November 09, 2026 – by Tamara Bandikova, C2SM Communications
Weather forecasts and climate projections rely on numerical models that simulate complex processes in the atmosphere and oceans. As scientists seek to better understand extreme rainfall, heatwaves, droughts and flooding, these models need to resolve the atmosphere and ocean with increasingly finer resolution. This increases not only the complexity of these models but also demands for ever more computing power. Increasing the ease of use of these models and improving their efficiency can therefore benefit both research and society.
A major software milestone
Now, researchers working in the Open ETH project EXCLAIM have modernized the dynamical core of the ICON weather and climate model. The dynamical core is the part of the model that solves the mathematical equations describing atmospheric motion. Led by Nicolas Gruber, professor of environmental physics at ETH Zurich, the team replaced a large part of the traditional implementation with a modern software approach based on GT4Py, a domain-specific Python library for weather and climate models developed by CSCS and ETH Zurich.
GT4Py enables scientists to write high-level Python code that is well readable by humans while automatically generating highly optimized code for a wide range of computing architectures. This makes it easier to adapt the model to new generations of supercomputers without rewriting large parts of the code. The new implementation was tested on the CSCS supercomputer and is already used for production climate simulations at C2SM.
Faster and more efficient simulations
The modernized dynamical core delivers up to 30 percent faster core performance than the previous implementation. End-to-end climate simulations run around 10 to 15 percent faster, allowing researchers to carry out more experiments and process larger datasets using fewer computing resources. The reduction in runtime also lowers the energy demand associated with large-scale simulations.
Benefits for science and society
Faster simulations also help researchers produce higher-resolution climate projections and improve forecasts of extreme weather events. More detailed simulations can support climate adaptation planning, regional risk assessments and weather warning systems. At the same time, the modular design makes the software easier to maintain, test and extend with new capabilities. These efforts also permitted researchers in the EXCLAIM project to fully benefit from the exceptional compute resources available on Alps, enabling them to undertake extended simulations of the fully coupled earth system (ocean – atmosphere – land surface – sea-ice, carbon cycle on land and in the ocean) at an unprecedented resolution of 2.5 kilometers.
A Swiss contribution with international impact
These achievements are the result of a collaboration between ETH Zurich, MeteoSwiss, the Scalable Parallel Computing Laboratory at ETH Zurich (SPCL), Empa and CSCS under the umbrella of C2SM. In 2025, C2SM joined the international ICON Partnership, which includes leading climate and weather institutions in Germany. The delivery of the new Python-based dynamical core represents one of the center’s first major contributions in this role and strengthens Switzerland's position in next-generation climate and weather modeling.
This article was first published on ETH News.