Corresponding author: M. Xu, School of Earth Sciences and Engineering, Nanjing University, 22th Hankou Rd., Nanjing 210093, China. (email@example.com); (firstname.lastname@example.org)
 A dense seismic array consisting of 28 temporary stations was deployed to study the crustal and upper-mantle deformations beneath eastern China. We measured the splitting parameters in the crust and mantle by analyzing P-to-S phases converted at the Moho discontinuity (called PmS phases) and the core-mantle boundary (i.e., core phases), respectively. The splitting parameters of core phases are retrieved at most stations while that of the PmS phases are retrieved at only a few stations. Distinct lateral variations of the fast polarization directions analyzed with the core phases are found in different tectonic blocks in eastern China. The delay times in the mantle and crust are moderately large (~1 s) and averagely smaller than 0.3 s, respectively. By the Fresnel-zone analysis, the laterally variant lithospheric anisotropy is revealed between the two sub-blocks (Southeast China Orogenic Belt and Yangtze Craton) of the eastern South China Block. In contrast, in the southeastern North China Craton, the anisotropy in a relatively deep layer contributes to the splitting observations.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Seismic anisotropy due to tectonic deformation is an ubiquitous property of the Earth's interior [Silver, 1996; Long and Silver, 2009]. In the upper crust the anisotropy is mainly caused by the alignment of vertically parallel microcracks [e.g., Crampin, 1984]. On the other hand, it is caused by the strain-induced lattice preferred orientation (LPO) of intrinsically anisotropic minerals in the lower crust (e.g., mica, amphibole) or the upper mantle (mainly olivine) [e.g., Karato et al., 2008]. Due to these relations between anisotropy and strain/stress states, anisotropy is undoubtedly a good indicator for the tectonic deformational processes.
 Shear-wave splitting is one of the most effective tools for characterizing anisotropy in the Earth [Silver, 1996; Savage, 1999; Long and Silver, 2009]. This phenomenon occurs when the shear wave travels through the anisotropic medium, resulting in the shear wave splitting into two orthogonally polarized quasi-shear waves with different velocities. Usually two splitting parameters are measured: the fast quasi-shear wave polarization direction “ϕ”, which indicates the geometry of anisotropy and the delay time “δt” between the two waves, which is related to the length of the paths the shear wave travels in the medium and the strength of anisotropy. Generally, the measured ϕ will be parallel with the a-axis of olivine and subparallel with the horizontal mantle flow direction or to the extension direction under dry conditions of olivine in the upper mantle [Zhang and Karato, 1995; Savage, 1999]. In a similar way, the LPO of crustal minerals such as amphibole in the form of ductile deformation or flow can also cause the fast direction parallel with the flow direction [e.g., Meissner et al., 2006].
 Continental regions are complex cases for studying seismic anisotropy because both anisotropy in the lithosphere and asthenosphere may contribute to the shear-wave splitting observations [e.g., Huang et al., 2011a]. The anisotropy in the lithosphere may be formed due to the present-day tectonic activities or reflect the fossil strains formed during complex past tectonic processes [e.g., Silver and Chan, 1991; Silver, 1996]. In contrast, the asthenospheric anisotropy may be related to the present-day mantle deformation induced by the horizontal mantle flow, absolute plate motion (APM) or small-scale mantle convection [e.g., Long and Silver, 2009]. Eastern China is located at the eastern margin of the Eurasian plate. It is composed of different tectonic blocks (Figure 1) that amalgamated before the Early Mesozoic. Subsequently it experienced multistage strong tectonic activities in the Mesozoic to Cenozoic [e.g., Li, 2000; Zhou and Li, 2000; Ren et al., 2002; Li and Li, 2007]. Therefore, the lithosphere may have experienced complex deformations during different geological periods. It is an interesting but very complex case for studying the anisotropy beneath eastern China.
 Many shear-wave splitting studies were carried out in eastern China [e.g., Huang et al., 2011a; Zhao et al., 2007, 2013]. Zhao et al. [2007, 2013] made shear-wave splitting measurements in eastern China and revealed the apparent diversity of fast directions in the adjacent tectonic domains. Huang et al. [2011a] measured core phase splitting parameters on large amounts of data to capture the main anisotropic characteristics in Mainland China. However, the results in eastern China are still sparse (Figure 1). The measured fast directions show diversity and complexity even at nearby locations. This may be caused by the limitation of the shear-wave splitting analysis in this particular region (will be discussed below), which hampers our understanding of the detailed upper mantle deformation beneath eastern China.
 To address this problem, we deployed a dense seismic array across eastern China (Figure 1). The array was designed in a line to be perpendicular to the trends of the main tectonic boundaries (Figure 1). With the data recorded by the array, we analyzed the splitting of the core phases (e.g., SKS) and the P-to-S converted phases at the Moho discontinuity (called PmS phases) to characterize the anisotropy beneath the stations. The present result provides new insights into the crustal and upper-mantle deformations beneath eastern China.
2 Geological Settings
 The study region is mainly located at the eastern South China Block (SCB) and partly at the southeastern North China Craton (NCC) (red dashed box, Figure 1). The SCB and the NCC collided to form the NWW-SEE trending Qinling-Dabie Orogenic Belt in the Early Mesozoic (Figure 1) [e.g., Ratschbacher et al., 2003]. The TanLu fault (TLF) is a large-scale NE-SW trending strike-slip fault that truncated at the east margin of the NCC (Figure 1) [Ren et al., 2002]. The SCB is subdivided into the Southeast China Orogenic belt (SCOB) and the Yangtze Craton (YC), which amalgamated in the Precambrian [e.g., Charvet et al., 1996]. The JiangShao fault was confirmed as the boundary between the SCOB and the YC from the magnetic anomaly, gravity measurements, and the wide-angle seismic profiling results [Zhang et al., 2005, and references therein].
 The significant geological features in the SCB are the widespread Late Mesozoic magmatism, which was induced by the NW/NWW-ward subduction of the paleo-Pacific plate [e.g., Zhou and Li, 2000; Zhou et al., 2006; Shu et al., 2009]. The change of the subduction angle was suggested to account for the temporal-spatial distributions and characteristics of Late Mesozoic igneous rocks [Zhou and Li, 2000; Zhou et al., 2006]. Inside the SCOB, the percentage of outcrops of Cretaceous volcanic rocks in the coastal area of Zhejiang province (east of the JiangShao fault) is particularly high, up to ~62% [Zhou and Li, 2000; Zhou et al., 2006]. On the other hand, the cratonic lithosphere of the eastern NCC was strongly destructed in the Mesozoic and thinned from ~200 km in the Archean to ~80 km in the present [e.g., Griffin et al., 1998; Zhao et al., 2001].
3 Data and Method
 A total of 28 temporary broadband seismic stations (yellow dots, Figure 1) were deployed across Anhui (AH) and Zhejiang (ZJ) provinces in eastern China by Nanjing University during December 2008 to April 2011. The stations were nearly in an NW-SE line across the important tectonic boundaries such as the TanLu fault and the JiangShao fault. Each station was equipped with a GMT-40 type, three-component seismometer and a Reftek 72A or 130 type digital acquisition system. The waveforms were recorded with the sampling rate of 40 sps. These stations operated in different periods with the effective recording times in the range of 4 to 16 months.
3.1 Core Phase Splitting
 The limitation of core phase splitting analysis is obvious in our study area, especially in the SCOB. Because most large earthquakes occurred at the Tonga subduction zone with the epicentral distances Δ around 85°, the SKS phases are likely contaminated by other shear phases. To reduce this limit, we checked many different core phases (e.g., SKS, PKS, SKKS, hereafter called XKS phases) from earthquakes with Δ > 85° and magnitudes Mw ≥ 5.9. Clear SKKS phases were found in the records of some earthquakes with extra-large distances (Δ > 165°) and were also included for analysis. Finally, XKS phases of 15 events (blue dots, inset map of Figure 1) yielded reliable shear-wave splitting measurements. The event backazimuths (Baz) are mostly in the ranges of 0°–50° and 110°–150°.
 The minimum energy method (SC) [Silver and Chan, 1991] and the rotation cross-correlation method (RC) [e.g., Bowman and Ando, 1987] were carried out simultaneously in the SplitLab software [Wüstefeld et al., 2008]. Both methods assume a single homogeneous layer of hexagonally symmetric material with a horizontal symmetry axis. The initial XKS waveforms were filtered with a third-order Butterworth band-pass filter. The upper and lower corner frequencies were manually adjusted (in the range of 0.02–0.08 Hz and 0.125–1 Hz, respectively) to optimum the waveform clarity and the reliability of results. The optimum pairs (ϕ, δt) were found by a grid search over trial ϕ and δt in the ϕ-δt domain by minimizing the energy of the reconstructed transverse (T) components (SC method) or maximizing the cross-correlation coefficient of the corrected fast and slow components (RC method). The grid-search was made with the increment of 1° for ϕ from –90° to 90° and 0.05 s for δt from 0 to 4 s. Errors with 95% confidence were estimated by the inverse F-test introduced by Silver and Chan . Figure 2 shows one example of the measurement with an SKS phase (Δ = 87.3°) at station AH04.
3.2 PmS Phase Splitting
 Because the XKS phase splitting can be attributed to the anisotropy anywhere along the wave path in the receiver side, we also analyzed the splitting of the PmS phases to estimate the contribution of anisotropy in the crust. To extract clear PmS signals, P wave receiver functions (RF) were calculated at each station in the time domain by the maximum entropy deconvolution method [Wu et al., 2003]. Gaussian low-pass filter with the parameter α equal to 2.5 or 3.0 was used to exclude high-frequency noise. Figure 3 shows the radial (R) and transverse (T) RFs at station ZJ05, for which coherent PmS phases with peak amplitudes at ~4.0 s on the R components are visible.
 The energy on the T-RFs may be caused by anisotropy, inclining layers or scattering from small-scale heterogeneities in the crust [McNamara and Owens, 1993; Savage, 1998]. For example in Figure 3, it is hard to distinguish these potential sources of energy on the T components. In the case of anisotropy the T component will be proportional to the time derivative of the R component (dR/dt) when the delay time is small compared with the dominant period [Chevrot, 2000]. Although the scattering can also yield elliptical particle motion, it will not cause the proportional characteristics between dR/dt and T waveforms [e.g., Alsina and Snieder, 1995]. Therefore during analysis we examined the similarity between the waveforms of dR/dt and T in the selected time window (e.g., Figure 4b), which was deemed as an auxiliary diagnostic of anisotropy to the initial elliptical particle motion. In the presence of noise signals, a third-order Butterworth band-pass filter within the range of 0.2–0.8 Hz was used to further enhance the signal-to-noise ratio (SNR) (e.g., Figures 4a and 4b). In total, 43 events were included in the PmS phase splitting analysis (red crosses, inset map of Figure 1).
 The best estimates of splitting parameters were obtained by the RC method because it is more stable compared with the SC method in the PmS phase splitting analysis [McNamara and Owens, 1993]. The grid-search was made with an increment of 1° for ϕ from –90° to 90° and 0.025 s for δt from 0 to 0.5 s. Figure 4 shows an example of the splitting measurement of a PmS phase at station AH01.
 In the analysis of XKS phases, we mainly rely on the SC results and inspect their consistence with the RC results as a criterion for reliability. Wüstefeld and Bokelmann  defined the quality of the measurements by checking the angular difference of ϕ (|Δϕ| = |ϕRC-ϕSC|) and the ratio of δt (ρ = δtRC/δtSC). Most of our non-null results have |Δϕ| smaller than 10° and some are in the range of 10°–23°, while ρ is in the range of 0.63–1.29 (e.g., Figure 2, listed in Table S1 in the Supporting Information). The null measurements, i.e., the incoming S wave polarization direction is (sub-) parallel or perpendicular to the fast axis of the anisotropic medium, have different results with the two methods and they generally follow the criterion: |Δϕ| = N∙45 ± 7° (N, positive integer) and 0 ≤ ρ ≤ 0.2 (Table S1) (see Wüstefeld and Bokelmann  for details). Finally, a total of 47 XKS splitting measurements (33 non-nulls, 14 nulls) are obtained (hereafter we call them (ϕm, δtm)) (see Supporting Information for all the measurements).
 The prominent features are the lateral variations of ϕm in the four subdomains of the study area: SCOB (mainly NE-SW), YC1 (NW-SE), YC2 (NEE-SWW), and eastern NCC (mainly NWW-SEE) (Figure 5). Note that for convenience station ZJ06 is classified into the YC1 for interpretation. The ϕm inferred from the null measurements generally agree well with the non-null measurements at most stations except for the SCOB region. In the SCOB, the Baz (~17°) of the null measurements (stations ZJ01–ZJ05) from event 2010.012 is neither parallel nor perpendicular to the non-null measurements at some stations (Figure 5). Being conservative about the measurement results, this may be due to the uncertainties of the individual measurements. Another possibility is the complex anisotropy beneath the SCOB (such as vertically anisotropic heterogeneities) which causes the observation deviating from the assumption of a single horizontal anisotropic layer with a horizontal axis of symmetry.
 Some previous results in our study area are also shown for comparison (gray bars in Figure 5) [Huang et al., 2011a; Becker et al., 2012; Zhao et al., 2013]. While the ϕm obtained in this study are nearly consistent with these results in most areas, they differ a lot from the results of Zhao et al.  in the YC2 (Figure 5). We note that the event Bazs they used are in the range of 127.4°–148.4°, close to ours (122.4°–144°). This discrepancy may arise from different preprocessing procedures, frequency content of waveforms or result quality criterions during the splitting measurements. Good event Baz coverage is needed to inspect the possibility of complex anisotropy in the YC2. In the following, however, we rely on our results for discussions due to the mutually consistent results given by the two methods, the coherent fast directions at nearby locations and also the high SNR of initial waveforms in the YC2 (e.g., in Figure 2a).
 The measured δtm ranges from 0.65 to 1.9 s in our study area. An additional inspection to the accuracy of the delay times is to check the angular differences between the event Baz and the measured ϕm (|ϕm - Baz|) [e.g., Huang et al., 2011a]. As shown in Figure 5, the bars with thickened outlines denote the 20 results for which the angular difference (projected to 0°–45° ) > 25°. Generally, small angular differences (< 15°) may overestimate the delay times in the presence of noise [Wüstefeld and Bokelmann, 2007]. The δtm of these 20 results are in the range of 0.65 to 1.4 s, with an average value of ~1.0 s. This is nearly equal to the global average delay times in the continents [Silver, 1996]. The largest delay times (1.4 s) are observed in the YC2 to the east of the TanLu fault (Figure 5).
 The estimates of 57 PmS splitting parameters (ϕc, δtc) beneath the 16 stations are listed in Table S2. The ϕc shows much scatter at some stations and the δtc is mostly smaller than 0.3 s except for few results with δtc > 0.4 s. Although the anisotropic strength in the lower crust can range up to 15%, the typical crustal delay times for the whole crust are < 0.3 s in most cases except for occasional values up to 0.5 s [Silver, 1996; Savage, 1999; Long and Silver, 2009], for example in the ~60 km thick Tibetan crust [e.g., Liu and Niu, 2012; Sun et al., 2012]. Previous studies show that the crust is ~32 km thick in our study area [e.g., Chen et al., 2010], for which the > 0.4 s δtc seems too large. Therefore relying on the individual PmS splitting measurement results can be highly risky [e.g., Nagaya et al., 2011; Liu and Niu, 2012] and the mean values of many measurements at a station or region are more reliable [Huang et al., 2011b], which is also supported by the synthetic tests [Wüstefeld and Bokelmann, 2007].
 The rose diagrams (i.e., the angular histograms) of the ϕc in each subdomain are also plotted on Figure 5 to compare with the ϕm. The number of measurements and stations in each subdomain are shown on the numerator and the denominator, respectively. The average ϕc in each subdomain according to Audoine et al.  is shown as orange bars (see captions of Table S2). Note that the results are relatively few and sparse in the YC2 and the eastern NCC, which may be caused by the relatively lower SNR of the PmS signals due to the influences of local complex tectonics on receiver functions. It seems that the average ϕc varies and follows the trend of the ϕm. Given the limited results for statistics and also the scatter of the ϕc, this trend needs to be further checked. However, this is consistent with the results of Zhang et al.  from the wide-angle seismic data and a P-wave anisotropic tomography study [Huang and Wang, 2011].
5.1 Depth Localization of Anisotropy
 Due to the P-to-S conversion at the core-mantle boundary (CMB) and the near-vertical incidence angle at the receiver side, the XKS phase splitting can be well attributed to the anisotropic medium right beneath the receiver side. However, it can be attributed to the anisotropy anywhere from the CMB to the receiver [e.g., Long and Silver, 2009]. According to our results, the average split times in the crust are smaller than 0.3 s, while typical ~1 s split times are observed for XKS phases beneath our study area. This indicates that the major part of anisotropy is located at the mantle.
 To constrain the depth localization of anisotropy, one efficient way is the Fresnel zone analysis on different XKS splitting observations [e.g., Alsina and Snieder, 1995; Rümpker and Ryberg, 2000]. The main criterion is that the Fresnel zones for XKS phases corresponding to different observations should not overlap significantly [Alsina and Snieder, 1995]. There are mainly two cases for this analysis: one is on the different splitting observations from the same events at nearby stations (for estimating the lower limit of the anisotropic layer) and the other is on the different observations from different incoming polarization directions at the same station (for estimating the upper limit of the anisotropic layer) [Alsina and Snieder, 1995].
 All our XKS splitting results from different events are summarized in Figure 6. For example in Figure 6a, the PKS phases from the same event 2010.012 yield different results at stations within the SCOB from those within the YC1. The corresponding initial R and T components of the PKS phases are shown in Figure 7. It is obvious that the particle motion is linear in the SCOB (stations ZJ01–ZJ05) while elliptical (stations ZJ06–ZJ10) or subelliptical (stations ZJ11, AH01 and AH02) in the YC1. Within short distances the Baz varies little in the range of ~14°–18° (Figure 7). Therefore this is a direct indicator of the laterally variant anisotropy beneath the SCOB and the YC1.
 To get a more quantitative estimation of the depth of anisotropy, we calculated the widths of the first Fresnel zones for corresponding XKS waves by [e.g., Chevrot et al., 2004], where TXKS the dominant period of the XKS wave (chosen as 8 s), v the shear wave velocity (chosen as 4.5 km/s above 200 km depth and 5.0 km/s at 410 km depth) and z the depth.
 Figure 8a shows the Fresnel zones at 50, 100, and 200 km depths for PKS waves from event 2010.012 recorded at four stations (ZJ04, ZJ06, ZJ09, and ZJ11). As depths become larger than 100 km, the Fresnel zones begin to overlap, implying that the variant shallow anisotropic structures at depths < ~100 km (approximately the lithospheric thickness beneath eastern China [An and Shi, 2006]) account for the variant splitting observations. This speculation is supported by the variant crustal anisotropy from the YC1 to the SCOB. Another example of the first case of Fresnel zone analysis is the different SKS splitting parameters from event 2010.246 at stations AH06 and AH10, which are near the TanLu fault (Figure 6b). The Fresnel zones for SKS waves are calculated at 100, 200, and 410 km (Figure 8b). Unfortunately, this does not well constrain the lower limit of the anisotropic layer, because the Fresnel zones only begin to touch at depths > ~410 km.
 Interestingly, at station AH09 where the fast directions change from the YC2 to the NCC (Figure 5), the measured ϕm depend on the event backazimuth. The event 2011.108 (SKS phase from southeast, Figure 6d) yields ϕm close to that in the YC2 while event 2011.001 (SKKS phase from north, Figure 6c) yields ϕm close to that in the NCC. Another SKKS phase from event 2011.001 was also clearly recorded at station AH09 (Figures 9a and 9b). For ease the two phases are denoted by SKKS1 and SKKS2 phases, respectively. With opposite backazimuths, the two phases have similar amplitudes on the R components while the amplitudes of the T components differ much (Figure 9a). The particle motions and the measurements indicate obvious and near-null splitting for the SKKS1 phase and the SKKS2 phase, respectively (Figures 9c and 9d).
 The variant splitting observations on the incoming polarization directions at station AH09 are deviating from the single-layer anisotropy with a horizontal axis of symmetry and may indicate lateral variations of anisotropy, multiple anisotropic layers or dipping symmetry axis of anisotropy [Silver and Savage, 1994; Rümpker and Silver, 1998; Chevrot and van der Hilst, 2003]. The poor backazimuthal coverage at station AH09 cannot allow distinguish these possibilities. However, the multiple-layer anisotropy will cause the apparent splitting parameters to vary with a 90° periodicity on incoming polarization directions [Silver and Savage, 1994; Rümpker and Silver, 1998]. The discrepant splitting of the SKKS1 and SKKS2 phases with opposite backazimuths is inconsistent with this periodicity. We consider the lateral variations of anisotropy to explain the splitting observations at station AH09. By the synthetic waveform modeling, Rümpker and Ryberg  show that for a model of laterally variant anisotropy at the same depth range with a sharp boundary, the apparent splitting parameters will vary 180°-periodically as function of initial polarizations when the receiver position is within the “sensitivity range.” This is either inconsistent with the SKKS1 and SKKS2 splitting discrepancy, therefore more complex mechanisms than this model are required to fit our observations.
 The Fresnel zones for the SKS, SKKS1, and SKKS2 phases are calculated at 50, 100, and 200 km depths (Figures 8c and 8d). As shown in Figure 8c, the Fresnel zones for SKKS1 and SKKS2 phases overlap at shallow depths and begin to separate at depths > ~200 km. Thus the different SKKS1 and SKKS2 splitting measurements suggest different anisotropy in a relatively deep layer in the mantle. Many recent studies have used the discrepant SKS - SKKS splitting to demonstrate the contribution from the lower mantle to splitting when the SK(K)S wave has propagated near-horizontally through the anisotropic lower mantle [e.g., Wang and Wen, 2007; Long, 2009; He and Long, 2011; Lynner and Long, 2012]. Is it possible that the SKKS1 and SKKS2 splitting discrepancy is caused by the contribution of the anisotropic lower mantle? The splitting parameters measured from other backazmiuths in the NCC (e.g., event 2010.246 shown in Figure 6b) are slightly different from the SKKS1 phase. Obviously the SKKS1 phase and those phases sample different regions of the lower mantle. Therefore, the measured splitting parameters in the NCC more likely reflect the upper mantle anisotropy. On the other hand, the Fresnel zones for the SKS and SKKS1 waves begin to separate at depths > ~200 km (Figure 8d), which also supports that a relatively deep anisotropic layer contributes to the SKKS1 splitting.
5.2 Mechanisms for Anisotropy in Different Tectonic Domains
 As shown above, from the Southeast China Orogenic Belt to the Yangtze Craton, the shallow anisotropic structures vary laterally, which may result from different deformational histories of the lithosphere. Generally the retained anisotropic structures in the lithosphere are related to the most recent tectonic activities strong enough to “erase” the older anisotropic fabrics [Silver and Chan, 1991; Silver, 1996]. Although the SCOB and the YC are two different blocks, which may have different inherent anisotropic properties after the collision in the Precambrian [e.g., Charvet et al., 1996], the anisotropic structures probably have been reformed during the strong Mesozoic tectonic activities.
 According to geological and geochemical studies the most recent significant tectonic activities were the Late Mesozoic magmatic activities in the SCOB. The remarkable features are the distributions of early Cretaceous volcanic rocks and the late Cretaceous back-arc basins in the SCOB and the YC1 (Figure 5) [Li, 2000; Zhou et al., 2006; Shu et al., 2009]. The model for these distributions involves the change of the subduction angle (from a low-angle to a high-angle) or slab retreat and rollback of the paleo-Pacific plate in the late Cretaceous, causing the formation of the late Cretaceous back-arc basins under the extensional environment (Figure 5) [e.g., Zhou et al., 2006; He and Xu, 2012]. Interestingly, the trend of the back-arc basin is almost normal to the measured fast directions at the adjacent stations ZJ06, ZJ08, and ZJ09 (Figure 5). We infer a possible relation between the variant anisotropy from the SCOB to the YC1 and the extensional environment in the late Cretaceous. In this case the local lithospheric extension made the fast directions parallel with the extensional direction [Savage, 1999]. It is interesting that the ϕc also seems to change and follow the trend of the ϕm from the SCOB to the YC1 (Figure 5), supporting the vertically coherent deformation model [e.g., Silver, 1996]. Additionally, tomography studies show the prominent low velocity anomalies in the upper mantle beneath the eastern SCB [e.g., Huang and Zhao, 2006; Huang et al., 2010; Obrebski et al., 2012; Zhao et al., 2012; Zhou et al., 2012]. Therefore the small-scale convection, vertical/horizontal flow may also exist beneath the eastern SCB to form some asthenospheric anisotropy.
 On the other hand, it is interesting to observe the abrupt change of anisotropic structures from the YC1 to the YC2 region within the Yangtze Craton. The Late Mesozoic magmatic activities are weak in the YC2 region. It is likely that the observed anisotropy in the YC2 is not significantly influenced by the Late Mesozoic magmatic activities as that in the YC1. The TanLu fault is a large-scale strike-slip fault and may cause the shear deformation of the lithosphere in the YC2. Geological studies show that the TanLu fault has experienced a complex geological evolutionary history which includes the transformation between sinistral and dextral strike-slip, and between transtension and transpression in the Mesozoic and Cenozoic [e.g., Ren et al., 2002]. Therefore, the NEE-SWW fast directions of anisotropy in the YC2 may be a compositive result of these transformations or the tectonic activities of the TanLu fault have few imprints on the frozen anisotropic fabrics formed in the Early Mesozoic.
 In contrast, in the southeastern NCC, the fast directions are subparallel with the trend of the Qinling-Dabie Orogenic Belt and also the local APM direction, which is –71.1° with a velocity of 2.2 cm/a (location: 32°N, 116°E, shown on Figure 5) in the HS3_NUVEL_1A frame [Gripp and Gordon, 2002]. Therefore, both the lithosphere and the asthenosphere may experience deformations that cause the observed anisotropy. Considering the variant crustal anisotropy across the TLF, the lithosphere probably experiences deformations that contribute significantly to the anisotropy in the southeastern NCC.
 However, as mentioned above, more complex anisotropy from the deeper layer beneath the southeastern NCC also contributes to the splitting observations. Geological studies showed that the lithosphere of the Yangtze Craton subducted northward beneath the North China Craton to the depth > 200 km during the amalgamation in the Early Mesozoic [Ye et al., 2000]. It is possible that some local deep heterogeneities exist and assemble strains that form anisotropy beneath the southeastern NCC. Another possibility is the LPO of olivine in the asthenosphere induced by the APM. However, in this case, it is not clear why the APM-induced anisotropy is only observed beneath the eastern NCC in our study area. One possible explanation is the strong mantle upwelling beneath the eastern SCB, which has destroyed the coherent APM-induced LPO of anisotropic minerals in the asthenosphere [Huang et al., 2011a]. Another possibility is the limited events used in this study not allowing to model two or more layers of anisotropy if the APM-induced anisotropy commonly exists beneath our study area. Therefore, although the asthenospheric anisotropy may exist, it does not dominate the splitting observations presented here.
 In this study, we obtained the splitting parameters by analyzing the P-to-S phases converted at the Moho discontinuity and the XKS phases converted at the core-mantle boundary to study the crustal and upper mantle deformations beneath eastern China. The results show distinct lateral variations of the fast directions in different tectonic blocks. The fast directions analyzed with core phases are: Southeast China Orogenic Belt (mainly NE-SW), Yangtze Craton (two groups: NW-SE and NEE-SWW) and eastern North China Craton (mainly NWW-SEE). The delay times in the mantle and crust are moderately large (~1 s) and averagely smaller than 0.3 s, respectively. The Fresnel zone analysis is used to constrain the depth localization of variant anisotropy. The result reveals the laterally variant anisotropy in the lithosphere between the two sub-blocks (Southeast China Orogenic Belt and Yangtze Craton) of the eastern South China Block. On the other hand, the anisotropy may have lateral variations in a relatively deep layer beneath the southeastern North China Craton.
 We thank Lingsen Meng for helpful suggestions on the early draft of this paper. Comments from James Tyburczy (Editor) and two anonymous reviewers greatly improved the manuscript. Figures 1, 5, 6 and 8 were made by GMT software [Wessel and Smith, 1998]. This work was supported by the Sinoprobe Project (02-03), the Project ZZKT-201105 and National Natural Science Foundation of China (Grants: 41204040 and 41174038).