[A-AS07] Weather, Climate, and Environmental Science Studies using High-Performance Computing

[AAS07-05] A 1000-member 14-km mesh global ensemble experiment to target typhoon FAXAI (2019)

*Yohei Yamada¹, Masuo Nakano¹, Tomoki Miyakawa², Chihiro Kodama¹, Ying-Wen Chen², Hisashi Yashiro³, Tomoe Nasuno¹, Masaki Satoh² (1.Japan Agency for Marine-Earth Science and Technology, 2.Atmosphere and Ocean Research Institute, The University of Tokyo, 3.National Institute for Environmental Studies)

Keywords:Large ensamble simulation, high resolution global nonhydrostatic model, Typhoon Faxai 2019

A thousand members ensemble experiment on the Supercomputer Fugaku was conducted by a nonhydrostatic icosahedral atmospheric model (NICAM) with a 14-km mesh, which was targeted to Typhoon Faxai of which the genesis time is 18UTC 4 September 2019. Ensemble simulations started from different initial dates from 26 August 2019 to 4 September (10 days). NICAM-LETKF JAXA Research Analysis (NEXRA) was used as the initial atmospheric condition of the ensemble experiment. The NEXRA provided 100 perturbed atmospheric conditions for each initial date. The total number of ensemble members was 1000 in the present study. The sea surface temperature was calculated by a slab ocean model, which was nudged to NOAA OI SST V2.1. The ensemble experiments ran for about 30 days. We simply regarded a simulated typhoon generated from 00UTC 3 September to 18UTC 7 September within 140E-170E as a Faxai-like typhoon.
In the ensemble experiment, 391 members reproduced the Faxai-like typhoon. Focusing on impact of lead time on the reproducibility of the genesis time and intensity, fewer number of ensemble members with the short lead time reproduced the Faxai-like typhoon than that of members with the long lead time. This is a possible reason why the short lead time is insufficient for a vortex being intensified, which may become a typhoon. As for tracks of the Faxai-like typhoon, these tracks differed between members of which even experiments started from the same initial date. The variations were more prominent in the long lead time experiments than those of the short lead time experiments.