Winner of the Gordon Bell Special Prize for High Performance Computing-Based COVID-19 Research! Prof. Makoto Tsubokura Talks About the Future Opened up by Droplet and Aerosol Dispersion Simulation
Prof. Tubokura from RIKEN and Kobe University, who is participating in this project, has been awarded the Gordon Bell Special Prize for High Performance Computing-Based COVID-19 Research. He will share the story behind how his award-winning “droplet/aerosol spread simulation” research came to be and its application to the current infection spreads.
How the Gordon Bell Prize winning “Droplet/Aerosol Dispersion Model” Came About
Dr. Tsubokura, why did you study a dispersion model of droplets and aerosols?
Recently there are people who think of me as an expert on droplets, but my original area of expertise is fluid dynamics, working with the flow of air and water, and for that research I have been using supercomputers. Before the coronavirus pandemic, we formed consortiums with various industries, and from the time of the K computer, the predecessor of the Fugaku supercomputer, our broad theme was to accumulate a lot of research results on how fluid simulation would benefit the industrial world.
One result of this work was our software, the analysis solver CUBE1. Just as we were finetuning this for the Society 5.0 era2, the coronavirus infection began to spread. In other words, the timing of the outbreak coincided perfectly with the timing of our simulation of advanced fluid dynamic behavior using Fugaku.
How did that lead to your droplet/aerosol dispersion model?
From the time of the Diamond Princess cluster outbreak, Japan was one of the first countries in the world where droplets and aerosols were considered an important factor in the spread of infection from an epidemiological point of view. This was the origin of the so-called “sanmitsu” or “three C’s” campaign.
The route of infection by the novel coronavirus can be explained by a model in which droplets dispersed through the mouth gradually become smaller as they evaporate, and finally spread as aerosol particles. In other words, the phenomenon can be broken down into the flow of air carrying the droplets, and the evaporation of the droplets at their interface with the surrounding air. This phenomenon is physically similar to the model we had been studying, in which gasoline vaporizes rapidly during piston movement in an automobile engine.
During the state of emergency declared in April 2020, our research members were working from home and watching the news about the novel coronavirus every day. Since this was a new virus for which there was no data, there was a great deal of speculation all over the world.
Just then, amidst public calls for the use of Fugaku, a young member of our team proposed via Slack (our communication tool) the idea of applying fluid mechanics to countermeasures for coronavirus infection. All our research members started working in unison, thinking that this was an opportunity to give back to the world the results of our research. Since none of us were experts in epidemiology or viruses, we asked some medical experts and companies for their collaboration, and by the end of the day, we had put together a team.
The novel coronavirus is a new disease that requires data, but without accurate information, people’s anxiety has led to negative speculation. On the other hand, simulation is a research method that you can carry out at home. The social challenges of dealing with COVID-19, the situation where we are required to refrain from taking action, and our simulation technology – the fact that everything fell into place was one of the main reasons for its speedy achievement.
“Useful Infection Risk Simulations” for Each Everyday Scenario
What kind of scenarios are these infection risk simulations for?
We believe that publishing our simulation results is a way of “returning the results of our past research to society.” The important thing is to publish the necessary output at the appropriate time. With the usual research process of “hypothesis, experimentation and verification, writing of the paper, and publication,” it is difficult to provide the necessary information at the time when it’s needed, as society is constantly changing along with the coronavirus infection situation.
For this reason, we kept an eye on trends in the number of the novel coronavirus infection and tried to present effective measures at the times when recommendations were most needed.
A supercomputer speeds up the research process, but it still takes from half a day at best to a week at worst to calculate the results. We therefore made a list in advance of important actions to take during periods of increased infection, and effective preventive measures to take during periods when infection numbers were down and people were more active in their daily lives. We ran simulations for various different scenarios, including public transportation and public facilities. By having a team system that could immediately put out messages according to the degree of spread of infection, we managed to communicate to society at the right times.
For example, what kind of simulations did you do?
When people were able to return to offices in May or June 2020, after the first state of emergency had been lifted, we published the results of our simulation to verify “the effects of partitions”3 . At the end of August that year, as the school summer holidays were coming to an end, we announced “the effects of opening classroom windows” .4。
When there was a significant spread of infection among younger people, we also conducted simulations on izakaya bars, karaoke boxes, the effectiveness of urethane masks, and so on. When many people began to travel using the Go To Travel campaign, we simulated the dispersion of aerosol virus particles in taxis and airplanes.
I think the difference from normal research was that “we prioritized putting out the necessary information at the time when it’s needed.” We issued a total of six press releases and got a lot of feedback.
What impact did that public feedback have on your simulation research?
When our simulation results were reported, we received emails and phone calls almost every day from people saying they were worried about the risk of infection in certain situations in their daily lives. For example, a member of a mothers’ choir contacted us because she was worried about droplet infection between members of the choir and their visitors.
Whether it’s about a sports game, an athletics meet, or some other event, we take a scientific approach in answering questions, such as “Is it safe to cheer if I wear a proper mask to control the spread of droplets somewhat?” and “Are your results different indoors and outdoors?” The various everyday situations in which coronavirus measures are required is something that not only we researchers think about, but that can also be found in public feedback obtained through the media. I don’t think this has ever happened before in science.
What do you think about people’s awareness being changed by the visualization of droplet infection phenomenon resulting from your simulations?
I think that showing videos of our simulation results makes it much easier to communicate. Some people said that simulations of droplet and aerosol dispersion only scared the public, but at the same time, we believe we have been able to educate society on the importance of measures such as opening windows, installing partitions, and wearing masks.
We have provided more scientific and quantitative data to be used for policy purposes, too. The difficulty is that although we can give objective data such as “the risk of infection of spending one hour in a particular place,” we cannot simply state “whether being in that place is dangerous or safe.”
That’s because it is not science, but a person’s assessment using scientific methods. Being able to think about it and make decisions on one’s own is also a way to reduce the risk of infection and act appropriately without being overly anxious as a result of speculative discussions.
Facing Novel Coronavirus Variants and Future Pandemics
Will droplet and aerosol dispersion simulations work for variants such as Omicron?
Simulations of droplet and aerosol dispersion focus on how droplets from the mouth travel through the air. The risk of infection from droplets inhaled or attached to the human body is calculated using the strength of infectivity as a coefficient, and so our simulations work for both Delta and Omicron variants.
Conversely, we still don’t know if the virus in the droplets is highly infectious. People generally think of droplets as coming from the mouth, but small aerosol particles are also dispersed from the vocal cords and alveoli. Since the number of viruses in the droplets varies depending on where they occur, the accuracy of future simulations will be enhanced by integrating the results of epidemiological and biological studies.
So, there is still room for growth in droplet and aerosol simulations?
When a virus comes into contact with the skin or a mucous membrane, if there are cells there that act as receptors (virus receptors), that’s where the virus will start to multiply in the human body. We have already begun a complex simulation of “where the virus starts to grow in the body” when droplets are inhaled, and “how that infected person will spread the infection as a result.”
When the modeling is this systematic, it is easy to imagine the need for the computing power of a supercomputer, but our published results to date are supported by more than 1,000 simulations. I believe that winning the 2021 Gordon Bell Special Prize for High Performance Computing-Based COVID-19 Research is the result of the judges’ evaluation of not only the effectiveness of our simulations but also the groundwork involved.
With the help of advanced simulations, how should we live with the coronavirus?
I think that all members of society need to consider their own individual actions based on the information they have obtained. If you are always asking for help because you feel anxious, you may become unable to cope when your location or situation changes.
For example, there are discussions about whether the best materials for masks are non-woven or woven. Non-woven fabrics have a higher filtering performance than woven cloth ones. So, a non-woven mask that is shaped to fit your face is very effective in preventing infection. However, it has a high resistance to the passage of air during breathing, so air tends to leak through any small gaps.
When worn by a child with a small face, a non-woven mask for adults will have larger gaps, which means it will not perform in the way it should. In such a case, a cloth mask that fits the child properly may be better.
By unpacking information to that level, each of us can make the best choice without getting caught up in a binary argument about whether something is good or bad.
The results of your research have helped to create a thriving dialogue between society and science. How significant do you feel this is?
What I have realized since presenting the results of our infection risk simulations is that social media now enables two-way communication between society and science.
This is very different from the past, and I feel that a connection between society and science has been created through social media. If there can be a constructive dialogue about scientific research results, such as “is there possibility of doing this or that?” I think both society and science will be able to improve the quality. The driver of that possibility, I believe, is for individual members of society to “think.”
Team Leader, Complex Phenomena Unified Simulation Research Team, RIKEN Center for Computational Science (R-CCS)
(Professor, Graduate School of System Informatics, Kobe University)
He graduated from Kyoto University in 1992 before going on to receive a Ph.D. in Mechanical Engineering from the Graduate School of Engineering, the University of Tokyo. At present, he has laboratories in both RIKEN and Kobe University. His research covers manufacturing technology (mechanical, electronic, and chemical engineering) and fluid engineering, as well as multi-objective optimization of simulation results, machine learning (AI), and technologies that integrate computational science and data science. At RIKEN, he has developed simulation technology using its flagship supercomputers K and Fugaku. Since the declaration of the state of emergency in April 2020, he has been using his previously developed research results and methods to control the spread of coronavirus infection, by conducting and publishing simulations of droplet and aerosol infection in scenarios such as public transportation and public facilities. He was awarded the 2021 Gordon Bell Special Prize for High Performance Computing-Based COVID-19 Research for his work in visualizing the behavior of droplets and aerosols and educating the world about the importance of understanding and taking measures against them.