Date depot: 11 avril 2023 Titre: Inversion of vertical profiles of phytoplankton from satellite images Directeur de thèse: Dominique BÉREZIAT (LIP6) Encadrant : Anastase CHARANTONIS (LAMME) Domaine scientifique: Sciences et technologies de l'information et de la communication Thématique CNRS : Images et vision Resumé: Retrieving the vertical profiles of bio-geophysical parameters is crucial in understanding the impacts of the changing climate on the ocean's biodiversity. Such retrieval is an arduous task given that in-situ observations of the vertical distributions are very sparse, and satellite imaging contains proxy information integrated over the top layer of the oceans. There exist numerical models that are plugged into global circulation models that simulate this evolution. These models are parametrized to fit specific time series of such vertical profiles, but they are highly uncertain. As such, the development of new methods that use deep learning advancements in order to combine all sources of information: in-situ, satellite, and numerically modeled, in order to reconstruct the vertical distributions are of big importance to both the bio-geophysics and data science community.