Predicting the Future Plankton Community in the High Arctic

Supervisors: The PhD candidate will be supervised by a multidisciplinary team within NORCE and the University of Bergen (UiB) that covers all expertise in this project, including phytoplankton ecology and microbial communities, numerical modelling and machine learning methods. The team: Xabier Davila (xada@norceresearch.no; Kyle Mayers (kyma@norceresearch.no); Filippa Fransner (UiB).

Project Background. Predicting changes in the Arctic Ocean marine planktonic ecosystem composition and their impact on biogeochemical cycles would be invaluable in supporting climate adaptation and sustainable development under reduced ice-cover conditions. However, to what extent the state of the Arctic planktonic ecosystem is predictable remains uncertain. The Arctic Ocean will likely be ice-free by 2050 (Jahn et al., 2024). While model predictions can be useful to identify key stressors, the taxonomic complexity of phytoplankton assemblages within diverse Arctic marine ecosystems remains difficult to model correctly. At the same time, observational datasets, which represent the changes in the taxonomic complexity, have limited coverage and are difficult to link to stressors (Ardyna and Arrigo 2020). In this project, the PhD candidate will benefit from the advantages of both models and observations, by combining them using machine learning methods.

Project Aims and Methods. The project seeks to answer three key questions:

i) What key processes control the phytoplankton community composition in the Arctic?

ii) How do those key processes affect the entire microbial ecosystem and how do they impact biogeochemistry?

iii) How will the Arctic planktonic ecosystem look like under ice-free conditions?

To answer these questions the PhD candidate will combine state-of-the-art models, a new observational dataset and novel machine learning techniques to maximize the advantage of each method and overcome their respective limitations.

The model. The candidate will analyse the 3D model output from the ECCO-Darwin model (Carroll et al., 2020) that includes five large-to-small phytoplankton functional types (diatoms, other large eukaryotes, Synechococcus, and low- and high-light adapted Prochlorococcus), along with two zooplankton types that graze preferentially on either large eukaryotes or small picoplankton. ECCO-Darwin also assimilates and models physical and biogeochemical variables. ECCO-Darwin will be used to understand the impact of different environmental stressors to the modelled phytoplankton groups and will answer question i).

The observational dataset. Over multiple research cruises between Svalbard and the Central Arctic Ocean (2014-2023), abundances of different phytoplankton types (dinoflagellates, nanophytoplankton, picophytoplankton, heterotrophic flagellates, in addition to those modelled by ECCO-Darwin) and even bacteria and viruses. Observations expand over multiple years, seasons and depths. Environmental variables such as temperature, salinity and nutrients measured at the same locations. These field observations will be used to train a machine learning model about the relationships between the different plankton groups and environmental variables. The trained machine learning model will be then applied to the ECCO-Darwin output to extrapolate its governing processes into higher complexity phytoplanktonic community, bacteria and viruses, thus filling the spatio-temporal gaps of the observational dataset and answering question ii).

The future Arctic phytoplankton community. The understanding that the PhD candidate will gain from answering questions i) and ii) will be applied to explore how the future Arctic phytoplankton community will look like. The trained machine learning model will be applied to future predictions from Earth System Models, potentially in combination with numerical model experiments (including the Norwegian Earth System Model). With this approach, the PhD candidate will model a future ice-free Arctic Ocean phytoplankton community and answer question iii).

Qualifications. The PhD candidate should have a background in earth system sciences, oceanography, biology, mathematics or similar. The candidate should be experienced in Python, Julia or Matlab (or others) and know how to analyse model outputs (4D datasets). Knowledge in machine learning methods, Arctic oceanography and phytoplankton communities, as well as phytoplankton counting methods would be an advantage.

The Training. The candidate will gain diverse experience in analysing model output and observations, as well as machine learning method and numerical modelling (depending on the chosen pathway). In addition, the candidate will gain experience in field work regarding phytoplankton counting methods (flow cytometry, pigment, microscopy, etc.) in preparation for joining one of the Arctic Ocean 2050 cruises.

References

Jahn, A., Holland, M.M. & Kay, J.E. Projections of an ice-free Arctic Ocean. Nat Rev Earth Environ 5, 164–176 (2024). https://doi.org/10.1038/s43017...

Ardyna, M., Arrigo, K.R. Phytoplankton dynamics in a changing Arctic Ocean. Nat. Clim. Chang. 10, 892–903 (2020). https://doi.org/10.1038/s41558...

Carroll, D., Menemenlis, D., Adkins, J. F., Bowman, K. W., Brix, H., & Dutkiewicz, S., et al. (2020). The ECCO-Darwin data-assimilative global ocean biogeochemistry model: Estimates of seasonal to multidecadal surface ocean pCO2 and air-sea CO2 flux. Journal of Advances in Modeling Earth Systems, 12, e2019MS001888. https://doi.org/10.1029/2019MS...