24–26 August 2026 · Bonn, Germany

Workshop on Machine Learning for Earth System Modelling

About the Workshop

We are pleased to announce the fourth Workshop on Machine Learning for Earth System Modelling, following a series of three successful workshops. The workshop brings together leading researchers working on large-scale machine learning approaches for weather and climate science. It will take place in Bonn, Germany, and follows the Hackathon on Machine Learning for the Earth System.

The workshop will take place on August 24 – 26, 2026 in the Universitätsclub Bonn, Germany.

The event brings together researchers from Earth system science, mathematics, and computational science to discuss recent advances in machine learning for weather and climate applications.
Topics include:

Machine learning-based numerical weather prediction
Climate emulators and machine learning-based climate models
Machine learning-based Earth system component models (e.g. ocean and land models)
Incorporation of Earth system observations into ML models
Evaluation and explainability of machine learning models
Datasets for model training, evaluation, and benchmarking

Read about our workshop last years, 2025 and 2024 .

Timeline, Important Dates & Abstract Submissions

Workshop Registration opens: 26 May 2026
Workshop dates: 24–26 August 2026
Abstract submission deadline: 01 June 2026
Notification of accepted abstracts: 20 July 2026

Abstract submission

We invite you to submit your abstracts with relevance to the topics to be discussed in this year’s workshop.

Submissions will only be via CMT* platform (https://cmt3.research.microsoft.com/About)

Authors will need to create an account on the platform (if they do not already have one) at: https://cmt3.research.microsoft.com/docs/help/general/account-creation.html

Each abstract should be a maximum of 1 page.

Notification of accepted abstracts was communicated on 20 Jul 2026.

MLESM Hackathon

The workshop follows the Hackathon on Machine Learning for the Earth System on 19-21 August 2026 in Bonn. Please find more information on how to apply here:

Workshop Registration

On-site participation will cost 240 €, including access to the venue, coffee breaks, lunches, and a conference dinner.

Up to 30 student seats are available at a reduced rate of 125 € (proof of student status required).

Online participation will cost 75 €.

Registration is closed.

Local Organising Committee

Scientific Review Committee:

Important Information

Information about the workshop for all registered participants will be sent via mlesm-workshop@cesoc.net or mlesm-registrations@cesoc.net. Please make sure to mark them as safe, i.e., not spam/junk.

No time to attend our MLESM event but interested anyway, stay tuned and subscribe at mlesm-workshop@listen.uni-bonn.de.

For questions please contact us via email: mlesm-workshop@cesoc.net

Supported by

We gratefully acknowledge the support by CESOC, ECMWF, the TRA Modelling at the University of Bonn, the Forschungszentrum Jülich and the University of Cologne.

Acknowledgement: The Microsoft CMT service will be used for managing the peer-reviewing process for this conference. This service was provided for free by Microsoft and they bore all expenses, including costs for Azure cloud services as well as for software development and support.

Programme 2026

Monday 24.08.2026

9:30 – 9:45	Welcome
9:45 – 10:45	Talks: High-resolution Forecasting Troy Arcomano: HiRO-ACE: Fast and skillful AI emulation and downscaling trained on a 3 km global storm-resolving model Sophie Buurman: Towards a data-driven weather model for forecasting on-demand extremes at hectometric scale Simon De Kock: RUSH: A Latent-Diffusion Framework for Rapid-Update Precipitation Nowcasting and Global-Guided Short-Range Forecasting Manvendra Janmaijaya & Tom Dunstan: FastNet: High-resolution Global and Regional Weather Prediction
10:45 – 11:00	Poster fast-forward
11:00 – 12:30	Posters and coffee : Aram Farhad Shafiq Salihi: Multi-domain: A dynamic way of training across domains and resolutions Maurice Schmeits: Training and verification of high-resolution European MLWP ensemble models within a stretched-grid framework Mariana Clare: A probabilistic Temporal Downscaler for hourly global weather forecasting Takuma Yoshida: Global and stretched-grid AI weather prediction models with JMA initial conditions David M. Hall: Beyond Forecasting: Hybrid Intelligence and the Next Shift in Earth System Research Lukas Kugler: Fully-Coupled Latent-Space Data Assimilation for the Earth-system Gian Luca Buono: Learning Data-driven Surrogate and Correction Models for Satellite Observations in Numerical Weather Prediction
12:30 – 14:00	Lunch
14:00 – 15:00	Talks: Probabilistic, Masked, and Hybrid Forecasting Adrien Audren: Lessons learnt from CRPS-based probabilistic weather forecasting models at high spatio-temporal resolution Syed Zahid Husain: Optimizing MLWP for Advancing Hybrid MLWP-NWP Forecasting Ando Shah: EarthDiT: Probabilistic Sub-km Weather Forecast Downscaling with Sparse Targets and Earth Observation Conditioning Jannik Thuemmel: Masked Token Models as a paradigm for probabilistic forecasts in weather and climate
15:00 – 15:15	Poster fast-forward
15:15 – 16:30	Posters and coffee Philipp Hess: Generating stable spatiotemporal dynamics with discriminator-guided diffusion models Christopher Subich: Rotary positional encoding on the sphere Simon Hentschel: Exploring Design Choices for Probabilistic Precipitation Downscaling with Corrective Diffusion Ayushya Pare: Machine learning on high performance computing systems \| HPC.nrw – a regional perspective Martje Buhr: Inverse Modelling of Emission Pathways from Target Temperature Fields using Bayesian Neural Networks Alix Lamotte: Accelerating Atmospheric-Composition Chemistry solver with Neural-Network Surrogates: From Chemistry-Transport Models to Operational Numerical Weather Predictions
16:30 – 17:30	Talks: Beyond 0.25 degree Charles Jones: High-resolution Probabilistic Forecasts of Fire Weather in California using Downscaling Machine Learning Models Thomas Hamill: Training Kilometer-scale Numerical Weather Prediction Models over the US. Haiyu Dong : MAI-Nowcast: Skillful 6h ensemble nowcasting of extreme precipitation via flow matching Björn Lütjens : Zero-shot regional weather forecasts
19:00	Dinner at Tuscolo

Tuesday 25.08.2026

9:30 – 10:30	Talks: ML and Physics Leonardo Trentini: Physics-Constrained Finetuning of Weather Foundation Models through GNSS Zenith Wet Delay Prediction Laura Budewig : Atmospheric physics on machine learning: exploring generalisable limited-area models Leonardo Olivetti : Beyond standard metrics: from hazard-based forecast skill to impact-based evaluation Sabine Scholle & Malin Braatz: Towards a Mechanistic Understanding of GraphCast Through Latent Analysis
10:30 – 10:45	Poster fast-forward
10:45 – 12:00	Coffee and posters Britta Seegebrecht: Beyond the RMSE: Scale dependent accuracy measures for a fair comparison between AI and NWP weather prediction models Cristina Iacomino: Assessing the physical consistency of high-resolution meteorological variables from a Downscaling Latent Diffusion Model Hyoungnyoun Kim: Beyond Pointwise Comparison: Diagnosing Forecast Evolution with Temporally Aligned Feature-Space Trajectories Zhenyi Zhang: PhyVapor-ESFM: Physics-Constrained Foundation Model for Multi-Source Water Vapor Fusion and Forecasting Victor Hertel: Extreme precipitation forecasting with the WeatherGenerator model: A pilot for humanitarian anticipatory action in Mozambique Jakob Schloer: AIFS-subs: Adapting the AIFS-CRPS for Sub-Seasonal Prediction
12:00 – 12:30	Talks: Climate Jasmin Lampert : Machine learning-based dynamic climate multi-model ensemble mixing for improved representation of high-impact extreme events Dale Durran : Success and Failure: Extreme Out-of-Sample Aqua-Planet Tests of AI Weather Prediction Models
12:30 – 14:00	Lunch
14:00 – 14:45	Talks: Emulators Simon Driscoll: A Machine Learning Emulator for Antarctic Ice Shelf Processes & Dynamical Systems Theory as a Diagnostic for Emulators of Physical Systems Gabriele Franch: MLCast: A Community Framework for Benchmarking Machine Learning Nowcasting Across European Earth System Observations
14:45-15:15	Discussion
15:15 – 15:30	Poster fast-forward
15:30 – 16:45	Coffee and poster: climate Hugo Germain: Machine learning-based online correction of a climate model using reanalysis Savvas Melidonis: HClimRep: A Foundation Model for Capturing the Atmosphere, Ocean, and Sea Ice Interactions Troy Arcomano: AIMIP Phase 1: systematic evaluations of AI weather and climate models Yueling Ma: Using Machine Learning to Explore Groundwater Responses under Various Climate Scenarios Jack Woodcock: Evaluation of European Temperature Extremes in the Machine-Learning-Based ACE2 Climate Emulator Steffen Tietsche: Harnessing the stratosphere to improve data-driven sub-seasonal predictions Matias Olmo: AI downscaling over the Iberian Peninsula using DUNES: Diffusion U-Net for Efficient Climate Downscaling Janika Rhein: RánCast: A Global 3D Ocean Emulator Trained on Kilometer-Scale Simulations with Atmospheric Forcing Nils Hutter: Learning sea-ice physics from data: towards a hybrid ML–numerical modelling framework Mohamad Hakam Shams Eddin: Distributed hydrological modeling in the feature space Robert Brunstein: Evaluating ArchesWeather and ArchesWeatherGen under Multi-Decadal AMIP-style climate simulations
16:45 – 17:15	Talks: Subseasonal to Seasonal Forecasting Robert Brunstein: S2S Forecasting with ArchesWeatherGen: Learning long-range spatio-temporal relationships Jonathan Weyn : Enhancing AI and Dynamical Subseasonal Forecasts with Probabilistic Bias Correction Piyush Garg: The Recipe Matters More Than the Kitchen: A Multi-Dimensional Diagnostic Framework for AI Weather Models

Wednesday 26.08.2026

9:15 – 10:00	Talks: Ocean Kacper Nowak: OceanRep: A Foundation Model for Ocean Dynamics Sara Hahner & Lorenzo Zampieri: Representing the Surface Ocean in ECMWF’s Data-driven Forecasting System AIFS Rachel Furner : Developing a data-driven 3d global ocean model at ECMWF
10:00 – 10:30	Coffee
10:30 – 11:15	Talks: Observations Christian Lessig: WeatherGenerator: benefits of multi-modality and pre-training Roope Tervo : Global GEO-Ring Radiance Data Record accompanied with Earth System Features Alistair White : AIDA: Operational AI Data Assimilation from Level 1 Observations
11:15 – 12:15	Final discussion
12:15 – 13:30	Lunch

Workshop Registration

On-site participation will cost 240 €, including access to the venue, coffee breaks, lunches, and a conference dinner.

Up to 30 student seats are available at a reduced rate of 125 € (proof of student status required).

Online participation will cost 75 €.

Registration is closed.