2. Flooding incident

Early on Sunday, April 21, 2024, residents of Samagaun in Chumanubri Rural Municipality, Gorkha, reported a sudden surge in the flow of the Budhi Gandaki River. This surge resulted in the destruction of a wooden bridge on the route from Samdo to Samagaun. Local people had already noticed increasing water levels in the river on Saturday, with a significant rise in flow and sediment load by Sunday morning. At around 09:30 local time on Sunday^[1], trekkers^[2] and residents of Samagaun witnessed massive ice avalanches from the Manaslu Glacier snout, which slid and dropped into Birendra Tal. Evidence shows that the displacement wave generated by the event led to the outflow from the lake and subsequent high river flow.

3. Mechanism behind the flood

A rapid assessment by ICIMOD and Kathmandu University (KU) with support from the University of Alaska Fairbanks, USA, using satellite images from Planet.com, corroborated the reports by local people, i.e. the flood resulted from the calving^[3] of the snout of Manaslu Glacier, which was steep and heavily crevassed, into the lake The current terminus of the glacier is about 600m above the Birendra Tal water level. This large ice mass swiftly flowed into the Birendra Tal, impacting its upper section and causing the lake water to overflow through the outlet. This resulted in a sudden increase in river flow in the Budhi Gandaki River and downstream.

Glaciers in Nepal are classed as both ‘summer accumulation’ and ‘summer ablation’ type. This means that predominant accumulation of snow and ice on the glaciers occurs during the summer (monsoon) season, while they also experience significant ablation (losing snow and ice, mainly through melting) during this season. As the summer approaches, the snow and then glacier ice begin to melt due to increasing temperatures.

The sequential satellite images show a rapid transformation on the landscape around Birendra Tal, showcasing the swift disappearance of fresh snow and frozen water surface in the area just within a span of a few weeks. The generally increasing temperatures during the period preceding the event are likely to have caused this rapid transformation. This is depicted in the comparative images from Sentinel and PlanetScope Figure 2 below.

Figure 2: Sentinel 2 MSI image dated 15 March 2024 (top left), 09 April 2024 (top right) and PlanetScope image dated 19 April 2024 (Bottom) shows the major disappearance of snow cover in the valley and glacier surface at a lower elevation and melting of the frozen lake surface.

As the snow cover on the glacier starts to melt, the underlying glacier surface, having a lower albedo (amount of light reflected back to the atmosphere from the surface) is exposed and also starts to melt. The snow avalanche from the upper reach is also visible in the image comparison of 19 and 20 April. The melting of both the snow surface and the glacier ice allowed meltwater to seep beneath the crevasses, potentially weakening and destabilising the glacier, leading to the detachment of ice masses or seracs – huge blocks or columns of ice that can topple without warning. We hypothesise that similar phenomena prevailed in this case. The ice mass then flowed downward over the exposed rock and through narrow dissected channels, flowing over the ice and debris mixed deposits (or ‘apron’) and plunging into the lake. This phenomenon was likely of a smaller scale and the flow was confined to the left-hand side of the rock slope and ice and debris apron, as indicated in Figure 3 and Figure 4 dated April 20, 2024, and did not cause significant displacement of the lake water. This corroborates the observation by residents of Samagaon.

The event of 21 April was likely of a much larger magnitude with the avalanche covering the whole width of the rocky slope. According to our analysis of the 21 April event, detachment of a massive ice mass from the extensively crevassed part of the glacier snout occurred. This is clearly seen in the textural changes in the PlanetScope satellite images from 19 to 21 April 2024 (the source area is labelled and marked by a orange arrow in the image dated 21 April; see Figure 3, Figure 4 and Figure 5). The propagation of ice in the lake water can be seen in the images of 21 and 22 April (Figure 5 and Figure 6). The displacement wave created by the huge mass of ice falling into the lake caused the water to overflow, which produced flooding in the immediate downstream area. The evidence from the image showing the number of icebergs floating on the lake water also proves that it was predominantly an avalanche of ice rather than snow, as some initial speculations had suggested.

×

4. Tracking GLOF evidence from the past

The high-resolution satellite images from Google Earth clearly depict a glacial lake outlet spanning approximately 30–50 metres wide (Figure 7). The V-shaped cross-section of the channel shown in Figure 8 presents a profile among the crest of the end moraine dam from ALOS 30m DEM; high-resolution satellite images in Google Earth show that an open channel outlet exists in the lake that is able to release a certain amount of flow safely (Figure 7 and Figure 8). The channel could have been created by lake outflow or by past GLOF events. The open channel is over 700 metres long measured from the rim of the lake to the moraine dam toe (Figure 9). The channel drops by 50m within this distance, giving it a slope of around 4 degrees. The outer side of the dam is covered by vegetation, indicating a more compact and stable slope. The present overflow was also from the same outlet with no significant impact on the dam, which is clearly visible in the images of 19 April, 2024 and 21 April 2024 shown in Figure 10 below.

To determine if the open channel was formed from past GLOF events, we analysed historical Landsat satellite images to track the formation of the V-shaped open channel on the moraine dam.

The Landsat image dated 28 October 1976 shows the distinct channel opening, whereas the image from 2 January 1973 (Figure 11) shows the position of the lake but the lake outlet is not distinctly visible. We further tracked the high resolution historical declassified satellite photographs from Corona (8 Nov 1967) and Hexagonal (16 Nov 1973), which confirm the open channel of the lake with erosion and deposition along the moraine dammed area (Figure 12 and Figure 13).  However, we cannot find any significant changes in the downstream river flow area, indicating erosion or deposition of flow sediments due to flash floods over time from the late 1960s onward. This suggests that the lake either experienced a breach prior to the late 1960s or the channel was opened during the glacial process of melting, deposition of debris and formation of lakes. Understanding and tracking these processes require obtaining and analysing further historical aerial photographs and other sources of satellite images before and after the lake formed.

The significant debris deposition on the left side of the lake is also observed, which flows from the left adjacent valley and through the left lateral moraine of the lake. There is also high potential for debris flow and ice and/or snow avalanches from the adjacent valley from the left.

Figure 10: PlanetScope image dated 19 April 2024 (left) shows the open outlet of the lake and an image dated 21 April 2024 taken at 10:54 NPT/05:09 UTC (right) shows an increase in the outlet water flow without significant impact on the dam.

×

5. Lake development and tracking of similar past events

The analysis of the Landsat and PlanetScope images spanning from 1973 to 2024 reveals significant fluctuations in the area covered by the Glacial Lake (Figure 14 and Figure 15). Initially measuring around 0.27 square kilometres (km²) in 1973, the lake area decreased to approximately 0.24 km² by 1976. Notably, a drastic reduction occurred, with the area halving to 0.11 km² by 1988 within a ten-year period. Following a period of stability until 1990, the lake area began to rapidly increase, reaching its maximum extent and covering 0.25 km² by 2000. Slight fluctuations in subsequent years are observed, which may be attributed to overflows caused by avalanches or debris flow from the glaciers and their adjacent valley. Some past events are also visible in the satellite images shown below (Figure 16 and Figure 17), and were also covered in the media^[4].

However, recent observations from 19 April 2024, just before the event, indicate a slight decrease in the lake area to approximately 0.24 km² and no significant changes in the lake area after the event based on a PlanetScope image dated 22 April 2024 (0.22 km²) except the ice deposition part on the upper reach of the lake. These findings underscore the dynamic nature of glacial systems and the importance of continuous monitoring to understand their response to changing environmental conditions.

The comparative images on 28 September, 14 and 30 October 2006, and 1, 8 and 9 April 2022 in Figure 16 and Figure 17 show evidence of similar events that occurred in the past. This serves as an illustration, to highlight the recurring nature of avalanches from the Manaslu Glacier. Such events are attributed to the morphological condition of the glacier, its processes, and presence of densely covered crevasses on its surface. The fluctuation in the lake surface area, shown in Figure 14 and Figure 15, also offers insights into the frequency of these kind of events. This underscores the importance of delving into the intricate physical process governing glacier movements as well as the dynamics of the formation and detachment of crevasses. Achieving this understanding necessitates continued monitoring of the glacier, utilising high spatio-temporal resolution satellite images alongside field-based monitoring such as time-lapse camera or ground-based interferometric synthetic aperture radar (InSAR) including monitoring of climate/weather in the catchment.

×

6. Estimation of peak outflow discharge

To estimate the peak discharge of the lake overflow, we conducted a back analysis using water level data from Ghap and Jagat, collected by a team from KU, and data from Arughat, collected by members of the Department of Hydrology and Meteorology (DHM), and we generated cross-section information utilising 2m DSM access through the SERVIR Program. These hydrological sites are located at 22, 48, and 87 km downstream from the lake, respectively (see Figure 18). The data shows that the flood reached Ghap at 10:45AM, Jagat at 01:15 PM, and Arughat at 05:30 PM after it was reported at 9:30 AM by an eyewitness^[5] (Figure 18).

We estimated the flow velocity, based on the travel time and abrupt water level rise measured by these hydrological stations and considering the average slope of the channel between the locations. The 1D-hydrodynamic model was set up in DHI Mike 11 software, integrating the cross-section of the river channel generated using 2m DSM and calibrated using the closest station data (Ghap station) to estimate the peak overflow discharge at the lake outlet. The results show that the peak flow discharge at the lake outlet was 32 cubic metres per second; this is shown in Figure 18. Additionally, the cross-sectional plot showing the simulated water level from just below the moraine dam, close to Sama village and at the Ghap station in Figure 19 and Figure 20 below, indicates that the flow is insufficient depending on channel morphology, to cause major damage in the downstream riverbank. However, factors such as flow velocity, river morphology and geomorphology, and sediment characteristics on the riverbed and banks also play a critical role in determining the impact of the flow along the river. This also clearly suggests that the present event was an overflow of lake water sparked exclusively by ice avalanches from the calving of the glacier snout, not by a breaching of the lake. Therefore, this event is not classed as a GLOF.

Figure 18: Location map (left) of the Birendra Tal and hydrological stations operated by KU (Ghap and Jagat) and DHM (Arughat); the water level measured by these stations (upper right graph), and the simulated peak flow discharge at the lake outlet based on the water level data measured at Ghap station at 22km downstream from the lake (lower right graph)

Figure 19: Cross-section (white line on the top image) and simulated water flow level (lower graphs) just below the moraine dam and near to Sama village, showing the water level increased to approximately 60cm (red dotted line indicates the simulated water level and the blue solid line is the base flow estimated using the high-resolution image fromGoogle Earth); flow width is around 22 metres.

Figure 20: Cross-section (white line in top image) and simulated water flow level near the Ghap station showing the water level increased to apprximately 60cm and flow width is around 20m (the red dotted line indicates the simulated water level and the blue solid line is the base flow estimated using the high resolution image from Google Earth)

×

7. Conclusion and recommendations

Based on preliminary analysis using satellite data, the flood event on 21 April 2024 originating from Birendra Tal, was attributed to the ice avalanches due to the dry calving of the Manaslu Glacier snout, which sparked the lake overflow.

The magnitude of the overflow was influenced by various factors including the volume of ice and snow released, the manner of the fall, distance from the drop point to the lake, and the channel geometry.

Although the 21 April event was moderate in scale and did not result in significant downstream damage, it’s essential to note that both the steep and crevassed frontal section and the crevasses over the whole of the Manaslu Glacier have a historical propensity for dry calving. With projected temperature increase and glacier retreat, future events could surpass the moderate impact observed in this instance.

There is also high potential for debris flow and ice and/or snow avalanches from the left adjacent valley of the lake. The significant debris deposition on the left side of the lake is also observed, which flows from the left adjacent valley and through the left lateral moraine of the lake.

Assessing the potential risk posed by avalanches from glaciers and debris flow from the adjacent valley into the lake and understanding the scale and extent of their impact on the lake, is essential for evaluating the potential magnitude of future disasters. This necessitates a detailed understanding of glacier flow velocity, crevasse formation dynamics, and ice detachment processes. Such information can be acquired through remote sensing techniques and field-based monitoring, such as Ground-based Interferometric Synthetic Aperture Radar (GB-InSAR) or the installation of time-lapse cameras.

To mitigate such risks, continuous monitoring of the lake using satellite imagery and field monitoring by installing a time lapse camera is recommended. Additionally, implementing in-situ sensors for lake-level monitoring and establishing a flood early warning system would enhance preparedness and response capabilities. As the results presented here are derived from rapid analysis, it is imperative that they be substantiated by more detailed assessments, potentially supported by field investigations.

Rising temperatures and rapid glacier retreat are growing concerns across the region. As they pose risks from various mountain hazards similar to the current avalanches triggering overflow from the lake, an extensive event (GLOF) could happen from much larger glacial lakes in the region. The countries in the region remain ill-prepared to cope with the rapid pace of these changes. Addressing this issue will requires urgent and strong political action to implement effective risk mitigation strategies as well as investment in scientific research to generate robust data. ICIMOD continues to work with our regional member countries to ensure line agencies have the required skillsets and access to the latest technology to monitor risks and analyse their data to feed into the decision-making process.

Acknowledgements

This rapid analysis has been conducted based on the Planet satellite imageries accessed through SERVIR-HKH. SERVIR is a joint initiative of NASA, USAID, and leading geospatial organisations in Asia, Africa, and Latin America. ICIMOD implements the SERVIR Hindu Kush Himalaya (SERVIR-HKH) Initiative regional hub in its regional member countries in the HKH region. We have also used other freely accessible satellite images and photographs through USGS portals (Glovis and Earth Explorer). The 2m Digital Surface Model (DSM) utilised for back analysis was accessed through SERVIR. Data from the hydrological stations were accessed through our close collaboration with KU and DHM.

Authors would also like to acknowledge Gillian Summers for editing the document and Chimi Seldon for coordinating science and Centre Communication Unit

References

[1] Ministry of Home Affairs urged to be vigilant in the coastal area of ​​Budhigandaki :: Setopati Correspondent :: Setopati (www-setopati-com.translate.goog)

[2] https://english.onlinekhabar.com/eyewitnesss-statement-we-saw-the-critical-situation-in-birendra-glacier-lake-we-were-saved-by-luck.html

[3] Avalanching glacier instabilities: Review on processes and early warning perspectives – Faillettaz – 2015 – Reviews of Geophysics – Wiley Online Library

[4] https://kathmandupost.com/national/2015/04/10/melting-of-glacial-lake-threatens-life-property

[5] https://shorturl.at/hPQS0