Projects
Topic 3: Uncertainty quantification of weather-society coupled simulation
澤田 洋平/Sawada Yohei
Associate Professor, Institute for Engineering Science, Graduate School of Engineering, The University of Tokyo
Interview
Q1 What are you specialized in?
I am working on the prediction of water-related disasters such as floods and droughts.
I am working on the prediction of water-related disasters such as floods and droughts.
As a part of the Moonshot Project, I am studying the statistical and mathematical techniques of data assimilation and uncertainty quantification.
Data assimilation is a technique that integrates computer simulations with observation data from the field.
Computer simulation is, so to speak, creating another Earth in a computer. This technology is indispensable for predicting the future of the earth in the modern era, but it has the disadvantage of not being very accurate. Apparently, the Earth created on a computer is different from the real Earth. Therefore, the results of the analysis will inevitably differ from reality.
In contrast to computer simulation, observation data are more accurate. However, because floods and droughts are natural disasters, they cannot be observed in detail. For example, we can observe temperature only at a few finite points in Tokyo.
One technique that can efficiently compensate for the shortcomings of both is data assimilation. Data assimilation is a technique that combines state-of-the-art simulations with state-of-the-art observational data to make better forecasts in general.
As a part of the Moonshot Project, I am studying the statistical and mathematical techniques of data assimilation and uncertainty quantification.
Data assimilation is a technique that integrates computer simulations with observation data from the field.
Computer simulation is, so to speak, creating another Earth in a computer. This technology is indispensable for predicting the future of the earth in the modern era, but it has the disadvantage of not being very accurate. Apparently, the Earth created on a computer is different from the real Earth. Therefore, the results of the analysis will inevitably differ from reality.
In contrast to computer simulation, observation data are more accurate. However, because floods and droughts are natural disasters, they cannot be observed in detail. For example, we can observe temperature only at a few finite points in Tokyo.
One technique that can efficiently compensate for the shortcomings of both is data assimilation. Data assimilation is a technique that combines state-of-the-art simulations with state-of-the-art observational data to make better forecasts in general.
Figure: Agricultural drought monitoring and forecasting in East Africa
The other technique, uncertainty quantification, numerically expresses uncertainty in a forecast or how reliable it is. This can be divided into four major tasks.
First, it is to understand why and where uncertainty arises. Second, it quantifies the magnitude of uncertainties in a forecast. Third, we aim to reduce the uncertainty in the forecast. However, it is impossible to perform forecasts with zero uncertainty, and forecasts always contain uncertainty. We make some decisions and act in the face of uncertainty. The final task is to provide scientific decision-making using uncertain predictions.
It is important to solve these four tasks efficiently and in "real time". In other words, it is better to avoid saying, "We should have done this” later. It is important to solve them in the present moment.
First, it is to understand why and where uncertainty arises. Second, it quantifies the magnitude of uncertainties in a forecast. Third, we aim to reduce the uncertainty in the forecast. However, it is impossible to perform forecasts with zero uncertainty, and forecasts always contain uncertainty. We make some decisions and act in the face of uncertainty. The final task is to provide scientific decision-making using uncertain predictions.
It is important to solve these four tasks efficiently and in "real time". In other words, it is better to avoid saying, "We should have done this” later. It is important to solve them in the present moment.
Q２ The term "Socio-meteorology" appears on your website. What made you think of linking your research with "society"?
First of all, as a premise, it must be understood that nature and society cannot be separated in the sciences of disasters.
If there existed an uninhabited earth, whether there were a typhoon or a tornado, it would not be considered a disaster. We call them disasters because nature affects human activity.
On the other hand, humans also influence nature by building dams and levees to cope with floods. Therefore, I believe that nature and society should not be considered separately, but rather that nature influences society and society influences nature.
Many researchers around the world think the same way. I feel that meteorologists should also think more about the interaction between nature and society. The final "outlet" for meteorologists to society is weather forecasting, but it is not necessarily true that they are satisfied if they successfully forecast the weather conditions. They forecast the weather because they want to improve people's behavior by weather forecasting, in other words, they want to improve society.
This brings us to the question of what kind of weather forecasting should be done considering its interactions with society.
While weather forecasting also exists, it is necessary to consider how other infrastructures such as levees and dams should be viewed in total, and my colleagues and I believe that this should be called Socio-meteorology. In addition, Japan is very strong in computing and strongly promoting projects to develop supercomputers. In particular, the computing resources that can be used for weather and disaster management are very large.
At the same time, however, I feel that there is still much room for improvements in effectively using the large computational resources and provided big data. I believe that there is a need and potential for Socio-meteorology in terms of pursuing how to make use of vast amounts of data for the benefit of society.
I am pursuing this research simply because they are interesting. It is interesting to do something so difficult that you think you will never be able to solve it. I would like to work on solving these difficult problems by myself, and I hope that students will also consider it an exciting topic to work on.
On the other hand, humans also influence nature by building dams and levees to cope with floods. Therefore, I believe that nature and society should not be considered separately, but rather that nature influences society and society influences nature.
Many researchers around the world think the same way. I feel that meteorologists should also think more about the interaction between nature and society. The final "outlet" for meteorologists to society is weather forecasting, but it is not necessarily true that they are satisfied if they successfully forecast the weather conditions. They forecast the weather because they want to improve people's behavior by weather forecasting, in other words, they want to improve society.
This brings us to the question of what kind of weather forecasting should be done considering its interactions with society.
While weather forecasting also exists, it is necessary to consider how other infrastructures such as levees and dams should be viewed in total, and my colleagues and I believe that this should be called Socio-meteorology. In addition, Japan is very strong in computing and strongly promoting projects to develop supercomputers. In particular, the computing resources that can be used for weather and disaster management are very large.
At the same time, however, I feel that there is still much room for improvements in effectively using the large computational resources and provided big data. I believe that there is a need and potential for Socio-meteorology in terms of pursuing how to make use of vast amounts of data for the benefit of society.
I am pursuing this research simply because they are interesting. It is interesting to do something so difficult that you think you will never be able to solve it. I would like to work on solving these difficult problems by myself, and I hope that students will also consider it an exciting topic to work on.
Q3 Tell us your motivation to join this project targeting weather control as a principal investigator.
As I mentioned earlier, the uncertainty of weather forecast can never be reduced to zero.
Even if a superior weather control method is established in this project, there will always be uncertainty when predicting the consequences of that control.
It is not simply a matter of saying if typhoons will weaken or deviate. We need to estimate what the results of weather control will bring about in society. This estimation of the impacts of control on society inevitably includes large uncertainties. We do not believe that weather control can be useful to the world without first correctly understanding this great uncertainty. In other words, we need to provide clear guidance on what kind of social decisions can be made based on a proper assessment of uncertainty.
I hope to be able to contribute to such areas as a principal investigator.
Figure: Advanced disaster prediction Optimization using precipitation in both ideal and observed experiments improved river flow prediction in the 2015 Kinugawa River flood case.
It is not simply a matter of saying if typhoons will weaken or deviate. We need to estimate what the results of weather control will bring about in society. This estimation of the impacts of control on society inevitably includes large uncertainties. We do not believe that weather control can be useful to the world without first correctly understanding this great uncertainty. In other words, we need to provide clear guidance on what kind of social decisions can be made based on a proper assessment of uncertainty.
I hope to be able to contribute to such areas as a principal investigator.
Figure: Advanced disaster prediction Optimization using precipitation in both ideal and observed experiments improved river flow prediction in the 2015 Kinugawa River flood case.
Q4 What are the barriers to research at this point when it comes to supporting social decision-making?
Compared to producing a single forecast, estimating the uncertainty of that forecast is very difficult. It is hundreds of times more computationally intensive.
For example, it would be quite difficult to comprehensively quantify the uncertainty of a complex phenomenon such as weather, even with the most currently advanced supercomputers, such as Fugaku.We can wait and hope that there will be better computers in the future, but I believe that the researchers' ideas will be able to solve this problem.
We could come up with more efficient algorithms and think about what the essential uncertainties are for us. I think we need to push through and concentrate on ideas in collaboration with the social sciences.
Q5 What do you think will happen in the future with the addition of a "socio-meteorology" perspective to weather control technology?
We aim not only to weaken the force of typhoons, but to lead to a better society.
This would raise the question of what a better society is, and even if it were defined, we must seriously consider the question of how to apply forces to nature to achieve a better society in the face of very complex uncertainties.
It is a problem that can never be solved simply by cowering in a cozy natural science box.
We believe it is important to solve problems within a broader framework, such as socio-meteorology.
Q6 Assuming that weather control is achieved in 2050, what kind of society do you hope it will be like there?
If we can control many things on Earth, we will be asked how we should design this society.
For example, if the accuracy of weather control becomes more and more precise and we can control the weather, something like the omnipotence of humankind will certainly increase.
In a sense, it seems a difficult problem to create a society in which each person can enjoy an interesting life even under control.
As an academic, I find it very exciting to think that these issues will become active in about 2050.
