Back to search

STIPINST-Stipendiatstillinger i instituttsektoren

Stipendiatstilling 1 NERSC (2020-2023)

Awarded: NOK 4.1 mill.

Algae blooms are a part of the natural annual cycle in marine and coastal waters and play an essential role in the marine ecosystem. The aim of this institutional doctoral study is to be able to predict the evolution of harmful algae blooms in Norwegian coastal waters from one to three weeks ahead of time (sub-seasonal predictions). Being able to provide early warnings will allow for increased monitoring and earlier detection of harmful algae blooms and will allows fisheries and aquaculture to plan for mitigation action. Phytoplankton blooms develop when phytoplankton grows rapidly. Phytoplankton are tiny organisms living in the water that are at the base of the marine food web and contribute to the carbon cycle. Due to their photosynthetic and accessory pigments, they are visible from space when present in sufficiently large amounts. In a study published in Frontiers in Marine Science, we apply a novel clustering technique to the 21 year-long record of satellite remote sensing measurements to investigate the phytoplankton bloom variability in the Barents, Norwegian, and North Seas. The study provided climatology, trends and interannual variability of the bloom phenology, i.e., when spring and summer blooms start, what intensity they reach, when they reach the highest biomass, and how long they last. Over the last two decades, the summer blooms are getting delayed and are lasting longer. Potential mechanisms driving the interannual variability were also investigated for the different ocean regions. The surface mixed layer depth stands out as the most important predictor of spring blooms onset and maximum biomass, and it can explain more than half of the interannual variability. This paves the way for potentially predicting changes in the bloom phenology for our region of interest, using both observations and modelling results. The Norwegian Food Safety Authority operates the 'Blåskjell Warning Service', issuing advice on toxins in mussels along the entire Norwegian coast to prevent seafood poisoning. The biomass of the Dinophysis species (a type of algae), as well as toxin concentration, are routinely monitored and available. Our follow-up study aims to predict harmful algae blooms of Dinophysis in the northern part of Norway. The main objective is to investigate the potential of a prediction system using a machine learning technique (Support vector machine) that uses in-situ data from past harmful bloom events and satellite remote sensing data of environmental conditions. The key environmental factors found to influence variations of harmful algae blooms for the specific region are sea surface temperature, photosynthetically available radiation, and surface wind speed. We tested how well the developed prediction system performed based on past independent events. The system is demonstrated to be skilful - being more accurate than both climatological and persistence forecasts - for up to 3 weeks. The identified environmental factors were found to be most important to predict the timing of the blooms, while the initial biomass is most important for predicting how strong the blooms will be. The results have been presented at several workshops and a manuscript is under review in Harmful algae journal. This activity has initiated a pilot project within the Norwegian Centre for Research-based Innovation 'Climate Futures' in collaboration with the Institute of Marine Research (IMR). Its objectives are i) to explore the potential of the tool with IMR, which leads the monitoring program and tailors predictions to the stakeholders' needs, ii) to explore measurements that can further improve the accuracy of predictions, and iii) to test how well the approach works for other regions. In a third study, the Norwegian Food Safety Authority 'Blåskjell Warning Service' data is used again to identify the sensitivity of each species of harmful algae in Norwegian water to environmental factors (e.g. sea surface temperature, mixed layer depth, ...). A statistical model is built using a neural network, and the model can predict the likelihood of experiencing a harmful algae event. A manuscript is currently being in preparation.

Funding scheme:

STIPINST-Stipendiatstillinger i instituttsektoren