Journal of Earth Science  2018, Vol. 29 Issue (6): 1340-1348   PDF    
0
Insight into Urban Faults by Wavelet Multi-Scale Analysis and Modeling of Gravity Data in Shenzhen, China
Chuang Xu1, Haihong Wang2,3, Zhicai Luo1, Hualiang Liu4, Xiangdong Liu1    
1. MOE Key Laboratory of Fundamental Physical Quantities Measurement, School of Physics, Huazhong University of Science and Technology, Wuhan 430074, China;
2. School of Geodesy and Geomatics, Wuhan University, Wuhan 430079, China;
3. Key Laboratory of Geospace Environment and Geodesy, Ministry of Education, Wuhan University, Wuhan 430079, China;
4. China Water Resources Beifang Investigation, Design and Research Co. Ltd., Tianjin 300222, China
ABSTRACT: Urban faults in Shenzhen are potential threats to city security and sustainable development. In consideration of the importance of the Shenzhen fault zone, the author provide a detailed interpretation on gravity data model. Bouguer gravity covering the whole Shenzhen City was calculated with a 1-km resolution. Wavelet multi-scale analysis (MSA) was applied to the Bouguer gravity data to obtain the multilayer residual anomalies corresponding to different depths. In addition, 2D gravity models were constructed along three profiles. The Bouguer gravity anomaly shows an NE-striking high-low-high pattern from northwest to southeast, strongly related to the main faults. According to the results of MSA, the correlation between gravity anomaly and faults is particularly significant from 4 to 12 km depth. The residual gravity with small amplitude in each layer indicates weak tectonic activity in the crust. In the upper layers, positive anomalies along most of faults reveal the upwelling of high-density materials during the past tectonic movements. The multilayer residual anomalies also yield important information about the faults, such as the vertical extension and the dip direction. The maximum depth of the faults is about 20 km. In general, NE-striking faults extend deeper than NW-striking faults and have a larger dip angle.
KEY WORDS: urban faults    Bouguer gravity anomaly    wavelet multi-scale analysis    gravity modeling    Shenzhen    

0 INTRODUCTION

Due to the lack of geological knowledge in the early period, many cities were established on fault zones. Nowadays, urban seismic hazards become one of the major threats for these cities. With the rapid urban development, more and more large infrastructures, such as ultrahigh buildings, supersized bridges, subway systems, are constructed to accommodate increasing population in the city. The large-scale infrastructures built on or near faults increase the chance of earthquake because the underground faults might gradually lose their balance due to the surface overload. By 2000, 281 out of 672 cities in China have suffered earthquakes (Mw > 4.0) during the last 4 300 years, in which 27 cities have even experienced earthquakes with Mw larger than 7.0 (Xu et al., 2004). Some earthquakes resulted from urban faults had caused tremendous casualties and economic loss, for example, Mw 7.8 Tangshan Earthquake in 1976, Mw 7.2 Kobe Earthquake in 1995, and Mw 7.6 Izmit Earthquake in 1999. Therefore, urban faults become great potential threats to human life and social wealth.

Delineation of urban active faults is of great significance for urban construction planning and disaster mitigation, hence has been emphasized by many countries and local governments. The Geospatial Information Authority of Japan (GSI) has published a series of 1 : 25 000 scale active fault maps in urban areas since 1996 (Iwahashi, 2010). The United States Geological Survey (USGS) compiled a database containing active faults and folds during the Quaternary (USGS, 200). Italy started the project ITHACA (Italy Hazard from Capable Faults) in order to build a tool for summarizing and making easily available information on capable faults, which are able to produce significant ruptures or deformations at or near the topographic surface. Based on the ITHACA database, a first assessment of fault displacement hazard in urban areas in Italy was made (Guerrieri et al., 2014). In China, an urban active fault surveying project was conducted by the China Earthquake Administration (CEA) in 2003. Active fault detection and earthquake hazard evaluation have been carried out in 20 metropolises in this project.

Complicated urban environment poses a challenge for fault detection. Most urban active faults belong to concealed faults, because they are typically overlain by Quaternary unconsolidated sediments. It is hard to investigate the concealed faults from surface geological and geomorphological characteristics. Several methods can be applied to detect concealed faults, including artificial seismic exploration (Sato et al., 2009; Karastathis et al., 2007; Hinsch et al., 2005; Fuis et al., 2001), electric prospecting (Sultan Araffa et al., 2012; Suski et al., 2010; Diaferia et al., 2006), ground-penetrating radar (Yalçiner et al., 2013; Bhosle et al., 2007; Slater and Niemi, 2003; Audru et al., 2001), and so on. However, not only the high density of population and buildings but also existence of various high- frequency vibrations and strong interfering noise in modern cities, make these geophysical methods inconvenient or inefficient in consideration of the construction condition, anti- interference and destructiveness. Besides, lack of information on deeper faults can not be supplied by these approaches.

Compared to the aforementioned methods, gravimetry is convenient to be implemented in urban areas (Xu et al., 2015; Pamukçu et al., 2014; Abbott and Louie, 2000). Gravity reflects the mass distribution in the Earth and hence can be used to infer subsurface structures, although gravity inversion is non-unique, similar to other geophysical problems. The spatial relationship between gravity anomalies and geologic structures provides clues concerning the distribution and evolution of structures (Selim and Aboud, 2012; Blakely et al., 1997). By far, lots of methods have been proposed for fault inversion and interpretation of gravity data (Zhang et al., 2004; Bansal and Dimri, 2001; Abdelrahman and El-Araby, 1993; Mareschal, 1985). In recent years multi-scale analysis (MSA) based on wavelet has become a favorite approach for gravity data processing and interpretation (Zamani et al., 2013; Jiang et al., 2012; Xu et al., 2009; Wang, 2005; Moreau et al., 1999; Hou and Yang, 1997). The MSA has advantages in separation of signals and identification of sources at different depths.

In this paper, gravity data are applied to delineate urban faults in Shenzhen, one of the fastest developing cities in China. Several faults go across the downtown of Shenzhen City, which severely affects the urban security and sustainable development. It is a common concern to clearly identify the faults and to assess their activities in this area. A number of investigations were carried out to survey the urban faults by diverse geophysical technologies (Xu, 2014; Yu, 2010; Ma and Chen, 2009a, b; Sun et al., 2007; Wang et al., 2005; Chen et al., 2001; Tan et al., 2000). The distribution and risk evaluation of main faults in Shenzhen City were preliminarily obtained. Studies indicate that the main faults are relatively stable with weak activity (Xu et al., 2015; Tan et al., 2000). Nevertheless, some critical information such as vertical extension and dip of these faults is still rather obscure. The present work aims to improve the knowledge on depth and dip angle of urban faults in Shenzhen based on MSA and modeling of gravity data.

1 GEOLOGICAL SETTING

The study area is located in the southeastern margin of the Cathysian Block, adjacent to the Philippine Sea Plate. Tectonically it belongs to the Zijin-Huiyang concave fold, a fourth grade tectonic unit of the South China Caledonian folded system (Sun et al., 2007). As the convergence center of three modern plates, namely, Indian Plate, Philippine Sea Plate and Pacific Plate, the Cathysian Block has been pushed by the southeastward uplifting of Qinghai-Tibet Plateau and northwestward subduction of Philippine Sea Plate (Yu, 2010). In this case, a number of large-scale NEE faults were formed in the Cathysian Block. Lianhuashan fault zone (LFZ), a deep and large fault about 360 km long from Fujian Province to Shenzhen, was formed in Mesozoic Yanshanian. Due to the compression and extension in Indo-Chinese epoch, Middle Jurassic, Late Jurassic and Cretaceous, Shenzhen fault (SF) and Lishui-Haifeng fault (LHF) were gradually developed (Huang and Zhang, 1990).

Shenzhen fault zone, regarded as the extreme southwest part of LFZ, was formed and evolved in several tectonic movements since Caledonian. The main faults in this area can be distinguished into three groups of NE, NW and EW trending (Yu, 2010). The NE faults with the largest scale and a Xi-type structure are composed of four nearly parallel faults (F1: Jiuweiling fault (JF), F2: Henggang-Luohu fault (HLF), F3: Liantang fault (LF), F4: Yantian fault (YF)). These faults cover the whole urban zone with the length of about 150 km, dominating the distribution of strata, intrusive bodies and metamorphic rocks in the study area. F1, with mainly anti-clockwise slip, has the multi-phase active characteristics of compression in the early stage and tension in the later stage. F2 is the axis of the NE faults, displaying as an "S" shape with a 38 km long segment across the downtown (Sun et al., 2007). Thousands of meters wide metamorphic zone and crush zone developed along F2 (Ma and Chen, 2009b; Sun et al., 2007). This fault is cut into several segments by NW faults during three strong tectonic movements since Late Jurassic. Most of F3 develops in the saddle or gentle slope zone of mountains, and its relative displacement twists in the clockwise direction. F4 is the southmost branch of Xi-type faults, which is shallower and smaller in scale than the others. F5–F8 (F5: Wentang-Guanlan fault (WGF), F6: Humen-Dongboliao fault (HDF), F7: Tiantangwei fault (TF), F8: Youganpu fault (YGF)) belong to the NW faults, developed later and with a smaller scale than the NE faults. Most parts of the NW-striking faults are located in the downtown area, among which F5 has the largest scale (Ma and Chen, 2009a). It is a normal fault with an anti-clockwise twist. The scale of F6–F8 is relatively smaller than F5. The EW faults have the smallest scale among the three groups, mostly lying to the east of Shenzhen. Therefore, the EW faults will not be discussed in this study. Small earthquakes (2.0 < Mw < 3.9) often occur in the cross zone of NE and NW faults, as well as in the intersecting zone between NE and EW faults (Yu, 2010).

Magmatic rocks of different periods are widely distributed in this region. It indicates that the magmatic activity is very intense during the long-term geological tectonic movement. Precambrian rock (An), the oldest rock in the study area, gradually disappears during the tectonic movement and distributes scatteredly in the northwestern and southern Shenzhen nowadays. The Upper Paleozoic (Pz), Mesozoic (Mz) and Yanshanian rocks (γ523) cover 90% of the area. In addition, there are also fragmented distributions of Caledonian (γ3), Indo-Chinese intrusive (γ51), Middle Pleistocene (Qp), and Holocene rocks (Qh) (Yu, 2010). The evolution of rocks is synchronous with the development of faults during tectonic movement in Shenzhen area. Many faults become the boundaries of rock distribution.

The terrain is high in the southeast and low in the west (Fig. 1). In southeastern area, small mountains are distributed along the coasts of Mirs Bay and Daya Bay. Low hills are the typical feature in the central and northwest part. Flat tablelands intersperse among the hills. To the southwest, there is the coastal plain. There is a certain relationship between the terrain and the distribution of faults. There are usually linear mountains, scarps and valleys along the urban faults. In the NE-SW and ENE-WSW direction, the distribution of intermountain basins and valleys of Quatemary are consistent with the trend of the main faults in Shenzhen. After strong karstification, the trace of faults in intermountain basins and valleys has not been obvious since Late Pleistocene and Holocene.

Download:
Figure 1. Topography from SRTM of Shenzhen. The dashed red box shows the study area. M1. Yinping Mountain; M2. Huangchao Mountain; M3. Guanyin Mountain; M4. Tiantou Mountain; M5. Paiya Mountain; M6. Qiniang Mountain; M7. Maluan Mountain; M8. Wutong Mountain; M9; Yangtai Mountain; M10. Shuilian Mountain.
2 DATA AND METHOD

In the present work, a 1-km grid of Bouguer gravity anomaly data is used (Fig. 2). The grid covers the whole Shenzhen City with the length of 100 km in EW direction and 42 km in NS direction, respectively. Bouguer gravity anomaly data were compiled from 3 609 land gravity observations and 1 262 marine gravity observations, with an average station density of one per km2. For the reduction, normal gravity was calculated from the WGS84 reference ellipsoid. Free air corrections were computed using a gradient of 0.308 6 mGal/m. Terrain corrections were applied on the basis of a digital terrain model with a resolution of 100 m. The reduction density was taken as 2.67 g/cm3. For the marine observations, the density of seawater was taken into consideration. For more details about the data, refer to Xu et al. (2015).

Download:
Figure 2. Bouguer gravity anomaly. Local coordinates are used here, and the coordinates of the bottom-left and top-right corners correspond to (113°43′E, 22°31′N and (114°40′E, 22°55′N), respectively. The white lines are coastlines. Dashed black lines (P1-P1', P2-P2' and P3-P3') represent the location of gravity profiles. The red lines are F1–F8 faults. The white dots are earthquake epicenters since 1990.

Negative anomalies throughout the Shenzhen area indicate density deficit. Features in the Bouguer gravity anomaly map show strong relationship with terrain and underground rock characteristics. The gravity lows and highs alternately distribute from the northwest to the southeast. The lowest gravity anomaly occurs in Yangtai Mountain (M9) with the altitude of 587 m a.s.l. in the southwest. The gravity lows appear in the hilly lands where Yanshannian rocks dominate. The gravity highs occur in low topography regions. The northwestern and southwestern gravity highs are consistent with alluvial deposits and erosional surface, respectively. In the southeastern corner, the distribution of gravity anomaly is scattered and the trap is small, indicating complex structure under this area.

Several NE and NW trending gravity anomaly gradient zones can be apparently observed in Fig. 2. These gradient zones are consistent with the distribution of the main faults. The steepest anomaly gradient lies at the north root of Maluan Mountain (M7) and Wutong Mountain (M8), ending in the Shenzhen Bay to the south. To the north, it intersects with an EW-trending gradient located at the north of Tiantou Mountain (M4). Lots of micro earthquakes happened near the two gradients. The NW- trending anomaly ridges can be seen where faults F6 and F7 are located. The main steep anomaly gradients are located at the boundaries of rocks. The gravity anomaly variations have strong correlation with the mapped geology at the surface.

Thanks to the close relationship between the gravity anomaly and geological structures, rich geophysical information of underground structures can be revealed from the gravity anomaly. However, surface gravity anomaly results from all of the structures in the Earth's interior. It should be decomposed to study the individual target at different depths. Traditionally, the anomaly is separated into regional and local anomalies by a certain method, for example, trend analysis, high-order derivative method or analytical continuation. Here, we applied wavelet-based MSA to decompose the Bouguer gravity anomaly in order to analyze the vertical extension of the main faults in Shenzhen. Due to the good localization feature and multiscale analysis function, MSA is capable of delineating structures at different depths and has been successfully used in potential fields recently. MSA can decompose the signal into a series of details and approximations. Approximations can be thought as the regional anomalies produced by deep large-scale geological bodies. Details correspond to the residual anomalies of high frequency produced by shallow small-scale geological bodies. The larger the order of the residual anomaly is, the deeper the source depth will be. With the order increasing, the anomalies related to the shallow and small structures will vanish (Jiang et al., 2012). The average source depth corresponding to the decomposed anomaly can be estimated using the logarithm power spectrum method (Cianciara and Marcak, 1976; Syberg, 1972; Spector and Grant, 1970). In this study, Coif3 wavelet basis was chosen to separate the gravity anomaly after comparing tens of candidate wavelets in synthetic experiments. The first five order residual gravity anomalies were used to study the fine crustal structures.

In order to better understand the primary faults of the study area, three gravity profiles transverse to the Shenzhen fault zone were selected for modeling. Figure 2 shows the locations of the three profiles. Profile P1-P1' is totally 53.5 km long from the gravity high in the northwest to the gravity low at Wutong Mountain (M8), nearly parallel to the gradient between fault F5 and F7. It is orthogonal to the NE-striking fault group (F1-F4) in the south end. Profile P2-P2' starts from the valley between Yinping Mountain (M1) and Huangchao Mountain (M2), and heads southeast to the Mirs Bay with the length of 46.6 km. It is also approximately perpendicular to the NE-striking fault group. Profile P3-P3' crosses the largest gravity low in a SW-NE direction, orthogonal to the NW-striking fault group with the length of 74.9 km. The modeling process was undertaken in 2D using interactive software GMS3.0, developed by China University of Geosciences (Wuhan).

3 RESULTS AND DISCUSSION 3.1 Analysis of Residual Gravity Anomalies

Figure 3 shows the residual gravity anomalies decomposed by MSA. Limited by the data length and resolution, only 5 order decompositions can be implemented. Based on the slope of the logarithm power spectrum of anomaly, the average source depth corresponding to each order residual anomalies is estimated (see Table 1). The largest average depth of the source is 20 km, estimated from the 5th order residuals. Comparing the residuals with the regionals, we found that the feature of the 5th order residuals is exactly the same as the 4th order regionals, whose corresponding depth is 34 km. The thickness of the crust in this area is about 30 km according to the deep seismic sounding result (Li et al., 2006). Therefore, the 1st- to 5th-order residuals can reveal density variations from the surface to the middle crust or even to the top of the lower crust.

Download:
Figure 3. The first five order residuals of Bouguer gravity anomaly are decomposed by MSA. The black dashed lines, black lines and white lines, for assistant analysis, are gravity profiles, F1–F8 faults and coastlines, respectively. The black circles from (a) to (e) represent the earthquake epicenter distribution at the depths of 0–3, 3–5, 5–7, 7–17 km and larger than 17 km, respectively.

The residual anomalies in each order vary slightly with the amplitude less than 3 mGal. The small anomaly variation indicates little structural diversity at different depths, which consists with the tectonic and seismic activity environment in this area. Although the anomaly amplitude is small, the main faults can be delineated from the residual anomaly maps.

Table 1 Average source depth corresponding to each order residual gravity anomalies

In the 1st-order residual map (Fig. 3a), gravity anomalies distribute dispersedly, showing lateral density inhomogeneity in the sedimentary layers and the crustal surface. Positive anomalies dominate in this layer. It only shows a few small negative anomaly circles at some mountains (M2, M5, M9). The signal corresponding to the faults is very vague because of the interference from the covered sediments. However, we can observe two obvious anomaly features from the map. One is a positive anomaly strip along the NE trending faults. The strip strikes NE in the south and turns to NEE in the north. The fault F2 is the axis of the strip. The faults F1 and F2 form the borderline of the strip. Obviously, the positive anomaly resulted from the upwelling of high-density materials along the fault zone, and the fault F2 controls the upwelling. Another feature is alternate negative and positive anomalies near the fault F6 in the southwest of the study area. The strike of these anomalies is parallel to the fault F6, which just lies in the negative-positive transition zone.

The mean source depth of the 2nd-order residual anomalies (Fig. 3b) is about 4 km. The anomalies in this layer begin to show strong correlation with the faults. The positive strip along the NE-striking faults becomes clearer, as well as the alternate negative and positive anomalies near F6. In the central area, another negative-positive-negative pattern appears, corresponding to the three NW-striking faults (F5, F7 and F8). A high gravity anomaly lies beneath the fault F7. The fault F8 traverses the low gravity anomaly L2 and deforms it in the southwest. The fault F4 passes through a low gravity anomaly near Mirs Bay. In the east, there are apparent nearly EW trending anomalies, indicating the existence of several EW faults parallel to each other. It is noteworthy that the 2nd-order residual anomalies are completely consistent with the distribution of the upper crustal rocks. Negative anomalies are strengthened and become more continuous. These gravity lows (L1, L2 and L3) are mainly caused by Yanshanian rocks (γ523). Accordingly, the gravity highs (H1, H2 and H3) appear where older rocks exist, for example, Mesozoic (Mz), Upper Paleozoic (Pz) and Holocene (Qh).

To the depth of 6 km (Fig. 3c), the correlation between anomalies and faults is still obvious. Negative anomalies are further strengthened and dominant. The low gravity anomaly L1 with the largest area is situated between the faults F5 and F6, centered by Yangtai Mountain (M9). The strongest low gravity anomaly L2 appears in the hilly area to the north of Henggang. The negative anomalies along the sides of the fault F4 in Fig. 3b merge together to form the gravity low L3 in Fig. 3c. It is deduced that the influence of the fault F4 does not reach this depth. The gravity low beneath the fault F8 in Fig. 3b has been completely merged into L2, indicating the maximum depth of F8 is about 6 km. The positive anomaly belt along the fault F2 is weakened due to the crush of NW-SE tectonic stress, while the anomalies in some intermountain basins and valleys reach their maximum (e.g., the gravity highs H1 to the north of Guanlan and H2 near Pingshan Basin in the east). Two corridors, formed by H1 and H3 of gravity highs, are consistent with the distribution of F5, F6 and F7. In addition, the strike of H1 and L2 suggests that there might be NEE-striking faults in this area. Compared with Fig. 3b, H3 shows different variation tendency. The magnitude decreases in its south part, while it increases in the north part where the Songgang Plain is located.

The gravity anomalies in the middle crust (Fig. 3d) are very smooth and can be roughly divided into three lows and three highs by the faults. In the southwest, the gravity low L1 is enhanced further and extends to the Shenzhen Bay. Correspondingly, the anomalies resulted from the fault F6 nearly disappear. The positive anomalies in the northwest join together to form a strong NE striking gravity high. In the central area, the gravity low L2 and the gravity high H1 apparently decrease in amplitude. Their strikes turn into NE direction and NW direction from NEE direction in upper layers, respectively. The gravity high H2 strongly dominates the eastern area, striking EW direction from Pingshan to the Daya Bay. In the south, the gravity low L3 moves toward south. The positive anomaly between L1 and L3 does not exist. A trend for consolidation between L1 and L3 can be seen. It also indicates that the southwestern ends of F1, F2 and F3 vanish at this depth. However, the existence of these NE striking faults can still be inferred from the weakening high-low gravity gradient belt. Another obvious gravity gradient zone occurs along the fault F5 between L1 and H1. The gradient along F5 in Fig. 3d is gentler than those in Figs. 3b and 3c.

The 5th-order residuals (Fig. 3e) reveal a fold density structure in the lower crust, consistent with the regional tectonic setting. Two high gravity anomalies lie in the northwestern plain area and in the southeastern bay area, respectively. The central area has a relatively lower gravity anomaly striking from southwest to northeast, implying it is less dense than the northwest and the southeast. The same feature can be observed from the Bouguer anomaly map in Fig. 2. It is inferred that the Moho interface in the center is slightly deeper than in the northwest and southeast area. In this layer, evidence for the existence of faults cannot be found any more. Therefore, the mean depth corresponding to the 5th-order residuals might be the maximum depth of main faults including F1, F2, F3 and F5.

According to the above analyses, it is concluded that the maximum depth of the fault is about 20 km for F1–F3 and F5, 12 km for F6 and F7, 6 km for F4 and F8, respectively. The main body of the Shenzhen faults is at the depth from 4 to 12 km. It can be confirmed from the epicenter distribution shown in Fig. 3. Most of the earthquakes since 1990 happened at about 6 km depth. Moreover, anomalies corresponding to the fault F5 and F6 move towards southwest with increasing depth (from Figs. 3a to 3c), relative to the fault line. It is inferred that the dip direction of the two faults is southwest. The faults F1-F3 might have a large dip angle because anomalies derived from them have no apparent shift with increasing depth.

3.2 Gravity Modeling

In order to provide insights into the crustal structure and the urban faults in Shenzhen, three gravity profiles aforementioned in Section 2 are selected to construct 2D underground model. A preliminary density model should be initialized for modeling. According to the seismic data, the crust in the study region can be divided into three layers (Yu, 2010). The surface layer is the Meso–Cenozoic continental deposit less than 3 km at depth with an average density of 2.22 g/cm3. The mean P-wave velocity in the surface layer is 4.92 km/s. The second layer is about 20 km thick, composed of metamorphic migmatite and granitoid. The average density is greater than 2.6 g/cm3 and the mean P-wave velocity is 6.13 km/s. The lowest layer corresponds to basalt with a thickness of 11 km. The low crust has a higher density ranging from 2.8 to 3.3 g/cm3, and a higher velocity of 6.83 km/s. In addition, a multilayer density model was derived from the multiscale decomposition of gravity data by Xu et al. (2015). The mean density at the depth of 4, 6, 12, and 20 km were set to 2.6, 2.7, 2.8, and 2.9 g/cm3, respectively. Furthermore, the decomposed gravity anomalies in Section 3.1 also reveal some information about density interface in various layers. All the prior information is taken into account in the preliminary model.

The preliminary models used for the three profiles are given in Table 2. The crust is divided into eight layers in the preliminary models. The scattered gravity anomalies in Fig. 3a indicate density inhomogeneity is not significant at the depth less than 3 km. Therefore the density inhomogeneity in the first three layers can be ignored. The lowest layer from 23 to 34 km is also considered without density discontinuity. In the middle four layers, different numbers of density discontinuity were assigned according to the number of gravity anomaly transition zones crossed by each profile (see Fig. 3). The position of the density discontinuity was expressed using the distance along the profile.

Table 2 Preliminary density model for three profiles

During the modeling process, trial-and-error method in modifying density and depth is applied to fit the gravity observations until the residual is satisfied. The modeling results are shown in Figs. 4, 5, and 6. Finally, the root-mean-square value of the residual is 0.32 mGal for profile P1-P1', 0.31 mGal for P2-P2' and 0.39 mGal for P3-P3', respectively.

Download:
Figure 4. 2D gravity model of profile P1-P1'. The upper one is the observed and calculated bouguer gravity anomaly along the profile. The lower one is gravity model constructed from bouguer gravity. Densities are labeled in the model and the unit of density is g/cm3.
Download:
Figure 5. 2D gravity model of profile P2-P2'. The upper one is the observed and calculated bouguer gravity anomaly along the profile. The lower one is gravity model constructed from bouguer gravity. Densities are labeled in the model and the unit of density is g/cm3.
Download:
Figure 6. 2D gravity model of profile P3-P3'. The upper one is observed and calculated bouguer gravity anomaly along the profile. The lower one is gravity model constructed from bouguer gravity. Densities are labeled in the model and the unit of density is g/cm3.

The gravity anomaly of the profile P1-P1' ranges from -8.8 to -22.1 mGal, decreasing gradually from northwest to southeast (Fig. 4). The gravity model shows that the decreasing trend can be mostly attributed to density decrease in NW-SE direction in the 6th and 7th layer between 7 and 23 km. The dip direction of the density discontinuity surface is SE in the northwest and NW in the southeast, consisting with the regional tectonic background. The NE-trending faults F1–F4 are located in the southeastern end of the profile. Their locations exactly confirm with the density discontinuity surfaces. In the upper crust from 3 to 7 km, the density in the area between F1 and F3 is higher than that in the two sides. It reveals that high-density materials upwelled along the fracture zone. In the model of P1-P1', the dip angles of F1–F4 are determined as 78.1°, 78.0°, 73.6° and 57.0°, respectively. The present results are consistent with the varying interval estimated by Yu (2010) (Table 3).

Table 3 Dip angles of the main faults in Shenzhen estimated by gravity modeling

The gravity anomaly of the profile P2-P2' varies from -16.7 to -22.5 mGal with several relative highs and lows. The relative gravity high near 30 km distance along the profile is located in Longgang Basin and Pingshan Basin, where faults F1–F3 traverse. Modeling results of P2-P2' are similar to those of P1-P1'. The location of faults F1–F4 also matches the density discontinuity surface. The density between F1 and F3 is obviously higher than other areas. Estimated dips of the four NE-striking faults are very close to the results in the model of P1-P1' (Table 3). However, the lateral density inhomogeneity along the profile is not apparent under 17 km depth.

The gravity anomaly of the profile P3-P3' ranges from -16.0 to -24.4 mGal. A low gravity anomaly exists in the southwest of the profile, corresponding to the relative low-density Yanshanian rocks. The minimum is located at Yangtai Mountain. The anomaly changes smoothly in the northeast part of the profile. Gravity model shows that the dip direction of most density discontinuity surface in the crust is southwestward. The density anomaly beside the NW-striking faults is comparatively small. The dips of F5–F8 are 51.4°, 75.0°, 54.0° and 70.3°, respectively, which are consistent with the results of Yu (2010).

4 CONCLUSIONS

Few studies on the urban faults in Shenzhen have been conducted based on gravity data in the past. In this paper, we have constructed a 1 km×1 km grid of Bouguer gravity anomaly. Multilayer anomalies were calculated using MSA method in order to analyze the crustal structures in Shenzhen fault zone, and three 2D gravity models were inversed. The work illustrates a gravimetric method to understand the Shenzhen fault zone. The Bouguer gravity anomaly shows an NE-striking high-low-high pattern from northwest to southeast. There is a strong correlation between gravity anomaly and main faults. The correlation is particularly significant from 4 to 12 km depth according to results of MSA. Positive anomalies can be observed along the faults in the upper layers, due to the upwelling of high-density materials. The residual gravity in each layer changes slightly with a small amplitude. It indicates that the tectonic activity is very weak in the study area. Furthermore, the multilayer gravity anomaly map (Fig. 3) can also provide important information about the faults, such as the vertical extension and the dip direction. In general, NE-striking faults extend deeper than NW-striking faults and have a larger dip angle. The gravity models we constructed also confirm the result. According to our study, features revealed by the multilayer gravity anomalies can be used as prior information for gravity modeling, which is helpful to improve the crustal density structure. In the future, 3D gravity modeling is expected to facilitate works on this topic. It is worth noting that although our method has only been applied to the urban faults in Shenzhen, it can also be used to investigate crustal structures anywhere.

ACKNOWLEDGMENTS

The authors would like to express their sincere thanks to the Urban Planning Land and Resources Commission of Shenzhen Municipality for supplying the gravity data and China University of Geosciences (Wuhan) for supplying GMS3.0 software. Thanks go to two reviewers for their constructive comments, which improved the manuscript. This study was supported by the National Natural Science Foundation of China (Nos. 41504015, 41429401), the National 973 Project of China (No. 2013CB733302), China Postdoctoral Science Foundation (No. 2015M572146), the National High Technology Research and Development Program of China (No. 2011AA060503), and the Surveying and Mapping Basic Research Program of National Administration of Surveying, Mapping and Geoinformation (No. 15-01-08). The final publication is available at Springer via https://doi.org/10.1007/s12583-017-0770-4.


REFERENCES CITED
Abbott, R. E., Louie, J. N., 2000. Depth to Bedrock Using Gravimetry in the Reno and Carson City, Nevada, Area Basins. Geophysics, 65(2): 340-350. DOI:10.1190/1.1444730
Abdelrahman, E. S. M., El-Araby, T. M., 1993. A Least-Squares Minimiza-tion Approach to Depth Determination from Moving Average Residual Gravity Anomalies. Geophysics, 58(12): 1779-1784. DOI:10.1190/1.1443392
Audru, J. C., Bano, M., Begg, J., et al., 2001. GPR Investigations on Active Faults in Urban Areas:The Georisc-NZ Project in Wellington, New Zealand. Comptes Rendus Geosciences, 333(8): 447-454.
Bansal, A. R., Dimri, V. P., 2001. Depth Estimation from the Scaling Power Spectral Density of Nonstationary Gravity Profile. Pure and Applied Geophysics, 158(4): 799-812. DOI:10.1007/PL00001204
Bhosle, B., Parkash, B., Awasthi, A. K., et al., 2007. Remote Sensing-GIS and GPR Studies of Two Active Faults, Western Gangetic Plains, India. Journal of Applied Geophysics, 61(2): 155-164.
Blakely, R. J., Christiansen, R. L., Guffanti, M., et al., 1997. Gravity Anomalies, Quaternary Vents, and Quaternary Faults in the Southern Cascade Range, Oregon and California:Implications for Arc and Backarc Evolution. Journal of Geophysical Research:Solid Earth, 102(B10): 22513-22527. DOI:10.1029/97JB01516
Chen, W., Zhao, H., Li, F., et al., 2001. Fault Activities and Their Influence upon Geologic Environment in Shenzhen City. Tropical Geography, 21(1): 45-50.
Cianciara, B., Marcak, H., 1976. Interpretation of Gravity Anomalies by Means of Local Power Spectra. Geophysical Prospecting, 24(2): 273-286. DOI:10.1111/gpr.1976.24.issue-2
Diaferia, I., Barchi, M., Loddo, M., et al., 2006. Detailed Imaging of Tectonic Structures by Multiscale Earth Resistivity Tomographies:The Colfiorito Normal Faults (Central Italy). Geophysical Research Letters, 33(9): L09305.
Fuis, G. S., Ryberg, T., Godfrey, N. J., et al., 2001. Crustal Structure and Tectonics from the Los Angeles Basin to the Mojave Desert, Southern California. Geology, 29(1): 15. DOI:10.1130/0091-7613(2001)029<0015:CSATFT>2.0.CO;2
Guerrieri, L., Leoni, G., Blumetti, A. M., et al., 2014. Fault Displacement Hazard in Urban Areas in Italy: A First Assessment. Proceedings of the 5th International INQUA Meeting on Paleoseismology, Active Tectonics and Archeoseismology (PATA). 21-27 September, 2014, Busan, Korea. 43-46
Hinsch, R., Decker, K., Wagreich, M., 2005. 3-D Mapping of Segmented Active Faults in the Southern Vienna Basin. Quaternary Science Reviews, 24(3/4): 321-336.
Hou, Z., Yang, W., 1997. Wavelet Transform and Multi-Scale Analysis on Gravity Anomalies of China. Acta Geophysica Sinica, 40(1): 85-95.
Huang, Y., Zhang, K., 1990. Some Characteristics of Neotectonic Movements in the Lianhuashan Fault Zone, Guangdong. South China Seismological Journal, 10(2): 25-34.
Iwahashi, J., 2010. 1:25 000-Scale Active Fault Map in Urban Areas Published by GSI. Bulletin of Geospatial Information Authority of Japan, 58: 29-37.
Jiang, W. L., Zhang, J. F., Tian, T., et al., 2012. Crustal Structure of Chuan-Dian Region Derived from Gravity Data and Its Tectonic Im-plications. Physics of the Earth and Planetary Interiors, 212/213(8): 76-87.
Karastathis, V. K., Ganas, A., Makris, J., et al., 2007. The Application of Shallow Seismic Techniques in the Study of Active Faults:The Atalanti Normal Fault, Central Greece. Journal of Applied Geophysics, 62(3): 215-233.
Li, S. L., Mooney, W. D., Fan, J. C., 2006. Crustal Structure of Mainland China from Deep Seismic Sounding Data. Tectonophysics, 420(1/2): 239-252.
Ma, H., Chen, P., 2009a. Quaternary Activity of Guanlan Fault in Shenzhen City. South China Journal of Seismology, 29(3): 17-24.
Ma, H., Chen, P., 2009b. The Quaternary Activity of Henggang-Luohu Fault in Shenzhen City. Technology for Earthquake Disaster Prevention, 4(3): 267-274.
Mareschal, J. C., 1985. Inversion of Potential Field Data in Fourier Transform Domain. Geophysics, 50(4): 685-691. DOI:10.1190/1.1441943
Moreau, F., Gibert, D., Holschneider, M., et al., 1999. Identification of Sources of Potential Fields with the Continuous Wavelet Transform:Basic Theory. Journal of Geophysical Research:Solid Earth, 104(B3): 5003-5013. DOI:10.1029/1998JB900106
Pamukçu, O., Gönenç, T., Uyanik, O., et al., 2014. A Microgravity Model for the City of İzmir (Western Anatolia) and Its Tectonic Implementations. Acta Geophysica, 62(4): 849-871. DOI:10.2478/s11600-014-0203-z
Sato, H., Ito, K., Abe, S., et al., 2009. Deep Seismic Reflection Profiling Across Active Reverse Faults in the Kinki Triangle, Central Japan. Tectonophysics, 472(1/2/3/4): 86-94.
Selim, E. S., Aboud, E., 2012. Determination of Sedimentary Cover and Structural Trends in the Central Sinai Area Using Gravity and Magnetic Data Analysis. Journal of Asian Earth Sciences, 43(1): 193-206. DOI:10.1016/j.jseaes.2011.09.010
Slater, L., Niemi, T. M., 2003. Ground-Penetrating Radar Investigation of Active Faults along the Dead Sea Transform and Implications for Seismic Hazards within the City of Aqaba, Jordan. Tectonophysics, 368(1/2/3/4): 33-50.
Spector, A., Grant, F. S., 1970. Statistical Models for Interpreting Aeromagnetic Data. Geophysics, 35(2): 293-302. DOI:10.1190/1.1440092
Sultan, Araffa S. A., Monteiro, Santos F. A., Arafa-Hamed, T., 2012. Delineating Active Faults by Using Integrated Geophysical Data at Northeastern Part of Cairo, Egypt. NRIAG Journal of Astronomy and Geophysics, 1(1): 33-44. DOI:10.1016/j.nrjag.2012.11.004
Sun, J., Jia, J., Zhan, W., et al., 2007. A Study of the Tectonic Activity of Shenzhen Fracture Zone. Advances in Earth Science, 22(3): 234-240.
Suski, B., Brocard, G., Authemayou, C., et al., 2010. Localization and Characterization of an Active Fault in an Urbanized Area in Central Guatemala by Means of Geoelectrical Imaging. Tectonophysics, 480(1/2/3/4): 88-98.
Syberg, F. J. R., 1972. A Fourier Method for the Regional-Residual Problem of Potential Fields. Geophysical Prospecting, 20(1): 47-75. DOI:10.1111/gpr.1972.20.issue-1
Tan, C., Sun, Y., Wang, R., 2000. Present Day Activity of the Shenzhen Fault Zone and Its Impact on the Safety of a Planned Diversion Tunnel in Shenzhen, China. Engineering Geology, 57(1/2): 73-80.
U.S. Geological Survey (USGS), 2006. Quaternary Fault and Fold Database of the United States.[2015-9-15]. http://earthquake.usgs.gov/haz-ards/qfaults/
Wang, H., 2005. Research on Applications of Wavelet Multiscale Analysis in the Earth's Gravity Field: [Dissertation]. Wuhan University, Wuhan (in Chinese with English Abstract)
Wang, Z., Zhu, Z., Yi, S., 2005. Finite Elements Analysis of Rheological Deformation of Huangbeiling F8 Fault Age in Shenzhen Luohu Fault Zone. Rock and Soil Mechanics, 26(Suppl): 211-214.
Xu, C., 2014. Study on Multi-Scale Gravity Inverse Method and Its Appli-cation to Detect Urban Active Faults: [Dissertation]. Wuhan University, Wuhan (in Chinese with English Abstract)
Xu, C., Wang, H. H., Luo, Z. C., et al., 2015. Multilayer Stress from Gravity and Its Tectonic Implications in Urban Active Fault Zone:A Case Study in Shenzhen, South China. Journal of Applied Geophysics, 114(8): 174-182.
Xu, W., Wang, J., Shi, P., et al., 2004. Hazard Degree Assessment of Urban Earthquake Disaster in China. Journal of Natural Disasters, 13(1): 9-15.
Xu, Y., Hao, T. Y., Li, Z. W., et al., 2009. Regional Gravity Anomaly Separation Using Wavelet Transform and Spectrum Analysis. Journal of Geophysics and Engineering, 6(3): 279-287. DOI:10.1088/1742-2132/6/3/007
Yalçiner, C. Ç., Altunel, E., Bano, M., et al., 2013. Application of GPR to Normal Faults in the Büyük Menderes Graben, Western Turkey. Journal of Geodynamics, 65(2): 218-227.
Yu, C. H., 2010. Research on the Faults Activity and Seismic Hazard in Shenzhen: [Dissertation]. Zhejiang University, Zhejiang (in Chinese with English Abstract)
Zamani, A., Samiee, J., Kirby, J. F., 2013. Estimating the Mechanical Anisotropy of the Iranian Lithosphere Using the Wavelet Coherence Method. Tectonophysics, 601(B14): 139-147.
Zhang, J., Wang, C. Y., Shi, Y. L., et al., 2004. Three-Dimensional Crustal Structure in Central Taiwan from Gravity Inversion with a Parallel Genetic Algorithm. Geophysics, 69(4): 917-924. DOI:10.1190/1.1778235