Budhi Gandaki flood poses no direct threat to Bangladesh, but it holds lessons for early warning

Saswata Sanyal: The Himalayas are a relatively young mountain range with steep slopes and high seismic activity and they have a long history of both types of disasters. The same Bhotekoshi corridor was hit in July 2025. At that time, NDRRMA reported a glacial lake outburst originating in the upper catchment.

The two events in 2026 also had different causes. The Rasuwa–Gyirong flood in August began with a high-altitude rock–ice collapse. In the case of the Budhi Gandaki River flood, the NDRRMA has said provisionally that a landslide blocked the flow of a tributary. The dam later broke, triggering the flood. No link between this event and glaciers or ice has yet been established.

Warming primarily increases risks in high mountain areas through glacier thinning, permafrost thaw and increased meltwater. To distinguish these effects from geological hazards, several types of evidence are needed:

• Using previous satellite data to compile a record of disasters over several decades, while correcting for gaps and biases in data collection;

• Regularly measuring changes in land and glacier elevation in the same areas, and monitoring landslide-prone slopes and glaciers using radar-based InSAR technology;

• Temperature and permafrost records at source elevations;

• Rather than directly attributing an individual disaster to climate change, examining the relationship between warming and the hazardous conditions that developed before the event.

Another reason for the increase in damage is the growth of settlements, roads and hydropower plants along river valleys. As a result, the number of people living in vulnerable areas and the amount of infrastructure exposed to these hazards have also increased.

Saswata Sanyal: On 11 October, residents began alerting one another at about 4:45 am. The Arughat police post issued the first official warning at 5:30 am. FFD and NDRRMA received initial reports around 6:00 am and sent mass SMS at 6:40 am. The nearest automatic water-level sensor is at Arughat, about 65 km downstream of the source, so the FFD relied on reports from residents and officials.

Once the warning system was activated, the dissemination of alerts worked properly. The question is not what failed, but it was to verify the detection, and the hand-off from a credible field report to an authorised alert. The idea is to not give false alerts from the government as well.

Saswata Sanyal: Warning time comes from detecting the hazard high in the catchment and acting fast. A layered system works best:

• Satellite screening: identifies growing lakes, slope creep and new landslide dams weeks to months ahead;

• Seismic detection: flags large collapses within minutes;

• Upstream network of sensors: placed at known glacial lakes and at newly formed dams once detected a network of different kind of sensors like cameras, microphone, geophones, wire sensors to focus on where exactly the breach is;

• Trigger mechanism and sacrificial infrastructure: Rate of rise of water takes place faster than the data is transmitted to the server. Therefore, rate-of-rise can’t be captured. Additionally, if the sensors stop transmitting that should also trigger alerts as it means they have been washed away. On 11th Oct we saw the water level drop as well uncharacteristically, this should also trigger alerts as it mean that there might have been a major landslide upstream. The trigger mechanism needs to be overhauled;

• Trained phone observers with satellite or radio backup: for when mobile towers fail.

Glacial lakes are largely known sites and can be instrumented. Landslide dams appear without notice, so they need rapid detection.

Saswata Sanyal: Not adequately for these kind of impact or hazards. Infrastructure is typically designed for clear-water floods of a given return period. The government RDNA puts recovery needs at about USD 4.74 billion, around 75 per cent of it for infrastructure. Most urgent changes:

• Multi-hazard risk informed land-use zoning to guide reconstruction, so recovery funds do not rebuild exposure;

• Relocating construction camps, worker housing and border posts off valley floors;

• Checking critical assets against debris-flow and outburst scenarios;

• hydropower licence conditions covering upstream multi-hazard risk monitoring and evacuation plans;

• Debris-tolerant river crossings.

Saswata Sanyal: There is no direct physical threat to Bangladesh. The Budhi Gandaki joins the Trishuli, which flows into the Narayani/Gandak and then the Ganges. Even the much larger August flood had attenuated to a routine level by Devghat in southern Nepal.

The relevance for Bangladesh lies in the lessons from Nepal’s gaps in detection, warning and planning:

• Layered warning systems for fast-onset floods, combining satellite screening, seismic detection and upstream sensors, rather than relying only on downstream gauges designed for slow, rainfall-driven rises;

• Alert triggers that respond to a sharp rate of rise, an unusual drop in water level or sensors going offline, not only to fixed danger levels;

• A fast, verified hand-off from community reports to official alerts, where Bangladesh’s Cyclone Preparedness Programme is itself a model for the region;

• Multi-hazard, risk-informed siting and design of riverside settlements, embankments and crossings, so that reconstruction does not rebuild exposure;

• Timely, real-time upstream hydrological data, alongside longer-term research on sediment pulses moving through shared river systems.

These are shared risks across the Hindu Kush Himalaya. Data-sharing arrangements are for governments to agree, with common standards developed regionally.

Saswata Sanyal: Drawing on the lessons above, three steps stand out:

Warning: a national alert-to-action protocol that links each warning level to fixed duties for every tier of government and every actor on a river, with a fast, verified route from community and field reports to pre-authorised alerts. Lead: NDRRMA, with DHM/FFD and local governments.

Detection: satellite screening of glacial lakes, unstable slopes and new landslide dams, starting with the most exposed river corridors, backed by upstream sensors and alert triggers that respond to rate of rise, sudden drops in water level and sensors going offline. Lead: DHM with universities and research institutes; ICIMOD supporting methods, regional data and training.

Planning and cooperation: multi-hazard, risk-informed zoning to guide reconstruction, debris-flow and outburst checks on critical assets, and hydropower licence conditions covering upstream multi-hazard monitoring and evacuation plans, alongside corridor-based data-sharing and warning arrangements with neighbouring countries. Lead: NDRRMA and local governments for zoning; the Department of Electricity Development for hydropower licences; governments through bilateral mechanisms for information-sharing, with ICIMOD and WMO supporting regional standards and platforms.