Earthquake Reaearch in China  2018, Vol. 32 Issue (1): 28-39
Tomography for Q of the Eastern Section of the Tianshan Area from High-frequency Attenuation of S-wave
Li Jin1, Zhou Longquan2, Wang Huilin3, Xiang Yuan1
1. Earthquake Agency of Xinjiang Uygur Autonomous Region, Urumqi 830011, China;
2. China Earthquake Networks Center, Beijing 100045, China;
3. Hainan Earthquake Agency, Haikou 570203, China
Abstract: Based on the waveform data of 5, 076 local earthquakes recorded at 25 stations in Xinjiang during the period from 2009 to 2014 and the observation reports provided by the Xinjiang Digital Seismic Network, a data set of 19, 140 attenuation factors t* is obtained by fitting the high-frequency attenuation of S-wave spectra with a genetic algorithm. The spatial distribution of QS is determined by inverting the t* data with seismic tomography. The results show that the average Q0 in eastern Tianshan is 520, and there is a significant correlation between the QS value distribution or attenuation characteristics it disclosed and the surface structure of the study area. The QS value is lower in the intersection area of the mountain basin which is located on the north and south sides of the Tianshan Mountains, and the high QS distribution is more concentrated inside the Tianshan orogenic belt. The M ≥ 6.0 earthquakes have been basically located in the Low-QS region since 1900. 24 high heat flow points in eastern Tianshan are located at the north and south of Tianshan Mountains where low QS exists, indicating a negative correlation. In addition, there is a positive correlation between the velocity structure and the attenuation structure in the study area, which reflects the consistency of the 2-D attenuation structure with the velocity structure and the two-dimensional density structure.
Key words: Eastern section of the Tianshan area     High-frequency attenuation of S-wave     Attenuation operator t*     Q tomography

INTRODUCTION

The Q value, reflecting the inelastic properties of the propagation medium, is a sensitive index for the crystal structure variation caused by temperature and phase change, which has important application value in understanding the inelastic properties of the Earth's interior medium and deducing its thermodynamic state (Hong Xuehai, 2003). According to the study by Chen Yong et al. (2009), seismic wave attenuation mainly depends on the microscopic nature of the rock, that is, the interaction between the density, distribution and structure of inner cracks of the rock and pore fluid. A physical mechanism study of seismic wave attenuation by Liu Jianhua et al. (2004a) finds that the main cause of seismic wave attenuation in the crust is that there are plentiful fractures in the Earth's crust, which are full of water, or partially contain water, and seismic wave propagation causes fluid motion in fractures, resulting in seismic wave attenuation. When a seismic wave passes through the active tectonic zone, the energy will be strongly attenuated and shown as a low Q value, while in some stable areas, attenuation is weak and the Q value is high. Therefore, with the study of the Q value of the crust, the characteristics, structure and change of the earth's interior medium can be understood more accurately.

At present, attenuation in the medium is studied by mainly using long-period surface waves and Lg coda waves, and a large number of research results have been obtained. Anderson D.L. et al. (1965) studied the average Q structure of the Earth by using long-period surface wave. Sato Y. (1958) studied the amplitude attenuation of surface wave using the Fourier transform. Solomon S. C. (1972) used the same method to determine the crust-mantle Q value between two network stations. Tsai Y.B. et al. (1969) established a method of single source-multiple stations to measure the average surface wave attenuation. Aki K. (1969) first proposed that coda wave of near earthquakes is caused by backscattering of seismic wave from countless discontinuity surfaces of the earth's crust and upper mantle distributed uniformly and randomly in an ellipse, and after that, Suteau A.M. et al. (1979) defined the scattering mechanism. Herrmann R. B. (1980) proposed a formula for measuring the Q value of the coda wave according to Aki's coda wave model and a scattering mechanism put forward by Suteau A.M. et al.(1979), and Xie J. et al.(1988) extended its application scope to the calculation of the Lg coda wave, and came up with the method of stack spectral ratio (SSR) to calculate the Q value; Cong Lianli et al. (2003) explored the distribution of Q of Lg waves in the Chinese mainland and its adjacent areas using the SSR method based on broadband vertical components of 785 shallow-focus earthquakes recorded by 10 seismic stations of the China Digital Seismograph Network and 5 seismic stations of the Global Seismic Network. Liu Jianhua et al.(2004b, 2004c) promoted the SSR method in the measurement of Q of Lg coda wave in small regions with epicentral distances less than 600km, and conducted tomography for the Q of coda waves in North China.

Current research on the Q value of the Tianshan area is focused mainly on the results of large regions or Chinese mainland and the average Q value obtained by joint inversion of multiple seismic stations and earthquakes, research of fine Q distribution has not been carried out. For instance, the results obtained by Wang Suyun et al. (2008) showed low attenuation in the west of Tarim Basin, with high attenuation in the east. Steinshouer D.W. et al. (1999) found that the west of the Tianshan Mountains presented high attenuation, while the eastern end of Tianshan Mountains presented low attenuation. The research results of the Q-value distribution of Lg coda waves in the Chinese mainland and its adjacent areas obtained by Cong Lianli et al. (2003) showed that the Q0 value of the Tarim platform was significantly higher, which was 350-425. Jiang Hui et al. (2007), by joint inversion using data from 52 earthquakes with ML2.5-5.0 in the northern Tianshan Mountains, obtained a Q0 value of the region, which was 344, and Zhao Cuiping et al. (2011) updated the Q0 in the region to 460 in subsequent studies. With the establishment and improvement of the Xinjiang regional digital seismic network, a large amount of seismic waveform data has been accumulated, providing favorable conditions for the study of tomography for Q of the crustal medium. In this article, high-frequency attenuation data fitting of the S-wave displacement spectra is used for Q tomography in the mid-eastern section of the Tianshan area on a resolution of 0.5°×0.5°, and the characteristics of the Q distribution in the region are analyzed.

1 METHOD AND PRINCIPLE

In the frequency domain, the amplitude spectrum of the ith event observed at the jth station can be expressed as

 ${A_{ij}}\left(f \right) = {S_i}\left(f \right){I_j}\left(f \right){R_j}\left(f \right){G_{ij}}\left(f \right){B_{ij}}\left(f \right)$ (1)

where, f denotes frequency, Si(f) source spectrum, Ij(f) instrument response function and Rj(f) the site response, which is used to describe the amplifying effect of the near-surface layer medium on seismic wave motion near seismic stations. Gij(r) denotes the geometric spreading factor along path r, and Bij(f) is attenuation spectrum.

Source spectrum can be described by the long-period spectrum Ω0 and corner frequency fc, and its mathematical expression is as below (Brune J. N., 1970)

 ${S_i}\left(f \right) = {\mathit{\Omega }_0}/\left({1 + \left({{f^2}/f_{\rm{c}}^2} \right)} \right)$ (2)

According to research by Wong Yuklung et al. (2003) and Su Youjin et al. (2006), the geometric spreading factor can be represented by three coherent geometric attenuation functions put forward by Atkinson G.M. et al. (1992), as below,

 ${G_{ij}}\left(r \right) = \left\{ {\begin{array}{*{20}{l}} {{r^{ - {b_1}}}, }&{r \le {R_{01}}}\\ {R_{01}^{ - {b_1}}R_{01}^{ - b2}{r^{ - {b_2}}}, }&{{R_{01}} < r \le {R_{02}}}\\ {R_{01}^{ - {b_1}}R_{01}^{ - b2}R_{02}^{ - b2}R_{02}^{ - b3}{r^{ - {b_3}}}, }&{r > {R_{02}}} \end{array}} \right.$ (3)

In the geometric attenuation model, coefficients b1, b2 and b3 are all considered to be independent of frequency. When rR01, it corresponds to the geometric attenuation of the direct wave. When R01rR02, in the transition zone and within the range of hypocentral distance, reflected waves from discontinuity surfaces and Moho surface in the crust are added to the direct wave; when r>R02, it corresponds to the S-wave attenuation after multiple reflection and refraction.

The attenuation spectrum along the entire ray path can be expressed by the following formula (Sanders C.O., 1993; Sherbaum F., 1990),

 ${B_{ij}}\left(f \right) = \exp \left({ - {\rm{ \mathsf{ π} }}f{t_{ij}}Q_{ij}^{ - 1}} \right) = \exp \left({ - {\rm{ \mathsf{ π} }}ft_{ij}^*} \right)$ (4)

where, tij denotes travel time along ray path, Qij dimensionless quality factor and tij* the operator along the entire ray path (namely, attenuation operator t*).

Instrument response can be directly deducted according to station parameters, and at this point the amplitude spectrum is expressed as Aij(f)=Aij(f)/Ij(f). Moreover, since most regional stations are built on bedrock, its field response can be assumed to be approximately 1. Therefore, equation (1) can be written as,

 $A{'_{ij}}\left(f \right) = {\mathit{\Omega }_0}\frac{{f_c^0}}{{f_c^0 + {f^2}}}G\left(r \right)\exp \left({ - {\rm{ \mathsf{ π} }}ft_{ij}^*} \right)$ (5)

In addition, operator tij* can be written as the projection of 1/(Q(S)v(S)) along the ray path from epicenter i to station j(Cormier V.F., 1982; Wittlinger G.H. et al., 1983), namely,

 $t_{ij}^* = \int {_{{\rm{path}}}} \frac{1}{{Q\left({\rm{S}} \right)\nu \left({\rm{S}} \right)}}{\rm{dS}}$ (6)

where, v denotes shear wave velocity, dS denotes ray path unit.

As shown in equation (5), there are 3 unknown variables in the equation, namely, Ω0, fc and tij*. According to equation (5), path attenuation tij* from the seismic source to each station can be determined by the inversion of displacement spectrum of waveform. Then, according to equation (6), the attenuation tomography is performed with the same method as the travel time tomography to determine Q.

1.1 Pretreatment of Observation Data

Waveforms selected in this study are from 25 digital seismic stations in mid-eastern section of the Tianshan area, and the locations of stations are shown in Fig. 1. Firstly, two horizontal components of waveform records are corrected with band-pass filtering, and then the S-wave window and noise window are taken. The time period (Tse-Tsn) from the start of S-wave to 90% energy of the S-wave is defined as the S-wave window. At near-source distances, S-wave window contains only a direct S-wave, while at greater distances, it incorporates reflected waves from discontinuity surfaces and the Moho surface in the crust, and at a further distance, it contains the seismic phase Sn and Lg. For the same earthquake, because of the difference in epicentral distance or (Sg-Pg), the S-wave window length also varies, and the two are generally proportional to each other. For the mid-eastern section of the Tianshan area, Liu Jianming et al. (2014) gave results obtained after fitting, which were y=0.793x+14.09, in which, y denotes the S-wave window, x denotes (Sg-Pg).

 Fig. 1 Distribution of seismic stations (red triangles), events (+) and rays in mid-eastern section of the Tianshan area

256 sampling points before the first motion of the P-wave are taken as the noise window (Rietbrock A., 2001; Hansen S. et al., 2004). The selection of the S-wave window and noise window is shown in Fig. 2. Fast Fourier transform is performed to deal with the intercepted S-wave window and noise window by the translational window spectrum method to convert the S-wave window waveform record and noise window noise record into the observed amplitude spectrum and noise spectrum respectively. Since all seismic station seismometers are velocimeters, the velocity amplitude spectrum should be converted into the displacement amplitude spectrum. After the above treatment for the two respective S-wave horizontal components, the synthetic displacement spectrum of the S-wave horizontal component is obtained (Su Youjin et al., 2006). In order to acquire a reliable signal-to-noise ratio, a frequency within 1-15Hz is selected and the amplitude spectrum is at least 3 times that of the noise spectrum. Fig. 1(a) shows waveforms of two horizontal components of the MS3.2 earthquake recorded by Urumqi seismic station (WMQ) on June 24, 2011 and the observed displacement spectrum and fitting spectrum at WMQ, STZ, GAZ, KMS and SCH station after geometric diffusion correction.

 Fig. 2 (a) Waveform of two horizontal components of a MS3.2 earthquake recorded by Urumqi station on June 24, 2011 (The left two dotted line segments are the range of noise window, followed by the S-wave window in the two dotted line segments). (b) Observed displacement spectrum and fitting spectrum at different stations after noise correction and geometric diffusion correction.(WMQ: Urumqi station, STZ: Shitizi station, GAZ: Gaoyazi station, KMS: Kumishi station, SCH: Shichang station)

According to formula (5), the principle of multi-station, multi-spectrum data joint inversion method (Zhou Longquan et al., 2009) is to assume that an earthquake is recorded by N stations, and that N+2 unknown variables need to be inverted, namely Ω0, fc, t1*, t2*…, tN*, 2N-2 variables are reduced compared with the single-station observation spectrum inversion method, thereby reducing the non-uniqueness of solutions. Given M frequency points, the value of observed spectrum of the ith station at the jth frequency point can be expressed as Aiobs(fj), and theoretical amplitude spectrum is Aical(fj). Multi-station observation spectrum inversion is the search of the unknown variable value which will generate the minimum residual $\sum\limits_{i = 1}^N {\sum\limits_{j = 1}^M {\lg |A_i^{{\rm{obs}}}\left({{f_j}} \right) - A_i^{{\rm{cal}}}} \left({{f_j}} \right)} |$. In order to find the global optimal solution, the genetic algorithm is used to invert N+2 unknowns, so that the sum of residuals between the multi-station observed spectrum and theoretical spectrum at different frequency points is the smallest (Liu Jie et al., 2003). As revealed by the inversion results of multi-station observation spectrum of the same source after noise and geometric correction, the high-frequency attenuation of displacement spectrum at different stations is affected mainly by the Q value in different paths.

1.2 Two-Dimensional QS Imaging

After obtaining attenuation operator t*, the QS value can be obtained by using the attenuation tomography method. First, the qualified attenuation operator t* is picked out, appropriate grid size is determined according to the average number of rays for resolution test of detecting board, and according to the test results of resolution of detecting board, grid size is repeatedly adjusted, and eventually proper resolution is determined. Then, by calculating the average QS of the crust as the initial input model and adopting the pseudo-bending ray tracing method, 10 iterative computations are carried out to obtain the QS value. To remove the influence of focal depth, focal distance is used to replace epicentral distance in the process of inversion of QS with t*.

2 TOMOGRAPHY FOR Q OF S-WAVE IN THE MID-EASTERN SECTION OF THE TIANSHAN AREA 2.1 Stability Analysis of Data and Solutions

Data used in this article incorporates 5, 076 localizable ML2.0-5.4 seismic events recorded by 25 digital seismic stations in mid-eastern section of the Tianshan area during January 1, 2009-December 31, 2014 (Fig. 1). According to formula (5), a data set of 44, 599 t* is obtained by inversion using the genetic algorithm. Because of large Q error, the data needs to be further screened, and 19, 140 t* values with error less than one time mean-square deviation are selected, and corresponding ray distribution is shown in Fig. 1. Using selected t* data, the crust of mid-eastern section of the Tianshan area (40.5°-45.5°N, 79°-90.5°E) is horizontally divided into a uniform grid of 0.5°×0.5° for QS inversion, and the average number of rays in the grid is 342, which is helpful to reduce the non-uniqueness of solutions.

Before the inversion, the initial average velocity of S-wave in the mid-eastern section of the Tianshan area is 3.406km/s by calculation. Assuming that the initial Q value of the region is uniform, according to equation (6), when Q is uniform, t* has a linear relationship with epicenter distance. The linear least square method is used for linear fitting of t* data (Fig. 3), and the average Q0 of the crust in the mid-eastern section of the Tianshan area is obtained, which is 520, slightly higher than that given by Zhao Cuiping et al.(2005, 2011). The obtained Q0 and initial average velocity are used as initial value for inversion and are input into the model. After 10 iterative inversions, the root mean square (RMS) residual of t* is reduced from 0.0255 before the inversion to 0.0226, and the distribution of residuals with epicentral distance before and after the inversion is shown in Fig. 4.

 Fig. 3 Linear fitting of t* and epicentral distance

 Fig. 4 (a) The distribution of residuals with epicentral distance before the inversion. (b) The distribution of residuals with epicentral distance after the inversion

To determine whether the inversion results of 0.5°×0.5° grid are reliable, the detecting board resolution test is performed (Fig. 5), and the results show that the study area is densely covered with rays, with good resolution effect, only in some marginal areas, due to station and earthquake distribution, the ray coverage is sparse, indicating that the inversion results are reliable in most areas.

 Fig. 5 The resolution test results of grid detecting board
2.2 The Tomographic Images for Q of S-wave

The crust of mid-eastern section of the Tianshan area (40.5°-45.5°N, 79°-90.5°E) is horizontally divided into a uniform grid of 0.5°×0.5°, and after 10 iterative inversions, the image of QS distribution of S-wave in mid-eastern section of the Tianshan area is obtained (6). The results show that QS of the mid-eastern section of the Tianshan area is between 380-790, with an average Q0 value of 520, slightly higher than that given by Zhao Cuiping et al. (2011), which is 460. Fig.(6) displays significant lateral variation of S-wave attenuation in mid-eastern section of the Tianshan area, which manifests the attenuation distribution of seismic energy in the earth's crust, and its size is closely related to the sedimentary layer thickness, tectonic activity intensity and crustal medium properties (Sun Lian et al., 2012). On the whole, the QS distribution and the attenuation variation characteristics it reveals are obviously related to the surface structure of the study area. For instance, the area between the Boluokenu and Beiluntai faults has a relatively higher QS of about 650-750, which happens to be the main area of the Tianshan Mountains. Low QS areas are mainly concentrated in the north and south sides of the Tianshan Mountains, that is, the intersection between the Tianshan Mountains and the Junggar Basin and the Tarim Basin, with QS generally under 550. Physical properties of the medium have changed in these areas, transitioning gradually from hard rocks in mountainous areas to sedimentary layers of basins. Low value areas mainly include the east of Ili River Valley, the north of Jinghe, Wusu, Shihezi, Urumqi, Kumux, Korla, Kuqa and Luntai regions. The relationship between QS and temperature is mainly manifested in the relation between regional QS and distribution of regional terrestrial heat flow. Terrestrial heat flow is the most direct display of thermal state and thermal structure of the Earth's interior on the Earth's surface. According to the Compilation of Heat Flow Data in the China Continental Area (3rd edition) published by Hu Shengbiao et al. (2001), data of 24 heat flow points within the scope of the study area (40.5°-45.5°N, 79°-90.5°E) are given in this paper. It can be seen from Fig.(6) that the QS distribution of the medium quality factor in mid-eastern section of the Tianshan area obtained from inversions has a certain relationship with the distribution of terrestrial heat flow. For instance, high heat flow points in the region are mostly distributed in low QS areas in the south and north sides of the Tianshan Mountains.

 Fig. 6 Distribution of Q of S-wave in mid-eastern section of the Tianshan area Red stars represent a high heat flow point and hollow circles denote the epicenter of M≥6.0 earthquake since 1900
3 CONCLUSIONS AND DISCUSSION

In this article, based on the waveform data of 5, 076 local earthquakes recorded at 25 digital seismic stations in mid-eastern section of the Tianshan area during the period from 2009-2014, tomography for the Q of mid-eastern section of the Tianshan area is performed at a resolution of 0.5°×0.5°, and the Q distribution characteristics in this region are analyzed. The main conclusions are as follows:

(1) QS value shows significant lateral variation in the mid-eastern section of the Tianshan area, and attenuation variation characteristics reflected by the QS value has an obvious correlation with the surface structure of this region, which is mainly manifested as such: ① Regions that happen to be the main area of the Tianshan Mountains have relatively higher QS value, about 650-750. ② QS value is lower in the intersection between the Tianshan Mountains and the Junggar Basin and the Tarim Basin which are located on the north and south sides of the Tianshan Mountains, and the farther from the mountainous areas, the lower the QS is, basically below 550, reflecting the gradual thickness of sedimentary layers during the transition from mountains to plains and basins.

(2) Since 1900, most M≥6.0 earthquakes (Fig. 6) have been located in low QS areas, such as the Boluokenu fault, the Qiulitag fault and Shihezi and Changji on the north slope of the Tianshan Mountains. Moreover, according to the Compilation of Heat Flow Data in the China Continental Area (3rd edition) published by Hu Shengbiao et al. (2001), 24 high heat flow points in mid-eastern section of the Tianshan area (40.5°-45.5°N, 79°-90.5°E) are basically located in low QS areas on the south and north sides of the Tianshan Mountains described above (Fig. 6), that is, the heat flow value is negatively correlated with attenuation value.

(3) Research on the velocity structure in mid-eastern section of the Tianshan area shows that north Tianshan and middle Tianshan area are uplifted regions with high P-wave velocity, and the Turpan Basin, the Kuqa depression and the south edge of the Junggar Basin constitute the piedmont low velocity zone on the north and south sides of the Tianshan Mountains (Xu Yi et al., 1994; Guo Biao et al., 2006; Qian Hui et al., 2011; Wang Zaihua et al., 2008). The research on S-wave velocity shows that there are many distinct S-wave low-velocity layers in the upper and middle crust of the Tianshan Mountains, which are located respectively at the intersection of basins and mountains on both sides of the Tianshan Mountains and the junction of different blocks of the Tianshan Mountains (Li Yu et al., 2007). The phase velocity images of a short period of 10s-20s show obvious low velocity anomaly areas in the Tarim Basin and the Junggar Basin, and the Tianshan Orogenic belt shows high velocity anomalies in the tomographic imaging of periods of 10s-16s, contrary to the low velocity anomalies of sedimentary rocks in the Tarim Basin and the Junggar Basin (Tang Xiaoyong et al., 2011). The research on the velocity structure of the mid-eastern section of the Tianshan area also shows that the velocity structure of this region is positively correlated with its attenuation structure, which proves that the two-dimensional attenuation structure is consistent with the velocity structure and two-dimensional density structure (Burtman V.S. et al., 1993). The primary cause of the lateral inhomogeneity of attenuation in the mid-eastern section of the Tianshan area is the uneven distribution of the medium. (Velocity structure and attenuation structure are two different things).

(4) The QS value inverted in this paper is frequency independent, and the initial average QS is slightly higher than the frequency-related inversion results (Zhao Cuiping et al., 2005, 2011). Although the frequency-independent assumption will affect the size of QS, its distribution will not change. Therefore, it still can be used to analyze the characteristics of physical property distribution of the medium (Zhou Longquan et al., 2009; Wang Huilin et al., 2012; Eberhart-Phillips D. et al., 2002).

ACKNOWLEDGEMENT

We are grateful to research professor Jiang Haikun for his guidance over the years. Assistant research professor Chen Xiangjun and Bao Cuiling, and assistant engineers Zhang Zhibin, Liu Shengmei and Wu Ni'er from the Earthquake Agency of Xinjiang Uygur Autonomous Region provided assistance in data sorting and other aspects. During the writing of this thesis, assistant research prefessor Liu Jianming and Dr.Ji Zhanbo provided much help and beneficial discussions, and to them we hereby express our sincere gratitude.

REFERENCES
 Aki K. Analysis of the seismic coda of local earthquakes as scattered waves[J]. Journal of Geophysical Research, 1969, 74(2): 615–631. DOI:10.1029/JB074i002p00615. Anderson D.L., Ben-Menabem A., Archambeau C.B. Attenuation of seismic energy in the upper mantle[J]. Journal of Geophysical Research, 1965, 70(6): 1441–1448. DOI:10.1029/JZ070i006p01441. Atkinson G.M., Mereu R.F. The shape of ground motion attenuation curves in southeastern Canada[J]. Bulletin of the Seismological Society of America, 1992, 82(5): 2014–2031. Brune J.N. Tectonic stress and the spectra of seismic shear waves from earthquakes[J]. Journal of Geophysical Research, 1970, 75(26): 4997–5009. DOI:10.1029/JB075i026p04997. Burtman V.S., Molnar P.. Geological and Geophysical Evidence for Deep Subduction of Continental Crust Beneath the Pamir[M]. Boulder, Colo.: Geological Society of America, 1993: 248-251. Chen Yong, Huang Tingfang, Liu Enru. Rock Physics[M]. Hefei: University of Science & Technology China Press, 2009 Cong Lianli, Hu Jiafu, Fu Zhuwu, Wen Yibo, Kang Guofa, Wu Xiaoping. Distribution of Lg coda Q in the Chinese continent and its adjacent regions[J]. Science in China (Series D), 2003, 46(6): 529–539. Cormier V.F. The effect of attenuation on seismic body waves[J]. Bulletin of the Seismological Society of America, 1982, 72(6): S169–S200. Eberhart-Phillips D., Chadwick M. Three-dimensional attenuation model of the shallow Hikurangi subduction zone in the Raukumara Peninsula, New Zealand[J]. Journal of Geophysical Research, 2002, 107(B2): ESE 3-1–ESE 3-15. Guo Biao, Liu Qiyuan, Chen Jiuhui, Zhao Dapeng, Li Shuncheng, Lai Yuangen. Seismic tomography of the crust and upper mantle structure underneath the Chinese Tianshan[J]. Chinese Journal of Geophysics, 2006, 49(6): 1693–1700. DOI:10.1002/cjg2.v49.6. Hansen S., Thurber C., Mandernach M., Haslinger F., Doran C. Seismic velocity and attenuation structure of the east rift zone and south Flank of Kilauea volcano, Hawaii[J]. Bulletin of the Seismological Society of America, 2004, 94(4): 1430–1440. DOI:10.1785/012003154. Herrmann R.B. Q estimates using the coda of local earthquakes[J]. Bulletin of the Seismological Society of America, 1980, 70(2): 447–468. Hong Xuehai, Zhu Jieshou, Cao Jiamin, Xu Zhuoqun. Tomography of the 3-D S-wave quality factor of the crust and upper mantle in China[J]. Chinese Journal of Geophysics, 2003, 46(5): 642–651. Hu Shengbiao, He Lijuan, Wang Jiyang. Compilation of heat flow data in the china continental area (3rd edition)[J]. Chinese Journal of Geophysics, 2001, 44(5): 611–626. Jiang Hui, Gao Mengtan, Yu Yanxiang, Tang Lihua. S wave inelastic attenuation and site effect in the northern Tianshan Area[J]. Acta Seismologica Sinica, 2007, 29(2): 181–186. Li Yu, Liu Qiyuan, Chen Jiuhui, Li Shuncheng, Guo Biao, Lai Yuangen. Shear wave velocity structure of the crust and upper mantle underneath the Tianshan orogenic belt[J]. Science in China (Series D: Earth Sciences), 2007, 50(3): 321–330. DOI:10.1007/s11430-007-0008-x. Liu Jianhua, Xu Yi, Hao Tianyao. Study on physical mechanism of the seismic wave attenuation[J]. Progress in Geophysics, 2004a, 19(1): 1–7. Liu Jianhua, Liu Futian, Yan Xiaowei, Xu Yi, Hao Tianyao. A study of Lg coda attenuation beneath North China: seismic imaging of Lg coda Q0[J][J]. Chinese Journal of Geophysics, 2004b, 47(6): 1044–1052. Liu Jianhua, Liu Futian, Yan Xiaowei, Xu Yi, Hao Tianyao. A study of Lg coda attenuation in North China — the measurement of Lg coda Q[J]. Chinese Journal of Geophysics, 2004c, 47(5): 822–831. Liu Jianming, Li Zhihai. Non-elastic attenuation, site response and source parameters in the northern Tianshan Mountain, Xinjiang of China[J]. Earthquake, 2014, 34(1): 77–86. Liu Jie, Zheng Sihua, Wong Yuk Lung. The inversion of non-elasticity coefficient, source parameters, site response using genetic algorithms[J]. Acta Seismologica Sinica, 2003, 25(3): 211–218. Qian Hui, Jiang Mei, Xiao Wenjiao, Zhao Dapeng, Wang Yu, Zhang Lishu, Zhao Lei. Seismic tomography of Tianshan-Junggar region and its lithospheric structure[J]. Acta Seismologica Sinica, 2011, 33(3): 327–341. Rietbrock A. P. wave attenuation structure in the fault area of the 1995 Kobe earthquake[J]. Journal of Geophysical Research, 2001, 106(B3): 4141–4154. DOI:10.1029/2000JB900234. Sanders C. O. Local Earthquake Tomography: Attenuation Theory and Results[M]. In: Iyer H., Hirahara K. (Editors). Seismic Tomography: Theory and Practice. New York: Chapman and Hall, 1993. 676-694. Sato Y. Attenuation, dispersion, and the wave guide of the G wave[J]. Bulletin of the Seismological Society of America, 1958, 48: 231–251. Sherbaum F. Combined inversion for the three-dimensional Q structure and source parameters using microearthquake spectra[J]. Journal of Geophysical Research, 1990, 95(B8): 12423–12438. DOI:10.1029/JB095iB08p12423. Solomon S.C. Seismic-wave attenuation and partial melting in the upper mantle of North America[J]. Journal of Geophysical Research, 1972, 77(8): 1483–1502. DOI:10.1029/JB077i008p01483. Steinshouer D. W., Qiang Jin, McCabe P. J., Ryder R. T. Maps Showing Geology, Oil and Gas Fields and Geologic Provinces of the Asia Pacific Region[R]. U. S. Geological Survey Open-file Report 97-470F, 1999. http://www.mendeley.com/catalog/maps-showing-geology-oil-gas-fields-geologic-provinces-asia-pacific-region/ Su Youjin, Liu Jie, Zheng Sihua, Liu Lifang, Fu Hong, Xu Yan. Q value of anelastic S-wave attenuation in Yunnan region[J]. Acta Seismologica Sinica, 2006, 28(2): 206–212. Sun Lian, Li Yonghua, Wu Qingju, Wang Suyun. Q0 tomography of S wave attenuation in Northeast China and adjacent regions[J]. Chinese Journal of Geophysics, 2012, 55(4): 1179–1185. Suteau A.M., Whitcomb J.H. A local earthquake coda magnitude and its relation to duration, moment M0, and local Richter magnitude ML[J]. Bulletin of the Seismological Society of America, 1979, 69(2): 353–368. Tang Xiaoyong, Fan Wenyuan, Feng Yongge, Tang Youcai, Chen Yongshun, Zhu Lixia. Phase velocity tomography of Rayleigh wave in Xinjiang from ambient noise[J]. Chinese Journal of Geophysics, 2011, 54(8): 2042–2049. Tsai Y.B., Aki K. Simultaneous determination of the seismic moment and attenuation of seismic surface waves[J]. Bulletin of the Seismological Society of America, 1969, 59(1): 275–287. Wang Huilin, Zhang Xiaodong, Zhou Longquan, Xu Xiaofeng, Yang Zhigao, Lu Xian, Li Zhirong. Study on the relationship between fluid infiltration and QS tomography of the crust in Zipingpu Reservoir Area[J]. Chinese Journal of Geophysics, 2012, 55(2): 526–537. Wang Suyun, Pei Shunping, Hearn T.M., Xu Zhonghuai, Ni J.F., Yu Yanxiang. Crustal S-wave estimating the ML amplitude Ⅱ: Q lateral variation in China[J]. Chinese Journal of Geophysics, 2008, 51(1): 133–139. Wang Zaihua, Liu Jie, Zhou Longquan, Wang Haitao. Tomography of 3-D velocity in earth's crust and upper mantle of middle Tianshan[J]. Inland Earthquake, 2008, 22(3): 203–211. Wittlinger G.H., Haessler H., Granet M. Three dimensional inversion of QP from low magnitude earthquakes analysis[J]. Ann. Geophys., 1983, 6: 427–438. Wong Yuklung, Zheng Sihua, Liu Jie, Zhao Xingquan, Kang Ying. Attenuation of ground motion and site response in Guangdong region[J]. Chinese Journal of Geophysics, 2003, 46(1): 54–61. Xie J., Nuttli O.W. Interpretation of high-frequency coda at large distances: stochastic modelling and method of inversion[J]. Geophysics Journal International, 1988, 95(3): 579–595. DOI:10.1111/j.1365-246X.1988.tb06705.x. Xu Yi, Zhu Jieshou, Liu Zhijian, Zhang Huaqing, Zhu Yan. Three-dimensional velocity image of the crust and upper mantle in the Tianshan, Xinjiang and its adjacent areas[J]. Acta Seismologica Sinica, 1994, 16(4): 480–487. Zhao Cuiping, Xia Aiguo, Zhen Shihua, Chen Zhangli. Study on source parameters of earthquakes occurred on the Middle East Area of Northern Tianshan in Xinjiang[J]. Earthquake Research in China, 2005, 21(1): 61–69. Zhao Cuiping, Chen Zhangli, Hua Wei, Wang Qincai, Li Zhixiong, Zheng Sihua. Study on source parameters of small to moderate earthquakes in the main seismic active regions, China mainland[J]. Chinese Journal of Geophysics, 2011, 54(6): 1478–1489. Zhou Longquan, Liu Jie, Su Youjin, Ma Hongsheng, Zhou Junjie. Tomography for Q of Yunnan region from high-frequency attenuation of S wave[J]. Chinese Journal of Geophysics, 2009, 52(6): 1500–1507.