Five dead, six fatal landslides — most headlines stopped at four
Typhoon Dujuan made landfall near Tokyo on September 22, triggering record rainfall across the Kanto region and forcing more than one million people under evacuation advisories. The NYT reported five deaths; the BBC, Guardian, and Reuters settled on four. Eos, a geoscience outlet that staffs the landslide beat, documented at least six separate fatal landslides — a detail absent from every major wire report. The story the wires told is a weather event. The story underneath is a landslide disaster that the generalist outlets weren't equipped to count.
Typhoon Dujuan hits Japan, causing flight cancellations and evacuations
6 sources · hover a dot to see coverage
What happened
Typhoon Dujuan struck Japan on Tuesday, September 22, 2026, brushing past Tokyo and dumping record rainfall across the Kanto region. More than one million people were placed under evacuation advisories, with over 74,000 remaining under formal evacuation orders in the Tokyo area as of Tuesday morning. Authorities warned of landslides and flooding. Hundreds of flights were cancelled, and power outages were reported across affected areas. The death toll varied by outlet: Reuters and NHK cited four killed; the NYT reported five. At least 22 people were injured. The storm pulled away from Japan after the initial landfall. Six outlets covered the event. The death count and the landslide mechanism behind it separated the coverage.
Eos counted six fatal landslides; no wire outlet matched that reporting
Eos, which publishes the American Geophysical Union's landslide blog, reported at least six separate fatal landslides triggered by Dujuan in the Kanto region over roughly 24 hours. That number appears in none of the wire or broadcast coverage. Reuters, BBC, and France 24 each mentioned landslides as a hazard or cause of death but did not enumerate them. The Eos piece treated the event as a landslide disaster that happened to arrive by typhoon. Every other outlet treated it as a typhoon that happened to cause some landslides. Those are different stories with different implications for where the missing and the dead are likely to be found.
The NYT's five-death figure went uncorroborated by any other outlet
The New York Times reported five deaths; Reuters, BBC, the Guardian, and France 24 each cited four, attributing the figure to NHK. The NYT did not explain the discrepancy or name a source for the fifth fatality. That one-person gap matters when landslides are still active and rescuers are still searching collapsed homes, as the Guardian noted. A fifth death that other outlets missed either reflects faster reporting or an uncorrected error. Neither the NYT nor any other outlet addressed the discrepancy directly.
What one side told you that the other didn't
Six fatal landslides documented. Zero wire outlets reported that number.
Eos identified at least six distinct fatal landslides in the Kanto region within 24 hours of Dujuan's passage. Reuters, BBC, France 24, and the Guardian each mentioned landslides in passing, as one item in a list of hazards alongside flooding and power outages. None counted them. The consequence is concrete: readers of the wire coverage have no basis for understanding the geographic spread of the disaster or the scale of the search still underway in collapsed structures.
The death toll varied by one, and nobody explained why.
Four outlets reported four dead, citing NHK. The NYT reported five dead without naming a source or flagging the discrepancy. In a fast-moving disaster with active landslide searches, a one-person gap in the official count is a signal worth explaining, not a rounding difference. None of the six outlets cross-referenced the competing figures. The reader is left to pick a number and trust it.
One million under advisories; 74,000 under orders. Most outlets used only one figure.
France 24 led with the broader figure: more than one million people under evacuation advisories. The NYT used the narrower, more urgent number: 74,000 remaining under formal evacuation orders in the Tokyo region. Both figures are real and compatible, but they describe different legal and risk thresholds. Advisories are recommendations; orders carry legal weight. Outlets that used only one figure gave readers no way to gauge the actual compulsory scope of the evacuation.
What to watch
If the landslide death toll rises in NHK or Japanese government updates through Thursday, September 25, watch whether the wire outlets revise their four-death figures upward and whether any of them retroactively cite the six-landslide count that Eos reported first. If the gap closes without attribution, that's worth noting.
This kind of gap — where one side reports a fact and the other doesn’t mention it — shows up in about 40% of major stories.