Champi weakens, but Taiwan’s danger persists: one dead, 94 injured

A man was swept away by storm-driven waves at a beach in Liuqiu Township, Pingtung County. His death is the clearest warning against reading Typhoon Champi’s weakening winds as a sign that Taiwan’s danger has passed. ³¹

By Friday, October 23, 2026, most schools and businesses across Taiwan were closed as Champi approached, with hundreds of flights canceled and more than 1,200 people reportedly out of their homes. The storm was disrupting daily life before any verified landfall: families faced closed classrooms and workplaces, while passengers lost both domestic and international connections. ¹

The human toll had also become more specific. Focus Taiwan, citing the Central Emergency Operations Center, reported one death and 94 injuries as of 6:30 a.m. Taiwan time on October 23. That is a time-stamped emergency count, not a final accounting. The available report does not establish the circumstances or severity of those injuries, or the name of the man killed at the beach. ³¹

Taiwan’s Central Weather Administration kept sea and land typhoon warnings active in its 12:15 p.m. bulletin that day. Champi was forecast to move northward east of the island. The established picture is of a dangerous approach and an active emergency response—not a confirmed strike on land. ^{23 26}

The reported evacuation count is substantial but less firmly documented than the casualty tally. AFP/France 24’s October 23 account put it at more than 1,200 people. An exact cutoff time and the locations making up that total have not been independently established, so it should be read as a reported figure rather than a complete map of displacement. ¹

BBC reporting also described evacuations along the northeast coast near a landslide dam at risk. The available account does not name the locality or say how many people were moved from there. It therefore cannot establish what share, if any, of the reported 1,200-plus evacuees came from that area. ³²

An earlier October 22 account had no reported evacuations. That absence does not mean nobody had left home; it means that account supplied no evacuation count. The later reporting makes displacement visible, but still leaves unanswered where all those families are and what conditions must be met before they can return. ^{2 1}

The warning geography is clearer. Taiwan issued a land warning on October 22 covering Yilan, Hualien, Taitung and Pingtung, according to the Taipei Times. The exact first issuance time remains uncertain. An earlier weather-administration analysis had projected possible warning windows; those forecasts should not be mistaken for the times warnings were actually issued. ^{25 26}

Air travel has been severely restricted. FMT’s October 23 report, citing Taiwan’s transport ministry, said all 262 domestic flights and more than 350 international flights had been canceled. Its account does not define the cancellation interval, leaving passengers with a clear measure of disruption but no verified timetable for full service to resume. ¹⁶

Those figures are not a clean running total alongside every earlier report. Global News’s October 22 account cited 125 cancellations, mostly domestic; the later AFP/France 24 account described hundreds across the region without defining the area or period. The figures cannot establish a precise like-for-like increase, and adding the broad regional description to Taiwan’s domestic and international counts would risk counting the same cancellations twice. ^{2 1 16}

Champi had previously reached super-typhoon strength, and the official wind sequence supports the description of a weakening storm. Maximum sustained winds fell from 55 metres per second on October 20 to 48 on October 22, then to 43 in the October 23 warning. These are sustained-wind figures, not the shorter bursts measured as gusts. ^{24 26 27 23}

The gust figures need their own explanation. The official-derived October 22 record gave 58 metres per second, equivalent to 208.8 kilometres an hour. The October 23 warning gave 53 metres per second, or 190.8 kilometres an hour. Both describe powerful bursts even as the storm’s sustained winds declined. ^{27 23}

CNA’s October 22 report had described gusts above 227 kilometres an hour. The time, location and measurement method behind that higher estimate remain unclear. The available evidence cannot determine whether the difference from the official figures reflects weakening, different measurement conventions or an error. Nor does the higher estimate establish that a surface gust above 227 kilometres an hour was measured in Taiwan. ^{24 27 23}

The distinction matters because a falling wind number is not an all-clear for the coast or for rain-exposed communities. Strong winds and heavy rain were forecast for Taiwan on Friday and Saturday, October 23 and 24. Southern Japan was also within the reported forecast area, but the available evidence does not establish the precise timing or realized impacts there. Forecast conditions should not be turned into damage already observed. ^{2 26}

The beach death makes the immediate stakes plain. Champi does not have to retain its highest wind speed for its approach to be deadly. For families already displaced—and those still in areas under warning—the decisive change will not be another lower storm reading, but conditions safe enough to go home.