Alfred Wegener Institute Helmholtz Centre for Polar and Marine Research (AWI), (Germany)
Local Project ID:
HPC Platform used:
Hazel Hen of HLRS
Results from high-resolution simulations with the sea ice-ocean model FESOM, formulated on unstructured meshes, are presented in which ocean eddies are resolved in the North Atlantic region. By resolving ocean eddies, these features are represented by the laws of physics rather than empirical rules of thumb, as done in most existing climate simulations. A comparison with satellite data suggests that the simulated eddy fields start to become indistinguishable from observations, showing that the model passes the climatic Turing Test. It is argued that these high-resolution resolution models have the potential to significantly increase our understanding of how the climate in general and the ocean in particular will be evolve in a warming world.
Despite of substantial progress in the area of climate modelling during the last five decades, state-of-the-art models still show substantial systematic errors. For some aspects, such as the deep North Atlantic Ocean, biases are much larger than the signals (natural variability and climate change) models are meant to predict. In nonlinear systems, such as the climate system, errors of that magnitude are cause for concern.
The resolution of models, which is determined by the size of the grid boxes, is believed to be a major source of such errors. This makes sense, given that processes that are smaller than (at least) twice the size of the grid boxes need to be represented by rules of thumb rather than the laws of physics. However, given that a doubling in resolution usually comes at least with a ten-fold increase in computational costs, resolving mesoscale processes in global climate modelling remains challenging. This is especially true for ocean eddies, whose size ranges from about 25km in the tropics and 10km in mid-latitudes to less than 1km over the shelves of the Arctic.
Given that climate models, which contributed to past IPCC Assessment Reports, employ resolutions of 100km and coarser, ocean eddies mostly still have to be parametrized (“rule of thumb”). Only as of recently is it possible to carry out sufficiently long simulations to explore the roles of ocean eddies in climate by using the latest high-performance computing systems.
In this study, high-resolution simulations with the Finite Element Sea Ice Ocean Model (FESOM) have been carried out to explore the impact of explicitly resolving ocean eddies. FESOM has been developed at the Alfred Wegner Institute Helmholtz Center for Polar and Marine Research and is the first mature global sea ice-ocean model formulated on unstructured meshes.
The “unstructuredness” gives a high degree of mesh flexibility, which allows to enhance resolution in dynamically active regions, while keeping a coarse-resolution setup elsewhere.
In this study, this flexibility is exploited to adjust the mesh size locally such that it matches the typical size of the ocean eddies (Fig 1), which, as a general tendency, decreases from the tropics towards high-latitudes. Furthermore, FESOM shows excellent scalability characteristics, which allows to make efficient use of many compute cores at the same time (here approximately up to 14,000).
FESOM has been run for many years on Hazel Hen of HLRS and thoroughly compared with observational data. Figure 2 shows snapshots of ocean surface currents from satellite data and two FESOM simulations and coarse and high resolution. Clearly, the coarse-resolution configuration fails in representing eddying motion in the North Atlantic, which in the satellite data maximizes along the Gulf Stream-North Atlantic Current extension. Interestingly, the high-resolution simulation manages to represent the key aspects of ocean eddies in the North Atlantic. In fact, one could argue that the high resolution model starts to become indistinguishable from the observation—that is, it starts to pass the “climatic Turing Test” suggested by Palmer (2016). When asked how to assess the skill of artificial intelligence, Allan Turing famously suggested to ask the machines and humans questions, and see whether you can tell them apart by the answers.
The ability of high-resolution models to realistically represent ocean eddies is intriguing, given that is allows, for the first time, to obtain a “space-type” view of the future ocean current and eddy field in a warming world.
In the meantime, a successor of FESOM called FESOM2.0 has been developed. The new dynamical core, which features a finite volume formulation, provides a speed up of a factor of 3, while retaining the same excellent scalability characteristics for O(104) computational cores. Running FESOM2.0 on supercomputers will allow to carry out eddy-resolving, coupled climate simulations with a throughput of 10 years. These kinds of experiments are expected to shed new light on the climate response to increased greenhouse gas concentrations while explicitly accounting for dynamical processes at the mesoscale.
Animations are available through the FESOM youtube.com channel
Danilov, S., D. Sidorenko, Q. Wang and T. Jung, 2017: The Finite-volumE Sea ice–Ocean Model (FESOM2). Geosci. Mod. Dev., 10, 765–789.
Palmer, T.N., 2016: A personal perspective on modelling the climate system. Proceedings of the Royal Society A., 472, doi.org/10.1098/rspa.2015.0772.
Sein, D., N.V. Koldunov, S. Danilov, Q. Wang, D. Sidorenko, I. Fast, T. Rackow, W. Cabos and T. Jung, 2017: Ocean modeling on a mesh with resolution following the local Rossby radius. J. Adv. Model. Earth Sys., 9, 2601–2614.
Sein, D., N.V. Koldunov, S. Danilov, D. Sidorenko, C. Wekerle, W. Cabos, T. Rackow, P. Scholz, T. Semmler, Q. Wang, and T. Jung, 2018: The relative influence of atmospheric and oceanic model resolution on the circulation of the North Atlantic Ocean in a coupled climate model. J. Adv. Model. Earth Sys., 10, 2026–2041, doi.org/10.1029/2018MS001327.
Prof. Dr. Thomas Jung
Alfred Wegener Institute
Helmholtz Centre for Polar and Marine Research (AWI)
Am Handelshafen 12, D-27570 Bremerhaven (Germany)
e-mail: Thomas.Jung [@] awi.de