- Email: [email protected]
Imaging the source area of the 1995 southern Hyogo (Kobe) earthquake (M7.3) using double-difference tomography
Earth and Planetary Science Letters 253 (2007) 143 – 150 www.elsevier.com/locate/epsl
Imaging the source area of the 1995 southern Hyogo (Kobe) earthquake (M7.3) using double-difference tomography T. Okada a,⁎, A. Hasegawa a , J. Suganomata a , D. Zhao a,b , H. Zhang a,c , C. Thurber a,c a
Research Center for Prediction of Earthquakes and Volcanic Eruptions, Tohoku University, Sendai 980-8578, Japan b Geodynamics Research Center, Ehime University, Japan c Department of Geology and Geophysics, University of Wisconsin—Madison, USA Received 14 April 2006; received in revised form 10 October 2006; accepted 11 October 2006 Available online 22 November 2006 Editor: Scott King
Abstract To understand the generation process of inland earthquake, we determined the seismic velocity structure in and around the source area of the 1995 southern Hyogo (Kobe) earthquake (M7.3) in SW Japan. We adopted the double-difference (DD) tomography method [Zhang, H. and C. Thurber. Double-Difference Tomography: the method and its application to the Hayward Fault,California. Bull Seism Soc Am 93 (2003) 1875–1889.]. We inverted arrival times recorded by a dense temporary seismic network for aftershocks and seismic networks routinely operated by Japanese Universities. Obtained results are summarized as follows: (1) Low-velocity zones of a few kilometers' width are distributed along the fault or along the aftershock alignment, suggesting that the fault of the 1995 earthquake is located primarily in a low-velocity zone. (2) Amount of velocity decrease within this low-velocity zone varies along the strike of the fault. Most of large slip areas (asperities) seem to correspond to higher velocity areas relative to the surroundings on the fault, rather than to lower velocity areas. © 2006 Elsevier B.V. All rights reserved. Keywords: fault; coseismic slip distribution; aftershock; asperity; tomography; the 1995 southern Hyogo (Kobe) earthquake
1. Introduction The 1995 southern Hyogo earthquake occurred on 17 January 1995 in the upper crust of the southern part of Hyogo prefecture, its fault extending from Kobe city to the northern part of Awaji Island. The focal mechanism of this earthquake was of the right-lateral strike–
⁎ Corresponding author. E-mail address: [email protected] (T. Okada). 0012-821X/$ - see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.epsl.2006.10.022
slip fault type [1]. Aftershock distributions [2] and surface ruptures in the northern part of Awaji Island [3] showed that this shallow inland earthquake was generated by reactivation of the Rokko fault system (the Suma and Suwayama faults) and the Nojima fault (see Fig. 1). On Awaji Island, in the southern part of the focal area of the southern Hyogo earthquake, surface ruptures were found on the Nojima fault. [3]. Maximum offset was 1.5 m. Source process studies e.g. [4–9]showed that a large quantity of slip was distributed along the surface rupture of the fault. In the northern part, a small amount of slip was distributed in the shallower part of the fault
144
T. Okada et al. / Earth and Planetary Science Letters 253 (2007) 143–150
Fig. 1. Locations of major active faults. Small dots show events.
and a large amount of slip occurred at depths near the mainshock hypocenter which is located at the deepest and central part of the fault plane. The nature of the large slip areas (asperities) of the fault is uncertain. It has become clear that asperities are common in several large or moderate-sized interplate recurrent earthquake pairs along the plate boundary in NE Japan [e.g. 10,11]. These observations possibly suggest that asperities are persistent features which reflect certain physical properties on the fault plane. Zhao et al. [12,13] have obtained three-dimensional seismic velocity structure in and around the 1995 southern Hyogo earthquake. They found low velocity zones in and around the hypocenter and the fault. However, the relation between the lateral coseismic slip distribution and the seismic velocity structure has not been discussed in their studies. In this study, we performed seismic tomography by the double-difference tomography method [14] to obtain the detailed seismic velocity structure in and around the source area of the 1995 southern Hyogo (Kobe) M7.3 earthquake in Japan and discuss its relation with the rupture process. 2. Methods and data For our analysis, we adopted the double-difference (DD) tomography method [14]. This method has the advantage of obtaining the seismic velocity structure at high spatial resolution for areas where hypocenters are densely distributed such as aftershock areas. For the DD tomography method, we used not only the absolute travel times, as in conventional tomography, but also the differential travel times between closely located events
in a manner comparable to DD location [15]. When differential data are used, the DD tomography method cancels the equations corresponding to the common parts of the ray paths, so we can estimate the velocity structure only around the hypocenter. The velocity model is parameterized as a set of grids, and the velocity at any location is calculated by linear interpolation between the grids. We calculated the velocity models with a grid of a few km in and around the focal area. Ray tracing is done by the pseudo-bending method [16]. After the southern Hyogo (Kobe) earthquake, 27 temporary seismic stations were deployed around the focal area. The data sets (Fig. 2) used in this study were derived mainly from NET-HYOGO (Urgent Joint Observation Network for the 1995 southern Hyogo earthquake) [2]. The NET-HYOGO deployed threecomponent short-period seismometers and telephone telemetry system. The inversion uses 47770 P-wave and 38616 S-wave arrival times, which are picked manually from 2390 earthquakes (M = N 1.5) recorded by this dense temporary seismic network for aftershocks and by the seismic networks routinely operated by Kyoto University, the University of Tokyo and Kochi University. Total number of the station used in this study is 113. Most of the events used are aftershocks of the Kobe earthquake. The analysis was carried out as follows. Hypocenters, one-dimensional velocity structure, and station correction values were first estimated simultaneously by the program VELEST [17]. This process yields a more suitable initial hypocenter locations and initial velocity structure. The three-dimensional velocity structure was then estimated by double difference tomography and used to refine the hypocenter locations. The grid points are set at intervals of 5–10 km NE–SW along the strike of the fault, 3–5 km NW–SE, and 3– 5 km in depth. Initial hypocenter locations are from [13]. Initial 1D velocity structure is also the same as that used in [13]. 3. Results Overall distribution of the obtained velocity structure is similar with those by [13]; the low-velocity zone detected along and around the mainshock fault, although detailed distribution of the low-velocity zones are different. We can see many aftershock clusters clearly because we also relocated the aftershocks by the doubledifference algorithm [15]. The obtained velocity structure shows complex patterns (Fig. 3); there are prominent low-velocity zones in and around much of the main shock fault. In the case of the 1995 southern Hyogo earthquake,
T. Okada et al. / Earth and Planetary Science Letters 253 (2007) 143–150
145
Fig. 2. Locations of observed stations (triangles), events (dots), and grids (crosses) used in this study. Red thin lines show the location of active faults.
aftershocks occurred over a 70 km length along the faults at depths between 5 and 15 km. Low-velocity (Vp and Vs) zones are distributed along the faults or along the aftershock alignment. In the Kobe area (the northeastern part of the fault), aftershocks are almost all aligned beneath the faults (Suma–Suwayama–Gosukebashi). In the northern part of the Awaji area (the southwestern part of the fault), most of aftershocks seem to be aligned beneath the Kariya fault and farther in the southwest are shifted to the west and aligned beneath the Nojima fault. Low-velocity zones seem to be aligned along some portions of aftershock alignment. Aftershocks generally occur on and around the mainshock fault plane. Therefore, the results presently obtained indicate that some portions of the fault plane of the present earthquake are located within low-velocity zones.
To check the resolution of the present inversions, we conducted a checkerboard resolution test (CBT) [18,19]. To make a checkerboard, we assigned positive and negative velocity anomalies of 5% to all grid nodes. Fig. 4 shows the results obtained by inverting synthetic data calculated from the original checkerboard model. We added Gaussian noise to the synthetic arrival times with a standard deviation of 0.1 s for P-waves and 0.2 s for S-waves. The checkerboard pattern and the absolute value of the velocity anomalies were well recovered where stations and earthquakes are densely distributed, although not where stations and earthquakes are less densely distributed and at the border of the study area. The CBT results indicated good resolution for both P- and S-waves in areas of width of at least about 20 km along the aftershock alignment, where high
146
T. Okada et al. / Earth and Planetary Science Letters 253 (2007) 143–150
Fig. 3. Horizontal section of (A) P-wave and (B) S-wave velocity perturbation at a depth of 5 km and (C) P-wave and (D) S-wave at a depth of 11 km. Note that depth interval is 3 km. Perturbation from the average velocity at a depth is shown by the color scale at the bottom. Deeper color surrounded by dashed white line means greater DWS value. Crosses denote epicenters of aftershocks at this depth.
Derivative Weighted Sum (DWS; [20]) values were distributed. The low-velocity zone detected along the mainshock fault changes in the amount of velocity decrease along its strike. For example, relatively high velocity areas are distributed primarily in the southwestern shallow part (X = − 30 km, Depth 0–5 km) and in the northeastern deeper part (X = − 10 to − 20 km, Depth 10–15 km). We compared the P-wave velocity distribution along the fault (Fig. 5(A)) with the slip distribution obtained by seismic and geodetic inversions [8]. Note that the
location of cross section in Fig. 5 is slightly different from Plate 1 in [13] and P-wave velocity distributions by this study and by [13] seem to be similar with each other along the same cross sections. Yoshida et al.'s results showed that there are two major asperities along the fault: the primary one is located to the northeast of the hypocenter in the deeper part and the secondary one is located to the southwest of the hypocenter in the shallower part, which corresponds with the Nojima fault. The comparison between P-wave velocity distribution and slip distribution shows that most of these primary
T. Okada et al. / Earth and Planetary Science Letters 253 (2007) 143–150
147
Fig. 4. Result of a checkerboard test for (A) P-wave and (B) S-wave at a depth of 5 km, and (C) P-wave and (D) S-wave at a depth of 11 km. Small dots denote the aftershocks. Triangles denote the stations.
and secondary large slip areas (asperities) tend to correspond to higher-velocity areas relative to the surroundings on the fault, rather than to marked low-velocity areas. Aftershocks are distributed within the areas of moderate velocity and few aftershocks are distributed within the areas of substantially low velocity. Fig. 5(B) shows the S-wave velocity distribution along the fault. The S-wave velocity distribution also shows that large slip areas (asperities) tend to avoid marked low-velocity areas although this tendency is less clear than for the Pwave distribution. It appears dVp is inversely correlated with dVs in some areas (e.g. in the northern part of the Awaji Island in Fig 3(C), (D)). This inverse correlation can be
interpret to be real to some extent from the results of the CBT (Fig. 4), and reflects the Vp/Vs distribution. Fig. 5 (C) shows variations in Vp/Vs along the fault plane. Low-Vp/Vs area and high-Vp/Vs area almost correspond to low-Vp area and low-Vs area, respectively. Most of the large slip areas (asperities) seem to avoid marked low-Vp/Vs (b1.65), low-Vp and/or high-Vp/Vs (N1.83), low-Vs areas. 4. Discussion and conclusions We could find the low-velocity zone along and around the mainshock fault using double-difference tomography, although the width of the low-velocity zones is
148
T. Okada et al. / Earth and Planetary Science Letters 253 (2007) 143–150
Fig. 5. Vertical cross-section of (A) P-wave velocity perturbation, (B) Vs perturbation and (C) Vp/Vs distribution along lines A–A′ shown in the bottom figure. Perturbations from the average velocity at each depth are shown. White contour lines show the slip distribution in an interval of 0.4 m [8]. Bold contour lines show the slip of 1.6 m. White stars denote hypocenters of the main shock. Deeper color means greater DWS value. Crosses denote aftershocks. Note that the average value of P- and S-wave velocity at each depth is re-calculated for these cross sections.
T. Okada et al. / Earth and Planetary Science Letters 253 (2007) 143–150
as large as a couple of kilometers. This is due to the resolution of the present inversions with a grid interval of 3 km and the low-velocity zones probably have a smaller width. [13] also detected a low-Vp and Vs zone along the fault zone of the 1995 southern Hyogo (Kobe) earthquake. In their case, the width of the low-velocity zone was as large as 10 km, which is again due to the resolution of their inversions. In other areas, for example, [21,22] imaged, by means of seismic tomography, a lowVp and high-Vp/Vs zone, possibly corresponding to inclusions of water, down to a depth of about 3 km within the fault zone of the San Andreas Fault. These low-Vp and Vs zones probably indicate a highly fractured damaged zone. Note that the surface lithology changes across the major faults in the focal area (cf. see Fig. 14 in [13]) but the lateral variation in the fault zone does not seem to be dominant and we could not see a specific correlation between the surface lithology and the seismic velocity structure in this study, although the depth extent of the surface lithology is unknown. We have also applied double-difference tomography to the data sets of some inland earthquakes that recently occurred in Japan. All the data sets consist of temporary aftershock recording networks densely deployed just above and around the focal area and the permanent stations routinely operated by several universities, JMA, and Hi-net, NIED, Japan. They are the 2003 northern Miyagi earthquake (M6.4) in NE Japan [23,24], the 2004 Mid Niigata earthquake in central Japan [25,26], and the 2000 western Tottori earthquake (M7.3) in SW Japan [27]. In all the cases, most of asperities seem to correspond to the areas of high velocity relative to other areas on the fault, same as in the case of the 1995 Kobe earthquake. This correspondence between high-velocity body and asperity has also observed in Parkfield, California [28,29]. Estimation of stress before the present earthquake [30] shows that the stress ratio value was large at the center of the fault near the asperities and smaller at both edges, where we obtained lower velocities. This stress distribution seems to support the correspondence between high-velocity body and asperity, if the welldeveloped areas with relatively lower velocity causes inelastic deformation to predominate and prevent the accumulation of large strain and only the less developed areas with relatively higher velocity on the fault can store large stress and release it as large coseismic slips, acting as asperities (cf. [26,31]). Further studies are necessary to verify whether this hypothesis is acceptable or not. In conclusion, we successfully obtained the detailed seismic velocity structures in and around the focal area of the 1995 southern Hyogo (Kobe) earthquake in SW
149
Japan by double difference tomography. The obtained results are summarized as follows: (1) Low-velocity zones are imaged along the fault of the present earthquake with a width of a couple of kilometers. (2) Asperities (large slip areas) correspond to areas of high seismic velocity relative to other areas along the fault. A correspondence between asperities and high-velocity areas was also observed for some other large shallow inland earthquakes (the 2000 M7.3 western Tottori, 2004 M6.8 Mid Niigata and 2003 M6.4 northern Miyagi earthquakes) that recently occurred in Japan. These observations will contribute to better understanding of the generation process of earthquake and may help to identify asperities. Acknowledgments We used data from a temporary seismological observation by the Japanese university group. We also used data from Kyoto University, the University of Tokyo, and Kochi University. We are grateful to all the team members for their efforts in obtaining the data. Discussions with Prof. K. Koketsu, Dr. F. Yamashita and Dr. K. Omura were valuable. We would like to thank Dr. Y. Iio, Dr. S. King and an anonymous reviewer for reviewing our manuscript and their valuable comments. This work was conducted as part of the 21st COE program, ‘Advanced Science and Technology Center for the Dynamic Earth’, at Tohoku University. This work was also partially supported by MEXT.KAKENHI (16740247) and JSPS. KAKENHI (15204037), Japan. References [1] H. Katao, N. Maeda, Y. Hiramatsu, Y. Iio, S. Nakao, Detailed mapping of focal mechanisms in/around the 1995 Hyogo-ken nanbu earthquake rupture zone, J. Phys. Earth 45 (1997) 105–119. [2] N. Hirata, S. Ohmi, S. Sakai, K. Katsumata, S. Matsumoto, T. Takanami, A. Yamamoto, T. Iidaka, T. Urabe, M. Sekine, T. Ooida, F. Yamazaki, H. Katao, Y. Umeda, M. Nakamura, T. Seto, T. Matsushima, H. Shimizu, Japanese University Group for the Urgent Joint Observation for the 1995 Hyogo-ken Nanbu Earthquake, Urgent joint observation of aftershocks of the 1995 Hyogo-ken Nanbu earthquake, J. Phys. Earth 44 (1996) 317–328. [3] T. Nakata, K. Yomogida, Surface fault characteristics of the 1995 Hyogoken-Nanbu earthquake, J. Nat. Disaster Sci. 16 (1995) 1–9. [4] H. Horikawa, K. Hirahara, Y. Umeda, M. Hashimoto, F. Kusano, Simultaneous inversion of geodetic and strong motion data for the source process of the Hyogo-ken Nanbu, Japan, earthquake, J. Phys. Earth 44 (1996) 455–471. [5] S. Ide, M. Takeo, Y. Yoshida, Source process of the 1995 Kobe earthquake: determination of spatio-temporal slip distribution by baysian modeling, Bull. Seismol. Soc. Am. 86 (1996) 547–566.
150
T. Okada et al. / Earth and Planetary Science Letters 253 (2007) 143–150
[6] H. Sekiguchi, K. Irikura, T. Iwata, Y. Kakehi, M. Hoshiba, Minute locating of fault planes and source process of the 1995 Hyogoken-nanbu, earthquake from the waveform inversion of strong ground motion, J. Phys. Earth 44 (1996) 473–487. [7] D. Wald, Slip history of the 1995 Kobe, Japan, earthquake determined from strong motion, teleseismic and geodetic data, J. Phys. Earth 44 (1996) 489–503. [8] S. Yoshida, K. Koketsu, B. Shibazaki, T. Sagiya, T. Kato, Y. Yoshida, Joint inversion of near-and far-field waveforms and geodetic data for the rupture process of the 1995 Kobe earthquake, J. Phys. Earth 44 (1996) 437–454. [9] K. Koketsu, T. Ohno, Y. Ikegami, Rupture process inversion using 3-D Green's functions: the 1995 Kobe earthquake revisited, Eos Trans AGU, Fall Meet Suppl, vol. 84, 2003, (Abstract S42H–Abstract S407. [10] T. Okada, T. Matsuzawa, A. Hasegawa, Comparison of source areas of M4.8+/− 0.1 repeating earthquakes off Kamaishi, NE Japan — are asperities persistent features? Earth Planet. Sci. Lett. 213 (2003) 361–374. [11] Hasegawa, A., N. Uchida, T. Igarashi, T. Matsuzawa, T. Okada, S. Miura, Y. Suwa. Asperities and quasi-static slip on the sudducting plate boundary east off Tohoku. NE Japan. SEIZE volume. Columbia Univ. Press., in press. [12] D. Zhao, H. Kanamori, H. Negishi, D. Wiens, Tomography of the source area of the 1995 Kobe earthquake: evidence for fluids at the hypocenter? Science 274 (1996) 1891–1894. [13] D. Zhao, H. Negishi, The 1995 Kobe earthquake: seismic imaging of the source zone and its implications for the rupture nucleation, J. Geophys. Res. 103 (1998) 9967–9986. [14] H. Zhang, C. Thurber, Double-difference tomography: the method and its application to the Hayward Fault,California, Bull. Seismol. Soc. Am. 93 (2003) 1875–1889. [15] F. Waldhauser, W.L. Ellsworth, A double-difference earthquake location algorithm: method and application to the Northern Hayward fault, Bull. Seismol. Soc. Am. 90 (2000) 1353–1368. [16] J. Um, C.H. Thurber, A fast algorithm for two-point seismic ray tracing, Bull. Seismol. Soc. Am. 77 (1987) 972–986. [17] E. Kissling, W. Ellsworth, D. Eberhart-Phillips, U. Kradolfer, Initial reference models in local earthquake tomography, J. Geophys. Res. 99 (1994) 19635–19646. [18] S. Grand, Tomographic inversion for shear wave velocity beneath the North-American plate, J. Geophys. Res. 92 (1987) 14065–14090. [19] E. Humphreys, R. Clayton, Adaptation of back projection tomography to seismic travel time problems, J. Geophys. Res. 93 (1988) 1073–1085. [20] C. Thurber, D. Eberhart-Phillips, Local earthquake tomography with flexible gridding, Comput. Geosci. 25 (1999) 809–818. [21] C. Thurber, S. Roecker, W. Ellsworth, Y. Chen, W. Lutter, R. Sessions, Two-dimensional seismic image of the San Andreas
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
Fault in the Northern Gablian Range, central California: evidence for fluids in the fault zone, Geophys. Res. Lett. 24 (1997) 1591–1594. C. Thurber, S. Roecker, K. Roberts, M. Gold, L. Powell, K. Rittger, Earthquake locations and three-dimensional fault zone structure along the creeping section of the San Andreas fault near Parkfield, CA: preparing for SAFOD, Geophys. Res. Lett. 30 (2003) (2002GL016004). T. Okada, A. Hasegawa, J. Suganomata, N. Umino, H. Zhang, and C. Thurber. Imaging the heterogeneous source area of the 2003 M6.4 northern Miyagi earthquake, NE Japan, by doubledifference tomography. Tectonophys. submitted for publication. T. Okada, A. Hasegawa, J. Suganomata, N. Umino, H. Zhang, C. Thurber, Imaging the fault plane of the 2003 M6.4 northern Miyagi, NE Japan, earthquake by double-difference tomography, Eos Trans AGU, West Pac Geophys Suppl, vol. 85, 2004, p. S42A21. T. Okada, N. Umino, T. Matsuzawa, J. Nakajima, N. Uchida, T. Nakayama, S. Hirahara, T. Sato, S. Hori, T. Kono, Y. Yabe, K. Ariyoshi, S. Gamage, J. Shimizu, J. Suganomata, S. Kita, S. Yui, M. Arao, S. Hondo, T. Mizukami, H. Tsushima, T. Yaginuma, A. Hasegawa, Y. Asano, H. Zhang, C. Thurber, Aftershock distribution and 3D seismic velocity structure in and around the focal area of the 2004 mid Niigata prefecture earthquake obtained by applying double-difference tomography to dense temporary seismic network data, Earth Planets Space 57 (2005) 435–440. T. Okada, T. Yaginuma, N. Umino, T. Matsuzawa, A. Hasegawa, H. Zhang, C. Thurber, Detailed imaging of the fault planes of the 2004 Niigata–Chuetsu, central Japan, earthquake sequence by double-difference tomography, Earth Planet. Sci. Lett. 244 (2006) 32–43. T. Okada, A. Hasegawa, H. Zhang, C. Thurber, Detailed seismic velocity structure around the focal area of the 2000 Western Tottori earthquake (M7.3) by double-difference tomography, 2004 Japan EPS Joint Meeting, 2004, J078-P001. A. Michael, D. Eberhart-Phillips, Relations among fault behavior, subsurface geology, and three-dimensional velocity models, Science 253 (1991) 651–654. D. Eberhart-Phillips, A. Michael, Three-dimensional velocity structure, seismicity, and fault structure in the Parkfield Region, Central California, J. Geophys. Res. 98 (1993) 15737–15758. F. Yamashita, E. Fukuyama, K. Omura, Estimation of fault strength: reconstruction of stress before the 1995 Kobe earthquake, Science 306 (2004) 261–263. A. Hasegawa, J. Nakajima, N. Umino, S. Miura, Deep structure of the northeastern Japan arc and its implications for crustal deformation and shallow seismic activity, Tectonophysics 403 (2005) 59–75.
Imaging the source area of the 1995 southern Hyogo (Kobe) earthquake (M7.3) using double-difference tomography T. Okada a,⁎, A. Hasegawa a , J. Suganomata a , D. Zhao a,b , H. Zhang a,c , C. Thurber a,c a
Research Center for Prediction of Earthquakes and Volcanic Eruptions, Tohoku University, Sendai 980-8578, Japan b Geodynamics Research Center, Ehime University, Japan c Department of Geology and Geophysics, University of Wisconsin—Madison, USA Received 14 April 2006; received in revised form 10 October 2006; accepted 11 October 2006 Available online 22 November 2006 Editor: Scott King
Abstract To understand the generation process of inland earthquake, we determined the seismic velocity structure in and around the source area of the 1995 southern Hyogo (Kobe) earthquake (M7.3) in SW Japan. We adopted the double-difference (DD) tomography method [Zhang, H. and C. Thurber. Double-Difference Tomography: the method and its application to the Hayward Fault,California. Bull Seism Soc Am 93 (2003) 1875–1889.]. We inverted arrival times recorded by a dense temporary seismic network for aftershocks and seismic networks routinely operated by Japanese Universities. Obtained results are summarized as follows: (1) Low-velocity zones of a few kilometers' width are distributed along the fault or along the aftershock alignment, suggesting that the fault of the 1995 earthquake is located primarily in a low-velocity zone. (2) Amount of velocity decrease within this low-velocity zone varies along the strike of the fault. Most of large slip areas (asperities) seem to correspond to higher velocity areas relative to the surroundings on the fault, rather than to lower velocity areas. © 2006 Elsevier B.V. All rights reserved. Keywords: fault; coseismic slip distribution; aftershock; asperity; tomography; the 1995 southern Hyogo (Kobe) earthquake
1. Introduction The 1995 southern Hyogo earthquake occurred on 17 January 1995 in the upper crust of the southern part of Hyogo prefecture, its fault extending from Kobe city to the northern part of Awaji Island. The focal mechanism of this earthquake was of the right-lateral strike–
⁎ Corresponding author. E-mail address: [email protected] (T. Okada). 0012-821X/$ - see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.epsl.2006.10.022
slip fault type [1]. Aftershock distributions [2] and surface ruptures in the northern part of Awaji Island [3] showed that this shallow inland earthquake was generated by reactivation of the Rokko fault system (the Suma and Suwayama faults) and the Nojima fault (see Fig. 1). On Awaji Island, in the southern part of the focal area of the southern Hyogo earthquake, surface ruptures were found on the Nojima fault. [3]. Maximum offset was 1.5 m. Source process studies e.g. [4–9]showed that a large quantity of slip was distributed along the surface rupture of the fault. In the northern part, a small amount of slip was distributed in the shallower part of the fault
144
T. Okada et al. / Earth and Planetary Science Letters 253 (2007) 143–150
Fig. 1. Locations of major active faults. Small dots show events.
and a large amount of slip occurred at depths near the mainshock hypocenter which is located at the deepest and central part of the fault plane. The nature of the large slip areas (asperities) of the fault is uncertain. It has become clear that asperities are common in several large or moderate-sized interplate recurrent earthquake pairs along the plate boundary in NE Japan [e.g. 10,11]. These observations possibly suggest that asperities are persistent features which reflect certain physical properties on the fault plane. Zhao et al. [12,13] have obtained three-dimensional seismic velocity structure in and around the 1995 southern Hyogo earthquake. They found low velocity zones in and around the hypocenter and the fault. However, the relation between the lateral coseismic slip distribution and the seismic velocity structure has not been discussed in their studies. In this study, we performed seismic tomography by the double-difference tomography method [14] to obtain the detailed seismic velocity structure in and around the source area of the 1995 southern Hyogo (Kobe) M7.3 earthquake in Japan and discuss its relation with the rupture process. 2. Methods and data For our analysis, we adopted the double-difference (DD) tomography method [14]. This method has the advantage of obtaining the seismic velocity structure at high spatial resolution for areas where hypocenters are densely distributed such as aftershock areas. For the DD tomography method, we used not only the absolute travel times, as in conventional tomography, but also the differential travel times between closely located events
in a manner comparable to DD location [15]. When differential data are used, the DD tomography method cancels the equations corresponding to the common parts of the ray paths, so we can estimate the velocity structure only around the hypocenter. The velocity model is parameterized as a set of grids, and the velocity at any location is calculated by linear interpolation between the grids. We calculated the velocity models with a grid of a few km in and around the focal area. Ray tracing is done by the pseudo-bending method [16]. After the southern Hyogo (Kobe) earthquake, 27 temporary seismic stations were deployed around the focal area. The data sets (Fig. 2) used in this study were derived mainly from NET-HYOGO (Urgent Joint Observation Network for the 1995 southern Hyogo earthquake) [2]. The NET-HYOGO deployed threecomponent short-period seismometers and telephone telemetry system. The inversion uses 47770 P-wave and 38616 S-wave arrival times, which are picked manually from 2390 earthquakes (M = N 1.5) recorded by this dense temporary seismic network for aftershocks and by the seismic networks routinely operated by Kyoto University, the University of Tokyo and Kochi University. Total number of the station used in this study is 113. Most of the events used are aftershocks of the Kobe earthquake. The analysis was carried out as follows. Hypocenters, one-dimensional velocity structure, and station correction values were first estimated simultaneously by the program VELEST [17]. This process yields a more suitable initial hypocenter locations and initial velocity structure. The three-dimensional velocity structure was then estimated by double difference tomography and used to refine the hypocenter locations. The grid points are set at intervals of 5–10 km NE–SW along the strike of the fault, 3–5 km NW–SE, and 3– 5 km in depth. Initial hypocenter locations are from [13]. Initial 1D velocity structure is also the same as that used in [13]. 3. Results Overall distribution of the obtained velocity structure is similar with those by [13]; the low-velocity zone detected along and around the mainshock fault, although detailed distribution of the low-velocity zones are different. We can see many aftershock clusters clearly because we also relocated the aftershocks by the doubledifference algorithm [15]. The obtained velocity structure shows complex patterns (Fig. 3); there are prominent low-velocity zones in and around much of the main shock fault. In the case of the 1995 southern Hyogo earthquake,
T. Okada et al. / Earth and Planetary Science Letters 253 (2007) 143–150
145
Fig. 2. Locations of observed stations (triangles), events (dots), and grids (crosses) used in this study. Red thin lines show the location of active faults.
aftershocks occurred over a 70 km length along the faults at depths between 5 and 15 km. Low-velocity (Vp and Vs) zones are distributed along the faults or along the aftershock alignment. In the Kobe area (the northeastern part of the fault), aftershocks are almost all aligned beneath the faults (Suma–Suwayama–Gosukebashi). In the northern part of the Awaji area (the southwestern part of the fault), most of aftershocks seem to be aligned beneath the Kariya fault and farther in the southwest are shifted to the west and aligned beneath the Nojima fault. Low-velocity zones seem to be aligned along some portions of aftershock alignment. Aftershocks generally occur on and around the mainshock fault plane. Therefore, the results presently obtained indicate that some portions of the fault plane of the present earthquake are located within low-velocity zones.
To check the resolution of the present inversions, we conducted a checkerboard resolution test (CBT) [18,19]. To make a checkerboard, we assigned positive and negative velocity anomalies of 5% to all grid nodes. Fig. 4 shows the results obtained by inverting synthetic data calculated from the original checkerboard model. We added Gaussian noise to the synthetic arrival times with a standard deviation of 0.1 s for P-waves and 0.2 s for S-waves. The checkerboard pattern and the absolute value of the velocity anomalies were well recovered where stations and earthquakes are densely distributed, although not where stations and earthquakes are less densely distributed and at the border of the study area. The CBT results indicated good resolution for both P- and S-waves in areas of width of at least about 20 km along the aftershock alignment, where high
146
T. Okada et al. / Earth and Planetary Science Letters 253 (2007) 143–150
Fig. 3. Horizontal section of (A) P-wave and (B) S-wave velocity perturbation at a depth of 5 km and (C) P-wave and (D) S-wave at a depth of 11 km. Note that depth interval is 3 km. Perturbation from the average velocity at a depth is shown by the color scale at the bottom. Deeper color surrounded by dashed white line means greater DWS value. Crosses denote epicenters of aftershocks at this depth.
Derivative Weighted Sum (DWS; [20]) values were distributed. The low-velocity zone detected along the mainshock fault changes in the amount of velocity decrease along its strike. For example, relatively high velocity areas are distributed primarily in the southwestern shallow part (X = − 30 km, Depth 0–5 km) and in the northeastern deeper part (X = − 10 to − 20 km, Depth 10–15 km). We compared the P-wave velocity distribution along the fault (Fig. 5(A)) with the slip distribution obtained by seismic and geodetic inversions [8]. Note that the
location of cross section in Fig. 5 is slightly different from Plate 1 in [13] and P-wave velocity distributions by this study and by [13] seem to be similar with each other along the same cross sections. Yoshida et al.'s results showed that there are two major asperities along the fault: the primary one is located to the northeast of the hypocenter in the deeper part and the secondary one is located to the southwest of the hypocenter in the shallower part, which corresponds with the Nojima fault. The comparison between P-wave velocity distribution and slip distribution shows that most of these primary
T. Okada et al. / Earth and Planetary Science Letters 253 (2007) 143–150
147
Fig. 4. Result of a checkerboard test for (A) P-wave and (B) S-wave at a depth of 5 km, and (C) P-wave and (D) S-wave at a depth of 11 km. Small dots denote the aftershocks. Triangles denote the stations.
and secondary large slip areas (asperities) tend to correspond to higher-velocity areas relative to the surroundings on the fault, rather than to marked low-velocity areas. Aftershocks are distributed within the areas of moderate velocity and few aftershocks are distributed within the areas of substantially low velocity. Fig. 5(B) shows the S-wave velocity distribution along the fault. The S-wave velocity distribution also shows that large slip areas (asperities) tend to avoid marked low-velocity areas although this tendency is less clear than for the Pwave distribution. It appears dVp is inversely correlated with dVs in some areas (e.g. in the northern part of the Awaji Island in Fig 3(C), (D)). This inverse correlation can be
interpret to be real to some extent from the results of the CBT (Fig. 4), and reflects the Vp/Vs distribution. Fig. 5 (C) shows variations in Vp/Vs along the fault plane. Low-Vp/Vs area and high-Vp/Vs area almost correspond to low-Vp area and low-Vs area, respectively. Most of the large slip areas (asperities) seem to avoid marked low-Vp/Vs (b1.65), low-Vp and/or high-Vp/Vs (N1.83), low-Vs areas. 4. Discussion and conclusions We could find the low-velocity zone along and around the mainshock fault using double-difference tomography, although the width of the low-velocity zones is
148
T. Okada et al. / Earth and Planetary Science Letters 253 (2007) 143–150
Fig. 5. Vertical cross-section of (A) P-wave velocity perturbation, (B) Vs perturbation and (C) Vp/Vs distribution along lines A–A′ shown in the bottom figure. Perturbations from the average velocity at each depth are shown. White contour lines show the slip distribution in an interval of 0.4 m [8]. Bold contour lines show the slip of 1.6 m. White stars denote hypocenters of the main shock. Deeper color means greater DWS value. Crosses denote aftershocks. Note that the average value of P- and S-wave velocity at each depth is re-calculated for these cross sections.
T. Okada et al. / Earth and Planetary Science Letters 253 (2007) 143–150
as large as a couple of kilometers. This is due to the resolution of the present inversions with a grid interval of 3 km and the low-velocity zones probably have a smaller width. [13] also detected a low-Vp and Vs zone along the fault zone of the 1995 southern Hyogo (Kobe) earthquake. In their case, the width of the low-velocity zone was as large as 10 km, which is again due to the resolution of their inversions. In other areas, for example, [21,22] imaged, by means of seismic tomography, a lowVp and high-Vp/Vs zone, possibly corresponding to inclusions of water, down to a depth of about 3 km within the fault zone of the San Andreas Fault. These low-Vp and Vs zones probably indicate a highly fractured damaged zone. Note that the surface lithology changes across the major faults in the focal area (cf. see Fig. 14 in [13]) but the lateral variation in the fault zone does not seem to be dominant and we could not see a specific correlation between the surface lithology and the seismic velocity structure in this study, although the depth extent of the surface lithology is unknown. We have also applied double-difference tomography to the data sets of some inland earthquakes that recently occurred in Japan. All the data sets consist of temporary aftershock recording networks densely deployed just above and around the focal area and the permanent stations routinely operated by several universities, JMA, and Hi-net, NIED, Japan. They are the 2003 northern Miyagi earthquake (M6.4) in NE Japan [23,24], the 2004 Mid Niigata earthquake in central Japan [25,26], and the 2000 western Tottori earthquake (M7.3) in SW Japan [27]. In all the cases, most of asperities seem to correspond to the areas of high velocity relative to other areas on the fault, same as in the case of the 1995 Kobe earthquake. This correspondence between high-velocity body and asperity has also observed in Parkfield, California [28,29]. Estimation of stress before the present earthquake [30] shows that the stress ratio value was large at the center of the fault near the asperities and smaller at both edges, where we obtained lower velocities. This stress distribution seems to support the correspondence between high-velocity body and asperity, if the welldeveloped areas with relatively lower velocity causes inelastic deformation to predominate and prevent the accumulation of large strain and only the less developed areas with relatively higher velocity on the fault can store large stress and release it as large coseismic slips, acting as asperities (cf. [26,31]). Further studies are necessary to verify whether this hypothesis is acceptable or not. In conclusion, we successfully obtained the detailed seismic velocity structures in and around the focal area of the 1995 southern Hyogo (Kobe) earthquake in SW
149
Japan by double difference tomography. The obtained results are summarized as follows: (1) Low-velocity zones are imaged along the fault of the present earthquake with a width of a couple of kilometers. (2) Asperities (large slip areas) correspond to areas of high seismic velocity relative to other areas along the fault. A correspondence between asperities and high-velocity areas was also observed for some other large shallow inland earthquakes (the 2000 M7.3 western Tottori, 2004 M6.8 Mid Niigata and 2003 M6.4 northern Miyagi earthquakes) that recently occurred in Japan. These observations will contribute to better understanding of the generation process of earthquake and may help to identify asperities. Acknowledgments We used data from a temporary seismological observation by the Japanese university group. We also used data from Kyoto University, the University of Tokyo, and Kochi University. We are grateful to all the team members for their efforts in obtaining the data. Discussions with Prof. K. Koketsu, Dr. F. Yamashita and Dr. K. Omura were valuable. We would like to thank Dr. Y. Iio, Dr. S. King and an anonymous reviewer for reviewing our manuscript and their valuable comments. This work was conducted as part of the 21st COE program, ‘Advanced Science and Technology Center for the Dynamic Earth’, at Tohoku University. This work was also partially supported by MEXT.KAKENHI (16740247) and JSPS. KAKENHI (15204037), Japan. References [1] H. Katao, N. Maeda, Y. Hiramatsu, Y. Iio, S. Nakao, Detailed mapping of focal mechanisms in/around the 1995 Hyogo-ken nanbu earthquake rupture zone, J. Phys. Earth 45 (1997) 105–119. [2] N. Hirata, S. Ohmi, S. Sakai, K. Katsumata, S. Matsumoto, T. Takanami, A. Yamamoto, T. Iidaka, T. Urabe, M. Sekine, T. Ooida, F. Yamazaki, H. Katao, Y. Umeda, M. Nakamura, T. Seto, T. Matsushima, H. Shimizu, Japanese University Group for the Urgent Joint Observation for the 1995 Hyogo-ken Nanbu Earthquake, Urgent joint observation of aftershocks of the 1995 Hyogo-ken Nanbu earthquake, J. Phys. Earth 44 (1996) 317–328. [3] T. Nakata, K. Yomogida, Surface fault characteristics of the 1995 Hyogoken-Nanbu earthquake, J. Nat. Disaster Sci. 16 (1995) 1–9. [4] H. Horikawa, K. Hirahara, Y. Umeda, M. Hashimoto, F. Kusano, Simultaneous inversion of geodetic and strong motion data for the source process of the Hyogo-ken Nanbu, Japan, earthquake, J. Phys. Earth 44 (1996) 455–471. [5] S. Ide, M. Takeo, Y. Yoshida, Source process of the 1995 Kobe earthquake: determination of spatio-temporal slip distribution by baysian modeling, Bull. Seismol. Soc. Am. 86 (1996) 547–566.
150
T. Okada et al. / Earth and Planetary Science Letters 253 (2007) 143–150
[6] H. Sekiguchi, K. Irikura, T. Iwata, Y. Kakehi, M. Hoshiba, Minute locating of fault planes and source process of the 1995 Hyogoken-nanbu, earthquake from the waveform inversion of strong ground motion, J. Phys. Earth 44 (1996) 473–487. [7] D. Wald, Slip history of the 1995 Kobe, Japan, earthquake determined from strong motion, teleseismic and geodetic data, J. Phys. Earth 44 (1996) 489–503. [8] S. Yoshida, K. Koketsu, B. Shibazaki, T. Sagiya, T. Kato, Y. Yoshida, Joint inversion of near-and far-field waveforms and geodetic data for the rupture process of the 1995 Kobe earthquake, J. Phys. Earth 44 (1996) 437–454. [9] K. Koketsu, T. Ohno, Y. Ikegami, Rupture process inversion using 3-D Green's functions: the 1995 Kobe earthquake revisited, Eos Trans AGU, Fall Meet Suppl, vol. 84, 2003, (Abstract S42H–Abstract S407. [10] T. Okada, T. Matsuzawa, A. Hasegawa, Comparison of source areas of M4.8+/− 0.1 repeating earthquakes off Kamaishi, NE Japan — are asperities persistent features? Earth Planet. Sci. Lett. 213 (2003) 361–374. [11] Hasegawa, A., N. Uchida, T. Igarashi, T. Matsuzawa, T. Okada, S. Miura, Y. Suwa. Asperities and quasi-static slip on the sudducting plate boundary east off Tohoku. NE Japan. SEIZE volume. Columbia Univ. Press., in press. [12] D. Zhao, H. Kanamori, H. Negishi, D. Wiens, Tomography of the source area of the 1995 Kobe earthquake: evidence for fluids at the hypocenter? Science 274 (1996) 1891–1894. [13] D. Zhao, H. Negishi, The 1995 Kobe earthquake: seismic imaging of the source zone and its implications for the rupture nucleation, J. Geophys. Res. 103 (1998) 9967–9986. [14] H. Zhang, C. Thurber, Double-difference tomography: the method and its application to the Hayward Fault,California, Bull. Seismol. Soc. Am. 93 (2003) 1875–1889. [15] F. Waldhauser, W.L. Ellsworth, A double-difference earthquake location algorithm: method and application to the Northern Hayward fault, Bull. Seismol. Soc. Am. 90 (2000) 1353–1368. [16] J. Um, C.H. Thurber, A fast algorithm for two-point seismic ray tracing, Bull. Seismol. Soc. Am. 77 (1987) 972–986. [17] E. Kissling, W. Ellsworth, D. Eberhart-Phillips, U. Kradolfer, Initial reference models in local earthquake tomography, J. Geophys. Res. 99 (1994) 19635–19646. [18] S. Grand, Tomographic inversion for shear wave velocity beneath the North-American plate, J. Geophys. Res. 92 (1987) 14065–14090. [19] E. Humphreys, R. Clayton, Adaptation of back projection tomography to seismic travel time problems, J. Geophys. Res. 93 (1988) 1073–1085. [20] C. Thurber, D. Eberhart-Phillips, Local earthquake tomography with flexible gridding, Comput. Geosci. 25 (1999) 809–818. [21] C. Thurber, S. Roecker, W. Ellsworth, Y. Chen, W. Lutter, R. Sessions, Two-dimensional seismic image of the San Andreas
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
Fault in the Northern Gablian Range, central California: evidence for fluids in the fault zone, Geophys. Res. Lett. 24 (1997) 1591–1594. C. Thurber, S. Roecker, K. Roberts, M. Gold, L. Powell, K. Rittger, Earthquake locations and three-dimensional fault zone structure along the creeping section of the San Andreas fault near Parkfield, CA: preparing for SAFOD, Geophys. Res. Lett. 30 (2003) (2002GL016004). T. Okada, A. Hasegawa, J. Suganomata, N. Umino, H. Zhang, and C. Thurber. Imaging the heterogeneous source area of the 2003 M6.4 northern Miyagi earthquake, NE Japan, by doubledifference tomography. Tectonophys. submitted for publication. T. Okada, A. Hasegawa, J. Suganomata, N. Umino, H. Zhang, C. Thurber, Imaging the fault plane of the 2003 M6.4 northern Miyagi, NE Japan, earthquake by double-difference tomography, Eos Trans AGU, West Pac Geophys Suppl, vol. 85, 2004, p. S42A21. T. Okada, N. Umino, T. Matsuzawa, J. Nakajima, N. Uchida, T. Nakayama, S. Hirahara, T. Sato, S. Hori, T. Kono, Y. Yabe, K. Ariyoshi, S. Gamage, J. Shimizu, J. Suganomata, S. Kita, S. Yui, M. Arao, S. Hondo, T. Mizukami, H. Tsushima, T. Yaginuma, A. Hasegawa, Y. Asano, H. Zhang, C. Thurber, Aftershock distribution and 3D seismic velocity structure in and around the focal area of the 2004 mid Niigata prefecture earthquake obtained by applying double-difference tomography to dense temporary seismic network data, Earth Planets Space 57 (2005) 435–440. T. Okada, T. Yaginuma, N. Umino, T. Matsuzawa, A. Hasegawa, H. Zhang, C. Thurber, Detailed imaging of the fault planes of the 2004 Niigata–Chuetsu, central Japan, earthquake sequence by double-difference tomography, Earth Planet. Sci. Lett. 244 (2006) 32–43. T. Okada, A. Hasegawa, H. Zhang, C. Thurber, Detailed seismic velocity structure around the focal area of the 2000 Western Tottori earthquake (M7.3) by double-difference tomography, 2004 Japan EPS Joint Meeting, 2004, J078-P001. A. Michael, D. Eberhart-Phillips, Relations among fault behavior, subsurface geology, and three-dimensional velocity models, Science 253 (1991) 651–654. D. Eberhart-Phillips, A. Michael, Three-dimensional velocity structure, seismicity, and fault structure in the Parkfield Region, Central California, J. Geophys. Res. 98 (1993) 15737–15758. F. Yamashita, E. Fukuyama, K. Omura, Estimation of fault strength: reconstruction of stress before the 1995 Kobe earthquake, Science 306 (2004) 261–263. A. Hasegawa, J. Nakajima, N. Umino, S. Miura, Deep structure of the northeastern Japan arc and its implications for crustal deformation and shallow seismic activity, Tectonophysics 403 (2005) 59–75.