17:15 〜 19:15
[AAS03-P16] High Resolution Numerical Simulation of Tornado Outbreak associated with Typhoon ShanShan 2024 using Supercomputer Fugaku
キーワード：竜巻、スーパーセル、アウターレインバンド、スーパーコンピュータ
Tornadoes associated with tropical cyclones cause significant damage in various countries, making their understanding and numerical forecast are a crucial challenge. This study aims to forecast tornadoes by a numerical model and conducts high-resolution numerical simulations at a horizontal resolution of 80 m in across the entire tropical cyclone using large-scale parallel computing on the Fugaku supercomputer.
The case study focuses on the tornado outbreak that occurred in the Kyushu region of Japan in association with Typhoon ShanShan 2024. This event was one of the most significant tornado outbreaks in Japan, with over ten instances of severe wind damages, including five tornadoes classified as the Japan Enhanced Fujita Scale (JEF) 2, with estimated wind speeds exceeding 65 m/s.
Numerical simulations were conducted using the Cloud Resolving Storm Simulator (CReSS, Tsuboki 2023), including a horizontal 500 m grid simulation and a one-way nested horizontal 80 m grid simulation, both of simulation covering the entire typhoon including outer rainband.
For this experiment, 8,192 nodes of the Fugaku supercomputer (approximately 5% of the total nodes) were utilized. To accelerate computation in large-scale parallel processing, simulation processing mapping for Fugaku's server network structure was optimized and overlapped execution of computation and file output were implemented.
As a result, mini-supercells continuously formed in the typhoon’s outer rainbands, and strong vertical vorticity (>0.5 /s) on the tornado scale was detected in the hook echo regions of multiple mini-supercells. Additionally, the 4-hour forecast time simulation and united files output was completed in approximately 1.5 hours elapsed time. These results indicate that predicting tornado forecast in associated with tropical cyclones are feasible with a sufficient lead time of more than two hour.
The case study focuses on the tornado outbreak that occurred in the Kyushu region of Japan in association with Typhoon ShanShan 2024. This event was one of the most significant tornado outbreaks in Japan, with over ten instances of severe wind damages, including five tornadoes classified as the Japan Enhanced Fujita Scale (JEF) 2, with estimated wind speeds exceeding 65 m/s.
Numerical simulations were conducted using the Cloud Resolving Storm Simulator (CReSS, Tsuboki 2023), including a horizontal 500 m grid simulation and a one-way nested horizontal 80 m grid simulation, both of simulation covering the entire typhoon including outer rainband.
For this experiment, 8,192 nodes of the Fugaku supercomputer (approximately 5% of the total nodes) were utilized. To accelerate computation in large-scale parallel processing, simulation processing mapping for Fugaku's server network structure was optimized and overlapped execution of computation and file output were implemented.
As a result, mini-supercells continuously formed in the typhoon’s outer rainbands, and strong vertical vorticity (>0.5 /s) on the tornado scale was detected in the hook echo regions of multiple mini-supercells. Additionally, the 4-hour forecast time simulation and united files output was completed in approximately 1.5 hours elapsed time. These results indicate that predicting tornado forecast in associated with tropical cyclones are feasible with a sufficient lead time of more than two hour.