4.3 Upward migration of the aftershocks along several planes
The aftershock sequence of the Kagoshima Bay earthquake sequence follows Omori’s law (Fig. 1d), suggesting that this sequence was triggered by the M5.3 mainshock. However, the aftershock sequence slightly deviates from the prediction based on the ETAS model (Figs 9b–c). The transient increase in the background seismicity rate suggests that the Kagoshima Bay earthquake sequence may have been affected by physical processes other than earthquake-to-earthquake interactions, especially during this period (20 to 40 days) and that these aseismic processes may have led to the largest aftershock (ML 4.4) that occurred 44 days after the mainshock. Based on the model simulations of fluid injection-induced seismicity, Hainzl and Ogata (2005) pointed out that the background seismicity rate of the ETAS model is sensitive to the amount of injected water. In previous studies, similar observations were made for fluid injection-induced seismicity and natural earthquake sequences (Llenos & Michael, 2013; Yoshida & Hasegawa, 2018b; Kumazawa et al., 2019).
Our results indicate that the aftershock hypocenters migrated toward the shallower portion on multiple planes. Such upward movements of hypocenters have been previously reported for earthquake swarms following nearby large earthquakes and it has been concluded that they reflect the upward pore pressure migration associated with the fault-valve behavior (Shelly et al., 2015; Ruhl et al., 2016; Yoshida & Hasegawa, 2018a, b). Examples are the earthquake swarms that occurred in northeastern Japan following the 2011 Tohoku-Oki earthquake (Yoshida et al., 2016a; Yoshida & Hasegawa, 2018a, b). The earthquake swarms might originate from the pore pressure increase because (1) they occurred in the stress shadow of the 2011 Tohoku-Oki earthquake with a time delay of a few weeks despite the reduction in the shear stress, (2) they are located beneath the caldera structures that are believed to host shallow igneous bodies, with hydrothermal fluids immediately below, (3) they are located a few kilometers above S-wave reflectors and the low-velocity zone including fluids, and (4) their hypocenters migrate upward (Yoshida & Hasegawa, 2018a; Yoshida et al., 2019a). The Kagoshima Bay swarm was also located beneath an ancient caldera and involved the upward migration of aftershocks, which can be explained by an increase in the pore pressure. Fluid paths in the crust may have expanded due to the deformation and shaking associated with the mainshock. Pore pressure migration may explain deviations in the seismicity rate from Omori’s law. These observations are consistent with the prediction based on the fault-valve model proposed in Sibson (1992), that is, upward fluid discharge after the mainshock. In recent geodetic studies, a porosity wave associated with the fault-valve action was detected (Rossi et al., 2016 and 2018).
We presume that the subducting Philippine Sea Plate is the source of fluids, similar to the model reported in Hasegawa et al. (2005), which is based on the geophysical and geological observations in northeastern Japan. This hypothesis is supported by seismic data obtained in Kyushu using tomography, which indicate that the existence of an inclined low-velocity layer continuously distributed in the mantle wedge and reaching right below the volcanic front as northeastern Japan (Zhao et al., 2012). The low-velocity zone is considered to represent the ascending flow portion of the secondary convection within the mantle wedge and therefore contains fluids from the slab and resultant melts (Hasegawa et al., 2005). The buoyancy facilitated the upward migration of the fluids, as shown in simulations (e.g., Iwamori, 1998; Wada et al., 2015; Horiuchi et al., 2016), and the fluids reached the source region of the Kagoshima Bay sequence.