ICON's full Earth system at 1.25 km resolution is awarded the Gordon Bell Prize for Climate Modelling
A research team led by the Max Planck Institute for Meteorology (MPI-M) and the German Climate Computing Center (DKRZ) has received the Gordon Bell Prize for Climate Modelling for their groundbreaking achievement in simulating the full Earth system with the ICON model at an unprecedented 1.25-kilometer resolution. The award was presented at the Supercomputing Conference 2025 (SC25) in St. Louis, Missouri (USA), the world’s leading annual HPC event.
Representatives of the research team from MPI-M, DKRZ, JSC, ETH Zurich, the Swiss National Supercomputing Centre, the University of Hamburg, and NVIDIA accepting the Gordon Bell Prize for Climate Modelling at the SC25. © DKRZ
Using the ICON Earth System Model, the team simulated atmosphere, ocean, land, and the full carbon cycle at a grid spacing of 1.25 km—fast enough to achieve a decade of climate simulation per month. This milestone, previously considered out of reach, opens the door to resolving key interactions between the carbon, water, and energy cycles with exceptional detail [1].
The 20-year development of ICON enabled the team to fully exploit NVIDIA GH200 Superchips on two of Europe’s most capable supercomputers: Alps in Switzerland and JUPITER in Germany. The resulting performance — 145.7 simulated days per real day using 85% of JUPITER’s resources — marks a new benchmark for Earth system simulations.
“This award testifies to the powerful synergy between science and technology and to exceptional teamwork. I am delighted and grateful for this wonderful recognition from ACM as well as from all our colleagues!” — Daniel Klocke, MPI-M researcher and project lead
To reach this level of throughput, the team introduced novel data-centric optimizations, distributing computational load between CPUs and GPUs in a heterogeneous setup. This not only boosted performance but significantly reduced energy use.
“Energy consumption in high-performance computing is becoming increasingly important, and here our approach delivers. We show that we can save more than a factor of four in energy compared to a classical, purely CPU-based setup.” — Jan Frederik Engels, DKRZ scientist
This project brought together experts from MPI-M, DKRZ, Jülich Supercomputing Centre, ETH Zurich, the Swiss National Supercomputing Centre, the University of Hamburg, and NVIDIA, reflecting the broad international collaboration behind this achievement. Earlier this week, the work also received the HPCwire Readers’ Choice Award for “Top HPC-Enabled Scientific Achievement.”
CO2 exchange between the air and the sea or the air and land in the Amazon basin at the unprecedented resolution achieved with ICON (green - CO2 uptake by the terrestrial biosphere or the ocean, brown - CO2 release into the atmosphere (the right side of the image is on the night side) © MPI/DKRZ/NVIDIA
Destination Earth Among the Finalists
A second project involving MPI-M and DKRZ — Destination Earth, led by CSC – IT Center for Science — was also among the finalists for the Gordon Bell Prize for Climate Modelling. Using the atmospheric model IFS, coupled to the ocean models NEMO and FESOM, the initiative [2] forms a cornerstone of Europe’s efforts to build a digital twin of the Earth, coordinated by ECMWF, ESA, and EUMETSAT.
The recognition of both ICON and Destination Earth highlights the deep and growing synergy between climate research and high-performance computing, pushing scientific and technological frontiers forward in tandem. Even more so, it recognizes the need for projects such as WarmWorld, which both changes the approach to climate model development and builds a trans-disciplinary community, now tackling challenges and celebrating such achievements together.
Further information
[1] ICON ESM publication: https://dl.acm.org/doi/10.1145/3712285.3771789
[2] Destination Earth publication: https://dl.acm.org/doi/10.1145/3712285.3771790
