17:15 〜 19:15
[SGD03-P19] Numerical experiments on estimating the viscosity from the postseismic deformation after a large earthquake
キーワード：粘弾性、データ同化、余効変動
Following large earthquakes, postseismic crustal deformations are often observed for years at the GNSS observational stations. The observed deformation at the ground surface includes the afterslip on the fault and the viscoelastic deformation of the crust and the upper mantle, activated by the coseismic stress change. The physical parameters, frictional properties of the fault and the rheological properties of the medium are the key to estimate the stress accumulation on the fault.
Data assimilation (DA) studies have attempted to estimate the frictional properties directly from the observational data. DA incorporates the observed data into the physics-based model to construct a more plausible model. When DA works well, we can obtain the physics-based model, including the physical properties, that can quantitatively explain the observed data. The constructed physics-based model can be used to simulate the evolution beyond the data period, i.e., prediction. Fukuda and Barbot (2021) employed Markov chain Monte Carlo (MCMC) method, one of the DA methods, to the postseismic crustal deformation after the 2011 Tohoku earthquake, and effectively distinguished between afterslip and viscoelastic deformation while estimating the associated physical parameters.
There are several types of DA technique including MCMC, and a sequential method called as Ensemble Kalman filter method (EnKF) and a variational method of adjoint method are the computationally efficient approaches that can apply to nonlinear systems. A sequential method called as Ensemble Kalman filter method (EnKF) is applied to the deformation data to estimate the physical variables on the fault and/or the medium (van Dinther et al., 2019, Hirahara and Nishikiori, 2019). A Variational method called as 4DVAR has been also applied to the fault slip considering afterslip assuming elastic medium to estimate the fault frictional properties (Kano et al., 2015; 2020). If the physics-based model under consideration is linear, the results of these methods should consistent. However, if the physics model is non-linear, the results are not necessarily same.
In this presentation, I apply both EnKF and 4DVAR to the viscoelastic deformation after a large earthquake, and compare the results to discuss the characteristics of the methods. The future goal is to apply these methods to the postseismic defomation including both afterslip and viscoelastic deformation.
I consider the 2D half-space medium consisted of elastic and inelastic domain. I calculate the development of the strain and the surface deformation after an earthquake using equivalent body force method (Barbot and Fiako, 2010). In this presentation, I perform several synthetic tests, called twin experiments, which use the same physics-model to simulate the “true case” and to estimate the model parameters and variables. I show the cases of considering linear viscosity and non-linear viscosity. Additionally, I discuss the effect of the spatial heterogeneity of the medium on estimation.
Data assimilation (DA) studies have attempted to estimate the frictional properties directly from the observational data. DA incorporates the observed data into the physics-based model to construct a more plausible model. When DA works well, we can obtain the physics-based model, including the physical properties, that can quantitatively explain the observed data. The constructed physics-based model can be used to simulate the evolution beyond the data period, i.e., prediction. Fukuda and Barbot (2021) employed Markov chain Monte Carlo (MCMC) method, one of the DA methods, to the postseismic crustal deformation after the 2011 Tohoku earthquake, and effectively distinguished between afterslip and viscoelastic deformation while estimating the associated physical parameters.
There are several types of DA technique including MCMC, and a sequential method called as Ensemble Kalman filter method (EnKF) and a variational method of adjoint method are the computationally efficient approaches that can apply to nonlinear systems. A sequential method called as Ensemble Kalman filter method (EnKF) is applied to the deformation data to estimate the physical variables on the fault and/or the medium (van Dinther et al., 2019, Hirahara and Nishikiori, 2019). A Variational method called as 4DVAR has been also applied to the fault slip considering afterslip assuming elastic medium to estimate the fault frictional properties (Kano et al., 2015; 2020). If the physics-based model under consideration is linear, the results of these methods should consistent. However, if the physics model is non-linear, the results are not necessarily same.
In this presentation, I apply both EnKF and 4DVAR to the viscoelastic deformation after a large earthquake, and compare the results to discuss the characteristics of the methods. The future goal is to apply these methods to the postseismic defomation including both afterslip and viscoelastic deformation.
I consider the 2D half-space medium consisted of elastic and inelastic domain. I calculate the development of the strain and the surface deformation after an earthquake using equivalent body force method (Barbot and Fiako, 2010). In this presentation, I perform several synthetic tests, called twin experiments, which use the same physics-model to simulate the “true case” and to estimate the model parameters and variables. I show the cases of considering linear viscosity and non-linear viscosity. Additionally, I discuss the effect of the spatial heterogeneity of the medium on estimation.