4:30 PM - 4:45 PM
[SSS05-10] Quasi-dynamic numerical simulation of earthquake generation cycles along the Japan Trench and Nankai Trough
★Invited Papers
We have conducted numerical simulations of earthquake generation cycles along the Japan Trench and the Nankai Trough [Nakata et al., 2012; 2014; 2016; 2021; 2023]. Here, we will report on the characteristics of the earthquake generation cycle simulation methods we have used, the results obtained, and future issues to be addressed.
Our simulation is based on the rate- and state-dependent friction law [Dieterich, 1979] that represents the process of stress accumulation and release on the plate interface. Seismic and aseismic events on a discretized geometry were modeled to represent the release of the slip deficit or backslip, which accumulates during interseismic periods [Rice, 1993]. Such spatiotemporal variations in the slip velocity were assumed to indicative of unstable slip with a frictional interface. We used a fault constitutive law [Nakatani, 2001] to determine the slip rate for a given stress and strength value. We used an aging law [Dieterich, 1979; Ruina, 1983] as the evolution law for the strength change. To treat the long cycle, we used the quasi-dynamic model [Rice, 1993] with a smaller seismic radiation damping term instead of a fully dynamic model. To solve these equations, we removed the time derivative of stress and obtained differential equations for slip rate and strength. The differential equations were solved using an adaptive time step fifth-order Runge–Kutta algorithm [Press et al. 1996]. For the boundary conditions, the slip velocity was assumed to be constant and equal to the plate convergence rate (Vpl) outside the model region. For the initial conditions, the slip velocity was assumed to be uniform at 0.9Vpl.
We discretized the curved subducting plate interface into small planar triangular elements. We treat three triangular elements as one rectangular-like subfault [Hyodo et al., 2016]. The time derivative of the shear stress at the i-th subfault due to the unit slip at the j-th subfault is represented by the combined effect of angular dislocations through three triangular elements within a rectanglar-like subfault [Comninou and Dundurs, 1975]. The stress change was estimated at the center of each subfault in the direction of slip acceleration in a homogeneous elastic half-space. We used the realistic 3D geometry of the subducting Philippine Sea Plate and Pacific Plate along the Japan Trench [Baba et al. 2003; 2006]. The free surface is shifted to the depth of the trough axis to avoid unrealistic stress concentration at the tip [Hyodo and Hori 2013]. The total number of approximately 1-km-long subfaults in the strike direction was up to approximately 520,000 in the area used in our previous study.
In our simulations, the frictional parameters A, B, and L were assumed to be constant over the earthquake cycle. In addition, we assumed a depth-dependent heterogeneity of A–B. The frictional parameters in the seismogenic zone are A–B < 0 (i.e., velocity weakening in steady-state slip). The computational resources of the Earth Simulator provided by JAMSTEC and of Cyberscience Center, Tohoku University were used for all simulations. As a result, when we assumed a single-circular patch, we obtained earthquakes with constant recurrence intervals. When we applied the multiscale-circular patch model, large to moderate earthquakes ruptured simultaneously with closely spaced and partially overlapping patches, showing the complex earthquake cycles with a wide range of magnitudes.
Our simulation is based on the rate- and state-dependent friction law [Dieterich, 1979] that represents the process of stress accumulation and release on the plate interface. Seismic and aseismic events on a discretized geometry were modeled to represent the release of the slip deficit or backslip, which accumulates during interseismic periods [Rice, 1993]. Such spatiotemporal variations in the slip velocity were assumed to indicative of unstable slip with a frictional interface. We used a fault constitutive law [Nakatani, 2001] to determine the slip rate for a given stress and strength value. We used an aging law [Dieterich, 1979; Ruina, 1983] as the evolution law for the strength change. To treat the long cycle, we used the quasi-dynamic model [Rice, 1993] with a smaller seismic radiation damping term instead of a fully dynamic model. To solve these equations, we removed the time derivative of stress and obtained differential equations for slip rate and strength. The differential equations were solved using an adaptive time step fifth-order Runge–Kutta algorithm [Press et al. 1996]. For the boundary conditions, the slip velocity was assumed to be constant and equal to the plate convergence rate (Vpl) outside the model region. For the initial conditions, the slip velocity was assumed to be uniform at 0.9Vpl.
We discretized the curved subducting plate interface into small planar triangular elements. We treat three triangular elements as one rectangular-like subfault [Hyodo et al., 2016]. The time derivative of the shear stress at the i-th subfault due to the unit slip at the j-th subfault is represented by the combined effect of angular dislocations through three triangular elements within a rectanglar-like subfault [Comninou and Dundurs, 1975]. The stress change was estimated at the center of each subfault in the direction of slip acceleration in a homogeneous elastic half-space. We used the realistic 3D geometry of the subducting Philippine Sea Plate and Pacific Plate along the Japan Trench [Baba et al. 2003; 2006]. The free surface is shifted to the depth of the trough axis to avoid unrealistic stress concentration at the tip [Hyodo and Hori 2013]. The total number of approximately 1-km-long subfaults in the strike direction was up to approximately 520,000 in the area used in our previous study.
In our simulations, the frictional parameters A, B, and L were assumed to be constant over the earthquake cycle. In addition, we assumed a depth-dependent heterogeneity of A–B. The frictional parameters in the seismogenic zone are A–B < 0 (i.e., velocity weakening in steady-state slip). The computational resources of the Earth Simulator provided by JAMSTEC and of Cyberscience Center, Tohoku University were used for all simulations. As a result, when we assumed a single-circular patch, we obtained earthquakes with constant recurrence intervals. When we applied the multiscale-circular patch model, large to moderate earthquakes ruptured simultaneously with closely spaced and partially overlapping patches, showing the complex earthquake cycles with a wide range of magnitudes.