In the COVID-19 era, many people are treating simple activities like getting gas or groceries with caution as they try to heed social distancing mandates and protect their own health. Still, significant uncertainty surrounds the relative risk of different activities, and conflicting information is prevalent. A team of Finnish researchers set out to address some of these uncertainties by running supercomputer simulations of how pathogens like COVID-19 spread in spaces like grocery stores.
The researchers hailed from four different organizations: Aalto University, the Finnish Meteorological Institute, the VTT Technical Research Centre of Finland and the University of Helsinki. First, they modeled a space with aisles between shelves, emulating the design of a typical grocery store. Then, they introduced human figures, one each in two adjacent aisles.
Then, they made one of them “cough,” spraying a cloud of aerosol particles smaller than 20 micrometers in front of them. The researchers observed how these particles – which move along air currents or linger in the air instead of sinking – moved throughout the simulated space.
Three of the organizations (Aalto, VTT and the Finish Meteorological Institute) conducted the modeling independently given a set of starting conditions. They all came to the same conclusions: after the coughing, the cloud of particles spread substantially through the area, eventually diluting itself over the course of several minutes. Based on this, the researchers concluded that the airborne particles could infect other shoppers during those crucial minutes, well after the coughing person had walked away.
“The preliminary results obtained by the consortium highlight the importance of our recommendations,” said Jussi Sane, chief specialist at the Finnish Institute for Health and Welfare. “[We recommend] that you stay at home if you are unwell and that you maintain physical distance with everyone. The instructions also include coughing into your sleeve or a tissue and taking care of good hand hygiene.” Sane also cautioned that the study, while helpful, is not yet grounds for issuing new or modified recommendations.
To run the simulations, the researchers used the Puhti supercomputer at CSC, a Finnish science and IT company owned by the state. Puhti is an Atos BullSequana X1000 system comprising 1,000 CPU nodes, each equipped with two Cascade Lake-era Intel Xeon processors and a mix of memory sizes. Puhti, which uses Mellanox HDR InfiniBand, also has an 80-node AI partition equipped with Nvidia Volta V100 GPUs. Excluding the AI partition, Puhti has a Linpack performance of 1.7 petaflops and ranks 232 on the current Top500 list.
The researchers say that Puhti – which was made available “at very short notice” – allowed them to complete their initial results and visualizations within a week. Now, the researchers are planning to continue work on the models, refining them and using them to answer broader questions about the behavior of aerosol particles.
To read CSC’s article discussing this research, click here.