Presenter
Matthew Chantry, ECMWF
Description
Last year ECMWF introduced the AIFS, building on a series of innovative papers which demonstrated the potential of machine learning in creating data-driven weather forecasting models.
Machine learning models typically build on open-source frameworks, such as Tensorflow, PyTorch and JAX which allow communities to work together and build on one another's work. Weather forecasting brings its own challenges for machine learning, with large datasets and spatial-temporal problems.
In this webinar, we will introduce Anemoi, the toolbox from which the AIFS is trained.
We have been developing this toolbox with a view to many users both inside and outside of ECMWF to use the toolbox to train both global and local weather forecasting models. The toolbox aims to allow users to bring their own datasets and customise their ML models.
We will introduce the toolbox, talk through the steps needed to train a data-driven model and explain how Anemoi helps with these tasks. Finally we will outline a roadmap for future Anemoi development.
Attendance
Please note that this webinar is open to employees of ECMWF Member and Co-operating States' Meteorological services.
If you would like to attend the webinar, please complete the registration form in the left-hand side menu. We will alert you of your acceptance to the webinar and send you all the joining information and Microsoft Teams meeting details in due course. A recording will be available for those unable to attend the event.
The webinar will be conducted in English.