2:15 PM - 2:30 PM
[SCG45-24] Numerical simulation with a multiscale circular patch model in the northern segment along the Japan Trench (III)
In subduction zone, various magnitude and types of earthquakes occurred such as megathrust earthquake, tsunami earthquakes, repeating earthquakes and quasi-repeating earthquakes. To reproduce some of them, we have approximated spatial heterogeneity as circular or rectangular patches and treated them as one- or two-hierarchical structure [Nakata et al., 2016; 2021]. However, it is important to explain various phenomena in a model by including hierarchical characteristics. Then, we conducted numerical simulations of earthquake generation cycles by adopting a multiscale patch model [Ide & Aochi, 2013]. Here, we performed off Sanriku region where various earthquakes have been observed. For example, in the northern segment of the Japan Trench, at the shallow area, M~8 earthquakes such as the 1896 Meiji-Sanriku earthquake (Mw8.1) [Tanioka & Satake, 1996; Satake et al., 2017] occurred, and at deeper area, smaller magnitude of earthquakes, repeating earthquakes, and slow slips [Uchida et al., 2016] occurred.
Our simulation is based on the rate- and state-dependent friction (RSF) law that represents the process of stress accumulation and release on the plate interface. We used the same equations, initial conditions, seismic radiation damping term, plate geometry, and plate convergence rate as in our previous study [Nakata et al., 2016; 2021] that treated M > 7 earthquakes. Here, a multiscale circular patch model [Ide & Aochi, 2013] represented the spatial heterogeneity of seismic events with M>5.5 in the northern segment. To reproduce M ~ 5 earthquakes in the numerical simulation, the plate interface weas discretized to a smaller size than in our previous studies. The time step of output in the numerical simulation was also fine-tuned.
For the Meiji Sanriku earthquake, we used that of the second-level patch from Ide & Aochi [2013]. For Mj=7.1–7.6 and 6.6–7.2 earthquakes, we used the third and fourth-level patches from Ide & Aochi [2013]. In addition, we added smaller earthquakes. For fifth-level patches, corresponding to events of Mj5.6–6.5 earthquakes, we used hypocenters determined by JMA since 1923. Each event was approximated in a four-hierarchical structure according to magnitude. For patches in four-size, we assumed four-level heterogeneity of the characteristic slip distance. We assumed uniform A-B regardless of the magnitude of the earthquake.
In tentative results, M5–8 earthquakes frequently occurred during 45 years. Earthquakes of smaller magnitude occurred mainly on the marginal area of the largest patch, because the large and small patches overlapped to each other. Then no earthquake occurred within the largest patch at the shallow part. This spatial distribution is inconsistent with the observations.
Then, patch of Level 3 decreased from six to three. Patches of Level 5 was reduced from 360 to 30 by grouping together, considering the uncertainty of the epicenter and the possibility that repeating earthquakes may be included. Then, the number of earthquakes included in the new model were 1, 3, 9, and 30 in Level 2, 3, 4, and 5, respectively. As a result, M5–8 earthquakes frequently occurred during 75 years. The magnitude-frequency distribution of obtained earthquakes partly followed the Gutenberg-Richter’s (GR) law. But few earthquake occurred within the largest patch at the shallow part.
Now we are examining how to load and rupture each patch separately. Simple approach of earthquake cycle simulation based on RSF law with multiscale patches may not be able to explain observed multiscale seismicity. It may require major modifications.
Our simulation is based on the rate- and state-dependent friction (RSF) law that represents the process of stress accumulation and release on the plate interface. We used the same equations, initial conditions, seismic radiation damping term, plate geometry, and plate convergence rate as in our previous study [Nakata et al., 2016; 2021] that treated M > 7 earthquakes. Here, a multiscale circular patch model [Ide & Aochi, 2013] represented the spatial heterogeneity of seismic events with M>5.5 in the northern segment. To reproduce M ~ 5 earthquakes in the numerical simulation, the plate interface weas discretized to a smaller size than in our previous studies. The time step of output in the numerical simulation was also fine-tuned.
For the Meiji Sanriku earthquake, we used that of the second-level patch from Ide & Aochi [2013]. For Mj=7.1–7.6 and 6.6–7.2 earthquakes, we used the third and fourth-level patches from Ide & Aochi [2013]. In addition, we added smaller earthquakes. For fifth-level patches, corresponding to events of Mj5.6–6.5 earthquakes, we used hypocenters determined by JMA since 1923. Each event was approximated in a four-hierarchical structure according to magnitude. For patches in four-size, we assumed four-level heterogeneity of the characteristic slip distance. We assumed uniform A-B regardless of the magnitude of the earthquake.
In tentative results, M5–8 earthquakes frequently occurred during 45 years. Earthquakes of smaller magnitude occurred mainly on the marginal area of the largest patch, because the large and small patches overlapped to each other. Then no earthquake occurred within the largest patch at the shallow part. This spatial distribution is inconsistent with the observations.
Then, patch of Level 3 decreased from six to three. Patches of Level 5 was reduced from 360 to 30 by grouping together, considering the uncertainty of the epicenter and the possibility that repeating earthquakes may be included. Then, the number of earthquakes included in the new model were 1, 3, 9, and 30 in Level 2, 3, 4, and 5, respectively. As a result, M5–8 earthquakes frequently occurred during 75 years. The magnitude-frequency distribution of obtained earthquakes partly followed the Gutenberg-Richter’s (GR) law. But few earthquake occurred within the largest patch at the shallow part.
Now we are examining how to load and rupture each patch separately. Simple approach of earthquake cycle simulation based on RSF law with multiscale patches may not be able to explain observed multiscale seismicity. It may require major modifications.