Abstract
In this study a new tomographic method is applied to over 43,400 high-quality absolute direct P arrival times and 200,660 relative P arrival times to determine detailed 3D crustal velocity structures as well as the absolute and relative hypocenter parameters of 2809 seismic events under the Beijing-Tianjin-Tangshan region. The inferred velocity model of the upper crust correlates well with the surface geological and topographic features in the BTT region. In the North China Basin, the depression and uplift areas are imaged as slow and fast velocities, respectively. After relocation, the double-difference tomography method provides a sharp picture of the seismicity in the BTT region, which is concentrated along with the major faults. A broad low-velocity anomaly exists in Tangshan and surrounding area from 20 km down to 30 km depth. Our results suggest that the top boundary of low-velocity anomalies is at about 25.4 km depth. The event relocations inverted from double-difference tomography are clusted tightly along the Tangshan-Dacheng Fault and form three clusters on the vertical slice. The maximum focal depth after relocation is about 25 km depth in the Tangshan area.
1. Introduction
Beijing-Tianjin-Tangshan (BTT) region (114°E~120°E, 37.5°N~41.5°N) is situated in the northern part of North China. Figure 1 shows the major geological structure in the BTT region. This study region is under complex tectonic process with the Taihangshan uplift in the west, the Yanshan uplift in the northeast, and the North China Basin in the middle portion, which is a large continental basin and is characterized as an alternate uplift and depression zone [7, 8]. As shown in Figure 1, in the North China Basin and Taihangshan uplift, there are several active Cenozoic faults, such as Weixian-Yanqing Fault, Tongxian-Nanyuan Fault, Xiadian-Fengheying-Caojiawu Fault, and Tangshan-Dacheng Fault, that are oriented in NE-SW direction. There are also some active faults in NW-SE direction in the BTT region, such as Western Luanxian Fault, Laishui Fault, Ninghe Fault, and Nankou-Sanhe Fault.
The BTT region is a very active area with high seismicity. In this region, earthquakes are concentrated in four seismic zones: Zhangjiakou-Bohai seismic zone, Tangshan-Hejian-Cixian seismic zone, Sanhe-Linshou seismic zone, and Huailai-Weixian seismic zone. The Zhangjiakou-Bohai seismic zone in NW-SE direction is most active with a majority of large earthquakes in the BTT region. The other three seismic zones are parallel to each other in the NE-SW direction. Historically, strong earthquakes occurred frequently in this region. So far, more than 100 earthquakes with magnitude equal to and larger than 5.0 have occurred there since 780 BC. Thirty-four of them are with magnitudes larger than 6.0 and seven with magnitudes larger than 7.0. The great Tangshan earthquake () in 1976 is one of the most destructive earthquakes in history, which totally destroyed Tangshan city and caused a casualty of ~240,000. Therefore, a detailed investigation of the crustal structure and seismicity of the BTT region is very important not only for the understanding of physics of continental earthquakes but also for the assessment and mitigation of seismic hazard.
A lot of studies have been performed in the past three decades to invert for the three-dimensional (3D) seismic velocity structure of the crust and upper mantle beneath this region using arrival times from local and/or teleseismic events [2, 9–15] as well as the seismicity in this region [16–24]. However, the spatial resolution and the accuracy of event location are limited by using only the absolute arrival times. To improve the resolution of regional tomography, some seismic tomographic techniques have been developed, such as double-difference tomography (tomoDD) [6], which uses both absolute and relative arrival time data to determine a 3D velocity structure jointly with the absolute and relative event location. Furthermore, most of the previous efforts have focused on the crustal and mantle structure [2, 9–15] whereas little effort has been devoted to the study of relationship between seismicity of small earthquakes and velocity anomaly. Where is the location of top boundary of low velocity beneath Tangshan area? What is the deepest focal depth of the relocated earthquakes inverted by tomography? What is the main dynamic source for the great Tangshan earthquake? The detailed structure of hypocentral area is still unclear beneath the Tangshan area. Determining a high-resolution 3D velocity model of the crust and upper mantle beneath this region is the key for answering these questions. The tomoDD imaging has the potential to bring substantial insight into them by using both absolute and relative arrival time data.
In this study, we apply the tomoDD method to combine absolute and relative arrival time data to invert for the detailed 3D crustal P wave velocity structure jointly with absolute and relative event locations in the BTT region. Our results will shed new light on the relationship between relocated seismicity and the 3D velocity structure.
2. Data and Method
Both absolute and relative arrival times are used in this study. We carefully select the data such that each event has at least 6 recordings (8 in the Tangshan area). The resulting data include over 43,400 high-quality absolute direct P arrival times and 200,660 relative P arrival times from 3,983 earthquakes recorded by one or more of the 112 stations of the North China Telemetry Seismic Network (NCTSN) and the Capital Digital Seismic Network (CDSN) from 1993 to 2004 in the BTT region (Figure 2). The accuracy of the first P arrival time picking is estimated to be 0.2~0.3 s. The focal depth varies from ground surface down to about 30 km depth. The ray path coverage is generally good except Bohai Bay where no seismic station is present (Figure 3).
The tomoDD, developed by Zhang and Thurber [6], is used in this study to determine a 3D velocity structure jointly with the absolute and relative event location, which is based on the hypoDD of Waldhauser and Ellsworth [5] and also uses both absolute and relative arrival time data. With standard tomography, event locations will be somewhat scattered due to imprecise picks and origin-time errors. The tomoDD method uses the differential arrival times which are free from origin-time errors, and thus it removes some fuzziness from the velocity model.
Our starting 1D model is inferred from a minimum 1D velocity model [2] for the crust (0~25 km depth) and from deep seismic soundings [25, 26] for the deeper crust and upper mantle (25~40 km depth) (Figure 4). A 3D regular grid is used in this study [27]. The velocity values are interpolated by using a trilinear interpolation method. The model has been parameterized into an optimal grid spacing of 50 km laterally and 5 km vertically after a number of resolution tests for different grid spacing. Distance weighting is used in this study to control the maximum separation between event pairs. For closer event pairs a larger weight is applied. Considering the trade-off between the roughness and the stabilization of the model, we choose the model using smoothing weight of 5 as the preferred model. Velocity structure and hypocentral parameters of the local earthquakes are all taken to be unknown parameters in the inversion. A detailed description of the method is given by Zhang and Thurber [6].
3. Seismic Tomography
Local (regional) earthquake tomography (LET) plays an important role in studying the velocity structure of the Earth’s interior, which has become a relatively routine application for use in seismically active regions covered by one or more dense seismic network.
We conduct many inversions using different values of damping parameter for the variance of the velocity perturbations and root-mean-square (rms) travel time residuals. We find that the best value of the damping parameter is 150. In order to confirm the main features of our tomographic image, we conduct a resolution test to assess the adequacy of the ray coverage and to evaluate the resolution [28, 29]. An initial checkerboard velocity model is created by assigning alternately positive and negative velocity anomalies (3%) to the 3D grid nodes in the model space. Synthetic travel times are calculated for the checkerboard model using the real event and receiver locations. A random noise with zero mean and standard deviation of 0.1 s is added to the synthetic data. The resolution is considered to be good for regions where the checkerboard image is well recovered. Figure 5 shows the result. The checkerboard pattern is recovered for almost the entire study region except for Bohai Bay and the edge of the BTT region (Figures 5(a)–5(f)). Areas with low resolution are excluded from the resulting tomographic images (Figure 6). The best resolution is in the depth range of 5–20 km (Figures 5(a)–5(d)), where the amplitude of velocity anomalies is well recovered across the whole region. The resolution is reduced below 30 km depth. But the checkerboard positive and negative patterns are basically recovered.
(a)
(b)
(c)
(d)
(e)
(f)
In order to show more clearly the continuous variations of velocity anomalies in the depth direction, our resulting tomographic images are presented in Figure 6. In general, the results reveal strong lateral heterogeneities in both of the crust and uppermost mantle. It is noted that the media beneath the Tangshan area are very different from adjacent areas throughout the crust and upper mantle. In the shallow depth (Figures 6(a) and 6(b)), the inversion results are consistent with the local geological settings and follow the trend of active faults in the BTT region. The tomographic images illustrate that the low P velocity (low-) anomalies exist beneath depressions and basins (such as the North China Basin) and high P velocity (high-) anomalies exist beneath mountains and uplifts (such as Yanshan uplift and Taihangshan uplift), which is consistent with the previous standard tomographic studies [2, 9–15]. But our model has sharper velocity contrasts near the boundary between low- and high- anomalies than previous tomography models do. Although the Tangshan area is located in the North China Basin, it is an uplifted block beside the Zhangjiakou-Bohai seismic zone [21]; hence, it shows up as high- anomaly. At the 35 km depth, our present result has revealed that a broad and prominent low- anomaly exists beneath the Taihangshan uplift area and the Tangshan area, and high- anomaly exists beneath the Yanshan uplift and the North China Basin. These results are consistent with the tomographic results [15, 30, 31].
Several cross-sections along the different longitudes (115.5°E, 116.5°E, 117.5°E, and 118.5°E) and latitudes (40.5°N, 39.8°N, and 39.0°N) are presented in Figure 7. At shallow depth (5~10 km), the boundary between the low- anomaly and high- anomaly is well consistent with the boundary between mountain/uplift and plain/basin, such as the 39.5°N area at the 115.5°E profile (Figure 5(a)) and 118.0°E area at the 39.8°N profile (Figure 7(b)). Our tomoDD model shows a high- anomaly of ~90 km length at 10~20 km depth under the Beijing, Tianjin, and the Tangshan area at the profile of 116.5°E, 117.5°E, and 118.5°E (Figure 5(a)). A prominent broad low- anomaly is discovered from 20 km to 30 km both beneath Yanshan uplift, the North China Basin area at the profile of 115.5°E (Figure 5(a)), and beneath the Tangshan area at the profile of 118.5°E (Figure 5(a)). At the profile of 39.8°N (Figure 5(b)), the P velocity is high beneath the east of the Tangshan area, where there is uplift block near the Zhangjiakou-Bohai seismic zone. Moreover, it can also be found that a broad high- anomaly beneath the Taihangshan uplift extends toward the east and down to ~20 km depth beneath the Beijing area.
(a)
(b)
4. Relationship between Seismicity and Tomography Image in the BTT Region
An advantage of the tomoDD is that it determines the 3D velocity model as well as the absolute and relative event location compared with standard tomography. We analyzed 3,983 earthquakes with magnitudes from M 1.0 to 6.2 recorded by 112 stations. An event will be excluded from the inversion if it cannot be connected to any other events, and as a result only 2,809 hypocentral parameters of both absolute and relative locations are given by the tomoDD. The weighted rms travel-time residuals decrease from 1.2 s to 0.3 s. Figure 2 shows the catalog locations, which are scattered along major active fault zone both in horizontal direction and depth direction due to imprecise picks, origin-time errors, and simple 1D velocity model. After relocation, the tomoDD method provides a sharp picture of the seismicity in the BTT region, which is concentrated along with the major faults in a shape of alignment (Figure 8).
To illuminate the relationship between seismicity and velocity anomaly, we present our tomographic images together with hypocentral locations of both relocated earthquakes within 5 km off each layer depth and historic earthquakes (M ≥ 6.0) that occurred in the BTT region (Figure 9). Although we do not know the accurate focal depths of the historic earthquakes, the statistic analysis of focal depth after the tomoDD relocation [32] suggests that most of earthquakes that occurred in the middle and lower crust under the BTT region and the North China are mainly clustered at 1~24 km depth. In the tomographic image of 10 km and 15 km depth (Figure 9), both the relocated earthquakes and historic earthquakes have a similar feature, that is, most of the earthquakes are located in the conjunctional areas of low- and high- anomalies. They are slightly closer to the high- anomaly areas. The epicentral location of the 1976 Tangshan earthquake, the 1976 Luanxian earthquake, and the 1679 Sanhe earthquake is in the transitional area closer to the high- anomalies. It is notable that the distribution of relocated small earthquakes is consistent with the trend of high- anomalies under the Beijing-Tangshan area. Maybe it suggests that the conjunctional zones of low- and high- anomalies represent weak sections of the seismogenic crust. The tectonic stresses are prone to being accumulated in the “brittle” high- anomalies area, and hence the earthquake ruptures happened closer to the high- anomalies zones. The locations of earthquakes, especially destructive earthquakes, are not random and are related closely to their deep structure of crust and upper mantle.
(a)
(b)
Figure 10(a) shows a cross-section along profile AA′ (Figure 10(c)) passing through the Tangshan-Hejian-Cixian seismic zone. A prominent high- anomaly zone about 100 km in length is visible from 10 km down to 20 km depth along the Tangshan-Tianjin area, while a broad low- anomaly exists in Tangshan and the north of the Tangshan area from 20 km down to 30 km, which is in agreement with the tomographic results of Huang and Zhao [15] using local crustal earthquakes, controlled seismic explosions, and quarry blasts. Due to the differential arrival time data used to improve the precision of event location in the tomoDD, we obtained similar tomographic image only with local earthquakes. In the upper crust, the cross-sectional images show that discontinuous low- anomalies exist under the Tangshan-Hejian-Cixian seismic zone, while, in the middle and lower crust, the low- anomalies change to high- anomalies. Under the Tangshan area, the maximum focal depth locates at the boundary of low- anomaly in the middle and lower crust.
(a)
(b)
(c)
Figure 10(b) shows a cross-section along profile BB′ passing through the Zhangiakou-Bohai seismic zone. Under the Tangshan area, our result displays a transitional zone of low- anomaly in the northwest and high- anomaly in the southeast in the upper crust, while a very prominent low- anomaly exists in the middle and lower crust (20~30 km depth). The focal depth of relocated earthquakes that occurred in the Tangshan area is distributed in the transitional zones of low- anomaly and high- anomaly. In the Zhangjiakou area, the northwest of profile BB′, the deepest focal depth of relocated earthquakes with magnitude ≥ 4.0 is about 15 km, which occurred on the margin of high- anomalies.
The Tangshan area, in the about 160 km southeast of Beijing, has the highest level of seismicity in BTT region. In this area numerous small earthquakes have occurred frequently since the great Tangshan earthquake in 1976. During 1993~2004, 118 earthquakes with ≥ 3.0 occurred in the area, 17 of them were larger than 4.0, such as the earthquake with 5.9 on 6 October 1995 and 5.0 on 20 January 2004 in the northeast of Tangshan.
Figure 11 shows the epicenters of earthquakes before (open circles) and after relocation (solid circles) using the tomoDD in the Tangshan area. Compared with the catalog locations, which are scattered along the fault zone, relocated hypocenters appear more clustered in the NE-SW direction along Tangshan-Dacheng fault. Figure 11 shows three clusters in different colors: the Tangshan cluster oriented in the NE-SW direction (grey solid circles), the Luanxian cluster oriented nearly in the E-W direction (green solid circles), and the Qian’an cluster oriented in the NE-SW direction (blue solid circles).
In the cross-section along profile TT′ (Figure 12), the distribution of hypocenters displays a big difference between the catalog location, which is layered and scattered without clear cluster characteristic (Figure 12(a)), and the relocated location with the tomoDD, which is clustered clearly as cluster A, cluster B, and cluster C (Figure 12(b)). The Tangshan cluster (Figures 11 and 12(b), grey solid circles) becomes two clusters in depth: cluster A and cluster B. In addition, an earthquake (5.9) without depth parameter before relocation is relocated in cluster A with focal depth 6.3 km.
(a)
(b)
In the cross-section along profile MN, perpendicular to TT′, the distribution of hypocenters before and after relocation using the tomoDD is also very different (Figure 13). Most of the tomoDD locations are centralized on a narrow zone within 10 km off the profile TT′ (Figure 13(b)). The earthquake relocation with ≥ 4.0 (Figure 13(b), stars) shows a near-vertical plane between 5 km and 15 km and a slight west dip between 15 km and 25 km, which is in agreement with the results of deep seismic soundings (DSSs) [33]. This indicates that the Tangshan fault is near-vertical in the shallow depth and west dip in the depth of about 22 km.
(a)
(b)
For comparison, we also relocate the events by using the DD location method, and extract the hypocenter parameters of earthquakes that occurred in the Tangshan area from previous results of the two standard tomography methods [2, 4]. The same minimum 1D velocity model [2] is used for both standard tomography and tomoDD as the initial model, which is also used for DD event location.
Figure 14 shows the event locations along profile TT′ in the Tangshan area inferred from different methods. Figure 14(a) shows the catalog locations by NCTSN. Figures 14(b) and 11(c) show the event relocations [2, 4] by the two standard tomography methods by Thurber [1] and Zhao et al. [3], respectively, where only the absolute arrival times were used for the inversion. The event locations are still scattered, similar to the catalog locations (Figure 14(a)).
(a)
(b)
(c)
(d)
(e)
Figures 14(d) and 11(e) show the event relocations by the DD location method [5] and the tomoDD method [6], respectively. In the DD location method, the weighted rms residuals decrease from 1.0 s to 0.6 s, while, in the tomoDD method, the weighted rms residuals decrease from 1.2 s to 0.3 s. After the relocation, both the DD methods provide similar features, three typical clusters under the Tangshan area. Although most of the relative event locations from the two DD methods are quite similar, there are some differences between them in detail. First, it can be noted that the absolute event locations with ≥ 4.0 are different between the two methods. The focal depth of earthquakes with ≥ 4.0 varies from 0 km to 20 km in the DD location method. In the tomoDD method, however, it varies from 5 km to 20 km. Second, in the tomoDD, earthquake relocations with ≥ 4.0 show a near-vertical plane between 5 km and 15 km and a slight west dip between 15 km and 25 km, which is in agreement with the results of deep seismic soundings (DSS) [33], while no such dipping is observed in the DD location method.
Figure 15 shows the velocity perturbations, resulting from the tomoDD, along the vertical cross-section of profile TT′ passing through the Tangshan area and the seismicity in this area. Only earthquakes that occurred within 0.25° off profile TT′ are chosen to project on the vertical slice. It can be clearly seen that all of the three clusters (Figure 12(b)) are relocated in the conjunctional area of low- and high- anomalies, slightly closer to the high- anomaly zones. Only a few earthquakes have ever occurred within the low- anomalies. The maximum focal depth (about 25.4 km) locates in the uppermost boundary of low- anomaly from 20 km to 30 km under the Tangshan area. Our results suggest that the top boundary of low- anomalies is at about 25.4 km depth in the Tangshan area, which is different from results obtained by previous standard tomography studies [15]. This indicates that the top boundary of low- anomalies is at 20 km depth in the Tangshan area, which is consistent with the maximum of focal depth of relocation by the tomoDD.
Our tomographic results show an evident low-velocity anomaly in the lower crust beneath the BTT region (Figures 6, 7, 10(a), and 10(b)), specially beneath the Tangshan area. The results of S structure by using receiver function method [34] showed that there exist obvious heterogeneous low- media in the upper mantle and middle crust and the crust-mantle boundary had an obvious uplift beneath the Tangshan area. Thus the existence of prominent low- anomalies in the lower crust may suggest that there is probably massive intrusion derived from the upper mantle beneath the Tangshan area. Our tomographic results of the crust and upper mantle support such a conclusion. The main dynamic source for the Tangshan earthquake is the vertical movement of the upper mantle, which leads to material and energy exchange between the crust and upper mantle [34]. The long-term influence of the upwelling of mantle materials on the seismogenic layer would change the stress distribution and compositional evolution of fault zones, and the stresses are easier to concentrate on the high- media, which would lead to the mechanical failure and the earthquake occurrence.
5. Conclusions
The tomoDD method is efficient in relocating a large number of earthquakes accurately and in characterizing the local velocity structure with high resolution. With this approach a high-resolution tomography model of crust and upper mantle under the BTT region has been obtained by using both absolute and relative arrival times of local earthquakes recorded by NCTSN or DCSN. Simultaneously, our results provide accurate hypocentral parameters of both absolute and relative event locations in the BTT region. The velocity images of the upper crust correlate well with the surface geological and topographic features. In the North China Basin, the depression and uplift areas are imaged as slow and fast velocities, respectively. A broad low- anomaly exists in Tangshan and the north of the Tangshan area from 20 km down to 30 km depth, which suggests that there is probably massive intrusion derived from the upper mantle beneath the Tangshan area. Our results suggest that the top boundary of low- anomalies is at about 25.4 km depth in the Tangshan area.
After relocation, the tomoDD method provides a sharp picture of the seismicity in the BTT region, which is concentrated along with the major faults in a shape of alignment. The seismicity of both the relocated earthquake hypocenters and the historic earthquakes shows that majority of the hypocenters are located in the conjunctional areas of low and high P wave velocity anomalies. And they are slightly closer to the high P wave velocity abnormal areas. Only a few earthquakes have epicenters in either high or low P wave velocity areas. It might suggest that the conjunctional zones of low- and high- anomalies represent weak sections of the seismogenic crust. The tectonic stresses are prone to being accumulated in the “brittle” high- anomalies area and hence the earthquake ruptures happened closer to the high- anomalies zones.
The surface event relocations in the Tangshan area are centered along the Tangshan-Dacheng fault. In the vertical slice along profile TT′, all the earthquakes are clustered in three clusters as shown in Figure 9(b), two clusters lie beneath the Tangshan-Ninghe fault, and another one lies beneath the Luanxian area. The maximum of focal depth of relocated earthquakes is 25 km, where there is the top boundary of low- anomalies beneath the Tangshan area.
Acknowledgments
The authors thank Dr. Haijiang Zhang for providing his tomoDD code and thoughtful discussions. This work was supported by a grant from the National Natural Science Foundations of China (40404002) to X. Yu as well as a grant from the National Natural Science Foundations of China (90715019) to W. Zhang and by the Special Research Project in Earthquake Science, CEA, to Y.-T. Chen (no. 200808068). The figures were made by using GMT [35]. They also thank the editor (Professor Y. Zhang) and four anonymous referees for their constructive comments and suggestions which improved the paper.