5th ECMWF-ESA Machine Learning Workshop
Download ProgrammeThe times below are displayed for Europe/London. We have detected that you are visiting us from .
Please select a timezone to show adjusted event times accordingly.
Location: Botte B4 pavilion - DAMA Technopole
Agenda time displays according to the selected time zone.
| Monday, 13 April 2026 | ||
|---|---|---|
| Opening and keynote | ||
| 13:00 to 13:10 | Opening and welcome |
Speaker: Massimo Bonavita (ECMWF)
Speaker: Rochelle Schneider (ESA)
|
| 13:10 to 13:40 |
AI for Humanity: Foundational Machine Learning and Data Assimilation for a Resilient Earth
Keynote speaker |
Speaker: Rossella Arcucci (Imperial College London)
|
| 13:40 to 14:10 |
Machine learning–driven advances in geophysical data assimilation
Keynote speaker |
Speaker: Marc Bocquet (École nationale des ponts et chaussées and ECMWF fellow)
|
| 14:10 to 17:10 | Poster session and ice breaker | |
Location: Botte B4 pavilion - DAMA Technopole
Agenda time displays according to the selected time zone.
| Tuesday, 14 April 2026 | ||
|---|---|---|
| Session 1.1: TA1 ML4DTE system | ||
| 08:00 to 08:30 |
CLIM4health: from Citizen Science, Machine Learning and Earth Observation towards Urban Climate Services
Keynote speaker |
Speaker: Ana Oliveira (CoLAB +ATLANTIC)
|
| 08:30 to 08:50 | The WeatherGenerator Project: Building a multi-modal, multi-resolution Foundation Model for Weather and Climate science |
Speaker: Ilaria Luise (ECMWF)
|
| 08:50 to 09:10 | Lessons from Weather: Adapting GraphCast for Global Ocean Forecasting |
Speaker: Stefano Campanella (National Institute of Oceanography and Applied Geophysics - OGS, University of Trieste)
|
| 09:10 to 09:30 | High-Resolution Thermal Mapping for Land Cover Intervention Planning |
Speaker: Hugo Poupard (Murmuration-SAS)
|
| 09:30 to 09:50 | Bris and Forecast-in-a-Box: Applications and results for Malawi. |
Speaker: Lene Østvand (The Norwegian Meteorological Institute)
|
| 09:50 to 10:20 | Coffee break | |
| Session 1.2: TA1 ML4DTE system | ||
| 10:20 to 10:40 | Probing sensitivities in sea ice surrogate models toward probabilistic forecasting |
Speaker: Flavia Porro (University of Bologna)
|
| 10:40 to 11:00 | Downscaling of vertical profiles of mean and turbulent atmospheric variables using deep learning |
Speaker: Maximilian Pierzyna (Technische Universiteit Delft)
|
| 11:00 to 11:20 |
Generative Adversarial Networks for Simulating Dynamic Interactions in Digital Twins of Coastal Ecosystems
Online presentation |
Speaker: Oscar Leung (HKU)
Speaker: Ye Ha Kim (UCL)
|
| 11:20 to 11:40 | Climate Emulators as Tools for Adaptation: Reducing Weather Risk Uncertainty under Climate Change with probabilistic emulators of CMIP6 GCMs |
Speaker: Amaury Lancelin (Laboratoire de Météorologie Dynamique (LMD), Ecole Normale Supérieure, RTE)
|
| 11:40 to 12:00 | Going with the flow: generalising physically consistent data-driven sea-ice models |
Speaker: Tobias Finn (CEREA, ENPC, EDF R&D, Institut Polytechnique de Paris)
|
| 12:00 to 13:30 | Lunch break | |
| Session 1.3: TA1 ML4DTE system | ||
| 13:30 to 13:50 | A New Concept for Comparing Satellite Observations and km-scale Atmospheric Simulations using Self-Supervised Machine Learning |
Speaker: Dwaipayan Chatterjee (Institute for Meteorology and Climate Research, Karlsruhe Institute for Technology)
|
| 13:50 to 14:10 | SHRUG-FM: Reliability-Aware Foundation Models for Earth Observation |
Speaker: Kai-Hendrik Cohrs (University of Valencia)
|
| 14:10 to 14:30 | Evaluation of GraphCast and PanguWeather over India and Downscaling using WRF for Simulating the 2023 North Indian Floods |
Speaker: Balu Shaharica (Indian Institute of Technology Roorkee)
|
| 14:30 to 15:10 | TA1 - Open discussion | |
| 15:10 to 15:40 | Coffee break | |
| Session 2.1: TA2 Hybrid ML-Physics based systems for DA and Weather and Climate prediction | ||
| 15:40 to 16:00 | Advancing Global Weather Prediction at ECCC Using Spectral-Nudging-Based Hybrid NWP-MLWP Modelling |
Speaker: Leo Separovic (Environment and Climate Change Canada)
|
| 16:00 to 16:20 | Hybrid NWP-AI at Météo-France : nudging AIFS large scales in the global NWP operational model ARPEGE |
Speaker: Vincent Chabot (Météo France)
|
| 16:20 to 16:40 | Enhanced land surface data exploitation using machine learning |
Speaker: Patricia de Rosnay (ECMWF)
|
| 16:40 to 17:00 | Post-training weather emulators using measurements |
Speaker: Radi Radev (MeteoSwiss)
|
Location: Botte B4 pavilion - DAMA Technopole
Agenda time displays according to the selected time zone.
| Wednesday, 15 April 2026 | ||
|---|---|---|
| Session 2.2: | ||
| 08:00 to 08:30 |
TBC
Keynote speaker |
Speaker: Steve Penny
|
| 08:30 to 08:50 | Development of an offline and online hybrid model for the Integrated Forecasting System |
Speaker: Alban Farchi (ECMWF)
|
| 08:50 to 09:10 | Online bias-correction in sea-ice models using machine learning and data assimilation |
Speaker: Giovanni De Cillis (Dept. of Physics and Astronomy “Augusto Righi”, University of Bologna, Bologna, Italy)
|
| 09:10 to 09:30 | RUSH: Rapid-Update High-Resolution precipitation nowcasting and Global AI/NWP downscaling in a single Latent-Diffusion model |
Speaker: Simon De Kock (Electronics and Informatics (ETRO), Vrije Universiteit Brussel & Royal Meteorological Institute of Belgium)
|
| 09:30 to 10:00 | Coffee break | |
| Session 2.3: TA2 | ||
| 10:00 to 10:20 | Hybrid machine learning and ensemble data assimilation |
Speaker: Wei Pan (ECMWF)
|
| 10:20 to 10:40 | Machine Learning-Driven Background Error Covariances for High-Resolution Data Assimilation |
Speaker: Ravi Shankar Nemani (School of Mathematical, Physical & Computational Sciences, University of Reading, Reading, UK and National Centre for Earth Observation, Reading, UK)
|
| 10:40 to 11:00 |
Multi-hazard nowcasting system over Romania
Online presentation |
Speaker: Claudiu Adam (National Meteorological Administration)
|
| 11:00 to 11:20 | Unmasking Compensating Biases: A Process-Partitioned Neural Network Approach |
Speaker: Yuiko Ichikawa (University of Cambridge)
|
| 11:20 to 12:00 | TA2 - Open discussion | |
| 12:00 to 13:30 | Lunch | |
| Session 3.1: TA3 ML applications for Earth system observations | ||
| 13:30 to 14:00 |
Developing a Pixel-Based Children's Climate Risk Index for Targeted Intervention
Keynote speaker |
Speaker: Do-hung Kim (UNICEF)
|
| 14:00 to 14:20 | Collaborative Labelling of Earth System Features in Satellite Data for Nowcasting and Climate Applications |
Speaker: Lauren Biermann
|
| 14:20 to 14:40 | Machine-learning-based observation operators for land-surface data assimilation |
Speaker: Peter Weston (ECMWF)
|
| 14:40 to 15:00 | A data-driven reanalysis of Amazon water levels with deep learning |
Speaker: Ruben Cartuyvels (ESA)
|
| 15:00 to 15:30 | Coffee break | |
| Session 3.2: TA3 | ||
| 15:30 to 15:50 | Benchmarking Earth Observation embeddings for large scale aboveground biomass mapping |
Speaker: Yu-Feng Ho (OpenGeoHub)
|
| 15:50 to 16:10 | Contrail Instance Segmentation on Geostationary Imagery |
Speaker: Ana Maria Pelin (EUROCONTROL & TU Delft)
|
| 16:10 to 16:30 |
GridShield™: A Hybrid ML-Satellite Fusion Framework for Real-Time Power Grid Resilience and Climate-Verified Carbon Credit Generation Using ESA Copernicus Sentinel Data
Presentation slides |
Speaker: Milagrosa Russelle Ballestar (iONTEK Power Solutions Corporation - Innovate with EU Space Data Mentoring Programme 2025)
|
| 16:30 to 16:50 | Data-Driven Eddy Detection from Wide-Swath Altimetry |
Speaker: Matteo Broccoli (CMCC Foundation, Italy)
|
| 16:50 to 17:10 | Developing a Multimodal AI Framework for Wetlands Mapping with Multi-Sensor Earth Observation Data |
Speaker: Khunsa Fatima (kf178@leicester.ac.uk)
|
Location: Botte B4 pavilion - DAMA Technopole
Agenda time displays according to the selected time zone.
| Thursday, 16 April 2026 | ||
|---|---|---|
| Session 3.3: TA3 | ||
| 08:00 to 08:20 | From Synthetic to Real: A Scalable Dataset and ML Framework for Global Methane Monitoring |
Speaker: Enno Tiemann (OHB Digital Connect / Technical University of Munich (TUM))
|
| 08:20 to 08:40 | Filling the Multisensor Gap: Multimodal Generative Models for Volcano Monitoring |
Speaker: Simona Cariello (INGV-OE Istituto Nazionale di Geofisica e Vulcanologia Osservatorio Etneo, Catania, Italia)
|
| 08:40 to 09:00 | Deep learning emulation of the far-infrared fast forward operator for ESA FORUM satellite |
Speaker: Simone Raciti (University of Bologna)
|
| 09:00 to 09:20 | Intelligent All-sky Cameras for Dense Mesoscale Observations |
Speaker: Ben Pickering (Wx Labs Ltd and NCAS)
|
| 09:20 to 10:00 | TA3 - Open discussion | |
| 10:00 to 10:20 | Coffee break | |
| Session 4.1: TA4 End-to-end ML systems for DA and Weather and Climate Prediction Multidomain ML4ESOP | ||
| 10:20 to 10:40 | GraphDOP: Are observations all you need? |
Speaker: Eulalie Boucher
|
| 10:40 to 11:00 | An End-to-End Observation-Centric Graph Neural Network Framework for Global Earth System Prediction |
Speaker: Azadeh Gholoubi (NOAA/NWS/NCEP/EMC)
|
| 11:00 to 11:20 | Duo-AttnOPNets: An End-to-End ML–4D-Var Framework for Atmospheric Composition Forecasting and Data Assimilation |
Speaker: Judongyang Zhou (Imperial College London)
|
| 11:20 to 11:40 | Extending the Aila AI Weather Model: A Multi-Decoder Approach for Variable Expansion |
Speaker: Mikko Partio (Finnish Meteorological Institute)
|
| 11:40 to 12:00 | AI-Based Weather Forecasting for Switzerland |
Speaker: Claire Merker (MeteoSwiss)
|
| 12:00 to 13:30 | Lunch break | |
| Session 4.2: TA4 | ||
| 13:30 to 14:00 |
From Real-Time Big Data Assimilation on Fugaku to Synergistic Development of DA and AI: Osaka Expo 2025 and Beyond
Keynote speaker |
Speaker: Takemasa Miyoshi (RIKEN)
|
| 14:00 to 14:20 | Beyond the atmosphere: Building a Data-Driven Earth System Model at ECMWF |
Speaker: Nina Raoult
|
| 14:20 to 14:40 | The physics of AI weather models |
Speaker: George Craig (Meteorological Institute, LMU Munich)
|
| 14:40 to 15:00 | Interpretability of AI Weather Models via Intermediate Decoding |
Speaker: Matthias Beylich (Meteorological Institute, LMU Munich)
|
| 15:00 to 15:30 | Coffee break | |
| Session 4.3: TA4 | ||
| 15:30 to 15:50 | Leveraging the kilometre-scale AROME reanalysis to develop AI-based regional forecasting systems over Western Europe |
Speaker: Thomas Rieutord (Météo-France)
|
| 15:50 to 16:10 | Probabilistic GAN for European Winter Storm Downscaling |
Speaker: Athul Rasheeda Satheesh (Karlsruhe Institute of Technology)
|
| 16:10 to 16:30 | Towards few-sample extreme event likelihoods with guided diffusion models |
Speaker: Peter Manshausen (NVIDIA)
|
| 16:30 to 17:00 | TA4 - Open discussion | |
Location: Botte B4 pavilion - DAMA Technopole
Agenda time displays according to the selected time zone.
| Friday, 17 April 2026 | ||
|---|---|---|
| Session 5.1: TA5 - HP and new computing technologies for ML applications in ESOP | ||
| 08:00 to 08:30 |
Combining HPC and AI to advance potential: Example from HLRS and HammerHAI
Keynote speaker |
Speaker: Bastian Koller (High Performance Computing Center Stuttgart)
|
| 08:30 to 08:50 | Scalable GPU-Accelerated Training of Graph Neural Networks for High-Resolution ESOP Simulations |
Speaker: Oscar Leung (HKU)
Speaker: Ye Ha Kim (UCL)
|
| 08:50 to 09:10 | Improving the performance and scalability of Anemoi |
Speaker: Jan Polster (ECMWF)
|
| 09:10 to 09:30 | Satellite Earth Observation and AI–QC-Physics methods for enhanced climate predictions and volcano–climate interaction modeling |
Speaker: Eleonora Amato (Istituto Nazionale di Geofisica e Vulcanologia (INGV), Sezione Osservatorio Etneo, Catania, Italy)
|
| 09:30 to 10:00 | Coffee break | |
| Session 5.2: TA5 | ||
| 10:00 to 10:20 | Optimising Anemoi for higher resolution data and next-generation systems |
Speaker: Cathal O Brien (ECMWF)
|
| 10:20 to 10:40 | Global Volcano Monitoring from Space with hybrid Quantum Convolutional Neural Networks |
Speaker: Federica Torrisi (INGV-OE Istituto Nazionale di Geofisica e Vulcanologia Osservatorio Etneo, Catania, Italia)
|
| 10:40 to 11:00 | Modular FAIR Components for Trust and Scalability in ML-Enabled Ecosystems |
Speaker: Ingo Simonis (OGC)
|
| 11:00 to 11:20 | Profiling Machine Learning Weather-Forecasting Models in Anemoi |
Speaker: Marieke Plesske (HPC Applications Team)
|
| 11:20 to 11:50 | TA5 - Open discussion | |
| 11:50 to 12:00 | Closing remarks | |