Phytoplankton Dynamics
Asses the role of surface layer water coloumn structure, freshwater fluxes and hydrography on modulating phytoplankton dynamics in the East Greenland Sea
We will analyse a comprehensive set of input variables including surface currents, sea surface temperature, salinity, fresh water content, and outputs from initial activities such as 3D stratification. Through a systematic approach, we will first refine and assimilate the available data, then employ classical correlation techniques to identify robust correlations and patterns. Subsequent adjustments to the data will facilitate data reduction and abstraction, enhancing our understanding of the underlying relationships.
Building upon these insights, we will train a Deep Learning (DL) model to predict phytoplankton blooms, leveraging the abstracted input variables for testing and validation. To gain further understanding of the model's workings and validate its predictions, we will utilise Explainable AI techniques, such as saliency maps and feature extraction, to visualise and interpret the learned correlations.
Building upon these insights, we will train a Deep Learning (DL) model to predict phytoplankton blooms, leveraging the abstracted input variables for testing and validation. To gain further understanding of the model's workings and validate its predictions, we will utilise Explainable AI techniques, such as saliency maps and feature extraction, to visualise and interpret the learned correlations.
