Development of acceleration and data assimilation techniques for wind digital twin in urban areas

Onodera, Naoyuki   ; Shimokawabe, Takashi*; Idomura, Yasuhiro   ; Kawamura, Takuma ; Hasegawa, Yuta   ; Ina, Takuya ; Inagaki, Atsushi*; Hirano, Kohin*; Shimose, Kenichi*; Oda, Ryoko*; Wada, Taisuke*

The project goal is to realize real-time wind prediction in urban areas by assimilating observed data into real-time wind simulations on GPU supercomputers. In FY2023, a data assimilation method based on the Local Ensemble Transform Kalman Filter (LETKF) was applied to CityLBM in order to reproduce local wind conditions with high accuracy. We validated the data assimilation method for two-dimensional forced isotropic turbulence. It was confirmed that the LETKF with 64 ensembles provides the same levels of accuracy with 1/16th of the coarse observation points compared to the nudging method. In addition, it was confirmed that the application of LETKF can reproduce the phase of the Kalman vortex with high accuracy in the verification of the flow around a three-dimensional square cylinder.