On 25 September 2026, the second session of the three-part “Rasuwa–Bhotekoshi Climate Catastrophe 2026: From Devastation to Resilience” webinar series brought together experts from government, hydrology, disaster risk management, avalanche science, geological monitoring, humanitarian mapping, artificial intelligence and frontier technologies to examine a central question:
What can we learn from the Rasuwa–Bhotekoshi event, and what should Nepal do differently next time?
Following the first webinar, which examined the physical basis of the catastrophe, cascading mountain hazards and climate attribution, the second session shifted the focus from understanding the event to strengthening the systems needed to anticipate, monitor, respond to and assess future disasters.
The discussion explored how hydrological monitoring, remote sensing, ground-based sensors, open mapping, artificial intelligence and community-based information can help Nepal move from early warning to early action, and from immediate disaster response towards comprehensive Loss and Damage assessment, recovery and long-term resilience.
A central message emerged across the presentations:
Technology alone does not create resilience. Technology becomes valuable when reliable information reaches the right people at the right time, is trusted and understood, and leads to timely decisions and action.
Watch the full recording of the webinar:
Keynote and Expert Contributors
- Dr. Dharam Raj Uprety — Chief Executive, National Disaster Risk Reduction and Management Authority (NDRRMA)
- Er. Sanjeeb Baral — Executive Director, Centre of Hydrology and Water Resources Research (CHWRR)
- Thomas Stucki — Head of Avalanche Warning Service, WSL Institute for Snow and Avalanche Research SLF, Switzerland
- Stefan Schneider — Branch Manager, Environment, Geology & Water, CSD Engineers Switzerland; former Head of Brienz Early Warning Service
- Rebecca Firth — Executive Director, Humanitarian OpenStreetMap Team (HOT)
- Jonas von Wartburg — Consultant, Monitoring Solutions, Geoprevent, Switzerland
- Dr. Ashok Dahal — Geo-AI Scientist and Assistant Professor, ITC, University of Twente
From Early Warning to Early Action: Er. Sanjeeb Baral
The technical discussion began with Er. Sanjeeb Baral, Executive Director of the Centre of Hydrology and Water Resources Research, who presented Nepal’s existing hydrological monitoring and early-warning capacity.
Nepal currently operates approximately 250 automatic water-level stations and around 15 high-altitude snow stations, including stations above 3,000 metres. These systems collect information on river levels, discharge, rainfall, snow conditions, weather and high-altitude lakes, supporting forecasting, hazard mapping, risk analysis and warning dissemination. Forecasting capability has also expanded, with general river forecasting moving from approximately three to five days, while flash-flood forecasting has been extended from roughly 24 hours towards 48 hours.
But the Rasuwa–Bhotekoshi catastrophe demonstrated why forecasting capacity alone is not enough. An effective warning system must connect:
Risk Knowledge → Monitoring → Forecasting → Warning → Communication → Preparedness → Early Action → Response → Recovery
The objective, therefore, is not simply to issue a warning, but to reduce loss and damage through timely action.
The event also exposed the limitations of relying on conventional flood-warning systems for complex mountain disasters. The Rasuwa–Bhotekoshi catastrophe involved interacting processes including landslides, flooding, river erosion, slope processes, seismic activity, infrastructure disruption and downstream impacts.
Mr. Sanjeeb therefore emphasized the need to move towards an integrated multi-hazard monitoring system combining hydrological sensors, meteorological and rainfall stations, satellite observations, camera-based monitoring, seismic monitoring, slope-deformation monitoring, land-settlement monitoring, glacier and glacial-lake monitoring, and snow and avalanche observations.
The key shift is from monitoring hazards individually to understanding how hazards interact and cascade. For high mountain areas, this means connecting observations from the upper catchment with downstream hydrological and hydraulic models, including dam-break and GLOF modelling where relevant.
Learning from Switzerland: Thomas Stucki
The webinar then turned to Switzerland, where Thomas Stucki, Head of Avalanche Warning Service at the WSL Institute for Snow and Avalanche Research SLF, presented lessons from Switzerland’s operational avalanche-warning system.
The Swiss experience demonstrates that effective early warning is not simply a technological achievement. It is the result of long-term investment in science, institutional arrangements, local observations, trained personnel and communication systems.
Responsibilities are distributed across federal, regional/cantonal and municipal levels. National institutions provide forecasts and technical expertise, while municipalities retain responsibility for local decisions and implementation. Local avalanche commissions contribute observations from the ground, creating continuous exchange between local and national levels.
The forecasting system draws on multiple information sources, including automatic monitoring stations, webcams, snowpack models, meteorological forecasts, avalanche detection systems; and public observations. Machine learning is increasingly used as an additional assessment tool, while expert judgement remains central.
For Nepal, the lesson is not simply to replicate Swiss technology. It is to build clear institutional responsibilities, local observation networks, trained personnel, communication pathways and decision-making authority around the technologies that are appropriate for Nepal’s hazards and terrain.
From Monitoring to Evacuation: Lessons from Brienz
Stefan Schneider, Branch Manager for Environment, Geology & Water at CSD Engineers Switzerland and former Head of the Brienz Early Warning Service, presented another dimension of early warning: how monitoring information becomes a decision to evacuate.
The Brienz system combined total stations, GNSS, ground-based interferometric radar, geological investigations, data hubs and web-based information platforms. But technology was only one part of the system. A major lesson was the importance of understanding the underlying geological process. Monitoring data becomes much more useful when interpreted within a geological model that allows experts to develop scenarios and understand how a situation could evolve.
Brienz also demonstrated the importance of connecting thresholds to predefined actions. As conditions escalated, the system could move from warnings to evacuation recommendations, access restrictions and transportation closures. This created a clear chain:
Data → Expert Interpretation → Recommendation → Decision → Communication → Action
The technical experts provided recommendations, while municipal leadership retained responsibility for decisions such as evacuation. The important lesson is that these roles and communication pathways need to be established before a crisis occurs.
This institutional clarity is particularly important in rapidly developing mountain hazards, where there may be little time to negotiate responsibilities once a warning is issued.
Monitoring Is More Than a Sensor: Jonas von Wartburg
Jonas von Wartburg of Geoprevent, Switzerland, brought the discussion towards the practical design of remote monitoring systems in complex mountain environments.
His central message was simple:
“No single instrument can capture a complex hazard.”
Different hazards require different technologies, including ground-based interferometric radar, Doppler radar, cameras, crack meters, impact sensors, water-level sensors, weather stations, GNSS and satellite communications.
But selecting the technology is only the beginning.
A reliable monitoring system must be designed around the physical hazard process, the location, the time available to react, the emergency organisation, power availability, communication bandwidth, data volume, accessibility, maintenance requirements, and local technical capacity.
As Jonas emphasized in his webinar contribution, a remote monitoring system is not simply a sensor. It is an architecture in which the sensor, location, power and communication system have to work together. This is particularly relevant to Nepal’s remote mountain terrain, where a technically sophisticated monitoring device can have limited value if it cannot transmit data, maintain power or be serviced when needed.
From Maps to Action: Rebecca Firth
The webinar then moved from physical monitoring to another fundamental question:
What information do responders need immediately after a disaster?
For Rebecca Firth, Executive Director of the Humanitarian OpenStreetMap Team, one of the most basic gaps can be the absence of an up-to-date map. After a disaster, buildings, roads, bridges, rivers and access to essential services can change rapidly. Responders need to understand what existed before, what has changed and where the greatest impacts are. Open and humanitarian mapping can help establish that baseline, particularly in places where existing geographic data are incomplete.
People + AI + People
One of the strongest messages from HOT’s presentation was that AI should not replace people.
Instead, the workflow is:
People → Data → AI → Human Validation → Improved Data
Human mappers create and validate information. That information can support AI models, which can then accelerate mapping over larger areas. The AI outputs are subsequently validated again by people.
This combination brings together:
speed + scale + local knowledge + quality control + transparency + trust.
The importance of local knowledge was particularly relevant to Nepal.
Buildings, agricultural structures, rooftops and other features in Nepal may look very different from those represented in datasets used to train models elsewhere. Local knowledge is therefore not simply an additional input, it can be an essential component of reliable AI-assisted disaster assessment.
This is also reflected in HOT’s broader experience in Nepal and Venezuela, where rapid assessment combined human mapping, AI-assisted analysis and human validation rather than treating AI as the final product.
Beyond Damage: Dr. Ashok Dahal on Loss and Damage
The discussion then moved from what was damaged to a broader question:
What was actually lost?
Dr. Ashok Dahal, Geo-AI Scientist and Assistant Professor at ITC, University of Twente, argued that conventional damage assessment captures only part of a disaster’s overall impact.
Deaths, injuries, destroyed buildings, damaged roads and hydropower facilities are essential indicators, but cascading disasters can affect much wider systems including agriculture, livelihoods, energy supply, trade, businesses, education, health, ecosystems, transportation; and regional economic activity. For example, damage to a hydropower facility is not only a loss of physical infrastructure. It can also affect electricity supply, businesses, industries, households and economic productivity.
Similarly, disruption of a major transport or trade corridor can produce consequences far beyond the road or bridge that was physically damaged. As Ashok noted in a quote shared following the webinar:
“Disaster impacts do not end when the event is over.”
From Early Warning to Loss and Damage: Dr. Dharam Raj Uprety
The keynote by Dr. Dharam Raj Uprety, Chief Executive of NDRRMA, brought together many of the themes raised throughout the webinar.
His central message was that disaster risk management should not treat warning, response, damage assessment and recovery as separate activities. Instead, Nepal needs a connected information and decision system:
Risk Knowledge → Monitoring → Forecasting → Early Warning → Anticipatory Action → Response → Impact & Loss and Damage → Recovery & Reconstruction → Adaptation & Resilience.
This also means strengthening the connection between data and decision-making.
As Dr. Uprety emphasized:
“Evidence must inform recovery, reconstruction, and future investment.”
Loss and Damage assessment, therefore, should not be treated as an exercise that begins only after the disaster and ends with a monetary estimate. The evidence generated during and after a disaster can also inform recovery, reconstruction, preparedness, adaptation and future investment.
What Should Nepal Do Differently Next Time?
Across the presentations and discussion, the webinar identified a set of practical priorities for Nepal.
1. Build an integrated multi-hazard monitoring framework
Hydrological, meteorological, seismic, satellite, GNSS, radar, camera, glacier, snow and slope-monitoring systems should increasingly feed into a common operational framework.
2. Move towards source-to-impact forecasting
The objective should not stop at predicting the hazard itself. The system should increasingly connect:
Source → Hazard → Exposure → Impact → Action.
3. Connect thresholds to predefined actions
Every warning threshold should answer:
Who receives the warning?
Who decides?
Who communicates it?
What action is triggered?
What happens if communication fails?
4. Strengthen last-mile communication
SMS, cell broadcasting, sirens, social media, local radio, community networks and local government structures should complement one another rather than relying on a single communication channel.
5. Build a national disaster data architecture
Real-time sensor data, satellite imagery, hazard maps, exposure information, infrastructure data, population data, damage assessments and historical disaster information should increasingly be connected through a shared data architecture.
6. Develop Nepal-specific AI
AI models should be trained and validated using locally generated data that reflect Nepal’s mountain terrain, building typologies, agricultural landscapes, infrastructure, languages and hazard characteristics.
7. Institutionalise community mapping
Baseline data should exist before a disaster occurs, with communities playing a role in mapping, validation, local observation and preparedness.
8. Expand Loss and Damage assessment
Assessment should capture human, physical, economic, social, environmental, cultural and livelihood impacts, rather than focusing only on physical destruction.
9. Continue monitoring after the disaster
Post-disaster monitoring cannot stop once the immediate emergency ends. Secondary hazards and impacts can persist long after the triggering event.
10. Invest in people as much as technology
Nepal will need sustained investment in data science, hydrology, meteorology, geology, remote sensing, AI, disaster management, risk communication and emergency management alongside institutional capacity, maintenance and operational systems.
Key Takeaways
1. Early warning is only effective when it leads to early action.
A warning is not the endpoint. Its value lies in whether it enables people and institutions to act before impacts occur.
2. Mountain disasters must be understood as cascading, multi-hazard events.
The Rasuwa–Bhotekoshi disaster demonstrated the limitations of looking at hazards in isolation.
3. Nepal needs integrated rather than fragmented monitoring systems.
Hydrology, meteorology, seismic observations, satellites, cameras, radar, GNSS, glacier and slope monitoring need to increasingly work together.
4. Source-to-impact forecasting should become a priority.
Understanding what happens at the hazard source is only the beginning. Systems must increasingly anticipate who and what will be affected and what action is required.
5. Technology must be designed around the hazard and local context.
A sophisticated sensor is not automatically an effective early-warning system. Power, communications, maintenance, institutional responsibility and reaction time matter just as much.
6. Local knowledge and community ownership are essential.
Communities can contribute observations, mapping, validation, preparedness and decision-making. They should not be treated simply as recipients of warnings.
7. AI can accelerate assessment, but human validation remains indispensable.
The most effective model is not AI instead of people, but people + data + AI + human validation.
8. Loss and Damage extends far beyond physical destruction.
Livelihoods, agriculture, energy, trade, health, education, ecosystems and regional economic activity all need to be considered.
9. Post-disaster monitoring must continue.
The hazard may end, but cascading processes, secondary hazards and socioeconomic impacts can continue long after the initial event.
10. Resilience requires sustained investment in people, institutions and technology.
Data and technology need institutional systems, skilled people, financing, maintenance, community ownership and regional cooperation to produce lasting resilience.
From Devastation to Resilience
The second webinar moved the conversation from what happened to what needs to change. The Rasuwa–Bhotekoshi catastrophe demonstrated the complexity of Himalayan risk. It showed that a mountain disaster cannot be understood or managed through a single hazard, a single sensor, a single dataset or a single institution.
The presentations instead pointed towards a more connected approach:
from isolated data to integrated disaster intelligence;
from hazard detection to source-to-impact forecasting;
from warning to early action;
from damage counting to comprehensive Loss and Damage assessment;
and from post-disaster response to long-term resilience.
The webinar also demonstrated that Nepal does not need to start from zero. The country already has hydrological monitoring infrastructure, forecasting systems, geospatial capabilities, disaster-management institutions, technical expertise and an expanding ecosystem of data and technology. The challenge is to connect these capabilities into systems that work together before, during and after a disaster.
Ultimately, the question is not simply whether Nepal has access to satellites, sensors, AI, drones or advanced monitoring technologies. The more important questions are:
Can we detect change early enough?
Can we understand what it means?
Can we communicate it to the right people?
Can institutions act on it?
Can communities trust and use the information?
And can the evidence generated today shape how we recover and reduce risk tomorrow?
That is the transition from data to action.
And it provides the foundation for the final part of the series: moving from evidence and technology towards climate justice, resilient development, institutional change and the financing required to build a safer Himalaya.