Monitoring birds around a floating offshore turbine: Fondation OPEN-C

Project overview

• First-of-its-kind upward-facing camera deployment on a floating offshore wind turbine in France.
• ~32,000 AI-detected bird trajectories over ~10 months.
• 1,690 expert-validated bird tracks reviewed in detail.
  • 80% of birds showed no interaction with the turbine. 20% showed avoidance behaviour.
  • Two confirmed collisions (0.1% of tracks), both involving gulls.
• 96% AI detection precision after site-specific retraining.
• No observable change in bird behaviour when the turbine was rotating versus stopped.

The challenge

Floating offshore wind is opening up new sites further from shore and in deeper water. It is also opening up a knowledge gap. Empirical data on how seabirds and bats interact with floating turbines is scarce, and existing monitoring methods are mostly designed for fixed-bottom turbines near the coast.

PIAFF&Co set out to address that gap at the SEM-REV test site, ~20 km offshore Le Croisic. Between 2022 and 2025, the project addressed three questions that matter for both regulators and developers:

1. Which species of birds and bats are present around a floating turbine at sea?
2. How do they use the platform itself?
3. How do they interact in flight with the rotating blades?

To get useful answers on the 3rd question regarding bird interactions, the team needed a monitoring approach that could run continuously for many months, capture fine-scale flight behaviour close to the blades, and stand up to independent scientific review.

The approach

The standard Spoor configuration was adapted in two important ways to suit the project design.

A customised camera setup. Spoor's normal setup uses a remote camera on one turbine looking sideways at a neighbouring turbine. For this project, a single camera was mounted on the FLOATGEN, a 2MW floating offshore wind turbine designed and operated by BW Ideol, looking upward toward the blades on a custom frame designed by OPEN-C. Power and data ran through a dedicated cabinet connected to the turbine's fibre optic backbone. The system operated during daylight hours.

Site-specific AI retraining. The upward angle and the floater's natural motion created conditions the standard model had not seen. Spoor retrained the detection model on data from the site, lifting precision to 96% for the operational monitoring period.

From there, experts reviewed 1,925 AI-detected tracks frame by frame, classified behaviours, and contextualised the results against metocean and turbine operating data. Spoor provided detection outputs as a structured CSV so scientists could run their own analyses outside the platform.

The results

Over the ~10-month campaign, the system produced ~32,000 AI-detected trajectories. Of these, 1,925 were selected for detailed expert review. 1,690 were confirmed as real bird tracks and formed the basis of the behavioural analysis. The picture that emerged is grounded in direct observation rather than modelling.

• Across the full project, 44 bird species and 6 bat species were recorded at the site.
• 80% of confirmed tracks showed no interaction with the turbine. Birds simply passed through the airspace.
• 20% showed avoidance behaviour. Of these, 80% were distance-based manoeuvres (meso-avoidance) and 20% were close-range adjustments around the blades (micro-avoidance).
• Two collisions were observed, equivalent to 0.1% of analysed tracks. Both involved gulls.
• Bird behaviour did not change measurably when the turbine was rotating versus stopped.

The project's full findings were presented in a public webinar and the full report.

What this means for the sector

There are two takeaways that the wider wind energy sector can act on.

Camera monitoring can capture the events that matter.
Collisions and avoidance behaviour on a floating turbine are observable with a camera-based approach. Direct observational evidence at this level of detail is still uncommon for floating offshore wind.

AI and expert review are stronger together.
Spoor's role was to filter ~32,000 raw detections into a manageable, well-structured dataset that scientists could interrogate. The behavioural interpretation required systematic human review of every confirmed track to confirm precision of automatic bird detections.

OPEN-C also flagged useful directions for future work: cross-comparison with other monitoring methods, longer time series, species-level identification on upward-facing setups, and dedicated bat monitoring.

PIAFF&Co was a first for floating offshore wind in France, and we needed a partner willing to adapt to a research setting rather than sell us an off-the-shelf product. Spoor adjusted both the hardware and the software so we could answer the questions our scientists actually wanted to answer. The result is a dataset that genuinely advances what we know about how seabirds interact with floating turbines.— Thomas Soulard, R&D Project Manager, Fondation OPEN-C

What the evidence shows

This project demonstrates that collisions occur, but remain rare – and that avoidance behaviour is far more common. That matters because collision risk in offshore wind has historically relied on conservative modelling rather than observation. Direct data, collected with the right technology over sufficient time, is what moves the sector from assumptions to evidence.

The same conclusion holds at Vattenfall's Aberdeen Bay – where Spoor's monitoring documented bird activity at scale with no confirmed collisions across 137,000+ detected birds. The right monitoring technology, applied rigorously, is what turns field observations into evidence the sector can act on.

Species detection

With support from Spoor’s ornithologists, multiple bird species were identified and documented. By analysing flight heights, Spoor was able to calculate the number of birds passing through potential collision zones.

Credits

PIAFF&Co was coordinated by Fondation OPEN-C with the Muséum national d'Histoire naturelle, BW Ideol, EDF Power Solutions and Centrale Nantes, with financial support from ADEME.