 M=7.6 Izmit earthquake, as well as the foreshocks, all of which struck near the future mainshock. The time series to the right highlights the foreshock sequence in the 44 minutes up until the mainshock. There were no shocks there during the preceding year.


Foreshocks are small earthquakes that precede a larger event. While that sounds simple, it’s important to know that the term ‘foreshocks’ is always applied after the mainshock strikes; no one has figured out how to identify a quake as a foreshock in advance. And so, the Bouchon et al finding that the shocks accompanied something quite rare and, in principal, detectable, could profoundly change their predictive value.

While less than one in twenty shocks is followed by a larger event, foreshocks can provide insight into the earthquake nucleation process. Today, there are two schools of thought about what a foreshock signals. One is that they accompany accelerating slip surrounding or near the eventual mainshock location. If this were the case, that slip could be used to forecast a future large event. The other view is that each foreshock is triggered due to the stress transferred by the preceding event, in a ‘cascade’ akin to falling dominoes. This is known as the cascade theory and implies that the foreshocks are no different than the mainshock; the mainshock just happens to be much larger.

The Osman Gazi Bridge which opened in 2016, crosses the 1999 Izmit rupture.



In the 2011 study, the scientists used seismic signals from a single station to conclude that all foreshocks showed the same waveforms, and that the same fault patch ruptured repeatedly, which they inferred had to be caused by slow slip. However, in the study published last week, by using seismic signals from nine stations, rather than one, Ellsworth and Bulut found that the foreshocks instead migrated toward the future epicenter, like falling dominoes.

“The authors of the 2011 Science paper were quite optimistic, but what they proposed had happened
did not happen,” Ellsworth said. “I have no argument with what Michel did: he proposed an
interesting and testable hypothesis. Wish we all did this all the time.”



While Ellsworth and Bulut argue they have falsified the theory that the Izmit earthquake was preceded by a period of slow slip less than an hour before the mainshock, others remain unconvinced. Michel Bouchon told us today that, “We respect the work of Ellsworth and Bulut on the Izmit earthquake nucleation, but we disagree with their interpretation that a succession of cascading ruptures, migrating along fault and driven only by coseismic stress transfer, led to the Izmit earthquake nucleation. For us the records of the foreshocks at the two best and closest stations to the epicenter do not show any significant migration of the foreshock source.” Because of this, Bouchon said that they stand by their original interpretation.

Bouchon added that, “We also believe that the long duration of the foreshock sequence (44 minutes) argues for a slower nucleation process than a purely elastic cascade where stress transfer would be partly controlled by elastic wave propagation. The striking acceleration of foreshock occurrences that we observed as the time of the earthquake approaches also supports, for us, a driving mechanism other than purely coseismic stress interactions.”

The stress interactions proposed in Ellsworth and Bulut have also been called into question by Professor David Marsan at the Institut des Sciences de la Terre of Université de Savoie Mont Blanc, who was not an author in the 2011 study. He pointed out that, “Of the 4 large foreshocks, 3 have stress drops in the 26 – 83 MPa range. If you compare these estimates with for example those of Californian earthquakes of the same magnitude range you find that they are anomalously large. I believe all the difference in interpretation only amounts to these very large stress drops.” Typical earthquake stress drops—independent of location or quake magnitude—are about 3 MPa or 30 bars. So, if Ellsworth and Bulut are right, the foreshocks have stress drops, a measure of the fault slip divided by its area, ten times larger than normal.


The search for precursors continues

While there is still debate as to the nature of the foreshocks preceding the Izmit earthquake, this new study is important because of the implications it has on earthquake forecast and mitigation. Migrations of small shocks are commonplace, and the vast majority of these sequences are not followed by large shocks. So if this were only a migrating, or cascading, sequence of small shocks, there would be nothing precursory about this process, nothing that points to the imminent occurrence of a great earthquake. In contrast, had the foreshocks accompanied a more profound acceleration of fault creep, or expansion of fault creep, creep episodes like this might be used to predict a coming large shock. Another possibility suggested by the new study might be that foreshocks have much higher stress drops than typical quakes, but a mountain of studies suggests this is not dependable.

Even though the 1999 Izmit case is now in doubt, there is still the possibility that foreshocks, coupled with slow slip, can lead up to a major event. the 2011 M=9.0 Tohoku earthquake was preceded by a M=7.3 foreshock in the same area two days before the mainshock (Hirose et al, 2011). Therefore, scientists will continue to look at precursory events to see if they can be used in earthquake forecasting and mitigation. Ellsworth told Stanford that they will not be giving up of foreshocks and that “We want to understand if they have predictive value and if not, why not.”


This figure from Kato et al., 2012 shows the Tokoku earthquake as well as the foreshock sequence that lasted 23 days. The blue box in the inset map is the area shown in the larger map.


What this means for us

Despite 50 years of efforts, we think it fair to say that we are nowhere close to predicting earthquakes. And so, we must accept that minutes to months of warning for strong shaking is a mirage. Instead, we need to be ready at any time for the next shock.



William L. Ellsworth and Fatih Bulut, Nucleation of the 1999 Izmit earthquake by a triggered cascade of foreshocks, Nature Geoscience (2018). DOI: 10.1038/s41561-018-0145-1

Michel Bouchon, Hayrullah Karabulut, Mustafa Aktar, Serdar Özalaybey, Jean Schmittbuhl, Marie-Paule Bouin, Extended Nucleation of the 1999 Mw 7.6 Izmit Earthquake, Science 18 Feb 2011: Vol. 331, Issue 6019, pp. 877-880

Michel Bouchon, Virginie Durand, David Marsan, Hayrullah Karabulut and Jean Schmittbuhl, The long precursory phase of most large interplate earthquakes, Nature Geoscience (2013)

Aitaro Kato, Kazushige Obara, Toshihiro Igarashi, Hiroshi Tsuruoka, Shigeki Nakagawa, Naoshi Hirata, Propagation of Slow Slip Leading Up to the 2011 Mw 9.0 Tohoku-Oki Earthquake, SCIENCE VOL 335 10 FEBRUARY 2012

Fuyuki Hirose, Kazuki Miyaoka, Naoki Hayashimoto, Takayuki Yamazaki and Masaki Nakamura, Outline of the 2011 off the Pacific coast of Tohoku Earthquake (Mw 9.0), Seismicity: foreshocks, mainshock, aftershocks, and induced activity, Earth, Planets and Space 201163:1, doi.org/10.5047/eps.2011.05.019