What happened in Nepal?
At about 8:37 in the morning, a slab of bedrock with glacier ice on top fell about 1,200 m from the north face of Langtang Lirung into the Lhende Khola, a small river that joins the Bhote Koshi at the border crossing of Rasuwagadhi (ICIMOD, 2026; Carbon Brief, 8 Sep 2026). The force was picked up by earthquake sensors. The US Geological Survey first logged it as a magnitude 4.4 quake, then said it was the collapse itself.
There had been almost no rain. Nepal's Department of Hydrology and Meteorology said rainfall in Rasuwa in the 24 hours before did not pass 7 mm (The Watchers, 26 Aug 2026). ICIMOD, the mountain research centre based in Kathmandu, put it simply: "This year there was no lake at the start, but a mountain face collapsed."
The debris flow tore through Timure and Syabrubesi, destroyed about 40 km of road and damaged more than a dozen hydropower plants. The flood wave was still moving at up to 30 km an hour when it reached Mugling, more than 130 km downstream, around 1 pm (HiRisk consortium, via Carbon Brief).
Nepal's government assessment puts damage and losses at about Rs 408 billion (about US$2.7 billion), and the cost of rebuilding at about Rs 723 billion (about US$4.8 billion) (Post-Disaster Needs Assessment, reported by Nepal News, 15 Sep 2026).
How much warning was there?
Very little. Nepal's Flood Forecasting Division heard about the surge roughly half an hour after the collapse and sent text alerts to six districts within about 40 minutes, reaching more than 600,000 phones (Nepal News, 15 Sep 2026). That helped people further down the valley. It came too late for the upper valley. "For an event of this magnitude, we're talking minutes, not hours or days," Jeff Da Costa of the University of Reading told the Kathmandu Post.
There was a second gap. The collapse started on the Nepali side of the border, then crossed into Tibet before the river turned back into Nepal. ICIMOD scientist Qianggong Zhang said this put the source outside the reach of China's warning system upstream. Nepal and China signed an agreement on sharing hazard information in 2019. ICIMOD called this flood "its first real test".
What happened in Hunza?
The Shishper glacier above Hassanabad village in Hunza began to surge, or slide forward fast, in 2018. It blocked the stream from a neighbouring glacier, and a lake now forms behind the ice each spring and bursts in early summer. The flood reaches Hassanabad about 15 minutes after a burst (Dialogue Earth).
On 7 May 2022, during a long and severe heatwave, the lake burst. It swept away a bridge on the Karakoram Highway, destroyed houses and damaged farmland, orchards and two hydropower projects. No lives were lost. Pakistan's Meteorological Department said warnings went out at least 24 hours before the outburst, and earlier warnings had gone out on 27 April (Dialogue Earth).
The warning worked because the lake can be watched. Gilgit-Baltistan's Disaster Management Authority (GBDMA) checks water flowing in and out every day, a camera watches the glacier, satellite images arrive every 16 days, and local teams visit the lake. Women, young people and volunteers in Hassanabad have practised evacuation drills. "The authorities and local community organisations were very vigilant and kept us informed," farmer Tariq Jamil told Dialogue Earth.
Is Hunza safe now?
Not fully. In August 2025, Dawn reported a glacial lake outburst flood from Shishper that damaged a newly built retaining wall and part of the Karakoram Highway and cut power to Hassanabad, a month after a similar July flood damaged four houses (Dawn, 8 Aug 2025). In July 2025, Dawn also reported that, according to unnamed sources, most warning systems in the region were "either non-functional, have been stolen, or lack the capacity to disseminate timely alerts" (Dawn, 17 Jul 2025). Hunza shows that warnings can save lives. It does not show that the job is finished.
Where sources disagree
| Question | Source A | Source B | Why they differ |
|---|---|---|---|
| Deaths in the Nepal–Tibet flood | About 1,399 dead and about 5,130 missing in Nepal (NDRRMA, via Nepal News, 15 Sep 2026) | 1,446 dead and at least 6,669 missing in Nepal and Tibet combined (AFP and Reuters, via Al Jazeera, 17 Sep 2026) | Two days apart, and one counts Tibet. Counts in Tibet also differ: 3 dead (Chinese authorities, 28 Aug) and 43 dead (Chinese authorities, mid-September) |
| Cost | At least US$2.5 billion in losses (NDRRMA chief to Reuters, 4 Sep 2026) | About US$4.8 billion (Post-Disaster Needs Assessment, Sep 2026) | The US$4.8 billion is the cost to rebuild. Damage and losses in the same assessment are about US$2.7 billion |
| Hydropower hit | 14 projects, 748 MW (Nepal Electricity Authority, early figures) | 13 projects, about 759 MW (needs assessment) | Early counts versus the later assessment |
| Was it a glacial lake outburst? | Early reports said a burst lake or a quake-made dam | ICIMOD and Carbon Brief: not a glacial lake outburst, but a rock and ice collapse that set off a chain of hazards | Early guesses before satellite images were studied |
| Was the summer 2025 Hunza damage a lake outburst? | Yes, from Shishper (Dawn, 8 Aug 2025) | No outburst that season, damage came from cloudbursts (GBDMA officials, via Lok Sujag, 12 Sep 2025) | Different officials and dates. We could not settle this |
| How many glacial lakes in northern Pakistan? | 3,044 in Gilgit-Baltistan and Khyber Pakhtunkhwa, 33 "most dangerous" (UNDP and Ministry of Climate Change, via Dialogue Earth) | "More than 7,000 lakes", about three dozen "highly dangerous" (GBDMA, via Lok Sujag, 2025) | The second may count all lakes, not only glacial ones |
What we conclude: the death toll is above 1,400 and still rising as the missing are confirmed, so we give it as "more than 1,400". The US$4.8 billion headline figure is the rebuilding bill, not the loss. On the 2025 Hunza floods, credible local reports point both ways, so we report both and do not merge the July and August damage figures. On lakes, about three dozen are considered highly dangerous in every count we found, and that is the number that matters most for warnings.
So was it climate change?
Partly, and the honest answer has two halves.
What scientists are confident about. The World Weather Attribution group found that human-caused warming made July and August temperatures around the collapse about 1.5°C hotter, and named thawing permafrost and thinning glaciers as key contributors (World Weather Attribution, 17 Sep 2026, via Carbon Brief). Glacial areas in the Langtang catchment warmed by about 0.31°C per decade between 1960 and 2023 (Silwal and others, Global and Planetary Change, 2026). The glacier on the slope had retreated about 450 m between 1990 and 2020 (Maria Shahgedanova, University of Reading, via Carbon Brief). Frozen ground acts like glue in a mountain. As it thaws, rock and ice come loose.
What they say cannot be claimed. Friederike Otto of Imperial College London, part of the World Weather Attribution team, said a full attribution of this disaster was not possible with current methods. The same analysis said the 2015 Nepal earthquake likely weakened the slope too. One researcher quoted by Carbon Brief put the caution plainly: "This could have happened without it." Climate scientist Zeke Hausfather wrote that a definitive single-event attribution "may never be possible", but that the scientists working on these hazards "point in the same direction".
Our reading: climate change did not cause this collapse on its own. It made a slope like this more likely to fail, and it is making such events more common across the Himalaya.
The wider trend is clear. ICIMOD's 2023 assessment found Hindu Kush Himalaya glaciers lost ice 65 per cent faster in 2011–2020 than in the decade before. It said they could lose a third to a half of their volume by 2100 if warming is held to 1.5–2°C, and up to 80 per cent on current emissions. Around 240 million people live in these mountains, and 1.65 billion more depend on the rivers below (ICIMOD, via Dawn and AFP, 21 Jun 2023).
What would help next time?
- Watch the slopes, not only the lakes. Most Himalayan warning systems wait for rain, a rising river or a filling lake. This flood had none of those at the start.
- Share data across borders in minutes. Many rivers that flood Nepal start in Tibet, and many that flood Pakistan and India cross borders too.
- Keep the equipment working. A sensor that has been stolen or switched off warns no one.
- Keep practising. Hunza's drills are part of why people moved in time in 2022.
A quick check you can copy
When you read about a "glacier flood", ask:
- Was there a lake that burst, or did ice and rock fall?
- How much warning did people get, and from whom?
- Is the death toll final, or are people still missing?
- Is the dollar figure the damage, or the cost to rebuild?
Key takeaways
- A rock and ice collapse on Langtang Lirung on 26 August 2026 sent a flood down Nepal's Bhote Koshi, killing more than 1,400 people.
- It was not a glacial lake outburst, and people in the upper valley had only minutes.
- In Hunza, Pakistan, a watched glacier lake gave at least 24 hours of warning in May 2022, and no one died.
- Climate change made the Nepal collapse more likely through warmer temperatures and thawing ground, but it was not the only cause.
- Warning systems save lives when hazards can be watched, and need to cover slopes and borders too.
Quick answers
What caused the 2026 Bhote Koshi flood in Nepal?
A slab of bedrock and glacier ice broke off Langtang Lirung on 26 August 2026 and fell into the Lhende Khola, setting off a debris flow and flood down the Bhote Koshi. There was almost no rain.
Was the Nepal flood caused by climate change?
Not on its own. Scientists found warming made local summer temperatures about 1.5°C hotter and helped weaken the slope through thawing permafrost and thinning ice, but they say one event cannot be fully attributed to climate change.
Was it a glacial lake outburst flood?
No. ICIMOD describes it as a rock and ice collapse with no lake at the start.
How did Hunza avoid deaths in 2022?
Officials watched the Shishper glacier lake daily and warned residents at least 24 hours before it burst on 7 May 2022, and the community had practised evacuating.
Read the first Quiet Crisis story. See how we check climate claims and our Key facts.
Where is the money for recovery? See our loss and damage fund ledger.
Spotted an error? Tell us at [email protected].


