It is common sense that model biases in ocean models are attributed to insufficient spatial resolution and shortcomings in parameterization, which need to be improved. However, existing capacity of modern supercomputers is still insufficient for running the models at eddy resolving scales. In the recent study, Dimitry Sidorenko and his colleagues from the Alfred Wegener Institute for Polar and Marine Research overcome this problem by developing a model which allows to use variable resolution in the ocean. This means that the resolution can be increased locally only in the regions where eddy dynamics is important while coarse resolution can be employed everywhere else in the ocean. In their recent study the authors validate the new version of the FESOM 2.0 (Finite Element Sea Ice-Ocean Model) and its applicability for simulating preindustrial, present and future climate.
Although climate models are being widely used for predicting climate and climate change, the model uncertainties are still large. Like this, state of the art climate models commonly operate on structured numerical grids and cannot deliver high resolution climate trajectories due to the limited computer resources. At the moment, the so-called nesting approaches, where two models: regional fine and global coarse are running in parallel and exchange information across their boundaries are widely used. This allows to simulate high activity regions with higher resolution (fine regional model) in the otherwise global context (coarse global model). The amount of required computing resources will be thus significantly reduced compared to using the fine resolution everywhere. Nesting approaches, however, have a lot of shortcomings arising from treating the exchange boundaries. That is why there is a growing interest in climate models which are based on so-called unstructured meshes and allow for variable resolution. This means that the resolution can be increased locally only in the regions where eddy dynamics is important while coarse resolution can be employed everywhere else in the ocean. One such model is the FESOM1.4 (Finite Element Sea Ice-Ocean Model).
The FESOM1.4 model has been used in different applications, however, it is relatively expensive and there is still growing interest in faster numerical solutions. This has led to the development of a newer version of FESOM, FESOM2.0. This version exploits Finite Volume discretization and has a three times faster computational rate which allows a throughput similar to that of regular-mesh models, while promising larger mesh flexibility and good scalability characteristics. The new abbreviation reads as follows: Finite-volumE Sea ice–Ocean Model (Compare: Finite Element Sea Ice-Ocean Model).
The aim of the paper from Sidorenko and his colleagues was to evaluate the performance of FESOM2.0 in the climate setting when it is coupled to atmospheric and land-surface components ECHAM6.3. The new setup has been analysed regarding: (1) the climate mean state and long-term drift under pre-industrial climate conditions, i.e. without the effects of the anthropogenic change; (2) the fidelity in simulating the historical warming over the period 1950-2014 in order inspect whether the model is capable to account for the anthropogenic change by comparing the results with data; (3) the influence of the oceanic resolution on the simulated climate and differences between coarse and eddy- resolving ocean configurations. Sidorenko and his colleagues increased the resolution in the dynamically active regions and analysed how the simulated climate changes compared to if it was not undertaken.
“The newer version of the model will allow the scientists to exploit the ultra-high spatial ocean resolution in climate studies at acceptable computational cost” explains Dimitry Sidorenko, who has been one of the main authors of the study.
The authors showed that climate setups with coarse FESOM configurations produce results within the range of existing models proving that FESOM2.0 can be used for climate studies. They also found that the oceanic mesh with a higher resolution has notable improvements regarding the simulation of oceanic sea surface temperature in the Southern Ocean when compared to low-resolution ocean meshes.
It is expected that through using local refinement in the area of the interest more reliable climate and climate change forecasts at acceptable computational cost can be made.
For further interest, the open access paper by Dimitry Sidorenko, Helge Goessling, Nikolay Koldunov, Patrick Scholz, Sergey Danilov, Dirk Barbi, William Cabos, Ozgur Gurses, Sven Harig, Claudia Hinrichs, Stephan Juricke, Gerrit Lohmann, Martin Losch, Longjang Mu, Thomas Rackow, Natalja Rakowsky, Dimitry Sein, Tido Semmler, Xiaoxu Shi, Christian Stepanek, Jan Streffing, Qiang Wang, Claudia Wekerle, Hu Yang and Thomas Jung is available here: https://zenodo.org/record/3545969#.XdKj4y1oQ1I