A glacier gave way. An hour later the valley was gone.

An hour of water, and no rain to explain it

At 08:37 Nepal time on 26 August 2026 (02:52 UTC), instruments recorded a signal in the high Langtang. The United States Geological Survey catalogued it at magnitude 5.2, at 28.271°N 85.515°E, at a depth of zero kilometres — and, critically, classified the event type as “landslide” rather than “earthquake”. That classification is visible in the live panel above and is the primary-source answer to a discrepancy that runs through the press coverage, where the same event is variously reported as M4.4 and M5.2. A zero-depth landslide solution is what a large mass of ice and rock striking a valley floor looks like on a seismometer; it is not a tectonic rupture.

The mass came off Langtang Lirung (7,245 m), whose northern face lies inside Nepal. A geologist at the University of Calgary identified in satellite imagery a section of glacier roughly 2,000 feet wide that detached and fell close to 4,000 feet. Imagery from Planet Labs placed the avalanche around 20 km north-east of Rasuwagadhi.

Why an avalanche became a flood

The debris landed in the Lhende Khola, a tributary along the Nepal–China border, and dammed it. Water and sediment ponded behind that dam — a landslide dam, not a pre-existing glacial lake — and then the dam failed. Preliminary imagery shared by Chinese authorities and circulated by ICIMOD put the blockage about 20 km upstream of the Miteri Bridge.

What came down was therefore not clear water. Binod Parajuli, who heads the Flood Forecasting Division at Nepal's Department of Hydrology and Meteorology, put the point directly: an icefall and landslide are visible in the imagery, an ice-and-debris lake formed and then burst, and — in his assessment — clear water alone would not have produced destruction on this scale. The sediment load is what turned a surge into something that removed settlements rather than flooding them.

How fast it travelled

Locally there had been almost no rain, so nothing about the morning warned anyone downstream. The river rose at Rasuwagadhi with what observers described as extraordinary speed, took Timure and Syabrubesi, then ran into the Trishuli. Gauges downstream registered a rise of nine metres at Galchhi and seven metres at Malekhu within thirty minutes. Alerts followed down the Trishuli into Nuwakot, Dhading, Gorkha and Chitwan.

What is still uncertain

• Whether a glacial lake was also involved. The working reading is an ice-and-rock avalanche and a debris dam. Rijan Bhakta Kayastha, a climate researcher at Kathmandu University, has said it could take several days to determine whether a glacial lake burst as well.
• Whether the seismic event triggered the avalanche or recorded it. Kayastha has noted the timing is suggestive but the causal direction is not established. The zero-depth “landslide” classification argues that the catalogue entry is the avalanche rather than a separate trigger.
• The full sequence. Sarthak Shrestha, a remote-sensing associate at ICIMOD, has put the time needed to establish the complete picture at another one to two weeks. Cloud cover has hampered satellite monitoring throughout.
• Whether the Lhende has fully cleared. Home Minister Sudan Gurung confirmed on 26 August that it had not.

We are stating these as open questions because they are open. Anyone presenting the trigger sequence as settled two days in is ahead of the evidence.

There is a second lake, and it is filling

The most important fact on this page is not in the casualty figures. China's Ministry of Water Resources has reported a large barrier lake formed near the confluence of the Chotchen and Purupe Tsangpo rivers in Tibet. By the morning of 27 August roughly two million cubic metres had accumulated behind it, with Chinese projections indicating a further three million cubic metres could enter within three days.

If that barrier fails the way the first one did, a second surge runs down the same valley — over rescue teams, over the survivors who have come back to look for people, and over a river corridor whose bridges are already gone. Nepali authorities have warned residents and rescue workers along the Bhote Koshi to stay off the banks. Chinese authorities have established continuous monitoring and run breach simulations. Cloud cover is limiting satellite observation of the site.

This is the single reason Humrahi is not sending volunteers, and we would say the same to anyone else considering it: the valley is not safe to enter, and untrained people in it are a liability to the responders.

This valley has now done this twice in fourteen months

On 8 July 2025, a flood came down the same stretch of the Bhotekoshi, swept away the Miteri Bridge linking Nepal and China, and damaged the Rasuwagadhi dry port. At least nine people were killed and nineteen went missing. Nepal's Department of Hydrology and Meteorology, working from Sentinel satellite imagery, attributed that event to a glacial lake outburst in the Lhende river near the border — a different mechanism from the 2026 avalanche-and-debris-dam sequence, in the same few kilometres of river.

The two events together describe the problem. The Bhotekoshi corridor sits directly beneath retreating Himalayan glaciers, and it carries a border crossing, a hydropower cascade and a string of settlements in a narrow gorge with nowhere to go sideways.

The wider glacial-lake picture

Research by ICIMOD with UNDP has catalogued 3,624 glacial lakes across the Nepal, India and Tibet portions of the Koshi, Gandaki and Karnali basins. Of these, 47 are classified as potentially dangerous — and 21 of those sit in Nepal. Those are the lakes. The 2026 event was not one of them: it began as an ice-and-rock avalanche, and the lake that killed people did not exist until the avalanche created it, minutes earlier.

That is the part worth carrying away. Hazard mapping built around known, catalogued lakes does not capture a dam that forms and fails inside an hour. As Kayastha has put it, rising temperatures increase the likelihood of glaciers melting and of ice and rock breaking apart — and an avalanche can manufacture a reservoir upstream of a populated valley with no warning at all. Nepal's early-warning systems on this corridor were built for a different failure mode.

Who is doing what

Nepal. The Nepal Army, the Armed Police Force and Nepal Police were mobilised on the day. The NDRRMA had recorded 113 people rescued from Timure, Dhunche, Syabrubesi, Uttargaya and Bidur by 27 August — 88 Nepali citizens, 11 Indian, 8 Chinese, 4 Korean and 2 American. The Home Ministry has deployed twelve coordinating teams. NDRRMA's spokesperson is Shanti Mahat; the Home Minister is Sudan Gurung.

International. India flew in 50 tonnes of relief on 26 August, and Prime Minister Modi spoke with Prime Minister Balen Shah. The United States committed $500,000 and a disaster response advisor; the United Kingdom pledged £5 million; the Red Cross released $1.4 million; the UN Secretary-General activated support. On the Tibetan side, roughly 800 emergency workers and 150 vehicles were deployed.

Where the gap is. Rescue is being handled by militaries and national agencies, which is correct — it needs helicopters and engineers, not NGOs. The gap opens afterwards: households in a gorge whose bridges, road, market, customs employment and hydropower jobs are gone at once, and who are not going to be reached by an airlift. That is the phase small organisations can actually serve, and it is what this appeal is being raised for.

Donate to the response →

For researchers and journalists

If you are working on this event, take the primary sources rather than this page. Everything below is open, needs no key, and is what the live tracker above queries.

Query these directly

• USGS earthquake catalogue (FDSNWS) — the seismic record of the collapse, including the type: landslide classification. CORS-open, no key.
  earthquake.usgs.gov/fdsnws/event/1/query?format=geojson&starttime=2026-08-20&minlatitude=27.6&maxlatitude=29.4&minlongitude=84.6&maxlongitude=86.4
• NDRRMA BIPAD portal — Nepal's official incident register, with per-incident loss breakdowns (deaths, missing, injured, families, infrastructure). CORS-open, no key. District IDs used here: Rasuwa 23, Nuwakot 25, Dhading 26, Gorkha 44, Chitwan 35, Nawalparasi East 481. Hazard IDs: Flood 11, Landslide 17, GLOF 26, Avalanche 3, Inundation 28.
  bipadportal.gov.np/api/v1/incident/?district=23&incident_on__gt=2026-08-25&expand=loss,hazard
  Caveat, and it matters: as of 27 August this register contains no Bhotekoshi records. It is filled in by local authorities, and Rasuwa cannot file. Do not treat its totals as a toll.
• ReliefWeb — situation reports from responding agencies. The v2 API requires a registered appname, so this page does not query it.
• ICIMOD — the remote-sensing analysis of the avalanche and the glacial-lake inventories. Department of Hydrology and Meteorology (DHM), Nepal — river gauges and the flood-forecasting division's assessments.
• Copernicus EMS and UNOSAT — for post-event satellite damage mapping, if activated. Planet Labs imagery is what the published avalanche identification rests on.

How we handle the figures on this page

• Every figure carries the source that published it and the date it was published. Where sources conflict, all values are shown — see the revision log in the live tracker.
• Superseded numbers are added to the log, never overwritten, so the revision history stays auditable.
• The two live panels query the APIs from your browser on each page load. Nothing is cached or transcribed by us, and each panel stamps the time it was queried.
• We do not publish a single “official” toll. Nepal Police, NDRRMA and news organisations are measuring different things; the spread between them is the honest uncertainty.
• Corrections: email a@myhumrahi.org. If you can show a figure here is wrong, we will change it and log the change.

Reusing this

The compiled figures, the revision log and the source list on this page may be reused freely with attribution to Humrahi Foundation and a link back. The underlying data belongs to the agencies listed above and carries their terms. The two Assam photographs are ours; the page carries no third-party photography, deliberately.