Dr. Anjal Prakash, Professor of Public Policy at FLAME University and formerly a Coordinating Lead Author on the IPCC’s Special Report on the Ocean and Cryosphere in a Changing Climate, has spent decades tracing water and risk across the Himalaya. Speaking to Naresh Nunna of Neo Science Hub, he cautioned against drawing one causal chain from a warming mountain to a flooded plain, arguing instead that the vegetation change, the Rasuwa collapse, and the floods in Assam and Bihar are distinct events, sharing only a common driver and a shared, unequal exposure to risk.
Your work spans high-altitude livelihoods and downstream water security. Which single link between them is best evidenced, and which is most often assumed?
The best-evidenced link is hydrological, and it is narrower than people imagine. In the Indus basin, snow and glacier melt supply a large share of river flow in the dry months before the monsoon. In the Ganga and the Brahmaputra, the monsoon does most of the work. By the time the Gandak or the Kosi reaches the Bihar plains, rain and groundwater carry the bulk of the water. So, the line that the glaciers feed a billion people holds well in the Indus and gets much weaker as you move east.
The link we assume most often is that what hill farmers do on their slopes decides whether the plains flood. That was the old theory of Himalayan degradation, which blamed mountain households cutting forests for floods in Bihar and Bangladesh. At basin scale, monsoon rain, natural erosion and the way we have embanked rivers on the plains matter far more than village land use. The idea survives because it is convenient. It puts the blame upstream, on poor people, and away from floodplain decisions taken in Patna and Delhi.
The three parts of this story run from a slow vegetation change to a sudden Nepal disaster to floods in Assam and Bihar. Would you draw one causal chain, or several?
I would draw several, and I would be careful with the arrows. The vegetation change is a slow shift in plant cover and snow at altitude, measured over decades. Rasuwa was something else entirely, a mass of rock and glacial ice broke off high on the Langtang range and fell into the Lhende River on the China border. It was a mountain cascade, and it happened on a morning when it was not raining in the valley. Assam’s floods this year came largely from heavy rain over Nagaland and Arunachal Pradesh feeding the south-bank tributaries. Bihar’s are monsoon floods in the Kosi, Gandak and Ganga catchments. The triggers are different, and the clocks they operate on differ as well.
What connects them is a common driver and a common exposure. Rising temperatures destabilize high-altitude slopes, migrate snowlines, and increase monsoon moisture. A physical connection also exists since the Trishuli joins the Narayani and enters Bihar as the river Gandak. Rasuwa’s water and sediment do flow into Valmikinagar. But tracing a direct route from mountain flora to Patna would be misleading. The complex reality is a mountain range evolving on multiple temporalities simultaneously, with downstream states vulnerable to each of them.
What did Rasuwa communities lose on 26 August that no casualty figure captures?
Small towns like Timure and Rasuwagadhi developed alongside the crossing to Kerung, building livelihoods in customs work, trucking, small hotels, and trade. Over the past fifteen years, many families had moved down to the road because that’s where the livelihoods were. The flood took that road livelihood away. This was the second such blow to the region in little over a year, a similar flood had hit the same river in July 2025. Those who rebuilt their lives and homes after 2025 chose to trust the valley again; that trust will not be bought with a relief package.
During the 2026 flood, when a body is not recovered, families are left to cope with what psychologists term as ‘ambiguous loss.’ Without last rites, there is essentially no death certificate, which prevents inheritance and land transfer, besides compensation. Many of those swept away were migrant workers at hydropower sites and traders from other districts, whose families are not in Rasuwa to grieve and count. And then there’s the land. The terraces and riverside pastures that are overwhelmed by metres of sludge do not easily return. The gompas and cremation grounds by the river are places of meaning for Tamang families. A death toll counts people. It cannot count for a community that loses the ground beneath its feet.
Who bears the cost of hazard in the Himalaya: the hill communities, the hydropower firms, downstream farmers or insurers?
Mostly the people, least able to pass it on. Hill households pay first, in lives, houses, land and the income that comes from roads and trade. They have almost no insurance, and state compensation in Nepal is small and slow to arrive. The hydropower companies look vulnerable, and the Nepal Electricity Authority said 14 projects were damaged in this earthquake. Many of these projects are financed by Nepali commercial banks, a few are publicly owned, and the NEA is the sole purchaser. When the plants go offline, the cost is borne by depositors, the public purse, and consumers who pay for imported electricity. A significant portion of it is socialized.
But downstream farmers face a cost of another kind, one that arrives almost every year on the Kosi and the Gandak. In 2018, Bihar withdrew from the national crop insurance scheme in favour of its own state scheme, and assistance goes significantly more readily to landowners than to sharecroppers, who cultivate much of north Bihar’s flood-prone land without title. Insurers carry the smallest share. Catastrophe cover in this region is thin, so they hold a sliver of a very large risk. The pattern is plain enough, the less able you are to price risk, the more of it you end up absorbing.
Hydropower is more than 95 per cent of Nepal’s electricity. What does concentrated exposure mean for adaptation?
It means Nepal’s power supply, its export earnings and a large share of its bank lending sit in the same valleys as its worst mountain hazards. Most plants are run-of-river projects on steep, narrow rivers fed by glaciated catchments, bunched into a few corridors such as the Trishuli and Bhote Koshi, the Sunkoshi and the Kaligandaki. One event can knock out several plants in a morning, as 26 August showed. A portfolio that looks diversified by the number of projects is not diversified by hazard at all.
Adaptation has to start before a licence is issued. That should make approval contingent on hazard mapping of the upstream glacial lakes, unstable slopes and debris-flow paths, including on the Tibetan side. Existing plants need warning systems at intakes and powerhouses tied to sensors upstream, and regular drills for workers, many of them migrants, who don’t know the valley. The grid also needs other sources: solar in the Terai, storage, and cross-border lines with India so the country can import when a valley goes dark. On the financial front, Nepal Rastra Bank should conduct stress test on lending against the loss of an entire corridor, and the sector should buy catastrophe cover instead of assuming the state will bail it out if things stay quiet. The cheapest adaptation of all is spreading the generation across more valleys.
How should governments weigh a rare, severe upstream event against annual downstream floods when allocating money?
Governments fund what they see year after year. The flood budget in Bihar always goes to repair of embankments and other structures, relief and compensation measures, for those floods happen every year and are politically visible. A collapse, once in decades, upriver is outside the yearly cycle and receives almost zero attention until that day arrives.
I wouldn’t be prepared to trade off with a single expected-loss figure. Expected annual loss will always be in favour of the regular flood, but it understates the rare event, which can kill more people in an hour than a decade of ordinary floods and destroy infrastructure that took years to finance. Loss of life and damage that cannot be undone deserve greater consideration in that calculus. A better approach is to layer the money. Frequent and moderate floods should be dealt with from the routine budget, spending more on preparedness, drainage and quick cash transfers instead of new embankments. Infrequent and serious events should be financed by catastrophe insurance, pre-arranged credit lines with the World Bank or the Asian Development Bank, and a disaster fund that can build up over easy years. Then there should be a small, protected line for keeping an eye on high-altitude risks. Sensors, satellite studies and information sharing agreements cost very little compared to either type of losses, or this is the first thing to get cut in a tight year. I would protect it.
What does a fair adaptation finance mechanism look like across Nepal, India and Bangladesh?
We must consider the stakeholders. Nepal hosts many of the hazard sources, and much of the monitoring burden lies on its shoulders. India and Bangladesh receive the floods, and India purchases an increasing share of Nepal’s power. China oversees the headwaters of a number of these rivers, including the one that failed in August, and deserves to be part of the conversation even if it remains outside a formal fund.
Contributions should flow along benefit and ability, which means that India, being the biggest economy and the largest downstream beneficiary of both monitoring and electricity, pays the biggest share. Nepal and Bangladesh can contribute partly to kind, through data, staff and access to sites. The initial funds must prioritize those things that all benefit from and no one can easily do alone: transboundary gauges, monitoring of glacial lakes and unstable slopes, a common data platform and agreed protocols on who alerts whom. Following this, a parametric risk pool modeled after the Caribbean’s CCRIF can disburse funds within a matter of weeks once a defined trigger is met, so districts are not left hanging on appeals.
Fairness also has much to do with where the money lands. It is Nepal’s municipalities and Bihar’s panchayats that respond on the ground, and a mechanism that links only national treasuries will have no relation with them. The new Fund for responding to Loss and Damage might contribute to capitalising the pool.
Why do warnings fail socially even when they succeed technically?
Warning is a communication between people, and the sensor is only the first step. There is a lot to go wrong after the signal is detected. Often these messages never reach those who are most at risk. Migrant workers at hydropower sites may not have locally registered phones, and no local contacts. Women without a personal phone hear these warnings late, and herders and people on the road entirely miss out on the sirens. Sometimes people hear the message but don’t believe it. After a few false alarms, when the river looks normal, the sky clear, like it had that morning in Rasuwa, people dismiss warnings. Sometimes, it’s the messenger who matters. A person well known in the ward, like the ward chair, will get people moving faster than those who get messages from an unknown phone number.
A warning needs to instruct people too. “High water expected” is not an instruction. People need to be told where to go, how much time is there, and what will happen to their animals and their shop. Even then, they may not be able to respond because they may lose a day’s wages or have to carry an elderly parent uphill. Cross-border hazards add one more gap. The Rasuwa flood originated on the Chinese side, and the question of who warns whom across that border is yet to be answered. Warnings work where the communities have rehearsed them.
What role should Indian cities, Hyderabad included, play in supporting upstream monitoring?
A small role, but a real one, through knowledge and some money. Hyderabad has unusual resources for this. The National Remote Sensing Centre of ISRO already provides maps of glacial lakes in the Himalayas and flood maps for the country. INCOIS runs India’s ocean hazard warnings and knows how to turn a detection into a public alert. Add the city’s universities, ISB among them, and a large data and software workforce, and you have what is needed to analyse satellite data, build forecasting models and train officials from Nepal and the northern states.
Cities can also help pay for it. Large municipal corporations face their own flood risk, as Hyderabad learnt in October 2020, and some now have climate budgets and access to green bonds. A small share of that could support shared monitoring alongside central and state money. My caution is about ownership. Monitoring in Nepal has to be run jointly with Nepali agencies, especially the Department of Hydrology and Meteorology, and with ICIMOD. If Indian institutions work alone on a neighbour’s mountains, it will be read as surveillance, and the arrangement will not last. Data must flow both ways, and training should build Nepali capacity. Hyderabad is most useful as a partner to answer requests that come from Kathmandu and Patna.
If you were writing one page of advice for a Bihar district magistrate, what would be on it?
I would start by instructing the DM to have familiarity with the river upstream. Collect daily data from the Valmikinagar and Birpur barrages and, through the Water Resources Department, from the gauges in Nepal, and assign an officer for this work from June to October with the working phone number of a counterpart across the border. Review the embankment breaches of the last two decades, identify the five most vulnerable spots, and deploy staff there whenever the river goes beyond the danger mark.
Prior to June 1, the district should have an updated list of households which are unable to migrate on their own, the elderly, disabled, pregnant women and families living in diara paddy between bunds. Boats, fuel, dry rations, ORS, chlorine tablets and fodder for cattle should already be in each block. Warnings should be issued in the vernacular through ASHA workers and the mukhiyas, text messages used as an aid but definitely not the sole means.
As the river water arrives, get relief by direct transfer within a week and ensure sharecroppers are on the list. Watch the weeks after the peak, when diarrhoea, snakebite, and drowning take lives, and protect the ghats during festivals. In September 2024, 46 people, mostly children, drowned in swollen rivers. Then help farmers clear sand off their fields quickly, and do not allow a rebuilding in breach paths or new schools and anganwadis on the lowest land. Finally, note who flooded, how deep and for how long. That record is the district’s best case for next year’s money.
