In an exciting development for the meteorological community, we are proud to announce that users can now run AI models using ECMWF's open data themselves. This breakthrough democratizes access to cutting-edge weather forecasting technology and opens new possibilities for research, education, and application development.
AI models at your fingertips
Building on our commitment to open science and the success of machine learning models like FourCastNet, Pangu-Weather, GraphCast and FuXi, ECMWF is now empowering users to harness these AI tools directly. This initiative allows researchers, meteorologists and enthusiasts to generate their own forecasts, experiment with model parameters, and contribute to the advancement of weather prediction.
The power of open data and AI
By combining ECMWF's high-quality open data with state-of-the-art AI models, users can:
- generate rapid forecasts on their own computer
- explore multi-model ensemble forecasting techniques with AI
- conduct research on model performance and intercomparison without any accounts/licences.
Getting started
To begin using AI models with ECMWF’s open data:
pip install ai-models pip install ai-models-panguweather # Or another model ai-models panguweather --input ecmwf-open-data
Plugins already exist for Pangu-Weather, FourCastNet (versions 1 and 2), GraphCast, FuXi and, most recently, Aurora. We hope that the creators of new models will build plugins themselves, contributing to accessible open science.
Please note that different licences are attached to each of the models, with many not permitting commercial use. Please check that your use is consistent with the licence of the models that you use.
Looking ahead
We look forward to reading about users sharing their findings, contributing models, and participating in the growing community of AI weather forecasters.
What about the AIFS? Well, you can already get AIFS open data forecasts from our website, but if you want to play with the model yourself, stay tuned for an announcement about that very soon.
Top banner image: NicoElNino/iStock/Getty Images Plus