Empowering weather and climate forecast

MAELSTROM's second dissemination workshop

ECMWF's impressive conference table

What it was about

This workshop started with an overview on the MAELSTROM project and continued with invited talks by experts from the three domains of machine learning, high-performance computing, and Earth sciences to summarise and disseminate the latest research and developments towards highly efficient, scalable machine learning tools that improve weather and climate predictions. The workshop was hybrid with the ability to join virtually from everywhere in the world and to follow the talks in person at ECMWF in Reading, UK.

We were honored to welcome 200 online participants and 36 in person.

The MAELSTROM Dissemination Workshop was organised back-to-back with the second MAELSTROM Boot Camp.

Drinks in ECMWF's weather room

Agenda and links to slides

7 November 2023
09:00 to 09:30 An introduction to MAELSTROM
Presentation slides
Peter Düben (ECMWF)
09:30 to 10:00 Performance models for machine learning
Presentation slides
Karthick Panner Selvam (University of Luxembourg)
10:00 to 10:30 Experiences with W&C ML Apps on AMD Instinct GPUs
Presentation slides
Stepan Nassyr (Jülich Supercomputing Center)
10:30 to 11:00 A machine learned weather forecast for Norway
Presentation slides
Thomas Nipen (Norwegian Meteorological Institute)
11:00 to 11:15 Coffee break
Partner EuroHPC Projects
11:15 to 11:45 SEA-Projects: towards a European heterogeneous system and SW architecture for Exascale and beyond Hans-Christian Hoppe (Jülich Supercomputing Center)
11:45 to 12:15 SparCity for Sparse Tensors: Study on Feature Extraction and Smart Tensor Generation
Presentation slides
Tugba Torun (Koç University)
12:15 to 12:45 Some recent improvements of parallel-in-time algorithm
Presentation slides
Daniel Ruprecht (Hamburg University of Technology)
12:45 to 14:00 Lunch break
Invited external speakers
14:00 to 14:30 Radiative transfer emulation: results so far and why we should move on to 3D
Presentation slides
Peter Ukkonen (DMI)
14:30 to 15:00 Deep Learning for regional ensemble forecasting : first results
Presentation slides
Laure Raynaud (Météo-France)
15:00 to 15:30 AIFS
Presentation slides
Simon Lang (ECMWF)
15:30 to 16:00 Coffee break
16:00 to 16:30 AtmoRep: Large-Scale Representation Learning of Atmospheric Dynamics
Presentation slides
Christian Lessig (ECMWF)
16:30 to 17:00 Physics-Constrained Deep Learning for Downscaling and Emulation
Presentation slides
Paula Harder (Fraunhofer Institute ITWM)
17:00 to 17:30 Towards km-scale AI emulation for weather and climate applications
Presentation slides
Karthik Kashinath (NVIDIA)
Drinks reception
18:00 to 19:30 Drinks reception in the Weather Room
Perfect double rainbow greeting our guests in Reading, UK.