Product
From satellite pass to expert decision.
Watchberg is a regional screening layer. It watches every slope in a region with open satellite data, finds the ones that start to behave differently, and hands the evidence to the people who decide what happens next.
Data
Open satellites, three bands, one time series per slope.
No sensor to install, no permit to obtain. Every data source below passes over every alpine slope on a fixed schedule.
| Source | Band | Cadence | Role |
|---|---|---|---|
| Sentinel-1C and 1D | C-band radar | 6-day repeat | Dense backbone time series |
| NISAR | L-band radar | 12-day repeat, public since July 2026 | Coherence on snow, ice and fast movers |
| Sentinel-2 | Optical, 10 m | 5-day revisit | Feature tracking, surface change, snow masks |
| ALOS-2 and SAOCOM archives | L-band radar | Historical | Back-tests of past events |
| Commercial X-band | High-resolution radar | On demand | Escalation on flagged slopes |
| ROSE-L | L-band radar | From 2028 | Second L-band source for continuity |
Terrain, glacier and permafrost inventories, and exposure layers (buildings, roads, rail, lifts, dams) come from Swiss federal and open data with attribution.
Pipeline
What happens between two satellite passes.
- 01
Catalogue
A slope catalogue for the region: glacier fronts, rock faces, permafrost zones and the corridors below them, each with its own geometry and visibility masks.
- 02
Process
Radar interferometry and offset tracking in both bands, optical feature tracking, atmospheric and geometric corrections, with layover and shadow masked out explicitly.
- 03
Baseline
Every slope is compared with its own history: seasonal cycles, normal glacier flow, winter data gaps. Nothing is flagged for being fast, only for being different.
- 04
Detect
Change-point detection on the velocity series across bands. Each detection carries a confidence value and a list of what the data could not see.
- 05
Rank
Acceleration evidence, confidence and prior susceptibility are combined with runout envelopes and the assets inside them. The list is ordered by what is at stake.
- 06
Brief
A grounded AI layer writes the briefing from the computed evidence only. An automated check rejects any sentence with a number that is not in the data.
The signal
What acceleration looks like.
Most alpine slopes move a little all the time. The question is never how fast, but whether the movement is changing. Watchberg models each slope's normal behaviour and raises a flag when the rate departs from it. The projected lead time between that flag and a ground-monitoring decision is years to weeks, to be validated in the pilot programme. Hours of warning still come from ground systems.
- Measured displacement rate
- Slope baseline
- Acceleration flagged (projected)
The briefing
Six parts, two pages, every number traceable.
A briefing is written for a geologist with a dozen other things to do today. It answers the questions in the order they get asked.
- 01
What changed
The slope, the period, and the movement in plain words with the figures behind it.
- 02
The evidence
Time series per band, the dates, the datasets and the processing version.
- 03
Confidence and gaps
How sure the detection is, and what the data could not see: snow, layover, missing passes.
- 04
What is exposed
The runout envelope and the assets inside it: settlements, roads, rail, lifts, dams.
- 05
Context
Prior susceptibility, past events on the same slope, and seasonal factors.
- 06
Recommended next step
Expert review, drone survey, ground radar, or no action. Addressed to the expert, never to the public.
What makes it different
Built for high altitude, not adapted to it.
Built for high altitude
Rock-ice interaction, permafrost, glacier fronts and seasonal snow are the baseline, not edge cases.
Multi-band by default
L-band NISAR, C-band Sentinel-1 and optical Sentinel-2 are fused, because C-band alone loses coherence on snow, ice and fast movers.
Acceleration, not velocity
Glaciers move all the time. The signal is when a slope starts moving differently from its own past.
Exposure-aware ranking
Every flagged slope is linked to what lies downstream: settlements, roads, rail, cable cars, dams.
Expert-ready briefings
A two-page briefing a geologist can read in minutes, every number traceable to a dataset, a date range and a processing version.
Hand-off built in
Every briefing ends with a recommended next step: expert review, drone survey, ground radar, and the partner who can deploy it.
Triage, not alarm
What Watchberg is, and what it is not.
Watchberg is
- A regional screening layer
- Expert decision support
- Years-to-weeks early detection, projected
- A hand-off into existing hazard workflows
Watchberg is not
- A replacement for ground radar, cameras or observers
- A public warning system
- Hours-ahead collapse prediction
- A competitor to the cantonal hazard system
Honest limits
What satellites can and cannot do.
- 01
Days, not hours
Satellites revisit every few days. Watchberg screens weekly to monthly. The final hours before a failure belong to ground radar, cameras and observers.
- 02
Winter is a gap
Seasonal snow degrades radar coherence for months. L-band helps but does not solve it. Gaps are shown, never hidden.
- 03
Timing stays hard
Even with long monitoring records, the exact timing and size of a collapse remain difficult to predict. Watchberg narrows where to look, not when.
- 04
Line of sight only
Radar measures motion towards or away from the satellite. Steep terrain causes layover and shadow. Both orbits are combined and unusable pixels are masked.
Roadmap
Switzerland first, then the Alps, then every range.
- 2027
Pilot programme in Switzerland
Back-tests on past events with no look-ahead, then live weekly screening of one region and one infrastructure corridor, used by pilot experts in three cantons.
- 2028
Operational service and the Alpine arc
Regional screening subscriptions for cantons and operators in Switzerland, first pilots in neighbouring Alpine countries.
- Later
High mountains worldwide
Himalaya, Andes and other ranges with local experts, operators and insurers, wherever the same satellites already look.
All dates are planned. Performance figures are projected until the pilot validates them.
See a briefing for your region.
We prepare an example briefing for one region or corridor you are responsible for, from open data, before any pilot discussion.
Pilot programme 2027. In discussion with hazard offices in three cantons.