Parameterization of subgrid-scale processes in finite element sea ice - ocean models

With the continuous improvement of models for better climate studies and predictions, simulating small scale physical processes remain a challenge. From this perspective, subgrid-scale parameterizations are used whenever such processes cannot be explicitly represented and when their inclusion is beneficial to simulations. However, the parameterizations result from approximations of the reality and also present drawbacks in specific circumstances. In order to prevent these parameterizations from shifting to unphysical behaviours, numerical artefacts are frequently used. In models based on the finite element method, these artefacts are distinctive due to the specificity of the formalism itself. The goal of this doctoral thesis is first to find the best compromise between these numerical artefacts and the preservation of physical processes, through the study of subgrid-scale oceanic parameterizations in a model based on the finite element method. Second, a coupled sea ice – ocean model is built from two disctinct models in order to take advantages of each of them, i.e., unstructured meshes and sophisticated representation of sea ice physics. Through this thesis, these different models are successively used in configurations of increasing complexity, in order to understand the impacts of parameterizations on the models physics accuracy and their skills with respect to observations.