R-CCS Cafe 第228回 第3部

Data Assimilation is a statistical method to incorporate observed real-world data to a simulation. Recent progress of atmospheric remote sensing techniques has provided us unprecedently abundant and detailed observation data. For example, phased array weather radar (PAWR) is a state-of-the-art weather radar which can provide reflectivity and doppler velocity observation every 30 seconds, 10 times more frequently than conventional weather radars. Such a rapid update allows us to capture very detailed structure of rapidly evolving localized convective rain. To make the best use of the new PAWR observation, we have developed the SCALE-LETKF, the system combining the regional numerical weather prediction model SCALE-RM and a data assimilation method LETKF. We improved the observation operator, which transforms the model state variables to the variables observed by the PAWR. We also enabled the SCALE-LETKF to assimilate the PAWR observation within 30 seconds with a high resolution and large number of ensemble size on Fugaku. Additionally, data assimilation with the LETKF provides the ensemble of analysis states which can be used to initialize ensemble forecasts to represent the uncertainty of the forecast growing with time. The capability of the SCALE-LETKF for the realtime precipitation forecasting was demonstrated in summer 2020 and 2021.