Shigenori Otsuka, RIKEN Advanced Institute for Computational Science
Space-time extrapolation of precipitation with data assimilation
Coauthors
Shunji Kotsuki, Takemasa MiyoshiAbstract:
Spatiotemporal extrapolation is a widely-used technique for precipitation nowcasting. Motions of precipitation systems are estimated using two or more consecutive observations of precipitation patterns by radars or satellites, and the precipitation patterns are extrapolated in time and space to obtain their future patterns. We have implemented a prototype extrapolation system for the JAXA’s Global Satellite Mapping of Precipitation (GSMaP) Near Real Time (NRT) product. Motion vectors based on satellite images have large uncertainties, and applying data assimilation will help improve the spatiotemporal structures of the motion vectors. The extrapolation system is implemented using the local ensemble transform Kalman filter, and tested using the real-case GSMaP global precipitation patterns. The same framework can be applied to ground-based weather radar networks. We will test the data assimilation system using three-dimensional volume scans from a ground-based Phased-Array Weather Radar (PAWR).