At 10:52 a.m. on 26 August 2026, a large mass of rock and ice failed in the high-altitude, snow- and ice-covered mountains close to the Nepal-China border. The rock-ice avalanche descended rapidly into the Lhende Khola valley floor where the combination of meltwater, intense erosion and sediment entrainment launched a massive debris flow that transitioned downstream to a dense, sediment-laden flood along the Lhende Khola and Trishuli rivers. The flood caused severe damage and major loss of life around the town of Gyirong Port and Nepal’s Rasuwa region.

Strong seismicity—equivalent to an M5.2 earthquake—was recorded during the event and has been attributed to the rock-ice avalanche itself.

1. Source-Area identification & propagation

Comparison of pre- and post-event satellite imagery indicates that the potential source area was a steep rock–ice slope in Nepal at an elevation of 5190 metres above sea level (masl). The town of Gyirong Port lies at about 1825 masl. This extreme relief, together with the confined channel geometry, provided the topographic conditions for rapid motion and long-distance propagation. Hydrological data from the International Centre for Integrated Mountain Development indicate that the water level at Galchhi station on the Trishuli River rose by approximately 9 m within 30 minutes, illustrating the exceptional magnitude and abruptness of the flood wave.

Pre-event imagery shows that the source was located within a high-altitude snow- and ice-covered zone, whereas the narrow valley below was bordered by exposed bedrock and abundant glacial deposits. Post-event imagery acquired on 26 August clearly reveals newly disturbed terrain near the source and a continuous runout path extending downstream (Fig. 1). The combined topographic, remote-sensing and seismic evidence suggests that the event was initiated by the failure of a high-elevation rock–ice mass.

The same catchment experienced a previous major hazard cascade on 8 July 2025. That event originated from the outburst of a supraglacial lake on the Purepu Glacier and subsequently developed into a flood and debris flow that destroyed the China–Nepal Friendship Bridge. Although the two events had different source areas and initiation mechanisms, they followed similar downstream pathways and both affected Gyirong Port and Rasuwa.

Figure 1. Overview of the rock–ice avalanche–debris-flow hazard cascade on the China–Nepal border (sourced from Chinese satellite imagery).

2. Seismic evidence for event initiation & process evolution

A seismic station located > 50 km from the potential source recorded the complete evolution of the hazard cascade. Continuous seismic monitoring provides second-level temporal resolution and all-weather observation, offering important constraints on the initiation time, process transitions and duration of the event.

Precise identification of the initiation time

At 10:52:12 a.m. (Beijing time), the recorded waveform increased abruptly from the background level, producing a clear signal onset. This indicates the sudden initiation of rapid, large-volume mass movement in the source area. An intelligent seismic-signal recognition model developed by our laboratory simultaneously returned a high probability of a geohazard event (Fig. 2).

Main energy-release stage of the rock–ice avalanche

The model identified an initial rock–ice avalanche stage lasting approximately 88 seconds. This stage was characterized by an abrupt onset, large amplitudes, concentrated broadband energy and prominent impulsive components, consistent with the failure, rapid descent and strong impact of the rock–ice mass.

Debris-flow & flooding

Following the main impulse, the seismic amplitude decreased markedly but did not immediately return to background levels. The time–frequency signature changed from a concentrated broadband pulse to a sustained, gradually decaying signal, indicating a transition from rock–ice avalanche motion to debris-flow activity.

After entering the channel, the fragmented material continued to move downstream, eroding the channel bed and entraining glacial deposits before transforming into a debris flow and sediment-laden flood. The associated seismic energy gradually decayed, with the different frequency bands returning to near-background levels approximately 2–4 hours after the event.

Figure 2. Seismic waveform, time–frequency evolution and intelligent classification of the hazard cascade. The upper panel shows the continuous seismic waveform and the principal event phases; the middle panel presents the time–frequency energy distribution; and the lower panel shows the probabilities of geohazard occurrence and process type predicted by an in-house model (at SKLGP)

3. Pre-event deformation of the source area

To reconstruct the pre-event activity of the potential source area, we conducted a preliminary SBAS-InSAR analysis using descending-orbit Sentinel-1 imagery acquired between 1 May and 24 August 2026.

The results reveal a spatially concentrated and relatively coherent line-of-sight deformation anomaly near the potential source. Maximum cumulative displacement at representative points approached 20 mm and appeared to accelerate after 12 August. The anomalous area corresponds closely with the source identified from post-event optical imagery, suggesting that the slope may have been undergoing localized deformation before failure.

Figure 3. SBAS-InSAR line-of-sight deformation velocity and displacement time series for representative points in the source area. The main panel shows the line-of-sight deformation velocity field derived from descending-orbit Sentinel-1 data. The inset presents the cumulative line-of-sight displacement at representative points P1–P3.

This text was written by GFÚ researcher, Dr John Jansen, based on the original version published on 28 August 2026 (in Chinese: https://mp.weixin.qq.com/s/hDtnRIKuuouzpZXErfcJbQ) by Dr Xin Wang, National Key Laboratory of Geohazard Prevention, Chengdu University of Technology, China.

Dr Jansen is currently a Distinguished Visiting Professor at SKLGP.