Precipitation nowcasting based on radar observations or satellite observations has been used for short-range precipitation predictions for lead times ranging from several tens of minutes to several hours. Conventional methods use algorithms such as cell tracking, cross correlation-based motion vectors, and advection. Recently, various deep learning-based precipitation nowcasting techniques have been proposed. Such techniques include CNN, UNet, LSTM, and GAN. In this presentation, some of these techniques will be reviewed, and an integrated DA-NWP-AI system will be discussed.