behnam-zakeri/awesome-weather-models

🌦️ A catalogue and categorization of AI-based weather forecasting models.

🌦️ awesome-weather-models

A catalogue and categorization of AI-based weather forecasting models.

Why this Page?

This page provide a catalogue and categorization of AI-based weather forecasting models. The aim is that this page will enable discovery and comparison of the different available model options.

Weather Models

The weather models are categorized according metadata found in the JSON schema specification (schema_ai_models.json). The table below (in alphabetical order) is extracted from the full categorization with columns defined as:

• Name: Name of the weather model.
• Organization: Organization that developed the weather model.
• Operational Data: If forecast data from the model is provided at an operational basis.
• Open Source: If the source code is provided as open source.
• Open Weights: If the model weights are provided as open weights.

Click the link of the model name to see the full model categorization.

Name	Organization	Operational Data	Open Source	Open Weights	Links
AIFS	ECMWF	✅ CC BY 4.0	❌	❌	[paper], [access]
ARCHESWEATHER‑L	INRIA	❌	✅ MIT	✅ MIT	[code], [paper]
ARCHESWEATHER‑M	INRIA	❌	✅ MIT	✅ MIT	[code], [paper]
ARCHESWEATHER‑S	INRIA	❌	✅ MIT	✅ MIT	[code], [paper]
Aurora	Microsoft	❌	✅ MIT	✅ MIT	[code], [paper], [docs], [pypi]
ClimaX‑H	Microsoft	❌	✅ MIT	❌	[code], [paper], [docs]
ClimaX‑L	Microsoft	❌	✅ MIT	❌	[code], [paper], [docs]
FengWu	OpenEarthLab	❌	✅ MIT	✅ None	[code], [paper]
FourCastNet	Nvidia	❌	✅ BSD 3-Clause	✅ BSD 3-Clause	[code], [paper]
GraphCast	Google-DeepMind	❌	✅ APACHE-2.0	✅ CC BY-NC-SA 4.0	[code], [paper], [blog]
MET Norway	MET Norway	❌	❌	❌	[paper]
Pangu‑Weather	Huawei	❌	✅ CC BY-NC-SA 4.0	✅ CC BY-NC-SA 4.0	[code], [paper]
Prithvi WxC	IBM and NASA	❌	✅ MIT	✅ MIT	[code], [paper], [weights]

How to Contribute?

Contributions are much welcome!

• Add a model to the list
• Update categorization and links
• Feedback on categorization/structure to make it more useful

Feel free to make a PR or issue and we will incorporate it.

For making PRs, here is how to make changes to the repo:

1. All updates/changes should be done to the following files:

• data_ai_models.json for updates/changes to model categorization.
• schema_ai_models.json for schema changes.
• README_no_table.md for changes to the README.

2. When updates/changes are completed run python validate_convert_insert.py. Make sure all JSON validations checks pass.

3. The files ai_model.md and ´README.md´ will be auto-generated from the script.

4. Add all the changed and generated files and submit the PR.