Guo-Yuan Lien, RIKEN Advanced Institute for Computational Science
Radar data assimilation at sub-kilometer scales
CoauthorsTakemasa Miyoshi, Seiya Nishizawa, Ryuji Yoshida, Hisashi Yashiro, Hirofumi Tomita
Abstract: The assimilation of reflectivity and Doppler velocity observations from meteorological radars has been widely studied and proved useful for analyses and short-range forecasts of convective storms. Most of these studies have been done at model resolution of 2-5 kilometers, which would be sufficient for capturing the general behavior of mesoscale convective systems but not necessarily individual convective cells. On the other hand, the resolution of the radar observation data can usually be higher than the model resolution; in particular, an advanced radar such as the phased array weather radar (PAWR) can scan the 3-dimensional volume at resolution as high as 100 meters. Therefore, to make use of these observation data in numerical weather prediction, there is an imperative need to explore the data assimilation at resolution comparable to the high-resolution radar observations.
We develop the SCALE-LETKF system to study the high-resolution data assimilation. It couples the Local Ensemble Transform Kalman Filter (LETKF) with a newly developed regional mesoscale model, the Scalable Computing for Advanced Library and Environment (SCALE)-LES model. Using this system, we assimilate the Osaka University PAWR data into a model at resolution ranging from 1 km to 100 m. We will discuss several issues raised by using such very high resolution model and observational data, including the procedure to spin up the high-resolution domains, and the localization and inflation settings. The results show that the relaxation-to-prior-spread method (Whitaker and Hamill 2012) would be a better covariance inflation scheme than other schemes at this scale.