Advanced supercomputers, including the K computer, make possible simulation of the global atmosphere by using a high-resolution global cloud-resolving model with a grid size of less than 1km. Since the global cloud-resolving model still cannot explicitly express sub-grid scale phenomena related to atmospheric small eddies and cloud generation, these phenomena are parameterized even in high-resolution models. The use of the parameterization scheme leads to uncertainty and/or bias in the simulated results. In the future, such uncertainty is expected to be reduced by using a global-LES (large eddy simulation) model relying on basic climate principles.
However, to realize global-LES simulations a number of challenges have to be overcome: understanding the spatial-resolution dependency of LES schemes, the development of a new theoretical basis of LES for real-atmosphere simulation, raising the sophistication level of physical process schemes, improving their computational performance, and preparation of a post-process library for analyzing a huge number of model outputs. To advance global-LES simulations, we will address several climate science targets: hierarchical structure of clouds, and exploration of multiple equilibrium solutions under ideal conditions, as well as regional climate studies under real atmospheric conditions.
