The BioBoost+ project is working to improve the way we monitor life in the sea. By combining advanced imaging methods with Artificial Intelligence (AI), we aim to make the identification of marine plants and animals faster, more cost-effective, and more reliable. These methods are applied across a wide range of ecosystems and species: seagrass meadows, shellfish beds, kelp forests, artic fjords, fish, lobsters, shorebirds, invasive species, and plankton.
At Nord University, our focus is on two long-term monitoring programs:
• Zooplankton, collected every year for over 40 years in two Norwegian fjords.
• Seabed fauna and fish, observed since 2023, through two continuous live-stream underwater video camera in Saltstraumen Marine Protected Area (MPA).
The project’s goal is to provide timely knowledge on biodiversity changes caused by human activities, climate change, and invasive species. BioBoost+ brings together eight organisations across seven countries, with study sites in the Mediterranean, North, and Norwegian Seas, including MPAs and sites with active habitat restoration. By speeding up data analysis, the project will deliver valuable information to scientists, governments, and citizens alike.
During the first year, the project team at Nord University focused on four main activities:
1. Cooperation across partners
We worked closely with colleagues in our consortium to harmonise approaches, ensuring that data and results can be compared and combined. This effort contributed to the first major project report on “Advances for species identification within Bioboost+ project” (Deliverable 1, scheduled for release in October 2025).
2. Underwater video monitoring in Saltstraumen
Two cabled underwater cameras have been recording continuously since 2023: one at 16 m depth in a kelp forest slope, and another at 25 m depth near a small cave regularly visited by wolffish. To build an initial training dataset for AI models, 10-minute video segments recorded every Sunday at noon from April 2023 to April 2024 were extracted.
• All visible fish on frame were labelled at species level. A total of 3,274 individual fish were annotated. Most belonged to saithe (Pollachius virens, 63.5%) and Atlantic cod (Gadus morhua, 35.2%). Less frequent sightings included wolffish (Anarhichas lupus), halibut (Hippoglossus hippoglossus), and flounder species.
• The strong imbalance in species annotations creates accuracy challenges for rare species and bias for computer vision applications. To address this, we will continue to expand the training set with additional annotations and using data augmentation techniques to improve future versions.
• Several object-detection models were tested. While initial results were promising, they did not provide the accuracy needed for our intended ecological applications. Further investigations showed that, in the context of Nord university, the existing TIVA object-detection model combined with custom training set will offers more efficient and accurate results. Implementation of this approach is now ongoing.
3. Long-term zooplankton monitoring (1983–2025)
Zooplankton samples from Saltfjord and Mistfjord, preserved since 1983, were carefully transferred from formaldehyde to ethanol to reduce health risks for researchers and prepare them for imaging and digital analysis. Samples were divided into two size fractions and subsampled to ensure optimal processing for scanning and computer vision applications. Scanning is ongoing.
4. Environmental data compilation
We standardised and collated environmental data collected alongside zooplankton surveys over the past four decades. To fill in gaps, collaboration with local fish farms was initiated. To date, one local fish farm has agreed to share its continuous environmental recordings (temperature, salinity, oxygen) from 2013 onwards, adding valuable environmental context for biodiversity analyses.
In its first year, the team at Nord University has laid the foundations for AI application on biodiversity monitoring: establishing international coordination, creating image training dataset, preparing plankton samples for digitalisation, and securing new environmental data partnerships. The next steps will focus on training and applying AI models at scale, allowing us to track changes in marine ecosystems more rapidly and support evidence-based management of the oceans.
BioBoost+ is designed to improve non-invasive, cost-effective, and high-frequency sampling and identification of marine plants and animals by applying state-of-the-art Artificial Intelligence (AI) technology with digital imagery including real-time monitoring via camera networks and involvement of citizen science groups. These methods are applied to a wide range of taxa, from habitat-forming species (seagrass meadows, shellfish beds, and macroalgal canopies), indicator species of ecological and economic importance (e.g., coastal fish, lobsters, shorebirds), invasive species, and understudied groups (from microscopic, free-floating animals to rare fish). BioBoost+ enhances biodiversity monitoring within European regional seas from the Mediterranean, North and Norwegian Seas to support ongoing conservation interventions such as active habitat restoration projects and Marine Protected Areas (MPAs). Improved biodiversity monitoring will allow a better understanding of large-scale phenomena, such as thresholds, regime shifts and species invasions in vulnerable ecosystems, improve the capacity to predict the impacts of multiple stressors, and to develop better indicators of marine ecosystem health.
