Clarification of Multiphase Turbulence Phenomena by Large-Scale Simulation

Phenomena such as typhoons, torrential rains during the rainy season, downbursts, and urban rainfall, which are localized and change rapidly, have a direct and serious impact on our daily lives. In order to predict these phenomena, it is said that at least a resolution of about 100m is required to handle the generation unit of each cloud. Supercomputers such as TSUBAME, the Earth Simulator, the K computer, and the Fugaku have made it possible to handle these phenomena more rigorously in a wider area surrounding the phenomena. By studying the formation, development, and maintenance of cumulus clouds and cumulonimbus clouds in more detail, we can expect to improve the prediction accuracy of meteorological and climatic phenomena that are closely related to our social lives. For super-high-resolution simulations, it is important not only to reduce the grid size but also to research and develop models that can capture physical phenomena on a detailed scale commensurate with the grid size. In our group, we are developing new models of physical processes, with a particular focus on turbulent flow phenomena.

Fig.1 : Multiphase turbulence and non-equilibrium turbulence phenomena in ocean earth science

For example, the formation of clusters (i.e., uneven distribution) of particles with inertia in turbulence is an important phenomenon observed in cloud turbulence. For particles smaller than the Kolmogorov scale (typically about 1 mm in atmospheric turbulence), micro-scale clustering can significantly increase the probability of collisions between particles. This enhancement effect is important for modeling the collisional growth of cloud particles and for modeling the formation of microplanets in protoplanetary disks.

For this micro-scale clustering phenomenon, for example, the effect of large scale flows, i.e., the Reynolds number (Re) dependence of the phenomenon, has been clarified for the first time in the world by large-scale simulation. We pointed out that the Re number dependence is not negligible in the case of the Earth's atmosphere and the formation process of microplanets in the protoplanetary disk, where the Re number is very large, and clarified the mechanism of the generation. In addition to discovering phenomena and clarifying the mechanism, we have achieved everything from basic to applied research by modeling this Reynolds number dependence and developing it into the development of a new cloud microphysics model.

We are interested not only in improving the performance of prediction models, but also in the predictability of phenomena. Will higher resolution of weather simulations simply improve prediction accuracy, or will they reveal the spatio-temporal limits of predictability? To answer these questions, we are developing a model (Lagrangian Cloud Simulator; LCS) that uses the Lagrangian method to track the position, motion, phase change growth, and even collision growth of individual particles. We have succeeded for the first time in the world in quantifying the fluctuations in bulk statistics due to slight differences in the initial particle positions. In addition, for the first time in the world, we have succeeded in tracing the movement and growth of all aerosol particles and reproducing the series of processes from the formation of tiny water droplets to their growth to raindrops and their arrival at the ground surface as precipitation particles.

Perspectives on cloud turbulence research using large-scale numerical simulations are introduced in Nagare, the journal of the Japan Society of Fluid Mechanics.(pdf)。

Tracked the movement and growth of all aerosol particles, and reproduced for the first time in the world a series of processes from the formation of tiny water droplets to their growth to raindrops and arrival at the ground surface as precipitation particles. (Kunishima & Onishi, Atmos. Chem. Phys., 2018)