This article is part of The Purple Report September 2026 | Disaster Forecasting and Predictions and focuses on glacier, landslide, GLOF and mountain-corridor risk.
The 26 August 2026 Gyirong–Rasuwa catastrophe changed how this project thinks about forecasting.
We already knew the ingredients separately: extreme Himalayan relief, changing glaciers, steep confined valleys, roads, hydropower, border infrastructure and vulnerable mountain settlements. What we did not do was ask whether those ingredients were converging strongly enough to deserve a higher observation priority before failure.
Mountain-disaster forecasting is not the prediction of one landslide. It is the detection of a susceptible landscape becoming more heavily loaded while people and critical systems occupy the runout path.
Quick Read
The September 2026 baseline tracks eight mountain/cryosphere systems:
- Gyirong–Rasuwa / Shigatse–Nepal corridor — HIGH-CONCERN
- Sikkim / Teesta basin — ELEVATED
- Uttarakhand–Himachal mountain corridors — ELEVATED
- Bhutan glacial-lake / hydropower valleys — ELEVATED
- Karakoram–Hunza–Chitral — ELEVATED
- Nepal Himalayan hydropower corridors — HIGH-CONCERN
- Cordillera Blanca / Peru — ELEVATED
- Taiwan central/eastern mountain corridors — ELEVATED
These states do not mean a catastrophe is forecast at all eight locations. They identify places where physical susceptibility, exposed infrastructure and changing environmental conditions deserve persistent observation.
Why Mountain Hazards Are Difficult to Forecast
A mountain slope can remain stable for decades and fail in seconds.
The trigger can be rainfall, snowmelt, glacier collapse, earthquake shaking, river erosion, freeze–thaw, slope cutting or several processes acting together.
Even the word landslide compresses very different processes: shallow rainfall-triggered slides, deep-seated slope deformation, rock avalanches, debris flows, glacier/ice-rock collapse and failures that block rivers and create secondary floods.
The correct forecast system therefore needs at least three layers:
- Static susceptibility: where can failure occur?
- Dynamic loading: where are rainfall, soil moisture, snowmelt, glacier change, river erosion or earthquakes increasing stress?
- Acute precursors: where do instruments show deformation, acceleration, cracking, lake growth or unusual flow?
NASA LHASA: From Susceptibility to Near-Real-Time Nowcast
NASA’s Landslide Hazard Assessment for Situational Awareness—LHASA—is a useful model for the live layer. It combines landslide susceptibility with changing inputs such as rainfall and soil moisture to estimate near-real-time landslide hazard globally.
The important conceptual move is that rainfall is not used alone.
Ten centimetres of rain falling on a dry, gentle, stable landscape is not the same hazard as the same rainfall falling on a steep, already saturated slope with weak geology and a road cut below it.
Trigger intensity matters. So does the state of the landscape before the trigger arrives.
NASA — Landslide Hazard Assessment for Situational Awareness
High Mountain Asia: The Structural Watch Is Already Changing
The World Meteorological Organization reports continued mass loss across all 23 monitored High Mountain Asian glaciers during the 2025 glaciological year, with multiple glacial-lake outburst floods and glacier collapses recorded across the region.
That observation does not mean every retreating glacier is becoming catastrophically unstable.
It does mean the substrate itself is changing.
As glaciers thin and retreat, lakes can expand, unsupported rock slopes can be exposed, drainage pathways can change and meltwater systems can reorganise. These changes occur inside a region already characterised by extreme relief, active tectonics and large downstream populations.
WMO — State of the Climate in Asia 2025
Gyirong: Our First Blind-Test Case
Gyirong is now the first retrospective blind-test site for the Purple Forecast.
The event has already happened, so the forecast must deliberately hide the answer.
We will reconstruct the information environment at several dates before 26 August 2026 and ask:
- Was the terrain structurally susceptible?
- What was known about glacier and upstream hazard conditions?
- What rainfall, temperature or hydrological loading was observable?
- Were roads, border facilities, power and communications concentrated in one corridor?
- Were alternative routes weak?
- Was the cross-border warning system sufficiently instrumented?
Then we will run the same model on comparable valleys that did not fail.
If Gyirong rises only because we already know the disaster happened, the model has learned nothing.
For the event reconstruction, see Civilisation Atlas | Shigatse and the Gyirong Disaster.
Sikkim and the Teesta: GLOF Plus Hydropower Cascade
Sikkim is a particularly important mountain-system watch because glacial lakes sit above narrow valleys containing settlements, bridges, roads and hydropower.
ICIMOD-hosted research has identified large numbers of glacial lakes in the state and ranked several as high or very-high risk based on lake, dam, glacier and downstream characteristics.
The Purple Forecast adds one more question:
If one lake fails, which power, bridge, road and rescue systems sit in the same downstream corridor?
A GLOF is not only a water event. It can become an energy, transport and communications event.
ICIMOD library — glacial-lake outburst flood risk in Sikkim
Uttarakhand and Himachal: Roads, Pilgrimage, Hydropower and Steep Terrain
The western Himalaya combines several kinds of loading.
- active tectonics;
- rapid river incision;
- glacier and snow change;
- intense monsoon rainfall;
- high mountain-road density;
- tunnel and hydropower construction;
- large seasonal pilgrimage and tourism flows.
The forecast challenge is not to declare the whole region unstable.
It is to identify corridor-level convergence: the specific valley or slope where dynamic loading intersects a road, town, dam, tunnel, bridge or pilgrimage route with limited redundancy.
This is where high-resolution satellite deformation, rainfall, soil moisture, landslide inventories and local engineering information become more valuable than a generic Himalayan risk label.
Bhutan: Dangerous Lakes Inside a Hydropower Nation
Bhutan’s glacial-lake monitoring matters not simply because the lakes are dangerous.
Hydropower, narrow inhabited valleys and transboundary rivers can connect a high-altitude failure to national electricity and downstream communities.
That makes Bhutan a strong example of why hazard mapping must be combined with downstream infrastructure and community exposure.
The glacier or lake is the physical object.
The forecast has to consider the whole downstream chain of consequences.
Karakoram–Hunza–Chitral: Road Is Part of the Hazard Model
Northern Pakistan contains glacierised valleys, glacial lakes, active slopes and critical mountain roads including parts of the Karakoram Highway system.
In this environment, a slope failure can block a river; a blocked river can create a lake; a lake can later fail; a road can be cut; and rescue or trade can lose its only practical route.
That creates a multi-stage hazard:
slope or glacier change → blockage → temporary storage → secondary flood → road isolation → delayed recovery
Forecasting must therefore keep temporary dams and new impoundments visible after the first landslide occurs.
Nepal Hydropower Corridors: Workers and Tunnels Change the Risk
The 2026 Himalayan disaster highlighted an exposed group that ordinary flood maps can miss: workers inside hydropower tunnels and construction sites.
A valley hazard can reach:
- surface settlements;
- construction camps;
- tunnels;
- access roads;
- powerhouses;
- transmission lines;
- rescue teams.
This changes evacuation geometry.
A warning that reaches a village may not reach a tunnel crew in time.
Infrastructure expansion therefore creates both capability and new exposure.
Cordillera Blanca: The Tropical-Andes Laboratory
Peru’s Cordillera Blanca is one of the world’s classic glacial-lake hazard regions.
Rapid tropical-glacier retreat has created and expanded lakes in steep valleys above populated areas, while earthquakes add another possible trigger pathway.
The region is valuable to this project because it allows comparison with the Himalaya:
- different climate;
- different institutions;
- similar glacier-lake and steep-valley physics;
- different warning and engineering histories.
If a forecasting method works only in the Himalaya, it may have learned the location rather than the mechanism.
Taiwan: Earthquake and Typhoon Can Attack the Same Slope
Taiwan is a useful compound-hazard laboratory because active tectonics and tropical rainfall repeatedly interact with the same mountain terrain.
An earthquake can weaken a slope without causing immediate collapse.
A later typhoon can deliver the rainfall that pushes that slope past failure.
This means the relevant state includes memory of previous loading.
A landslide nowcast should not treat today’s rainfall as if the mountain arrived fresh this morning.
What the Mountain Forecast Tracks
| Layer | Examples of useful evidence |
|---|---|
| Structural susceptibility | slope, relief, geology, faults, known landslides, glacier/lake geometry, river confinement |
| Dynamic loading | rainfall intensity/duration, antecedent soil moisture, snowmelt, temperature anomaly, river erosion, earthquakes |
| Precursors | InSAR/GNSS deformation, glacier velocity, cracks, lake growth, pore pressure, river-stage change |
| Exposure | settlements, roads, tunnels, hydropower, border facilities, tourists/workers |
| Cascade | river blockage, secondary flood, road isolation, grid loss, communications loss |
| Observability | satellite revisit, weather station coverage, lake gauges, slope sensors, local reporting |
| Repair | alternate roads, helicopters, satellite communications, bridge stock, early-warning protocol |
What Would Raise a Hotspot?
- rapid lake expansion;
- measured slope or glacier acceleration;
- large cracks or deformation;
- extreme or prolonged rainfall over already saturated ground;
- rapid snow/ice melt;
- recent strong earthquake loading;
- new river blockage;
- critical road or hydropower dependence with no alternate route;
- loss of monitoring or communications during a high-loading period.
What Would Lower It?
- dynamic loading falls;
- measured deformation stops or reverses under expert interpretation;
- lake levels stabilise;
- drainage or engineering work reduces the credible failure mode;
- route redundancy improves;
- early-warning and evacuation systems are installed and verified;
- new evidence shows the assumed mechanism was wrong.
The Purple Report must be willing to downgrade a dramatic hotspot when reality becomes less concerning.
September 2026 Mountain Ledger
| System | State | Main reason |
|---|---|---|
| Gyirong–Rasuwa | HIGH-CONCERN | post-event instability + destroyed critical corridor + reconstruction |
| Sikkim / Teesta | ELEVATED | dangerous glacial lakes + hydropower/road exposure |
| Uttarakhand–Himachal | ELEVATED | extreme relief + monsoon + roads/hydropower/pilgrimage |
| Bhutan glacier valleys | ELEVATED | GLOF susceptibility + hydropower and narrow-valley exposure |
| Karakoram–Hunza–Chitral | ELEVATED | glacier/lake/slope hazards + critical road corridors |
| Nepal hydropower corridors | HIGH-CONCERN | recent disaster + tunnel/worker/power exposure |
| Cordillera Blanca | ELEVATED | glacial lakes + steep downstream settlements |
| Taiwan mountain corridors | ELEVATED | earthquake + typhoon rainfall compound slope loading |
Final Rule
We cannot reliably predict every mountain failure. We can become much better at recognising when a vulnerable mountain corridor is carrying more risk than usual—and at seeing what will be lost if that corridor fails.