Shear-wave splitting in the crust: Regional compressive stress from polarizations of fast shear-waves

1.   Introduction

Earth is naturally anisotropic. Seismic anisotropy is widely observed in the crust (Crampin, 1978) and the mantle (Fukao, 1984). However the crustal anisotropy is generally much more complicated possibly due to short wavelength structure and stress variations within a thin crustal layer. Information about crustal anisotropy, however, is important for interpretations of various observations of crustal activities (Crampin, 1978; Zhang et al., 2000). Seismic anisotropy in the crust typically originates from stress-aligned fluid-saturated grainboundary cracks and pore throats, which have been called extensive-dilatancy anisotropy (EDA) cracks(Crampin et al., 1984). The EDA-cracks tend to be aligned perpendicular to the direction of minimum compressive stress. At depths below the critical depth (usually between 500 and 1 000 m), where the increasing vertical compressive stress equals to the minimum horizontal compressive stress, the minimum stress is horizontal and cracks or microcracks tend to be vertical with strikes parallel to the maximum horizontal stress.

Shear-waves propagating through anisotropic rocks in the crust will split into faster and slower components with almost orthogonal polarizations (Crampin, 1981), that is shear-wave splitting (SWS). Theoretical studies demonstrate that stress-aligned SWS is the most diagnostic phenomenon for recognizing anisotropy in rocks in the crust (Crampin, 1978, 1981, 1994; Hudson, 1981; Thomsen, 1995). SWS is highly sensitive to the crack distribution and orientations in response to the applied stress system (Crampin, 1978). This means that orientations of the polarizations of fast shear-waves (PFS) generally provide information about the anisotropic sym-metry and stress directions (Crampin and Peacock, 2005; Gao et al., 1995; Zhang et al., 2009). However the stress can change anisotropic structure and symme try, and can make the EDA-crack-induced anisotropy evolve, which will further change stress distribution. Its process can be simulated theoretically by the anisotrop ic poro-elasticity model (Zatsepin and Crampin, 1997; Crampin and Zatsepin, 1997). It suggests that observed crustal seismic anisotropy could indicate the stress state and stress change (Gao and Wu, 2008; Gao et al., 2010; Gao and Crampin, 2008). Accordingly, SWS has been studied extensively to obtain information on local (or in situ) stress state and stress orientation (Crampin, 1978; Li et al., 1994; Gao et al., 1995, 1999; Crampin and Peacock, 2005; Mizuno et al., 2005; Musumeci et al., 2005; Cochran et al., 2006; Shi et al., 2009a, b) and information on regional stress field and stress-buildup (and/or stress relaxation) processes (Gao et al., 1998; Gao and Crampin, 2003, 2004, 2006, 2008; Gerst and Savage, 2004; Crampin and Peacock, 2005; Wu et al., 2007; Gao et al., 2011). It is definite that SWS could be adopted to infer regional stress field as well as local stress disturbance due to active faults, only if we have enough dense regional seismography networks and enough data for SWS analysis (Gao et al., 2011).

2.   Principles and methods

SWS can be observed within the shear-wave window for almost all small earthquakes (Figure 1). The shear-wave window at a horizontal free-surface is the solid angle of incidence directions with radius arcsin(v_S/v_P)≈35° (for a Poisson's ratio of 0.25), within which the apparent velocity of shear-waves, parallel to the surface, is so great that S-to-P conversions cannot occur for incident SV-waves on a (horizontal) free-surface (Crampin and Peacock, 2005). The faster shear-wave is generally polarized approximately parallel to the orientation of maximum horizontal stress. Such stress-aligned SWS is caused by the distributions of stress-aligned fluid-saturated parallel vertical cracks or microcracks, i.e. EDA cracks, pervading almost all in situ rocks at depths below the critical depth (Crampin, 1994). It is a kind of transverse isotropy (hexagonal symmetry) with a horizontal axis of symmetry (TIHanisotropy). This anisotropic symmetry system, or a minor variation, could display such parallel PFS, within a large part of shear-wave window (Crampin, 1981; Gao and Crampin, 2006). Distribution of stress-aligned EDA cracks or microcracks is the best rock mechanism to describe the geological configuration with such symmetry phenomena, common to both crystalline and sedimentary rocks (Crampin, 1994). Therefore, the PFS becomes an access to determination of the stress in the crust.

Figure  1.  An example of shear-wave splitting (SWS) analysis. A small M0.81 earthquake was recorded within the shear-wave window at station KRI in the SIL seismic network in Iceland. The time axis of the 100 samples/s seismogram is in seconds and it is filtered with a Butterworth 3.0–50.0 Hz bandpass. Vertical bars spanning the S-wave arrivals (from SIL earthquake catalogue data) mark (0.1 s) time-intervals in Figure 2. The top-right header describes information of basic parameters and analysis results for this event by a semi-automatic SWS program based on an expert system technique (Gao et al., 2006). From top to bottom are respectively EW, NS, vertical and two rotated components. Vertical long red bars point to arrivals of fast (right) and slow (left) shear-waves.

DownLoad: Full-Size Img PowerPoint

There have been many crustal SWS measurement techniques. However, at the present time, the best is still the semi-automatic visual measurements (Crampin and Gao, 2006). Automatic measurement is much quicker, but there exist lots of mis-estimations which results in wrong conclusions. The technique of polarization diagram is a basic method (Crampin, 1978). Figure 2 is an example of polarization analysis, which is calculated by a program of SWS analysis using an expert system developed by Gao et al. (2006). This program needs all available parameters of small earthquakes, such as in Iceland. However under most conditions, due to lack of enough parameter service, a more adaptive analysis technique or program is necessary. We adopt a systematic analysis method (SAM) to analyze local seismic waveform data recorded in seismic networks (Gao et al., 1998, 2008). SAM includes three steps: calculation of cross-correlation functions, measurement and elimination of time-delays, and measurement of polarizations (Gao et al., 1998, 2008, 2011).

Figure  2.  Polarization analysis of the M0.81 earthquake at station KRI. Polarization diagrams are shown in three profiles, i.e., a horizontal and two vertical ones. The enlarged polarization diagram at bottom indicates clear abrupt changes in particle-motion directions with linear motion between the fast and slow shear-wave arrivals, where green straight line is fitting line. The analysis results are shown at the top, the same as Figure 1.

DownLoad: Full-Size Img PowerPoint

3.   Crustal compressive stress from PFS data

3.1   North China

There were once a Capital Area Seismograph Network (CASN) at the north and middle of North China, which ran from 1 October 2001, consisted of 107 seismograph stations. It was the largest, most closely-spaced regional seismograph network in China, with 500 km in east-to-west, 400 km in north-to-south. Now CASN has been separated into several regional seismic networks, including Beijing, Tianjin and Hebei. The records of 2 548 local events well located nearby stations are examined in a period from January 2002 to August 2005. Finally, we obtain 513 records of good waveforms from 64 stations for SWS analysis (Gao and Wu, 2008; Gao et al., 2011; also see Wu et al., 2007, 2009). PFS could be obtained by SWS analysis. Figure 3a shows a superimposed equal-area rose diagram of all PFS measurements from 64 stations, indicating a predominant nearly EW orientation (average N86°E±41°). PFS are critically dependent on ray path angles to the free surface and hence are critically dependent on surface and subsurface topography, and the large standard errors suggest the influence of the complicated tectonics, although they should be almost parallel to each other within shear wave window, at least at large part of shear-wave win dow. The study area consists of three active neo-tectonic elements, Yanshan uplift, Taihang uplift and North Chi na basin. These 64 stations are mainly in the North China basin around the two big cities of Beijing and Tianjin, partly in the Taihang uplift and at the western boundary of the Yanshan uplift, and partly around the Shijiazhuang city in the conjunction zone between the Taihang uplift and the North China basin (see Figure 3b).

Figure  3.  Fast polarization of shear-wave in the crust in North China.

(a) PFS equal-area projection rose diagram. (b) Spatial smooth fitting of all PFS data, shown in short yellow straight lines. Red triangles are stations. White arrows indicate the direction of the maximum compressive strain from GPS data, black arrows indicate the orientation of regional compressive stress from focal mechanism and borehole drilling data (Zhang et al., 2004), and red arrows indicate the direction of the maximum compressive stress deduced from PFS data (Gao et al., 2011). Black dots are cities.

DownLoad: Full-Size Img PowerPoint

Figure 3b shows spatially smooth fitting to all PFS data and the present compressive stress field distribu tion in North China, which indicates the influence of both regional compression and local compressive field. For example, the compressive stress at about 40.0°N and l16.5°E, around Beijing city, is nearly EW. Howev er, at 50–100 km east to Beijing, the compressive stress directions are distributed along NE, which suggests the influence of local faults which strike NE orientation. At about 100–150 km, W and NW to Beijing, the compressive stresses are along NW or nearly WNW, which again suggests the influence of local faults striking WNW. Note that there are several faults near Beijing striking NE and the tectonic line between Taihang uplift and North China basin also strikes NE or NNE, there is also WNW faults through this zone (also see Gao et al., 2011). It suggests overlapped or compositive influences of faults, tectonics even heterogenous structure on crustal stress in this area.

In North China, there is a Zhangjiakou-Bohaistrong earthquake occurrence zone, also called the Zhangjiakou-Bohai tectonic belt or sometimes Zhangjiakou-Penglai tectonic belt (e.g., Xu et al., 2002). There are also full of Holocene and late-Pleistocene ac tive faults, early- and medium-Pleistocene faults, and pre-Quaternary fault, in striking of NE (or ENE) and NW (or WNW), widely through the uplifts and the North China basin. In general, these complicated tec tonics results in complicated pattern of PFSs (also see Wu et al., 2007; Gao et al., 2011). In reverse, com plicated distribution of orientations of PFSs suggests complexity of tectonics, therefore complexity of stress field.

3.2   Longmenshan fault zone (east edge of Ti betan plateau)

The Longmenshan fault zone is located at the middle of the south-to-north seismic zone in the Chi nese mainland. Its southeast is stable Sichuan basin, its northwest is steep uplift of Qinghai-Xizang plateau (Tibetan plateau), where is the east edge of Qinghai-Xizang plateau. The M_S8.0 Wenchuan earthquake occurred on 12 May 2008 at Longmenshan fault zone. Since the pe riod of the Indosinian movement, especially due to the Himalayan tectonic evolution, with the Qinghai-Xizang plateau intensely lifts, there were collisions and extru sions of continental crusts in different degrees. These tectonic movements formed the large scale of Longmen shan fault zone with mainly thrust mechanism and slight dextral strike-slip. According to the focal mecha nisms, the orientation of contemporary principal com pressive stress axis is at WNW direction on the Long menshan fault zone (Deng et al., 1994). The direction of maximum horizontal principal stress obtained by in situ hydrofracturing stress measurement is at NW direction, with slightly difference in each segment on the Long menshan fault zone (An et al., 2004). Longmenshan fault zone is large scale reverse faults.

After 2008 M_S8.0 Wenchuan earthquake, many temporary seismic networks were set up around Long menshan fault zone. It provides us with a chance to detect the crustal anisotropy along the faults by using data from temporary seismic networks plus regional per manent seismic networks. After strict data selection of shear-wave window and high-quality waveforms, 2 068 available SWS data from 24 seismic stations are used to obtain SWS parameters (Shi et al., 2009a). Accord ing to results of PFSs, the study area could be divided into two parts along the Longmenshan fault zone (Figure 4). In southwest part (segment B in Figure 4a), predominant PFS is basically in NW direction (see Figure 4b), which is consistent to the directions of regional principal compressive stress and regional principal GPS compressive strain. They are perpendicular to the strike of the faults, coinciding with the reverse Longmenshan fault zone. At northeast part (segment A in Figure 4a), predominant PFS is basically in ENE direction (see Figure 4b), where the Longmenshan fault zone is of strikeslip characteristics. It seems nearly perpendicular to the direction of principal GPS compressive strain and the direction of regional principal compressive stress before the earthquake occurred. It is consistent with the strikes of the Longmenshan fault zone, coinciding with horizontal P axis distribution by the focal mechanisms of local strong aftershocks (Hu et al., 2008). It suggests, at least after Wenchuan M_S8.0, the orientation of regional principal compressive stress is divided into two parts. Along Longmenshan fault zone, principal compressive stress is respectively in NW in southwest part and in ENE in northeast part (Shi et al., 2009a). However the fact is that, predominant NW orientation of PFS in southwest part is consistent with local and regional stress field, consistent with focal mechanism of large Wenchuan M_S8.0 earthquake; predominant ENE orientation of PFS in northeast part coincides with strike of faults at northeast of Longmenshan fault zone, which gradually changes into strike-slip, also coincides with imaging of rupture process.

Figure  4.  The PFS directions at seismic stations along the Longmenshan fault zone, southwestern China. Data came from aftershocks of Wenchuan M_S8.0 earthquake (Shi et al., 2009a).(a) Black dashed straight line divides the studied zone into segments A and B. (b) PFS equal-area projection rose diagrams for two segments.

DownLoad: Full-Size Img PowerPoint

3.3   Yunnan region (southeast edge of Tibetan plateau)

Yunnan is a province in southwestern China, which is located in southeast edge of Tibetan plateau. Due to strong collision between the Indian plate and the Eurasian plate, there are many large active faults and lots of strong earthquakes with magnitudes larger than M_S5.0 in this area. Shi et al. (2009b) used 15 stations of Yunnan Seismic Networks to study the crustal anisotropy by SWS using data from January 2000 to May 2005. Now Yunnan Seismic Networks consists of more stations. Here, we analyze data from January 2000 to October 2010, recorded by 32 stations in this area. Shi et al. (2012) used these data to study seismic anisotropy. In this study, we discuss crustal stress deduced by PFSs of SWS.

Figure 5 shows spatial smooth fitting of PFSs obtained by SWS in the crust. Most faults are in strike of nearly NS in the study area, but some of them are in strike of NE in north and some of them are in strike of NW in south (Figure 5). The topography generally is higher in the north and lower in the south. The complicated tectonics directly influences crustal stress. See blue short straight lines in Figure 5, they are distribu tion of orientations of principal compressive stress field, obtained in this study. By Shi et al. (2012), the equal area projection rose diagram for all available records shows predominant PFS in NNW direction, which suggests NS or NNW direction of principal compressive stress. It matches the results of Han et al. (1977). However we must carefully analyze the deduced compressive stress direction because there are lots of irregular faults across, intense change of topography, different tectonic sub-zones cut by large faults. In order to understand relationship between deformation in the crust and in the mantle (at least in the upper mantle), we also image the lithospheric anisotropy by XKS (SKS, SKKS, PKS phases) splitting. This study uses more data and more phases although there was basic results only by SKS splitting (Chang et al., 2006). Figure 6 shows spatial smooth fitting of fast wave polarizations of XKS. XKS travels through the crust and the whole mantle. However studies indicated XKS can image azimuthal anisotropy in the upper mantle and the crust, which mainly focuses on the lithosphere and the upmost asthenosphere (Liu et al., 2008). See red short straight lines in Figure 6, they are distribution of directions of fast wave polarizations of XKS, obtained in this study. Clear stream-like traces indicate deep deformation or upper mantle stream. The directions are absolutely dissimilar to each other between PFSs in the crust and those of XKS through the upper mantle and the crust. Using only GPS and SKS data, Wang et al. (2008) concluded the decoupling between the crust and the mantle in this area. Although it was not enough sometimes by only GPS data and SKS splitting data to make conclusions on coupling or decoupling (Gao et al., 2010), however results in this study generally support this con clusion. Due to the decoupling between the crust and the mantle, the crustal stress or crustal deformation is basically independent on mantle deformation. In southwest China, at least in Yunnan, complicated compressive stress in Figure 5 suggests crustal stress is controlled mainly by tectonics, faults and crustal structure.

Figure  5.  Spatial smooth fitting of PFSs in Yunnan of China, southeast edge of Tibetan plateau. Black triangles are seismic stations.

DownLoad: Full-Size Img PowerPoint

Figure  6.  Spatial smooth fitting of fast wave polarizations of XKS phases traveling through whole mantle in Yunnan of China, southeast edge of Tibetan plateau. Black triangles are seismic stations.

DownLoad: Full-Size Img PowerPoint

3.4   Stress change from SWS in the crust

Lots of studies reported that SWS could indicate stress change with seismicity or volcano eruption (Crampin et al., 1990, 2003; Booth et al., 1990; Gao et al., 1998; Gao and Crampin; 2003, 2004, 2006; Bianco et al., 2006). One of the typical examples is temporal change of time-delay of SWS before a M5.0 earthquake in 1998 in Iceland (Crampin et al., 1999). The PFS can change in orientation with volcano activity (Miller and Savage, 2001; Gerst and Savage, 2004; Bianco et al., 2006). The rock experiment also verifies SWS is sensitive to indicate the change of differential stress (Gao, 2001; Gao and Crampin, 2003). Because of change of time-delay of SWS with stress or seismicity, it is applicable to earthquake forecasting (Crampin et al., 2008; Gao and Crampin, 2008; Crampin and Gao, 2010). Here we do not discuss temporal change of time-delay of SWS in detail, but focus on just change of PFS.

In November 29, 1999, a M_L5.7 (M_S5.9) earthquake occurred in Xiuyan of Liaoning province, northeastern China. A seismic station YK near to Xiuyan earthquake, about 50 km away, recorded seismic waveforms of small earthquake events before and after the M_L5.7 mainshock of Xiuyan earthquake sequence. In order to discuss possible temporal variation of PFS, histogram of monthly average PFS at station YK is shown in Figure 7. Because of lack of data in the period of about one month immediately before mainshock, there is no chance to study detailed change of SWS parameters immediately before a strong earthquake. However, it could still be seen that there seems to be about 10°–15° PFS change within 3 months before the mainshock of Xiuyan earthquake.

Figure  7.  Average PFSs monthly at station YK. The Xiuyan earthquake occurred on November 29, 1999. The ordinate denotes average PFS in degree and the abscissa denotes time in year and month.

DownLoad: Full-Size Img PowerPoint

In an early study, Gao and Feng (1993) once used near-field earthquake events with magnitude larger than 3.0 to study SWS in Datong, North China, and figured out orientation change of regional principal compressive stress by PFSs. The mainshock of magnitude M_S5.8, occurred on October 19, 1989, is about 50 km away from a seismic station in Datong. The shear-waves used in the study were possibly Sn phase since the epicentral distances of source events were about 200 km. Because of weak shear-wave phase, the study adopted polarization diagram plus filter in data analysis (Gao and Feng, 1993). It is early preliminary discussion of temporal change in polarization pattern although reliability requires further confirmation.

4.   Conclusions

It is certain that anisotropic crust results in SWS and SWS parameters also indicate information of structure and stress. Since PFS is always parallel to the orientation of in situ principal compressive stress, naturally PFS could be applied to study stress in the crust. In the zone where there is a large quantities of dense seismic stations, it is possible to deduce local or regional compressive stress in the crust. At the first time, PFS is adopted to study compressive stress in North China (Gao and Wu, 2008). Now it is also applied to discuss southwest zone in this study. These results match those of other independent measurements using traditional stress measurement techniques (Zhang et al., 2004). It is acceptable to use PFS data to deduce the compressive stress in some zones, such as North China. But, in extremely-complicated tectonic zones, such as in Yunnan region, southeast edge of Tibetan plateau, it needs carefully application by dividing the region into smaller tectonic sub-zones under excellent data coverage. Temporal changes of SWS parameters suggest change of stress. However the biggest limitation of SWS technique is the difficulty in recording enough available data at seismic stations. Lots of regional dense seismic networks provide opportunity to detect stress by SWS analysis, although we know that lots of factors, such as category of crustal rocks, composition of liquids in the crust, dis tribution and natures of faults, influence stress in the crust. At least, it is a new path to discuss or detect stress in the crust.