CaltechAUTHORS
  A Caltech Library Service

FourCastNet: A Global Data-driven High-resolution Weather Model using Adaptive Fourier Neural Operators

Pathak, Jaideep and Subramanian, Shashank and Harrington, Peter and Raja, Sanjeev and Chattopadhyay, Ashesh and Mardani, Morteza and Kurth, Thorsten and Hall, David and Li, Zongyi and Azizzadenesheli, Kamyar and Hassanzadeh, Pedram and Kashinath, Karthik and Anandkumar, Animashree (2022) FourCastNet: A Global Data-driven High-resolution Weather Model using Adaptive Fourier Neural Operators. . (Unpublished) https://resolver.caltech.edu/CaltechAUTHORS:20220714-224617950

[img] PDF - Submitted Version
See Usage Policy.

8MB

Use this Persistent URL to link to this item: https://resolver.caltech.edu/CaltechAUTHORS:20220714-224617950

Abstract

FourCastNet, short for Fourier Forecasting Neural Network, is a global data-driven weather forecasting model that provides accurate short to medium-range global predictions at 0.25∘ resolution. FourCastNet accurately forecasts high-resolution, fast-timescale variables such as the surface wind speed, precipitation, and atmospheric water vapor. It has important implications for planning wind energy resources, predicting extreme weather events such as tropical cyclones, extra-tropical cyclones, and atmospheric rivers. FourCastNet matches the forecasting accuracy of the ECMWF Integrated Forecasting System (IFS), a state-of-the-art Numerical Weather Prediction (NWP) model, at short lead times for large-scale variables, while outperforming IFS for variables with complex fine-scale structure, including precipitation. FourCastNet generates a week-long forecast in less than 2 seconds, orders of magnitude faster than IFS. The speed of FourCastNet enables the creation of rapid and inexpensive large-ensemble forecasts with thousands of ensemble-members for improving probabilistic forecasting. We discuss how data-driven deep learning models such as FourCastNet are a valuable addition to the meteorology toolkit to aid and augment NWP models.


Item Type:Report or Paper (Discussion Paper)
Related URLs:
URLURL TypeDescription
https://doi.org/10.48550/arXiv.2202.11214arXivDiscussion Paper
ORCID:
AuthorORCID
Li, Zongyi0000-0003-2081-9665
Azizzadenesheli, Kamyar0000-0001-8507-1868
Hassanzadeh, Pedram0000-0001-9425-8085
Kashinath, Karthik0000-0002-9311-5215
Anandkumar, Animashree0000-0002-6974-6797
Additional Information:We would like to acknowledge helpful comments and suggestions by Peter Dueben from ECMWF. We thank the researchers at ECMWF for their open data sharing and maintaining the ERA5 dataset without which this work would not have been possible. This research used resources of the National Energy Research Scientific Computing Center (NERSC), a U.S. Department of Energy Office of Science User Facility located at Lawrence Berkeley National Laboratory, operated under Contract No. DE-AC02-05CH11231. A.K. and P. Hassanzadeh were partially supported by ONR grant N00014-20-1-2722. We thank the staff and administrators of the Perlmutter computing cluster at NERSC, NVIDIA Selene computing cluster administrators, Atos, and Jülich Supercomputing Center for providing computing support. JP and KK would like to thank Sanjay Choudhry and the NVIDIA Modulus team for their support. JP, SS, P. Harrington and KK would like to thank Wahid Bhimji for helpful comments.
Funders:
Funding AgencyGrant Number
Department of Energy (DOE)DE-AC02-05CH11231
Office of Naval Research (ONR)N00014-20-1-2722
Subject Keywords:Numerical Weather Prediction · Deep Learning · Adaptive Fourier Neural Operator · Transformer
Record Number:CaltechAUTHORS:20220714-224617950
Persistent URL:https://resolver.caltech.edu/CaltechAUTHORS:20220714-224617950
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:115597
Collection:CaltechAUTHORS
Deposited By: George Porter
Deposited On:15 Jul 2022 23:27
Last Modified:15 Jul 2022 23:27

Repository Staff Only: item control page