Discussions

The BP results of 2013 Okhotsk deep earthquake sequence are summarized in the Figure 3.11. As the mainshock, the integrated P- and pP-waves 3D BP images suggest that this most recent largest deep earthquake consisted of two subhorizontal ruptures toward the opposite directions. The en-echelon like anti-parallel ruptures separate at least 10~15 km in depth as the cross-sections show in Figure 3.12. Other than that, based on the focal depth of aftershocks collected from the ISC catalog (International Seismological Centre, 2015) and relocated by Chen et al. (2014) could be divided into two groups. The shallower aftershock group roughly has comparable focal depth of the mainshock; whereas there is another deeper group near the terminus of the 2^nd S-ward rupture. The Figure 3.12 demonstrates the coincident distribution of the 3D BP results

of mainshock and the aftershocks. The depth aperture of ~15 km between the upper and lower rupture zone is also observed by Chen et al. (2014) who obtained one subevent starting at 12 s after the earthquake nucleated and having focal depth about 16.7 km deeper than hypocenter by using multiple point source inversion technique.

The 3D BP results and the aftershocks distributions clarify that the rupture process of great 2013 Okhotsk deep earthquake consists two subhorizontal shear zones. The

paired shear ruptures have been observed in other large deep earthquakes. For example, Chen et al. (1996) utilized the P and SH waveform inversion method and discovered the “en-chelon” rupture feature for several large deep earthquakes such as the 1994 great Bolivia earthquake and the 1994 Japan deep earthquake. Moreover, the deep earthquakes showing double rupture zones are also observed in Tonga, Izo-bonin subduction zones (Furukawa, 1994; Iidaka and Furukawa, 1994; Wiens et al., 1994).

For the 2013 Okhotsk mainshock, the 3D BP results also revealed different rupture speeds of two-stages of en-chelon like anti-parallel ruptures. The 1^st NE-ward rupture almost reaches the northern edge of the Pacific slab and propagates in much slower velocity (~0.64Vs) than the 2^nd S-ward rupture stage and two aftershocks (>0.82Vs). Varying rupture speeds along different section of the same subducting plate implies the importance of the slab temperature which might affect the earthquake rupture behaviors, because the northernmost slab edge is believed to be warmed by both the ambient mantle and corner asthenosphere flow (Peyton et al., 2001; Park et al., 2002; Davaille and Lees, 2004a). The shorter NE-ward rupture length also denotes the northern terminus of the seismogenic zone inside the subducting Pacific slab.

Figure 3.11 The tectonic setting of northern Kuril subduction zone and the BP imaging results of 2013 great Okhotsk deep earthquake sequence, including the mainshock and two largest aftershocks.

The rupture propagations for mainshock and two aftershocks are shown in circles with brownish and purplish color-bars. The colors and sizes of the circles represent the elapsed time after the onset of P-wave and the pseudo-spectrum power of BP images. The rupture fault planes of mainshock and aftershocks are highlighted by thick lines in beach balls reported by the Global CMT solutions. Gray circles are the aftershocks in 6 month after mainshock. Gray diamonds show the relocated aftershocks done by Chen et al., 2014. The white circles show the background seismicity since 1900. The blue contours are the slab contours of the subducting Pacific Sea Plate from slab 1.0 model.

Figure 3.12 Cross-sections of the 3D BP results of 2013 Okhotsk mainshock.

Cross-sections along the profiles parallel to the (a) subduction direction (AA’) and (b) slab contour (BB’) as indicated in Figure 3.11. Right panels show the enlarge part of the cross-sections. The background image are the P wave tomography from Fukao and Obayashi (2013). The symbols configurations are the same with those in Figure 3.11.

When an earthquake occurred, the static stress drop could be estimated directly with the confidential rupture length, width of the rupture plane and seismic moment.

According to Starr (1928), the stress drop of a buried dip slip fault is given by

∆𝜎 = 16𝑀₀/(3𝜋𝑆𝑊), where M0, S and W are the seismic moment, fault area and rupture width respectively. The BP imaging represents the spatiotemporal slip

distribution on the fault plane (Fukahata et al., 2014). Therefore the high resolution BP imaging has confidential estimation on the minimum rupture length. According to our 3D BP results, the total rupture length of two anti-parallel ruptures is about 120 km. As the rupture width, it can be roughly assumed by setting the aspect ratio of the

rectangular fault plane from 0.1 to 1.0 to obtain the exaggerated range of stress drop estimations. For aspect ratio of 0.1 and 1; the stress drop are 344 and 3.44 MPa, respectively. By taking the seismogenic width of 40~60 km simulated by numerical modeling of rheological structure of subducted oceanic plate (Karato et al., 2001) as the rupture width approximation, then the static stress drop ranges from 31 to 13 MPa for width. Further compare with the result from teleseismic frequency spectrum analysis, Ye et al. (2013) suggested the stress drop is about 12~15 MPa which is close to the estimation by setting the rupture width of 60 km.

The global slab geometry model, the slab 1.0 (Hayes et al., 2012), optimized by the seismicity may have some uncertainties on sketching the slab geometry for the regions having rare deep-focus seismicity, such as the northern part of Kuril subduction zone. As shown in the cross-sections of Figure 3.3 and Figure 3.12, the 2013 Okhotsk mainshock nucleated outside the slab boundary of slab1.0 model. The high P-wave speed anomaly from the tomographic images by Fukao and Obayashi (2013) illuminates the colder subducted Pacific plate (Figure 3.12) clearly. Due to the limited resolution of the tomography image, the metastable olivine wedge inside the slab could not be discernible here. According to the thermo-kinetic subduction models proposed by Mosenfelder et al. (2001), the maximum depth of the olivine metastable tip should be less than 600 km if the subducting plate having thermal parameters of 4000 to 6000 km. Moreover, the numerical modeling for olivine-spinel transformation

(Marone and Liu, 1997) suggested an extreme narrow seismogenic zone (5~15 km) outside the olivine metastable wedge. Several studies also tried to determine the width and extension depth of metastable olivine wedge by detecting the seismic velocity anomalies (Koper and Wiens, 2000; Kaneshima et al., 2007; Jiang et al., 2008; Kubo et al., 2009; Kawakatsu and Yoshioka, 2011). However, the rupture area estimated by our 3D BP images indicates that 2013 Okhotsk mainshock ruptures beyond the entire metastable wedge and implies that thermal shear instability remains the most plausible mechanism of 2013 Okhotsk deep earthquake.

The triggering of deep earthquakes at great distances has been observed in several subduction systems including Tonga, Japan and Bolivia (Myers et al., 1995; Tibi et al., 2003a, 2003b). The first aftershock “2013AS1” was also dynamic-triggered because of the static stress decreases to only 0.08 MPa by the assumption that two-stage ruptures have equal moment magnitude. Dynamic stress triggering induced by maximum passing seismic wavefields has been proposed to explain the triggered earthquake at great distance for both shallow and deep earthquakes (Kilb et al., 2000; Tibi et al., 2003a). The dynamic stress disturbance may supply the extra stress or pressure at the region facilitating the nucleation of earthquakes. However, much longer delay times of the triggered events do not accommodate with the travel times of the passing seismic wavefields. The first MW 6.7 aftershock,“2013AS1”, located 300 km away in south was triggered in only 9 hours; however the second one half way close to the mainshock occurred after 4 months. We suspect great different time delays responding to the dynamic stress disturbance of two aftershocks is more related to the ambient seismogenic circumstance than the distance to mainshock or the stress changes inherited from the mainshock. As the cross-sections shown in Figure 3.13, the

background seismicity in the north (prof. C) is higher than in the south (prof. D). The event “2013AS1” locates in the vicinity with lower seismicity where may have accumulated vulnerable stress and closed to a more critical circumstance to nucleate earthquake as soon as the area inherits the stress disturbance. Other than the

background seismicity, the variation of thermal structure along the deepest part of the subducting slab may also affect the response time to the stress; however more

seismological observations, dynamic rupture models and slab geodynamic simulations are necessary for further examinations.

Figure 3.13 Cross-sections of the 3D BP results of two aftershocks.

Cross-sections along the profiles of (a) C, (b) D and (c) E as shown in Figure 3.11. The background

image and symbol configurations are the same with those in Figure 3.11 and Figure 3.12.

3.7 Conclusions

We revealed the rupture properties of 2013 Okhotsk deep earthquake sequence by applying the 3D MUSIC BP method. As the mainshock, the low-frequency pP-wave BP image is integrated to the P-wave BP images to ascertain the focal depth difference between two-stage en-echelon rupture feature. We also found both largest aftershock rupture downward in fast rupture speed. In summary, along the same subduction plate, we observe two different rupture speeds: the slow NE-ward rupture of mainshock and the fast 2^nd rupture stage of mainshock and two aftershocks. The significant variation of rupture speeds in the subducting Pacific slab suggests the difference of thermal structure along deepest part of the slab. Also the NE-ward rupture may reach the terminus of the seismogenic zone of the norther tip of the slab which is warmed by the ambient mantle and asthenosphere corner flow.

We also perform a suit of earthquake rupture model tests to assure the reliability of the 3D BP images and approve the capability on diminishing the smearing effects by adding the P-wave and pP-wave BP images.

Despite of the fact that aftershock 2013AS1 located far from the mainshock in the central part of Kuril, it is triggered much faster than the other closed aftershock

2013AS2. The response time to the dynamic stress perturbation induced by the mainshock seems related with the backgrounds seismicity around two aftershocks or the pressure-temperature distribution inside the slab which need to be examined further by geodynamic or earthquake rupture dynamic simulations.

Chapter 4 Rupture characteristics of the 2016 Meinong earthquake