Researchers at Northwestern University have published a study that may help solve one of seismology’s biggest challenges – predicting when the next big earthquake will hit a fault. Seismologists have traditionally believed that powerful earthquakes on faults occur regularly and occur over the same period of time between the previous two. However, Earth doesn’t always adapt, as sometimes earthquakes can happen sooner or later than expected. So far, seismologists have been unable to explain this unpredictability.

They do now. A Northwest research team of seismologists and statisticians has developed a more complete and realistic earthquake probability model than currently available. Instead of using the mean time between past earthquakes to predict the next one, the new model takes into account the specific sequence and timing of previous earthquakes. This helps explain the surprising fact that earthquakes sometimes occur in clusters – groups with relatively short periods in between and separated by longer earthquake-free periods.

“Taking into account the exact history of earthquakes, not just the time average and the time since the last one, will greatly help us predict future earthquakes,” said Seth Stein, William Deering Professor of Earth and Planetary Sciences. Weinberg College of Arts and Sciences. “When trying to determine a team’s chances of winning a game, you don’t want to just look at the last game and the long-term average. A review of recent games can also be helpful. Now we can do the same for earthquakes.”

The research was recently published Bulletins of the Seismological Society of America. The study’s authors are Stein, Northwestern professor Bruce D. Spencer, and most recently Ph.D. alumni James S. Neely and Leah Soldich. Stein is a Faculty Member at the Northwestern Institute for Policy Research (IPR) and Spencer is an IPR Faculty Member.

“Earthquakes act like an unreliable bus,” said Neely, who is now at the University of Chicago. “The bus comes every 30 minutes, but sometimes it’s too late, sometimes too early. Seismologists have suggested that even if the earthquake is late, the next one probably won’t happen sooner. In contrast, in our model, if it’s late, it’s more likely to occur sooner than now. And the later the bus leaves, the faster the next bus comes.”

Traditional model and new model

The traditional model used after the great 1906 earthquake that ravaged San Francisco suggests that slow motions along a fault create tension, all of which are released during a major earthquake. In other words, the fault has only a short-term memory – it only “remembers” the last earthquake and “forgets” previous earthquakes. This assumption is used to predict future earthquakes and then create hazard maps that estimate the shaking level at which earthquake resistant buildings should be designed.

But “big earthquakes don’t go like clockwork,” Neely said. “Sometimes we see several large earthquakes occurring in a relatively short period of time, followed by nothing for a long time. Conventional models cannot handle this type of behavior.”

Instead, the new model suggests that earthquake faults are smarter — they have long-term memory — than seismologists thought. The long-term memory of a fault is sometimes caused by an earthquake not releasing all the stress that has built up on the fault over time, so some of it remains after a major earthquake and may trigger another. This explains earthquakes that sometimes occur in groups.

“Earthquake clusters mean that faults have long-term memory,” said Soldich, who now works at the U.S. Geological Survey. “If it’s been a long time since a major earthquake, even after another earthquake has occurred, the earthquake sometimes doesn’t erase the ‘memory’ of the fault, leaving residual stresses and increasing the probability of it happening again. This is how our new model calculates the probability of an earthquake.”

For example, while major earthquakes occur on the Mojave section of the San Andreas fault on average every 135 years, the last one occurred in 1857, just 45 years after the 1812 earthquake. While this was not expected using the traditional model, the new model shows that residual stress caused the 1857 earthquake earlier than average because the 1812 earthquake occurred after a 304-year gap since the previous 1508 earthquake.

“It’s clear that the specific order and timing of past earthquakes is important,” said Spencer, professor of statistics. “The behavior of many systems is dependent on their history over a long period of time. For example, the risk of an ankle sprain depends not only on the most recent sprain, but also on previous sprains.”