Finland has become host to the world's third-fastest supercomputer, dubbed Lumi. Half the size of a basketball court, it operates in a former paper mill in the eastern city of Kajaani.

Lumi is a supercomputer project involving 10 European countries and the European Union, representing an investment of more than 200 million euros. The supercomputer itself is worth some 145 million euros and is the most powerful such machine in Europe.

"It would take a couple of million regular laptops to match the same power. That would be a 25-kilometre-high tower of MacBooks," director of LUMI Pekka Manninen told Yle.

The supercomputer was inaugurated in June. This autumn, its full computing power should be available to scientific communities throughout Europe, meaning researchers across the continent can access Lumi's resources and benefit from this new research tool.

The name is an acronym of Large Unified Modern Infrastructure, while 'Lumi' is also the Finnish word for snow. The supercomputer has an expected lifespan of about five years and ranks third among the world's supercomputers, behind Frontier in the US state of Tennessee and Fugaku in Kobe, Japan.

Kajaani as supercomputer host

Lumi is located in Kajaani, in the Kainuu region of eastern Finland, and is managed by the CSC - IT Center for Science Ltd.

Located in the machine room of a former paper mill, Lumi occasionally gets so hot that it has to be water-cooled, although eastern Finland's generally cool climate reduces the need for this procedure.

The heat generated by the supercomputer is fed into the regional heating network, providing about one fifth of Kajaani's district heating needs.

The supercomputer was awarded to Finland through a bidding process. A major advantage for Kajaani was the fact that the old paper mill is an ideal place to host the computer, Manninen explained.

"The benefits of a supercomputer come not only from computing power, but also from versatility. The machine can be used in a variety of disciplines, from physics to human sciences," he noted.

"With some simplification, Lumi works like any computer and the mechanics are no different," he added, "making the machine an enabling technology."

For example, Lumi can help researchers carry out their work with unprecedented speed and accuracy. The societal value of the supercomputer is directly linked to the value of scientific research and innovation research.

"Lumi will be a powerful catalyst," Manninen told Yle News.

"Ambitious" climate goals

Three international organisations - the European Centre for Medium-Range Weather Forecasts (ECMWF), the European Space Agency (Esa) and the European Meteorological Satellite Organisation (EUMETSAT) - have been tasked with using Lumi to build a digital twin model of Earth known as Destination Earth (DestinE). The project is scheduled to begin later this year.

Funded through the Digital Europe programme of the European Commission, the simulation aims to contribute significantly to achieving the EU's Green Development Agenda.

DestinE is intended as an information system and a research infrastructure to help societies deal with weather events and make weather predictions and forecasts. The model will be used to monitor natural phenomena and human activities.

The Destination Earth project is ambitious, Manninen said, as it attempts to bring different climate science models under one umbrella. This is a multi-year project and will take several years of determined development work, with large swaths of previously collected weather and climate data built into the simulation.

If the simulation is used to support policy making, understanding the associated inaccuracies and imprecisions of its predictions is very important, he added.

"Systematic estimations of uncertainty are very important. We can see if the predictions are sufficiently reliable, and not a dice roll."

Computer simulations of earth system are widely used in climate science and are basic building blocks of weather and climate predictions. For example, the Intergovernmental Panel on Climate Change (IPCC) already constructs its reports based on multiple supercomputer simulations made by research groups all over the world, according to Sami Niemelä, Director of Meteorological and Marine Research Programme at the Finnish Meteorological Institute (FMI).

However, Lumi will be able to better tackle research questions that are currently limited by computing power with current technology. It uses GPU processors specifically targeted for scientific computing that are faster and more energy-efficient.

"Lumi is first and foremost a research infrastructure. The earth system models that are used in weather and climate predictions have been continuously developed during the recent decades. Such models contains huge amount of legacy codes with very much relevant science encoded in it," Niemelä said.

"We need to adapt the models to this new technology in order to meet the new scientific challenges. Lumi allows scientists to access this new technology and to test algorithms on larger scales than ever before."