1 Introduction
An earthquake is a natural phenomenon during which a high-speed rupture propagates along a fault. Two factors control the occurrence of an earthquake: an increase in the shear stress acting on the fault and a decrease in the fault strength. The results of previous studies suggested that the increase in the pore pressure plays an important role in the earthquake occurrence (e.g., Hasegawa, 2017; Hubbert & Rubey, 1959; Nur & Booker, 1972; Sibson, 1992; Rice, 1992) because it reduces the fault strength.
A well-known example of fluid-driven seismicity is the seismicity induced by fluid injection for engineering purposes (e.g., Ellsworth, 2013). There is also growing evidence that natural earthquake swarms are closely related to fluid movement at depth. In fact, the characteristics of many natural seismic swarms are similar to those of fluid injection-induced seismicity including the migration behavior of the earthquake hypocenter (e.g., Fischer and Horálek, 2003; Parotidis et al., 2003; Bianco et al., 2004; Yukutake et al., 2011; Shelly et al., 2016; Yoshida et al., 2016a; Ruhl et al., 2016; De Barros et al., 2019). Based on the determination of the hypocenters and focal mechanisms of earthquake swarm at the 2009 Hakone volcano, the diffusion of high-pressure fluid triggered the swarm (Yukutake et al., 2011). The spatiotemporal evolution of seismic activity in the Long Valley Caldera, California, indicates that a pore pressure transient with a low-viscosity fluid initiated and sustained the swarm in 2014 (Shelly et al., 2016). It has been hypothesized that several earthquake swarms that occurred after the 2011 Tohoku-Oki earthquake were triggered by a decrease in the fault strength due to upward pore pressure migration (Terakawa et al., 2013; Okada et al., 2016; Yoshida et al., 2016a, 2019a).
Not only earthquake swarms but also foreshock–mainshock–aftershock sequences may be closely related to the fluid behavior in the Earth interior. Sibson (1992) established the fault-valve model in which the pore pressure cycle controls the earthquake cycle due to overpressurized fluids that rise from the deeper portion of the fault. In this model, fault ruptures create a transient fracture permeability within the fault zone, which acts as a valve, promoting the upward discharge of fluids from deeper portions of the crust. This model is supported by various geological and geophysical observations (Sibson, 2020). Hasegawa et al. (2005) proposed a model for the deformation process in a subduction zone based on various geophysical observations including seismic tomography data obtained for northeastern Japan. In this model, fluids expelled from the subducting slab migrate upward, reach the crust, and cause anelastic crustal deformation including earthquakes.
The migration characteristics of earthquake hypocenters can be used to infer the origin of the seismicity (e.g., Yukutake et al., 2011; Ruhl et al., 2016; Yoshida & Hasegawa, 2018a,b; De Barros et al., 2019). Pore pressure migration and aseismic slip propagation are typical mechanisms attributed to the migration of earthquakes. In the former mechanism, the hypocenter migration is presumed to reflect the migration of fluids (e.g., Shapiro et al., 1997; Talwani et al., 2007). In the latter mechanism, the hypocenter migration is presumed to be a result of aseismic slip propagation (e.g., Lohman & McGuire, 2007; Roland & McGuire, 2009). The spatiotemporal distribution of earthquake hypocenters can be more precisely estimated than other seismological characteristics such as the three-dimensional seismic velocity structure. By examining relocated hypocenters, we may extract information on aseismic physical processes controlling earthquakes, which is crucial to understanding the earthquake generation. The results of previous studies showed that the seismic activity caused by aseismic processes differs from that of the mainshock–aftershock sequence (e.g., Hainzl & Ogata, 2005; Roland & McGuire, 2009; Kumazawa & Ogata, 2013; Yoshida & Hasegawa, 2018b). This suggests that investigations of the seismicity may provide clues about aseismic processes governing earthquakes.
The volcanic front on Kyushu Island in southern Japan formed due to the subduction of the Philippine Sea Plate. Several of the most active volcanoes in Japan are distributed along this volcanic front (e.g., Sakurajima and Aso). Kagoshima Bay is located at this volcanic front (Fig. S1), which is characterized by a low-gravity anomaly that extends from north to south. On July 11, 2017, an ML 5.3 strike-slip earthquake occurred at a depth of ~10 km in Kagoshima Bay (Fig. 1). Seismicity activity had been recorded near the mainshock hypocenter since December 2016 (Fig. 1c). In total, 1.843 foreshock events were recorded and listed in the Japan Meteorological Agency (JMA) unified seismic catalogue. The seismicity increased after the mainshock; 12.595 events are listed in the JMA catalogue. Based on the focal mechanisms of earthquakes in this region, these events were of strike-slip type with a NW–SE P-axis (Fig. 1b). Only a small coseismic step was detected by the national GNSS (Global Navigation Satellite System) network (Fig. S2). Based on the spatiotemporal variation in the b-value and the migration of the hypocenters, Nanjo et al. (2018) suggested that fluid movement caused the earthquake sequence in Kagoshima Bay, but the detailed physical process controlling this foreshock–mainshock–aftershock sequence remains unclear.