India can see the danger in the Himalaya. Can it manage the risk?

On August 26, when a rapid slope failure involving a glacier occurred high on Langtang Lirung near the Nepal-China border, the disaster that followed travelled astonishingly far. The U.S. Geological Survey (USGS) says in a report on its website that the debris flow and flood were likely triggered by rapid slope failure involving a glacier. The slope failure generated seismic energy equivalent to a magnitude 5.2 earthquake, and the flow travelled nearly 100 kilometres.It is important to separate that event from the subject of this story. The Nepal disaster has not been established as a glacial lake outburst flood, or GLOF. The latest USGS assessment says it is still unclear whether the initial failure was a landslide incorporating part of a glacier or a glacial collapse. Yet the catastrophe has drawn attention back to a related transformation across the Himalaya: as glaciers retreat, lakes are appearing where ice once existed, while some existing lakes are becoming larger.India can now watch many of these changes from space. The harder question is what to do with what the satellites see.Are the lakes really growing?A glacial lake is, in simple terms, a body of water associated with a glacier or the landscape created by one. When a glacier retreats, hollows exposed by disappearing ice can fill with meltwater. Some lakes are held back by bedrock. Others are contained by moraines, piles of rock, gravel, and sediment left behind by glaciers. A moraine can act like a dam, except that it was assembled by nature rather than engineered to hold a reservoir.India’s most comprehensive official inventory was prepared by the National Remote Sensing Centre (NRSC) under ISRO. Using high-resolution satellite imagery principally from 2016-17, it mapped 28,043 glacial lakes of at least 0.25 hectares across the Himalayan river basins. Of these, 7,570 were within India and 20,473 in transboundary portions of those basins.Also Read | Nepal floods leave a trail of death and devastationThe distribution inside India is highly uneven. Ladakh accounts for 3,219 mapped lakes; Arunachal Pradesh, 2,188; Sikkim, 733; Jammu and Kashmir, 546; Himachal Pradesh, 537; and Uttarakhand, 347. The 7,570 figure should not be mistaken for a live 2026 census. Lakes can form, merge, shrink, and expand. The atlas is a baseline against which change can be measured.

Army officers carry relief material during a rescue operation in Trishuli on September 1. | Photo Credit: PTI And change is plainly visible. ISRO examined satellite archives from 1984 to 2023 for 2,431 lakes larger than 10 hectares that were identified in 2016-17. It found that 676 had expanded significantly since 1984. Of those expanding lakes, 130 are within India: 65 in the Indus basin, 58 in the Brahmaputra, and 7 in the Ganga basin.One of the clearest examples is Ghepang Ghat in Himachal Pradesh. Its surface area increased from 36.49 hectares in 1989 to 101.30 hectares in 2022, a rise of about 178 per cent. That is not proof that the lake is about to burst. It is evidence of how quickly the high-mountain water landscape can change.The Central Water Commission’s (CWC) expanded monitoring programme adds another layer. From 2025, it has monitored monthly during the monsoon season 2,843 glacial lakes and other water bodies larger than 10 hectares across the Himalayan river basins. Of these, 2,485 are glacial lakes and 358 are other water bodies. There are 681 monitored glacial lakes within India. In June 2025, 432 of those 681 showed an increase in water spread relative to the reference used by the CWC, leading the commission to say they required vigorous monitoring for disaster purposes. Again, expansion is a screening signal, not a prediction of failure.Is a growing lake dangerous?This is where counting lakes becomes inadequate. Two lakes of exactly the same size can present completely different risks. One may sit behind solid rock in an uninhabited valley. Another may be held by loose glacial debris beneath an unstable mountainside, with a town, bridge and hydropower plant directly downstream.The physical question is therefore not simply: how much water is there? It is also: what holds the water in; what could fall into it; how stable are the slopes; how rapidly is the lake changing; where would the water travel if the barrier failed; and what would it encounter on the way down?The CWC’s risk-indexing framework makes this distinction explicit. It uses 12 criteria: 4 for the lake itself, 3 for the upstream setting, 3 for downstream exposure, plus historical outbursts and seismicity. The framework gives 30 of 100 marks to downstream vulnerability, including the distance to habitation, dams, and bridges. Lake size, change in lake size, and proximity to the nearest dam each receives a maximum weight of 15 marks.That is a significant conceptual shift. A glacial lake becomes a disaster risk not merely because it exists or grows, but because hazard meets exposure.

Rescue personnel work at a hydropower tunnel site, in the aftermath of deadly flash floods and a mudslide, in Rasuwa district of Nepal. | Photo Credit: Naveen Srivastava via X via REUTERS The natural dam matters particularly. Water can gradually erode a moraine. Ice buried within it can melt. But failure can also be triggered suddenly when an avalanche, landslide, or large mass of glacier ice falls into a lake. The resulting wave can overtop the barrier. Once water begins cutting through loose moraine material, the opening can widen quickly and release much more water.The surrounding mountains are therefore part of the lake’s risk. So is permafrost, ground that remains frozen for at least two consecutive years. Ice within rock and sediment can help bind high slopes together; warming and thawing can alter that stability. The blue patch visible in a satellite image is only one piece of the hazard.South Lhonak: How a disaster can be built in stagesThe catastrophic outburst of South Lhonak Lake in Sikkim on the night of October 3, 2023, provides the clearest Indian example of why a lake has to be understood as part of a larger mountain system.A major study published in Science led by Ashim Sattar of IIT Bhubaneswar and involving 34 researchers from 9 countries and published in 2025, combined satellite imagery, seismic observations, field evidence, and computer modelling. It reconstructed a sequence more complicated than a natural dam simply giving way.About 14.7 million cubic metres of frozen lateral moraine collapsed into South Lhonak. The impact produced a tsunami-like wave about 20 metres high. That wave overtopped and breached the frontal moraine, releasing roughly 50 million cubic metres of lake water. The flood then scoured its way downstream and eroded about 270 million cubic metres of sediment, greatly increasing the destructive mass moving through the Teesta valley.The slope that collapsed had not been motionless. According to the IIT Bhubaneswar account of the study, parts of it had been moving for years, reaching a maximum velocity of 15 metres a year between 2016 and 2023. That matters because monitoring only the lake surface would have captured one part of the developing hazard, not the unstable material beside it.Kristen Cook, a co-author of the study, put the lesson simply: “Anticipating future GLOFs requires understanding cascading hazards like landslides and sediment mobilization.” The phrase “cascading hazard” describes the sequence itself: a slope fails, a wave forms, a natural dam is overtopped or breached, the released flood erodes the valley, and the enlarged flow meets settlements and infrastructure.Christian Huggel of the University of Zurich highlighted another part of the same problem: “The thawing of permafrost and the instability of rock, ice and moraine structures pose major risks.” The University of Zurich summary of the study argues that high-resolution remote sensing is crucial precisely because several parts of the mountain system may be changing at once.The downstream hazard had also been modelled before the disaster. A 2019 hazard-modelling study by Sattar, Ajanta Goswami, and Anil V. Kulkarni modelled possible South Lhonak breach scenarios and flood routing downstream towards Lachen and Chungthang. The 2023 disaster did not reproduce the modelled scenario exactly, but the research had already established that failure of the lake could create serious downstream exposure.After the 2023 reconstruction, Sattar again stressed the policy implication: the disaster showed the urgency of early-warning systems in valleys exposed to glacier hazards. The warning is not just to watch lakes more frequently. It is to watch the processes around them.Transboundary geographyThe most striking number in the NRSC atlas may not be 7,570. It may be 20,473. Of the 28,043 lakes mapped across the Himalayan river basins, almost 73 per cent lie outside Indian territory in transboundary portions of those catchments.That does not mean all of those lakes threaten India. Many have no plausible flood pathway affecting Indian settlements. But it means India’s glacial-lake problem cannot be understood from an India-only map. The CWC, therefore, monitors lakes and water bodies beyond the country’s borders as well as those inside India.The difficulty is obvious. India may be downstream from a hazardous lake whose physical condition, weather, and surrounding slopes are in another jurisdiction. Satellite images can cross a political boundary effortlessly. Field teams, instruments, and administrative warning chains cannot.The Himalaya has already demonstrated the problem. On July 5, 2016, the small Gongbatongsha glacial lake in Tibet’s Poiqu basin failed after heavy precipitation contributed to slope failure above the lake. A 2022 peer-reviewed study published in Scientific Reports found that the outburst crossed into Nepal, damaging the Arniko Highway, buildings, and the Upper Bhotekoshi hydropower plant. Its peak discharge increased dramatically downstream as the flow picked up river water and large quantities of sediment already stored in the valley.The lesson is counterintuitive: a small lake can create a large disaster if the landscape below amplifies it. Research led by Simon K. Allen and Sattar has modelled the warning problem in the same transboundary region. In worst-case avalanche-triggered scenarios, according to a 2022 peer-reviewed study published in Natural Hazards and Earth System Sciences”, flood waves reached Nyalam in 6 to 11 minutes and the Nepal border in roughly 30 minutes. The authors argued that such short lead times make early warning alone insufficient; it has to be coupled with land-use zoning and local capacity to act.When the hazardous lake is upstream in another country, data sharing is therefore part of the warning system. Minutes lost in detecting, communicating, or verifying an event can become the difference between an alert and an arrival.Can observation be turned into warning?India’s response has accelerated since South Lhonak. The National Disaster Management Authority (NDMA) has integrated assessments by several agencies into a dynamic list of 195 high-risk glacial lakes. The list is intended to focus field expeditions and mitigation on lakes showing characteristics that justify closer attention.Separately, the CWC has finalised its 12-factor risk-indexing method and applied it to 100 Indian glacial lakes. That exercise is important precisely because it moves the discussion from a count of lakes towards a ranking of risk.The government has also approved a Rs. 150-crore National Glacial Lake Outburst Flood Risk Mitigation Programme for Arunachal Pradesh, Himachal Pradesh, Sikkim, and Uttarakhand. Its components include scientific assessment, automatic weather and water monitoring, early-warning systems, mitigation measures, and community preparedness.Hydropower is increasingly being brought into the equation. After the Teesta-III collapse, according to a Ministry of Jalshakti release published by the Press Information Bureau,the CWC decided to review the design flood of vulnerable existing and under-construction dams. The Central Electricity Authority identified 47 commissionedor under-construction dams that could potentially be affected by GLOFs originating from lakes in Indian territory; studies had been completed for 31 projects when the figure was reported.

A drone view of the “green miracle house” standing amid debris in Nepal’s Nuwakot district on September 1. The house withstood the flood and the mudslide in a landscape devastated by the natural disaster. | Photo Credit: FRANCIS MASCARENHAS/REUTERS The monitoring technology is also becoming more sophisticated. In June 2026, the Centre for Development of Advanced Computing (C-DAC) handed Sikkim an indigenous glacial-lake profiling technology suite that includes an autonomous vessel for depth mapping and three-dimensional volumetric profiling at high altitude.A month earlier, the Union Home Minister directed that a plan to develop early-warning systems for 30 high-risk lakes in Jammu and Kashmir, Ladakh, Uttarakhand, Himachal Pradesh, Arunachal Pradesh, and Sikkim be expanded to at least 60 lakes. That is a policy target, not evidence that 60 complete warning systems are already operational.And this distinction returns the story to its central question. A satellite can show that a lake has become larger. Radar can reveal movement on a slope. Field teams can measure lake depth. Models can show where water might travel. But none of these by itself gets a family out of a valley.An effective warning chain must connect observation to a threshold for action, then to district administrations, project operators, sirens or phone alerts, evacuation routes, and communities that know what the warning means. Scientists cannot yet predict the exact date and time at which a particular glacial lake will fail. Monitoring, therefore, works less like forecasting tomorrow’s rain and more like continually looking for changes that indicate growing instability.Also Read | When a flood crosses three countries, why can’t South Asia’s disaster plans?The Himalaya does not lack warning signs. Some appear over decades as a glacier retreats and a lake grows. Some appear over years as a slope begins to move. Others unfold in minutes when millions of tonnes of material fall into water. South Lhonak showed why those different clocks have to be read together.India’s satellites can now identify thousands of lakes and repeatedly measure hundreds of the larger ones. Its agencies have begun ranking them by risk, modelling downstream flood paths, inspecting selected lakes on the ground and developing warning technology. The next step is more demanding than producing a better map: it is ensuring that a change detected at 5,000 metres above sea level becomes a decision downstream, across districts, infrastructure systems, and sometimes national borders, before the water arrives.N.V. Geetha is a former Senior Technical Director at the National Informatics Centre. She writes data-based stories.References:1. National Remote Sensing Centre (2020-2023): Glacial Lake Atlas, National Hydrology Project.2. ISRO (2024): Satellite Insights: Expanding Glacial Lakes in the Indian Himalayas.3. Central Water Commission (June 2025): Report on Monitoring of Glacial Lakes & Water Bodies in the Himalayan Region of Indian River Basins.4. Central Water Commission (2024): Criteria for Risk Indexing of Glacial Lakes in Indian Himalayan Region.5. Ashim Sattar et al. (2025): “The Sikkim flood of October 2023: Drivers, causes and impacts of a multihazard cascade”, Science, Vol. 387, Issue 6740.6. Allen, Simon K., Ashim Sattar, Owen King, Guoqing Zhang, Atanu Bhattacharya, Tandong Yao, and Tobias Bolch (2022): “Glacial lake outburst flood hazard under current and future conditions: worst-case scenarios in a transboundary Himalayan basin”, Natural Hazards and Earth System Sciences, Vol. 22, No. 11, November 23.7. U.S. Geological Survey: M 5.2 Landslide - 55 km NW of Kodāri̇̄, Nepal, https://earthquake.usgs.gov/earthquakes/eventpage/us7000tbwb/executive.