Ocean modelling

The strength of unstructured meshes lies in their multi-resolution ability. First, this can be used with a purely geometrical focus in mind, as to resolve the Canadian Archipelago (Wekerle 2013), ice cavities around the Antarctic, the Gibraltar or the straits connecting the Baltic and North seas. Second, many processes in the ocean depend on eddies, generated by various instabilities. Eddies contribute into lateral spreading of watermasses and also shape the mean currents through eddy-mean flow interactions, especially in the vicinity of topography. In almost any place of the ocean the presence of eddies is indispensable, the prominent examples being the Gulf Stream separation and path, and the simulation of the North-West corner. 
The  ability of the code to simulate eddies depends on the mesh resolution which has to provide several mesh element per the internal Rossby radius. A closely related aspect is that the viscosity and diffusivity of numerical codes are always selected so that the codes are able to dump the variance at the grid scale to warrant model stability. Local refinement implies the reduction in dissipation, and this alone may lead to an improvement even if the Rossby radius is only marginally resolved (an eddy-permitting resolution). This is the direction of effective nesting, and the unstructured-mesh capability of FESOM allows one to carry out long locally eddy-resolving or eddy-permitting simulations very efficiently.
In the context of ocean modeling, the outstanding question is the path of Atlantic Wwater (AW) in the Arctic ocean, and its fate in future climate (“Antlantification of Arctic Atlantification”). Resolution of 4.5 km is only eddy- permitting for the Arctic, but the accompanying reduction in dissipation contributes immensely in to the ability to model the AW layer (Wang et al 2017, Wekerle et al. 2017). Locally resolving eddies in Fram strait leads to essential the ability to simulate the observed watermass properties there (Wekerle, 2017, submitted).

In the context of climate modeling, the challenge is in correctingto correct surface temperature biases by using locally eddy resolving meshes (Sein et al., 2016, Sein et al. 2017, submitted). What is the most efficient way of using model resolution, while keeping the total cost at the a level that is affordable for climate simulations? It was proposed to rely on the observed eddy-variability pattern, but our group this approach should will be further elaborate this approach to reach an even better efficiency.