Deadly Langtang rock-ice avalanche in Nepal followed decades of glacier retreat and exceptional high-altitude heat

The August 26 failure provides a case in which long-term cryosphere change can be compared with conditions immediately before a major high-mountain disaster. HiRISK’s rapid hazard assessment found that glacier ice and underlying bedrock detached across a zone approximately 1–1.3 km (0.6–0.8 miles) wide near 5 200 m (17 060 feet), descended to the valley floor below 3 000 m (9 840 feet), temporarily blocked the Lhende Khola, and released a flood when the deposit failed. HiRISK classified the event as a rock-ice avalanche and flood rather than a conventional glacial lake outburst flood.

The physical sequence is increasingly well constrained; the trigger is not. Post-event satellite imagery showed that both glacier ice and bedrock had failed, while HiRISK’s assessment stopped short of attributing the collapse to climate change. Climate-related weakening processes remain possible contributors rather than demonstrated causes of the August 26 failure.

Silwal et al. (2026) reconstructed glacier change across the Langtang Catchment from the Little Ice Age, around 1815, to 2023. Total glacier area decreased by 78.46 ± 4.38 km² (30.3 ± 1.7 mi²), or about 41.5%, during that period. Glacier-area loss accelerated from 0.11% per year between the reconstructed Little Ice Age extent and 1964 to 0.49% per year from 1964–2023, with still faster losses during parts of the post-2000 period.

One result can appear counterintuitive without explanation: the number of mapped glaciers increased from 58 to 115 even as total glacier area declined sharply. The catchment did not gain ice. Large glaciers separated into smaller units as thinning and retreat disconnected formerly continuous sections, increasing the glacier count while reducing the total area covered by ice (Silwal et al., 2026).

The study also documented widespread glacier thinning, slower surface flow and rising equilibrium-line altitudes. The equilibrium-line altitude, or ELA, is the elevation where annual snow and ice gain approximately balances annual loss. When the ELA moves uphill, a larger proportion of a glacier lies in the lower, melt-dominated zone. In Langtang, rising ELAs increasingly intersected steep, thin, and heavily crevassed icefalls, contributing to glacier fragmentation and disconnection (Silwal et al., 2026).

Warming was strongest at higher elevations. Silwal et al. (2026) found warming of about 0.3°C (0.54°F) per decade above 4 000 m (13 120 feet) from 1964–2023, compared with about 0.1°C (0.18°F) per decade at 3 865 m (12 680 feet). This elevation-dependent warming means that high mountain terrain has warmed faster than the lower-elevation reference site, changing the thermal conditions that control snow persistence, glacier melt, and frozen mountain ground.

HiRISK documented a more localized structural change at the August failure site. The glacier retreated approximately 450 m (1 480 feet) between 1990 and 2020 and moved across a bergschrund, the large fracture that commonly separates moving glacier ice from more stationary ice or rock near the head of a glacier.

The assessment said long-term retreat had probably produced a debuttressing effect on the underlying bedrock slope. Debuttressing occurs when retreating ice removes some of the mechanical support previously exerted against a steep rock face, changing the stresses acting on the exposed slope.

HiRISK also noted that permafrost is generally warming in mountain terrain and that recent glacier recession may have facilitated the failure process. Permafrost is rock or ground that remains frozen for at least two consecutive years. In fractured mountain rock, ice inside cracks can add cohesion; thawing can reduce that frozen bonding and allow more liquid water to enter fractures. HiRISK did not report direct measurements showing a site-specific permafrost temperature trend on the failed slope.

A separate short-term temperature reconstruction adds another layer to the event. Berkeley Earth chief scientist Robert Rohde used ERA5 reanalysis together with observations from high-altitude weather stations to estimate conditions near the failure zone at about 5 200 m (17 060 feet).

Rohde estimated an average temperature of about 5°C (41°F) from August 21–26. AFP reported that corresponding six-day averages in his reconstruction remained below 4°C (39.2°F) throughout more than 55 years covered by the series. Rohde also estimated that the six-day mean at the site had warmed by about 1.8°C (3.2°F) since the mid-20th century.

The values are reconstructed estimates rather than direct measurements from a weather station at the failure point. Rohde’s analysis therefore establishes the thermal conditions estimated for the period before the failure, but it does not constitute an event-attribution study demonstrating that the heat caused the avalanche.

Scientists interviewed by AFP identified several physical processes through which unusually warm conditions can influence steep ice-rock slopes. Valérie Masson-Delmotte said high temperatures can contribute to permafrost thaw, while CNRS glaciologist Étienne Berthier said satellite imagery showed substantial snowmelt before the disaster, providing liquid water capable of entering mountain fractures.

Additional meltwater can alter water pressure and friction inside fractures or beneath glacier ice. Thawing permafrost can weaken ice-filled cracks, while glacier thinning and retreat can change the stresses acting on steep rock and ice. These mechanisms are physically plausible, but their relative importance in the August 26 failure has not been established.

That distinction separates what was observed from what is inferred. Post-event imagery documented the loss of glacier ice and bedrock. Silwal et al. (2026) measured and reconstructed long-term glacier retreat, fragmentation and elevation-dependent warming across the catchment. Rohde reconstructed the short-term temperature conditions before the failure. Debuttressing, meltwater infiltration and permafrost degradation remain mechanisms that may have contributed to instability rather than demonstrated triggers of this particular event.

The event also exposed a monitoring challenge. HiRISK found that no operational early-warning system for glacier-related hazards covered the failure site. Retrospective evidence included faster ice-surface movement toward the eventual failure area shortly before the event, discolored meltwater by August 24 and possible crack development extending into surrounding bedrock.

The avalanche also produced a strong seismic signal, which HiRISK identified as potentially useful for future regional monitoring. Conventional river gauges close to the source would provide little warning because the initial avalanche and flood moved rapidly, while monitoring farther downstream could offer a longer response window.

HiRISK estimated upper-reach flood speeds above 30 m/s (67 mph) and reported that the flood wave reached Rasuwagadhi at about 08:44 LT (02:59 UTC). The rapid propagation illustrates why monitoring the unstable slope, glacier movement, and seismic signals may be more useful close to the source than relying only on downstream detection.

The remaining scientific question is narrower than it was immediately after the disaster. Researchers have increasingly constrained what failed and how the rock-ice avalanche developed into a flood, but the process that pushed the slope across its failure threshold on August 26 remains unresolved.

Answering that question will require combining pre-failure deformation, fracture development, meltwater conditions, glacier geometry, permafrost state and mechanical modelling. Until those constraints are available, decades of glacier retreat, high-elevation warming, and the exceptional short-term temperature reconstruction are supported parts of the event’s physical context, not proof that climatic warming triggered the August 26 collapse.

References:

¹ Silwal, G., Davies, B., Carr, J. R., King, O., Buzzard, S., Carrivick, J. L., & Baral, P. (2026). Glacier changes in Langtang Catchment, Central Nepalese Himalaya from the Little Ice Age (∼1815 CE) to 2023 CE. Global and Planetary Change, 264, Article 105555. https://doi.org/10.1016/j.gloplacha.2026.105555

² Liu, Q., Khatri, A., Lord, A., Ghouli, M., Thapa, N., Steiner, J., Immerzeel, W. W., & Allen, S. (2026). Rapid Hazard Assessment Report No. NP3: Lirung massif, Lendhe Catchment, Upper Trishuli Watershed. HiRISK. https://hirisk.org/wp-content/uploads/2026/08/RHA_NP3_RasuwaFlood-1.pdf