Copyright : Laboratoire LEMAR- 2018
Laurent Memery (50%), Ronan Fablet (50%)
CNRS INEE
Context. Major player in climate and biogeochemical cycles, the Ocean Carbon Pump represents the processes that regulate the absorption and storage of atmospheric CO2 in the deep ocean. Until the 2000s, observations of the deep ocean carbon pump were limited to a few dozen sediment traps on fixed moorings, unable to strongly constrain biogeochemical models. Since the 2010s, deep ocean observation has undergone a tremendous qualitative and quantitative evolution. In fact, new high-frequency devices on autonomous platforms are shedding light on processes that take place in the ocean depths, which have been difficult to access until now. Together with acoustic data, the observations (including imaging – UVP) obtained with Argo floats are new data, non-existent until now, that should constrain these little considered processes (more specifically particle dynamics and zooplankton distribution).
Objectives. The main objective of this thesis is to develop a new and efficient methodology to better constrain the parameters of BGC models, in particular for biogeochemical processes in the mesopelagic layer, using these emerging observations. New approaches are needed. They must simplify the models while retaining the relevant processes and scales, develop efficient tools to allow exchanges between heterogeneous observations and models, and rigorously quantify uncertainties. In this regard, Artificial Intelligence has recently become an emerging field in oceanography. Overall, the differentiable emulators can bridge the gap between current operational systems and AI approaches. The proposed approach will first rely on Observing System Simulation Experiments (OSSEs) to design and evaluate the proposed methodology before its application to real data. The proposed methodology will be applied to the NEMO/LIM3/PISCES model, considered as a model of intermediate complexity
2025
Chibido
