Radar echo extrapolation based on deep learning is an important method for conducting precipitation nowcasting. Radar echo sequence data have spatiotemporal correlations and nonrigid movements of the radar echo. According… Click to show full abstract
Radar echo extrapolation based on deep learning is an important method for conducting precipitation nowcasting. Radar echo sequence data have spatiotemporal correlations and nonrigid movements of the radar echo. According to the characteristics of radar data, this study proposes a new spatiotemporal fusion neural network called STUNNER. STUNNER implements a two-stream spatiotemporal fusion strategy to extract and fuse spatial and temporal signals. Specifically, it uses a novel cross-network embedding method to achieve efficient spatiotemporal fusion; the fusion integrates a temporal differencing network (TDN) and a spatiotemporal trajectory network (STTN). The TDN models the high-order nonstationarity of the radar sequence to learn the motion trend. The STTN optimizes the convolution operator in a spatiotemporal long short-term memory model to extract the transient variations from the radar images. We compare STUNNER with the other five models on two public datasets. On the radar echo extrapolation task, STUNNER achieves the best performance.
               
Click one of the above tabs to view related content.