I was sitting in one of the small restaurants off the road near the Teaching Hospital in Kathmandu, arguing with a young man from the city, barely into his twenties, about whether Nepal's problem was that it had built too much or too little, when his phone began to fill up with the first videos from Rasuwa. Brown water, moving far too fast, taking a road with it. He had been making the case, with the fluency of his generation, that the country had thrown concrete at its rivers and was now paying for it. Within an hour, the argument we were having had been overtaken by an event that did not fit either of our positions, and it is that mismatch, between the frame we reach for and the thing that actually happened, that I have not been able to put down since.
What actually came down the Lhende River
The flood that struck Rasuwa on August 26 was not a monsoon flood. Nepal's Department of Hydrology and Meteorology reported no heavy rainfall in the affected area immediately before it, which removes the explanation everyone reaches for first. What the evidence points to instead is a chain: ice collapsed from around 5,000 metres down to 3,000 metres, dammed the Lhende River near the Chinese border, and when that impounded lake burst, a surge of debris, mud and boulders came down through Rasuwagadhi. The US Geological Survey has since reassessed the tremor recorded minutes earlier as a magnitude 5.2 glacial collapse rather than an earthquake, and Jeffrey Kargel of the Planetary Science Institute puts the wave's average speed over the first 22 kilometres at about 193 kilometres an hour. Downstream, monitoring stations recorded the Trishuli rising by up to nine metres in under thirty minutes at Galchchi. By September 1, the toll had crossed 1000, with the number of missing still far higher.
That sequence matters, because it is not a glacial lake outburst flood in the ordinary sense, and calling it one will send policy in the wrong direction. Lake inventories, the instrument South Asian governments have spent two decades building, would not have flagged this slope at all. What failed was rock and ice on a mountainside, and what killed people downstream was a temporary landslide dam and the sediment it mobilised. Sikkim in October 2023 followed the same logic: 14.7 million cubic metres of frozen lateral moraine collapsed into South Lhonak lake, generating a wave of roughly 20 metres that breached the moraine and released about 50 million cubic metres of water, destroying the Teesta III dam at Chungthang. Chamoli in 2021 was rock and ice too. The science has already moved from monitoring lakes to monitoring slopes. Himalayan policy has not.
The same corridor, wrecked three times in a decade
Crisis in Gaza: The Western Hypocrisy and The Unfolding Tragedy
Nothing about this was unforeseeable, and that is the hardest part. On July 8, 2025, a flood originating in the Lhende surged into the Bhote Koshi, killing at least nine and leaving nineteen missing, sweeping away the Friendship Bridge that links Nepal and China, and damaging the Rasuwagadhi dry port along with hydropower infrastructure. That one was traced to a supraglacial lake at about 5,150 metres, some 36 kilometres north of the border. The corridor has now been destroyed three times in ten years, by the 2015 earthquake, by the 2025 outburst and by this, and the road between Syabrubesi and Rasuwagadhi, still under reconstruction from last year's damage, was damaged again.
The response to 2025 was, on its own terms, the sensible one. By late August 2025, Nepal and China had agreed to share cross-border information on glacial lake outburst risk across the Tibetan plateau. Thirteen months later, that arrangement produced nothing usable. Nepali authorities were unaware as the water crossed the border; the Department of Hydrology and Meteorology learned of it roughly sixteen minutes afterwards, routed from the Rasuwa district administration through the disaster authority, and the SMS warnings went out only once the flood was already moving through upper Rasuwa. The automated gauging stations at Bhote Koshi and Syabrubesi went offline after 8:40 that morning. In effect, the sensing network was destroyed by the event it existed to detect, which is a question of redundancy rather than of budget. What did move afterwards moved through the scientific channel, satellite imagery passed to Nepali glaciologists by Chinese counterparts through ICIMOD, and a Chinese notice about a new barrier lake of roughly two million cubic metres sitting north of the border. That is courtesy after the fact, not an operational warning system, and the distinction between the two is the whole argument.
Two entirely different systems
Here is the comparison that should be doing the work in this debate, and is not. Nepal sits between two upstream and downstream relationships, and they are not remotely alike.
To the south, the machinery exists and it functions. Under the arrangement with India, forecasts for the common rivers are built from Nepali DHM data, covering 42 sites in Nepal and 18 hydrological sites in India, according to the Central Water Commission's 2023-24 annual report. During the monsoon these readings are updated hourly and feed 24-hour forecasts prepared by the CWC's divisional flood control rooms at Lucknow and Patna, and the two governments run a Joint Team on Flood Forecasting alongside joint committees on the Koshi and Gandak projects, on inundation and flood management, and on water resources. It is not a perfect system and Bihar has plenty to say about how the barrages are run, but it is a standing, year-on-year, institutionally staffed channel, and it worked this week. The CWC had an advisory out on August 26 warning that the Trishuli wave could travel down into the Gandak, and Bihar and Uttar Pradesh had a day to move.
To the north, none of this exists. Nepal has no equivalent of the Joint Team on Flood Forecasting with China, no hourly telemetry from Tibetan catchments, no standing technical committee that meets whether or not the political weather is good. India's own experience on that frontier shows what the northern model actually offers. Hydro-diplomacy with China remains confined to the 2002 memorandum, renewed in 2008, 2013 and 2018, with a Sutlej companion from 2004, and reporting since 2023 describes both as lapsed. Even at their best they delivered water level, discharge and rainfall from three stations on the Yarlung Tsangpo, twice a day, during the high-flow season only, for an annual payment of about a crore, and were withheld altogether in 2017 during Doklam. Forty-two stations reporting hourly against three reporting twice daily is not a difference of degree.
So the model Nepal needs is not something it must invent. It has one, on its southern border, and the task is to extend it northwards. That said, the southern system also has to be honest about its own limits, because it worked this week only because the Gandak is two hundred kilometres and many hours downstream of Rasuwa. A network designed for slow-onset monsoon inundation in the Terai has nothing to offer a village twenty kilometres from an ice fall. Extension northwards and upgrading for hazard type are two separate jobs, and both are outstanding.
Why the overdevelopment argument does not survive contact with this flood
So, let’s talk about that argument in the restaurant. The real problem started out twenty kilometres inside Tibet, at five thousand metres, where there wasn’t much Nepali development to cause trouble. Even if Rasuwa had been more careful, that slope would have been impossible to control. Nepal isn’t a country that has built too much; its electricity use and road network are both low, its economy relies on money sent home and its pollution is tiny compared to the warming that is making its mountains unstable. It’s a country that takes risks, not one that creates them, and the ideas of reducing consumption and protecting the planet that come from wealthy countries don’t really work well in a place where what matters most is how well you can do things.
The Nepal Electricity Authority reported six major facilities damaged, an estimated 431 megawatts off the grid and a further 470 megawatts of projects under construction affected, the Rasuwagadhi customs office destroyed with more than 300 vehicles swept away, while on the Chinese side road access, communications and power to the Gyirong Port area were cut. The distinction that matters is between sited and unsited development, not between development and its absence.
The seismic signal was present; the USGS detected the collapse before the water reached the area. However, the missing component was an institution capable of translating that signal into an alert. This would necessitate the installation of telemetered gauges outside the flood path, equipped with redundant power and automated change detection on slopes, not just on lakes. Additionally, seismic and infrasound monitoring should be linked to an authority with the authority to sound a siren, along with a communication system that reaches people on the banks of the Trishuli in their native language. Considering the proximity to a border post, a road that has been constructed twice, and nearly nine hundred megawatts of generation, this is a cost-effective solution.
Kathmandu needs to be upfront about a key challenge: the vital sensors are situated upstream of an international border, which means we’ll need to work together with Beijing to sort out the technical details. That’s why the request needs to change. Nepal has been working for years to get water rights. Now, it should be pushing for, with the Indian states that are downstream of the same rivers, what it already has with Delhi but doesn’t have with Beijing: a permanent, staffed, real-time information line that stays up and running, no matter what happens politically.
