What if the road could warn the car before the moose arrives?

A wildlife crossing is an architectural answer to a biological problem: an animal needs to move through a landscape that a road has cut in two. But concrete bridges and kilometres of fencing are not the only possible interface.

SINTEF’s WILDETECT project is mapping how Norway currently detects and responds to wildlife collision risk. The research points to a fragmented system: population data, collision records, vehicle information, sensors and user reports exist across different organisations and in different formats.

From reacting to predicting

The project’s researchers describe much of the current system as reactive. They are exploring more flexible approaches built around earlier detection and more precise warning. One possible concept is direct animal warnings to vehicles in a relevant area — something that would require roads, cars and data systems to communicate.

The design problem is not simply how to move an animal across a road. It is how to make transport infrastructure perceive the living landscape around it.

The constraints are physical. A train cannot swerve and needs a long stopping distance. A car has more options, but a warning must arrive at the right time and place. Animals also adapt to sounds and lights, so an intervention that works initially may lose effect.

Infrastructure as a sensing system

SINTEF points to sensors, vehicle data, communication systems and drones as technologies with potential, while stressing data quality, standardisation, traffic safety, ethics and animal welfare. The interesting future is therefore not one gadget. It is an interoperable system that can notice risk early enough for both human and animal behaviour to change.

A moose does not know where the transport system begins. It knows the forest, the smell of other animals, forage, snow, fear and the routes it has learned to move through. The road arrives later — a strip of engineered certainty pushed through a landscape that is still alive.

Norway’s collision problem sits at that boundary. SINTEF reports that 9,350 deer died in collisions during the 2024/2025 hunting year. Roads and railways fragment habitat as well as move people, creating places where human speed and animal movement repeatedly intersect.

The conventional responses are familiar: fencing, crossings, signage and clearing vegetation. They can help, but they are expensive and their effectiveness varies by place and species. WILDETECT, a Norwegian research project led by SINTEF, is asking whether the next generation of prevention could be more dynamic — a system that detects risk earlier and delivers warnings where they are actually useful.

The problem is not a missing gadget

SINTEF’s first finding is organisational. Responsibility and data are spread across road authorities, rail operators, wildlife managers, researchers, municipalities, vehicle manufacturers, technology companies and others. Collision records exist. Population data exist. Sensors, vehicle data and user reports exist. What is missing is a clean way to make those pieces work together.

That fragmentation matters because different data answer different questions. Historical records can indicate where collisions tend to occur. Real-time sensors may tell you that an animal is present now. Vehicle systems know where traffic is. Wildlife knowledge helps interpret behaviour. No single dataset is the solution, but together they could create a more responsive warning system.

The design problem is not simply how to move an animal across a road. It is how to make transport infrastructure perceive the living landscape around it.

A warning has to arrive in time

Physics makes the problem uneven. A train cannot swerve and needs a long distance to stop, so a driver seeing an animal may already be too late. A car can brake or steer, but only if the warning reaches the driver early enough and is precise enough to be trusted. Constant false alarms would quickly train people to ignore the system.

Animals learn too. SINTEF notes that wildlife can habituate to traffic sounds and light, and reactions can differ between species and situations. An acoustic deterrent that works at first may become background noise later. That makes this a behavioural system as much as a sensor system.

From roadside sign to network

One possibility raised by the researchers is direct animal warnings to vehicles in a defined area. That requires roads and cars to exchange information — effectively turning wildlife risk into another layer of intelligent transport infrastructure. Drones, roadside sensors and communication systems are also being explored as potential components.

The technical ambition is straightforward to describe and difficult to govern. Data quality has to be good enough. Systems need common standards. Traffic safety cannot be compromised by experimental alerts. Animal welfare and privacy questions have to be handled. And the system has to keep working across agencies that do not naturally share one operational interface.

The bigger design question

For most of the last century, transport intelligence focused overwhelmingly on people and vehicles. The landscape was treated as the fixed background. WILDETECT points towards a different model: infrastructure that can sense the living world it cuts through and adjust behaviour before impact.

That does not eliminate the need for crossings, fencing or better land-use planning. It adds another layer. The future road may need to be judged not only by how efficiently it moves traffic, but by whether it can perceive enough of the surrounding ecosystem to stop treating wildlife as an unpredictable obstacle that appears seconds before a collision.