Large-Scale Deep Learning for the Earth System
Supported by
The second workshop on Large-Scale Deep Learning for the Earth System took place on August 29 – 30, 2024 in the Universitätsclub Bonn, Germany.
The program is available below.
A short post is available at: https://cesoc.net/workshop-on-large-scale-deep-learning-for-the-earth-system-a-recap/
The workshop brought together the leading groups that develop large-scale neural networks for weather and climate.
Topics include:
-
machine learning-based numerical weather prediction
-
climate emulators
-
AI-based Earth system component models
-
machine learning from observations
-
foundation models for the Earth system
-
evaluation and explainability of large-scale machine learning models for weather and climate
Would you like to stay updated and receive information about future workshops? then Subscribe HERE
For questions please contact us via email:
lsdles-workshop@cesoc.net
Information
When: August 29 – 30, 2024
Where: Universitätsclub Bonn, Konviktstr. 9, 53113 Bonn
Who:
Scientific Organizers: Martin Schultz (Jülich Supercomputing Center), Christian Lessig (ECMWF), Jürgen Gall (University of Bonn), Dale Durran (University of Washington), Matthew Chantry (ECWMF)
The local organizing committee: Hana Mohammed and Sybille Y. Scheele (CESOC coordination office), Daniel Minge (TRA1 Modelling, University of Bonn)
We gratefully acknowledge the support by
CESOC, ECMWF, the TRA Modelling at the University of Bonn, the Forschungszentrum Jülich and the University of Cologne.
Abstracts
Abstracts Submission closed. Notification has been sent out to the authors.
For the abstracts of the talks, please have a look at the program below and click the titles.
Registration
Registration is closed
Registration fees and deadlines:
| registration type | registration fee | deadline, 2024 |
| in-person, student reduced price (with verification) |
50 € (registration deadline ended) |
June 15 |
| in-person, full price | 120 € (registration deadline ended) |
August 7 |
| online | 20 € (registration deadline ended) |
August 27 |
-
In-Person registration fees include access to the venue, extended coffee breaks and lunch for the 2 days.
-
Online registration fee includes professionally filmed Livestream. Link will be sent within 24 hours before start of the event.
-
All registered attendees will be provided with recorded lectures after the event.
Program
(All times are in CEST (UTC+2))
The access to the sessions recordings is password-protected and restricted to registered participants (passwords shared via email)
Thursday, August 29, 2024
8:45 – 9:20 h Registration
9:20 – 9:30 h Welcome (co-chairs)
9:30 – 10:30 h Talks Weather
- Simon Lang: AIFS, ECMWF’s data driven forecast model (20 min)
- Magnus Sikora Ingstad: Regional data-driven modelling with a global stretched-grid approach (10 min)
- Guillaume Couairon: ArchesWeather (10 min)
- Jonathan Weyn: Aurora (15min)
Session recording: Link to watch here
10:30 – 11:00 h Coffee break
11:00 – 12:30 h Talks Weather
- Leo Separovic: Leveraging data-driven weather forecasting for improving numerical weather prediction skill through large-scale spectral nudging (20 min)
- Anna Vaughan: Aardvark Weather: End-to-end data driven weather prediction (10 min)
- Peter Watson: Machine learning for predicting high-resolution extreme precipitation (10 min)
- Michael Langguth: DownscaleBench (10 min)
- Lei Bai: FengWu-GHR: Learning the Kilometer-scale Medium-range Global Weather Forecasting (10 min)
- Troy Arcomano: Stormer – A state-of-the-art transformer for medium-range weather forecasting (20 min)
Session recording: Link to watch here
12:30 – 13:45 h Lunch
13:45 – 15:15 h Talks Weather
- Joel Oskarsson: Probabilistic Weather Forecasting with Hierarchical Graph Neural Networks (10 min)
- Martin Andrae: Continuous Ensemble Forecasting with Diffusion (10 min)
- Christopher Subich: Efficiently fine-tuning 37-level GraphCast for the Canadian GDPS (10 min)
- Ankur Mahesh: Design and generation of ensemble weather forecasts using Spherical Fourier Neural Operators (10 min)
- Jun Wang: Machine Learning Weather Prediction Model Development for Global Forecast (10 min)
- Jan Keller: AI-Var – A data-driven data assimilation system (10 min)
- Christian Lessig: Skillful multi-day forecasts directly from observations (20 min)
Session recording: Link to watch here
15:15 – 16:30 h Coffee break and poster session (1:15 h)
16:30 – 17:30 h Perspective talks from weather centers
- Sabrina Wahl: AICON – A data-driven weather forecasting model based on ICON (10 min)
- Daryl Kleist: Overview on progress toward AI-driven Numerical Weather Prediction at NOAA (10 min)
- Laure Raynaud: Data-driven weather forecasting at kilometre scale: towards new Arome-AI systems at Météo-France (10 min)
- Stephane Beauregard: AI/ML Initiatives at the Canadian Meteorological Centre (10 min)
Session recording: Link to watch here
17:30 h Adjourn
19:30 h (self-paid) Conference dinner at Delik‘Art restaurant – Colmantstraße 14-16, 53115 Bonn (set menu, 40€ per person plus costs for your choice of drinks)
Friday, August 30, 2024
09:00 – 10:15 h Talks Climate
- Salva Rühling Cachay: Probabilistic Emulation of a Global Climate Model with Spherical DYffusion (20 min)
- Spencer Clark: Coupling the AI2 Climate Emulator to a slab ocean and learning the sensitivity to changes in CO2 (20 min)
- Jakob Schloer: A Hybrid Model for El Niño Southern Oscillation Dynamics (10 min)
- Jannik Thuemmel: Probabilistic Representations of Subseasonal to Annual Ocean Dynamics (10 min)
Session recording: Link to watch here
10:15 – 11:30 h Coffee break and poster session (1:15 h)
11:30 – 12:30 h Talks Climate
- Johannes Jakubik: Prithvi WxC: A Multi Regional Foundation Model for Weather and Climate (20 min)
- Nikolay Koldunov: Using AI-based numerical weather prediction models for climate applications (10 min)
- Sebastian Hickman: Learning causal representations of climate model data (10 min)
Session recording: Link to watch here
12:30 – 13:45 h Lunch
13:45 – 15:00 h Talks Climate
- Catherine de Burgh-Day: Evaluation of data-driven models on S2S timescales (10 min)
- Troy Arcomano: LUCIE – A Lightweight Uncoupled ClImate Emulator with long-term stability and physical consistency for O(1000)-member ensembles (10 min)
- Julia Kaltenborn: ClimateSet: A Large-Scale Consistent Climate Model Dataset for Machine Learning (10 min)
- Dale Durran: Coupled atmosphere-ocean simulations with a parsimonious deep learning model (20 min)
15:00 – 15:20 h General discussion: where do we come from, where do we go?
15:20 – 15:40 h Wrap-up, announcement of 2025 workshop
15:40 h Adjourn
Session recording: Link to watch here
List of Poster
Weather
Climate
Bonn has a variety of hotels at various price points. We recommend to book early since Bonn has many business travelers and hotels tend to become booked out and/or expensive.
Some recommendations:
-
Motel one Bonn | cheap budget design hotel & very close to the workshop location (motel-one.com)
-
IntercityHotel Bonn (online-reservations.com)
-
Hotel Kurfürstenhof (kurfuerstenhof-bonn.de)