Data Assimilation in Volcano Deformation Using Fast Finite Element Analysis with High Fidelity Model

Estimation of the inner state of volcanoes are important for understanding the mechanism of eruption and reduction of disaster risk. With the improvement in observation networks, data assimilation of internal magma state using time-history crustal deformation data observed at the surface is expected to be suitable for solving such problems. Using finite-element methods capable of modeling complex geometry is desirable for modeling the three-dimensional heterogeneous crust structure, and nonlinear time-history analysis is required for considering the change in material properties due to the movement of magma. Thus, many cases of large-scale finite-element analysis is required, and the computational cost incurred is expected to become a bottleneck. As a basic study towards data assimilation of internal magma state considering change in material properties of the crust, we demonstrated that many case analyses of volcano deformation problems can be conducted in a reasonable time frame by development of a crustal deformation analysis method accelerated by GPUs. For verification of the data assimilation method, we estimated the magma trend in an actual three-dimensional heterogeneous crust structure without temporal change in material properties. We confirmed that the magma movement trend can be reproduced using the model considering crust heterogeneity, while models disregarding three-dimensional crust structure resulted in wrong estimations. Thus, we can see that using finite-element methods capable of modeling three-dimensional heterogeneity for crustal deformation analysis is important for accurate magma state estimation.