Presentation Information
[SCG56-52]Numerical study of estimating time interval between successive megathrust earthquakes along the Nankai Trough, Japan: Incorporating viscoelastic crustal deformation
*Sota Murakami1, Ryota Kaneko2, Yoshihisa Hiyoshi1, Takane Hori1, Kohei Fujita2, Tsuyoshi Ichimura2, Takeshi Iinuma1 (1.Japan Agency for Marine-Earth Science and Technology, 2.Earthquake Research Institute, The University of Tokyo)
Along the Nankai Trough, the plate interface extending from off the Tokai through off the Kii Peninsula to off Shikoku is known to rupture either as a single sequence of great earthquakes or as separate eastern and western events segmented near the Kii Peninsula. In the latter case, the time interval between the eastern and western earthquakes varies widely from about one day to more than two years. Estimating this interval is important for planning post-disaster response and recovery activities following the first event. Hori et al. (2014) proposed a method to estimate the time interval between successive earthquakes based on seafloor crustal deformation observations. In this study, we incorporate recent locked-zone estimates (Sato et al., 2025) with a three-dimensional viscoelastic crustal deformation analysis method and conduct numerical experiments using a model that more realistically reflects the spatial distribution of locked zones and the viscoelastic structure. Based on this framework, we perform a preliminary investigation of time-interval estimation.
We focus on the case in which rupture first occurs in the eastern locked zone and perform earthquake cycle simulations while varying the critical slip distance L near the tip of the Kii Peninsula. As L increases, the region functions as a stronger barrier, resulting in reduced coseismic slip. Consequently, the stress perturbation around the western locked zone becomes smaller, the increase in slip velocity in the western locked zone is suppressed, and the time interval becomes longer. As a result, many scenarios exhibiting time intervals ranging from one day to two years were obtained.
Next, using the time evolution of slip derived from the cycle simulations, we computed crustal deformation at ocean-bottom stations using a viscoelastic crustal deformation analysis that incorporates detailed crustal structure. Synthetic data were generated by selecting a true case from the calculated scenarios and adding observational errors. The likelihood of each scenario was then evaluated against the synthetic data. Based on these likelihoods, we calculated the expected value of the time interval and the total probability for each time interval range. The results show that the time interval can be constrained to some extent from observation data, as short-interval cases (within one month after the first earthquake) and long-interval cases (three months or more after the first earthquake) exhibit markedly different afterslip behavior and associated crustal deformation. Furthermore, when the wrong viscoelastic structure was assumed, the resulting modeling errors became significantly larger than the differences among scenarios, leading to degraded estimation performance. These results suggest that more accurate identification of the viscoelastic structure is essential for more reliable estimation of the time interval between successive earthquakes.
We focus on the case in which rupture first occurs in the eastern locked zone and perform earthquake cycle simulations while varying the critical slip distance L near the tip of the Kii Peninsula. As L increases, the region functions as a stronger barrier, resulting in reduced coseismic slip. Consequently, the stress perturbation around the western locked zone becomes smaller, the increase in slip velocity in the western locked zone is suppressed, and the time interval becomes longer. As a result, many scenarios exhibiting time intervals ranging from one day to two years were obtained.
Next, using the time evolution of slip derived from the cycle simulations, we computed crustal deformation at ocean-bottom stations using a viscoelastic crustal deformation analysis that incorporates detailed crustal structure. Synthetic data were generated by selecting a true case from the calculated scenarios and adding observational errors. The likelihood of each scenario was then evaluated against the synthetic data. Based on these likelihoods, we calculated the expected value of the time interval and the total probability for each time interval range. The results show that the time interval can be constrained to some extent from observation data, as short-interval cases (within one month after the first earthquake) and long-interval cases (three months or more after the first earthquake) exhibit markedly different afterslip behavior and associated crustal deformation. Furthermore, when the wrong viscoelastic structure was assumed, the resulting modeling errors became significantly larger than the differences among scenarios, leading to degraded estimation performance. These results suggest that more accurate identification of the viscoelastic structure is essential for more reliable estimation of the time interval between successive earthquakes.