Alka Tripathy-Lang, Ph.D. (@DrAlkaTrip)
A new comprehensive study of almost 100,000 earthquakes in Japan suggests that large or small, quakes look alike in the first 0.2 seconds.
Citation: Tripathy-Lang, Alka (2019), Earthquake size is not foretold in the first second of rupture, study finds, shortening warning times, Temblor, http://doi.org/10.32858/temblor.048
One of the many unanswered questions in earthquake science is how earthquakes grow, and whether predicting the final magnitude of an earthquake is possible when it is still small and developing. Earthquake Early Warning systems were originally based on the hotly debated premise that the rate of growth in the initial 4 seconds was enough to estimate an earthquake’s ultimate magnitude [Allen and Melgar, 2019]. Now, the systems initially use a half-second to 4 seconds of data and are constantly updated as a quake progresses. A key to understanding how an earthquake grows, and thus how big it will eventually be, is an even more basic question: How does an earthquake begin?
In a recent Nature paper, seismologist Satoshi Ide of the University of Tokyo explores what earthquakes look like when they begin to rupture. He found that along approximately 1,100 kilometers (660 miles) of the Japan Trench, some earthquakes of different final magnitudes start in the same place and can look almost identical during the very earliest part of rupture. In fact, he says, “what I show in this paper is the limitation of [Earthquake Early Warning systems].”
A hundred thousand earthquakes
Ide began by combing through 15 years’ worth of data and examined records of more than 100,000 events along the Japan Trench. His goal was to pair large earthquakes (greater than M4.5) with small ones (between M2.0 and 4.0).
Each paired large and small earthquake must have a magnitude difference of greater than 0.5. After winnowing the quakes to those close together and well recorded, Ide was left with 1,654 large and 95,000 small earthquakes for comparison.
Since the duration — the time it takes for a fault to break — of an M4 earthquake is less than 0.2 seconds, Ide looked at only the first 0.2 seconds of data for every quake in this analysis. Ide found 390 pairs of large and small earthquakes with nearly indistinguishable seismograms, which means that even though one grew large and one was stunted, they started in exactly the same way.
The simplicity of comparing seismograms from the same seismic stations eliminates any artifacts that might be introduced during processing, says Men-Andrin Meier, a seismologist and Earthquake Early Warning expert at Caltech’s Seismological Laboratory, who was not an author of the study. “That’s as simple as it gets.”
Pairing earthquakes is unique, says Rachel Abercrombie, a seismologist at Boston University and author of a Nature News and Views article about Ide’s work. “By considering co-located earthquakes, [Ide] eliminates bias from waves traveling their own course through the Earth.”
Cascading quakes or precursory creep?
To explain his findings, Ide suggests a model of differently sized nested patches along a fault system. When a small patch ruptures, it may trigger larger patches in a cascade effect that could result in a large earthquake. How these patches are arranged and stressed influences the cascade, so the final earthquake magnitude cannot be foreseen. Abercrombie likens this model to dominoes: Once the first one goes, the pattern depends on the rest of the dominoes.
Competing with the cascade model is the “slow-slip-start” or “precursory creep” model, Abercrombie says. According to the precursory creep model, with support from laboratory-scale quakes, rocks slip very slowly on either side of a fault. This movement is imperceptible to seismometers, and only gradually speeds up to become an earthquake. In this model, the area of the precursory creep and its nonseismic acceleration would give away its ultimate size — if that creep could be detected.
Both models speak to the question of earthquake determinism: When do we know how large an earthquake will be? Seismologists Diego Melgar of the University of Oregon and Gavin Hayes of the U.S. Geological Survey suggest that the answer is after about 15 seconds for very large earthquakes, based on work they published in Science Advances earlier this year. Melgar says that Ide’s results are consistent with their findings: that in the earliest seconds of an earthquake, different magnitudes are indistinguishable. “I don’t think we know yet one way or another which model holds up to scrutiny,” Hayes says. But Melgar and Hayes did not look at small shocks, did not use shocks in the same locations recorded by the seismometers, and had a much smaller sample size, and so their study did not have the resolving power of Ide’s, says Ross Stein, a geophysicist, and Temblor CEO.
Critical trade-offs in Earthquake Early Warning systems
Because Earthquake Early Warning systems warn and protect people soon after an earthquake begins, these systems rely on an accurate prediction of the final magnitude as soon as possible. Fortunately, the most damaging (“surface”) waves come several seconds after the initial (“P”) waves, so many early warning systems exploit this lag time to save lives. Most early warning systems use the first few seconds of data to send the alert. That time lag is critical, Stein says: The earlier the alert, the more uncertain the shaking will be if the final magnitude is unknown. The later the alert is issued, the more accurate the forecast, he says, but the “blind spot” where no warning is possible will also be much larger.
The last word
The fundamental message of the new study is that accurate Earthquake Early Warning will always be a matter of seconds, not minutes, Stein says. That’s because if it takes 15 seconds before the magnitude can be estimated, the damaging surface waves will have already raced outward 40 kilometers (25 miles) from the epicenter, and everyone within that circle will get a precious little useful warning, he says. It should be possible to issue a warning much earlier than 15 seconds, but the uncertainty on the shaking intensity will be very large, he notes. So, you can have an early but uncertain alert, or a late but accurate one, he says.
Ide’s results are nevertheless “a key step toward a better understanding of earthquake initiation,” Abercrombie says. “The more we do know about the beginning, the higher the chances are that we can do a better job with early warning.”
Abercrombie, R., 2019, Small and large earthquakes can have similar starts (News and Views), Nature, 573, 42-43. doi: 10.1038/d41586-019-02613-5
Allen, R. M. and Melgar, D., 2019, Earthquake Early Warning: Advances, scientific challenges and societal needs, Annual Review of Earth and Planetary Sciences, 47, 361-388. doi: 10.1146/annurev-earth-053018-060457
Ide, S., 2019, Frequent observations of identical onsets of large and small earthquakes: Nature, 573, 112-116. doi: 10.1038/s41586-019-1508-5
Melgar, D., and Hayes, G., 2019, Characterizing large earthquakes before rupture is complete: Science Advances, 5. doi: 10.1126/sciadv.aav2032
Tripathy-Lang A. (2019), Can the size of a large earthquake be foretold just 10 seconds after it starts?, Temblor, http://doi.org/10.32858/temblor.029