[S-CG50] Dynamics in mobile belts

[SCG50-05] Deciphering 3-dimensional crustal movements before and after the 2011 Tohoku-oki earthquake using principal component analysis

Shuntaro Fukaya², *Takeshi Sagiya¹ (1.Disaster Mitigation Research Center, Nagoya University, 2.Graduate School of Environmental Studies, Nagoya University)

Keywords:principal component analysis, crustal deformation, GNSS, 2011 Tohoku-oki earthquake, inelastic deformaiton, postseismic deformation

In northeast Japan, significant crustal deformation has been observed by the continuous GNSS network (GEONET) and its pattern drastically changed before and after the 2011 Tohoku-oki earthquake, reflecting various processes such as locking or slipping on the plate subduction interface, activities of inland active faults and volcanoes, and the mechanical properties of the island arc crust. In this study, we attempt to separate the components caused by different deformation sources by applying the principal component analysis (PCA) to 3-dimensional GNSS data. In this study, data matrices for the PCA analysis are created for the preseismic (2004/1-2010/12) and postseismic period (2011/5-2019/12), respectively. As a result, crustal deformation pattern in each time period is represented as a superposition of multiple modes. Each mode is represented by a simple product of a spatial mode representing a spatial pattern of the displacement and a temporal mode representing temporal changes common to all the GNSS stations and their components. The first mode of the preseismic period shows steady-state deformation due to interplate coupling. E-W strain shows a broad contraction field and a localized large contraction (-2.1×10^-7/year) along the Ou-backbone range. In contrast, in the postseismic period, relaxation processes with different time constants are found for the first two modes. Depending on the time constant and spatial distribution, the first mode corresponds to steady-state displacement and viscoelastic relaxation with a larger time constant (~213 days), while the second mod corresponds to afterslip with a smaller time constant (~58 days). The contribution of the first mode accounts for 99.8% of the original data, representing the overwhelming dominance of a single mechanism in the postseismic deformation. The E-W strains of the first two modes show extensional strains in the wide area pattern. However, along the Ou-backbone range, the first mode shows a contraction while the second mode shows a larger extension than the surrounding area. The former implies contribution of an inelastic deformation driven by the absolute stress which does not change significantly before and after the earthquake, while the latter reflects an elastic response of elastically heterogeneous medium to rapid stress changes.