BIG BRUVER IS WATCHING Utilising Low-Cost Pelagic Baited Remote Underwater Video Stations to Assess Fish Assemblages at a Swedish Seaweed Farm

Sammanfattning

Seaweed aquaculture has gained considerable attention in recent decades due to its low environmental impact and potential to provide a more sustainably produced food source. But seaweed farms can also provide supporting ecosystem services. This is due to its three-dimensional structure, which can be used as a habitat and as nursery grounds. However, the understanding of how seaweed farms impact the local biodiversity in Sweden remains limited. Understanding how the addition of a seaweed farm interacts with the surrounding environment is crucial, as more and more businesses are adopting a nature-positive approach. The primary aims of this study were to enhance our understanding of the supporting ecosystem services provided by seaweed aquaculture by designing a pelagic Baited Remote Underwater Video System (BRUVS) and using the recorded footage to train an object detection model to recognise and classify the most encountered species. While the pelagic BRUVS proved to be an effective and durable low-cost sampling platform, the lack of fish observed in the collected footage suggests that the three-dimensional habitat provided by the Swedish seaweed farms does not enhance local biodiversity during cultivation. Due to the absence of fish in the footage, an object detection model was trained on old BRUVS footage recorded at a Swedish mussel farm, which proved effective, achieving high recall while maintaining high precision. Although the footage was recorded on the seafloor, the results highlight the potential of combining BRUVS with deep learning for environmental impact assessment. Future research should prioritise more frequent samples over a longer period of time, as it could capture possible seasonal trends.