S wave tomography of the crust and uppermost mantle in China

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 113,, doi:1.129/28jb5836, 28 S wave tomography of the crust and uppermost mantle in China Youshun Sun, 1 M. Nafi Toksöz, 1 Shunping Pei, 2 Dapeng Zhao, 3 F. Dale Morgan, 1 and Anca Rosca Received 29 May 28; revised September 28; accepted 19 September 28; published 21 November 28. [1] We use 28, local and regional arrival times from 13,15 local earthquakes recorded by 22 seismic stations to determine a three-dimensional S wave velocity structure of the crust and uppermost mantle under China and surrounding regions. We use a traveltime tomography code that accommodates a heterogeneous crust and varying Moho depth. The grid for tomographic inversion is 1 1 in the horizontal direction and 1 km in depth. Our results show that large velocity variations of more than 6% exist in the crust and upper mantle in the China region. The velocity image of the upper crust correlates with surface geological features. Crustal heterogeneity is clearly observed, and velocity changes are visible across some of the large fault zones. In Tibet, low velocities exist in the thick crust, and high Sn velocities exist in the southern and eastern parts. Low-velocity zones beneath volcanic sites and the rifts are clearly observed in our tomographic results. Under the Tengchong, Hainan, and Changbai volcanic areas, strong low velocity zones are visible down to the upper mantle, indicating the existence of magma chambers beneath the volcanoes. Citation: Sun, Y., M. N. Toksöz, S. Pei, D. Zhao, F. D. Morgan, and A. Rosca (28), S wave tomography of the crust and uppermost mantle in China, J. Geophys. Res., 113,, doi:1.129/28jb Introduction [2] In this paper we present a 3-D shear wave velocity structure for the crust and uppermost mantle under China and surrounding regions (Figure 1). The velocity structure is obtained using local and regional traveltimes and a tomographic inversion method. [3] The available S wave velocity models of the crust and upper mantle in China and the surrounding area have been obtained using different approaches. Global models such as CUB 1. [Shapiro and Ritzwoller, 22] and the SAIC 1 1 model [Stevens et al., 21] were constructed from group and phase velocity dispersion measurements of surface waves. The global model CRUST 2. [Bassin et al., 2] was constructed from seismic refraction data and developed from the CRUST 5.1 model [Mooney et al., 1998] and a 1 1 sediment map [Laske and Masters, 1997]. Only P wave velocities are inverted by traveltime tomography. The S wave velocities in the model are obtained by empirical Vp/Vs ratios or compiled from other sources. For East Asia, mantle S velocity models were obtained from shear and surface waveforms [Friederich, 23]. Lateral spatial resolution of the models is larger than 2 km. 1 Earth Resources Laboratory, Department of Earth, Atmospheric, and Planetary Sciences, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA. 2 Institute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing, China. 3 Department of Geophysics, Tohoku University, Sendai, Japan. Chevron-Texaco, San Ramon, California, USA. Copyright 28 by the American Geophysical Union /8/28JB5836$9. [] Regional models were constructed by Sn and/or Pn tomography [Hearn and Ni, 21; Ritzwoller et al., 22; Liang et al., 2; Hearn et al., 2; Liang and Song, 26; Phillips et al., 27; Pei et al., 2, 27] and from surface waves [Song et al., 1991; Wu et al., 1997; Zhu et al., 22; Huang et al., 23; Lebedev and Nolet, 23]. P and S wave tomography have been performed in several local regions in China [Xu et al., 22; Yu et al., 23; Jia et al., 25; Liu et al., 25; Chang et al., 27]. Xu et al. [22] used P and S wave arrival times of local, regional, and teleseismic events recorded by Chinese and Kyrgyzstan seismic networks to derive crust and upper mantle velocity structures beneath western China. The grid spacing is in the horizontal direction and about 1 km (in the crust) and km (in the uppermost mantle) in the vertical direction. Crustal structure models beneath north and east China, including Beijing and surrounding regions, were obtained by Zhu et al. [199], Yu et al. [23], and Jia et al. [25] with P and S wave tomography. Liu et al. [25] inverted the lithospheric structure in southwest China from local-regional traveltime data. A number of studies carried out in the Tien Shan (also known as Tianshan) [Roecker et al., 1993; Poupinet et al., 22; Vinnik et al., 22, 2; Lei and Zhao, 27] and Hindu Kush [Koulakov and Sobolev, 26] regions, which include parts of western China, show heterogeneous crust and upper mantle structures with crustal thickness varying between and 7 km. All of these models provide detailed crustal structures in specific regions. A detailed map for the whole China area remains to be developed. [5] During the last 2 decades, many digital seismic stations were installed in China. The large database of highquality recorded arrival times provides an unprecedented 1of1

2 MONGOLIA SoB Tian Shan on rat CB Basin Tarim OB LM S Tibetan Plateau SB TC o -K o Sin C rea JAPAN SEA on rat ze C t g n Ya hina th C Sou lock B INDIA STB IB HN PHILIPPINE SEA KB Figure 1. The 512 earthquakes (M > 6. from January 1978 to May 2), 22 stations, active faults, and major tectonic boundaries in China and the surrounding area. Earthquake epicenters are shown by red dots, and stations are shown by red triangles. The yellow line shows the boundary of China. Active faults in the China area are shown by purple lines, and tectonic sutures are shown by blue lines. SoB, Songliao Basin; OB, Ordos Basin; SB, Sichuan Basin; CB, Changbai; HN, Hainan; LMS, Longmenshan; KB, Khorat Basin; STB, Shan Thai Block. opportunity to determine a detailed 3-D crustal structure under the region. Using these data, Sun and Tokso z [26] derived a 3-D P wave velocity structure and Sun et al. [28] presented a 3-D shear velocity model for the crust and uppermost mantle. Using the adaptive moving window method [Sun et al., 2], Sun et al. [28] obtained the 3-D shear velocity model by combining and smoothing the 1-D layered models at about 2 selected locations. Each 1-D profile was determined independently. The Moho depths and Sn velocities were constrained using the first arrivals within a window centered at each selected location. Sun et al. s [28] 3-D S wave velocity model, however, was not constructed using a full 3-D approach. Both vertical and horizontal resolutions need to be improved. In the present work, we use Sun et al. s [28] S wave velocity model as a starting model and use the tomographic constraint from the P wave study along with the S wave traveltimes to determine a full 3-D shear wave velocity structure of the crust and upper mantle under this region. 2. Data and Method [6] For this work, we selected shear wave arrival time data from the Annual Bulletin of Chinese Earthquakes (ABCE) (Institute of Geophysics, China Seismological Bureau, ). In this database there are 33, earthquakes, 22 stations, and 8, S wave raypaths in China and the surrounding areas in the January 199 to December 26 time period. Figure 2a shows earthquake epicenters and stations in China. The selection of earthquakes is based on the following criteria: (1) events that occurred in the study area with magnitudes greater than M 3. (except for the northwest area with less seismicity, where we included earthquakes with M ); (2) all selected events were recorded by at least 15 stations; and (3) earthquakes were selected to provide a uniform distribution of hypocenters in the study area [Sun and Tokso z, 26]. With these criteria, 28, S wave arrival times from 13,15 earthquakes were selected to determine the S wave velocity structure. The reading accuracy of S wave arrival times is, in general,.1. s (P. Chen, personal communication, 27). Figure 2b shows the epicentral distribution of the 13,15 selected earthquakes. [7] To determine the shear velocity structure from arrival time data, we used the tomographic method of Zhao et al. [1992]. This method has the following features. [8] 1. It accommodates discontinuities such as the Moho. A 3-D grid is set up in the model to express the 3-D structure (Figure 3), and velocity perturbations at the grid nodes are the unknown parameters. The velocity perturbation at any point in the model is calculated by linearly interpolating the velocity perturbation at the eight grid nodes surrounding that point. [9] 2. It calculates traveltimes and raypaths using an efficient 3-D ray-tracing technique [Zhao et al., 1992], and it iteratively uses the pseudobending technique of Um and Thurber [1987]. Station elevations are also taken into account in the ray tracing. [1] 3. It uses a LSQR algorithm [Paige and Saunders, 1982] with a damping regularization to solve the large and sparse system of equations inherent in the tomographic problem. The LSQR algorithm has been used by a number of researchers [Nolet, 1985; Spakman and Nolet, 1988; Papazachos and Nolet, 1997] and is an efficient algorithm for solving problems with large sparse systems. 2 of 1

3 Urumqi Depth (km) 12 8 Lhasa Chengdu Guangzhou Beijing Figure 3. Three-dimensional configuration of the grid adopted in the present study. traveltimes. Finally, we obtained the 1-D model shown in Figure, which gave the best fit to the observed data. We used this averaged 1-D model as the reference velocity model for our tomographic inversions. We put into the starting model the Moho depth obtained from the P wave tomography, which is consistent with the Moho depths in this region obtained by different researchers [Mooney et al., 1998; Hearn et al., 2; Sun et al., 2; Sun and Toksöz, 26; Sun et al., 28]. In the study area, the Moho depth ranges from 3 km in the east to 78 km. During the Figure 2. (a) The 33, earthquakes, 22 stations, and 8, S wave raypaths in China and the surrounding area. Earthquake epicenters are shown by red circles, and stations are shown by red triangles. The green line shows the boundary of China. (b) Distribution of seismic stations and epicenters of selected earthquakes used in this study. Totally, 13,15 events from M 3. and above are selected and plotted by red circles. All of the 22 stations are used and represented by red triangles. [11]. The nonlinear tomographic problem is solved by conducting linear inversions iteratively. [12] For the shear wave tomography we follow a procedure very similar to the one used for the P wave tomography [Sun and Toksöz, 26]. We adopt a grid spacing of 1 1 in the horizontal direction and 1 km in depth. We also set grids at the depths of 1, 2, 5, and 7 km to sample the sediment layer. We use the SEAPS shear wave velocity model [Sun et al., 28] as a starting 3-D model. The SEAPS models were obtained from inverting and combining 2 1-D P wave [Sun et al., 2] and S wave velocity profiles in China [Sun et al., 28]. The input and output 3-D models are represented by velocity perturbations relative to an averaged 1-D reference model in the study area. We inverted the arrival time data in our selected data set for a 1-D velocity model representing the whole study area by minimizing the root-mean-square (RMS) error of the Figure. The averaged 1-D velocity model used for the tomographic inversion. 3of1

4 Figure 5. Trade-off curve for the variance of the velocity perturbations and root-mean-square traveltime residuals. Numbers beside the black dots denote the damping parameters adopted for the inversions. The largest black dot denotes the optimal damping parameter for the final tomographic model. tomographic inversion, the Moho geometry is fixed and only the velocities at grid nodes are determined. [13] The earthquake hypocentral parameters are those determined by P wave tomography [Sun and Toksöz, 26]. They are kept constant during S wave inversion. We chose the value of 25. for the damping parameters to balance between the reduction of traveltime residuals and the smoothness of the 3-D velocity model obtained (Figure 5). The RMS traveltime residual was reduced by 35% from.98 to.72 s by the inversion (Figure 6). An average of 7% RMS reduction has been observed, compared to the average 1-D velocity model in the study area. Spatial resolution was evaluated using a checkerboard test [Spakman and Nolet, 1988; Zhao et al., 1992]. To make a checkerboard, positive and negative 3% velocity perturbations are assigned to 3-D grid nodes that are arranged in an alternating pattern in the model space. The traveltimes are calculated and inverted as if they were data. Examining the image of the synthetic inversion of the checkerboard, one can observe where the resolution is good and where it is poor. Random errors in a normal distribution with a standard deviation of.3 s are added to the theoretical traveltimes calculated for the synthetic models. Figure 7 shows the checker board test results. The resolution is quite good for most of China at 2,, and 6 km depths. Resolution is poorer below 8 km. Resolution deteriorates toward the edge of the model space, where ray density is lower. 3. Results [1] The 3-D shear wave velocity models are shown in Figures 8, 9, 1, 11, and 12. Figure 8 shows the Moho depth map taken from P wave tomography and used as an input to S wave tomography. On the right of Figure 8 are the Sn velocities. Depth slices of shear velocities are shown in Figure 9. N-S and E-W cross sections (shown in Figure 1) of shear velocities are shown in Figures 11 and 12, respectively. [15] Strong S wave velocity lateral variations of more than 6% are found in the crust in the study area, indicating the existence of significant structural heterogeneities in the crust and upper mantle in this region. At a depth of 1 km, the North China Block (also known as the Sino-Korea craton), the South China Block, the Sichuan Basin, the Ordos Basin, the Tarim Basin, and the Tianshan region exhibit distinct velocity patterns (Figure 9). In the North China Block, a low-velocity (low-v) anomaly is found beneath the northern Songliao Basin and the Bohai Gulf, and a high-velocity (high-v) anomaly is imaged beneath the region to the east of the Bohai Gulf. In the South China Block, a high-v anomaly is found in the south, while low-v zones are imaged in the north. The Tianshan region and the Tarim Basin exhibit low-v patterns; high-v zones exist in the Sichuan Basin and the Ordos Basin. The velocity images of shallow layers correlate well with the surface geology and topography. [16] Low-V anomaly beneath the N-S seismic zone (98 16 E, 22 N) is shown at depths of 2, 3, and km, indicating that the boundary between Tibet and eastern China is characterized by low-velocity perturbation [Liu et al., 25]. At the depth of km, low-v zones are visible in the western part of China and high-v zones exist in the eastern part with a clear dividing line appearing in the N-S seismic zone. At a depth of 6 km, the Tibetan plateau is clearly outlined as a low-v zone, suggesting a thickened lower crust of the Tibetan plateau [Liu et al., 25]. The high-velocity anomaly in the Sichuan Basin and adjacent region indicates that this region is tectonically stable [Liang et al., 2]. At a depth of 8 km, the root of the Tibetan plateau disappears and high-v zones appear in the western and southern parts of the Tibetan plateau. There are a few low-v zones sandwiched between the high-v zones. The two low-v zones, (98 12 E, N) and (98 13 E, 3 33 N), exist in the uppermost mantle. A local tomographic study by Liu et al. [2, 25] suggests that these zones continue to a larger depth of 12 km and are connected with the mantle upwelling from a depth of km or deeper. The deeper velocity slices (6 and 8 km) show Number of Rays Travel Time Residual (s) Figure 6. Number of rays versus traveltime residuals. The red line denotes the results for the averaged 1-D model in the study area. The green line denotes the result for the starting model (before the inversion). The blue line denotes the result after the inversion. of1

5 Figure 7. The input checkerboard when the grid shown in Figure 3 is adopted. The depth of the layer is shown at the lower left corner of each map. Blue and red squares denote high and low velocities, respectively. The velocity perturbation scale is shown at the bottom. 5of1

6 km/s Figure 8. Depth distribution of (left) the Moho discontinuities and (right) Sn velocities in the present study area. The Moho depths are shown in contours. Here 1, Tarim Basin; 2, Ordos Basin; 3, Songliao Basin;, Sichuan Basin; 5, Shan Thai Block; 6, Khorat Basin. scattered low-v zones in eastern China, indicating the presence of tectonic extension in the area [Sun et al., 28]. [17] In an earlier study [Sun and Toksöz, 26], 3-D P wave velocity models were obtained through a tomographic inversion similar to those used in this study for S waves. It is informative to compare the results. Figure 13 shows N-S cross sections of P and S wave velocities. There is good correlation between the P and S wave velocities under Tibet. Lower velocities in the midcrust appear in both the P and S tomograms. The low-velocity zone is more prominent in the S wave velocities.. Discussion and Conclusions [18] It has been found that large crustal earthquakes (M , depth 2 km) occurred in or around low-v zones or at the boundary between low-v and high-v zones from 1885 to 1999 in Japan [Zhao et al., 2]. In the China region we also found that large fault zones, including the Longmenshan Fault zone and most of the large crustal earthquakes such as the M 7.9 Wenchuan earthquake, are located in the boundary areas between high-v and low-v zones in the crust. Some of them occurred above the low-v zones of the lower crust to the uppermost mantle. The Wenchuan earthquake (M 7.9), which occurred at (13. E, 3.5 N) in Sichuan Province at a depth of about 1 km on 12 May 28, is located on the Longmenshan Fault zone between the low-v eastern Tibetan plateau and the high-v Sichuan Basin. The low-v zones feature low electric resistivity, negative gravity anomaly, and high heat flow [Huang et al., 22]. Therefore, they may indicate high temperature anomalies or Figure 9. S wave velocity image at each depth slice. The depth of each layer is shown at the lower left corner of each map. 6of1

7 N M L K J I H G A B C D E F Figure 1. Locations of the vertical cross sections shown in Figures 11 and 12. magma chambers or fluid reservoirs, which may cause weakening of the seismogenic layer in the upper crust [Zhao et al., 1996, 2]. [19] Low-V zones also exist beneath the rifts and the volcanic sites such as the Tengchong, Changbai, and Hainan areas shown in Figure 1. The most recent eruption in Tengchong occurred in 169 [Qin et al., 1996]. Many hot springs exist in this region. This volcanic area is situated between N latitude and E longitude and covers a rectangular area of 9 km km [Huang et al., 22]. Huang et al. [22] show that a larger portion of the low-v zone exists under the southern part of the volcanic center than under its northern part. Similar patterns are found in our study in Figure 1 (cross section A). The Changbai area is another active volcanic region in China. A low-v zone beneath Changbai found by Lei and Zhao [25] extended to the depth of 1 km. Our tomographic study also shows a large low-v zone beneath Changbai from the surface to the uppermost mantle (Figure 1, cross section C). Evidence of the presence of the Hainan plume has been shown by Lebedev and Nolet [23]. Our study shows a low-v zone beneath Hainan extending from the crust to the mantle (Figure 1, cross section B). At the Moho beneath Hainan, however, the low-v zone is not as clear as in the crust or mantle. [2] The Sn velocities obtained by 3-D tomography and shown in Figure 8 are similar to the Sn model developed by Pei et al. [27] and Sun et al. [28]. The prominent low Sn zones in eastern China exist consistently in both Sn models. Southern and eastern Tibet clearly show high Sn anomaly. We also observe high Sn anomaly beneath the Sichuan Basin, the Ordos Basin, and the Tianshan area. The discrepancies between the two Sn models are minor. Some parts of the Kunlun area show high Sn in our results but not in Pei et al. s model. This discrepancy might be caused by the overaveraging effect of our initial 1-D inversion. The high Sn in eastern Tibet was carried over to the western part because of the sparse ray coverage in the western part; thus, very large windows were selected to provide averaged 1-D profiles in the Monte Carlo inversion. Figure 11. Vertical cross sections of S wave velocity when the Moho depth variations (Figure 8) are taken into account. The surface topography along each profile is shown on the top of each cross section. The black curved lines show the Conrad (dashed) and Moho (solid) discontinuities. Each grid in the region between the two white lines has a raypath hit count of 2 or above. 7of1

8 Figure 12. Similar to Figure 11. Cross sections G through N are plotted. Figure 13. (left) P and (right) S vertical velocity profiles of the crust and uppermost mantle. The selected profile locations are shown in Figure 1. 8of1

9 [21] Another important comparison is between Pn and Sn velocities. Figure 15 shows Pn and Sn velocities. These are very similar. All major spatial features are well correlated. The percentage variation of Sn velocities is greater than that of Pn. [22] The 3-D shear wave velocity models of the crust and upper mantle, obtained by S wave traveltime tomography, reveal pronounced lateral heterogeneities under China and surrounding regions. Prominent features in the study area are: (1) the low velocities in the thick crust under Tibet; (2) low velocities in the crust and uppermost mantle beneath volcanic sites and rift zones; (3) high Sn velocities in southern and eastern Tibet; and () low Sn velocities in km/s km/s 3.6 Urumqi C Depth Depth Depth Lhasa A A Tengchong 8 Chengdu Beijing B Guangzhou Tibetan Plateau S. China B Sea 8 Hainan C Changbai Heilongjiang Latitude Figure 1. Three vertical velocity profiles of the crust and uppermost mantle. (top) The locations of the vertical cross sections A, B, and C. The recent volcanic activities in China are shown by red dots. % % % Figure 15. (top) Pn and (bottom) Sn velocities in the present study area. Here 1, Tarim Basin; 2, Ordos Basin; 3, Songliao Basin;, Sichuan Basin; 5, Shan Thai Block; 6, Khorat Basin. the rift zone located in northeastern China. There is good correlation between the P and S velocities under Tibet. Lower velocities in the midcrust appear both in the P and S tomograms. The low-velocity zone is more prominent in the S velocities. [23] Acknowledgments. This work was partially supported by the Defense Threat Reduction Agency under contract DTRA1--C-2 and by Air Force Research Laboratory under contract FA8718--C-18. We thank John Chen from Peking University, Beijing, China, for his suggestions and comments. Michael Fehler, Mark Willis, and Haijiang Zhang from MIT-ERL provided great help on improving the manuscript. The constructive reviews from Jeffrey Stevens, Xiaodong Song, and the associate editor have provided valuable comments on the manuscript. All the figures in this work were made using GMT. References Bassin, C., G. Laske, and G. Masters (2), The current limits of resolution for surface wave tomography in North America, Eos Trans. AGU, 81(8), Fall Meet. Suppl., Abstract S12A-3. Chang, X., Y. Liu, J. He, and H. Sun (27), Lower velocities beneath the Taihang Mountains, northeastern China, Bull. Seismol. Soc. Am., 97, , doi:1.1785/ Friederich, W. (23), The S-velocity structure of the East Asian mantle from inversion of shear and surface waveforms, Geophys. J. Int., 153, 88 12, doi:1.16/j x x. Hearn, T. M., and J. F. Ni (21), Tomography and location problems in China using regional travel-time data, in Proceedings of the 23rd Seismic Research Review: Worldwide Monitoring of Nuclear Explosions, Rep. LA-UR-1-5, pp. 37 5, Los Alamos Natl. Lab., Los Alamos, N. M. (Available at researchreview/index.cgi?year=21) Hearn, T. M., S. Wang, J. F. Ni, Z. Xu, Y. Yu, and X. Zhang (2), Uppermost mantle velocities beneath China and surrounding regions, J. Geophys. Res., 19, B1131, doi:1.129/23jb of1

10 Huang, J., D. Zhao, and S. Zheng (22), Lithospheric structure and its relationship to seismic and volcanic activity in southwest China, J. Geophys. Res., 17(B1), 2255, doi:1.129/2jb137. Huang, Z., W. Su, Y. Peng, Y. Zheng, and H. Li (23), Rayleigh wave tomography of China and adjacent regions, J. Geophys. Res., 18(B2), 273, doi:1.129/21jb1696. Jia, S., C. Qi, F. Wang, Q. Chen, X. Zhang, and Y. Chen (25), Threedimensional crustal gridded structure of the capital area, Chin. J. Geophys., 8, Koulakov, I., and S. V. Sobolev (26), A tomographic image of Indian lithosphere break-off beneath the Pamir-Hindukush region, Geophys. J. Int., 16, 25, doi:1.1111/j x x. Laske, G., and G. Masters (1997), A global digital map of sediments thickness, Eos Trans. AGU, 78(6), F83, Fall Meet. Suppl.. Lebedev, S., and G. Nolet (23), Upper mantle beneath Southeast Asia from S velocity tomography, J. Geophys. Res., 18(B1), 28, doi:1.129/2jb73. Lei, J., and D. Zhao (25), P wave tomography and origin of the Changbai intraplate volcano in northeast Asia, Tectonophysics, 397, , doi:1.116/j.tecto Lei, J., and D. Zhao (27), Teleseismic P wave tomography and the upper mantle structure of the central Tien Shan orogenic belt, Phys. Earth Planet. Inter., 162, , doi:1.116/j.pepi Liang, C., and X. Song (26), A low velocity belt beneath northern and eastern Tibetan plateau from Pn tomography, Geophys. Res. Lett., 33, L2236, doi:1.129/26gl Liang, C., X. Song, and J. Huang (2), Tomographic inversion of Pn travel times in China, J. Geophys. Res., 19, B113, doi:1.129/ 23JB2789. Liu, F., J. Liu, and D. Zhong (2), The subducted slab of Yangtze continental block beneath the Tethyan orogen in western Yunnan, Chin. Sci. Bull., 5, 66 69, doi:1.17/bf Liu, Y., C. Xu, J. He, F. Liu, and H. Sun (25), Three-dimensional velocity images of the crust and upper mantle beneath the north-south zone in China, Bull. Seismol. Soc. Am., 95, , doi:1.1785/ Mooney, W. D., G. Laske, and T. G. Masters (1998), CRUST 5.1: A global crustal model at 5 5, J. Geophys. Res., 13, , doi:1.129/ 97JB2122. Nolet, G. (1985), Solving and resolving inadequate and noisy tomographic systems, J. Comput. Phys., 61, 63 82, doi:1.116/ (85)975-. Paige, C., and M. Saunders (1982), LSQR: An algorithm for sparse linear equations and sparse least squares, Trans. Math. Software, 8, Papazachos, C., and G. Nolet (1997), P and S deep velocity structure of the Hellenic area obtained by robust nonlinear inversion of travel times, J. Geophys. Res., 12, , doi:1.129/96jb373. Pei, S., Z. Xu, and S. Wang (2), Discussion on origin of Pn velocity variation in China and adjacent region (in Chinese), Acta Seismol. Sin., 17, Pei, S., J. Zhao, Y. Sun, Z. Xu, S. Wang, H. Liu, C. A. Rowe, M. N. Toksöz, and X. Gao (27), Upper mantle seismic velocities and anisotropy in China determined through Pn and Sn tomography, J. Geophys. Res., 112, B5312, doi:1.129/26jb9. Phillips, W. S., M. L. Begnaud, C. A. Rowe, L. K. Steck, S. C. Myers, M. E. Pasyanos, and S. Ballard (27), Accounting for lateral variations of the upper mantle gradient in Pn tomography studies, Geophys. Res. Lett., 3, L1312, doi:1.129/27gl Poupinet, G., et al. (22), Intracontinental subduction and Paleozoic inheritance of the lithosphere suggested by a teleseismic experiment across the Chinese Tien Shan, Terra Nova, 1, 18 2, doi:1.16/ j x. Qin, J., X. Qian, and G. Huangpu (1996), The seismicity feature of the volcanic area in Tengchong (in Chinese), Seismol. Geomagn. Obs. Res., 17, Ritzwoller, M. H., M. P. Barmin, A. Villasenor, A. L. Levshin, and E. R. Engdahl (22), Pn and Sn tomography across Eurasia to improve regional seismic event locations, Tectonophysics, 358, 39 55, doi:1.116/ S-1951(2)16-X. Roecker, S. W., T. M. Sabitova, L. P. Vinnik, Y. A. Burmakov, M. I. Golvanov, R. Mamatkanova, and L. Munirova (1993), Three-dimensional elastic wave velocity structure of the western and central Tien Shan, J. Geophys. Res., 98, 15,779 15,795, doi:1.129/93jb156. Shapiro, N. M., and M. H. Ritzwoller (22), Monte Carlo inversion for a global shear velocity model of the crust and upper mantle, Geophys. J. Int., 151, 88 15, doi:1.16/j x x. Song, Z. H., C. Q. An, G. Y. Chen, L. H. Chen, Z. Zhuang, Z. W. Fu, and J. F. Hu (1991), Study on 3-D velocity structure and anisotropy beneath West China from the Love wave dispersion (in Chinese), Acta Geophys. Sin., 3, Spakman, W., and G. Nolet (1988), Imaging algorithms, accuracy and resolution in delay time tomography, in Mathematical Geophysics, edited by N. Vlaar et al., pp , D. Reidel, Norwell, Mass. Stevens, J. L., D. A. Adams, and G. E. Baker (21), Improved surface wave detection and measurement using phase-matched filtering with a global one-degree dispersion model, in Proceedings of the 23rd Seismic Research Review: Worldwide Monitoring of Nuclear Explosions, Rep. LA-UR-1-5, pp. 2 3, Los Alamos Natl. Lab., Los Alamos, N. M. (Available at researchreview/index.cgi?year=21) Sun, Y., and M. N. Toksöz (26), Crustal structure of China and surrounding regions from P wave traveltime tomography, J. Geophys. Res., 111, B331, doi:1.129/25jb3962. Sun, Y., X. Li, S. Kuleli, F. D. Morgan, and M. N. Toksöz (2), Adaptive moving window method for 3-D P-velocity tomography and its application in China, Bull. Seismol. Soc. Am., 9, 7 76, doi:1.1785/ Sun, Y., M. N. Toksöz, S. Pei, and F. D. Morgan (28), The layered shear wave velocity structure of the crust and uppermost mantle in China, Bull. Seismol. Soc. Am., 98, , doi:1.1785/ Um, J., and C. Thurber (1987), A fast algorithm for two-point seismic ray tracing, Bull. Seismol. Soc. Am., 77, Vinnik,L.P.,S.W.Roecker,G.L.Kosarev,S.I.Oreshin,andI.Y. Koulakov (22), Crustal structure and dynamics of the Tien Shan, Geophys. Res. Lett., 29(22), 27, doi:1.129/22gl Vinnik, L. P., C. Reigber, I. M. Aleshin, G. L. Koserev, M. K. Kaban, S. I. Oreshin, and S. W. Roecker (2), Receiver function tomography of the central Tien Shan, Earth Planet. Sci. Lett., 225, , doi:1.116/ j.epsl Wu, F. T., A. L. Levshin, and V. M. Kozhevnikov (1997), Rayleigh wave group velocity tomography of Siberia, China, and vicinity, Pure Appl. Geophys., 19, 7 73, doi:1.17/s235. Xu, Y., F. Liu, J. Liu, and H. Chen (22), Crust and upper mantle structure beneath western China from P wave travel time tomography, J. Geophys. Res., 17(B1), 222, doi:1.129/21jb2. Yu, X., Y. Chen, and P. Wang (23), Three-dimensional P velocity structure in Beijing area, Acta Seismol. Sin., 16, 1 15, doi:1.17/ s Zhao, D., A. Hasegawa, and S. Horiuchi (1992), Tomographic imaging of P and S wave velocity structure beneath northeastern Japan, J. Geophys. Res., 97, 19,99 19,928, doi:1.129/92jb63. Zhao, D., F. Ochi, A. Hasegawa, and A. Yamamoto (2), Evidence for the location and cause of large crustal earthquakes in Japan, J. Geophys. Res., 15, 13,579 13,59, doi:1.129/2jb926. Zhao, Z., X. Han, and R. Zhou (1996), Geothermal structure of lithosphere in Tengchong region, Yunnan Province (in Chinese), Seismol. Geomagn. Obs. Res., 17, 3 1. Zhu, J., J. Cao, X. Cai, Z. Yan, and X. Cao (22), High resolution surface wave tomography in East Asia and West Pacific marginal sea, Chin. J. Geophys., Engl. Transl., 5, Zhu, L., R. Zeng, and F. Liu (199), 3-D P wave velocity structure under the Beijing network area (in Chinese), Acta Geophys. Sin., 33, F. D. Morgan, Y. Sun, and M. N. Toksöz, Earth Resources Laboratory, Department of Earth, Atmospheric, and Planetary Sciences, Massachusetts Institute of Technology, Cambridge, MA 2139, USA. (youshun@mit.edu) S. Pei, Institute of Tibetan Plateau Research, Chinese Academy of Sciences, No. 18 Shuangqing Road, Haidian District, Beijing 185, China. A. Rosca, Chevron-Texaco, 61 Bollinger Canyon Road, San Ramon, CA 9583, USA. D. Zhao, Department of Geophysics, Tohoku University, Sendai , Japan. 1 of 1