Three-dimensional Attenuation Structure beneath the Tokai Region, Central Japan Derived Using Local Earthquake Spectra

Abstract

Long term slow slip (LTSS) and non-volcanic low frequency earthquakes (LFEs) were reported in the central part of the Tokai district, central Japan. Such LTSS and LFE events are considered to take place at transition zone from stick-slip zone to stable sliding zone and to be associated with fluids on the subducting Philippine Sea plate’s surface. To clarify the spatial variation of the physical properties in this region, we estimated a three dimensional seismic attenuation structure using joint inversion method. In this study, we used 3688 spectra of 140 earthquakes which were observed by both temporary stations conducted from April to August in 2008 and permanent stations. Frequency band was divided equally among 24 between 0.78125 and 18.75 Hz and equally among 8 between 18.75 and 31.25 Hz. We gave Q blocks by dividing study area into 7 in the N-S direction between 137E and 138.5E degree, into 6 blocks in E-W direction between 34.5N and 35.7N, and 6 depth layers. We estimated frequency independent Q value of each block. In Q zone located along the Median tectonic line which divides the southwestern Japan into two parts; a old geologic belt and a new accretionary belt. In the lower crust of the land plate at the depths of 17 to 25km, a very high Q zone (about 2000) exists just above the region where large slip rate was observed in LTSS between 2001 and 2005. Since very few earthquakes occur in this high Q zone, that portion might consist of harder rocks than surroundings. On the contrary, the region just beneath the large slip zone has lower Q than surrounding area. Comparing our results with the seismic velocity structure derived from travel time tomography, we found the high Q zone approximately coincides with relatively high velocity zone, and lower Q zone corresponds to the relatively low velocity and high Vp/Vs region. As mentioned in previous studies, low Q zone with low velocity and high Vp/Vs is interpreted as the zone which involves high-pressure fluid. Probably the high Q zone above the large slip zone works as a cap rock and prevents the fluid moving toward the shallow part, then the fluid pressure becomes high and it affects the occurrence of slow slip in this region. Slow earthquakes called episodic tremor and slip (ETS) propagate over 100 kilometers at low velocities, ˜10 kilometers per day, along several plate interfaces. These low velocities differentiate slow earthquakes from ordinary earthquakes, and thus understanding their propagation processes is fundamental to understand the diversity and universality of earthquake processes. Com-prehensive modeling and previously-unreported correlations of migration patterns with energetics of tremor observed in Japan show that rheological fault heterogeneity essentially governs ETS propagations. The fault has persistent small-scale segmenta-tion, where the propagations always energetically started in brittle sections and decelerated in the ductile sections; spontaneous rupture calculations constrain the ductility to that caused by the Newtonian plastic flow or dilatant strengthening, but not by large-scale fluid flows. Activities of shallow very low frequency earthquakes (VLFEs) have been reported around the trench axis of the Nankai subduction zone and the off-Tokachi region (e.g., Obara and Ito, 2005; Asano et al., 2008). In NIED, epicenters of seismic sources including VLFEs are routinely located by an array analysis technique using Hi-net high-sensitivity accelerometers (Asano et al., 2008). Some of the epicenters are located off the Pacific coast of Tohoku, though the number in this region is much smaller than that in the off-Tokachi region. However, these events are not fully examined, as most of these sources have been thought to arise from earthquakes and microseisms. In addition, a slow slip event has been recently reported in the off-Sanriku region (Ito et al., 2010). In this study, we aim to detect VLFEs off Sanriku with a waveform correlation analysis. The array analysis with Hi-net high-sensitivity accelerometers detected a seismic source off Sanriku at 22h, 10 March, 2011. We estimated a CMT solution of this event using F-net broadband seismometers and Hi-net high-sensitivity accelerometers (Ito and Obara, 2006). The result shows a reverse fault mechanism located at the depth of 18 km with Mw3.5. This event is not clear at the frequency of several Hz, and dominant at 0.05-0.1 Hz, though a typical dominant frequency of regular earthquakes with similar magnitude and close hypocenter is several Hz. Therefore, this event is considered as a VLFE. Other VLFEs off Sanriku are detected by a waveform correlation analysis adopted in Asano et al. (2010). Averaged cross correlation values are calculated using broadband seismograms at six F-net stations which are bandpass-filtered between 0.02 to 0.1 Hz. The VLFE at 22h, 10 March 2011 are adopted as a template event in this correlation analysis. Epicenters of similar VLFEs are searched within the range of one degree both in the longitudinal and latitudinal direction. If the averaged cross correlation value is over 0.3, we manually check the waveforms and select the events which are not attributed to near- or far-field earthquakes and microseisms. Applying this technique to the period between 2005 to 14:46, 11 March 2011, we detected two VLFEs on 12 December 2007, one VLFE on 5 July 2009, and another VLFE on 10 March 2011. Activities of VLFEs off Sanriku are much lower than that off Tokachi, where accumulated counts of VLFEs are about six thousand (Asano, 2011). These VLFEs are located at the north of the aftershock area of the M7.3 off-Sanriku earthquake on 9 March 2011. In addition, the activity of regular earthquakes is low in this region. It is not revealed whether the VLFEs occur on the plate interface or within the overriding plate in this region. If we assume that the VLFEs are slip of plate interface, our result implies that frictional property shows stable sliding at this region in a usual state. Shibazaki et al. (2011) numerically reproduced large and great earthquakes which recur at the intervals of one hundred and several hundred years, respectively. In their model, it is assumed that the surrounding region of asperities shows a velocity-strengthening behavior at low and intermediate slip velocity, and strong velocity-weakening at high slip velocity. The detected VLFEs are located at the surrounding region of the asperity of the M 7.3 off-Sanriku earthquake on 9 March 2011, and at the region with large slip of the 2011 off the Pacific coast of Tohoku earthquake. Perhaps, the frictional property assumed in Shibazaki et al. (2011) may be actually important in the seismic cycles of the Tohoku region. A further observation at ocean bottom in this region would reveal the detailed activity of VLFEs. During the 2011 Tohoku-oki earthquake, large slips occurred in the region near the trench off Miyagi (e.g. Fujii et al. 2011). Hasegawa et al. (2011) showed that during the earthquake, the background deviatoric stress was completely released. This result suggests that the frictional strength decreased considerably. Recent studies on the fault rheology show that a considerable weakening can occur at a high slip velocity because of thermal pressurization or thermal weakening processes (Tanikawa and Shimamoto, 2009; Di Toro et al., 2011; Tsutsumi et al., 2011). Noda and Lapusta (2010) performed 3D simulations of earthquake sequences with evolving temperature and pore pressure resulting from shear heating, and they found that regions of more efficient thermal pressurization produce relatively large slips, resulting in large events with long interseismic periods. Mitsui et al. (2012) developed a 2D quasi-dynamic earthquake cycle model of the Tohoku-oki earthquake by considering thermal pressurization. The present study develops a 3D quasi-dynamic earthquake cycle model of the Tohoku-oki earthquake by considering thermal pressurization. We use a spectral solver for 1D diffusion problem developed by Noda and Lapusta (2010) to efficiently calculate the temperature and pore pressure evolution on a fault plane. We set several asperities in the regions off Miyagi, off Fukushima, and off Ibaraki, and set long asperities near the trench. We set the frictional properties of velocity weakening in the asperities; however, we set velocity strengthening outside of the asperities. Further, we set a low value for hydraulic diffusivity in the shallower part of the plate interface off Miyagi. The preliminary results show that M7.5 class earthquakes occur at the zone with relatively large hydraulic diffusivity. When rupture occurs around the low hydraulic diffusivity zone, significant thermal pressurization occurs and results in large and fast slips. This rupture propagates to the surrounding region and to the asperities of M7.5 earthquakes, because thermal pressurization occurs as a result of large slip even in the region of large hydraulic diffusivity. In order to forecast the occurrence of large events in the Earth’s crust, we need to understand their preparation process. Although some precursory phenomena have been proposed as preparation process for large events, most of their mechanical background is not clear. To understand the mechanical processes before large scale events, we examine numerical experiments in which multi-scale events spontaneously occur using discrete element method. The results show that before the occurrence of large events, the deviation of the direction of principal stress axes becomes small in a surrounding area of the large events. This represents a kind of homogenization of the stress field before a large event. After the large event, the stress distribution becomes scattered where only small events can occur. Rupture propagation on a fault plane forms microcracks outside the slip zone. Formation of microcracks consumes the energy for the rupture propagation. Off-fault microcracks are thus important and ha