Skip to content
Watchberg
Articles

8 October 2026/5 min read

Acceleration, not velocity: the signal that matters

Alpine slopes move all the time. The warning sign is not how fast a slope moves but when it starts moving differently.

By Julian Walder

A geologist who receives a satellite velocity map of an alpine valley sees a lot of colour. Rock glaciers creep. Glacier tongues flow. Deep-seated slopes move a few centimetres a year and have done so since the last ice age. Permafrost terrain heaves and settles with the seasons. If every moving pixel were an alert, the map would be useless on day one.

That is why the question is never "how fast". It is "faster than before".

Slopes that always move

Switzerland's permafrost monitoring network publishes rock glacier velocities as a matter of routine, and the national Sentinel-1 motion product maps points that move at steady rates year after year. Steady is the key word. A slope that has moved at the same rate for a decade is, from a triage perspective, low priority. Its behaviour is known, its hazard has usually been mapped, and the people responsible for it have already decided what to do.

The slope that deserves attention is the one whose rate changes.

Brienz: a rate that kept rising

The village of Brienz/Brinzauls in Graubünden sits on a large landslide that has been measured since 2009. The numbers from the monitoring programme tell the story more clearly than any map. In 2020 the fastest part of the slide was moving at 3 to 5 metres per year. By the end of 2022 it had reached up to 20 metres per year. In the days before 15 June 2023, when about 1.2 million cubic metres came down as a debris flow, local rates exceeded 40 metres per day.

Each of those steps was an acceleration, and each was visible in the data before the next. The same pattern appears at Blatten, where published L-band analysis shows the Kleines Nesthorn moving at about 50 centimetres per year in 2023 and at more than 150 centimetres per year by August 2024, three times faster within a year and nine months before the collapse.

The logic also works in the other direction. At Brienz a drainage tunnel, started in June 2024, cut the village's own motion from about 250 centimetres per year in autumn 2024 to well under 100 centimetres per year by August 2025. A system that only ranked by velocity would still have Brienz near the top of the list. A system that ranks by change in rate would correctly show it calming down.

Two ideas in plain words

Two established methods underpin the detection of acceleration, and neither is complicated to describe.

Change-point detection asks a simple statistical question of a time series: is there a date after which the data are better explained by a different trend than before? Apply it to a slope's displacement history and it returns the moment the slope stopped behaving as it used to, together with a measure of how confident that conclusion is. A seasonal cycle, such as a glacier speeding up every summer, is first modelled and removed, so that normal behaviour is not mistaken for change.

The inverse-velocity method, developed by Fukuzono for slopes in creep, observes that in many slopes approaching failure the velocity increases in such a way that one divided by velocity falls roughly along a straight line towards zero. Extrapolating that line gives a rough idea of when the velocity would become unbounded. Ground-monitoring teams use it on slopes that are already in an accelerating phase. It is not a clock: a 2026 review of collapsing mountains is clear that the exact timing and size of a failure remain hard to predict even with long-term monitoring. But it turns "faster" into "how much faster, and on what trajectory", which is the information an expert needs to decide whether to escalate.

Velocity tells you a slope is alive. Acceleration tells you something has changed.

How the Italian regions do it

The closest public analogue to this logic already runs south of the Alps. The regions of Tuscany, Valle d'Aosta and Veneto operate continuous Sentinel-1 point monitoring with the University of Florence, updated every 6 to 12 days. A point is flagged as an anomaly automatically when its velocity changes by more than 10 millimetres per year over the last 150 days. Flagged points then go to remote and field validation by the regional geological service. The trigger is the change, not the absolute rate. The service is C-band only and points-based, so it does not reach onto snow and ice, but the workflow is the right one.

What Watchberg ranks by

Watchberg applies the same principle to high-alpine terrain, with three additions. First, each slope gets its own baseline: its normal rate, its seasonal cycle and its measurement noise, estimated from its own history, so that a glacier flowing at its usual summer speed is not an alert. Second, detection runs across Sentinel-1 C-band, NISAR L-band and Sentinel-2 optical offsets, so a slope that saturates in one sensor can still be followed in another. Third, the triage score combines four things: the strength of the acceleration evidence, the confidence of the measurement, the prior susceptibility of the terrain (permafrost, slope angle, glacier loading) and the exposure below, meaning the settlements, roads, rail lines, cable cars and dams inside a simple runout envelope.

The ranked list that results is not a list of the fastest slopes. It is a list of slopes whose behaviour has changed, ordered by how much is at stake downstream and how sure the data are. Each entry comes with its evidence, its confidence and its gaps, and a recommended next step for the expert who owns the decision.

How well this performs on real Swiss cases is a projection until it is tested. The back-tests planned for the 2027 pilot programme, run with no look-ahead on data that would have been available at the time, are designed to measure exactly that: how early acceleration is detected, how often an alert is wrong, and how the known cases rank against the quiet majority.

Sources

  • InSAR
  • Brienz
  • Methods

Related

Join the 2027 pilot programme.

Watchberg is preparing pilots with hazard offices and operators in three Swiss cantons. If you carry responsibility for alpine slopes, let us show you what a weekly briefing would look like for your region.

Request a pilot briefingRead the articles

Pilot programme 2027. In discussion with hazard offices in three cantons.