Three-dimensional distributions of S wave reflectors in the source region of the 2000 Western Tottori Earthquake
Abstract
[1] We conducted a reflection analysis of the aftershock waveforms in the source region of the 2000 Western Tottori Earthquake. We obtained three-dimensional distributions of S wave reflectors in the source region by using ∼9000 waveform traces. In the cross-section projected along the mainshock fault plane, we found three major reflection zones at depths of 15–25 km, 30–40 km and 50–60 km. The first and second are thought to correspond to the Conrad and the Moho discontinuities. The depth of these reflective zones seems to change along the fault direction. Also, cross-sections perpendicular to the fault plane show that the strengths of the reflected waves are different between the two sides of the fault plane. This result suggests that the mainshock fault plane is located at the boundary of medium properties and its near-vertical downward extension in the lower crust.
1. Introduction
[2] On October 6, 2000, the Mw 6.6 Western Tottori Earthquake [Fukuyama et al., 2001] occurred in the Chugoku Region of the western part of Japan. 100 people were injured in this earthquake and hundreds of buildings collapsed. Lineaments, fracture zones with fault gouge related to shallow fault ruptures were found on the surface above the source region [Inoue et al., 2002; Fusejima et al., 2003]. However, an active fault had not been identified before the earthquake. For such cases, knowing the characteristic structure in the source region may be useful for evaluating the possibility of future large earthquakes. Also, it is important for understanding the earthquake generating process to elucidate the structures of the crust and the uppermost mantle in the source region from the viewpoint of both local heterogeneities and stress accumulation near the source.
[3] Several past studies have suggested the downward extension of faults into the lower crust. For example, Parsons [1998] and Parsons and Hart [1999] found reflection phases from the impedance contrast dipping with high angle in the lower crust, and they interpreted the reflection planes to be the Hayward fault and the San Andreas Fault in the north California. Zhu [2000] also suggested that the San Andreas Fault extends through the whole crust by showing, from receiver function analysis, that the Moho is disrupted under the Eastern California Shear Zone. On the other hand, Nakamura et al. [2002] and Imanishi et al. [2002] found that the Nagamachi-Rifu Fault behaves like a nearly horizontal detachment fault in the crust, at depths of 10–14 km. It is also important for understanding the mechanisms of earthquake generating and stress accumulating process, to confirm the downward extension of the fault for other crustal earthquakes.
[4] Aftershock observations using a dense array of stations were carried out cooperatively by universities in Japan [Group for the Dense Aftershock Observations of the 2000 Western Tottori Earthquake, 2001] after the mainshock. The waveform data from these observations enable us to investigate the detailed heterogeneities in the crust and the upper mantle of this earthquake source region. We estimated the distribution and the strength of the reflectors by using aftershock seismograms. Nakagawa et al. [2005] analyzed the waveform data recorded by a 12-km-long array with 50 m spacing above the hypocenter and estimated the reflection structure in the upper crust. In this study, we used about 60 stations distributed in and around the source region for obtaining the reflection structure to the depth of the uppermost mantle, both along and perpendicular to the fault plane. The results of this study will show the characteristic heterogeneous structure beneath the source region of the 2000 Western Tottori Earthquake.
2. Data
[5] The dense aftershock observations were conducted from 15 October to 6 December 2000, following the 2000 Western Tottori Earthquake. 57 temporary seismic stations were installed, as shown by the solid squares in Figure 1. The stations were placed with intervals of about 4 km covering the source area and aftershock region of this earthquake. Waveforms were recorded continuously with 0.01 s sampling interval and 16-bit resolution. The natural periods of the seismometers were 0.5–4.5 s (except for 0.22 s at two stations and 20 s at two stations). More than 1000 hypocenters with the magnitude ranging from 0.7 to 3.6 were determined using these temporary stations together with the permanent network from 15 to 25 October 2000. The hypocenters were very accurately determined with relative errors calculated to be within 50 m in horizontal and 100 m in vertical directions [Group for the Dense Aftershock Observations of the 2000 Western Tottori Earthquake, 2001; Shibutani et al., 2005].
[6] In this study, we used the waveform data from 741 earthquakes (M 1.4–3.6), shown in Figure 1, recorded at the temporary stations in order to estimate the distributions of the reflectors. 8735 waveform traces with good signal to noise ratios were analyzed.
3. Analyses
[7] In the seismic reflection survey, we use artificial sources and stations aligned on the ground. On the other hand, Inamori et al. [1992] showed the normal moveout (NMO) correction analysis that uses waveforms from natural earthquakes, taking the source depths into consideration. They applied this method to the aftershocks of the 1984 Western Nagano Prefecture Earthquake and detected two reflection planes. We applied this NMO correction method to the seismogram stacks of aftershocks of the 2000 Western Tottori Earthquake. In the NMO correction, we assume that coda waves after the direct S waves are composed of S to S wave reflections at a horizontal plane, and we convert the travel times of coda waves to the depths where they should reflect. After NMO correction, the reflected phases from the same horizontal plane are aligned as coherent horizontal bands in the record section. In the present study, we will estimate a three-dimensional distribution of reflectors by stacking the energies of reflected waves. This enables us to estimate the crustal heterogeneity in more detail.
[8] We assumed the velocity structure that is used in the routine hypocenter determination at the Tottori Observatory, Kyoto University, i.e., 3.2 km/s, 3.5 km/s, 3.8 km/s and 4.6 km/s at depths of 0–3 km, 3–16 km, 16–32 km, and >32 km, for S wave, respectively. This velocity structure was determined from explosion experiments along the Kurayoshi-Hanabusa Line [Yoshii et al., 1974]. Although the experimental line did not exactly include our study area, we think this velocity model is reasonable, because the Conrad depth of this velocity model is the same as in the 1-D model derived from a Joint Hypocenter Determination (JHD) study [Shibutani et al., 2005] and the differences in S wave velocity between these two models in the upper crust is within 0.15 km/s.
[9] In the procedures of NMO correction, we also corrected the amplitudes of the seismograms as follows. First, the two horizontal components were converted to the transverse direction to detect S waves. Therefore, P to S converted reflections are thought to have little effect if the reflection planes are nearly horizontal. A 5 to 25 Hz bandpass filter was applied so that we could examine the features over a broad frequency range. We corrected the geometrical spreading effect by multiplying by the reflection path distance. For the correction of the anelastic effect, we used a Q value of 780 for S waves so that the amplitudes are nearly constant after the geometrical spreading and anelastic attenuation corrections. We also applied a coda normalization method [Aki, 1980] to compare the amplitudes among different waveform traces: we normalized every waveform trace by its average amplitude at an appropriate time window in enough later part of coda. We chose the time window at 21–23 s after the event origin time, because of no significant later phases in it for a large number of seismograms inspected. In order to avoid the large amplitudes of the direct S waves, we used the data 1.5 s after the direct S waves.
[10] We evaluated the amplitudes of these wave traces in order to obtain three-dimensional reflector distributions. We cannot stack the amplitudes of the wave traces because of their incoherency due to much larger station spacing compared with the seismic wavelengths analyzed. Therefore we stacked the squared amplitudes as follows. We set X, Y, Z axes as fault-parallel, fault-perpendicular and vertical directions, respectively, with the origin at the mainshock epicenter, as shown in Figure 1. The analysis area includes the source region, with the size of 35 km in X, 36 km in Y, and 60 km in Z directions, respectively. This area was divided into 27,300 blocks with an interval of 1 km along the X-axis, 1–6 km along the Y-axis, and 1 km along the Z-axis. The block division is not uniform in the Y direction as shown in Figure 1. This is because higher resolution is expected near Y = 0 with most earthquakes located near Y = 0. For each block thus assigned, we stacked and averaged the squared amplitudes of the reflected waves to estimate a three-dimensional distribution of reflectors. We defined these averaged amplitudes as the “reflection strength”.
4. Results
[11] The reflection strengths projected onto vertical planes parallel and perpendicular to the fault strike are shown in Figure 2. In all the figures, contour lines show the hit count of 10, which is the number of source-station pairs used to obtain the reflection strength for each block. We assumed that averaged values of reflection strengths are reliable for hit counts greater than 10, because reflection strengths show rather stable images within contours of hit count 10. The reflection strengths are the same at the depth of about 40 km because coda normalization was applied.
[12] For the fault-parallel planes, shown in Figure 2b, we detected three major reflection zones at depths of 15–25 km, 30–40 km, and 50–60 km, as denoted by the pairs of brackets. We call these three reflection zones, Zone A, B and C, respectively.
[13] For the fault-perpendicular planes, the results show higher reflection strength in the northeastern side of the aftershock distribution, at depths from 10 to 25 km, as shown by a thick broken line in Figure 2e. The reflection strengths show a sharp contrast between the two sides of the fault.
[14] Hypocenters of aftershocks used in this study are located near the mainshock fault (i.e., Y = 0), but more events seem to be distributed on the northeastern side of the fault, as seen in Figures 1 and 2. The stations are distributed rather widely on both sides of the fault, and therefore we consider that the remarkable contrast of reflection strength across the mainshock fault plane is not affected by the asymmetric distribution of aftershocks.
[15] We assumed Q = 780 in correcting anelastic attenuation of coda waves. If we assume Q = 500, the amplitudes of coda part corresponding to the reflection at 55 km depth should be larger with a factor of 1.75. However, the amplitudes get larger gradually as increasing lapse times of coda part, and therefore the assumption of Q value does not cause serious effect in the estimate of reflection strengths.
5. Discussion
5.1. Depth Change of the Reflective Zones Along the Fault
[16] In this study, we found three major reflection Zones A, B, and C at depths of 15–25 km, 30–40 km, and 50–60 km, respectively. If we assume the error in the velocity structure used in the analysis to be ∼0.2 km/s, the estimated reflection strength has an error in its depth, for example, approximately 1.8 km at the Moho. Also, if we assume the location and the origin time errors of aftershocks used in the analysis to be less than 2 km in depth and less than 0.5 s, respectively, the estimated reflection strength has an error within ∼3 km in depth.
[17] Shibutani et al. [2005] obtained a 1-D velocity model from a JHD analysis with the Conrad discontinuity at a depth of 16 km. Yoshii et al. [1974] modeled the Moho at a depth of 32 km beneath Kurayoshi, which is located about 40 km east of the source of the 2000 Western Tottori Earthquake. Yamauchi et al. [2003] showed that the Conrad and the Moho discontinuities lie at depths of 15–20 km and 30–35 km, respectively, in this region from a receiver function analysis using 1.6 Hz low-pass-filtered waveform data. Comparing our results with these previous studies in this region, Zone A and B likely corresponds to the Conrad and the Moho discontinuities, respectively. Zones A, B, and C estimated in our analysis have a thickness of ∼5 km, as shown in Figure 2b. This is due to rather long duration (∼1–1.5 s) of reflection phases, which may be partly caused by a site effect of stations.
[18] We assumed a horizontal reflection plane in the NMO correction. Also the correction for the dipping reflection plane was not made, and therefore the shape of reflection planes shown in Figure 2b may not be exactly true. However, we consider that the reflection Zones A, B, and C seem to have variations in their depth along the fault.
5.2. Downward Extension of the Fault
[19] As we mentioned above, reflection strength shows quite a different pattern on either side of the vertical hypocentral distribution. The mainshock fault is thought to be a vertical strike slip fault, from the hypocentral distribution and the focal mechanisms [Shibutani et al., 2005]. The aftershock distributions extend to a depth of ∼13 km but the contrast of the reflection strength suggests that this fault has vertical downward extension deeper than the seismogenic zone, as shown in Figure 2e. Similar downward extension of the fault, into the lower crust, was also shown for the San Andreas Fault system [e.g., Parsons and Hart, 1999; Zhu, 2000].
5.3. Relation to the Philippine Sea Plate
[20] Zone C is the northwest-dipping reflection zone which is found beneath the entire source region, as shown in Figures 2b and 2c. This zone is more sensitive to lower frequency waves (5–10 Hz) than higher frequencies (10–25 Hz), which means that it has larger scale (>10 km) heterogeneities. In southwest Japan, the Philippines Sea Plate (PHP) is subducting beneath the Eurasian Plate. The results of several studies showed the possibility that the PHP reaches beneath the source region of the 2000 Western Tottori Earthquake. Yamane et al. [2000] found a high velocity region from travel time tomography at depths of 70–90 km beneath the source region of this earthquake. Shiomi et al. [2004] traced the PHP from receiver function analyses to the depth of 60 km beneath a region ∼30 km south of this study area. Ueno [2006] showed the deeper extension of the oceanic Moho discontinuity at depths of 60–80 km beneath the source region of the 2000 Western Tottori Earthquake, from receiver function analyses using a dense station network. Doi and Nishigami [2007] found north-dipping reflection zones at depths of 50–70 km along the Japan Sea coastline. From these studies, we consider that Zone C is related to the PHP, i.e., the plate surface or mantle wedges are the candidates of Zone C.
5.4. Reflection Strength in the Source Region of DLF Events
[21] It is reported that deep low frequency (DLF) events occur beneath the source region of this earthquake sequences [Ohmi et al., 2002]. As shown in Figure 2, the reflection strength is relatively high around the source region of the DLF events, which is represented by an ellipse. Ohmi and Obara [2002] showed the properties of the DLF events: they have predominant frequencies at 2–4 Hz and their amplitudes are dominant in S waves. Their durations are one minute or longer and, in some case, several minutes. They also showed that a single force model is more preferable than a double couple model from the amplitude ratios of P and S waves. DLF events are often observed near the volcanoes [Hasegawa and Yamamoto, 1994], and it is thought that they may be related to the existence of fluids.
6. Conclusions
[22] We estimated a detailed three-dimensional distribution of S wave reflectors in the source region of the 2000 Western Tottori Earthquake, by analyzing ∼9000 waveforms obtained in the dense aftershock observations. We found the following:
[23] 1. There are three reflection zones beneath the source region, which are located at depths of 15–25 km, 30–40 km and 50–60 km. The first and the second reflectors are thought to correspond to the Conrad and the Moho discontinuities, respectively.
[24] 2. The depth of the reflective zones seems to change along the fault strike direction.
[25] 3. The reflection strength shows a remarkable contrast between the two sides of the fault plane at depths of 10–25 km. This suggests that the fault, related to the Western Tottori Earthquake, has vertical downward extension below the seismogenic depth.
Acknowledgments
[26] We used the waveform data of the Group for the Dense Aftershock Observations of the 2000 Western Tottori Earthquake. Comments by two anonymous reviewers and James Mori were helpful in improving the manuscript. We used the Generic Mapping Tools [Wessel and Smith, 1991] for drawing all the figures.
