10:15 AM - 10:30 AM
[AAS07-06] A 1024-Member Data Assimilation and Forecast Experiment with NICAM-LETKF Using Fugaku: A Heavy Rainfall Event in Kyushu in July 2020
Keywords:Data assimilation, Fugaku, Large ensembles, Predictability
Recently, the occurrence of severe heavy rainfall has been increasing. More accurate predictions are essential to reduce the risk of disasters caused by heavy rainfall. In July 2020, heavy rainfall caused severe disasters in Japan. At that time, heavy rainfall occurred across a wide area of western and eastern Japan, and total precipitation amounts reached more than half of the annual average in some parts of Kyushu from 3 to 14 July 2020. This study aims to investigate the predictability and cause of this heavy rainfall event. We use a global numerical weather prediction system composed of the NICAM (non-hydrostatic icosahedral atmospheric model: Satoh et al. 2014) and the LETKF (local ensemble transform Kalman filter: Hunt et al. 2007; Miyoshi and Yamane 2007). The NICAM-LETKF has been optimized for Fugaku to perform at high resolution and large ensemble sizes.
We performed ensemble data assimilation and forecast experiments using the NICAM-LETKF with 1,024 members and 56-km-horizontal resolution using 2,048 nodes of Fugaku. The results show that the large ensemble size contributes to significantly more accurate analysis compared with a typical ensemble size of less than 100. In addition, 1,024-member ensemble forecasts capture the probability of heavy rainfall in Kyushu about 5 days before it happens, although 10-day-lead forecast is difficult. Ensemble-based lag-correlation analyses with the large ensemble size show very small sampling errors in the correlation patterns and show that the moist air inflow in the lower troposphere and the location of a trough in the upper troposphere play an essential role in causing the heavy rainfall in Kyushu.
We performed ensemble data assimilation and forecast experiments using the NICAM-LETKF with 1,024 members and 56-km-horizontal resolution using 2,048 nodes of Fugaku. The results show that the large ensemble size contributes to significantly more accurate analysis compared with a typical ensemble size of less than 100. In addition, 1,024-member ensemble forecasts capture the probability of heavy rainfall in Kyushu about 5 days before it happens, although 10-day-lead forecast is difficult. Ensemble-based lag-correlation analyses with the large ensemble size show very small sampling errors in the correlation patterns and show that the moist air inflow in the lower troposphere and the location of a trough in the upper troposphere play an essential role in causing the heavy rainfall in Kyushu.