Water Column Structure

The 3D subsurface water column structure will be recontructed from 2D Earth Observations data and in-situ observations using an innovative machine learning methodology. We will develop a Convolutional Neural Network (CNN) framework for the direct reconstruction of salinity and temperature profiles within selected study sites.

Through a comprehensive analysis and systematic evaluation of various input variables, coupled with a rigorous sensitivity analysis, we will provide robust and adaptable solutions for reconstructing essential oceanic variables in East Greenland, which can be further tested for other areas in the Arctic. Additionally, Fresh4Bio will explore the influence of training set size and geographical distribution on reconstructed variables, thereby offering valuable insights into the factors shaping model performance across distinct regions of the Atlantic Ocean.