ALBERT

Description: Models from phase 5 of the Coupled Model Inter-comparison Project (CMIP5) show substantial biases in the deep ocean that are larger than the level of natural variability and the response to enhanced greenhouse gas concentrations. Here, we analyze the influence of horizontal resolution in a hierarchy of five multi-resolution simulations with the AWI Climate Model (AWI-CM), the climate model used at the Al-fred Wegener Institute, Helmholtz Centre for Polar and Ma-rine Research, which employs a sea ice–ocean model com-ponent formulated on unstructured meshes. The ocean grid sizes considered range from a nominal resolution of ∼ 1◦ (CMIP5 type) up to locally eddy resolving. We show that increasing ocean resolution locally to resolve ocean eddies leads to reductions in deep ocean biases, although these im-provements are not strictly monotonic for the five different ocean grids. A detailed diagnosis of the simulations allows to identify the origins of the biases. We find that two key re-gions at the surface are responsible for the development of the deep bias in the Atlantic Ocean: the northeastern North Atlantic and the region adjacent to the Strait of Gibraltar. Furthermore, the Southern Ocean density structure is equally improved with locally explicitly resolved eddies compared to parameterized eddies. Part of the bias reduction can be traced back towards improved surface biases over outcrop-ping regions, which are in contact with deeper ocean layers along isopycnal surfaces. Our prototype simulations provide guidance for the optimal choice of ocean grids for AWI-CM to be used in the final runs for phase 6 of CMIP (CMIP6) and for the related flagship simulations in the High Resolution Model Intercomparison Project (HighResMIP). Quite remarkably, retaining resolution only in areas of high eddy activity along with excellent scalability characteristics of the unstructured-mesh sea ice–ocean model enables us to per-form the multi-centennial climate simulations needed in a CMIP context at (locally) eddy-resolving resolution with a throughput of 5–6 simulated years per day.

Description: Ocean model biases such as the North West corner cold bias connected to the location of the Gulf Stream path, the warm bias in upwelling zones, the warm bias in the Southern Ocean, and model drift like the deep ocean warm bias which tends to peak in around 800 to 1000 m depth in the Atlantic Ocean are issues common among state-of-the-art ocean models. These issues are often amplified when the ocean model is coupled to an atmosphere model to perform climate simulations. Furthermore, unrealistic freezing of the Labrador Sea is an issue in various climate models. With the unstructured mesh approach in our Finite Element Sea ice Ocean Model (FESOM) we are able to systematically investigate the benefits of local refinement of the ocean model grid both in an uncoupled set-up (sea-ice ocean only) as well as in a fully coupled climate model (atmosphere- land-sea ice-ocean). While the horizontal ocean model resolution is 25 km on average in the finer grids, we refine the grids in some key areas to up to 5 km. Therefore we can explicitly resolve ocean eddies and simulate eddy-mean flow interactions in these key areas. The atmosphere-land component of our AWI-CM (Alfred Wegener Institute Climate Model) is ECHAM6-JSBACH developed at the Max-Planck-Institute for Meteorology in Hamburg, Germany. Here we present results of century-long uncoupled and coupled simulations on ocean model grids with different local refinements while keeping the atmosphere resolution constant in the coupled simulations. Results indicate that high horizontal resolutions in key regions such as the Gulf Stream / North Atlantic Current area or the Agulhas Stream can reduce biases such as the North West corner cold bias and the deep ocean model drift.

Description: A study of the scalability of the Finite-volumE Sea ice–Ocean circulation Model, Version 2.0 (FESOM2), the first mature global model of its kind formulated on unstructured meshes, is presented. This study includes an analysis of the main computational kernels with a special focus on bottlenecks in parallel scalability. Several model enhancements improving this scalability for large numbers of processes are described and tested. Model grids at different resolutions are used on four high-performance computing (HPC) systems with differing computational and communication hardware to demonstrate the model's scalability and throughput. Furthermore, strategies for improvements in parallel performance are presented and assessed. We show that, in terms of throughput, FESOM2 is on a par with state-of-the-art structured ocean models and, in a realistic eddy-resolving configuration (1/10∘ resolution), can achieve about 16 years per day on 14 000 cores. This suggests that unstructured-mesh models are becoming very competitive tools in high-resolution climate modeling. We show that the main bottlenecks of FESOM2 parallel scalability are the two-dimensional components of the model, namely the computations of the external (barotropic) mode and the sea-ice model. It is argued that these bottlenecks are shared with other general ocean circulation models.