Tomography of the Source Area of the 1995 Kobe Earthquake: Evidence for Fluids at the Hypocenter?

Kanamori H., Seismol. Res. Lett. 66, 6 (1995);

Somerville P., Eos 76, 49 (1995).

Hagiwara T., Ed., Earthquakes in the Japan Islands: Seismotectonics and Structure (Kashima Press, Tokyo, 1990). The Kobe epicenter is about 200 km north of the Nankai Trough (the major plate boundary between the Philippine Sea plate and the Eurasian plate, Fig. 1C), and about 40 km from the Median Tectonic Line, which is a large strike-slip fault zone in southwestern Japan. Numerous active Quaternary faults exist in the region. In the past century, four major intraplate earthquakes have occurred in central to western Japan: the 1891 Nobi (M = 8), 1927 Tango (M = 7.3), 1943 Tottori (M = 7.2), and 1948 Fukui (M = 7.1) earthquakes. Two historical earthquakes occurred in the vicinity of the Kobe earthquake, M > 7 in 868, and M > 6 in 1916.

Tsuboi S., et al., J. Seismol. Soc. Japan 42, 277 (1989). More than 90 permanent seismic stations are operated in southwestern Japan by Kyoto University, Nagoya University, Kochi University, Kyushu University, and the University of Tokyo, which are used to detect and record micro-earthquakes in this region. Among the 37 stations used in this study (Fig. 1B), 33 are equipped with three-component seismographs and four with one-component seismometers. All the P- and S-wave arrival times were double-checked by the analysts of the permanent networks.

Ohmi S., Progr. Abstr. Fall Meet. Seismol. Soc. Japan, A39 (1995);

Umeda Y., Proc. 1996 Japan Earth Planet. Sci. Meet., 36 (1996);

N. Hirata the Urgent Observation Group (GROUP-95) for the 1995 Hyogo-Ken Nanbu Earthquake, J. Phys. Earth, in press. All 30 portable stations are equipped with three-component seismometers. All the P- and S-wave arrival times were double-checked by the GROUP-95 analysts and the present authors. More than 98% of the P- and S-wave data used in this study was generated by the Kobe aftershocks and local micro-earthquakes with M < 3. The dominant frequency was 8 to 10 Hz for P waves and 5 to 8 Hz for S waves, and the Fresnel zones did not exceed 0.8 km, which is much smaller than the grid spacing we adopted in this study

Lees J. M., Geophys. Res. Lett. 17, 1433 (1990);

Lees J. M., Nicholson C. E., Geology 21, 387 (1993);

Foxall W., Michelini A., McEvilly T. V., J. Geophys. Res. 98, 17691 (1993);

Zhao D., Kanamori H., Geophys. Res. Lett. 20, 1083 (1993);

ibid 22, 763 (1995).

Eberhart-Phillips D., Michael A. J., J. Geophys. Res. 98, 15737 (1993);

Johnson P. A., McEvilly T. V., ibid. 100, 12937 (1995);

Fuis G. S., et al., Eos 77, 173 (1996).

Gupta H. K., Sarma S. V. S., Harinarayana T., Virupakshi G., Geophys. Res. Lett. 23, 1569 (1996). Seismologic and magnetotelluric studies revealed a low-velocity, high conductive anomaly near the hypocenter of the 1993 Latur earthquake in India. The anomaly was interpreted as a fluid-filled, fractured rock matrix that may have contributed to the nucleation of the Latur earthquake.

Zhao D., Hasegawa A., Horiuchi S., J. Geophys. Res. 97, 19909 (1992). We set up a 3D grid in the study area with a grid spacing of 4 to 5 km in the horizontal direction and 2 to 4 km in depth. Hypocenter locations and velocities at the grid nodes are taken as unknown parameters. The velocity at any point in the model is calculated by linearly interpolating the velocities at the eight grid nodes surrounding that point. An efficient 3D ray tracing technique is used to compute travel times and ray paths. We used a conjugate gradients algorithm to invert the large and sparse system of observation equations that relate the observed arrival-time data to the unknown parameters. P- and S-wave velocity structure and hypocenter parameters are simultaneously determined in an iterative inversion process. Then we use the relation (V_p/V_s) ² = 2(1-δ)/(1-2δ), to determine distribution. The tomographic inversion was conducted for the area shown in Fig. 1. Only the ray paths within that area were used in the inversion. We did not use station corrections in our parameterization; any local anomalies close to stations can be resolved as velocity anomalies at the grid nodes of the top grid mesh layer at a depth of 0 km. Elevations of stations are included in the 3D ray tracing. The initial one-dimensional velocity model consists of four layers that are separated by the Conrad and Moho discontinuities and a velocity boundary at a depth of 3 km. In the four layers, Vp is 4.8, 6.0, 6.7, and 7.8 km/s. Vp/Vs ratio is 1.73. The Conrad discontinuity is at a depth of 15 to 17 km, and the Moho is at a depth of 34 to 35 km in the Kobe area (

Zhao D., Horiuchi S., Hasegawa A., Tectonophysics 212, 289 (1992). We have taken into account the Conrad and Moho depth changes in the tomographic inversions and found that the velocity changes due to the Conrad and Moho depth variations are less than 0.5% in the tomographic images (Fig. 2, Fig. 3, Fig. 4).

Zhao D., Kanamori H., Wiens D., Negishi H., Eos (fall suppl.) 76, 378 (1995).

Fault zones are typically less than 1 km wide and composed of highly fractured material, fault gouge, and fluids; see Scholz C. H., Geology 15, 493 (1987);Cox S., Scholz C. H., J. Struct. Geol. 10, 413 (1988);and Forster C. B., Evans J. P., Geophys. Res. Lett. 18, 979 (1991).. Since the spacing of grid nodes used in our tomographic inversion is 4 to 5 km, the Kobe earthquake fault zone was imaged as a low-velocity zone not narrower than 4 km

Hirata N., et al., Proc. 1996 Japan Earth Planet. Sci. Meet., 38 (1996);

Tsumura N., et al., ibid., p. 39.

O'Connell R. J., Budiansky B., J. Geophys. Res. 79, 5412 (1974);

Toksoz M. N., Cheng C. H., Timur A., Geophysics 41, 621 (1976);

Moos D., Zoback M. D., J. Geophys. Res. 88, 2345 (1983).

To make a checkerboard, we assigned positive and negative velocity anomalies to the 3D grid nodes. Synthetic data are calculated for the checkerboard model. Then we added random errors of 0.05 to 0.15 s to the synthetic data and inverted them with the same algorithm as we did for the observed data. The inverted image of the checkerboard suggests where the resolution is good and where it is poor

Mamada Y., et al., Progr. Abstr. Fall Meet. Seismol. Soc. Japan, P37 (1995).

Fukuyama E., ibid., p. A87;

Tadokoro K., Umeda Y., Ando M., Proc. 1996 Japan Earth Planet. Sci. Meet., 44 (1996). The direction of the fastest S waves is generally oriented east-west in the Nojima fault zone. The time difference between the two splitting S waves is generally 0.1 s and does not exceed 0.15 s. This means that the effect of S-wave splitting on the accuracy of S-wave arrival-time data is not larger than 0.15 s.

Ito K., Umeda Y., Ohmi S., Ohigashi A., Matsumura K., Progr. Abstr. Fall Meet. Seismol. Soc. Japan, A79 (1995).

Umeda Y., Yamashita T., Ito K., Horikawa H., ibid., p. A78.

Katao H., Maeda N., Hiramatsu Y., Iio Y., Nakao S., ibid., p. A74.

Obara K., et al., Proc. 1996 Japan Earth Planet. Sci. Meet., 46 (1996);

Matsumoto S., et al., ibid., p. 46.

Chatterjee S. N., Pitt A. M., Iyer H. M., J. Volcanol. Geotherm. Res. 26, 213 (1985);

Walck M. C., J. Geophys. Res. 93, 2047 (1988);

Hasegawa A., Zhao D., in Magmatic Systems, , Ryan M. P., Ed. (Academic Press, San Diego, 1994), pp. 179–195

Yokoyama I., Aramaki S., Nakamura K., Volcanoes (Iwanami Press, Tokyo, 1987).

Okubo Y., Tsu H., Ogawa K., Tectonophysics 159, 279 (1989).

Kerrich R., La Tour T. E., Willmore L., J. Geophys. Res. 89, 4331 (1984). Fluids may exist in the middle to lower crust in fault zone environments.

Hirahara K., Tectonophysics 79, 1 (1981);

Zhao D., Hasegawa A., Kanamori H., J. Geophys. Res. 99, 22313 (1994).

Sibson R. H., in Earthquake Prediction: An International Review, Maurice Ewing Ser., vol. 4, , Simpson D. W., Richards P. G., Eds. (American Geophysical Union, Washington, DC, 1981), pp. 593–603;

Sibson R. H., Tectonophysics 211, 283 (1992);

Hickman S., Sibson R. H., Bruhn R., J. Geophys. Res. 100, 12831 (1995).

We are grateful to the members of the Urgent Observation Group for the 1995 Hyogo-Ken Nanbu Earthquake, who operated the portable stations and picked the P- and S-wave arrival times, and to the staff members of the seismic networks of Kyoto University, Nagoya University, Kochi University, Kyushu University, and the University of Tokyo for providing the data recorded by their permanent networks that were used in this study. J. Vidale and an anonymous referee provided thoughtful reviews, which improved the manuscript. This work was partially supported by a grant from the National Science Foundation (EAR-9526810) to D. Zhao. This paper is Contribution 5785, Division of Geological and Planetary Sciences, California Institute of Technology