Models

We set up and perform computationally demanding ocean-sea ice models based on the community codes NEMO and FESOM. Since an important focus of our research is on the importance of mesoscale and submesoscale processes, our configurations not only need to represent the large-scale ocean circulation but also resolve part of the smaller scales. In NEMO, this is realized through an adaptive grid refinement, leading to nests down to the km-scale within global models at coarser resolution. FESOM instead uses a flexible mesh refinement through unstructured triangular volumes where resolution can be flexibly placed wherever necessary.

Our models do not only represent ocean dynamics but also air-sea coupled processes (coupling to atmospheric models, such as OpenIFS and ECHAM, in FOCI), sea-ice dynamics or the impact of the ocean on the uptake of carbon and trace gases (through biogeochemical models). We use idealized configurations such as ocean channels or double gyre configurations to focus on specific oceanic regimes, gain process understanding and further develop ocean and climate models. Hybrid numerical machine learning models are a rapidly developing field of research to which we also contribute in order to improve model capabilities.

Owing to the computational demand, our numerical models are typically performed at high-performance centers such as the Helmholtz-funded Jülich Supercomputing Centre (JSC) and the German Climate Computing Centre (DKRZ), as well as other national computing centers such as the National High Performance Computing (NHR) and the local computing center at Kiel University.