Slow Earthquakes in Cascadia are Predictable

Evidence mounts that slow-slip seismic events follow a deterministic pattern

If there is one word you are not supposed to use when discussing serious earthquake science, it is "predict." Seismologists cannot predict earthquakes; instead they calculate how likely major earthquakes are to occur along a certain fault over a given period of time.

It is a matter of debate among seismologists whether the process that drives earthquakes—the loading of strain along a fault followed by the sudden, sharp release of energy as two tectonic plates grind against one another—is a stochastic (random) process, for which only an estimate of the probability of occurrence can be made, or whether it is a deterministic, and potentially predictable, process.

Seismologists at Caltech studied a decade's worth of so-called "slow-slip events," which result from episodic fault slip like regular earthquakes but only generate barely perceptible tremors, in the Cascadia region of the Pacific Northwest. Their analysis shows that this particular type of seismic event is deterministic and potentially could be predictable days or even weeks in advance.

A paper about the work was published in the journal Science Advances on July 1.

"Deterministic chaotic systems, despite the name, do have some predictability. This study is a proof of concept to show that friction at the natural scale behaves like a chaotic system, and consequently has some degree of predictability," says Adriano Gualandi, the lead and corresponding author of the paper. Gualandi was a postdoctoral scholar in the lab of Jean-Philippe Avouac, the Earle C. Anthony Professor of Geology and Mechanical and Civil Engineering, while working on this research. Gualandi and Avouac collaborated with Sylvain Michel, who worked on this project as a graduate student at Caltech, and Davide Faranda of Institut Pierre Simon Laplace in France on the study.

Slow-slip events were first noted about two decades ago by geoscientists tracking otherwise imperceptible shifts in the earth using global positioning system (GPS) technology. The events occur when tectonic plates grind incredibly slowly against each other, like an earthquake in slow motion. A slow-slip event that occurs over the course of weeks might release the same amount of energy as a one-minute-long magnitude 7.0 earthquake. However, because these quakes release energy so slowly, the deformation that they cause at the surface is on the scale of millimeters, despite affecting areas that may span thousands of square kilometers.

As such, slow-slip events were only discovered when GPS technology was refined to the point that it could track those very minute shifts. Slow-slip events also do not occur along every fault; so far, they have been spotted in just a handful of locations including the Pacific Northwest, Japan, Mexico, and New Zealand.

Slow-slip events are useful to researchers because they build up and reoccur frequently, making it possible to study how strain loads and releases along a fault. Over a 10-year period, 10 magnitude 7.0 or greater slow-slip earthquakes might occur along a given fault. By contrast, most regular earthquakes of that magnitude only reoccur on the order of hundreds of years. Because of this time lag between regular large earthquakes and the lack of instrumental records from hundreds of years ago, it is impossible to precisely compare past events with recent ones.

^{GPS stations reveal activity beneath Cascadia where the oceanic floor slides beneath North America. The plate interface is locked at shallow depths (the shaded area), but we see recurring slow-slip events (in blue) that unzip the plate interface, generating tremors (the black dots).}

Despite their name, slow-slip events offer seismologists a way to press "fast-forward" on the loading/slipping process that drives earthquakes. In a short time frame of around 10 years, seismologists using state-of-the-art GPS equipment can observe the cycle repeat itself several times.

Slow-slip events represent what is known as a "forced non-linear dynamical system." The motion of the tectonic plates is the force driving the system, while the friction between the plates, which causes pressure to build up and then eventually be released in a slip event, makes the system non-linear; in a non-linear system, the change in output is not proportional to the change in input. Despite the fact that both the motion and the friction can be modeled using fully deterministic differential equations, the starting conditions of the system—how much strain the fault is already under, for example—have a significant impact on long-term outcomes. Not knowing those exact starting conditions is one of the possible reasons that the overall system is unpredictable in the long run. However, an examination of the fault slip history can reveal how often and for how long similar patterns repeated over time. In this way, the team was able to assess the predictability horizon time of slow-slip events.

"This result is very encouraging," Gualandi says. "It shows that we are on the right track and, if we manage to get more precise data, we could attempt some real-time prediction experiments for slow earthquakes."

Gualandi likens the potential prediction of a slow-slip event to the current science of forecasting the weather, which also involves predictions about a complex, chaotic process (and similarly falls off in accuracy after a week or so). "We already know that approximately every 12 to 14 months there will be a new slow earthquake, but we do not know exactly when it will happen. What we have shown is that it seems to be possible to determine when the fault will slip some days before it happens, similar to the way weather can be forecast fairly accurately a couple days in advance."

One key question is whether the findings for slow-slip quakes can translate to the regular earthquakes that shake cities and endanger lives and property. Last year Michel, Avouac, and Gualandi reported evidence that slow-slip earthquakes are a good analogue for their more destructive cousins.

"If the analogy that we're drawing between slow earthquakes and regular earthquakes is correct, then regular earthquakes are predictable," Avouac says. "But even if regular earthquakes are deterministic, the predictability horizon may be very short, possibly on the order of a few seconds, which may be of limited utility. We don't know yet."

The paper is titled "The Predictable Chaos of Slow Earthquakes." This research was funded by the National Science Foundation.