LUMI offers groundbreaking computing methods to climate research

14.5.2025

the LUMI data center

Research professors Christina Williamson and Risto Makkonen from the Finnish Meteorological Institute (FMI) and the University of Helsinki are working together with partners from the University of Colorado to conduct research on aerosols.

– Our collaboration project focuses on understanding and constraining aerosol forcing, which is a key uncertainty in anthropogenic climate forcing. In the first steps of the project, we are generating a Perturbed Parameter Ensemble with Earth System Model EC-Earth. Since it is computationally heavy to simulate EC-Earth for each possible parameter combination, we will develop emulators based on simulated parameter and model output space. Later, we will develop and apply methods to understand parametric uncertainties in simulated aerosol forcing and further constrain the forcing with a comprehensive suite of real-world observations of atmospheric aerosols. No PPE studies have previously been done with EC-Earth, so parametric aerosol forcing uncertainty in the model is currently unknown.’

Global collaboration enables richer datasets for research

The collaboration between the FMI researchers and the Coloradan partners is driven by the need for global coverage of observations in order to constrain aerosol processes and forcing at a global scale.

– We have a lot of aerosol measurement data in-house both at the Finnish Meteorological Institute and the University of Helsinki, but we are also bringing in valuable datasets from our international partners. Our Coloradan partners have made aerosol observations on ATom (the Atmospheric Tomography Mission), including microphysical, chemical and optical properties, black carbon and sulphur dioxide by NOAA CSL (National Oceanic and Atmospheric Administration’s Chemical Sciences Laboratory), aerosol composition by University of Colorado, and relevant gas phase-species by NCAR (National Center for Atmospheric Research). These datasets will enable us to make precise and relevant constraints for simulated and emulated aerosol forcing.’

LUMI brings climate modelling research to the next level – but it is not always easy to utilize computing capacity of that scale

This is the first time the researchers from FMI are exploiting the potential of LUMI with the EC-Earth model. Running these models in large computing systems is, however, challenging traditionally.

– Most global Earth System Models are not ready to utilize massive GPU infrastructures such as in LUMI. Individual parts of the model components have been tested on GPU environments, but with major technical challenges and sometimes without substantial gains to traditional CPU platforms. However, in this project, we are porting model post-processing workflows and emulator to the LUMI GPU partition.

Despite the challenges, there are great prospects and clear advantages of these new methods to the research efforts. Also, interest towards the utilization of GPU infrastructures in EC-Earth models is very high globally.

– EC-Earth is a large international consortium with more than 30 partners, and many institutes would be interested in running EC-Earth on the EuroHPC systems. Before the project, we were not running EC-Earth on LUMI, but on Puhti (national supercomputer operated by CSC). We are excited to see how EC-Earth performs on the machine: our project is proposing massive simulations, and LUMI would be an excellent addition to climate modeling infrastructure.’

Authors: Maari Alanko, Elisa Halonen and Pihla Kauranen, CSC – IT Center for Science