Machine Learning for Precipitation Nowcasting from Radar Images

Shreya Agrawal, Luke Barrington, Carla Bromberg, John Burge, Cenk Gazen, Jason Hickey

arXiv preprint · 11 December 2019 · Google Research

Why it matters. Early influential UNet radar nowcasting result beating HRRR and optical flow at 1 km; precursor to later operational nowcasting models.

In one sentence. A U-Net treats 1-hour, 1 km precipitation nowcasting as image-to-image translation and outperforms optical flow, persistence and HRRR.

Abstract

High-resolution nowcasting is an essential tool needed for effective adaptation to climate change, particularly for extreme weather. As Deep Learning (DL) techniques have shown dramatic promise in many domains, including the geosciences, we present an application of DL to the problem of precipitation nowcasting, i.e., high-resolution (1 km x 1 km) short-term (1 hour) predictions of precipitation. We treat forecasting as an image-to-image translation problem and leverage the power of the ubiquitous UNET convolutional neural network. We find this performs favorably when compared to three commonly used models: optical flow, persistence and NOAA's numerical one-hour HRRR nowcasting prediction.

Labels

Labels and summary are generated by Claude and reviewed weekly. Wrong? Edit this entry.