Marc Hendrickx and Associates Pty Ltd

This report presents a preliminary model for the evolution of the Langtang/Tsangbu rock avalanche and debris flow event. This intuitive model is intended to stimulate further analysis and discussion of this extraordinary and tragic event. Having spent some time looking over the disaster I thought I would condense my analysis. Our thoughts are with the 1000s of people affected.

The analysis below benefits from much public discussion of the event by other experts. In particular, I point to early analysis by Dan Shugar, Shawn Willsey and Philip Prince. Any errors below belong to me.

Initiating as a progressive creep failure throughout the first eight months of 2026 (or perhaps earlier), the mass experienced a catastrophic release on August 26, 2026. The model presented covers the first 22km of the event from source to the Border station at Rasuwagadhi. Over this distance the event transformed from a solid rock-ice slope collapse – sturzstrom – into an ultra-fast, highly bulked, hyper-mobile debris flow. Five phases are defined (Figure 1).

Figure 1. Google earth images showing location of landslide phases discussed in text. Phase 0 in source area.

Phase 0: Pre-Failure Progressive Creep and Collapse

The disaster is inferred to have initiated early in 2026 (or earlier) through progressive structural degradation along a 45° discontinuous defect (joint or fault) in the bedrock gneiss, leading to total collapse on August 26, 2026. Several stages of collapse are expected following generalized models for large rock failures (eg https://doi.org/10.1139/t02-085).

The Frictional Steady-State (January – May): Instability initiated in or prior to January 2026. Google images covering the bedrock from January 7, 2026 show linear features along bedrock defects suggestive of tension cracks (Figure 2).  These may be simply due to snow accumulation on linear bedrock features or snow depressions caused by movement on recently formed tension cracks. Further analysis is required to establish whether they are early indicators of movement. It is possible movement was initiated earlier.

Figure 2. Google earth imagery 7/1/2026 and 10/10/2017. Linear features following bedrock defects are suggestive of presence of tension cracks.

Under a generalized model during this early Pre-Monsoon period, the rock mass remained in a secondary creep phase, perhaps moving at a stable rate of ~1.5 mm/day. Sub-surface ice within the bedrock joints acted as a structural binding agent, keeping the mass in a temporary state of equilibrium.

• The Onset of Acceleration (May 1 – June 1): As temperatures rose during the late Pre-Monsoon spring thaw, meltwater began infiltrating deep structural fissures, elevating internal pore-fluid pressures. The rock mass likely broke its linear displacement trend in early May. Intact rock bridges supporting the mass began failing in shear, driving creep velocity to ~4 mm/day by early June.
• The Monsoonal Hyper-Acceleration (June 1 – August 25): The arrival of the continuous monsoon on June 1 introduced an unbroken, multi-month window of precipitation. Water infiltration flooded now open tension cracks, driving pore pressures to critical levels and drastically reducing effective friction along the main 45° failure plane. Displacement likely accelerated over this period with indicative values in order of 10 mm/day by July 1 and 20 mm/day by earlyAugust. By late August, the slope transitioned into an uncontrollable asymptotic tertiary creep runaway, with displacement velocity accelerating exponentially to over 400 mm/day by August 25.
• Catastrophic Release: On August 26, 2026, the remaining structural rock bridges experienced total failure, instantly transitioning the static slope into a dynamic 132 million cubic metre sturzstrom.
•  
• A graphical representation of suggested movement phases is shown in Figure 3.

Figure 3 Generalised displacement – time model for large rock slope failure applied to Langtang/Tsangbu rock avalanche. Acceleration driven by monsoonal rains and increased summer temperatures. AI generated.

Phase 1: Catastrophic Collapse, Seismogenesis, & Rebound

The sudden release of the 132 million cubic metre mass—comprised of 87% gneiss rock and 13% ice by volume—plunged 1,200 vertical metres down the steep slope, releasing 3.84 × 10¹⁵ Joules (3.84 Petajoules) of gravitational potential energy (Figure 4 and 5). The failure has a general translational/ wedge style mechanic bound by structural defects. The area under the glacial appears to have failed at an earlier stage – perhaps also as a large rock slide, or incrementally in smaller sections. The area of rock appears to taper to the north, under the upper glacier.

Figure 4. Satellite imagery pre and post event and google earth imagery – backscarp to failure indicated by red line.

Figure 5 Cross sections and volume estimate for Langtang/Tsangbu rock avalanche – initial ice content put at about13%

• Impact Seismicity & Flash Melting: Upon striking the hard rock valley floor, the immense impact generated a magnitude 5.2 earthquake. Accounting for a seismic efficiency of 0.1% (3.98 × 10¹² J radiated as seismic waves), the remaining 99.9% of the total potential energy converted instantly into frictional heat and mechanical rock crushing. This thermal energy flash-melted 11.5 billion kilograms (73.2%) of the internal ice into liquid water.
• Opposite Slope Run-Up: Driven by severe forward momentum, the mass surged 500 metres up the opposite 45° valley wall and slid back down. This U-turn generated an additional 3.39 × 10¹⁴ to 6.78 × 10¹⁴ Joules of frictional work. The shearing action liquefied another 1.02 to 2.03 billion kilograms of ice. By the time the avalanche returned to the valley floor for the second time, 86% of its original ice assets had converted into liquid water.

Phase 2: Glacial Chute Entrainment & Acceleration

Rebounding off the opposite wall, the avalanche entered a 2-kilometre-long remnant glacier chute with an estimated average thickness of 15 metres with approximate slope of 25°.

• Mass Bulking: Assuming a representative flow path width of 500 metres, the massive shear forces at the base of the flow completely scoured and incorporated the glacier, entraining an additional 15 million cubic metres (1.38 × 10¹⁰ kg) of ice into the landslide matrix.
• Chute Acceleration: The vertical drop down this 25° glacial chute released an additional 2.76 × 10¹⁵ Joules of potential energy. Intense rock-on-rock grinding and fluid turbulence melted up to 8.26 billion kilograms of ice entrained by the lower glacier.
• Velocity Metrics: Highly lubricated by the basal melted ice slush, the flow rapidly accelerated to a terminal velocity ranging between 75 and 125 metres per second (270 to 450 km/h). At the exit of the 2 km chute, the total moving volume stood at 147 million cubic metres, with liquid water comprising 9.8% to 14.3% of the flow by volume.

Phase 3: Intermediate Valley Flattening & Complete Liquefaction

The flow exited the steep glacier chute and entered a 4-kilometre-long intermediate valley section characterized by a constrained width of 300 metres and a gentler 10% average slope.

• Deceleration Profile: The sharp drop in gradient and high boundary resistance of the 300-metre wide channel slow the flowing mass. The flow shed its excessive entry velocity, settling into a stable, slope-dictated terminal velocity of 25 to 40 metres per second (90 to 144 km/h).
• Phase Transition to Debris Flow: The continuous grinding down this 4 km stretch released enough frictional heat to flash-melt 100% of the remaining ice within the matrix. Total liquid water rose to 29.5 billion kilograms. This completed the mechanical transition of the disaster from a rock avalanche into a fluidised debris flow with a water content of 8.7% by mass and 20.4% by volume.

Phase 4: Confined River Canyon & Intense Soil Scouring

In this section, the debris flow entered the Lhende River and surged into a steeply confined valley for a distance of 14 kilometres on an average slope of about 12%.

• Hydro-Dynamic Interaction: Assuming the canyon contained an active river flowing at 20 m³/s. Over the brief duration of the surge event, the river contributed roughly 150,000 metric tonnes of water—a negligible addition to the existing 29.5 billion kilograms of water, resulting in no meaningful fluid dilution.
• Re-Acceleration: The narrow canyon walls prevented lateral spreading, causing the flow depth to swell to an estimated 25–35 metres. This severe vertical bulking reduced relative basal drag, allowing the flow to re-accelerate to a terminal speed of 30 to 45 metres per second (110 to 160 km/h) as it passed the border station.
• Debris Erosion & Final Fluid Composition: The high shear stress of the deep flow aggressively scoured the canyon floor, entraining an estimated 2,000 cubic metres of soil and rock debris for every metre travelled. This process added 28 million cubic metres (75.6 × 10⁹ kg) of material to the avalanche. Accounting for a 5% ambient water content by mass within the scoured valley soil, an additional 3.78 billion kilograms of liquid water was injected into the flow alongside the dry solids.

Summary of System State at Rasuwanadhi Border crossing point

Parameter	Metric Value	Analysis & Physical Significance
Total Moving Volume	175 million m³	A 32.5% increase from the initial collapse volume via intense glacier and valley floor erosion.
Total System Mass	4.16 × 10¹¹ kg	416 million metric tonnes of moving rock, scoured soil, and fluid.
Total Liquid Water Mass	33.43 × 10⁹ kg	Derived from the complete melting of all initial and entrained ice, river volume, and soil moisture.
Exit Terminal Velocity	30 to 45 m/s (110–160 km/h)	High-velocity fluid motion capable of total structural destruction upon impact.
Final Water Fraction (Mass)	8.0%	The mass proportion of liquid phase to solid phase across the bulk slurry.
Final Water Fraction (Volume)	18.7%	Nearly one-fifth of the total volume exists as a liquid matrix, maintaining fluidization.

Discussion

Water volume: The model presented above accounts somewhat for potential water sources and required volume to sustain a debris flow. The initial estimate of ice volume of the upper glacier does not appear sufficient to sustain a debris flow on its own, and without addition of water from the lower glacier it is conceivable that the event could have remained dominantly a “dry” rock avalanche and been stopped in the upper reaches of the Lhende River creating a massive landslide dam.

Runout for very large volume, dry landslides is proportional to the square root of the starting volume.  Based on a detached volume of 132Mm³  the runout would be expected to be about 11km on a flat surface. If it did not turn into a debris flow and remained mainly a rock avalanche this dam would have been huge and would have threatened development downstream.

With the addition of water from melting of the lower glacier total water volume available shifts to the 20% range and a debris flow can be sustained. It is interesting to note that water in the river did not add that much to the overall water content.

More detailed analysis is required to more accurately assess the rock and ice volume involved in the event.

Transition from Sturzstrom to debris flow: There appears to be sufficient kinetic energy in the system to provide for complete melting of any entrained ice prior to the flow reaching the Lhende River. The transition from rock – ice avalanche to debris flow likely occurred in the valley below the lower glacier.

Comparison to other large landslides: The Chamoli Landslide (https://science.sciencemag.org/content/early/2021/06/09/science.abh4455) in India in Feb 2021 shares many features with the Langtang/Tsangbu rock avalanche and debris flow event. Chamoli was 27Mm³ in volume (~5x less than Nepal based on estimates above) with a debris flow run out of over 17km. Chamoli had a starting rock/ice ratio with 20% ice and kinetic energy turned this to water in the initial fall perhaps providing for more rapid transiton to a debris flow. Both highly destructive events. There were indications of precursor movement at Chamoli.

Any warning?: Large rock falls of this nature do not generally occur without precursor movements. It appears that if movement had been detected much earlier in 2026, the area could have been monitored in more detail, similar to events preceding the Blatten Landslide in Switzerland in May, 2025. Follow-up on potential cracks suggested in Imagery from early January is required to determine when movement initiated.

In my view it would have been incredibly difficult to accurately predict the behaviour of the Langtang/Tsangbu rock avalanche event. Early warnings allowing people to get out to higher ground would have been substantially reduced the toll. There were some seismic signals detected prior to 5.2M seismic event that marked the collapse. Recognition of these (given they are connected to the event) may have provided more warning time downstream. It appears that a wider network of seismic stations in the region may assist in earlier recognition of these events. There is a risk of false positives that could have a huge disruptive effect. This requires much more detailed investigation.

The Himalayas are vast and monitoring rock slopes represents an immense challenge. Remote monitoring of rock slopes via InSAR or other satellite technology to catch events like this is feasible but would require substantial resources to setup. For the Himalayas a cooperative centre for “mountain movement” could be established with resources from China, India, Pakistan and Nepal and other partners. The process of systematic comparisons between datasets to detect slope movement is perhaps something an AI could be trained to do, to speed up the process – flagging issues for more detailed expert analysis and potential follow-up. Or local higher-risk areas could be flagged for detailed satellite monitoring to reduce the scale of the task. This would also help identify other hazards (eg GLOF).  This requires substantial further investigation but would certainly help reduce the human toll of such events.

Future development in the high Himalayas needs to take these events into account. There is already too much infrastructure exposed.

This event represents an immense human tragedy. We need to learn from it to reduce the impact of future events like this.