Tackling Turbidity: BiOceanOr Wins Jury Prize
Tackling Turbidity: BiOceanOr Wins Jury Prize in Antwerp-Bruges for Ports & Coastal Authorities
Harmful algal blooms (HABs) pose a major environmental and economic challenge in aquaculture. Predicting these events before they escalate is essential to protecting fish farming operations. That is why BiOceanOr is supporting a doctoral research project, conducted within and under the scientific direction of the Laboratoire Environnement & Ressources at Ifremer in Boulogne-sur-Mer (LERBL). The thesis asks: can AI predict HAB risk early enough to act, and explain why?
Combining Environmental Data Sources With AI
The research project integrates multiple data sources:
in-situ meteorological, hydrological, and physicochemical data
phytoplankton cell counts
satellite observation data
numerical model outputs
Based on artificial neural networks, machine learning and explainable AI, this research aims to predict the probability of a HAB occurring in real time. To address the black-box nature often associated with machine learning models, a SHAP (SHapley Additive exPlanations) analysis is used to identify which environmental factors are driving the risk, so predictions can be easily understood and trusted.
Curious about how BiOceanOr supports scientific research?
From the English Channel to Salmon Producing Chile
The methodology is first developed in the eastern English Channel, which benefits from a dense observation network (the French Research Infrastructure SNO COAST-HF) and long-term phytoplankton data from the SNO PhytObs network and the SRN-REPHY monitoring programs.
As it continues, the research will investigate the transferability of the methodology to ecosystems where monitoring is sparser and historical data more limited. This second stage will assess the model’s adaptability and robustness under different environmental and data-availability conditions.
Chile offers a particularly relevant case study for this, as it is:
the world’s second-largest salmon producing country.
a place where the need for such predictions is real. In 2016, a bloom of Pseudochattonella verruculosa wiped out more than 40,000 tonnes of salmon (León Muñoz et al., 2018), and major events have occurred more recently in 2021, 2023, and 2024.
a region where BiOceanOr is already active and works alongside several local producers who are eager to test the solutions.
From Scientific Research to Applied Early Warning for Aquaculture
By combining multimodal data fusion, machine learning and explainable AI, this research aims to advance predictive tools that help stakeholders anticipate HAB events, improve environmental risk assessment and make better-informed decisions, for more resilient and sustainable aquaculture.
Close ties with the scientific community have been at the heart of BiOceanOr’s project from day one, and turning the latest research into operational tools is central to our mission. This thesis is a natural continuation of that commitment, and of our ongoing work on HAB.
Want to see how these advances could apply to your own sites?
See also
Blumar: Anticipate Environmental Challenges with BiOceanOr
Blumar: Anticipate Environmental Challenges with BiOceanOr’s Predictive Insights
From Aquaculture to Broader Water Quality Monitoring
From Aquaculture to Broader Water Quality Monitoring: BiOceanOr’s Work with the City of Cannes