Historic earthquake relocation
Transcription
Historic earthquake relocation
investigation Historic earthquake relocation Yunong Nina Lin Institute of Geosciences, National Taiwan University, No. 1, Sec. 4, Roosevelt Rd., Taipei, Taiwan, R.O.C. The related subsurface fault planes of the Hsinchu-Taichung earthquake, Taiwan in 1935 remains a puzzle till now. Although according to some previous literature, they were recommended to be a dextral fault with almost vertical fault plane and a back thrust dipping to the west, we have no strong seismic proof of the existence of the back thrust fault so far. In this study, we tried to relocate the micro-earthquakes recorded in 1935, and figured out what the possible fault plane geometry of the back thrust is. The result shows a west-dipping and deep-seated fault plane. This fault plane may have a gradually changing dip from 60° at the top to 45° at the bottom (~10 km). A big earthquake (MGR=7.1)1 took place in northwestern Taiwan on April 21st, 1935. This earthquake daylighted two surface ruptures: The Tuntzuchiao Fault in the south, which is a dextral strike-slip fault, and the Siko Fault in the north, which is a thrust fault. Based on the some wave-form and numerical modelings done by Sheu et al. (1982) 2, Lin (1987) 3 and Huang & Yeh (1994)4, the fault plane geometry of the Siko Fault is west dipping, concave upward and deep-seated (up to 10 km in depth). In Taiwan, a west-dipping fault plane infers a hinterland sequence under regional tectonic setting, which is not a unique phenomenon here. However, a “west dipping” and “deep-seated” fault plane in Taiwan is rare, for most west dipping fault planes in Taiwan are shallow-seated. Therefore, the existence of the Siko fault has a strong influence on the reconstruction of the subsurface structural geology, if the fault-plane geometry is real. Nevertheless, except those modeling results, there are no any other strong physical proofs, such as seismicities, standing for this fault plane geometry so far. The purpose of this study is to re-image the fault plane of the Siko Fault via seismic data. Data A B Chinshui Miaoli Tunglo Chuhuangkeng NTU|20 JUNE 2004 We searched the catalog of Central Weather Bureau (CWB) in Taiwan, through a period from 1973 to 2003. Unfortunately, in these latest two decades this once seismically active area seems going through a quiescence stage (Fig. 1). Since it is not possible to re-image the fault plane via such few earthquakes, we turned to the historic seismic record. In 1936, Nasu (1936)5 and his colleagues came to Taiwan doing some micro-earthquake observations from August to December. They set up 4 seismic stations in Miaoli, Chuhaungkeng, Tunglo and Chinshui. Totally 55 earthquakes with better quality were published by ERI, including the P-S travel time residual, maximum acceleration and initially-calculated depth. We took advantage of these recorded P-S travel time residual and tried to relocate these earthquakes. Velocity model Because only P-S travel time residual was provided, we need to have the P- and S- wave velocity in order to do the relocation. Besides of the 1D P-wave velocity structure published by CWB, we also tried the velocity model specific for the Hsinchu-Taichung area done by Lin et al. (1989, with both P-wave and S-wave velocity model)6, Lee (1998, S-wave velocity model) 7 and Sheng (1999, P-wave velocity mode) 8. We combined the models of Lee and Sheng to be ShL model, and then evaluated these three models of CWB, Lin and ShL. A FORTRAN program VELEST was introduced here to do this evaluation. This freeware could be obtained on the website http://quake.usgs.gov/research/software/index.html Figure 1. The map showing the regional tectonic setting and all the earthquakes from 1973 to 2003 in CWB catalog. A: The regional tectonic feature. Our study area is located at the northwestern Taiwan. B: The catalog earthquakes relocated by hypoDD. The big blue circle represents the 1935 mainshock. Two blue lines represent the Tuntzuchiao Fault in the south and the Siko Fault in the north. Obviously, only very few earthquakes happened near the Siko Fault in these two decades. ©2004 NTU publishing Group 1 investigation 2.4 and 3.4 seconds at distance = 0 km). Finally we decided to take the CWB model for further relocation job. A Relocation method & result The method we used is the most basic inversion due to the lack of other information necessary for complicated inversion. A FORTRAN subroutine called CPTIME was engaged here, with velocity model, depth and distance as the inputs and travel time as the output. We applied the CWB velocity model, initially-calculated depth and travel time residuals to make the distance as output. We also tried to use a uniform P-S velocity difference to do the relocation. The result shows very similar patterns but different scattering diameters of these micro-earthquakes (Fig. 4). We made a cross-section from (120.5°E, 25.25°N) to (121.05°E, 24.75°N) and projected all the B A B Figure 2. A: The velocity structure models tested in this study. Solid lines: the original models; dashed lines: the modified models after running with the program VELEST. B: Model RMS testing. The . average RMS values for the models from top to bottom are 0.13, 0.19 and 0.19. The upper one (CWB) model has the lowest proportion of higher RMS values. For more usage and details, please also refer to the user guide in the package. We input the earthquakes from CWB catalog to be the testing dataset (1501 earthquakes), and tested the three velocity models respectively (Fig. 2). In the Figure 2-(A), the solid line means the original model while the dash line means the modified model. Figure 2-(B) shows the RMS distribution of the three models. Obviously after the VELEST, each model shows extremely low RMS level, which means good fits to the CWB dataset. Among these three, the CWB model has the lowest average (~0.13) and the smallest proportion of the higher RMS (which means, a smaller tail in the higher RMS). We also plotted the travel time curve of the three models (Fig. 3). We found that the P-S travel time residual at smaller distance of the CWB model fits our the historic data best (travel time residual = 0.8 second at distance = 0 km), for most of the recorded travel time residuals ranging from 2.0-5.5 seconds, which are already beyond the resolution of the other two models (the smallest travel time residuals of the other two models are NTU|20 JUNE 2004 C Figure 3. The travel time curve derived from the three velocity models by CPTIME calculation. Among these three, only the CWB model has the P-curve (blue) and the S-curve (red) get very close to each other at the distance of 0 km. For the other two, the time residual between P and S at distance = 0 km is already too large to be applied to those historic earthquakes. ©2004 NTU publishing Group 2 investigation earthquakes onto a vertical plane (Fig. 5). It is clearly seen that in Fig. 5-(A), a west-dipping plane gradually changing dip form nearly 60° at the top and 45° at the bottom (~ 10 km deep) is imaged. From Fig. 5-(B) to 5-(F), a similar west-dipping deep-seated plane is also recognized, while the character of dipping may slightly differ. This allows us to assure the existence of such a special fault plane. Discussion The purpose of this short study, to re-image the fault plane of the Siko Fault, seems to be satisfied here. However, some sources of errors should be taken into consideration. First, the exact locations of the 4 seismic stations remain unknown. The X, Y, and Z coordinates of the stations are arbitrarily read from the 1/25000 map. Second, the CWB velocity model only provide with the P-wave velocity. As for the S-wave velocity, we simply calculated by the formula Vs = 3Vp . This may represent the true S-wave velocity model and cause some potential errors. Third, the quality of the data itself. The recorded precision of the P-S time residual is up to 0.01 second, which might already be the best available precision in 1935. NTU|20 JUNE 2004 However, it is still not precise enough to reduce the relocation RMS to an extremely small extent. We finally set the RMS threshold for the summation of the X, Y, and Z dimension to be 1.5 km, and consequently lost 14 earthquakes after setting this RMS threshold. Even though, we still could not deny the imaged geometry presented by these earthquakes. The result reminds us to re-examine the role and importance of this deep-seated back thrust in the northwester Taiwan. 1. 2. 3. 4. 5. 6. 7. 8. Gutenberg, B. & Riichter C.F. Seismicity of the earth and associated phenomena. Priceton University Press. 310p (1949). Sheu, H.C., Kosuga, M., & Sato, H. Mechanism and fault model of the Hsinchu-Taichung (Taiwan) earthquake of 1935. Zisin II 35, 567-574 (1982). Lin, D.H. Mechanism of the Hsinchu-Taichung, Taiwan, earthquake of 1935. M.S. Thesis, National Central Univ. 88p (1987) Huang, B.S. & Yeh, Y.T. Source geometry and slip distribution of the April 21, 1935 Hsinchu-Taichung, Taiwan earthquake. Tectonophysics 210, 77-90 (1992). Nasu, N. The after-shocks of the Formosa earthquake of 1935. Bulletin of Earthquake Research Institute Suppl. 3, 10-21 (1936). Lin, C.H., Yeh, Y.H., & Roecker, S.W. Seismic velocity structures in the Sanyi-Fengyuan area, central Taiwan. Proc. Geol. Soc. China 32, 1, 101-120 (1989). Sheng, C.H. The study of the velocity structure of the northwestern linear seismic zone by ray tracing. M.S. Thesis, National Central Univ. 103p (1999). Lee, H.C. The S-wave velocity structure under center-west Taiwan by strong ground motion waveform modeling. M.S. Thesis, National Central Univ. 138p (1998). ©2004 NTU publishing Group 3 investigation A B C D E F Figure 4 The relocated earthquakes. A: CWB velocity structures. B: uniform velocity of 4.0 km/s. C: uniform velocity of 4.5 km/s. D: uniform velocity of 5.0 km/s. E: uniform velocity of 5.5 km/s. F: uniform velocity of 6.0 km/s. Blue lines: the Tuntzuchiao Fault (south) and the Siko Fault (north). The coordinate system is the Taiwan TM-2 Coordinate system, not the longitude and latitude system. Figure 5 The earthquakes projected onto the vertical plane with two end points as (120.5°E, 25.25°N, left) and (121.05°E, 24.75°N, right). Notice that from A to F, a west-dipping fault plane is imaged and has a concave upward geometry. This fault plane extends to about 10 km in depth. NTU|20 JUNE 2004 A B C D E F ©2004 NTU publishing Group 4