BY KEIICHI OZAWA, CHUNICHI SHIMBUN
This is the second in a series of investigative reports from the Chunichi Shimbun about how figures for the probability of a Nankai Trough earthquake were manipulated for a government expert panel report released in February 2018.
Seismologists in a subcommittee of a government panel on earthquakes had been critical of a special calculation method used only for a Nankai Trough earthquake, and minutes of discussions revealed the skepticism they had from the start.
By using the method, called the “time-predictable model,” the probability of a huge earthquake striking the region within the next 30 years was assessed to be up to 80 percent. But the figure would be lowered to 20 percent if a different method used for all other quakes in Japan was applied. The time-predictable model therefore was considered unreliable by seismologists at a subcommittee on ocean-trench quakes under the government’s Earthquake Research Committee.
“The time-predictable model is used only for the Nankai Trough and it is strange to present the figure in a map showing earthquake predictions across Japan (using other methods),” one member was quoted as saying in the minutes. “The map surely shows the (Nankai Trough) area in red (meaning the probability is high), but it is necessary to precisely review whether this is scientifically accurate.”
Another member said: “The last time the evaluation for the earthquake probability figures was released (by the committee) was in 2001. I did some research and found a paper published in (British science journal) Nature in 2002, stipulating that the Parkfield earthquake in the United States showed the time-predictable model is a failure.”
Data from one location
Many members criticized the model, saying it was impossible to predict when an earthquake would strike simply through ground upheaval and subsidence data from a single location at Murotsu Port in Muroto, Kochi Prefecture.
The evaluation paper written in 2001 condemned the credibility of the time-predictable model as low.
“The previous evaluation paper says the time-predictable model is applicable only to Murotsu Port and not for other data,” one member pointed out.
Then why was the model adopted, despite the fact that its credibility was questioned 19 years ago? The answer may lie in how the time-predictable model was introduced in the first place, decades ago.
The model was formulated in 1980 by Kunihiko Shimazaki, who now serves as professor emeritus at the University of Tokyo. When methods to calculate earthquake probabilities were evaluated in 2001, he was serving as the head of the Earthquake Research Committee’s subcommittee for long-term evaluations, with a big say on the adoption of the time-predictable model.
In the minutes from those evaluations, a person who appears to be Shimazaki said: “At first, we were considering adopting the renewal process model (used for other earthquakes). But there were several reasons that made us think it is more reasonable to adopt the time-predictable model. The main method (to be introduced to calculate the Nankai Trough area) swayed to the time-predictable model.”
In an email interview with the Chunichi Shimbun, Shimazaki admitted that he made the remarks, adding that, as the formulator of the time-predictable model, he “tried to remain neutral.”
He said the members of the subcommittee in 2001 agreed to adopt the time-predictable model in the end following a suggestion by Masataka Ando, one of the members.
Ando, former project professor at Shizuoka University, told Chunichi Shimbun: “I don’t remember strongly recommending the time-predictable model myself.” Then he said, “I think the members were thinking it would be more acceptable if we made (the prediction for) the size (of the earthquake) big and its time frame short.”
Kenshiro Tsumura, who was serving as the head of the Earthquake Research Committee in 2001 and had final authority over the evaluation paper, admitted that he was the one who pushed for adopting the time-predictable model.
“Initially, the evaluation paper was to state ‘there is a high possibility of (a Nankai Trough earthquake) occurring within the 21st century,’” said Tsumura, who left the post in 2006 and was formerly the head of the weather agency’s earthquake and volcano division.
“But, historically, Nankai Trough quakes occurred about every 90 years, at the shortest gap after the previous one, and nearly 60 years had passed since the last quake at the time,” he said. “I thought a better expression was needed to emphasize the urgency and drive people to prepare for disasters.”
Then he noticed that the calculations using the time-predictable model resulted in a strong probability of a quake occurring in 2034.
“The model was already famous at the time and I thought the result would look imminent,” he said.
A theory advocated since before World War II says that after a major earthquake, the time interval until the next one will be longer and the interval will be shorter after a small one. The time-predictable model is in line with this theory and Tsumura was influenced by it as well.
Surprisingly, however, both Ando and Tsumura said they currently don’t think the model is credible in any way.
“The time-predictable model is meaningless. It is (unscientific) to assess the probability of the Nankai Trough earthquake by using only the data observed at Murotsu Port. Many people are saying this,” Ando said. “But the probability was pushed up (intentionally) in 2013. This is extremely misleading and bad science.”
Tsumura said: “I didn’t imagine they were still using the same model. If I had been heading the committee now, maybe I wouldn’t have approved it.”
Then why was the model adopted in 2013?
I contacted a total of 17 sources who have direct knowledge of the meetings, including scholars and government officials, to find out what happened.
Among them was professor Manabu Hashimoto, current director of the Disaster Prevention Research Institute at Kyoto University, who recounted the meetings with considerable candor.
“Our seismological assessment should’ve been based solely on science. That’s what we, as a panel of experts, should’ve pursued,” Hashimoto said. “But in reality, we disregarded science. It was a fiasco.”
Another set of internal documents obtained by Chunichi Shimbun shows that Hashimoto, in an email with his fellow panel members, sounded the alarm about the consequences of releasing only the high-risk version of the estimates.
“It is egregious to hide the fact that the probability would be lowered if calculated in accordance with today’s science. I firmly believe that we should honestly tell the public that the conventional time-predictable model is flawed, and that our thorough discussions on the matter and resulting re-calculation have led to a much lower probability,” he wrote.
“Even if we keep this a secret, someone is eventually bound to find out, and when that happens there will be a huge scandal about an information cover-up. I believe the ensuing loss of public trust will be fatal,” he added.
But in the end, it came down to fierce resistance by scholars and officials who argued that the release of a low probability estimate would hinder efforts toward disaster prevention, Hashimoto said.
What happened was that the decision of a subcommittee on ocean-trench quakes not to apply the time-predictable model to the Nankai Trough earthquake was later overturned by disaster prevention experts.
During a November 2012 meeting of the subcommittee on ocean-trench quakes, experts vehemently opposed the application of the time-predictable model.
Such a chorus of disapproval led to the panel deciding at one point to balance out its estimates — by presenting both the high-risk version and a significantly lower probability of just 8 to 20 percent that was calculated based on a different, more widely used model.
Normally, the ocean-trench quakes subcommittee tasked with predicting earthquake occurrence is responsible for vetting and releasing a 30-year probability of the Nankai Trough earthquake occurrence. But at the time, a separate panel tasked with disaster prevention — the policy committee — was consulted.
Later, while attending the policy committee meeting, seismologists proposed presenting the two different probabilities. They were upfront about the possible downsides of doing so: presenting two vastly different estimates, they admitted, could cause confusion and lower public vigilance regarding a Nankai Trough earthquake.
Their suggestion was swiftly met with a volley of protests from those on the disaster prevention committee, with the minutes showing one of them was particularly vocal.
“We need to urgently prepare for the Nankai Trough earthquake. A high probability of its occurrence is needed to gain understanding” on disaster prevention policies, the person said.
“If the probability is lowered, it will trigger criticism that measures taken to prevent (a Nankai Trough earthquake) should not be prioritized, and that there is no need to pour taxpayers’ money into hammering out measures,” the person said.
The minutes suggested the same person further went on to make their case.
“To get the ball rolling, we need to secure budgets first. We’ve been trying so hard to get this done, but our efforts will go down the drain if you start saying (the probability could be much lower),” they said.
Heated discussions continued between seismologists and disaster-prevention experts during meetings in December 2012 and February 2013.
“A model used to calculate the probability of a Nankai Trough earthquake is different from one used for other ocean-trench earthquakes and active faults. The model, in fact, is increasingly seen as scientifically untrustworthy, with some people saying its application should be discontinued,” said education ministry official Yasuhiro Yoshida, who represented the seismologists, referring to the time-predictable model.
He concluded by saying a probability of about 20 percent in the next 30 years, which was calculated based on a separate model applied to other earthquakes, was “the most valid figure that can be gleaned from today’s science.”
As if anticipating a backlash from disaster-prevention experts, Yoshida then conceded that any retreat from the much-hyped figure of over 70 percent would risk weakening public vigilance toward an earthquake.
As a compromise, he proposed four possible ways of releasing its probability. These were: not to disclose any probability at all, to disclose only a probability of 20 percent that was the most scientifically accurate figure at that moment, to disclose both the 20 percent figure and the high-risk version based on the time-predictable model, and to disclose both scenarios, but with an asterisk indicating one was more official than the other.
Yoshida’s proposal was followed by another panel member saying, “I think it’s the most honest thing to do — to disclose both scenarios.”
It goes without saying Yoshida’s bombshell sent a shock wave throughout the community of disaster-prevention experts.
“We have been led to believe in the time-predictable model. There must be lots of people out there who feel the same way,” one said. “This would cause huge confusion in society,” said another.
Yoshida’s proposal blindsided those on the disaster-prevention side so much that they initially reacted with pure bewilderment, rather than trying to hit back with logical counterargument, the minutes show.
In an apparent bid to ease their shock, one seismologist ventured a further explanation, saying Yoshida’s proposal reflected a harsh lesson learned from the devastating 2011 earthquake and tsunami in the Tohoku region, which widely caught seismologists off-guard.
The proposal is “one example of how we’re trying to be upfront about what we don’t know about. In truth, we don’t really know why and how this time-predictable model works. When I speak at a lecture, I frequently refer to the year 2034, which is presumably when this model tells us the (Nankai Trough) earthquake will occur, but at the same time, I also say just how much margin of error there can be. It’s something we just don’t know yet,” the seismologist said.
“It’s an utter mystery. There is a lot we still don’t know, although we seismologists often get scolded for saying things like ‘we don’t know.’”
The scholar’s confession of weakness, honest as it was, provided ammunition to those on the opposing team, with one of them launching into a counterattack.
“The way we understand it, Nankai Trough is the earthquake we have the largest volume of accumulated data on, and is something we know best about. You must’ve been aware all along of what kind of results the time-predictable model would bring,” the person said.
When grilled, the seismologist who “confessed” earlier, the minutes suggested, suddenly turned apologetic, saying: “I’m sorry I wasn’t clear. I thought I made myself clear when I said on a number of occasions that we have no clue at all.”
Another disaster-prevention expert piled on, saying, “but there’s no way your short speeches could’ve conveyed all these minute nuances of your message.”
Even the moderator took sides: “If you failed to convey your message, then that’s your fault. I’m sorry if I’m being harsh, but that’s the reality.”
That’s when the disaster prevention side ended the discussion.
“Please go ahead with the high-risk version of the probability. Given how much progress we’ve made on disaster prevention efforts, I think that’s the kind of figure more in line with public sentiment, unless we have scientific grounds convincing enough to override it.”
With that, a seismologist caved. “Given how far we’ve come, no change of course can be justified just because we’re seeing some problems, so I would say we should go ahead and adopt” a probability based on the time-predictable model.
Taking this as a cue, the moderator chimed in. “Well, I think that made it pretty clear that many people are actually opposed to the idea of lowering” the probability.
At this point, the science was beaten. The relentless attack by disaster-prevention experts drove a wedge between seismologists who had previously been united against what they felt to be the selective release of an unscientific probability estimate. All the seismologists could do was defend themselves as best they could.
Keiichi Ozawa is a Chunichi Shimbun reporter who covered nuclear power and earthquakes after an earthquake-triggered tsunami caused meltdowns at the Fukushima No. 1 nuclear power plant in March 2011. Ozawa received the science journalist award, presented by the Japanese Association of Science & Technology Journalists in June, for this series published between October and December 2019.