[A-AS08] 高性能計算で拓く気象・気候・環境科学

[AAS08-08] Challenge to Kilometer-scale Atmospheric Climate Simulations with NICAM

*高須賀 大輔^1,2、小玉 知央²、末松 環³、大野 知紀⁴、山田 洋平²、清木 達也²、八代 尚⁵、中野 満寿男²、三浦 裕亮⁶、野田 暁²、那須野 智江²、宮川 知己¹、升永 竜介² (1.東京大学大気海洋研究所、2.海洋研究開発機構、3.理化学研究所計算科学研究センター、4.気象研究所、5.国立環境研究所、6.東京大学大学院理学系研究科地球惑星科学専攻)

キーワード：全球雲解像モデル、気候シミュレーション

The recent increase in computing power enables us to perform a global kilometer-scale (K-scale) “convection-resolving” model (GCRM) simulation on the climate time scale. While this approach is expected to reduce some uncertainties related to unresolved cloud processes in the climate system, we also found systematic model biases in both climatological statistics and variabilities, irrespective of grid spacing at least for O(1-10)-km scale. Considering that this situation partly comes from the poor constraint of a moisture-convection relation depending on model physics at a K-scale, we have revised the treatment of cloud microphysics and unresolved turbulent diffusion in the Nonhydrostatic ICosahedral Atmospheric Model (NICAM) for improved climate simulations on the K-scale. After this revision, we are now performing a 10-year AGCM-type simulation at 3.5-km horizontal mesh from January 2011 on the supercomputer “Fugaku”, and we have accomplished about 5-year integration. A preliminary analysis reveals that both the model revisions and 3.5-km horizontal resolution can seamlessly improve the reproducibility of atmospheric fields over a wide range of spatio-temporal scales: the climatological mean precipitation and OLR distributions, mid-latitude circulations, the Asian summer monsoon, spontaneous MJO realization, tropical cyclone intensity, and precipitation diurnal cycle. However, we also face challenges associated with the poor representation of low clouds in higher resolution simulations, which should be resolved in the ocean-coupled climate run in terms of the energy balance. This work can be a clue for understanding the merits and issues of K-scale GCRM climate simulations and providing a possible strategy for the future model development.