ICE 2 Water
Nepal's glacial lakes are expanding. Three and a half decades of satellite imagery — from 1989 to 2025 — reveal how the cryosphere is being converted to water at an accelerating pace, reshaping downstream hydrology and intensifying Glacial Lake Outburst Flood risk across the Third Pole.
Twelve major glacial lakes across Nepal's high-mountain river basins are tracked continuously using Google Earth Engine and multi-mission Landsat imagery. The resulting analysis — lake boundary polygons, area time series, and before/after comparisons — is published on a live interactive platform open to researchers, planners, students, and policymakers worldwide.
Open ICE 2 Water Platform
36 Years of Change
Nepal's glaciers are retreating — and the lakes they leave behind are growing fast
Glaciers are slow-moving giants that have shaped the Himalayan landscape over millennia. As global temperatures rise, they are retreating at accelerating rates, leaving behind overdeepened basins, hollows, and valleys. These depressions fill with meltwater to form glacial lakes — bodies of water that grow year by year as the ice above them continues to shrink.
In Nepal, this is not a distant future concern. It is already happening at high elevation, across dozens of watersheds, with consequences that reach far downstream. Communities, infrastructure, hydropower stations, and farmland all sit in the paths of rivers fed by glacial terrain. As these lakes grow, so does the probability of a sudden, catastrophic outburst — a Glacial Lake Outburst Flood.
A GLOF occurs when the natural dam holding back a glacial lake — made of moraine, ice, or bedrock — fails catastrophically. The stored water is released in a sudden surge, carrying sediment, boulders, and debris downstream at enormous velocity. GLOFs can devastate communities, destroy bridges and roads, damage hydropower infrastructure, and alter river ecosystems for years. Key triggers include moraine dam erosion from rising water pressure, ice avalanches falling into the lake, seismic activity, and permafrost degradation weakening dam structures.
The challenge is not a lack of data — satellite imagery covering these lakes has existed since the late 1980s. The challenge is converting that raw archive into something legible, accessible, and immediately useful to those who need it. ICE2Water was built to close exactly that gap: from raw Landsat pixels to an open, interactive platform anyone can explore.
Six ways glacial lakes form — each with a different risk profile
Not all glacial lakes are the same. The type of dam, the position relative to the glacier, and the drainage mechanism together determine how a lake behaves and how dangerous it may become. The platform's Learn section explains all six types in full.
Meltwater pools behind a dam composed of glacial debris — rock, sediment, and ice cores — deposited by the retreating glacier. These are the most common lake type in Nepal and carry the highest GLOF risk. The moraine wall can be porous, ice-cored, and susceptible to overtopping or piping failure as the lake grows.
Highest GLOF riskFormed when advancing or retreating glacier ice blocks the natural flow of a river or valley, trapping water behind a wall of ice. These lakes can drain with almost no warning if the ice dam weakens, floats free, or is undermined by subglacial drainage — releasing water in sudden, violent pulses.
Sudden drainage riskFound directly in front of a glacier's terminal end. As the glacier retreats, it exposes the basin it has eroded over centuries, which fills with meltwater. The majority of lakes monitored on the ICE2Water platform are proglacial. They grow continuously as glacier retreat accelerates.
Most common in NepalFormed in bowl-shaped hollows called cirques, carved by glaciers at high elevations. After the glacier melts away, the cirque fills with water. Generally more structurally stable than moraine-dammed lakes, but can drain if the bedrock rim is fractured or overtopped during extreme events.
Generally stableBodies of water that form on the surface of a glacier or ice sheet during summer melt seasons by accumulating meltwater in surface depressions. They can drain rapidly through crevasses or moulins, and their sudden drainage can destabilise the glacier structure itself, accelerating calving and ice loss.
Seasonal, rapid drainageHidden beneath thick ice sheets or glaciers, maintained in liquid form by geothermal heat and the insulating pressure of overlying ice. Not visible from the surface or detectable by optical remote sensing. Their dynamics are relevant to understanding long-term basal melt rates and glacier stability globally.
Hidden beneath the iceTwelve glacial lakes across Nepal — from Solukhumbu to Bajhang
The platform monitors twelve major glacial lakes spanning Nepal's high-mountain regions — from the eastern Khumbu and Makalu-Barun areas through the central Rolwaling and Manaslu ranges to the far-western Saipal Himal. Each lake is tracked with annual or near-annual satellite imagery, with lake boundary delineation and area calculated using Google Earth Engine.
| Lake | Area (start → 2025) | Expansion |
|---|---|---|
| Imja Glacial LakeSolukhumbu · Everest Region · Dudh Koshi | 0.64 → 2.28 km² | +1,167% |
| Nunekhara Glacial LakeBajura · Saipal Himal · Kuwadi–Karnali | 0.03 → 0.40 km² | +1,033% |
| Glacial Lake — Gatlang ValleyDhading · Ganesh Himal · Bagmati Province | 0.02 → 0.08 km² | +300% |
| Tsho RolpaDolakha · Rolwaling Valley · Central Himalaya | 0.001 → 1.78 km² | +295% |
| Lower Barun Glacial LakeSankhuwasabha · Barun Valley · Makalu-Barun | 0.78 → 2.76 km² | +254% |
| Raidhungi LakeBajhang · Thado Khola · Sudurpashchim | 0.18 → 0.55 km² | +206% |
| Hongu 2Solukhumbu · Koshi Province · Everest region | 0.47 → 1.16 km² | +147% |
| Lumding TshoSolukhumbu · Dudh Koshi watershed | 0.71 → 1.55 km² | +118% |
| Birendra LakeGorkha · Manaslu Glacier · Gandaki Province | 0.14 → 0.30 km² | +114% |
| Thulagi (Dona) Glacial LakeManang · Marshyangdi Basin · Gandaki Province | 0.75 → 1.05 km² | +33% |
| Barun Glacial LakeSankhuwasabha · Makalu-Barun · Koshi Province | 0.31 → 0.39 km² | +23% |
| Glacial Lake — Head of Seti RiverBajhang · Saipal Himal · Sudurpashchim | ~0.00 → 0.10 km² | Newly formed |
Among the most closely watched: Tsho Rolpa in Dolakha is Nepal's highest-risk glacial lake — a risk-reduction sluice was constructed in 2000, yet the lake continues to expand. Thulagi (Dona) in Manang carries a modelled potential economic loss of US $406.73 million in the event of a GLOF, with peak discharges of 1,400–5,400 m³/s and downstream exposure for approximately 165,000 residents — as well as direct threat to the Marsyangdi (69 MW) and Middle Marsyangdi (70 MW) hydropower stations.
A reproducible GEE pipeline — raw satellite to mapped lake boundary
Every lake boundary, every area measurement, and every before/after image on the platform was produced using a reproducible, code-driven pipeline built entirely in Google Earth Engine. The full methodology is documented and the GEE scripts are publicly shared on the platform's Source Code page — written to be reusable for any glacial lake region with minimal modification.
Historical imagery from 1984–2012 uses Landsat 5 TM. Contemporary imagery from 2013–2021 uses Landsat 8 OLI, and from 2022 onward Landsat 9 OLI-2. The pipeline automatically selects the correct sensor for each year, ensuring a consistent 30-metre spatial resolution across the entire 36-year time series — even as the satellite constellation changes overhead.
Imagery is filtered to the post-monsoon season — typically October to November — when seasonal snow cover is minimal and atmospheric transparency is highest. A maximum cloud cover threshold of 20% is applied per scene. When multiple qualifying images exist for a given year, a pixel-wise median composite is computed, effectively removing residual clouds, cloud shadows, and transient surface outliers.
To isolate open water from surrounding terrain, the Normalised Difference Water Index is calculated from Green and Near-Infrared (NIR) bands: NDWI = (Green − NIR) / (Green + NIR). Pixels with NDWI above 0.3 are classified as open water. This index is highly effective for glacial lake delineation because water, ice, snow, and dark rock surfaces occupy distinctly different positions in the Green–NIR spectral space.
The binary water mask is applied within a defined region of interest around each lake. The total number of water-classified pixels is multiplied by the pixel ground area (30 m × 30 m = 900 m²) to produce the lake surface area in square metres, then converted to km². This calculation is automated and run for every year in the monitoring period, producing the annual lake area time series displayed on the Explore page.
The water mask is converted to vector polygon outlines using GEE's reduceToVectors function with 4-connected pixel clustering, producing clean lake boundary delineations. Boundary polygons and true-colour median composite images are exported to Google Drive as GeoTIFFs at 30-metre resolution, then post-processed and rendered as PNG images overlaid with lake boundaries for display on the platform.
Six sections — exploration, science, education, and connection
ICE2Water is not a static map or a single dataset. It is a fully navigable web platform with six distinct sections, each serving a different kind of user — from the researcher who needs the GEE source code, to the community member who wants to recommend a lake for addition to the monitoring network.
A Leaflet.js map of Nepal with clickable markers for all twelve monitored glacial lakes. Each marker links to the lake's dedicated monitoring page. A guided walkthrough modal introduces first-time visitors to the platform. The map is the spatial entry point — users explore by geography, not by list.
Each lake has its own dedicated monitoring page with side-by-side before/after satellite imagery, lake coordinates and elevation, formation type, area data for earliest and most recent years, and a detailed description of expansion history and GLOF risk profile. Data spans up to 36 years of imagery per lake.
A comprehensive educational section covering what glaciers are, how glacial lakes form, all six lake types, their importance and ecosystem services — freshwater provision, hydroelectricity, biodiversity, and climate indication — and the causes and downstream impacts of Glacial Lake Outburst Floods. Accessible to non-specialists.
The platform's mission — bringing clarity and awareness to glacial lake change from 1989 to 2025, and helping cryospheric workshop attendees understand the vulnerabilities of downstream communities — is set out alongside full team profiles, acknowledgments, and institutional partnerships.
A live contact form allowing anyone — a researcher, NGO, local community member, or field scientist — to recommend a glacial lake for addition to the monitoring network. Fields include lake name, coordinates, district, and description. This keeps the platform growing through open, collaborative contribution.
The complete GEE scripts — Landsat time-series exporter, NDWI water mask generator, and lake boundary vectorisation — are published with detailed explanatory comments. Written to be reusable for any glacial lake study region with minimal modification, enabling full methodological reproducibility.
Why glacial lakes matter — numbers from the region
As Himalayan glaciers retreat under rising temperatures, meltwater accumulates in moraine-dammed lakes at an accelerating rate. The Hindu Kush Himalaya region is warming at 1.5 times the global average — a rate that is compressing centuries of glacial change into a matter of decades. When natural dams fail — triggered by landslides, earthquakes, ice avalanches, or the growing weight of water — the stored volume is released in hours, not years.
Downstream communities, farmland, bridges, and hydropower infrastructure across Nepal face direct exposure. Long-term, reproducible monitoring of lake area is the first and most essential step in understanding, anticipating, and communicating these risks to the people who need to act on them.
Beyond the hazard — why these lakes also sustain life downstream
Glacial lakes are not only potential hazards — they are ecologically and economically vital systems. The platform's Learn section situates GLOF risk within a fuller picture of what these lakes provide, and why their preservation and careful monitoring matters far beyond disaster preparedness alone.
Glacial lakes provide freshwater to communities downstream and act as natural buffers that regulate seasonal river flow. During dry seasons, they sustain river levels that agriculture, drinking water systems, and downstream ecosystems depend on — making their health directly linked to human water security across the Himalayan region.
Nepal's hydropower potential — one of the largest in the world — is directly tied to glacially-fed river systems. Several major installed and planned hydropower projects sit downstream of monitored lakes. Understanding lake dynamics is essential for ensuring the long-term viability of these critical energy assets.
The expansion of glacial lakes is one of the most spatially precise and temporally consistent signals of climate change in high mountain environments. Their growth rate, elevation distribution, and formation type encode information about temperature trends, precipitation phase shifts, and glacier mass balance across the region.
Glacial lakes create unique aquatic ecosystems at extreme elevation. They support specialist aquatic species and serve as critical resting habitat for migratory birds along Himalayan flight corridors. As these lakes grow and new ones form, they open new ecological niches in previously uninhabited high-altitude terrain.
Built for everyone who needs to understand the ice above them
A reproducible GEE-based methodology, annual lake area time series, and published source code enable researchers to validate, extend, or compare findings with their own work on Himalayan cryosphere dynamics and GLOF hazard assessment.
Infrastructure professionals involved in road, bridge, and hydropower development downstream of glacial terrain can use the platform to understand historical lake growth trends, spatial risk profiles, and the watersheds most urgently requiring monitoring attention.
Accessible, visual, and grounded in satellite data — the platform translates remote sensing science into policy-relevant knowledge about GLOF risk, early warning needs, and national adaptation priorities across Nepal's mountain regions.
The Learn section and open source code make ICE2Water a teaching resource. Students can engage with real satellite data, understand remote sensing methodology, and study a concrete, live example of applied climate science in one of the world's most vulnerable mountain environments.
Exhibited at Faces of Ice — in support of the International Year of Glaciers' Preservation 2025
ICE2Water was originally developed and exhibited as part of the Faces of Ice initiative — an exhibition that provided the platform and opportunity to showcase science-driven tools about glacier change to a broader international audience. The platform directly supports the International Year of Glaciers' Preservation (IYGP) 2025, a globally coordinated effort to raise awareness about accelerating glacier loss and its consequences for human and ecological systems worldwide.
Institutional development and support was provided by The Small Earth Nepal (SEN), with scientific guidance from Dr. Dhiraj Pradhananga (Glaciologist), whose continuous encouragement and expertise were instrumental in shaping the platform's scientific framing and bringing the work to fruition.
Explore the live platform
The ICE 2 Water platform is hosted online and continuously displays satellite imagery, per-lake area trends, and interactive maps for all twelve study sites. Built and maintained by the team at The Small Earth Nepal.
Open ICE 2 Water