A Radar Echo Extrapolation Method Based on the ARUR Network with Multi-Source Data

Accurate nowcasting provides key information for disaster weather warning. Nowcasting is mostly based on radar echo extrapolation, where the echoes evolution results from complex interactions among cloud systems and various thermal-dynamic features of the weather background. However,  high spatiotemporal resolution weather background information has not been fully integrated into the existing nowcasting algorithms. In this study, several multiple data fusion units (Radarcells) integrating the radar mosaic and rapid-refresh data are constructed as input for the radar echo extrapolation network architecture designed according to Attention-Resnet Unet with Radarcells (ARUR). In addition, a self-defined loss function combining weighted mean square error and structural similarity index is further incorporated to improve the extrapolation effect. The rapid-refresh data include relative humidity, zonal wind, meridional wind, and vertical velocity within the range of 38.81°N–40.81°N, 115.48°E–117.48°E,  during June–August of 2018–2021. Four ARUR-based models with different Radarcells as input and an ARU-based model (Attention-Resnet Unet, without fusing physical data) are trained for 120-min extrapolation, respectively. The models are evaluated with the critical success index (CSI), probability of detection (POD), and false alarm rate (FAR) based on the test dataset. The results show that the radar  echo prediction by the ARUR-based models is better and less time-consuming than that by the ARU-based models, especially for strong echoes. Under the reflectivity thresholds of 25 and 35 dBZ, the average values of CSI, FAR, and POD calculated by the ARUR-based models are improved by 8.42%, 7.76%, 8.52% and 10.36%, 7.36 %, 9.10% than those by ARU-based, respectively. The results suggest that integrating weather background information can significantly enhance the effect of extrapolation through reducing echo blurriness and better capturing the echoes formation and dissipation compared with the previous deep learning-based models.