How Nepal Lowers Dangerous Glacial Lakes: GLOF Mitigation Explained

A glacial lake outburst flood (GLOF) occurs when a lake ponded behind a glacier's moraine — a natural dam of loose rock and buried ice — suddenly breaches. A trigger as small as an ice avalanche, an earthquake or simple overtopping can cut through the dam in minutes, sending millions of cubic metres of water and debris down narrow Himalayan valleys. As the climate warms, Nepal's meltwater lakes are growing in number and size.

The hazard was quantified in a 2020 (2077 BS) inventory by the International Centre for Integrated Mountain Development (ICIMOD) and the United Nations Development Programme (UNDP). It mapped 3,624 glacial lakes in the Koshi, Gandaki and Karnali basins — 2,070 inside Nepal, 1,509 in the Tibet Autonomous Region of China and 45 in India — and found 1,410 large enough (0.02 sq km or more) to cause damage if breached. Of these, 47 were classified as potentially dangerous: 42 in the Koshi basin alone, with 21 of the 47 inside Nepal, 25 in Tibet and 1 in India.

Nepal knows the danger first-hand: the same inventory records at least 26 GLOF events affecting the country since 1977. The defining disaster came on 4 August 1985 (2042 BS), when an ice avalanche into Dig Tsho sent roughly five million cubic metres of water down the Bhote Koshi and Dudh Koshi rivers, destroying the nearly completed Namche small hydropower plant, around 14 bridges and about 30 houses. That flood pushed Nepal to become one of the first countries in the world to physically lower a dangerous glacial lake.

Lake lowering does not empty the lake — that is usually impossible. The aim is to drop the water level by a few metres in a slow, controlled way: even a 3-metre drawdown removes millions of cubic metres, reduces pressure on the moraine dam and creates freeboard so an avalanche-triggered wave is less likely to overtop it. Engineers choose between two techniques, often combining them.

Siphoning lays large pipes over the crest of the moraine; once primed, gravity pulls water continuously over the dam with no excavation, making siphons a quick, low-risk first step. Their weakness is capacity: high-altitude siphons freeze, clog and often discharge barely faster than glacier melt refills the lake, so they rarely achieve large drawdowns alone.

The permanent solution is an open outlet channel: a spillway excavated through the lowest section of the terminal moraine, fitted with gates so outflow can be regulated. Because cutting into a dam holding back tens of millions of cubic metres is delicate, the channel is dug above the waterline first and the lake then released gradually through the gate while instruments watch for erosion. Nepal used this method at both Tsho Rolpa and Imja Tsho.

Tsho Rolpa, at about 4,580 metres in the Rolwaling valley of Dolakha district, is Nepal's largest glacial lake — roughly 1.5 square kilometres, up to 135 metres deep and holding around 86 million cubic metres of water. It grew from a small pond in the late 1950s, and by the 1990s was rated the most dangerous glacial lake in the country, threatening villages and hydropower along the Tama Koshi corridor.

Test siphons laid over the moraine in the mid-1990s with Dutch technical support could not discharge water much faster than the lake refilled. Nepal therefore moved to a permanent fix: between mid-1999 and 2000 (2056-2057 BS) a roughly 70-metre gated open channel was cut through the terminal moraine, lowering the lake by 3 metres. The works cost close to US$3 million, funded mainly by the Government of the Netherlands — among the first engineered lowerings of a dangerous glacial lake anywhere in the world.

A GLOF early-warning system installed in 1998 (2055 BS) with World Bank support relayed sensor alarms by VHF radio to siren stations along roughly 100 kilometres of the Rolwaling and Tama Koshi valleys. It worked for only a few years: batteries and solar panels were stolen or failed during the conflict period, and by 2002 the network had collapsed — a cautionary tale about hardware installed without local ownership. The Department of Hydrology and Meteorology (DHM), with UNDP and Practical Action, revived monitoring from 2015-2016 with automated stations and a community-based warning chain.

Imja Tsho, just above 5,000 metres near Island Peak in Solukhumbu district, barely existed in the early 1960s, when satellite images show only small meltwater ponds. By a 2014 bathymetric survey it covered about 1.28 square kilometres, reached 149 metres deep and held around 75 million cubic metres. The 2015 Gorkha earthquake added urgency to plans under the Community-Based Flood and GLOF Risk Reduction Project (CFGORRP, 2013-2017), a Government of Nepal-UNDP project financed by the Global Environment Facility's Least Developed Countries Fund and implemented through DHM.

In the spring of 2016 (2073 BS), DHM commissioned the Nepal Army's engineering corps to carry out the lowering. Over more than six months, about 150 people — roughly 40 army personnel supported by local Sherpa high-altitude workers — installed siphons for initial relief and built a 45-metre gate-controlled outlet canal through the moraine. Controlled drainage began in late September 2016, and on 23 November 2016 (8 Mangsir 2073 BS) the government announced the lake had been lowered by 3.4 metres, releasing almost four million cubic metres of water safely down the Dudh Koshi.

The army described it as the highest-altitude disaster-risk-reduction work ever performed by a military anywhere. Project assessments counted about 71,750 people directly vulnerable along the first 50 kilometres downstream — including the Everest-trail hubs of Dingboche, Pangboche and Phakding — with wider estimates exceeding 96,000. Imja remains monitored rather than "solved": meltwater keeps flowing in, and the gates and gauges need ongoing maintenance.

The Department of Hydrology and Meteorology, under the Ministry of Energy, Water Resources and Irrigation, is Nepal's mandated agency for hydrological and meteorological monitoring, flood forecasting and public warnings. It operates the national gauging network and has led every major GLOF risk-reduction project from Tsho Rolpa in the 1990s to Imja in 2016.

A GLOF early-warning system follows a simple chain: detect, transmit, alert, respond. Sensors at the lake outlet and downstream detect the abnormal surge of a breach; telemetry — VHF radio in the 1990s, satellite and mobile links today — relays the signal within seconds; automated sirens sound in downstream settlements while SMS alerts reach residents and officials. Lead time is short: minutes for the nearest villages, an hour or more further down. The 2016 Imja system pairs lake sensors and hydro-meteorological stations with automated sirens in six settlements along about 50 kilometres of the valley, plus an SMS alert service.

Tsho Rolpa's 1998 collapse taught Nepal that sensors and sirens are the easy part. Modern installations budget for maintenance, train local operators and community caretakers, and run periodic evacuation drills so that a siren at 2 a.m. produces movement rather than confusion. Redundant sensors must corroborate each other before an alarm fires, protecting community trust from false alarms.

The CFGORRP (2013-2017 / 2070-2074 BS) deliberately paired the Imja works with a flood component in the Terai, where most of Nepal's flood deaths occur. It worked in four flood-prone districts — Mahottari, Siraha, Saptari and Udayapur — across the Ratu, Khando and Gagan river basins and the Hadiya and Kong tributaries of the Trijuga, rivers that shift course and burst their banks as they descend from the fragile Chure (Siwalik) hills.

Rather than relying on concrete alone, the project used eco-engineering: over 13 kilometres of embankment combined gabion revetment with bioengineering — planted grasses, bamboo and trees whose roots bind the soil. Such "bio-dykes" are cheaper than hard structures, can be built and repaired by communities themselves, and grow stronger as vegetation matures while also yielding fodder and fibre.

The human infrastructure gets equal weight. Community disaster-management committees were formed and trained, task forces assigned for warning dissemination, search and rescue and first aid, evacuation centres established, and drills run before each monsoon. This community-based disaster risk reduction (CBDRR) model is now standard in Nepali policy, coordinated since 2019 by the National Disaster Risk Reduction and Management Authority (NDRRMA) under the Disaster Risk Reduction and Management Act 2074 (2017).

On 16 August 2024 (2081 BS), a flood tore through Thame village in the Khumbu. The source was the Thyanbo supraglacial lake, a pond of only about 0.05 square kilometres that appeared on no danger list. Initial assessments recorded 14 properties destroyed — a school, a health post, five hotels and seven homes — and about 135 people displaced; timely daytime evacuation meant no lives were lost. ICIMOD's lesson: small, short-lived lakes can be more hazardous than previously assumed, because debris-charged floods amplify in steep channels.

The big monitored lakes remain on watch. Thulagi, below Manaslu in the Marsyangdi basin, and Lower Barun, near Makalu in the Arun basin, rank among Nepal's highest-priority dangerous lakes and have been studied for possible future lowering. The harder problem is transboundary: 25 of the 47 potentially dangerous lakes lie in Tibet, upstream of Nepali rivers, making data-sharing and joint monitoring with China essential and, so far, incomplete.

Nepal's toolkit is nevertheless among the most tested in the world: two engineered lake lowerings, rebuilt early-warning networks, an expanding automated hydromet system under DHM, satellite lake monitoring with ICIMOD, and a community-based response model studied by other mountain countries. The constraint is money and maintenance — a lowering costs millions of dollars per lake and warning systems die without upkeep — so Nepal continues to press for international climate-adaptation finance.