Data Assimilation Seminar
Dr. Shigenori Otsuka (August 21, 2019, 16:00-)
|Title||RIKEN Nowcast: optical flow and machine learning|
|Abstract||Precipitation nowcasting is widely used for various purposes such as
disaster prevention and economic decision making. Typically,
ground-based radar observations are used as the input, and other data
sources such as numerical weather prediction can also be used. At RIKEN, two optical-flow-based systems are running: Phased-Array Weather Radar (PAWR) three-dimensional nowcast, and Global Satellite Mapping of Precipitation (GSMaP) RIKEN Nowcast (RNC). Recently, deep learning was applied to various data. The optical flow has a limitation due to the Lagrangian persistence assumption and
nonlinear evolution of precipitation systems, where deep learning would be effective. Therefore, we applied the Convolutional Long Short-Term Memory (Conv-LSTM) algorithm to the PAWR data. In this presentation, preliminary results of real-time experiments with Conv-LSTM will be shown.