Spherical Fourier Neural Operator model checkpoint package. This is a deep learning weather model that predicts 73 atmospheric/surface variables in an autoregressive fashion.
Details
This NGC asset is a Spherical Fourier Neural Operator model checkpoint package for predicting atmospheric dynamics. Model checkpoint package refers to the set of artifacts needed to run inference using pre-trained model which includes the model checkpoint and required metadata and arrays.
Architecture
Please refer to the reference paper to learn about the model architecture. In training this checkpoint, Spherical Fourier Neural Operator (SFNO) is applied to forecasting atmospheric dynamics, and demonstrates stable autoregressive rollouts for a year of simulated time (1,460 steps), while retaining physically plausible dynamics. The SFNO has important implications for machine learning-based simulation of climate dynamics that could eventually help accelerate our response to climate change.
This model predicts the following variables surface variables: "u10m", "v10m", "u100m", "v100m", "t2m", "sp", "msl", "tcwv" and the variables u, v, t, z, q at pressure levels 1000, 925, 850, 700, 600, 500, 400, 300, 250, 200, 150, 100, 50. See the model Lexicon for more details.
Training
The model is trained on a 73-channel subset of the ERA5 reanalysis data on single levels and pressure levels that is pre-processed and stored into HDF5 files. This model was trained in Modulus-Makani
How to use?
This model package can be used with Modulus, Modulus-Makani, and Earth-2 Studio.
References
Spherical Fourier Neural Operators: Learning Stable Dynamics on the Sphere
License
This checkpoint is distributed under the Apache 2.0 license.