Imaging lithospheric structure of the eastern Himalayan syntaxis: New insights from receiver function analysis

Authors

  • Qiang Xu,

    1. Key Laboratory of Continental Collision and Plateau Uplift, Institute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing, China
    2. State Key Laboratory of Geological Processes and Mineral Resources, China University of Geosciences, Beijing, China
    Search for more papers by this author
  • Junmeng Zhao,

    1. Key Laboratory of Continental Collision and Plateau Uplift, Institute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing, China
    Search for more papers by this author
  • Shunping Pei,

    1. Key Laboratory of Continental Collision and Plateau Uplift, Institute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing, China
    Search for more papers by this author
  • Hongbing Liu

    1. Key Laboratory of Continental Collision and Plateau Uplift, Institute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing, China
    Search for more papers by this author

Corresponding author: Q. Xu, Key Laboratory of Continental Collision and Plateau Uplift, Institute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing 100101, China. (xuqiang@itpcas.ac.cn)

Abstract

[1] We employ the P and S receiver function technique to data from the 44 seismic stations deployed in the eastern Himalayan syntaxis to investigate the crustal thickness, the average Poisson's ratio, and the depth of the lithosphere-asthenosphere boundary (LAB). The observed crustal thickness exhibits an overall NE-deepening trend, varying from 55 to 75 km. Two anomalous areas lie in the west and east of the Namche Barwa syntaxis characterized by thinner and thicker crust, respectively. The average Poisson's ratios within the study area are low in the north and moderate elsewhere with some high values in the south, consistent with felsic and intermediate rocks forming the crust. Our migrated images reveal that (1) the LAB of the Tibetan plate exists at relatively shallow depths (~110 km) and exhibits a gap beneath the Namche Barwa syntaxis, which may have formed by the delamination of mantle lithosphere due to local mantle upwelling, and (2) the LAB of the Asian plate is observed at a depth of ~180 km, which implies that the Asian plate has advanced southward to about 30°N under the Lhasa terrane. Our results provide new insights into the understanding of continental subduction and lithospheric deformation of the eastern Himalayan syntaxis.

1 Introduction

[2] The eastern Himalayan syntaxis (EHS) is located at the eastern termination of the Himalaya orogenic belt (Figure 1), which results from the Indian-Asian collision and is marked by intensive tectonic deformation, extremely rapid uplift and denudation rates, and extensive Cenozoic metamorphism and magmatism [Xu et al., 2012]. Since ~50 Ma, continuous northward movement of the Indian plate has generated the dominant east-west structural grain of the Himalayan Mountains. However, an eastward transition into north-south trending structures occurs in the Namche Barwa region of the EHS, which has resulted in the U turn of the Indus-Yarlung suture (IYS) [Ding et al., 2001; Xu et al., 2012; Zhang et al., 2004]. The EHS consists of the Qiangtang, Lhasa, and Himalayan terranes, which are divided on the north by the Bangong-Nujiang suture (BNS) and the IYS to the south [Searle et al., 2011].

Figure 1.

Topographic map showing the major tectonic boundaries and seismic stations (blue triangles) within the eastern Himalayan syntaxis. Also shown are the abbreviated station names by removing the first two characters. The red box in the inset map shows the location of the study area with respect to the Tibetan Plateau. IYS, Indus-Yarlung suture; BNS, Bangong-Nujiang suture; JRS, Jinsha river suture; NBS, Namche Barwa Syntaxis.

[3] Previous detailed petrological and geochronological analyses of xenoliths show that the EHS experienced high- and ultrahigh-pressure metamorphism during the early Tertiary [Ding and Zhong, 1999; Zhang et al., 2008]. However, the knowledge on the geodynamic processes of the continental subduction and exhumation, and the mechanism of lithospheric deformation are still unclear. Two representative models are proposed to explain the crustal deformation mechanisms of the EHS and the adjacent region: (1) middle/lower crustal flow [Clark and Royden, 2000; Sol et al., 2007] and (2) vertical coherent deformation [Sol et al., 2007]. The former suggests that the low-viscosity middle/lower crust flows ductilely and is decoupled from the upper crust and the underlying lithospheric mantle, which are supported by the existence of the low-velocity regions identified by receiver function methods and ambient noise tomography, and the low-resistivity zones revealed by Magnetotelluric data [e.g., Bai et al., 2010; Xu et al., 2007; Yao et al., 2008]. However, the distributions of the low-velocity anomalies are complex. In addition to correlating to the pattern of surficial faulting, the anomalies are located at varying depths. Moreover, the existence of a low-velocity zone (LVZ) does not in itself indicate crustal flow. In contrast, vertical coherent lithospheric deformation (i.e., the crust and lithospheric mantle deform almost homogeneously and are mechanically coupled) is strongly supported by the strikes of the major faults, stress field distributions, GPS velocity measurements, and fast directions from shear wave splitting analysis which show a clockwise motion around the EHS and adjacent regions [Sol et al., 2007]. However, this model is in conflict with Rayleigh wave azimuthal anisotropy, whose properties vary with depth, implying a different deformation pattern between the crust and the upper mantle [Yao et al., 2010]. Therefore, additional constraints on the crustal properties of the EHS are necessary to solve this contradiction.

[4] The fate of the mantle lithosphere at the EHS and surrounding area remains controversial. The pattern of the high and low Pn velocities in the southern Tibetan Plateau indicates that the Indian lithosphere in the eastern part (90°E–96°E) extends south of the Jinsha River suture (~33.0°N) [Liang and Song, 2006]. In contrast, the results from finite frequency tomography suggest that the Indian lithosphere experienced delamination and advances only as far as the central Lhasa terrane [Ren and Shen, 2008]. However, this interpretation disagrees with travel time tomography [Li et al., 2008], which suggests the front of the Indian plate subducted to depths of 200–400 km beneath the Himalayan terrane, whereas interpretation of Rayleigh wave tomography suggests that the high-velocity anomalies on either side of the BNS represent the Indian lithosphere to the south and the Asian lithosphere to the north [Fu et al., 2010]. However, it should be noted that body wave and surface wave tomographic methods are not sensitive to the intralithospheric mantle discontinuities.

[5] In this paper, we explore the detailed depth variations of the velocity discontinuities (Moho and lithosphere-asthenosphere boundary (LAB)) in the crust and upper mantle of the EHS using teleseismic waveform data recorded at 44 portable stations of the Namche Barwa Project (Figure 1). In addition, simultaneously obtained Poisson's ratios provide valuable constraints on the bulk composition of crustal rocks. We primarily employ the station-based converted wave techniques, namely P-to-S (P receiver function, PRF) and S-to-P (S receiver function, SRF), for these aims. Finally, we discuss the geodynamic implications suggested by these estimated seismic parameters.

2 Data

[6] The waveform data utilized in this study come from seismic events with magnitudes larger than 5.5 recorded by the 44 high-quality broadband stations of the Namche Barwa Project between July 2003 and October 2004 conducted by Lehigh University. The broadband stations used in this study are equipped with 33 Streckeisen STS-2, 9 Gurlap CMG3-ESP and 2 Gurlap CMG3T sensors, and RefTek 72A data acquisition system recorded continuously at 40 samples per second. For PRF analysis, we selected P phases with epicentral distance between 30° and 95°, while SRF calculation was restricted to S phases ranging from 60° to 85°, and SKS phases ranging from 85° to 120°. We hereafter denote both S and SKS phases by S phases [Kumar et al., 2006; Yuan et al., 1997, 2006]. The waveform data of each teleseismic event are cut in a time window of 120 s, which is 20 s before the theoretical P wave onset for PRF and 100 s before the predicted S-phase arrival for SRF, respectively. After low-pass filtering with a corner frequency of 1 Hz to avoid aliasing, the traces are demeaned, detrended, and decimated to a sampling rate of 10 samples per second. In addition to these described processing, we also visually pick events that clearly show P phases at vertical components and S phases at radial components at around their theoretical onset times for further analysis. Figure 2 illustrates the distribution of the 185 teleseismic events used in this study according to the above-mentioned criteria, most of which are confined to the circum-Pacific seismogenic belt.

Figure 2.

Epicentral locations of the teleseismic events used in the P and S receiver function analysis. Contours show the distance in degree to the approximate position of the network.

3 Methods

3.1 Receiver Function Calculations

[7] Using the P-to-s (Ps) converted waves, PRF has become one of the dominant techniques for investigating the seismic parameters of the velocity discontinuities in the crust and upper mantle, whereas the recently developed SRF employing the S-to-p (Sp) converted waves has proven to be very effective to determine the uppermost mantle discontinuities such as LAB in different tectonic regions of the world [Abt et al., 2010; Fischer et al., 2010; Rychert and Shearer, 2009; Wittlinger et al., 2004a], because the Sp phases from the deeper discontinuities are not disturbed by the multiples originating from the shallow discontinuities. By contrast, the Ps phases from the deeper discontinuities and the shallow crustal multiples almost arrive in the same time window, which hinders the identification of the significant phases in the PRF technique [Kumar and Kawakatsu, 2011]. However, it is indisputable that PRF is more helpful for the detailed crustal structure analysis than SRF due to the higher-frequency content.

[8] Here we briefly outline the basic processing steps of the receiver function calculations. For a detailed description of the procedure, see the recently published review paper [Kind et al., 2012]. The process begins by rotating the raw Z, NS, and EW (ZNE) components into the local P-SV-SH ray-based coordinate system using theoretical back azimuth and incident angles. For PRF, the incident angles are determined by a table of angles of incidence of P waves given by the theoretical calculation [Pho and Behe, 1972]. For SRF, we consider the optimal angle which minimizes the amplitude of the P component at a time widow of ±1 s of the S-phase arrival time by systematically varying the rotation angles as the appropriate angle of incidence [e.g., Kumar and Kawakatsu, 2011; Xu et al., 2011].The PRF and SRF are computed in ray coordinate system, which has the advantage of better confining P, SV, and SH energy than ZRT system. Then, deconvolution is used as a source normalization procedure to remove the influence of the source and propagation path. Here deconvolution is preformed by a time domain Wiener filtering deconvolution method [Yuan et al., 1997]. After deconvolution, all components are aligned and normalized to the maximum amplitude of the spike in the P components for PRF and SV components for SRF. The resulting deconvolved SV (P) from P (SV) is called PRF (SRF). We also reverse the polarity of the SRF amplitudes, as well as the time axis, in order to compare directly with the PRF. Finally, an additional zero-phase Butterworth band-pass filtering with corner frequencies of 0.03 and 1 Hz is also applied to the resultant PRF and SRF to eliminate low-frequency noise.

3.2 Joint Application of Semblance Analysis and Moveout Correction

[9] Based on the delay times and amplitudes of the Ps phase from the Moho and its crustal multiples (PpPs and PsPs + PpSs), the H-κ stacking method has been introduced to jointly determine both the Moho depth and Vp/Vs ratio at each station [Zhu and Kanamori, 2000]. Due to the fact that the crustal thickness determined only from the delay time of the Moho Ps converted phase trades off strongly with the crustal Vp/Vs ratio, this algorithm resolves the ambiguity by incorporating the later multiples. To make more reliable estimates of Moho depths and Vp/Vs, we utilize the joint application of semblance analysis and moveout correction. The detailed processing procedure is outlined below.

[10] First, the so-called semblance analysis adds the semblance parameter into the objective function of the widely used H-κ stacking approach [Eaton et al., 2006]. The semblance parameter serves as an additional weighting function suppressing the incoherent noise, defined as follows:

display math(1)

where ri(t) are the amplitudes for the ith receiver function and tj are predicted delay times of the Moho Ps, PpPs, and PsPs + PpSs phases. Accordingly, the updated objective function, which sums the weighted amplitudes of each phase at the predicted arrival times for different values of crustal thickness (H) and Vp/Vs ratio (κ), becomes

display math(2)

where wj are the weighting factors (∑ wj = 1) for each phase. The combination of H and κ that maximize s′ (H,κ) is considered as the optimal solution. For this study, an average crustal Vp of 6.2 km/s is used based on the results from deep seismic sounding[Wang et al., 2007] and previous studies in eastern Tibetan plateau [Wang et al., 2010; Xu et al., 2007], and the weights of 0.5, 0.5, and 0 are set for Ps, PpPs, and PsPs + PpSs, respectively. As noted by Zhu and Kanamori [2000], the H and κ are relatively insensitive to Vp, since the differential traveltimes are used for the stacking [Wittlinger et al., 2009]. A 3% uncertainty in Vp changes ~2 km in H and ~0.01 in κ after testing data of station ES02 (Figure 3). The weight of the second multiple is set to zero because it is almost invisible on the PRFs at every station, possibly resulting from the thicker crust and strong crustal attenuation in Tibetan plateau. The uncertainties in H and κ are determined by employing a bootstrap method [Finotello et al., 2011].

Figure 3.

An example showing the procedure for obtaining the Moho depth (H) and Vp/Vs ratio (κ) at each station. (a) Individual receiver functions for station ES02, sorted by the ray parameter. (b) Summed trace of all the receiver functions after moveout correction for the PpPs phase. The black triangle marks the predicted delay time of the Moho PpPs phase according to the obtained H and κ shown in (c) from the semblance analysis.

[11] Second, we stack all the PRFs after moveout correction for PpPs phase at each station. In contrast with the prominent Ps phases, the PpPs phases are usually weak and have low amplitude. After applying this correction to the PpPs phase, it is obvious that stacking the distance (ray parameter) equalized PRFs enhances the amplitude of PpPs phase. If the PpPs phase is properly identified (i.e., a presence of the prominent positive amplitude arrival in a time window of ±1 s around the predicted time for the obtained H and κ) in the sum trace, we adopt the satisfactory H and κ for further interpretation. A reference slowness of 6.4s/° is applied for moveout correction [Yuan et al., 1997, 2002]. The Poisson's ratio (σ) associated with the crustal rock properties can be computed from the Vp/Vs ratio (κ) by the formula σ = 0.5*[1 − 1/(κ2 − 1)].

[12] Using station ES02 as an example, Figure 3a illustrates the individual P receiver functions sorted by the ray parameters. The maximum semblance in Figure 3c shows a Moho depth of 68.9 km and average Vp/Vs ratio of 1.77. The summation trace of moveout-corrected PRFs for the PpPs phase in Figure 3b indicates the clear arrival of the PpPs phase, which agrees well with the predicted delay time of the Moho PpPs phase marked by the black triangle according to the obtained H and κ.

[13] Lastly, we roughly estimate H using the picked delay time of the Moho Ps phase for these stations with no acceptable results, which imply that no clear PpPs phases are observed in the sum traces or multiple maxima appear in stacking matrices, assuming an average crustal Vp of 6.2 km/s and Vp/Vs ratio of 1.732. We do not attempt to determine the stratified Vp/Vs by utilizing the multiple maxima in stacking, owing to the absence of a coherent Ps conversion from the intracrustal layer at most stations as shown in Figure 4 [Vergne et al., 2002; Wittlinger et al., 2009; Wittlinger et al., 2004b]. In addition to the deviation from the real κ with respect to 1.732, the uncertainty in H is also positively correlated with the Moho Ps time [Xu et al., 2010].

Figure 4.

(a) Stacked traces of the Moho Ps (left) and PpPs (right) phases for the stations with ideal H and κ, sorted by the delay time of the Moho Ps phase. Black triangles mark the predicted delay times of the Moho PpPs phase corresponding to the obtained H and κ. (b) Summed traces of the Moho Ps phase for other stations, where H is estimated. Station names are given on the left of each trace.

3.3 Depth Migration

[14] To attain the needed spatial subsurface images, we conduct the depth profiles by using the common conversion point (CCP) stacking algorithm, which has been successfully applied in many studies of the Tibetan plateau [Kind et al., 2002; Zhao et al., 2010; Zhao et al., 2011]. The CCP stacking is done along the 2-D profile, which is 100 km in width, with a horizontal and vertical grid spacing of 2 km. For each profile, the Ps (Sp) amplitudes of each receiver function are traced back to their true spatial locations where the conversions occur along the raypath based on the ray-tracing algorithm. Then, all the signal amplitudes are stacked along the direction normal to the profile to form a vertical cross section, with lateral and depth grids. Finally, we achieve the horizontal stacking along the profile to produce a smooth image. The horizontal stacking distance depends on the size of the Fresnel zone at different conversion depths, which can be calculated using the math formula, where λ is its wavelength and z is depth. At a 60 km depth, this width is ~15 km for PRF and ~43 km for SRF. For simplicity, we choose the modified IASP91 model with the crustal velocity structure from ambient noise and surface wave inversion as the background velocity model [Yao et al., 2008]. Similar to many previous receiver function studies with good imaging results [Dzierma et al., 2010; Eagar et al., 2011; Rychert and Shearer, 2009], the 1-D velocity model for ray tracing, instead of a more sophisticated 3-D tomographic model, seems rational because the amplitudes in transverse PRFs at almost all stations are so small that the dipping structures have little effect on our final images. On the other hand, a maximum error of 3 km in the Moho depth and 5 km in the LAB depth can be introduced if we use another tomographic model derived from P wave travel times [Sun and Toksoz, 2006], but we think that these differences do not change our final geological interpretations. Considering that the dominant periods of PRF and SRF (1 and 4 s, respectively) limit a maximum vertical depth resolution of 1 km for Ps and 6 km for Sp conversions, the errors in the depth estimation are less than 5 km using PRF and about 10 km in the LAB depths determined by SRF [Sodoudi et al., 2011].

4 Results and Discussion

4.1 Crustal Thickness and Poisson's Ratio

[15] We produced 3613 PRFs from the 44 stations utilized in this study. Figure 4 illustrates the sum traces for all stations, consisting of (a) the sum traces of the Moho Ps and PpPs phases for these stations with ideal results of H and κ sorted by the delay time of the Moho Ps phase, and (b) the stacked traces of the Moho Ps phase for the remaining stations (gray dotted line) with the unique estimated H. Before stacking, the moveout correction technique and two low-pass filters of 2 and 3 s were applied to all the PRFs for the Ps and PpPs phases, respectively. Overall, the large Ps phases generated at the Moho are clear in all the sum traces between 6.8 and 9.2 s, and the subsequent PpPs phases, which vary from 23.9 to 29.8 s, are well resolved for most stations and are consistent with the predicted delay times of the Moho phases (Figure 4) corresponding to the obtained H and κ. Another Ps conversion directly after the P wave arrival at 0 s in Figure 4 may be attributed to the mutual interference between converted energy at shallow depth and some remaining P wave energy from imperfect rotation of the coordinate system [Yuan et al., 2002].

[16] Table 1 lists the obtained H and κ at each station, and these observations are plotted in Figure 5 in color to visualize the specific variations. In addition, results of H (white circles) are imposed on the four depth profiles constructed from the CCP stacking approach for mutual confirmation and comparison, and are shown in Figures 7 and 8. Even though both the results show reasonable consistency in the variation pattern of the Moho depth, some local deviations still exist due to the assumption of a layer over a half-space model used in the semblance analysis and the unknown velocity model of time-depth conversion.

Table 1. Summary of the Moho Depth (H), Vp/Vs (κ), and Poisson's Ratio (σ)a
StationκH1σnStationκH1σn
  1. aH1 is the results of the Moho depth from the joint application of the semblance and moveout correction. H2 is the roughly estimated Moho depth from the delay time of the Moho Ps phase. n is the number of PRF at each station.
ES011.77 ± 0.0471.2 ± 3.00.266115ES371.79 ± 0.0355.1 ± 2.50.273143
ES021.77 ± 0.0168.9 ± 0.50.26667ES381.72 ± 0.0163.6 ± 0.60.245110
ES031.78 ± 0.0166.8 ± 0.50.26962ES391.73 ± 0.0163.4 ± 0.50.249132
ES041.72 ± 0.0368.1 ± 2.90.24590ES401.76 ± 0.0264.6 ± 1.80.26291
ES061.72 ± 0.0568.8 ± 4.30.24513ES411.79 ± 0.0263.9 ± 1.00.27372
ES081.67 ± 0.0171.2 ± 0.40.220109ES431.85 ± 0.0263.3 ± 1.30.294108
ES091.81 ± 0.0168.2 ± 0.50.280101ES44A1.75 ± 0.0462.8 ± 3.00.25849
ES101.68 ± 0.0172.5 ± 0.50.22671ES481.85 ± 0.0152.3 ± 0.60.29468
ES111.74 ± 0.0169.5 ± 0.50.253129     
ES121.68 ± 0.0169.1 ± 0.50.226110StationPs(s)H2σn
ES131.70 ± 0.0270.7 ± 1.60.235120ES058.771.0 65
ES141.74 ± 0.0171.4 ± 0.50.253117ES078.065.2 64
ES151.76 ± 0.0171.7 ± 0.50.262101ES168.972.9 57
ES171.67 ± 0.0372.3 ± 2.00.22044ES198.569.3 97
ES181.69 ± 0.0171.1 ± 0.70.23171ES218.771.0 26
ES201.70 ± 0.0170.5 ± 0.50.23591ES249.275.8 21
ES231.69 ± 0.0174.6 ± 0.40.23198ES278.266.8 58
ES251.72 ± 0.0173.2 ± 0.50.24588ES308.872.0 95
ES261.75 ± 0.0273.0 ± 1.40.258115ES318.771.0 45
ES281.83 ± 0.0265.9 ± 1.40.28735ES328.367.6 70
ES291.76 ± 0.0169.9 ± 1.00.262101ES338.569.3 103
ES341.72 ± 0.0165.0 ± 0.50.245104ES458.166.0 38
ES351.75 ± 0.0160.2 ± 0.40.25869ES466.855.4 80
Figure 5.

Map of our estimates of the (a) Moho depth and (b) Poisson's ratio.

[17] Our results reveal a NE-deepening trend of Moho depths beneath the EHS. As indicated in Figure 5, the Moho depth gradually deepens from ~60 km beneath Lhasa terrane to ~72 km within the Qiangtang terrane. Two anomalous areas are observed in our findings. One prominent shallower zone of the Moho at a depth of 55–60 km is located in the west of the highly exhumed Namche Barwa Massif, which is not observed in the results of Zurek [2008]. We attribute this significant discrepancy to the different methods used in both investigations. In addition, the agreement of our results with those of Zurek [2008] is reasonable at most other stations. Nevertheless, the combination of semblance analysis and moveout correction makes our results more reliable than the only H-κ stacking. The other anomalous area with thicker crust above 73 km appears in the eastern part of the study region along the BNS. One possible explanation for this feature may be correlated to the abrupt change of subduction direction of the Indian plate from NE to E. The same variation pattern of the Moho depth is shown by independent Rayleigh wave tomography [Fu et al., 2010], but the obtained crustal thicknesses from the tomography technique may be less credible than our results due to the associated lower resolution and integration effects across the crust.

[18] The estimated Poisson's ratios range between 0.22 and 0.29 and have an average value of 0.26, which represent no pronounced variation tendency. The Poisson's ratio is low in the north along the BNS and moderate elsewhere with some high values in the south of the IYS. In comparison with P- or S-wave velocity alone, the Poisson's ratio is a better diagnostic tool for assessing crustal composition. The crust generally consists of four compositional rock types, which include felsic, intermediate, mafic, and fluid-filled porous/fractured or partially molten rocks. These rocks correspond to low (σ < 0.26 or Vp/Vs < 1.76), medium (0.26 ≤ σ < 0.28 or 1.76 ≤ Vp/Vs < 1.81), high (0.28 ≤ σ < 0.30 or 1.81 ≤ Vp/Vs < 1.88), and very high (σ ≥ 0.30 or Vp/Vs ≥ 1.88) Poisson's ratio, respectively [Ji et al., 2009]. Our estimated Poisson's ratios within the study area correspond to the low and moderate values, suggesting that the crust is mainly composed of felsic and intermediate rocks, which is consistent with the conclusion reached by previous receiver function studies [Zurek, 2008]. However, the felsic and intermediate crust does not exclude the local existence of thin layers of partial melt. As shown in Figures 7 and 8, the patches of negative velocity contrasts at ~20–40 km depth mark the top of local low-velocity zones (LVZs). These midcrustal LVZs represent local partial melting and/or aqueous fluids probably caused by the overthickened crust with high heat production, whose properties are consistent with the low-resistivity layers and some negative polarity bright spots recognized in the INDEPTH survey [Nelson et al., 1996]. In accord with the findings of Hi-CLIMB seismic array [Xu et al., 2013], the fragmental LVZs are incompatible with the hypothesis of continuous flow channels as suggested by Clark and Royden [2000]. Accordingly, we argue that crustal thickening maybe achieved by extrusion of the mafic lower crust and/or thrusting and folding of the felsic upper and middle crust.

4.2 Images of the LAB

[19] We generate depth-migrated images of the SRF and the PRF along four profiles by the CCP stacking approach (Figures 7 and 8), which include 1821 SRFs. Figure 6 shows the profile locations, which consist of two NE-trending transects (A and B) and two approximate E-W trending transects (C and D). These profiles include all the stations and pass through the best ray coverage region, which can be evaluated by the distribution of the piercing points of the Sp phase at a depth of 100 km. Thus, these receiver function images clearly show all of the structural features in the study region. In the migrated images, red and blue colors indicate the positive (velocity increase with increasing depth) and negative (velocity decrease with increasing depth) amplitudes of the receiver function, respectively.

Figure 6.

The red lines labeled A–D are the profiles along which the obtained migrated images have been analyzed in Figures 7 and 8. The black crosses denote the piercing points for S-to-p conversion at 100 km depth. The dashed blue line indicates the orientation of the interpretive cartoon shown in Figure 9.

[20] The Moho is the most prominent seismic convertor within the crust and is coherently observed in all the migrated profiles (Figures 7 and 8). Due to the fact that the PRF determines the Moho depth more effectively than the SRF, we draw the lateral location variation of the Moho with black dashed lines following the maximum value of the phases of our concern in the PRF migrated images. In contrast to the detailed crustal features inferred from the PRF profiles, we mainly analyze the SRF images in this study. For comparison with the PRF images, identical dashed lines along the same profiles are overlain on the SRF images. Given that SRF has longer periods, the location of the Moho in both PRF and SRF images fit reasonably well, enabling us to increase our confidence in interpreting the SRF images below the Moho.

Figure 7.

Migrated P and S receiver function images along the NE-SW trending profiles A and B (Figure 6). (top) The dashed lines in SRF images mark the Moho and lithosphere-asthenosphere boundaries of the Tibetan plate (TP), the Asian plate (AP), and the postulated fragment of the detached lithosphere (X). (bottom) The white circles mark the Moho depths derived from the combined application of the semblance analysis and moveout correction for comparison with the PRF images.

Figure 8.

E-W trending profiles C and D (see Figure 7 caption for description).

[21] A negative conversion at a depth of ~110 km beneath the Moho is clearly observed in the SRF images (Figures 7 and 8). This boundary is almost horizontal and continuous in profile A, but only a small portion of the boundary appears in the vicinity of the BNS in profile B. It is intriguing to note that there is a large gap between 95°E and 97°E for this boundary in the E-W–oriented profiles (Figures 8c and 8d). This negative Sp conversion can be interpreted as either the LAB or a midlithospheric discontinuity (MLD), but we prefer to consider it as the LAB of the Tibetan plate according to the available observations at the present time [Fu et al., 2010; Hu et al., 2011; Zhao et al., 2011]. The reliable constraints on the depth of the LAB are provided by the absolute shear-wave velocity structures in the same region from the Rayleigh wave tomography, which reveal the only pronounced velocity decrease at ~120 km depth [Fu et al., 2010]. On the other hand, the interpretation of the LAB of the TP coincides with the independent results derived from SRF, which suggests that a separate and relatively thin (100–120 km) Tibetan lithosphere overrides the southward subduction of the flat Asian lithosphere beneath the central and northern Tibet, and the LAB depth is located at a depth of 100–120 km on the eastern margin of the study region [Hu et al., 2011; Zhao et al., 2011]. Furthermore, the interpretation of a negative Sp phase at ~110 km as an MLD may be appropriate for the cratons where a thicker lithosphere is expected. In contrast, negative Sp conversions at ~110 km depth represent the LAB in the tectonically active EHS, which resembles the features of lithospheric discontinuity structure in the western United States and Phanerozoic southern and eastern United States imaged by Ps and Sp receiver functions [Abt et al., 2010; Fischer et al., 2010; Rychert and Shearer, 2009]. In agreement with the previous studies, the mechanism of the shallow LAB with sharp velocity contrast cannot be explained by thermal gradient alone and probably requires a contrast in composition, hydration, melting, or vertical anisotropy [Fischer et al., 2010; Sodoudi et al., 2011].

[22] The remarkable gap in the LAB of the TP is located within the Namche Barwa syntaxis and coincides with the high-low-high anomaly pattern of the shear wave absolutely velocity to a depth of ~120 km, inverted from the Rayleigh wave tomography [Fu et al., 2010]. Thus, the formation of this gap may have resulted from the delamination of the mantle lithosphere due to local mantle upwelling at the that location (as shown in Figure 9), which is supported by the existence of a low-velocity anomaly in the upper mantle, observed by the Rayleigh wave and body wave tomography [Fu et al., 2010]. Consequently, delamination of the Tibetan mantle lithosphere may have driven the rapid uplift and denudation rates indicated by geological and geochronological investigations in the Namche Barwa syntaxis [Ding et al., 2001; Zhang et al., 2004]. Our identified LAB of the Tibetan lithosphere is unclear in the body wave tomographic image [Ren and Shen, 2008; Zhang et al., 2011, 2012], which probably arises from the applied smoothing and smearing effect in the structure.

Figure 9.

Cartoon showing the geodynamic setting on the formation of the Namche Barwa syntaxis. The approximate location of this cartoon is shown in Figure 6. The delamination of the Tibetan plate and southward subduction of the Asian plate beneath the Namche Barwa syntaxis are suggested based on the migrated SRF images.

[23] We also distinguish another negative boundary located at a depth of ~180 km in the SRF images. In NE-trending profiles A and B, this layer extends from 30°N to north of the BNS, and is discontinuous but reliably recognized in E-W trending profiles C and D. Given that the existence of the LAB of the TP and the complex collision environment, we attribute this discontinuity to represent the LAB of the Asian plate (AP). For this interpretation, we mainly refer to the previous reports from the receiver function studies indicating that the Asian LAB is imaged at a depth of 140–180 km beneath central-northern Tibet [Kumar et al., 2006; Zhao et al., 2011]. From this interpretation, we conclude that the southward subduction of the Asian lithosphere, with different distances in an E-W direction, has reached to about 30°N under the Lhasa terrane (Figures 7-9).

[24] It is noteworthy that a third negative boundary, which is at a depth of ~150 km and lies in the vicinity of BNS, also appears in the SRF images. This boundary is shown as a small area in the NE-SW–oriented profiles A and B, and spans ~200 km in the E-W profiles C and D. At present, this discontinuity is ambiguous and cannot be easily explained, so we tentatively link it with the LAB of a fragment of the detached lithosphere denoted by X in Figures 7-9.

[25] Our results fail to image the subduction of the Indian plate, but they do not deny its presence. The lack of resolution of the Indian plate is mainly attributed to (1) a lack of the essential data coverage in the Himalayan terrane, (2) the subducting Indian plate does not go beyond the IYS [Li et al., 2008], and (3) the difficulty of identifying the Indian plate because of the complicated structure and low conversion signals.

5 Conclusions

[26] By analyzing the P and S receiver functions calculated from 44 seismic stations in the eastern Himalayan syntaxis, we highlight the detailed crustal structure and the LAB in our migrated images. The main conclusions that contribute to understand the geodynamic process of continental subduction and lithospheric deformation are summarized as follows:

  1. The crustal thickness shows a NE-deepening trend beneath the EHS, ranging from 55 to 75 km. Two anomalous areas with thinner and thicker curst are observed in the west and east of the Namche Barwa syntaxism, respectively. The Poisson's ratios within the EHS fall between 0.20 and 0.29 and are low in north along BNS and moderate elsewhere with some high values in the south of the IYS, suggesting that the crust consists of felsic and intermediate rocks. Together with the patches of the low-velocity zones, we infer that crustal thickening is achieved by extrusion of the mafic lower crust and/or thrusting and folding of the felsic upper and middle crust.
  2. Our migrated images reveal that the LAB of the Tibetan plate is located at a depth of ~110 km, indicating a gap beneath the Namche Barwa syntaxis. The gap is likely the result of delamination, which occurred in response to local asthenosphere upwelling.
  3. We also image the LAB of the Asian plate at a depth of ~180 km, which implies that the Asian plate has advanced southward to about 30°N under the Lhasa terrane.

Acknowledgments

[27] We sincerely acknowledge IRIS, Prof. Anne Meltzer, and Project team of Namche Barwar seismic experiment deployed by Lehigh University for making the seismic data available. We thank the Associate Editor and two anonymous reviewers for their constructive suggestions that help to improve this paper. This research is funded by National Natural Science Foundation of China (grants 41104055, 40930317, 41174036, and 41174069), the Open Research Program of State Key Laboratory of Geological Processes and Mineral Resources (grant GPMR201036), the NSFC Innovation Research Group Fund (grant 41021001), and Project SinoProbe-02-03. Xiaohui Yuan kindly provides his codes for mostly receiver function analysis. We have made use of the Seismic Handler (SH) for processing the seismic data and the General Mapping Tool (GMT) for preparing the figures.