Relationship between the detailed hypocenter distribution and the velocity structure in the western Nagano Prefecture, central Japan, derived from travel time tomography using dense array data

Issei DOI1, Shunta NODA1,3, Yoshihisa IIO1, Shigeki Horiuchi2 and Shoji SEKIGUCHI2

1Disaster Prevention Research Institute, Kyoto University 2National Research Institute for Earth Science and Disaster Prevention 3 Railway Technical Research Institute

We conducted three-dimensional travel time tomography in and around the source region of the 1984 Western Nagano Prefecture to investigate the generation process of the and swarm activity near the source. As many as about 250,000 travel time data of good quality (1 ms error) from a dense network were compiled and we obtained hypocentral distribution and three-dimensional velocity structure with high accuracy and resolution. Most of the estimated hypocenters were aligned in lines or planes, not in the form of masses. We found the hypocenter distribution corresponding to the mainshock plane, and it was sandwiched by high-velocity regions on both horizontal sides. A low-velocity region spreading horizontally at the bottom of this hypocenter alignment was also seen. In the northeastern side of the mainshock fault, where there is swarm activity, we detected several other alignments of hypocenters, most of which may be located at the boundaries of geological strata. A low-velocity region was also found at the bottom of some of these alignments. This low-velocity region may be due to fluids from below the seismogenic zone. The generation of both the mainshock and the swarm activity might be related to fluid intrusions from the lower side of the hypocenter alignments.

1. Introduction 2. Data The 1984 Western Nagano Prefecture Earthquake (Mj 6.8) occurred about 10 km southeast of Mt. A dense earthquake observation is conducted Ontake, an active volcano in central Japan, on from 1995 to the present in and around the source September 14, 1984, and caused huge damage, region of the 1984 Western Nagano Prefecture including landslides. The earthquake fault estimated Earthquake 3). This observation network has 57 is right lateral 1), trending ENE-WSW. Ooida et al. stations, shown by green squares in Figure 1, which [1989] 2) pointed out high seismic activity in the are settled on hard rock with plaster so that artificial hypocentral region of this earthquake even before noise is very small. The station spacing is 1–4 km, the mainshock (at least from May, 1978, when Mt. covering the source regions of the mainshock and Ontake erupted). We can see not only seismic swarm activity. Waveforms are recorded with 10 activity distributed along the mainshock fault plane, kHz sampling and 16-bit resolution on high- and but swarm-like activity can also be seen in the low-gain channels. The clocks are corrected by GPS eastern parts of the source region. every two hours, and the error in the accuracy of the The dense seismic observation 3) is conducted in absolute time is less than 1 ms. There is no and around the source region of the 1984 Western semi-stationary seismic network with high density Nagano Earthquake. The data from this observation and sampling in the world such as this one. enable us to estimate more precise hypocenter We used 14,226 hypocenters from October 1995 locations and more detailed velocity structures than to February 2005, indicated by solid circles in before. In this study, we estimate a detailed Figure 2. We used 250,212 P wave and 209,656 S three-dimensional velocity structure with a wave travel times from these events, which had resolution ~1 km and discuss the relationship of the more than ten travel times of both P and S waves. hypocenter distribution and fluids to elucidate the All P and arrival times were manually generation process of . picked. Transverse components were used for S wave picking to reduce the effect of converted distribution near the mainshock fault plane 1), i.e., 4 waves such as sP phases. The accuracy of P and S < Y < 6 [shown by arrows A in Figure 2(2), (3)] at arrival times are ~1 and ~30 ms, respectively. Such depths of 2–6 km. At the bottom of these a high accuracy level could be achieved because of alignments, hypocenter distributions perpendicular the high sampling rate of 10 kHz and an adequate to these are detected with south dipping, shown by S/N ratio. arrows B in Figure 2(3), to a depth of 8 km. In the eastern part of these hypocenter 3. Method distributions, where swarm activities are observed, we found hypocenter alignments with the This study uses the following three steps to northeastern dips (30–60 degrees) which were split obtain three-dimensional velocity structures. First, into several parallel planes [shown by arrows C1 we determined the initial hypocenter locations and and C2 in Figure 2(4)]. These alignments were times with a fixed initial velocity structure. We used 5–10 km long. the velocity model of Hirahara et al. [1992] 4) as the initial structure, which employed approximately (2) Features of the three-dimensional velocity 7000 picks from the observation data to structure conduct travel time tomography. For the next step, We detected two low-velocity regions on the we conducted one-dimensional inversion by northern side of the mainshock fault at a depth of 2 compiling the hypocenter information obtained in km, shown in Figure 2(1) as L1 and L2. The eastern the first step and estimated the one-dimensional region, L1, at a depth of 1 km to 6 km, is connected velocity profiles and station corrections. Finally, we to the northeastern side at depths of 6 km. The performed three-dimensional seismic tomography western one, L2, was imaged sharply at depths of to obtain three-dimensional velocity perturbations 2–3 km. A low-velocity region L2’ was spreading at and recalculated station corrections. These three a depth of 6 km in the west of the swarm region steps yielded more precise hypocenter locations and [Figure 2(2)]. velocity structure than those obtained by directly On the other hand, several high-velocity regions putting the data into three-dimensional tomography were found in the analysis area. High-velocity 5). We repeated every step four times to obtain the region H1 is located at depths of 2–6 km, separating final model. We referred only to the P wave velocity the low-velocity regions L1 and L2 [Figure 2(1), structure, considering the magnitudes of the errors (2)]. As can be seen from Figure 2(4), H1 includes in picking. the two northwestern-dipping high-velocity layers In the ray calculation, the pseudo-bending H1a and H1b, which are found in the swarm method 6) was used: the initial path is perturbed by a activity region. We could find two high-velocity geometric interpretation of the ray equations, and regions H2 and H3 at depths of 3–4 km near the the travel time along the path is minimized in a mainshock fault [Figure 2(3)]. piecewise fashion. The estimated error in this Hypocenters are mainly distributed outside the method is about 1 ms. In the matrix calculation, we main low-velocity regions L1 and L2. Hypocenter used the LSQR method 7). distribution A was located in a narrow, relatively Figure 1 shows the lateral grid locations used for low-velocity zone sandwiched by two high-velocity the three-dimensional inversion. We set the X, Y, regions H2 and H3 on either horizontal sides, while and Z axes as the strike direction of the mainshock hypocenter distribution B was located in or on the (N70 E 1)), the perpendicular, and depth directions, top of the low-velocity regions [Figure 2(3)]. respectively. High-velocity layers H1a and H1b have a dip To check the resolutions of the obtained model, parallel to hypocenter alignments C1 and C2, and we conducted a checkerboard test 8). We made low-velocity patches are found along C1, between synthetic travel time data by adding 5% velocity these high-velocity layers. We found a low-velocity perturbations to the initial velocity model and region, the lower end of L1, at the bottoms of C1 random noise. and C2.

4. Results 5. Discussion

(1) Features of hypocenter distribution (1) The fault plane of the mainshock Most hypocenters are determined in lines or We detected near-vertical hypocenter alignment planes, not in the form of masses, even in the A near the hypocenter of the 1984 Western Nagano swarm activity region. Earthquake. We think that this alignment We found the 5 km long, near-vertical hypocenter corresponds to the mainshock fault plane because it shows a dip and strike similar to the mainshock study area, which are considered to be generated by . This alignment also has a similar fluid movement 13). strike and dip, compared to the fault plane 1), but In the swarm activity region, some of the was located at a distance of 1–2 km in the –Y hypocenter alignments (C1 and C2) and velocity direction, as shown in Figure 2(3). This shift layers (H1a and H1b) are sub-parallel to each other, appeared possibly because they used data only from suggesting that those hypocenters are located at the permanent stations of the Japan Meteorological boundaries of geological strata [Figure 2(4)]. Agency (JMA), which are spaced about 50 km apart. Low-velocity region L1 is located at the lower limit We could recognize clear hypocenter alignments by of hypocenter distributions C1 and C2. This fact determining the very precise hypocenters implies that fluids may pour into the boundaries of determined by the dense network, contributing to geological strata, generating these earthquakes of accurate estimation of the fault plane. reverse fault type 2).

(2) Relation between mainshock generation and 6. Conclusions velocity structure As described in Section 4, the mainshock fault We conducted three-dimensional travel time plane (denoted by arrows A) corresponds to a tomography in and around the source region of the relatively low-velocity region sandwiched between 1984 Western Nagano Earthquake using about high-velocity regions H2 and H3. We detected a 250,000 highly accurate picks (errors less than 1 low-velocity region L2’ with a small southern dip, ms) from a dense network (station spacing 1–4 km). spreading at the bottom of hypocenter alignment A, We obtained very accurate hypocenter distributions at depths of 6–8 km [Figure 2(4)]. Inamori et al. and detailed three-dimensional velocity structure at [1992] found an S wave reflection plane dipping to depths of 0–6 km with high resolutions and found: the south at depths of 8–14 km beneath the i) Most of the hypocenters were aligned in lines mainshock fault and suggested the existence of or planes, not in the form of masses, even in the fluid through discussion of the amplitudes of the swarm activity region. reflection phases. This reflection plane is located in ii) We found hypocenter alignments corresponding and above the low-velocity region L2’ and is to the mainshock fault. These alignments have a dip thought to be produced by the impedance contrast and strike similar to those in the fault zone 1), but of its boundary. The region L2’ may be considered are located 1–2 km southward. The mainshock fault to be filled with fluids. zone corresponds to the narrow, relatively It is considered possible that region L2’ is related low-velocity zone horizontally sandwiched between to mainshock generation. Fluids might leak into the two high-velocity regions. A low-velocity region is mainshock fault zone at the bottom. The fault zone found at the lower limit of the hypocentral generally contains many cracks 9), allowing the distribution of the mainshock fault. fluids to intrude. The fluids pouring into the fault iii) We found several hypocentral alignments with zone increase the pore pressure and decrease the smaller dips than the mainshock fault in the swarm effective normal stress. activity region. We also detected a broad low-velocity region at a depth of 2 km that comes (3) Relationship between swarm activity and the upward from a depth of at least 6 km. Based on the velocity structure results from earlier studies it can be said that this Low-velocity region L1 arises from a depth low-velocity region is formed by fluids from the of 6 km to 1–2 km on the northeastern side of the mantle, and located at the lower limit of these swarm activity region. Kasaya et al. [2002] 10) hypocenter distributions. found a low-resistivity region at a depth of 2 km iv) Both the mainshock and the swarm activity that continues till a depth of 10 km. This may be induced by fluid intrusions into low-resistivity region almost corresponds to high-velocity regions from the lower limit of the low-velocity regions L1 and L2. Fluids originating hypocentral distributions. in the mantle were found in hot springs 11), and surface uplift is observed in the swarm activity region 12), where we detected low-velocity region ACKNOWLEDGMENT: We thank Shiro Ohmi L1. This suggests that L1 corresponds to the fluid for developing the novel data processing system path through which it rises upward from the mantle used in the first several years. We also thank Eiji and is accumulated at depths of 1–2 km. This is Yamamoto, Hisao Ito, Yasuto Kuwahara, Takao supported by the fact that deep low-frequency Ohminato for their great contributions for borehole events occur at depths of 15–30 km just beneath the observations. We are grateful for Kaori Takai for her great help for data processing. This work was Sano, 2003, Volatile element isotope systematics partially supported by JSPS.KAKENHI (19204043), at Ontake volcano, Jpn, Geochem. J., 37, Japan. 299–310. 12) Kimata, F., R. Miyajima, M. Murase, D. REFERENCES Darwaman, T. Ito, Y. Ohta, M. Irwan, K. Takano, 1) Yoshida, S., and K. Koketsu,1990, Simultaneous F. Ibrahim, E. Koyama, H. Tsuji, T. Takayama, K. inversion of waveform and geodetic data for the Uchida, J. Okada, D. Solim, and H. Anderson, rupture process of the 1984 Naganoken-Seibu, 2004, Ground uplift detected by precise leveling Japan, earthquake, Geophys J. Int., 103, 355-362. in the Ontake earthquakes swarm area, central 2) Ooida, T., F. Yamazaki, I. Fujii, and H. Aoki, Japan in 2002-2004, Earth Planets Space, 56, 1989, Aftershock activity of the 1984 Western e45-e48. Nagano Prefecture Earthquake, central Japan, 13) Ohmi, S., and K. Obara, 2002, Deep and its relation to earthquakes swarms, J. Phys. Low-Frequency Earthquakes beneath the Focal Earth, 37, 401-416. Region of the Mw 6.7 2000 Western Tottori 3) Iio, Y., S. Ohmi, R. Ikeda, E. Yamamoto, H. Ito, Earthquake, Geophys. Res. Lett., 29, No.16, H. Sato, Y. Kuwahara, T. Ohminato, B. Shibazaki, 10.1029/2001GL014469. and M. Ando, 1999, Slow initial phase generated by microearthquakes occurring in the Western Nagano Prefecture, Japan. - The source effect -, Geophys. Res. Lett., 26, 1969-1972. 4) Hirahara, K., and The Members of the 1986 Joint Seismological Research In Western Nagano Prefecture, 1992, Three-dimensional P and S wave velocity structure in the focal region of the 1984 Western Nagano Prefecture Earthquake, J. Phys. Earth, 40, 343-360. 5) Shibutani, T., H. Katao, and Group for the dense aftershock observations of the 2000 Western Tottori Earthquake, 2005, Very dense aftershock observations of the 2000 Western Tottori Figure 1 Hypocenter (black dots) and station (green Earthquake (Mj=7.3) in southwestern Honshu, squares) distributions used in this study. We show Japan: high resolution aftershock distribution, the horizontal locations of the grids by crosses. focal mechanisms and 3-D velocity structure in Red star and black rectangles denote the the source region, Earth Planets Space, 57, mainshock and fault plane of the 1984 825-838. Western Nagano Prefecture Earthquake 1). Two 6) Um, J., and C. Thurber, 1987, A fast algorithm lines denote the cross-sections for Figure 2. for two-point seismic ray tracing, Bull. Seism. Soc. Am., 77, 972-986. 7) Paige, C. C., and M. A. Saunders, 1982, LSQR: an algorithm for sparse linear equations and sparse least squares, ACM Trans. Math. Software, 8, 43-71. 8) Inoue, H., Y. Fukao, K. Tanabe, and K. Ogata, 1990, Whole mantle P-wave travel time tomography, Phys Earth Planet. Inter., 59, 294-328. 9) Chester, F.M., J.P. Evens, and R.L. Biegel, 1993, Internal structure and weakening mechanisms of the San Andreas fault, J. geophys. Res., 98, 777–786. 10) Kasaya, T., N. Oshiman, N. Sumitomo, M. Uyeshima, Y. Iio, and D. Uehara, 2002, Figure 2 (1)(2) X-Y and (3)(4) X-Z cross sections Resistivity structure around hypocental area of of the velocity perturbation models. Black dots the 1984 Nagano Prefecture earthquake in central denote the hypocenters. We show the region with Japan, Earth Planets Space, 54, 107–118. a restoration rate of more than 0.3 in the 11) Takahata, N., R. Yokochi, Y. Nishio, and Y. checkerboard test by solid lines.