AI4Carbon: Machine Learning for Carbon Cycle Science
A community effort
Announcements
| Jan 12, 2027 |  Call for Abstracts: submit to our EGU27 session Monitoring the Carbon Cycle with Machine Learning: Opportunities for Science, Policy and Industry – deadline 12 January 2027, 13:00 CET. | ||
|---|---|---|---|
| Nov 19, 2026 |  Our 2nd virtual workshop, The Next Generation of Atmospheric Transport Models?, takes place on Thursday, Nov 19th, 2026 – 4pm CET / 3pm GMT / 10am EST / 7am PST / 11pm AWST. Details & registration | ||
| May 05, 2026 |
| ||
| Jan 15, 2026 |  Please consider submitting an abstract to the EGU26 session Machine learning and hybrid modelling for carbon cycle science, monitoring and carbon market policy – see you in Vienna! | ||
| Jan 10, 2025 | Please consider submitting an abstract to the EGU25 session Machine Learning for Carbon Cycle Research – see you in Vienna! |
EGU26 Update: We have a splinter meeting Tuesday, 5 May, 16:15–18:00 CEST (Room 2.43). SPM74: AI4Carbon: Splinter meeting to discuss potential future work on AI for atmospheric transport and inverse modeling. 🌍 Advancing AI for Carbon Cycle Science
The AI4Carbon Initiative is a community-driven effort to leverage cutting-edge machine learning for understanding and monitoring carbon cycles. We bridge top-down atmospheric inversion approaches with bottom-up ecosystem modeling through AI, advancing both atmospheric transport science and carbon accounting.
Stay Connected
📧 Join Our Mailing List
The Challenge
Artificial intelligence has revolutionized weather prediction and other scientific domains. Yet the carbon cycle—critical for understanding climate and supporting climate policy—remains largely untouched by these AI advances. AI4Carbon works to close this gap by leveraging machine learning across both atmospheric-based (top-down) and ecosystem-based (bottom-up) approaches to carbon science.
🔬 The Top-Down Challenge
Atmospheric inverse modeling relies on coarse-resolution transport models, leading to systematic errors in retrieving surface carbon fluxes. Higher-resolution models exist in research but are computationally prohibitive for operational use.
📊 The Data Gap
Unlike weather prediction, there is no consensus benchmark dataset for training machine learning models on atmospheric CO₂ transport. This lack of standardization hinders progress and collaboration across the research community.
🤝 The Bottom-Up Complement
Ecosystem models and direct measurements provide valuable constraints, but integrating them with atmospheric observations remains challenging. AI can help unify these perspectives by learning patterns across scales and bridging gaps between different data sources.
Our Mission
📈 Build Consensus
Establish benchmark datasets and evaluation frameworks for AI in carbon cycle research.
🔗 Foster Community
Connect researchers across top-down atmospheric science, bottom-up ecosystem modeling, and machine learning through workshops and collaborative projects.
🎯 Drive Impact
Develop and validate AI methods that improve carbon flux estimates and support climate monitoring.
🌐 Enable Action
Provide tools and techniques that support policy-relevant carbon accounting and the Global Greenhouse Gas Watch.
Upcoming Events
Join us at upcoming workshops and conferences to learn more about AI for carbon cycle science:
📅 View All Workshops 📚 Explore Resources
Community
AI4Carbon is an open community, not a fixed committee. Each workshop, session or meeting is organized by whoever wants to drive it, and you can find the organizers of every event on the workshops page. Want to organize something under the AI4Carbon umbrella, or give a talk? Just get in touch!
A big thank you to everyone who has organized, presented, advised or otherwise helped shape AI4Carbon since 2024:
…and every participant in our workshops and sessions. Missing from this list? Let us know!
Contact
If you are interested in joining the collaborative effort, please reach out to vbenson (at) bgc-jena (dot) mpg (dot) de!