P-38

Development of the SCALE-LETKF system for radar data assimilation

G.-Y. Lien (RIKEN), T. Miyoshi (RIKEN), S. Nishizawa (RIKEN), H. Yashiro (RIKEN), R. Yoshida (RIKEN), and H. Tomita (RIKEN)

High resolution regional models are useful for short-range weather prediction and disaster prevention. However, it is not easy to optimally use dense and frequent observations in these models. One of the difficulties is the lack of knowledge of the mesoscale error covariances that relate the observation data to the high-resolution model grids; another difficulty is the large computational cost to process the large amount of data at high frequency and resolution.

Using the highest-end K computer, we explore the use of the local ensemble transform Kalman filter (LETKF) for high resolution numerical weather prediction. We develop a SCALE-LETKF system that couples the Scalable Computing for Advanced Library and Environment (SCALE)-LES model, a large-eddy-simulation weather model, with the LETKF. To run the system for large problems, the parallelization scheme of the LETKF and the data I/O flow of the system are carefully designed to improve the computational scalability. The concept of the SCALE-LETKF system design and the preliminary results of weather radar data assimilation using the system up to the time of the symposium will be presented.