[M-IS16] Geophysical fluid dynamics-Transfield approach to geoscience

[MIS16-01] Reconstruction and prediction of thermally-driven flows using Lagrangian particle data assimilation

*Atsushi Nakao^1,2, Daisuke Noto³, Takatoshi Yanagisawa^2,4, Yuji Tasaka^2,4, Tatsu Kuwatani² (1.Graduate School of Engineering Science, Akita University, 2.Research Institute for Marine Geodynamics, Japan Agency for Marine-Earth Science and Technology, 3.Department of Earth and Environmental Science, University of Pennsylvania, 4.Faculty of Engineering, Hokkaido University)

Keywords:Adjoint method, Marker in cell method, Rayleigh-Bénard convection, Inverse problem

Lagrangian particles passively following fluid motions often record information about the surrounding flow, e.g. volcanic ash distributions reflect eruption intensity and wind speed, and metamorphic rocks contain flow processes of the surrounding mantle. In this study, we demonstrate the effectiveness of data assimilation in exploiting such particle-recorded information by applying it to particle images obtained from a laboratory experiment. Long-term particle trajectories were obtained from video recordings of Rayleigh-Bénard convection in a high-viscosity fluid containing tracer particles in a thin rectangular water tank. We developed four-dimensional variational data assimilation (4D-Var) for a marker and cell system and applied it to the particle trajectory data to reconstruct the thermal convection. As a result, we successfully estimated the temperature, velocity, particle trajectories, and Rayleigh number with high accuracy. Furthermore, the obtained solution was able to predict the fluid behavior beyond the assimilation time window. These results suggest that 4D-Var data assimilation can be used not only to reconstruct past processes and estimate system parameters, but also to predict future states in the absence of direct observational data.