Global statistics of tropical convective circulation simulated in the global 220-m mesh simulation

Global nonhydrostatic models that cover the globe with a kilometer (km)-scale mesh have been developed by various organizations worldwide and are expected as next-generation models that can explicitly calculate deep convective clouds. However, it is known that convective upward motions are not sufficiently represented at the km-scale resolution, and the mesh size of O(100m) is required to obtain convergence of upward motions. To understand the limitation of global km-scale models, we investigate how tropical convection changes with the resolution in the global simulations with the sub-km-scales.

We conduct the global atmospheric simulations by the Nonhydrostatic Icosahedral Atmospheric Model (NICAM) for the mesh size of 3.5 km, 1.7 km, 870 m, 440 m, and 220 m using the flagship Supercomputer in Japan, "Fugaku." The global simulation with a mesh size of 220 m was possible by taking full advantage of Fugaku's performance using 81920 nodes. The 3.5 km experiment started on August 1, 2016, the same day as DYAMOND-summer, and the next higher resolution was run using the lower resolution simulation results as initial conditions. We analyzed data on August 5, 2016. We conducted the global 220m simulation for 8 hours.

The resolution dependence of tropical convection was investigated. The precipitation distribution and zonal mean humidity do not change significantly, but the precipitation intensity changes with resolution. For the grid spacing of less than km, it eliminates overconcentration of precipitation, and the rain area widens as the resolution becomes finer. The coarse-grained rainfall distribution is smoother in the sub-km scale model than in the km scale model. The maximum and 99th percentile values of vertical wind speed at grid point scales increase with increasing resolution. However, when horizontally averaged over a few-degree grid, the vertical wind speed decreases, and the circulation becomes weaker with higher resolution. A finer scale convection core is reproduced in the sub-km scale model. We found that the km-scale model may be creating too much convection on large scales.