Deep learning for twelve hour precipitation forecasts.
Lasse Espeholt, Shreya Agrawal, Casper Sønderby, Manoj Kumar, Jonathan Heek, Carla Bromberg, Cenk Gazen, Rob Carver, Marcin Andrychowicz, Jason Hickey, Aaron Bell, Nal Kalchbrenner
Author Information
Lasse Espeholt: Google Research, Google Inc, 1600 Amphitheatre Pkwy, Mountain View, CA, 94043, USA. lespeholt@google.com.
Shreya Agrawal: Google Research, Google Inc, 1600 Amphitheatre Pkwy, Mountain View, CA, 94043, USA.
Casper Sønderby: Google Research, Google Inc, 1600 Amphitheatre Pkwy, Mountain View, CA, 94043, USA.
Manoj Kumar: Google Research, Google Inc, 1600 Amphitheatre Pkwy, Mountain View, CA, 94043, USA.
Jonathan Heek: Google Research, Google Inc, 1600 Amphitheatre Pkwy, Mountain View, CA, 94043, USA.
Carla Bromberg: Google Research, Google Inc, 1600 Amphitheatre Pkwy, Mountain View, CA, 94043, USA.
Cenk Gazen: Google Research, Google Inc, 1600 Amphitheatre Pkwy, Mountain View, CA, 94043, USA.
Rob Carver: Google Research, Google Inc, 1600 Amphitheatre Pkwy, Mountain View, CA, 94043, USA. ORCID
Marcin Andrychowicz: Google Research, Google Inc, 1600 Amphitheatre Pkwy, Mountain View, CA, 94043, USA.
Jason Hickey: Google Research, Google Inc, 1600 Amphitheatre Pkwy, Mountain View, CA, 94043, USA.
Aaron Bell: Google Research, Google Inc, 1600 Amphitheatre Pkwy, Mountain View, CA, 94043, USA.
Nal Kalchbrenner: Google Research, Google Inc, 1600 Amphitheatre Pkwy, Mountain View, CA, 94043, USA. nalk@google.com. ORCID
Existing weather forecasting models are based on physics and use supercomputers to evolve the atmosphere into the future. Better physics-based forecasts require improved atmospheric models, which can be difficult to discover and develop, or increasing the resolution underlying the simulation, which can be computationally prohibitive. An emerging class of weather models based on neural networks overcome these limitations by learning the required transformations from data instead of relying on hand-coded physics and by running efficiently in parallel. Here we present a neural network capable of predicting precipitation at a high resolution up to 12 h ahead. The model predicts raw precipitation targets and outperforms for up to 12 h of lead time state-of-the-art physics-based models currently operating in the Continental United States. The results represent a substantial step towards validating the new class of neural weather models.
