We are pleased to announce the launch of the Atmosphere Machine Learning Emulation Challenge (AMLEC), taking place from April 21 to June 30, 2025, as part of the ECML-PKDD 2025 competition program. Led by the Image and Signal Processing Group (ISP) at the University of Valencia (Spain), this challenge focuses on a timely and critical task: emulating Radiative Transfer Models (RTMs) — essential tools in climate and Earth observation sciences — with fast, accurate, and generalizable machine learning models.

🌍 Challenge Goal
Participants are invited to develop surrogate models that can accurately replicate the behavior of computationally expensive RTMs. The task is particularly relevant for hyperspectral satellite missions, where real-time data processing and efficient atmospheric correction are crucial.

📊 Resources & Evaluation
Participants will use RTM datasets provided on HuggingFace, where detailed instructions, evaluation metrics, and submission guidelines are also available. Submissions are evaluated via predefined error metrics. No formal registration is required—just a HuggingFace account to submit your results.

🧠 Why Participate?

• Contribute to physics-aware AI and climate science

• Present your work at ECML-PKDD 2025

• Be part of a research publication summarizing the challenge and top models

🔗 Useful Links

• Challenge Dataset & Info: HuggingFace RTM Emulation

• Follow us for updates on:

  

   

📅 Key Dates

• Challenge opens: April 21, 2025

• Submission deadline: June 30, 2025

We encourage all researchers, students, and practitioners in machine learning, climate science, and remote sensing to join us in this initiative.

For any queries, please contact Jorge Vicent Servera (jorge.vicent@uv.es).

Best regards,
Jorge Vicent Servera, Gustau Camps Valls, Cesar Aybar, Julio Contreras
ISP – University of Valencia