Use of Deep Learning for Weather Radar Nowcasting

Abstract Nowcasting based on weather radar uses the current and past observations to make estimations of future echoes. There are many types operationally deployed nowcasting systems, but none them currently deep learning, despite it being an active area research in last few years. This paper explores learning models as alternatives methods by proposing different architectures comparing against some operational systems. The proposed here, harnessing residual convolutional encoder-decoder architectures, reach a level performance expected systems certain scenarios can even outperform them. Finally, potential drawbacks using analyzed. No decay geographical from where were trained was found. edge or checkerboard artifact, common operations, found that affects metrics.