Figure 5a shows the schematic relationship between the low-resistivity zones and earthquake magnitude. A localized stress accumulation around the mechanically weak low-resistivity zones, and/or fluid supply from the low-resistivity zone have been proposed as potential crustal earthquake nucleation mechanisms around low-resistivity zones (e.g., Ogawa et al., 2001; Ichihara et al., 2014; Aizawa et al., 2017; Cai et al., 2017). However, these proposed mechanisms might not account for the relationship between the low-resistivity zones and the final earthquake magnitudes. The large earthquake did not occur distal to the low-resistivity zones, even though the stress field was estimated high there (Matsumoto et al., 2016). Therefore, we argue that the pore-fluid pressure—which is considered to be high inside and near low-resistivity zones, such as deep magmatic fluid zones (Fournier, 1999; Lee et al., 2020) or a fracture zone that transports magmatic volatiles (Aizawa et al., 2016; Lee et al., 2020)—plays an important role in the evolution of crustal earthquake rupture. We hypothesize that the pre-failure pressure/temperature (PT) gradient (spatial difference) of the pore fluids contributes to the propagation and arrest of earthquake rupture. This proposed mechanism is based on Griffith's criterion, which states that real rocks must contain flaws, such as cracks and voids. Note that the observed resistivity values (1–20,000 Ωm range; Figs. 2 and 3) suggest that fluid-filled cracks can exist anywhere in the crust because dry granite and gabbro possess electrical resistivities of > 100,000 Ωm at < 400 °C (Kariya and Shankland, 1983; Fuji-ta et al., 2004). When the rupture nucleates at one of the cracks, it can propagate along other cracks, and these rupturing cracks can coalesce. Such crack coalescence can enhance both the rupture magnitude and propagation speed, culminating in macroscopic rupture (Ashby and Hallam, 1986; Kame and Yamashita, 1997). Tensile cracking may be generated when the rupture (slip) nucleates at one of the cracks (Fig. 5b) (Ashby and Hallam, 1986; Liu et al., 2017), which is consistent with the novel view of crustal earthquakes, whereby rupture nucleation not only induces slip but also accompanies the opening of tensile cracks due to the presence of localized high-pressure fluids (Hayashida et al., 2020). If the rupture nucleates at a crack near the edge of a low-resistivity zone and crack coalescence occurs outward from the low-resistivity zones, then the pre-failure PT gradient in the pore fluids may promote pore-fluid migration into the crack, which may enhance crack propagation and widening at the tip of the coalesced crack (Fig. 5b). This process likely occurs successively at various spatial scales, and subsequently tends to advance the rupture front; it also yields a high probability of resulting in a large earthquake. Conversely, if the rupture nucleated either within or distal to the low-resistivity zone, then the small PT gradient in the pore fluids may be less likely to promote rupture growth, resulting in a small earthquake. The relative pore-pressure difference would be maintained based on the phase diagram of water (Chen et al., 2017), even if water vaporization occurs along the edge of the newly opened crack.
The PT gradient may also contribute to the arrest of the rupture. The presence of high pore-fluid pressure inherently acts to slow rupture growth due to dilatant hardening (French and Zhu, 2017). Furthermore, rupture propagation from a zone of low PT conditions to a zone of high PT conditions is considered to be reduced by the opposite mechanism, as shown in Fig. 5b. This means that the cracks with high PT fluids act as barriers to rupture propagation. However, the high PT condition may act to advance the rupture front away from the low-resistivity zones once the rupture is well within these zones. C1N arrested the rupture of the Mw 6.2 foreshock of the 2016 Kumamoto earthquake sequence, but this zone was highly damaged by the foreshock and subsequent aftershocks, which reduced the pore fluid pressure around C1N and potentially placed it in a vulnerable state for the next large rupture propagation. Furthermore, the spatial scale and slip amount at the mainshock rupture front (10 km and 2 m, respectively) were far larger than those at the foreshock rupture front (5 km and 0.5 m) when the rupture reached C1N; therefore, the mainshock rupture might be able to propagate deep into C1N, resulted in the promotion of the mainshock.
We found changes in the seismicity before and after the 2016 Kumamoto earthquake sequence. The number of earthquakes that nucleated within the low-resistivity zones before the earthquake sequence was approximately 16% of the total number of the earthquakes; this number decreased to 8% after the earthquake sequence. This observation may indicate that pore pressure in the low-resistivity zones decreased during the 2016 Kumamoto earthquake sequence due to pore-fluid pressure diffusion via the mechanism proposed in Fig. 5b. Large (M > 5) earthquakes only occurred at the top of C1 and C2 before the 2016 Kumamoto earthquake sequence, whereas they also occurred along the outer edges of the shallow sections of C1S and C1N after the earthquake sequence. The observed changes in the hypocenter locations of the large earthquakes are also consistent with the speculation that the high pore-fluid pressure diffused from the low-resistivity zones to the surrounding high-resistivity zones.
Another change in the seismicity before and after the 2016 Kumamoto earthquake sequence was observed at the eastern edge of C1. No earthquakes occurred at > 15 km depth before the 2016 Kumamoto earthquake sequence, whereas the Mw 6.2 foreshock (18 km depth slice in Fig. 2) marked a shift to the nucleation of deeper earthquakes. The deep aftershocks are actually located along the eastern edge of C1 (Fig. 2), and magmatic fluids likely moved horizontally along the rupture formed by these deep aftershocks. This may indicate that the lateral expansion of C1 in the lower crust was stimulated by the rupture of a large crustal earthquake. Deep low-resistivity zones, similar to C1, have been found beneath active faults (e.g., Ogawa et al., 2001; Yoshimura et al., 2009; Becken et al., 2011; Cai et al., 2017). These results suggest that there may be a positive feedback whereby deep fluidized zones generate large crustal earthquakes and induce the evolution of high-temperature fluidized zones in the mid-to-lower crust.
The low-resistivity zones and their observed relationship to local seismicity suggest that the mid-crustal fluid distribution controls the nucleation, growth, and arrest of crustal earthquake rupture. These results demonstrate that the 3-D imaging of electrical low-resistivity zones using spatially dense MT observations provides valuable information for assessing the locations and maximum magnitudes of future earthquakes, particularly since some large (M 6–7) intraplate earthquakes have occurred in zones with no significant surface fault traces (Semmane et al., 2005). It has recently been suggested that the strong frictional coupling (plate locking) of megathrust interplate earthquakes occurs in relatively high-resistivity zones, which are interpreted to be fluid- and sediment-sparse zones (Wannamaker et al., 2014; Heise et al., 2017). Therefore, evaluations of the potential for large earthquakes and their locations in a given region may be refined when we take into account the locations of low-resistivity fluidized zones, regardless of earthquake type.