The Slope That Moved — and the System That Caught It

Kashish — Thu, 04 Jun 2026 12:26:15 +0000

How a single GEO T20 device gave a mountain road authority 11 days of warning before a slope failure nobody expected.

The road authority didn’t think the slope was particularly dangerous. It had been stable for years — no visible cracks, no reported movement, nothing that flagged in the annual inspection. The GEO T20 disagreed. Quietly, without any drama, it started logging a 4 millimetre displacement in the upper section of the slope. Then 6mm. Then the alerts started. Eleven days later, the slope failed.

Table of Contents

1. The Site and the Problem

The site was a regional mountain road in an Alpine corridor — the kind of road that serves a handful of villages and a ski resort during winter. Around 1,800 vehicles used it daily during peak season, dropping to a few hundred in the shoulder months. The slope in question was a cut face running about 80 metres above the road on the uphill side, made up of weathered limestone with clay intercalations — not ideal, but not something that had caused any serious issues in the past.

The road authority had done a geotechnical survey three years earlier. The conclusion was that the slope was in a stable condition with low risk of failure. No monitoring had been installed. The only maintenance regime in place was an annual visual inspection — someone walking the road, looking up, noting anything obviously wrong.

The problem with that approach — and this comes up again and again — is that slopes don’t announce their intentions visually until very late in the process. What you can see with your eyes, standing on a road, is almost never the early warning signal. The early signal is almost always subsurface, or so subtle that you need precise measurement to catch it.

The key risk factor

An unusually wet autumn had raised groundwater levels in the slope significantly above seasonal norms. This wasn’t visible from the road — but it was silently reducing the effective friction angle of the clay layer at depth. The slope was moving. Nobody knew.

2. How the System Was Set Up

The installation happened in September — before the wet season, as part of a regional risk assessment programme that covered several roads in the area. The brief was to deploy continuous visual monitoring on slopes that had been flagged as “low to medium risk” in recent surveys. Not high priority. Not emergency situations. Just ongoing observation.

A single Avacam GEO T20 was mounted on a pole on the opposite side of the road, positioned to cover the full extent of the cut face at a distance of about 110 metres. The lens was adjusted remotely during commissioning — the Avacam engineer never needed to visit the site after the initial installation day. A 4G connection transmitted images to the cloud platform every 30 minutes during daylight hours, and every 2 hours overnight.

Device = Avacam GEO T20 — 20 megapixel full-resolution timelapse camera, IP68-rated enclosure.

Connectivity = 4G/LTE connection. Images transmitted to cloud platform every 30 minutes. Remote lens adjustment at commissioning.

Power = Mains-connected. No solar needed at this elevation. Backup battery for 72-hour outage continuity.

Alert thresholds = Displacement alerts set at 3mm (notification), 8mm (amber warning), 15mm (red — road closure protocol).

The alert thresholds were set in consultation with the road authority’s geotechnical team. Three millimetres would trigger a notification email. Eight millimetres would trigger an amber warning requiring a site inspection within 24 hours. Fifteen millimetres would trigger the red alert — the pre-agreed protocol called for immediate road closure and emergency assessment. These numbers were conservative by design. Better to inspect something that turns out fine than miss something that doesn’t.

3. What the Monitoring System Detected — Day by Day

For the first six weeks after installation, nothing of note happened. Images came in on schedule, the slope looked as expected, and the platform logged a baseline that would later prove invaluable. Then, in the third week of October, something changed.

Day 1 — October 21 = First displacement detected: 4mm

The AI flagged a 4mm displacement in the upper-left quadrant of the monitored face. A notification email went to the road authority’s monitoring inbox. At this point, 4mm is within the “watch” category — notable but not alarming. The duty engineer logged it and scheduled a visual check during the next routine patrol.

Day 4 — October 24 = Displacement increases to 7mm. Rate accelerating.

Three days later, the same area showed 7mm of cumulative displacement. More importantly, the rate of change was accelerating — not linear. The platform flagged this as an anomalous progression pattern. A second notification went out. The duty engineer — who had seen the first one — marked the site for priority inspection.

Day 6 — October 26 = Amber alert: 9mm displacement. On-site inspection ordered.

The 8mm amber threshold was crossed. An automated amber alert triggered the pre-agreed protocol: the geotechnical consultant was contacted and a site inspection was arranged for the following morning. The road remained open — the displacement was real but not yet at the level that justified closure.

Day 7 — October 27 = On-site inspection. Hairline crack found at slope crest.

The consultant visited the site. Using the platform’s time-lapse to pinpoint exactly where to look, she found a hairline tension crack at the slope crest — invisible from the road, only visible from above. Without the monitoring data directing her to the exact location, she may not have found it at all. The crack measured approximately 3cm wide and ran for about 4 metres. This changed everything. The assessment was upgraded to high risk.

Day 9 — October 29 = Red alert: 16mm. Road closed. Emergency measures begin.

Displacement reached 16mm. The red alert protocol activated. The road authority issued an immediate closure order — the road was blocked at both ends with barriers and diversion signage. An emergency contractor was mobilised to begin rock anchoring and drainage works. The closure caused significant disruption to local traffic, but the decision was straightforward given the data in front of them.

Day 11 — October 31 = Slope failure at 02:14. Road closed. Nobody hurt.

At 02:14 in the morning, approximately 850 cubic metres of material detached from the slope and came down onto the road. The debris covered a 60-metre section of carriageway to a depth of up to 2.5 metres. The road was already closed. Not a single vehicle was on it. The monitoring cameras captured the failure sequence in full — images that would later form the core of the incident investigation report.

4. The Response: What Happened When the Alerts Fired

Something that gets underappreciated in monitoring case studies is how much the quality of the alert matters — not just whether one fires. A vague alert saying “something might be happening” is not the same as an alert that says “4mm displacement detected in sector B3, see attached comparison images from 72 hours ago.”

Every alert the Avacam platform sent included a link directly to the image comparison view — the engineer receiving the notification could immediately see a side-by-side of the current image and the baseline, with the displacement zone highlighted. No login confusion, no searching through folders, no waiting. The data was there, readable, and actionable within seconds of opening the email.

“When the amber alert came through, I had the comparison images open on my phone within about 30 seconds. I could see exactly where the movement was before I even made a phone call.”

— Geotechnical Consultant, on-site response team

That speed mattered. The decision to escalate, to bring in the geotechnical consultant, to find the crack at the crest — all of that happened faster because the data was immediately legible. In an emergency, every hour of decision-making delay is a real cost.4

5. The Slope Fails — and Nobody Is on the Road

The failure itself happened in the early hours of October 31st. The GEO T20 was still running. It captured the failure sequence across four consecutive image frames — the progressive detachment of the upper block, the debris run, and the final deposition pattern on the road below. These images were automatically stored in the permanent cloud archive.

What would have happened without monitoring

Based on the failure timing — 02:14 on a midweek morning — traffic modelling suggests approximately 12–18 vehicles would have been on that section of road during the 30-minute window around the failure event under normal operating conditions. The road was closed 48 hours before the failure occurred. No traffic. No casualties. No near misses.

The debris volume — around 850 cubic metres — was substantial enough to have caused fatal accidents had any vehicle been caught underneath it. The cleanup and road reinstatement took 19 days and cost approximately €380,000. The monitoring system, including hardware, installation, and 12 months of platform subscription, cost a fraction of that.

6. After the Event: The Data That Told the Full Story

One of the things that surprised the road authority most was how useful the monitoring data was after the event — not just before it. The complete image archive, going back to the installation date six weeks earlier, gave the investigation team a precise record of when movement started, how fast it progressed, and what the failure mechanism looked like in sequence.

This data did three important things:

It supported the insurance claim — the road authority had documented evidence of the entire event, from first detection to failure, with timestamps on every image. The claim was settled without dispute.
It informed the remediation design — the engineers designing the rock anchoring and drainage solution could see exactly where the failure initiated and how it propagated. That’s far more useful than a post-failure inspection of a debris pile.
It changed the authority’s monitoring programme — within three months of the event, the authority commissioned Avacam systems on four additional slopes in the same corridor. The budget came from the insurance payout.

Outcome summary = zero casualties. Road closed 48 hours before failure. Full failure sequence documented. Insurance claim settled in full. Four additional monitoring sites commissioned within 90 days of the event.

7. What This Case Teaches Us

We’ve written this up not to pat ourselves on the back, but because the lessons here apply to a lot of sites that are currently unmonitored. A few things stand out clearly.

Visual inspection alone is not enough

The annual inspection had passed this slope as low risk just three years earlier. That assessment wasn’t wrong at the time — but conditions change, groundwater levels change, and the slope was already moving before anyone could see it from the road. The 4mm displacement that triggered the first notification was completely invisible to the naked eye at 110 metres distance. Only a 20-megapixel camera with baseline comparison could catch it.

Early warning only works if thresholds are set correctly

The 3mm notification threshold was conservative. Some engineers pushed back on it during setup — “you’ll get too many false alerts.” In this case, that 3mm alert was the thing that put the slope on the authority’s radar six days before the red alert fired. If the lowest threshold had been set at 10mm, the first notification would have come 48 hours before closure — not six days. The margin for decision-making would have been much smaller.

The response protocol matters as much as the detection

The road authority had agreed on the response protocol before anything happened. When the amber alert came, everyone knew what to do. No committee meeting, no escalation debate, no “let’s wait and see.” The protocol said: amber alert means geotechnical inspection within 24 hours. That happened. That decision directly led to finding the tension crack — the piece of evidence that justified the eventual road closure.

Post-event data has serious financial value

The monitoring archive was worth real money in the insurance claim. An undocumented slope failure leaves the road authority trying to reconstruct a timeline from memory and post-event inspection. A fully documented one — with timestamped images showing every stage of the progression — is a completely different conversation with an insurer.

11 = Days from first alert to failure

4mm = First detected displacement

850m³ = Volume of debris on road

€0 = Cost of casualties

8. Key Takeaways

Monitoring doesn’t prevent slope failures. Nothing does — geology does what geology does. What monitoring gives you is time. Time to inspect, time to decide, time to close a road before 850 cubic metres of rock and clay lands on it. In this case, 11 days was more than enough. On a different site, with a faster-moving failure mechanism, even 48 hours might be the difference between a closed road and a disaster. The cost of a GEO T20, installed and running for 12 months, is a rounding error compared to what was avoided here. That’s the argument, and we think it’s a straightforward one.

Have a Slope That Needs Watching?

Tell us about your site. We’ll recommend the right device, help define your alert thresholds, and have it transmitting data within days of installation.

Talk to Our Team

L'articolo The Slope That Moved — and the System That Caught It proviene da Avacam.

Monitoring System Archivi - Avacam