Design of High-Resolution Grid Systems for the Next Generation Coupled Climate System Models and Earth System Models

Shiming Xu,Chungang Chen,Hengbin An,Xingliang Li,Yiyuan Li,J Wright,Kun Xia,Xu Wang,Ye Pu

DOI: https://doi.org/10.3969/j.issn.1009-2412.2017.05.007

2017-01-01

Abstract:The grid system is the basis of the geophyisical flu-id dynamic models.Coupled climate models and the at-mospheric and oceanic components carry out the spatial discretization based on model grids.Therefore the grid is definitive of the spatial resolution of the model, and have direct influence over the model accuracy,the time stepping and the computational performance.High-reso-lution modeling brings both oppurtunity and challenges to climate studies, and how to effectively utilize the hugh number of freedom of degrees that is enabled by high-resolution grids is the central task.For atmospher-ic models,the research frontier is on the design of glob-al quasi-uniform grids and the accompanying algo-rithms, including the icosahedral grid, the cubed-sphere grid,etc.Specifically,the high-order of accura-cy of the algorithm and the mitigation or elimination of grid imprints is the key research focus.In this project we carry out the optimization of quasi-uniform grids and propose the use of multi-moment constrained finite vol-ume method to improve accuracy and reduce grid noise. For the oceanic grid, due to the comparatively higher resolution as required for the atmospheric grid,the main focus of research is the multi-scale modeling and the im-provement of key ocean-atmosphere interaction regions. We propose the use of complex conformal mappings based on Schwarz-Christoffel mapping to generate multi-scale grids,with the co-design of atmospheric and oce-anic components for maritime continent and the Arctic region.These two regions are both key climate regions and important with societal impacts.High-effiency par-allelization of the new grids is carried out to incorporate spatial refinement and dynamic load balancing.The proposed grid generation methods and related algorithms are developed by Chinese scientists, and have advanta-ges over key apsects over counterparts in the world, in terms of accuracy and computational characteristics. Applications to both domestic and widely used models (such as CESM)are also carried out for the verification in terms of both scientific and computational aspects. Through open-source grid generation software, sample grids and supportive datasets, we provide the open use of the aforementioned algorithms.