Jump to Content

Machine Learning for Precipitation Nowcasting from Radar Images

Carla L. Bromberg
Cenk Gazen
Jason J. Hickey
John Burge
Luke Barrington
Machine Learning and the Physical Sciences Workshop at the 33rd Conference on Neural Information Processing Systems (NeurIPS) (2019), pp. 4
Google Scholar


In recent years, Deep Learning techniques have shown dramatic promise in many domains, including the geosciences. We continue this trend by investigating the application of Deep Learning techniques to the problem of \emph{precipitation nowcasting], i.e., the short-term prediction of precipitation at high spatial resolution. We treat forecasting as a image-to-image translation problem, and leverage the power of the ubiquitous UNET autoencoder to make our predictions. We find our straight-forward approach performs favorably to the commonly used HRRR numerical nowcast. Such numerical methods provide strong longer-term predictions (e.g., next-day predictions), but due to their computational complexity, struggle to make effective short-term predictions--an issue deep learning techniques don’t suffer from.