AI & ML interests
Destination Earth (DestinE) is an ambitious initiative of the European Union to create a highly accurate digital replica of Earth exploring coupled earth system machine learning models
Recent Activity
Welcome to Destination Earth on Hugging Face!
Destination Earth (DestinE) is a flagship initiative of the European Commission to develop a highly accurate digital model of the Earth. It is implemented by ECMWF, ESA and EUMETSAT.
This organisation hosts the artificial intelligence (AI) and machine learning (ML) work led by ECMWF and its partners within DestinE. AI and ML play an increasingly central role in DestinE: ML-driven models and AI-based software boost the ability to quantify uncertainty in digital twin simulations and make the DestinE system more interactive, using the supercomputing infrastructure of the EuroHPC Joint Undertaking.
This work builds on ECMWF's own ML efforts with its Member States, in particular:
- Anemoi, the open-source ML framework now powering several state-of-the-art data-driven forecasting models in Europe.
- The Artificial Intelligence Forecasting System (AIFS), ECMWF's operational AI-based weather forecasting system.
- AIFL (AI for Flood Forecasting), ECMWF's global deep learning model for daily streamflow at catchment scale, including in ungauged basins.
Objectives
In Phase II of DestinE, AI activities focus on:
- Developing ML-based Earth system model components, working towards a European ML-based Earth system model.
- Enhancing the Digital Twin Engine to support training and running ML models on EuroHPC systems.
- Applying AI to support uncertainty quantification of digital twin simulations.
- Using AI to make the system more interactive and digital twin data easier to use.
ML Earth system components
ECMWF is developing ML-based Earth system components that will be coupled with the AIFS to form a fully AI-driven Earth system model:
| Component | Description |
|---|---|
| Land | aiLand, a global deep learning emulator of land surface processes, trained on ECMWF's physical land model. |
| Waves | An AI wave model trained on over 40 years of ECMWF physical wave model output, driven by the AIFS. |
| Hydrology | Builds on ECMWF's AIFL model, developed as part of ECMWF's wider ML activities. |
| Sea ice | Learns sea ice and surface ocean evolution from decades of reanalysis data, capturing seasonal cycles and ice-edge behaviour. |
| Ocean | Models the full-depth ocean: temperature, salinity, currents and sea surface height anomaly. |
Learn more
- AI in DestinE: news, videos and podcasts on AI and ML in DestinE.
- AI in DestinE: the explainer
- Forecast-in-a-Box: portable AI forecasting workflows within the Digital Twin Engine.
- AIFS blog series
Get involved
- Training: ECMWF offers advanced ML and AI training, including the Machine Learning for Earth System Modelling course series. See the ML training page.
- AI Weather Quest: an international competition to advance weather forecasting with AI and ML. See the AI Weather Quest.
- Tenders: ECMWF regularly opens invitations to tender for AI and ML development in DestinE. See DestinE procurement and the opportunities to be involved.
Support
For questions about DestinE, please use the DestinE contact page. For the ECMWF tools behind this work:
- To follow announcements and discuss with other users and ECMWF staff, visit the ECMWF User Forum.
- To report a bug or contribute to Anemoi, open a GitHub issue in line with the contribution guidelines.
If you spot an issue specific to our entries on Hugging Face, you may raise it through Hugging Face discussions. Please note that we do not guarantee support through Hugging Face.
Destination Earth is funded by the European Union and implemented by ECMWF, ESA and EUMETSAT.