17:15 〜 18:45
[SSS05-P13] How the choice of an evolution law for an RSF fault affects the acceleration and localization behaviors of preseismic slip: Insight from dynamic earthquake cycle simulation
キーワード：地震前滑り、地震核形成、動的地震サイクルシミュレーション、速度状態依存摩擦
Investigating characteristic fault behaviors preceding large earthquakes is important. Seismological observations have demonstrated that the seismicity rate, a potential indicator of preseismic slip, increases before not all but some large earthquakes. Such acceleration may be modeled by the inverse Omori law, where the seismicity rate is inversely proportional to the power of the time-to-failure. The localization of the seismic activity toward the hypocenter of a mainshock is another unique preseismic behavior. Ben-Zion and Zaliapin (2020) (hereafter, BZ2020) reported various localization behaviors in that the localization degree (i) progressively increases, (ii) increases and subsequently keeps almost constant, or (iii) increases and subsequently decreases, before some large earthquakes in the southern California, USA.
Rate- and state-dependent friction (RSF) is probably the most popular constitutive law for modeling the fault motion. Noda et al. (2013a) performed three-dimensional dynamic earthquake cycle simulations (ECS) of a circular RSF fault with the aging law (AG) and showed that the preseismic moment rate is inversely proportional to the time-to-failure. Noda et al. (2013b) also reported milder preseismic acceleration for the slip law (SL) in a quasi-dynamic framework. They also demonstrated that the high-slip-rate region with SL is progressively migrated and localized while that with AG keeps almost constant size without migration. The results of SL can qualitatively explain the localization behavior of the case (i) of BZ2020, whereas not the other two cases. Previous quasi-static modeling showed the progressive expansion and shrink of the high-slip-rate region for AG and SL, respectively. Similar quasi-static simulations with the Nagata law (NGT), which accounts for the stress-weakening effect, showed the slip behaviors between those of AG and SL. Therefore, the localization behaviors of the cases (ii) and (iii) of BZ2020 may be realized with NGT, although the dynamic ECS with NGT is necessary. The acceleration behaviors with NGT are also of great interest.
In the present study, we performed dynamic ECS of an RSF fault with different evolution laws, with a particular interest in NGT, and investigate how the choice of an evolution law affects the acceleration and localization behaviors of the preseismic slip. Since the dynamic ECS with SL require impractically large computational resources, we adopted the modified slip law (MSL) instead of SL as in Noda (2022). We consider the elastodynamic, two-dimensional, and anti-plane problem where a linear fault is embedded in a linearly elastic infinite medium. We first implemented NGT into the code developed by Noda (2021, 2022) for the dynamic ECS with the spectral boundary integral equation method (SBIEM). The second-order accuracy was realized in our quasi-dynamic convergence test, which supports the validity of our implementation of NGT. Then, we performed the dynamic ECS with the SBIEM by employing AG, MSL, and NGT. As far as we know, the present study is the first to perform the dynamic ECS with NGT and to investigate in detail the preseismic behaviors with an SL-like evolution law.
The fault behaviors obtained with NGT are between those of AG and MSL, as suggested by the previous quasi-static modeling. The preseismic moment rates of AG and NGT were inversely proportional to the time-to-failure, whereas that of MSL showed milder acceleration. Among three evolution laws, only SL showed the dependence on the background slip rate. The absolute level of the preseismic moment rate was AG>NGT>MSL, indicating that the difficulty in the imminent earthquake prediction is AG<NGT<MSL. The size of high-slip-rate region keeps almost constant with AG, firstly decreases and then keeps almost constant with NGT, and progressively decreases with MSL. Therefore, the localization behavior of the case (ii) of BZ2020 can be qualitatively explained by NGT. For further investigating the localization at the final stage of the earthquake nucleation, we examined the creeping and nucleation regions at the onset of earthquakes. The creeping and nucleation widths, as well as the ratio of the latter to the former, were the largest for AG and the smallest for MSL. Those of NGT were between these two laws, and the stronger stress-weakening effect resulted in the smaller values. The creeping widths required to make a certain size of the nucleation were AG<NGT<MSL, indicating that the spatial density of the observatories required for geodetically resolving the signals is AG>NGT>MSL.
Rate- and state-dependent friction (RSF) is probably the most popular constitutive law for modeling the fault motion. Noda et al. (2013a) performed three-dimensional dynamic earthquake cycle simulations (ECS) of a circular RSF fault with the aging law (AG) and showed that the preseismic moment rate is inversely proportional to the time-to-failure. Noda et al. (2013b) also reported milder preseismic acceleration for the slip law (SL) in a quasi-dynamic framework. They also demonstrated that the high-slip-rate region with SL is progressively migrated and localized while that with AG keeps almost constant size without migration. The results of SL can qualitatively explain the localization behavior of the case (i) of BZ2020, whereas not the other two cases. Previous quasi-static modeling showed the progressive expansion and shrink of the high-slip-rate region for AG and SL, respectively. Similar quasi-static simulations with the Nagata law (NGT), which accounts for the stress-weakening effect, showed the slip behaviors between those of AG and SL. Therefore, the localization behaviors of the cases (ii) and (iii) of BZ2020 may be realized with NGT, although the dynamic ECS with NGT is necessary. The acceleration behaviors with NGT are also of great interest.
In the present study, we performed dynamic ECS of an RSF fault with different evolution laws, with a particular interest in NGT, and investigate how the choice of an evolution law affects the acceleration and localization behaviors of the preseismic slip. Since the dynamic ECS with SL require impractically large computational resources, we adopted the modified slip law (MSL) instead of SL as in Noda (2022). We consider the elastodynamic, two-dimensional, and anti-plane problem where a linear fault is embedded in a linearly elastic infinite medium. We first implemented NGT into the code developed by Noda (2021, 2022) for the dynamic ECS with the spectral boundary integral equation method (SBIEM). The second-order accuracy was realized in our quasi-dynamic convergence test, which supports the validity of our implementation of NGT. Then, we performed the dynamic ECS with the SBIEM by employing AG, MSL, and NGT. As far as we know, the present study is the first to perform the dynamic ECS with NGT and to investigate in detail the preseismic behaviors with an SL-like evolution law.
The fault behaviors obtained with NGT are between those of AG and MSL, as suggested by the previous quasi-static modeling. The preseismic moment rates of AG and NGT were inversely proportional to the time-to-failure, whereas that of MSL showed milder acceleration. Among three evolution laws, only SL showed the dependence on the background slip rate. The absolute level of the preseismic moment rate was AG>NGT>MSL, indicating that the difficulty in the imminent earthquake prediction is AG<NGT<MSL. The size of high-slip-rate region keeps almost constant with AG, firstly decreases and then keeps almost constant with NGT, and progressively decreases with MSL. Therefore, the localization behavior of the case (ii) of BZ2020 can be qualitatively explained by NGT. For further investigating the localization at the final stage of the earthquake nucleation, we examined the creeping and nucleation regions at the onset of earthquakes. The creeping and nucleation widths, as well as the ratio of the latter to the former, were the largest for AG and the smallest for MSL. Those of NGT were between these two laws, and the stronger stress-weakening effect resulted in the smaller values. The creeping widths required to make a certain size of the nucleation were AG<NGT<MSL, indicating that the spatial density of the observatories required for geodetically resolving the signals is AG>NGT>MSL.