Figure 10. Temporal evolution of the distances between the foreshocks and initial hypocenter. Blue circles represent the hypocenters expressed by the size corresponding to the JMA magnitude. The black curves show the fluid diffusion models with Dh= 0.01, 0.03, and 0.05 m2/s. Gray straight lines show the linear spread model with migration speeds of d = 0.001, 0.003, and 0.005 km/h.
The pore pressure diffusion model with a hydraulic diffusion coefficient of ~0.05 m2/s matches the observations better than the linear spread model. In previous studies, it has been estimated that the hydraulic diffusion coefficient in the crust ranges from ~0.01–10 m2/s (e.g., Talwani et al., 2007; Shelly et al., 2016; Yoshida & Hasegawa, 2018a), which is similar to the foreshock migration speed of the M5.3 Kagoshima Bay earthquake sequence. Based on the linear spread model, the propagation velocity is ~0.001–0.005 km/h. Based on previous studies, the migration speed of aseismic slip propagation ranges from 0.1–1.0 km/h (e.g., Lohman & McGuire, 2007; Kato et al., 2016), which is significantly higher than the migration speed of the present foreshock activity. If we advance the initiation timing of propagation, the propagation speed decreases. Thus, according to the migration speed and spatiotemporal pattern of the foreshocks, the pore pressure diffusion model better explains the overall migration of the foreshock hypocenters.
Aseismic creep related to the nucleation process of the mainshock might be involved in the migration of the foreshocks. In fact, physical simulations indicate interseismic creep in seismogenic patches from external stable-slip regions before the occurrence of unstable slip (Tse & Rice, 1986). Such an expansion of quasi-static slip prior to the mainshock may explain the current migration of the foreshocks (e.g., Dodge et al., 1996; Yabe & Ide, 2018). However, the source of the mainshock is smaller than the foreshock area (Fig. 6a), which contradicts the hypothesis because the mainshock rupture zone should include the nucleation area. Note that the size of the source of the mainshock was estimated based on source models assuming a constant rupture velocity (subshear rupture propagation). If the assumptions differ from reality (e.g., supershear rupture propagation), the source size may differ from our estimation, explaining this contradiction. However, the aftershocks migrate upward on multiple planes (Fig. 8d), which can be explained with the foreshock sequence if the pore pressure migration model is adopted. Thus, we prefer the hypothesis that pore pressure migration is primarily responsible for the generation of the 2017 M5.3 Kagoshima Bay earthquake sequence. The heterogeneity in the permeability and/or pore pressure along the fault may explain the up- and downward movement of the hypocenter along the plane (Fig. 8c).
However, recent observations of fluid injection-induced seismicity and natural earthquake swarms suggest that an increase in the pore pressure can cause aseismic slip (Cornet et al., 1997; Guglielmi et al., 2015; Ruhl et al., 2016; Yoshida & Hasegawa, 2018a; De Barros et al., 2020). In the presence of fluids, the effective normal stress decreases and the critical nucleation size increases; thus, the occurrence of aseismic slip is likely (e.g., Scholz, 1998). The increase in the pore pressure also accelerates creep in the stable-slip segment of the fault. Both aseismic slip and fluid movement may have contributed to the occurrence of foreshocks. Furthermore, the poroelastic effects associated with pore pressure migration (Segall, 1989; Goebel et al., 2018) and the earthquake-to-earthquake interaction (Helmstetter, 2002) may contribute to the occurrence of earthquakes.