The blue arrows on the map splay out like uncooked spaghetti tossed into a pot as geophysicist Jonathan Bedford presses play.
Ten years after the March 11, 2011, Great East Japan Earthquake — which upended long-held views about where megaquakes were likely to strike — scientists like Bedford, a researcher with the GFZ German Research Center for Geosciences in Potsdam, are developing new tools to better understand the threat, while others are digging deep into the past in a search for geological clues.
Each arrow on Bedford’s animated analysis represents tiny movements over time — most at a speed of well under 1 millimeter per day — of the Japanese archipelago, as detected by the thousands of highly accurate GPS stations that dot one of the world’s most seismically active countries. Red circles spring up now and then to mark an earthquake.
There’s no visible pattern to the shifting landscape as the months go by and the analysis moves through 2007, 2008 and beyond.
Then, in October 2010, the map springs to life.
From Hokkaido all the way to the southern region of Kyushu, the arrows rapidly dart toward the northeast in a seemingly choreographed movement, then sway back to the west in November before their dance continues in an easterly direction in the early part of 2011. The arrows elongate, indicating accelerated movement, as March 11 — and a megathrust earthquake and resulting tsunami that would kill thousands — draws nearer.
“The motion in the final couple of months before (the quake) was quite unusual,” says Bedford.
In a study he led that was published last year in the journal Nature, Bedford calls this unusual shift of tectonic plates, also observed prior to a massive 2010 quake off Chile, a “wobble,” and concludes that it was a precursor to the 9.0 magnitude Great East Japan Earthquake.
The potential of the discovery is clear: If the science and the work of Bedford’s team holds up, researchers envision a future in which megathrust earthquakes — which cause some of the world’s biggest calamities — are not only less unexpected, but perhaps, to a certain degree, predictable.
On March 11, 2011, marine geologist Chris Goldfinger was listening to a presentation at an earthquake conference in Chiba Prefecture when the room began to shake.
“It went on for 10 or 15 seconds, and you know, for a 6 or 7 earthquake that’s so common in Japan, you kind of figure it would stop … and it didn’t stop,” recalls Goldfinger, an Oregon State University professor and leading researcher of subduction zone earthquakes.
“We’re a minute into it, and it’s getting bigger,” Goldfinger says. “And then it began to dawn on us that we were seeing a paradigm change — because by the time we hit minute two we were already past the biggest expected earthquake for northeast Japan. And by the time we hit minute three … we’re into magnitude 9 range. And everybody knew it in real time.”
That paradigm, based on science laid out in the early days of plate tectonics, tried to predict which subduction zones — places where one plate is slowly sliding below another — could produce magnitude 9 earthquakes and which could not, Goldfinger says. The idea was that very old plates, which cool as they age and grow heavier, begin to sink rather than forming a connection with the upper plate, a condition that could lead to a quake.
“So they went around and they just discounted all the places with really old subducting plates in the world,” Goldfinger says.
In 2004, the magnitude 9.1 earthquake off Sumatra, Indonesia, showed that the theory had holes, as the subducting plate was older. And Japan, Goldfinger says, “has the oldest subducting crust around the globe.”
“Sumatra proved that theory had at least one large problem,” Goldfinger says. “And then Japan just killed it completely.”
Now scientists warn of the potential for large earthquakes in all subduction zones, while actively looking for ways to assess when and where one might occur.
“We’ve gone from thinking we know what’s going on here to knowing nothing in a very short span of time,” Goldfinger says.
Stress along the subduction zone off the Tohoku region was building long before it ruptured.
The Pacific Plate — the cause of much of the earthquake activity in the seismic hot spots, known as the Ring of Fire, around the edges of the world’s largest ocean — was slowly subducting under the Okhotsk Plate, which lies under northeastern Honshu, Hokkaido and Russia’s Kamchatka Peninsula, and stress was increasing along a point of contact between the two plates that was locked together.
Earthquake scientists refer to these points as “asperities.”
“You can only really get a very large earthquake when you’ve had an asperity that’s stuck together for a long time,” Bedford says.
And asperities all have one thing in common.
“They eventually fail,” Bedford says.
Transient activity often increases the closer a mature asperity gets toward failure, but since GPS only registers the location of a station, separating tectonic shifts from other factors that affect the data — from tidal changes to rain and snow — is challenging.
“A lot of what we do as earthquake scientists using GPS is looking into (the data) to try and figure out ‘where are the tectonic signals?’” Bedford says.
To isolate tectonic activity, Bedford developed an algorithm that assumed that such movements, like the wobble detected before the Tohoku quake, are relatively rare compared to the ebb and flow of seasonal loading caused by rain and snow.
“(The algorithm) does quite a good job of allowing us to see what is seasonal oscillation and what is this other, probably tectonic, signal,” Bedford explains.
Thus the unusual activity in the months prior to the March 2011 quake was uncovered and, like investigators examining a plane’s black box after a crash, potential clues to the impending calamity became visible.
The whole of Japan swayed by millimeters during this time period — sufficient movement to show up in the GPS data but not enough to be felt by people. The start of the wobble, according to Bedford, seemed to coincide with a slow-slip originating southwest of Kyushu in October 2010 that traveled along the plate interface toward the northeast. Soon after the slip began, the Philippine Sea Plate, which is also subducting under Japan, appears to have accelerated downward and then sprung back up. This caused the Pacific Plate — which meets the Philippine Sea Plate off Chiba Prefecture — to also slip, hinting at a surprising interaction between the two plates in the months prior to the quake.
It’s not known how widespread the phenomenon is, but similar wobbling was observed in the months prior to an 8.8 magnitude quake off Chile in 2010, hinting that it may be common prior to megathrust temblors. Bedford, who has shared his model with others in the field for further study, is in the process of applying the algorithm to data prior to other large earthquakes and says he has found other signs of wobbles, although more research is needed before a conclusion can be formed.
Goldfinger, who wasn’t involved in Bedford’s study, says the research shows promise.
“(It was) one of the first sort of forward-looking, hopeful things about subduction zone earthquakes that’s happened in a long time,” he says.
Yosuke Aoki, an associate professor with the University of Tokyo’s Earthquake Research Institute who also was not involved in the study, says that it has “no obvious flaws” but stressed that more research is needed.
“Many scientists are stimulated by (Bedford’s) study and many people are working on developing algorithms to detect precursory signals more effectively,” Aoki says.
“What they found was so amazing, so odd, so counterintuitive, so I think we need further investigation.”
“Scientists have to always be accepting that they can be wrong, even if what they show … seems like a great breakthrough.”
Clues further back
Japan’s extensive network of GPS stations, which was built out following the Great Hanshin Earthquake that devastated Kobe in 1995, may have provided clues about the Tohoku quake going back even further than the wobbles picked up by Bedford and his team.
In fact, the GPS network allowed scientists to pick up on subtle tectonic shifts, which occurred at a much slower pace than the wobble months prior to the quake, over a decade before March 2011, according to Aoki.
But at the time, with the widespread use of GPS only in its infancy, scientists didn’t have any evidence as to whether or not these long-term slips, or periods of precursory deformation, were indications that a major earthquake was likely.
“In theory the deformation data could serve as a warning sign in future earthquakes, but we still have some challenges. We don’t really know if such precursory deformation occurs everywhere or every time before such great earthquakes of magnitude 8 or magnitude 9.Even now, while Aoki has no doubt that precursory deformation existed in the years leading up to the Tohoku quake based on the extensive data available, it’s unclear how strong the link is to major earthquakes.
“We need to observe another precursory deformation before such an earthquake, but the problem is we don’t have (major) earthquakes so often.”
Bedford also notes that the observation window available to researchers through GPS is minuscule when considering the history of the planet.
“When we look at transients in this observation window, we don’t really have anything to compare further back,” Bedford says. “And there’s been a heck of a lot of time further back.”
Diving into the past
Goldfinger, while hopeful about what future tools may bring, is more focused on going further back in time. He is among a small but growing number of scientists around the world practicing paleoseismology, which looks to geologic samples from deep below the ocean floor for signs of ancient earthquakes and tsunamis.
The goal is to chart a history of major seismic events to gain a better understanding of recurrence patterns. He and others have used the approach to study the Cascadia subduction zone off the west coast of North America, where the threat of a large quake has come into greater focus in recent years.
“No matter where you are — Japan or Indonesia or North America — the past we know about is probably not long enough,” Goldfinger says.
To go back further, scientists set out on research vessels and take columns of sediment from the seafloor, aiming for samples that can tell a geological history of perhaps up to 20,000 years.
Goldfinger’s counterpart in Japan is Ken Ikehara of the government-affiliated Geological Survey of Japan. This spring, Ikehara will be among the leaders of a mission seeking to develop research methods for deep-sea paleoseismology in the Japan Trench. Expedition 386, delayed a year due to the pandemic, will include a 60-day cruise to collect core samples and a 30-day period of onshore analysis.
The work doesn’t directly link to predicting large quakes, Ikehara stresses. But the data can be one piece of the puzzle when it comes to hazard mitigation.
“We need past earthquake records as long as possible,” Ikehara says.
In Oregon, Goldfinger agrees.
“It’s hands down the best tool for assessing a hazard like that in terms that can be used by governments to set policy,” he says. “It’s muddy, and it’s not sexy, but it works.”
The road ahead
Will wobbles like those detected by Bedford one day be part of a system to issue earthquake warnings weeks or even months ahead of time?
Perhaps, but not quite yet.
“In earthquake science, prediction is sort of the holy grail, and we call it the ‘p’ word, and nobody wants to talk too loudly about it because it attracts a lot of flak, usually,” Goldfinger says. “But that’s becoming less so over time, because people are discovering that there were precursory bits of evidence, looking back retrospectively.”
Bedford, naturally, sees potential in his own study, but he also sees the challenges in terms of getting from scientific discoveries to real-world application.
“When we see transients, are these transients just normal phenomena that are not anything to worry about or are they really special and showing us that a big failure is imminent?” Bedford asks. “This is a major question and perhaps in the next 10, 20 years we will be able to figure this out.”
In a time of both misinformation and too much information, quality journalism is more crucial than ever.
By subscribing, you can help us get the story right.