Application of a Radar Echo Extrapolation‐Based Deep Learning Method in Strong Convection Nowcasting

Strong convection nowcasting has been gaining importance in operational weather forecasting. Recently, deep learning methods have used to meet the increasing requirement for precise and timely nowcasting. One of promising models is convolutional gated recurrent unit (ConvGRU), which proven perform better than traditional strong Despite its encouraging performance, ConvGRU tends produce blurry radar echo images fails model intensities that multi-modal skewed distributions. To overcome these disadvantages, we tested structural similarity (SSIM) multiscale (MS-SSIM) indexes as loss functions. The SSIM MS-SSIM functions are composed luminance, contrast, structure provide more information about intensity, grade, shape echo, can reduce blurring. Due multi-layer downscaling, extracted characteristics, extrapolation was most realistic accurate among all function schemes. Only scheme successfully predicted echoes after 2 h, especially those at rainstorm level.