Articles | Volume 19, issue 2
https://doi.org/10.5194/gmd-19-755-2026
© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
https://doi.org/10.5194/gmd-19-755-2026
© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Development and technical paper
| Highlight paper
| 
23 Jan 2026
Development and technical paper | Highlight paper |
| 23 Jan 2026
| 23 Jan 2026
Operational numerical weather prediction with ICON on GPUs (version 2024.10)
Xavier Lapillonne
CORRESPONDING AUTHOR
Federal Office of Meteorology and Climatology MeteoSwiss, Zurich, Switzerland
Daniel Hupp
Federal Office of Meteorology and Climatology MeteoSwiss, Zurich, Switzerland
Fabian Gessler
Center for Climate Systems Modeling C2SM, ETH Zurich, Zürich, Switzerland
André Walser
Federal Office of Meteorology and Climatology MeteoSwiss, Zurich, Switzerland
deceased, 19 February 2025
Andreas Pauling
Federal Office of Meteorology and Climatology MeteoSwiss, Zurich, Switzerland
Annika Lauber
Center for Climate Systems Modeling C2SM, ETH Zurich, Zürich, Switzerland
Benjamin Cumming
Swiss National Supercomputing Centre, ETH Zurich, Zürich, Switzerland
Carlos Osuna
Federal Office of Meteorology and Climatology MeteoSwiss, Zurich, Switzerland
Christoph Müller
Federal Office of Meteorology and Climatology MeteoSwiss, Zurich, Switzerland
Claire Merker
Federal Office of Meteorology and Climatology MeteoSwiss, Zurich, Switzerland
Daniel Leuenberger
Federal Office of Meteorology and Climatology MeteoSwiss, Zurich, Switzerland
David Leutwyler
Federal Office of Meteorology and Climatology MeteoSwiss, Zurich, Switzerland
Dmitry Alexeev
NVIDIA, Zürich, Switzerland
Gabriel Vollenweider
Federal Office of Meteorology and Climatology MeteoSwiss, Zurich, Switzerland
Guillaume Van Parys
Federal Office of Meteorology and Climatology MeteoSwiss, Zurich, Switzerland
Jonas Jucker
Center for Climate Systems Modeling C2SM, ETH Zurich, Zürich, Switzerland
Lukas Jansing
Federal Office of Meteorology and Climatology MeteoSwiss, Zurich, Switzerland
Marco Arpagaus
Federal Office of Meteorology and Climatology MeteoSwiss, Zurich, Switzerland
Marco Induni
Swiss National Supercomputing Centre, ETH Zurich, Zürich, Switzerland
Marek Jacob
Deutscher Wetterdienst, Offenbach am Main, Germany
Matthias Kraushaar
Swiss National Supercomputing Centre, ETH Zurich, Zürich, Switzerland
Michael Jähn
Center for Climate Systems Modeling C2SM, ETH Zurich, Zürich, Switzerland
Mikael Stellio
Center for Climate Systems Modeling C2SM, ETH Zurich, Zürich, Switzerland
Oliver Fuhrer
Federal Office of Meteorology and Climatology MeteoSwiss, Zurich, Switzerland
Petra Baumann
Federal Office of Meteorology and Climatology MeteoSwiss, Zurich, Switzerland
Philippe Steiner
Federal Office of Meteorology and Climatology MeteoSwiss, Zurich, Switzerland
Pirmin Kaufmann
Federal Office of Meteorology and Climatology MeteoSwiss, Zurich, Switzerland
Remo Dietlicher
Center for Climate Systems Modeling C2SM, ETH Zurich, Zürich, Switzerland
Ralf Müller
Deutsches Klimarechenzentrum, Hamburg, Germany
Sergey Kosukhin
Max Planck Institute for Meteorology, Hamburg, Germany
Thomas C. Schulthess
Swiss National Supercomputing Centre, ETH Zurich, Zürich, Switzerland
Ulrich Schättler
Deutscher Wetterdienst, Offenbach am Main, Germany
Victoria Cherkas
Center for Climate Systems Modeling C2SM, ETH Zurich, Zürich, Switzerland
William Sawyer
Swiss National Supercomputing Centre, ETH Zurich, Zürich, Switzerland
Data sets
ICON-CH1-EPS input and verification data Federal Office of Meteorology and Climatology MeteoSwiss https://doi.org/10.5281/zenodo.16760638
Model code and software
ICON model, 2024.10 ICON partnership https://doi.org/10.35089/WDCC/IconRelease2024.10
ICON model with Data Assimilation, 2024.10_withDA MeteoSwiss https://doi.org/10.5281/zenodo.15674269
Executive editor
Increasingly, new supercomputers depend on GPUs for the vast bulk of their processing power. This makes the effective exploitation of GPUs an imperative across geoscientific modelling. This paper presents the port of a full numerical weather prediction system to GPU. It provides an excellent example of how such a port can be achieved in practice while delivering significant performance benefits. As such, this work offers particularly valuable guidance for the wider modelling community.
Increasingly, new supercomputers depend on GPUs for the vast bulk of their processing power....
Short summary
The ICON climate and numerical weather prediction model was fully ported to Graphical Processing Units (GPUs) using OpenACC compiler directives, covering all components required for operational weather prediction. The GPU port together with several performance optimizations led to a speed-up of 5.6× when comparing to traditional Central Processing Units (CPUs) . Thanks to this adaptation effort, MeteoSwiss became the first national weather service to run the ICON model operationally on GPUs.
The ICON climate and numerical weather prediction model was fully ported to Graphical Processing...
