Built for the Wrong Hazard: The Bhote Koshi Flood and Himalayan Early Warning
On the morning of 26 August, a mass of ice and rock sheared off a slope near the Nepal-China border and became a debris flow that travelled close to a hundred kilometres down the Lhende Khola into the Bhote Koshi and then the Trishuli rivers. Water levels rose by as much as nine metres in half an hour. By 31 August, Nepal’s disaster authority counted 903 dead and more than 4,300 missing.
At Galchhi, the Trishuli River rose by as much as 9 metres in just 30 minutes. Yet the systems meant to detect and communicate the danger failed to keep pace with the rapidly rising water.
This is the aspect of the Bhote Koshi disaster that should most concern anyone thinking about Himalayan security, and it is not, at bottom, a scientific failure. Nepal and China have maintained a flash flood early warning arrangement in this basin for years, and it has prevented casualties before. It failed on 26 August because it had been built for a different hazard. Monitoring was calibrated for monsoon and perennial flooding, not for a cascading cryospheric event in which ice collapse, landslide and debris flow arrive together. Upstream stations were destroyed before they could transmit. Downstream, Nepal’s gauges had been mounted on bridges on the reasonable assumption that bridges are sturdy things. The bridges went, and the gauges with them. The corridor lost its capacity to warn in the same half hour that it lost its road.
Nor was this an unimaginable event. On 8 July 2025 a flood from the same Lhende Khola killed at least eight people, destroyed the Miteri Bridge to China and damaged the Rasuwagadhi dry port. The road was rebuilt within about a month; the wider reconstruction was still unfinished 13 months later, when the water returned with ice in it. A specialist at the Stimson Center has since described the area as a monitoring blind spot. It was a blind spot that had already announced itself once.
Nepali lawmakers have responded by demanding a cross-border information-sharing mechanism with China. India should pay close attention, because the same architecture governs its own upstream exposure, and it is in worse repair. The Expert Level Mechanism on trans-border rivers, agreed in 2006, remains the principal channel through which India obtains hydrological data from China. The Sutlej memorandum lapsed in 2020 and the Brahmaputra memorandum has since expired as well; neither has been renewed. When they functioned, they delivered water level, discharge and rainfall figures twice daily, but only between May and October, and only for the monsoon. During the Doklam standoff in 2017 between India and China, the latter discontinued Brahmaputra data altogether. After multiple rounds of talks in recent months, the two governments have agreed to convene the mechanism in September and discuss renewal.
This meeting is now the most consequential item on the regional disaster agenda, and it should not be treated as a routine technical renewal. Three things need to change. First, the seasonal frame is obsolete. A flood that begins with a glacier does not consult the monsoon calendar, as Chamoli in 2021, the South Lhonak outburst that took out the Teesta-III dam in Sikkim in 2023, and Dharali in Uttarkashi last year, all demonstrated. Data sharing must be year-round and must cover cryospheric hazards, glacial lakes and slope instability, not river discharge alone.
Second, emergency notification must be insulated from the state of bilateral relations. A hazard alert is not a concession in a negotiation, and treating it as one is what makes the arrangement fail precisely when it is needed. A standing Himalayan multi-hazard notification protocol, binding on all parties and carved out of wider disputes on humanitarian grounds, is a modest ask compared with what a single unannounced surge costs.
Third, and least discussed, is the last mile. A warning that reaches a ministry in Kathmandu or New Delhi but not a Tamang-speaking porter at Timure is not a warning; it is a record. Rasuwa is served by community radio stations, among them Radio Langtang and Radio Rasuwa at Dhunche. After the 2015 earthquake, when the road to Dhunche stayed blocked for 10 days, stations like these were the most reliable information channel people in the district had. Yet, last week, they were largely absent from the response, while families searching for the missing worked through handwritten lists photographed and posted on Facebook and a torrent of AI-generated footage. India, which has its own community radio network across the Himalayan districts, has made the same omission repeatedly.
The technical case for transboundary early warning in the Himalaya has been made for two decades and is not seriously contested. What the Bhote Koshi flood exposes is that early warning is not an engineering asset that states own. It is a communication system, and a communication system is only as good as its weakest institutional link – an expired memorandum, a seasonal window, a gauge bolted to a bridge, a broadcaster nobody thought to call. Disasters in this region can cross borders in minutes but our warnings still travel at the speed of diplomacy.
About the Author: Aniruddha Jena is Assistant Professor in the School of Management at the Indian Institute of Technology Mandi. More about his works at aniruddhajena.in.