# How Landslides Work

> On 26 August 2026 a mass of rock and glacier ice fell from Langtang Lirung in the Himalaya and swept down valleys in Nepal and Tibet; scientists from World Weather Attribution called it a compound crisis of warming, thawing ice and unstable rock. This film explains the established science of landslides behind that news: the balance of gravity against friction and cohesion on a slope, the factor of safety, why water in the ground is the commonest trigger, the main kinds of mass movement from rockfalls to debris flows and creep, the triggers from rain and earthquakes to volcanoes and people, landslide dams and the floods they release, the Vajont disaster of 1963, and how places like Hong Kong learned to watch their slopes.

Interactive film: https://www.planetpast.com/stories/how-landslides-work/  
Publisher: PlanetPast (https://www.planetpast.com/) · d => T('readPublished', l, { d: dateIn(d, l) }) 2026-11-04 · Earth science, Physics, Society

## Prologue

### When the mountain lets go
*Earth science*

How landslides work, what sets them off, and how we watch for them.

On the morning of 26 August 2026, a huge mass of rock and glacier ice broke off Langtang Lirung, a Himalayan peak near Nepal's border with Tibet. It became a torrent of water, ice and boulders that wrecked villages on both sides of the border. By mid-September more than fourteen hundred people were confirmed dead, and thousands were still missing. Six weeks later, another landslide blocked the Budhi Gandaki River in Gorkha, then burst through. Scientists called the first disaster a compound crisis. To see why, we need the physics of a slope.

![Mud, ice blocks and boulders fill a valley floor below a scarred peak. AI-generated still.](https://assets.planetpast.com/a/ae/ae3ee1431f5376e7.jpg)

- **1,446**: dead in Nepal and China
- **6,669+**: still missing

![A lake backs up behind fresh landslide debris in a river gorge. AI-generated still.](https://assets.planetpast.com/a/82/82a4a7616029ca6b.jpg)

- Why a slope stands, and when it cannot
- Falls, slides, flows and creep
- Rain, quakes, volcanoes, people, ice
- How warnings work

Prologue

#### Nepal’s year of falling mountains

On 26 August 2026 a rock-ice avalanche from the Lirung massif became a fast, long-runout debris flow that destroyed the Gyirong border crossing and villages in Nepal and Tibet. Al Jazeera, reporting a World Weather Attribution study on 17 September, gave 1,446 dead and at least 6,669 missing in the two countries. On 11 October a landslide dammed the Budhi Gandaki River in Gorkha district; the dam broke at about 6 am and a debris flood ran downstream (Dave Petley, The Landslide Blog).

Casualty figures changed for weeks and differ between sources; those shown are attributed with their date. The rest of the film is established landslide science.

The 3D scenes are simplified teaching models, not to scale.

Sources: [Al Jazeera (17 Sep 2026): Climate change had major role in triggering Nepal floods, scientists say (World Weather Attribution)](https://www.aljazeera.com/news/2026/9/17/climate-change-had-major-role-in-triggering-nepal-floods-scientists-say); [Wikipedia: 2026 Nepal–Tibet floods (Langtang Lirung rock-ice avalanche, 26 August 2026)](https://en.wikipedia.org/wiki/2026_Nepal%E2%80%93Tibet_floods); [Dave Petley, The Landslide Blog, Eos/AGU (12 Oct 2026): the 11 October 2026 landslide and debris flow on the Budhi Gandaki River, Gorkha, Nepal](https://eos.org/thelandslideblog/the-11-october-2026-landslide-and-debris-flow-flood-on-the-budhi-gandaki-river-in-gorkha-district-nepal)

### 

Geologists use the word landslide for any movement of rock, soil or debris down a slope. It can fall, topple, slide, spread or flow. Landslides happen wherever there are slopes. One study counted 4,862 fatal landslides between 2004 and 2016 that were not caused by earthquakes. They killed 55,997 people. About three quarters were in Asia, and the number set off by people was rising.

![Torn turf and roots at the top edge, where the ground broke away. AI-generated still.](https://assets.planetpast.com/a/4c/4c7e512ffb727e68.jpg)

- **4,862**: fatal landslides, 2004–2016, earthquakes excluded
- **55,997**: deaths

- **~75%**: of fatal landslides were in Asia, 2004–2016

![Homes, roads and slopes share the same steep ground. AI-generated still.](https://assets.planetpast.com/a/f1/f175e1a950e326ad.jpg)

Prologue

#### What counts as a landslide

The USGS defines a landslide as the downslope movement of rock, soil or artificial fill; the material may fall, topple, slide, spread or flow. Froude and Petley (2018) compiled 4,862 distinct fatal non-seismic landslide events from January 2004 to December 2016, which killed 55,997 people. Asia had about 75% of them, and the authors found landslides triggered by human activity, mainly construction, illegal mining and hill cutting, were increasing.

Sources: [USGS Fact Sheet 2004-3072: Landslide Types and Processes](https://pubs.usgs.gov/fs/2004/3072/fs-2004-3072.html); [Froude & Petley (2018), Global fatal landslide occurrence from 2004 to 2016, Natural Hazards and Earth System Sciences](https://nhess.copernicus.org/articles/18/2161/2018/nhess-18-2161-2018.html)

## Chapter I · Gravity against grip

### 

Picture a block of soil on a hillside. Gravity pulls it straight down, but on a slope that pull splits in two. One part presses the block into the hill. The other drags it downhill. The steeper the slope, the bigger the downhill part. Tilt the slope yourself, and watch the two arrows trade places.

$$\text{downhill} = W\sin\theta$$
$$\text{into the slope} = W\cos\theta$$
- $W$: the block’s weight
- $\theta$: slope angle

![On the steepest stretch of the hillside the soil has already given way. AI-generated still.](https://assets.planetpast.com/a/dd/dd5976a9eec50bfb.jpg)

- 10°: 17%
- 20°: 34%
- 35°: 57%
- 50°: 77%

Chapter I · Gravity against grip

#### Two parts of one pull

On a plane tilted at angle θ, the weight W of a block resolves into W sin θ along the slope, which drives it downhill, and W cos θ at right angles to the slope, which presses it in. At 10° about 17% of the weight pulls downhill; at 50° about 77%. The figures are sin θ.

The 3D scenes are simplified teaching models, not to scale.

Sources: [Portland State University, Earth Surface Processes: slope stability and the factor of safety](https://pdx.pressbooks.pub/earthsurface/?p=27)

### 

Holding the block back are two kinds of grip. Friction grows the harder the block is pressed into the slope. Cohesion is the glue between grains, from clay, roots or natural cement. Engineers compare grip and pull in one number, the factor of safety: the grip resisting movement, divided by the pull driving it. Above one, the slope stands. At one, it is on the edge. Below one, it must move. Dry sand has no glue at all. Its grip comes only from grains locking together, and that sets a limit on how steep it can stand, called its friction angle. So a pile of dry sand settles at about the same steepness each time.

$$FS = \frac{\text{strength resisting}}{\text{stress driving}}$$
- $FS>1$: stable
- $FS<1$: failing

- **> 1**: stands
- **= 1**: on the edge
- **< 1**: moves

![Grains trickle down the cone’s sides until the slope stops steepening. AI-generated still.](https://assets.planetpast.com/a/1a/1a11c99670b04479.jpg)

Chapter I · Gravity against grip

#### The factor of safety

The factor of safety is the ratio of the strength available to resist failure to the stress driving it. F > 1 is stable, F = 1 is the threshold and F < 1 means the slope should be failing. Strength has two parts in the usual Mohr–Coulomb description: cohesion, the strength present with no normal stress, and friction, which adds strength in proportion to the effective normal stress through tan φ, where φ is the friction angle. For a long uniform slope with no cohesion and no water, the same equations give F = tan φ / tan θ, so the slope is on the edge when its angle equals φ; the steepness at which a pile of dry sand comes to rest is close to this, varying with the size and shape of the grains.

The 3D scenes are simplified teaching models, not to scale.

Sources: [Portland State University, Earth Surface Processes: slope stability and the factor of safety](https://pdx.pressbooks.pub/earthsurface/?p=27)

### 

Now add water. Rain soaks in and fills the spaces between the grains, and that water pushes outward on them. This pore pressure props the grains apart, so they press on each other less. Engineers call what is left the effective stress, and friction depends on it. The slope weighs the same, but its grip has weakened. The US Geological Survey calls saturation a primary cause of landslides. Press rain, and watch the factor of safety sink.

![Water seeps from between grains and pebbles in a dark, wet soil face. AI-generated still.](https://assets.planetpast.com/a/71/716353f1fd3a4350.jpg)

$$\sigma' = \sigma - u$$
- $\sigma$: stress pressing grains together
- $u$: pore water pressure
- $\sigma'$: what friction actually feels

- **1.31**: factor of safety, dry
- **0.71**: factor of safety, soaked

![Rills of runoff cut across a stripped hillside. AI-generated still.](https://assets.planetpast.com/a/5b/5b3aa5a6dc6d5925.jpg)

Chapter I · Gravity against grip

#### Why water is the commonest trigger

Effective normal stress is the normal stress minus the pore water pressure. When the ground is saturated, pore pressure becomes positive and reduces the effective stress, so the frictional part of the strength falls. Water does not change the cohesion or the friction angle themselves; it acts through the pore pressure. The USGS lists saturation by water as a primary cause of landslides, from intense rain, snowmelt and changes in groundwater or in water levels along coasts, dams and riverbanks.

The 3D scenes are simplified teaching models, not to scale.

Sources: [Portland State University, Earth Surface Processes: slope stability and the factor of safety](https://pdx.pressbooks.pub/earthsurface/?p=27); [USGS Fact Sheet 2004-3072: Landslide Types and Processes](https://pubs.usgs.gov/fs/2004/3072/fs-2004-3072.html)

## Chapter II · Five ways to fall

### 

When a slope fails, how it moves depends on what it is made of, how wet it is and how steep. Geologists sort landslides by the motion. In a fall, rock breaks from a cliff and drops through the air, bouncing and rolling. In a topple, a column tips forward over a pivot near its base. Over the years the fallen blocks pile up as a fan of rubble at the foot of the cliff.

- Falls and topples
- Slides, rotational or translational
- Flows: debris, mud, earth
- Creep and lateral spreads

![Blocks break from a limestone face in a haze of dust. AI-generated still.](https://assets.planetpast.com/a/de/de262a6ffcc0be8d.jpg)

![A jumble of broken rock fills a gorge floor. AI-generated still.](https://assets.planetpast.com/a/11/1197893e12a81b03.jpg)

Chapter II · Five ways to fall

#### Falls and topples

In the USGS classification, falls are abrupt movements of rock or boulders that detach from steep slopes or cliffs and move by free fall, bouncing and rolling. Topples are the forward rotation of a unit about a pivot point below or low in the unit.

The 3D scenes are simplified teaching models, not to scale.

Sources: [USGS Fact Sheet 2004-3072: Landslide Types and Processes](https://pubs.usgs.gov/fs/2004/3072/fs-2004-3072.html)

### 

In a slide, a mass moves along a surface of weakness beneath it. If that surface is curved like the bowl of a spoon, the block rotates as it sinks, and leaves a crescent-shaped scarp at the top. That is a rotational slide, or slump. If the surface is a flat layer, such as a bed of clay between rocks, the mass glides along it with little turning. That is a translational slide.

![The block tilts back as it drops, so its top ends up sloping into the hill. AI-generated still.](https://assets.planetpast.com/a/4c/4c8874d175be12a3.jpg)

- A curved headscarp
- Cracks across the slope
- Bulging ground at the toe

![Layers dip toward the valley, the kind of surface a slab can glide on. AI-generated still.](https://assets.planetpast.com/a/14/1466c7e5b5db264b.jpg)

Chapter II · Five ways to fall

#### Rotational and translational slides

A slide has a distinct zone of weakness that separates the moving material from the more stable material below. In a rotational slide the surface of rupture is curved concavely upward and the movement is roughly rotational about an axis parallel to the slope. In a translational slide the mass moves out or down along a roughly planar surface with little rotation; a block slide is a translational slide of one or a few coherent units.

The 3D scenes are simplified teaching models, not to scale.

Sources: [USGS Fact Sheet 2004-3072: Landslide Types and Processes](https://pubs.usgs.gov/fs/2004/3072/fs-2004-3072.html)

### 

Add enough water, and the moving mass stops behaving like a solid and starts to flow. A debris flow is a fast slurry of soil, rock, water and air that pours down channels. With more fine silt and clay, it is a mudflow, what the news often calls a mudslide. At the other extreme is creep: soil moving too slowly to see, except in fence posts and tree trunks bent downhill over the years.

![Boulders and broken trees ride a brown surge down a gully. AI-generated still.](https://assets.planetpast.com/a/5e/5ed7a27cf0b8ecc3.jpg)

![Flow ridges in glossy mud that has spread across a valley floor. AI-generated still.](https://assets.planetpast.com/a/10/10905cfab3f3d8e1.jpg)

![Posts lean downhill and trunks curve at their base. AI-generated still.](https://assets.planetpast.com/a/5e/5e35d32ad4139286.jpg)

- **< 50%**: fine grains in a debris flow
- **≥ 50%**: sand, silt and clay in a mudflow

Chapter II · Five ways to fall

#### Flows, creep and spreads

The USGS describes a debris flow as a rapid mass movement in which loose soil, rock, organic matter, air and water form a slurry; less than half of it is fine material. A debris avalanche is a very rapid to extremely rapid debris flow. A mudflow is an earthflow wet enough to flow rapidly, with at least 50% sand, silt and clay; news reports often call these, and debris flows, mudslides. Creep is imperceptibly slow, steady downslope movement, shown by curved tree trunks and tilted poles or fences. Lateral spreads occur on very gentle slopes or flat ground, usually when saturated loose sediments liquefy, often during earthquakes.

The 3D scenes are simplified teaching models, not to scale.

Sources: [USGS Fact Sheet 2004-3072: Landslide Types and Processes](https://pubs.usgs.gov/fs/2004/3072/fs-2004-3072.html)

## Chapter III · What tips the balance

### 

Water is the commonest trigger. In March 2014, after a February and March with up to twice the usual rain, a hillside of sand and glacial clay above Oso, in Washington State, gave way. Its debris averaged about sixty-five kilometres an hour, crossed a valley about 800 metres wide and killed 43 people. As it moved, the soaked soil liquefied. Ground that has been wet for months needs only one last storm.

- **150–200%**: of normal rain in Feb–Mar
- **~19 million US tons**: of sand and glacial till

![Runoff streams over ground that can hold no more water. AI-generated still.](https://assets.planetpast.com/a/5b/5b3aa5a6dc6d5925.jpg)

![Grey sand and clay spread across a forested valley and pond the river. AI-generated still.](https://assets.planetpast.com/a/d0/d0b15c7ae21c5046.jpg)

- **2**: stages of failure, in about a minute
- **500–6,000**: years ago, earlier slides of the same kind

Chapter III · What tips the balance

#### Oso, 2014

On 22 March 2014 about 19 million tons of sand and glacial till failed above the Steelhead Haven community near Oso, Washington, killing 43 people. February and March rainfall was 150–200% of the long-term average, with the heaviest rain late in the wet season when the ground was nearly saturated. The USGS estimates the debris averaged about 40 mph; it failed in two stages over about a minute, the soil liquefied in the second, and it crossed a valley about half a mile wide. Similar slides happened there 500 to 6,000 years ago.

The 3D scenes are simplified teaching models, not to scale.

Sources: [USGS: Five years later, the Oso SR 530 landslide, Washington](https://www.usgs.gov/news/featured-story/five-years-later-oso-sr-530-landslide-washington)

### 

Earthquakes shake slopes loose. In Nepal's 2015 earthquake, landslides buried the village of Langtang, in the same valley as Langtang Lirung, and killed more than three hundred and fifty people. Volcanoes can bring down whole mountainsides. On 18 May 1980, the north side of Mount St. Helens collapsed: about two and a half cubic kilometres of rock, the largest landslide recorded in historic time. It raced more than twenty kilometres down the valley, at well over two hundred kilometres an hour.

![A fan of boulders and ice fills the floor between steep snowy walls. AI-generated still.](https://assets.planetpast.com/a/0c/0c1745ee5547355a.jpg)

- **350+**: killed by earthquake-triggered landslides

![A horseshoe-shaped crater opens toward a hummocky field of debris. AI-generated still.](https://assets.planetpast.com/a/3a/3a4d82090abb04f9.jpg)

- Mount St. Helens, 1980: ~2.5
- Langtang Lirung, 2026: 0.2–0.5
- Vajont, 1963: ~0.26

- **~2.5 km³**: of rock
- **~24 km**: down the North Fork Toutle
- **>240 km/h**: top speed

Chapter III · What tips the balance

#### Earthquakes and volcanoes

Shaking can trigger slides directly and also loosens soil so that water infiltrates faster; the USGS notes that earthquakes in steep landslide-prone areas greatly increase the likelihood of landslides. In 2015, earthquake-triggered landsliding at Langtang killed more than 350 people. At Mount St. Helens on 18 May 1980, the USGS describes the rockslide-debris avalanche, about 2.5 km³ by one estimate, as the largest landslide-debris avalanche recorded in historic time; it travelled about 24 km down the North Fork Toutle River at more than 150 mph (about 240 km/h).

The 3D scenes are simplified teaching models, not to scale.

Sources: [USGS Fact Sheet 2004-3072: Landslide Types and Processes](https://pubs.usgs.gov/fs/2004/3072/fs-2004-3072.html); [Wikipedia: 2026 Nepal–Tibet floods (Langtang Lirung rock-ice avalanche, 26 August 2026)](https://en.wikipedia.org/wiki/2026_Nepal%E2%80%93Tibet_floods); [USGS Cascades Volcano Observatory: the 18 May 1980 debris avalanche at Mount St. Helens](https://webarchive.library.unt.edu/eot2008/20090115210113/http://vulcan.wr.usgs.gov/Volcanoes/MSH/DebrisAval/description_msh_debris_aval.html)

### 

People tip the balance too: by cutting into the foot of a slope for a road, loading its top with buildings, or clearing the trees whose roots bind the soil. In November 1988, floods and landslides struck villages in Nakhon Si Thammarat, in southern Thailand. After those floods, Thailand banned logging. Worldwide, landslides set off by construction, mining and hill cutting are becoming more common.

![Soil and stones spill onto a mountain road below a steep cut. AI-generated still.](https://assets.planetpast.com/a/8a/8af705e2b903f469.jpg)

- 22 Nov 1988: Kathun, Phipun district, Nakhon Si Thammarat
- after 1988: logging banned after the floods
- 11 Aug 2001: Nam Ko, Lom Sak, Phetchabun: 136 dead

- Excavating the toe of a slope
- Loading the crest
- Clearing forest
- Leaking pipes, irrigation, mining

Chapter III · What tips the balance

#### Human triggers

The USGS lists human causes including excavation of a slope or its toe, loading of the slope or its crest, drawdown of reservoirs, deforestation, irrigation, mining, artificial vibration and leaking water services. Froude and Petley (2018) found landslides triggered by construction, illegal mining and hill cutting increasing worldwide. In Thailand, landslides struck Ban Kathun, Phipun district, Nakhon Si Thammarat, on 22 November 1988; after the massive 1988 flooding the government banned logging. On 11 August 2001 a landslide at Nam Ko, Lom Sak, Phetchabun, killed 136.

The 3D scenes are simplified teaching models, not to scale.

Sources: [USGS Fact Sheet 2004-3072: Landslide Types and Processes](https://pubs.usgs.gov/fs/2004/3072/fs-2004-3072.html); [Froude & Petley (2018), Global fatal landslide occurrence from 2004 to 2016, Natural Hazards and Earth System Sciences](https://nhess.copernicus.org/articles/18/2161/2018/nhess-18-2161-2018.html); [Wikipedia: List of disasters in Thailand (Kathun 1988, Nam Ko 2001)](https://en.wikipedia.org/wiki/List_of_disasters_in_Thailand); [PBS Commanding Heights: Thailand, environment (logging ban after the 1988 floods)](https://www.pbs.org/wgbh/commandingheights/lo/countries/th/th_env.html)

## Chapter IV · A river blocked

### 

When a landslide falls into a narrow valley, it can block the river with a dam of loose rubble. Water piles up behind it. A dam that was never compacted often leaks or overflows, and it can fail all at once. That is what happened on the Budhi Gandaki on 11 October 2026. A landslide dammed the river at around four in the morning. Around six the dam broke, and a debris flood swept downstream. Warnings reached people downstream, but about five people are thought to have died. Press breach to watch the lake empty.

![Muddy water backs up behind a ridge of fresh debris. AI-generated still.](https://assets.planetpast.com/a/82/82a4a7616029ca6b.jpg)

- **~7.6 m**: deep where the river was blocked (25 ft)
- **~4 km**: long lake (2.5 miles)

- ~04:22: seismic signal of the landslide
- ~06:00: the landslide dam breaches
- after: houses, bridges and a hydropower camp damaged

- Houses and roads
- Bridges
- A hydropower project camp

![A brown, debris-laden surge tears past a broken footbridge. AI-generated still.](https://assets.planetpast.com/a/5a/5ab9025a77db241a.jpg)

Chapter IV · A river blocked

#### Landslide dams

The USGS notes that landslides can dam rivers, causing backwater flooding upstream and flooding downstream when the dam fails. On the Budhi Gandaki in Gorkha district, Nepal, Dave Petley reports a landslide recorded seismically at about 04:22 local time on 11 October 2026, possibly a rock avalanche near Chumchet (the timing preliminary), followed by a breach of the landslide dam at about 6 am. Houses, roads, bridges and a camp for the Budhi Gandaki hydropower project were damaged; warnings reportedly reached people, and he wrote that five people had probably been killed. At Oso in 2014 the debris dammed the North Fork Stillaguamish River up to 25 feet deep, creating a lake about 2.5 miles long that drained by mid-May.

The 3D scenes are simplified teaching models, not to scale.

Sources: [USGS Fact Sheet 2004-3072: Landslide Types and Processes](https://pubs.usgs.gov/fs/2004/3072/fs-2004-3072.html); [Dave Petley, The Landslide Blog, Eos/AGU (12 Oct 2026): the 11 October 2026 landslide and debris flow on the Budhi Gandaki River, Gorkha, Nepal](https://eos.org/thelandslideblog/the-11-october-2026-landslide-and-debris-flow-flood-on-the-budhi-gandaki-river-in-gorkha-district-nepal); [USGS: Five years later, the Oso SR 530 landslide, Washington](https://www.usgs.gov/news/featured-story/five-years-later-oso-sr-530-landslide-washington)

### 

A landslide into a lake can be just as deadly. On the night of 9 October 1963, about two hundred and sixty million cubic metres of rock slid into the reservoir behind the Vajont Dam in Italy, 262 metres tall. It pushed out a wave that rose far above the dam and poured over its crest into the valley below. The wave destroyed the town of Longarone and other villages. According to Italy's Civil Protection Department, 1,917 people died. The dam itself was left standing.

![A tall concrete arch wedged in a narrow limestone gorge. AI-generated still.](https://assets.planetpast.com/a/bb/bb5bd236d64c20f8.jpg)

![Broken rock chokes a narrow gorge between bare walls. AI-generated still.](https://assets.planetpast.com/a/11/1197893e12a81b03.jpg)

- **~260 million m³**: of rock (Britannica)
- **150–200 m**: the wave rose above the dam (Britannica)

![A wide scoured floor of gravel and new grass in the valley beneath the arch. AI-generated still.](https://assets.planetpast.com/a/bf/bf40b36a44683b9b.jpg)

- **1,917**: dead
- **400**: of them never found

Chapter IV · A river blocked

#### Vajont, 1963

Britannica gives the Vaiont (Vajont) Dam a height of 262 m and describes about 260 million m³ of rock sliding into its reservoir on 9 October 1963, sending a wave 150–200 m above the dam that spilled over the crest and flooded the valley; it gives about 2,000 dead. Italy’s Civil Protection Department gives 1,917 victims, 400 of whom were never found, and names Erto, Casso and Longarone among the places hit.

The 3D scenes are simplified teaching models, not to scale.

Sources: [Britannica: Vaiont Dam](https://www.britannica.com/topic/Vaiont-Dam); [Italian Civil Protection Department: the Vajont landslide](https://servizio-nazionale.protezionecivile.gov.it/en/pagina-base/vajont-landslide/)

## Chapter V · Ice that held the rock

### 

Now back to Langtang Lirung. High in the Himalaya, ice and frozen ground work like cement in cracked rock. By one geologist's estimate, a slab of ice and rock covering about a fifth of a square kilometre dropped around one point two kilometres. Early analyses put the flow that followed at around fifty metres a second, running up to a hundred kilometres downstream. The US Geological Survey found that the shaking recorded that morning came from the collapse itself, not from an earthquake. World Weather Attribution scientists found that warming had raised the freezing level, thawing ice and frozen ground after record summer heat. Damage from the 2015 earthquake may have played a part too. They called it a compound crisis.

![Ice cliffs cling to a steep face; dark wet rock shows where ice has melted back. AI-generated still.](https://assets.planetpast.com/a/3a/3a34fdb1fea5afa0.jpg)

- **~0.2 km²**: of ice and rock (Daniel Shugar)
- **~1.2 km**: vertical drop
- **~52 m/s**: flow speed, early analyses

![A cloud of ice and rock dust pours down a giant peak. AI-generated still.](https://assets.planetpast.com/a/74/744de66af1bed62e.jpg)

> “There is no doubt that climate change is one of the drivers.” (quoted by Al Jazeera, 17 September 2026)

- Long-term warming and thaw
- Record heat and extra meltwater
- Rock already weakened, perhaps by the 2015 quake

Chapter V · Ice that held the rock

#### A compound crisis

The collapse began as a rock-ice avalanche from the Lirung massif at 5,200–5,400 m and became a fast, long-runout debris flow; geologist Daniel Shugar estimated about 0.2 km² of ice and rock fell roughly 1.2 km vertically; other analyses cited by Wikipedia put the debris at about 52 m/s, travelling up to 100 km. The USGS found the seismic signal was generated by the collapse, not by a preceding earthquake. A World Weather Attribution report (17 September 2026) called it a compound crisis: long-term warming raised the freezing level, thawing ice and permafrost; record July–August heat about 5 °C above normal and heavy snow in late 2025 added meltwater; the 2015 magnitude 7.8 earthquake may have weakened the slope, though its role is uncertain. The region’s flood early-warning system was configured for monsoon and perennial floods and did not detect the event in time.

Estimates of volume and speed come from early analyses and may be revised.

The 3D scenes are simplified teaching models, not to scale.

Sources: [Wikipedia: 2026 Nepal–Tibet floods (Langtang Lirung rock-ice avalanche, 26 August 2026)](https://en.wikipedia.org/wiki/2026_Nepal%E2%80%93Tibet_floods); [Al Jazeera (17 Sep 2026): Climate change had major role in triggering Nepal floods, scientists say (World Weather Attribution)](https://www.aljazeera.com/news/2026/9/17/climate-change-had-major-role-in-triggering-nepal-floods-scientists-say)

## Chapter VI · Watching the slopes

### 

Landslides cannot be stopped everywhere, but the risk can be cut. After deadly landslides in the 1970s, Hong Kong set up its Geotechnical Engineering Office in 1977, to control how hillsides are built on, checked and kept up. Engineers drain slopes so water cannot build up, pin rock with anchors and nets, and plant vegetation to bind the soil. Press drain, and see the water table fall. Warnings are built on rain. Scientists match past storms with the landslides they set off, and draw a threshold: when this much rain falls in this many hours, slopes start to fail. The Langtang collapse showed the gap. The region's flood warning system was set up for monsoon and other floods, and it did not catch this collapse in time.

![Concrete-faced terraces, drains and planted slopes between tall apartment blocks. AI-generated still.](https://assets.planetpast.com/a/f1/f175e1a950e326ad.jpg)

- Contain the risk from new building
- Reduce the risk to what is already built
- Lessen the harm when slides happen

![Anchored mesh drapes a rock cut; a catch fence holds a few boulders. AI-generated still.](https://assets.planetpast.com/a/33/3351e58c36294bb2.jpg)

$$A = 0.4\,D^{0.5}$$
- $A$: rain so far in the storm
- $D$: hours it has rained

The lowest of three curves; heavier rain brings bigger debris flows.

![An automatic rain gauge with a small solar panel on a forested slope. AI-generated still.](https://assets.planetpast.com/a/f3/f3da4d6856690c5a.jpg)

Chapter VI · Watching the slopes

#### Living with landslides

Hong Kong’s Geotechnical Engineering Office was formed in 1977, after the Sau Mau Ping landslide of 1976, on the advice of an independent panel of international experts, to regulate hillside development and the design, construction and maintenance of slopes. Its strategy contains the risk from new development, reduces the risk to existing development and minimises the consequences of landslides, including by issuing landslip warnings. The USGS derives rainfall intensity–duration thresholds such as A = 0.4D^0.5 (accumulation A against duration D) from past storms and debris-flow responses on recently burned slopes in the San Gabriel Mountains, and uses them in an emergency-response decision chart.

The 3D scenes are simplified teaching models, not to scale.

Sources: [Hong Kong Slope Safety (CEDD): the slope safety system and the Geotechnical Engineering Office](https://hkss.cedd.gov.hk/hkss/en/slope-safety-in-hong-kong/slope-safety-system/overview/index.html); [USGS: Rainfall intensity–duration thresholds for postfire debris-flow emergency response planning](https://www.usgs.gov/publications/rainfall-intensity-duration-thresholds-postfire-debris-flow-emergency-response); [Wikipedia: 2026 Nepal–Tibet floods (Langtang Lirung rock-ice avalanche, 26 August 2026)](https://en.wikipedia.org/wiki/2026_Nepal%E2%80%93Tibet_floods)

## Epilogue

### 

Every disaster in this film began the same way: on one slope, on one day, the pull downhill grew bigger than the grip. Many slopes warn before they go: new cracks or bulges in the ground, tilting trees and poles, water ponding where it never did, a stream that suddenly rises or falls. If you see them, tell the local authorities, and leave if it is safe. Mountains will keep letting go. The science is about seeing it coming.

- 1963: Vajont, Italy
- 1977: Hong Kong’s Geotechnical Engineering Office
- 1980: Mount St. Helens, USA
- 1988: Nakhon Si Thammarat, Thailand
- 2014: Oso, USA
- 2015: Langtang village, Nepal
- 2026: Langtang Lirung and the Budhi Gandaki, Nepal

![Leaning poles and fence posts on a slope that is slowly moving. AI-generated still.](https://assets.planetpast.com/a/5e/5e35d32ad4139286.jpg)

- Cracking, knocking or groaning sounds
- A rumble like a passing freight train
- A stream that suddenly rises or falls in a storm

- How fast is mountain permafrost thawing?
- Can satellites spot slopes speeding up in time?
- How to warn for rare, fast events?

![A pale scar and a fan of debris reach a winding river at dawn. AI-generated still.](https://assets.planetpast.com/a/29/293e668fe7f2f740.jpg)

Epilogue · Sources

#### Sources

Every fact in this film comes from the sources below. The chapter stills and pictures are AI-generated artist’s impressions; they show no real places, people or events. The slope, the landforms, the dam and the valleys in 3D are simplified teaching models, not to scale. Casualty figures for 2026 are attributed with their date and may be revised. Warning signs are from the USGS and are general; follow local authorities’ advice.

Sources: [Al Jazeera (17 Sep 2026): Climate change had major role in triggering Nepal floods, scientists say (World Weather Attribution)](https://www.aljazeera.com/news/2026/9/17/climate-change-had-major-role-in-triggering-nepal-floods-scientists-say); [Wikipedia: 2026 Nepal–Tibet floods (Langtang Lirung rock-ice avalanche, 26 August 2026)](https://en.wikipedia.org/wiki/2026_Nepal%E2%80%93Tibet_floods); [Dave Petley, The Landslide Blog, Eos/AGU (12 Oct 2026): the 11 October 2026 landslide and debris flow on the Budhi Gandaki River, Gorkha, Nepal](https://eos.org/thelandslideblog/the-11-october-2026-landslide-and-debris-flow-flood-on-the-budhi-gandaki-river-in-gorkha-district-nepal); [USGS Fact Sheet 2004-3072: Landslide Types and Processes](https://pubs.usgs.gov/fs/2004/3072/fs-2004-3072.html); [Portland State University, Earth Surface Processes: slope stability and the factor of safety](https://pdx.pressbooks.pub/earthsurface/?p=27); [Froude & Petley (2018), Global fatal landslide occurrence from 2004 to 2016, Natural Hazards and Earth System Sciences](https://nhess.copernicus.org/articles/18/2161/2018/nhess-18-2161-2018.html); [USGS: Five years later, the Oso SR 530 landslide, Washington](https://www.usgs.gov/news/featured-story/five-years-later-oso-sr-530-landslide-washington); [USGS Cascades Volcano Observatory: the 18 May 1980 debris avalanche at Mount St. Helens](https://webarchive.library.unt.edu/eot2008/20090115210113/http://vulcan.wr.usgs.gov/Volcanoes/MSH/DebrisAval/description_msh_debris_aval.html); [Britannica: Vaiont Dam](https://www.britannica.com/topic/Vaiont-Dam); [Italian Civil Protection Department: the Vajont landslide](https://servizio-nazionale.protezionecivile.gov.it/en/pagina-base/vajont-landslide/); [Hong Kong Slope Safety (CEDD): the slope safety system and the Geotechnical Engineering Office](https://hkss.cedd.gov.hk/hkss/en/slope-safety-in-hong-kong/slope-safety-system/overview/index.html); [USGS: Rainfall intensity–duration thresholds for postfire debris-flow emergency response planning](https://www.usgs.gov/publications/rainfall-intensity-duration-thresholds-postfire-debris-flow-emergency-response); [USGS: What are the signs of landslide development, and what do I do if a landslide occurs?](https://usgs.gov/programs/landslide-hazards/what-are-signs-landslide-development-what-do-i-do-if-a-landslide-occurs); [Wikipedia: List of disasters in Thailand (Kathun 1988, Nam Ko 2001)](https://en.wikipedia.org/wiki/List_of_disasters_in_Thailand); [PBS Commanding Heights: Thailand, environment (logging ban after the 1988 floods)](https://www.pbs.org/wgbh/commandingheights/lo/countries/th/th_env.html)

