An earthquake registering the top of Japan’s seismic intensity scale struck Kumamoto Prefecture on the island of Kyushu at 16:27 on Tuesday afternoon, and by ten minutes before midnight the country’s Fire and Disaster Management Agency had 354,284 people living under its two highest categories of evacuation order. The agency has formally named the event the 2026 Kumamoto Earthquake. Shindo 7 — the maximum reading, denoting shaking in which it is impossible to move without crawling and poorly built structures can collapse outright — was recorded in exactly two municipalities: Uki City and Hikawa Town.
The evacuation figures are split across two tiers that mean different things. Emergency Safety Assurance, the highest, covering people who can no longer safely leave and are told to move to the sturdiest available space where they are, applied to 46,610 households and 92,743 people in two Kumamoto municipalities. Evacuation Instructions, the tier below, covered 125,325 households and 261,541 people across four municipalities — three in Kumamoto and one across the water in Nagasaki Prefecture. A tsunami advisory, not a warning, went out for the Ariake and Yatsushiro seas two minutes after the shaking and was lifted at 18:10, one hour and 41 minutes later.
The human toll recorded in that same report is a single light injury, in Oita City. It should not be read as the casualty count. The document is marked, in its own header, as a preliminary report whose figures may still change, and it lists two serious incidents that were still being worked when it was published: a factory chimney had collapsed in Yatsushiro City, and in Kashima Town the second floor of a commercial building had given way with, in the agency’s words, many people trapped inside. The Fukuoka prefectural battalion of the Emergency Fire Response Team was working the Kashima site. Twenty-two rescue operations and eight fires had been logged across ten municipalities; the fires in Uki and Kumamoto City were out, one of the two in Yatsushiro was contained rather than extinguished.
The mobilisation kept widening for more than five hours. The FDMA commissioner issued his first call for Emergency Fire Response Teams at 16:27, the minute of the earthquake, drawing on Fukuoka, Okayama, Hiroshima, Oita and Miyazaki. Saga was added at 17:08. Then came three more requests — 20:15, 20:28 and 21:45 — pulling in air units from Miyazaki, Kitakyushu, Nagasaki and Saga and prefectural battalions from Okayama, Hiroshima, Yamaguchi and Kagoshima. An escalation on that schedule is not what a single-injury event looks like.
The two agencies that measure Japanese earthquakes have not published the same number, and the difference is worth explaining rather than reporting as a dispute. The United States Geological Survey puts the event at moment magnitude 6.8 at a depth of 10 kilometres, five kilometres east of Uto, and has named it the 2026 Uto earthquake. The Japan Meteorological Agency puts it at M7.1 and, at 20:30, issued a formal notice revising its hypocentre to 16 kilometres deep. These are different scales — the JMA magnitude and the moment magnitude are computed from different measurements and routinely diverge by a few tenths on the same event — so neither figure is wrong, and a reader who sees both has not caught anyone out.
What the USGS adds is a loss model, and it ran hot. PAGER, the agency’s automated impact system, raised a red alert for economic losses, its highest, with a modelled central estimate of just under $9.95bn, and an orange alert for fatalities. Its exposure tables put 173,148 people in the zone of violent shaking and a further 553,231 one step below that. It estimated peak ground acceleration at 0.88 g and modelled 557 deaths on its empirical method. Japan’s own fire agency, seven hours in, had logged one light injury and two active rescues.
That gap is the story, and part of it is visible inside the model itself. PAGER offers readers a short list of historical analogues to calibrate expectations, and for this earthquake it lists three: a magnitude 5.8 in Yamaguchi that killed nobody, a 6.6 in Fukuoka that killed one, and a 6.8 in Geiyo that killed two. It does not list the 2016 Kumamoto sequence. Querying the USGS’s own catalogue for everything of magnitude 5.5 or greater within 60 kilometres of Tuesday’s epicentre since 1990 returns six events, and all six are from April 2016 — including the magnitude 7.0 mainshock 12.5 kilometres away, and a magnitude 6.0 foreshock that struck 1.7 kilometres from where the ground broke on Tuesday. The nearest and largest precedent available is missing from the comparison set the model shows.
We do not know the rule PAGER uses to pick those analogues, and it may weight modelled loss similarity rather than geography, so this is an observation about the product rather than an accusation of error. But it matters for how the numbers are read. The three earthquakes the model does offer killed nobody, one person and two people respectively, which sits oddly beside an orange fatality alert, and the one it leaves out is the one that would tell a reader most about what a large earthquake does to this specific stretch of Kyushu.
The aftershocks have been heavy. The JMA located 145 events between the mainshock and 03:35 on Wednesday morning, the largest a magnitude 6.1 at 17:08 that itself reached Shindo 5-lower, followed by a magnitude 5.0 ten minutes later. Shaking was measured in 553 municipalities across 26 prefectures, reaching as far east as Fukui, Gifu, Mie, Shiga, Kyoto and Nara. Both Kumamoto and Nagasaki stood up prefectural disaster response headquarters at 16:27, and three FDMA staff were dispatched to the Kumamoto prefectural government at 17:30. As of the latest report, no damage had been reported to any facility the agency classifies as critical.