CREDIT npj Climate and Atmospheric Science Article
CREDIT is an open software platform to train and deploy AI atmospheric prediction models. CREDIT offers fast models that can be flexibly configured both in terms of input data and neural network architecture. The interface is designed to be user-friendly and enable fast spin-up and iteration. CREDIT is backed by the AI and atmospheric science expertise of the MILES group and the NSF National Center for Atmospheric Research, leading to design choices that balance advanced AI/ML with our physical knowledge of the atmosphere.
CREDIT has reached its first stable release with a full set of models, training, and deployment options. It continues to be under active development. Please contact the MILES group if you have any questions about CREDIT.
New to CREDIT? See QUICKSTART.md to get up and running locally, or the full online Quickstart for detailed guidance.
Install CREDIT and launch the interactive config wizard:
conda create -n credit -y python=3.13 uv
conda activate credit
uv pip install miles-credit
credit begincredit begin walks you through creating a starter config, then validates it
with credit check and prints the credit preprocess / credit train
commands to run next. See QUICKSTART.md for the NCAR
Casper/Derecho install paths and the rest of the workflow.
The full documentation covers installation, training and deploying models, the config file schema, datasets, and the gen2 pipeline, along with complete API docs.
If you are interested in using CREDIT as part of your research, please cite the following paper: Schreck, J.S., Sha, Y., Chapman, W. et al. Community Research Earth Digital Intelligence Twin: a scalable framework for AI-driven Earth System Modeling. npj Clim Atmos Sci 8, 239 (2025). https://doi.org/10.1038/s41612-025-01125-6
Model weights for the CREDIT 6-hour WXFormer and FuXi models, the 1-hour WXFormer, and the CAMulator climate emulator are available on huggingface.
Processed ERA5 Zarr Data are available for download through Globus (requires free account) through the CREDIT ERA5 Zarr Files collection.
Scaling/transform values for normalizing the data are available through Globus here.
CREDIT also supports realtime runs generated from deterministic Google Cloud GFS files and raw cube sphere GEFS files.
This software is based upon work supported by the NSF National Center for Atmospheric Research, a major facility sponsored by the U.S. National Science Foundation under Cooperative Agreement No. 1852977 and managed by the University Corporation for Atmospheric Research. Any opinions, findings and conclusions or recommendations expressed in this material do not necessarily reflect the views of NSF. Additional support for development was provided by The NSF AI Institute for Research on Trustworthy AI for Weather, Climate, and Coastal Oceanography (AI2ES) with grant number RISE-2019758, NSF Grant RISE-2425659, and by Schmidt Sciences, LLC.