Fibre in the ground: the quiet engine of AI cities

A lot of smart city AI projects fail for a boring reason: the city can’t “feel” what’s happening fast enough, reliably enough, or cheaply enough. Cameras and standalone sensors help, but they’re expensive to deploy at scale, they need power and maintenance, and they often create governance headaches.

Meanwhile, many cities already have something incredibly valuable sitting under their streets: fibre-optic cable. The SmartCitiesWorld podcast episode “Maximising the use of fibre in the ground” (with Maria Shiao of Fotech) points to a practical idea that deserves more attention in the public sector: distributed acoustic sensing (DAS), a technology that can turn existing fibre into a long, continuous sensor.

This matters for our series “Mākslīgais intelekts publiskajā sektorā un viedajās pilsētās” because AI is only as good as the data pipeline beneath it. Fibre plus DAS isn’t a flashy AI feature. It’s the enabling infrastructure that makes AI-driven urban management credible: better data, earlier warnings, and a lower cost per kilometre of coverage.

Distributed acoustic sensing (DAS): what it does and why cities should care

DAS turns fibre-optic cable into a vibration and sound sensor along its entire length. In practical terms, instead of placing a sensor every 50–200 meters, the fibre itself becomes the sensor.

Here’s the idea in plain language: a DAS unit sends pulses of light down the fibre and measures tiny reflections. When something nearby causes vibrations—traffic, digging, a leak, footsteps, a train—the vibration changes the reflection pattern. Software translates those patterns into signals you can classify and act on.

For city leaders, the attraction is simple:

• Coverage density: kilometres of “sensing” without installing kilometres of new devices.
• Speed: near real-time detection and alerting.
• No roadside power: fibre is passive in the ground; the active equipment sits at the end points.
• Use what you already own: municipal or utility fibre, or fibre deployed through public-private partnerships.

DAS is not “another sensor project”

Most companies get this wrong: they treat sensing as a set of isolated pilots. DAS flips the model. It’s a platform capability—a way to extend situational awareness across big parts of the city using infrastructure already in place.

That’s exactly how public sector AI programs should think: build durable data capabilities, then add use cases on top.

Why fibre is an AI infrastructure asset (not just connectivity)

Fibre is usually framed as broadband and backhaul. That’s only half the story. In an AI-enabled municipality, fibre plays three roles:

1. Transport: moving video, IoT telemetry, and operational data securely to compute.
2. Timing and reliability: enabling consistent, low-latency operations for critical services.
3. Sensing (with DAS): creating new data streams without deploying more street hardware.

The third role is the sleeper hit. When a city can sense disturbances along fibre corridors, it can feed AI models that support:

• Predictive maintenance for roads, rail, pipelines, and ducts
• Incident detection (construction encroachment, suspicious activity near sensitive assets)
• Operational optimisation (traffic flow inference, work-zone monitoring)

The cost problem DAS helps solve

Traditional city sensing is expensive in three places: installation, power, and maintenance. DAS concentrates cost into fewer points (DAS interrogators and analytics), which can improve the economics of scaling.

No technology is free—DAS still requires investment and integration—but it often changes the question from:

“Can we afford to instrument the whole city?”

to:

“Which fibre routes should we activate first to get the biggest operational impact?”

High-value public sector use cases for DAS (and the AI layer on top)

DAS is most valuable where the city wants continuous monitoring across distance. That’s why it’s often discussed in the context of critical infrastructure, transport corridors, and utility networks.

1) Construction and excavation detection (protect what’s underground)

The quickest ROI is often preventing damage from accidental digging. Fibre routes frequently run alongside other valuable assets: water mains, gas lines, district heating, and power.

With DAS, a city or utility can detect characteristic vibration patterns associated with excavation. The AI layer then classifies and prioritises alerts:

• Is it heavy machinery or normal traffic?
• Is it near a high-risk segment?
• Is it inside permitted work hours or outside?

This supports faster dispatch and fewer outages. It also improves governance: you can connect alerts to permit systems and contractor accountability.

2) Water leakage and pipeline anomalies (listen for what you can’t see)

Many leaks start small and become expensive because they go unnoticed. Acoustic signatures can be detectable along fibre routes that run near pipelines.

AI helps by:

• filtering out false positives (traffic, weather, background noise)
• learning the “normal” acoustic profile of a location
• highlighting deviations that merit inspection

For municipalities thinking about AI in infrastructure management, this is the model to copy: continuous sensing → anomaly detection → targeted field work.

3) Rail and road corridor monitoring (operations, safety, and planning)

Along rail lines, DAS can detect train movements and trackside disturbances. Along major roads, it can infer patterns tied to traffic flow and incidents.

AI-driven urban management benefits in two ways:

• Operational awareness: detect disruptions quickly (stopped trains, unusual vibration events, accidents).
• Planning intelligence: aggregate patterns over time to support capital planning and safety interventions.

Even if a city already has traffic cameras, DAS can complement them by providing non-visual, privacy-friendlier signals across long stretches.

4) Perimeter and asset security (public safety without more cameras)

Cities have sensitive sites: depots, substations, water treatment facilities, tunnels, and bridges. Fibre routes near these sites can act as an “early warning line.”

AI classification matters here because raw DAS signals can be noisy. The difference between “someone walking” and “a maintenance vehicle” is a machine learning problem.

Done well, this becomes a practical public safety tool that avoids some of the citizen trust issues tied to adding more video.

How to integrate DAS into an AI-enabled smart city architecture

DAS becomes valuable when it’s integrated like a data product, not treated like a standalone tool. Here’s a city-ready blueprint.

Data flow: from fibre signals to operational decisions

A workable architecture typically includes:

1. DAS interrogator hardware connected to a fibre segment
2. Signal processing and feature extraction (turn vibrations into measurable features)
3. AI/ML models for classification and anomaly detection
4. Event management layer (alert routing, severity, incident correlation)
5. Workflow integration (dispatch, work orders, permits, ticketing)
6. Dashboards and reporting for operations and leadership

The reality? The hardest part is often step 5. Cities don’t need more alerts; they need alerts that trigger the right action with clear ownership.

Governance: treat DAS as critical infrastructure data

Because this series focuses on AI in the public sector, governance can’t be an afterthought. A sensible DAS governance approach includes:

• Data ownership: who owns the fibre, the signals, and the derived events?
• Retention rules: how long raw waveforms vs. derived events are stored
• Access control: role-based access for security-sensitive corridors
• Procurement clarity: avoid vendor lock-in by specifying data interfaces and exportability
• Accountability: an agreed on-call model (who responds to alerts, and in what SLA)

If you’re building e-governance maturity, this is a strong test case: it forces departments to align around shared infrastructure and shared operational outcomes.

A practical adoption roadmap for municipalities (what to do in 90 days)

You don’t need to “instrument the whole city” to start. The smart move is to pick corridors where you can measure operational impact quickly.

Step 1: Map “fibre you can activate”

Start with a joint workshop across IT, utilities, transport, and public works:

• where municipal fibre runs today
• which segments are accessible for DAS interrogators
• which corridors align with high-risk assets (pipelines, rail, tunnels)

Deliverable: a shortlist of 3–5 candidate segments with clear ownership.

Step 2: Choose one use case with a measurable outcome

Good first use cases share three traits: clear baseline costs, frequent incidents, and fast validation.

Examples of measurable outcomes:

• reduced time-to-detect excavation near protected assets
• fewer false call-outs through better classification
• fewer service interruptions tied to third-party damage

Step 3: Run a pilot that looks like production

A pilot should include:

• integration into an existing operations workflow (not an isolated dashboard)
• alert thresholds agreed with the team who will respond
• a simple metric pack reviewed weekly

If your pilot can’t answer “who gets the alert and what happens next?”, it’s not a pilot—it’s a demo.

Step 4: Expand by corridor, not by department

Cities scale faster when they expand along physical networks (corridors, districts) rather than organisational silos. Fibre already maps to geography—use that.

Common questions city leaders ask (and straight answers)

Can DAS replace cameras and IoT sensors?

No—and it shouldn’t. DAS is strongest for continuous linear coverage and detecting vibration/acoustic events. Cameras are better for visual confirmation; point sensors are better for precise local measurements. The winning approach is sensor fusion, where DAS provides early warning and cameras confirm when needed.

Is DAS privacy-friendly?

It can be, if you design it that way. DAS generally detects vibrations and acoustic signatures, not identities. Still, governance matters: define what you store, how you classify, and how you restrict access—especially near sensitive locations.

What’s the biggest implementation risk?

Operational overload. Early systems can generate too many alerts if classification and thresholds aren’t tuned. Budget time for model refinement and field validation, and measure false positives as seriously as detections.

Where this fits in “Mākslīgais intelekts publiskajā sektorā un viedajās pilsētās”

AI in public administration often gets framed as chatbots, document automation, and analytics dashboards. Useful, yes. But the cities that improve service delivery year after year also invest in the unglamorous layer: infrastructure that produces trustworthy, timely signals.

Maximising the use of fibre in the ground—especially with distributed acoustic sensing—fits that pattern. It gives municipalities a way to expand real-time awareness, support data-driven decision-making, and strengthen urban resilience without multiplying street-level hardware.

If you’re planning your 2026 smart city roadmap, here’s the question worth asking internally: which fibre corridors could become your next city-scale sensor network—and what AI decisions would you trust more if you had that data tomorrow?