For a Distant Tsunami That Brings No Shaking, Only Trust in the Warning Drives Evacuation — the Dilemma of False Alarms and Misses

The notice the government sent to municipalities

In June 2026, an earthquake near the Philippines put a tsunami advisory on Japan's Pacific coast, by way of the Meteorological Agency's advisory[1]. Some municipalities then issued wide-area evacuation orders, and the government notified them that a wide-area evacuation order on a mere advisory could, on the contrary, undermine trust in the warning. Urging evacuation too strongly can backfire.

An "evacuation rate" cannot be read as it stands

People do not readily flee even when a tsunami comes. But the figures that circulate as "evacuation rates" refer to different things depending on how they were gathered.

For the 2024 Noto Peninsula earthquake there is a survey of 258 people carried out at evacuation shelters. "I thought of the tsunami and evacuated at once" came to 52.3%[2], and "I did not evacuate at once but did so as a precaution" to 7.0% — together, 59.3% evacuated with the tsunami in mind. The paper describes this as "just under six in ten of the whole."

What matters here is that this 59.3% is not the evacuation rate of Ishikawa's population. The survey was conducted at shelters, so respondents had already evacuated for one reason or another. The 59.3% is closer to "the share of those who evacuated who did so because of the tsunami." The rest had evacuated for other reasons, such as their house collapsing.

The same survey asked whether people expected a tsunami to reach them: 41.1%[2] said "a tsunami will come and reach where I am," 41.5% "a tsunami will come but not reach me," and 15.9% "no tsunami will occur." Only four in ten thought the danger would reach them (the structural limits of that recovery are treated in a separate piece（このサイトの記事）). In the 2011 Great East Japan Earthquake, of the 870 people interviewed, 496 fled as soon as the shaking stopped[3] — short of six in ten. And for a distant tsunami with no felt shaking, it is lower still: on the Pacific coast in the 2006 Kuril Islands earthquake it was 19.9%[4].

If it shakes, people flee — not necessarily

Why the wide variation? Whether the source is near or far matters a great deal. With a near-field tsunami and strong shaking, the shaking itself can prompt evacuation: in the 2003 Tokachi-oki earthquake the rate reached 92.9% in Toyokoro[4]. Yet in the same earthquake it was 46.8% in Akkeshi[4], below half. Hiroo 86.2%, Niikappu 71.3%, Hamanaka 80.4%, Erimo 52.3% — nearly a twofold spread between municipalities. The same shaking and the same forecast wave height do not produce the same share fleeing.

And "if it shakes, people flee" is contradicted by the very paper cited. In the 2003 Miyagi-oki earthquake, shaking reached lower 5 on the Japanese intensity scale, yet no tsunami warning was issued, and the evacuation rate among Kesennuma residents was 1.7%[4]. Katada and Murasawa conclude that "large shaking alone tends not to bring residents to evacuate; for residents to evacuate, a tsunami warning must be issued or an evacuation advisory given." Both the shaking and the warning are needed.

What people gave as their prompt bears this out. Among those who evacuated in the 2003 Tokachi-oki earthquake, 63.8%[4] said "the way it shook made me think a tsunami was coming," but 54.2% cited "the city or town called for evacuation" and 51.1% "I heard the tsunami warning" — more than half pointed to information.

But a distant tsunami arrives with only the wave, no strong shaking. For a far-off earthquake off the Philippines or Tonga, the natural trigger is absent. What remains is trust in the warning.

The evacuation rate for a distant tsunami is about 20%

The November 2006 Kuril Islands earthquake is a case of a distant tsunami. Shaking registered 1 to 2 on the Japanese scale, and the tsunami warning came 14 minutes after the earthquake[4]. The 2003 Tokachi-oki near-field event (up to lower 6 in intensity, warning after six minutes) differs in both shaking and speed of information.

The evacuation rate among Pacific-coast residents this time came to about 20%[4] — far below the 46.8% to 92.9% of Tokachi-oki.

There was an exception. On the Sea of Okhotsk coast, where residents had never experienced a tsunami warning, the rate was 63.5%[4], higher than on the Pacific coast. Katada and Murasawa read this as heightened alarm at a warning issued suddenly, and for the first time, amid slight shaking. The region less used to warnings was the one that followed them.

False alarms wear trust down

Here the problem of "false-alarm fatigue" enters. When evacuation is urged but no tsunami comes, or only a tiny one, and this repeats, people stop believing the warning — the tale of the boy who cried wolf. Research by Toshitaka Katada and colleagues shows that after the 2006 Kuril Islands earthquake, a single false alarm markedly lowered the evacuation rate. Trust is shaved away by false alarms.

Nor is issuing regardless of false alarms the answer

So should warnings be issued freely, unafraid of false alarms? It is not that simple. Both misses and false alarms are unavoidable, from the technical limits of disaster information. The government has set a policy of "issue early, unafraid of a false alarm," yet there is experimental evidence that the value of this strategy of issuing frequently, unafraid of false alarms, is granted in the short term but hard to grant beyond it[5]. Oikawa and Katada note prior work that finds the boy-who-cried-wolf effect empirically weak, and there is no academic consensus. Cutting false alarms, or issuing freely, is not on its own decisive.

The same "evacuation rate" counts different things

Every figure set out above goes by the name "evacuation rate." But Noto's 59.3% is a breakdown of reasons among people surveyed at shelters; Tokachi-oki's 92.9% is a resident survey at municipal level; Kesennuma's 1.7% is the figure where no warning was issued. The denominators, the questions, and the conditions all differ. Before ranking them, one has to establish what each of them counted. What gets lost when a figure is compressed into one number is the same point set out in an introduction to statistical literacy（このサイトの記事）.

The trade-off between accuracy and trust

What lies here is a trade-off between accuracy and trust. Issue too many warnings, and more false alarms make people stop believing them. But keep issuing for fear of a miss, and the value is limited. Both misses and false alarms are a technical dilemma inherent to disaster information; either way there is a cost. And for a distant tsunami, this trust is the sole trigger for evacuation. With no natural signal from shaking, the design of information decides, directly, who lives and who dies.

So the question is not only "how to cut false alarms." Within the dilemma, it is "how to keep trust." After the 2011 disaster, the Meteorological Agency rebuilt its tsunami warnings in 2013, cutting the categories from eight to five[6] and, for a giant earthquake, conveying not a number for the height but a word like "massive." A turn from competing on numerical accuracy toward a design that gets people to flee first. The government's notice to curb excessive wide-area evacuation orders on advisories can be read as part of the same trial and error in designing trust.

Distant tsunamis of the Philippine or Tongan type will draw more attention. When only the wave comes, with no felt shaking, what can move people is trust in the warning alone. That trust wavers with a false alarm, and it cannot be held by issuing warnings freely for fear of a miss either. Beyond the question of predicting the height to the centimeter lies the design of how a warning is believed. It decides the number of people who flee.

Further Reading

• 人が死なない防災 (Disaster Prevention Where No One Dies)（外部サイト、新しいタブで開きます） (Toshitaka Katada, Shueisha Shinsho, March 2012) — a tsunami-evacuation researcher rethinks disaster prevention not from warnings and information alone but from the side of how people actually flee. (In Japanese.)

References

Tsunami Evacuation in the 2024 Noto Peninsula Earthquake: A Survey of Residents — Maho Ishibashi, Shinya Yasumoto, Sayaka Irie, Yoshikazu Kagi, Naoya Sekiya. Journal of Japan Society for Natural Disaster Science 43-3

Present Conditions and Issues of Correspondence of Administrations and Inhabitants for Distant Tsunami — Toshitaka Katada, Naoki Murasawa. Journal of Disaster Information Studies, No.7

Effects of Repetitive False Evacuation Advisory on Residents' Behavior — Yasushi Oikawa, Toshitaka Katada. Journal of Disaster Information Studies, No.14

Committee on Earthquake and Tsunami Measures Learning from the 2011 Tohoku Earthquake (7th Meeting), Document 2 — Cabinet Office, Central Disaster Management Council. Cabinet Office