Indian mantle corner flow at southern Tibet revealed by shear wave splitting measurements



[1] We constrain anistropic seismic structure in the southernmost Tibet and the Tethyan Himalays using SKS and SKKS phases recorded from a temporary seismic network, operated from July 2004 to August 2005 in conjunction with the Hi-Climb project. Shear wave splitting is not detected at 16 stations, most of which are located near and to the north of the Indus-Yalong suture (IYS). For the first time anisotropy with an N-S fast direction and 0.4 ∼ 1 s delay times is observed to the region about 50 km south of the IYS. This weak anisotropy correlates with the flat part of the subducting Indian lithosphere and could be caused by northward asthenospheric flow and shear at the base of the Indian lithosphere. The null measurements in the vicinity of the IYS are most likely the result of a vertical asthenospheric flow, the corner flow induced by the subvertical subduction of the Indian lithosphere just south of the Bangong-Nujiang suture (BNS). Observations of the systematic changes in mantle anisotropy from southern Tibet to the central Tibet provide evidence for different mantle flow fields between the Indian and the Eurasian asthenospheric mantle, which are separated by a subvertical Indian lithosphere at the BNS.

1. Introduction

[2] The Himalayas and the Tibetan plateau are a result of continental collision between the Eurasian and Indian plates since 55 Ma. However, the mechanism that created the majestic mountains and plateau is still open to debate. Various hypotheses have been proposed to explain the crust thickening of the Tibetan Plateau such as the subhorizontally underthrusting model [Argand, 1924; Ni and Barazangi, 1984], the convective thickening model [Houseman et al., 1981; Molnar et al., 1993], and the lower crust injection model [Zhao and Morgan, 1987].

[3] Measurements of seismic anisotropy can provide insight into the nature of the finite strain field in the lithosphere and uppermost mantle because the anisotropy appears to result from the strain-induced lattice preferred orientation of anisotropic minerals, such as olivine and orthopyroxene [e.g., Silver and Chan, 1991]. Shear wave splitting measurements, mostly from SKS and SKKS phases, have been reported for India, the western Himalayas, and Tibet [Sandvol et al., 1994; McNamara et al., 1994; Lavé et al., 1996; Sandvol et al., 1997; Huang et al., 2000]. Since the scale of lateral variations and the depth distribution of anisotropy are usually not well constrained by the SKS data alone, interpretations of shear wave splitting results are not unique.

[4] In this study, we have analyzed splitting of shear waves recorded at 24 portable seismic stations in southern Tibet, where anisotropy measurements from previous studies are spares. Our results provide new constraints on azimuthal anisotropy in Himalayas and across the IYS and new insights into the dynamic process of continental collision in southern Tibet.

2. Data and Method

[5] From July 2004 to August 2005, a portable broadband seismic array of 24 stations, equipped with Guralp 3ESP sensors was deployed in southern Tibet from the Himalayas to 150 km north of the IYS as part of the second phase of the international collaborative Hi-CLIMB project [Nábělek et al., 2005] (Figure 1). We performed shear wave splitting analyses on both SKS and SKKS phases from teleseismic events with epicentral distances between 86 degree and 149 degree and Mb of 5.5 and above. Figure 1 shows the epicenters of 32 events that yield all reliable measurements at the stations. Although majority events are from the Tonga subduction zone in the southwest Pacific, several earthquakes from northwestern America and south Indian Ocean have broaden the azimuthal coverage, which improves the resolution of the shear wave splitting parameters (Figure 2).

Figure 1.

Shaded relief map of the array showing the results of the shear wave splitting analyses. Filled black circles denote stations with clear splitting and the bars attached to the black circles indicate both the orientation of the fast directions and the delay times. Filled white circles with two orthogonal lines represent the absence of splitting from one back azimuth or two orthogonal back azimuths. Splitting at these stations is possible if the fast direction lies along either of the two directions. Filled white circles alone denote the absence of splitting from at least two non-orthogonal back azimuths. Insert shows the distribution of 32 teleseismic events used in this study.

Figure 2.

Seismograms of SKS phases from 3 events with different back azimuths recorded at station T032 of the Hi-CLIMB array. Note very little energy of SKS phase on the tangential components.

[6] The method of Silver and Chan [1991] was adopted to determine the splitting parameters, fast direction (ϕ) and delay time (δt), which characterize seismic anisotropy. This method involves searching throughout the shear wave splitting parameter space for the fast and slow coordinate system that comes closest to having a singular covariance matrix of two horizontal components. The covariance matrix from the two horizontal components of a shear wave will be singular if that shear wave does not split. Thus the most optimum splitting parameters can be obtained by minimizing one of the two eigenvalues of the covariance matrix. One example is shown in Figure 3. Splitting of SKS or SKKS is reliable when the tangential component shows substantial energy. Here the shapes of the splitting fast and slow waves are remarkably similar, and the particle motion is close to elliptic. When the effect of splitting is removed, the energy on the tangential component becomes negligible and the particle motion is almost linear. The minimum energy on the tangential component is well localized on the contour plot as shown in Figure 3, which yields the optimum values for the splitting parameters of fast direction and delay time.

Figure 3.

Splitting analysis of the SKS arrival recorded at station T014 from the October 23 2004 Tonga event. (a) The original horizontal waveforms; (b) the horizontal waveforms corrected for the determined fast direction and delay time; (c) uncorrected fast and slow components; (d) corrected fast and slow components for the delay time; (e) uncorrected horizontal particle motion of SKS waves; (f) corrected horizontal particle motion of SKS waves for the delay time; and (g) contour plot for the optimum splitting parameters (black dot) in the fast direction and delay time space. The smallest contour indicates the 95% confidence level.

3. Results

[7] The average polarization direction of fast shear wave and the average delay time at each station are shown in Figure 1 and Table 1. Most of stations with null measurements are located close to the IYS from 29°N to 30°N (Figure 1). Sixteen stations exhibit no detectable shear wave splitting. The lack of azimuthal anisotropy is confirmed at four stations (T032, T028, T019 and T007) where null measurements are obtained for events from a broad back-azimuth range. Null measurements at the twelve stations (white circles with crosses in Figure 1) are calculated from events that are from one or two orthogonal back azimuth groups. These null results either reflect an apparent isotropic structure beneath the stations or suggest that fast directions are parallel or orthogonal to the event back azimuths at these stations, where the majority of the SKS data are polarized along an approximately ESE-WNW direction with a back-azimuth range of 105–120 degrees from north.

Table 1. Shear Wave Splitting Parameters Averaged by Stationa
StationLat.Lon.ϕ, °δt, sNo.StationLat.Lon.ϕ, °δt, sNo.
  • a

    Uncertainties are 95% confidence limits. Delay time and fast direction are not given for stations where no splitting was observed.

T03830.4486.9335 ± 51 ± 0.28T01928.8988.789
T03229.7788.3519T01428.5487.764 ± 60.7 ± 0.19
T02829.4388.238T00828.2688.996 ± 100.7 ± 0.211
T02529.1488.166 ± 80.6 ± 0.217T00628.1688.106 ± 100.4 ± 0.24
T02128.9089.584T00428.1687.369 ± 50.7 ± 0.36
T02028.9489.1652 ± 60.7 ± 0.210T00128.0187.683 ± 41.1 ± 0.36

[8] Despite the dominant nulls in our overall results, we observed shear wave splitting at eight stations (Figure 1). Seven of them are located south of the IYS and one station (T038) is located 100 km north of the IYS. The delay times are in general less than 1 s, which is almost half of the delay times (∼1.5 s) observed at the INDEPTH III stations in central Tibet [Huang et al., 2000]. A complex pattern of anisotropy is observed in the region immediately south of the IYS, where both fast directions in NS (T025) and NE-SW (T020) are observed along with the 12 stations with null results. There is an apparent change in both the fast direction and delay time north of 29°N. South of this latitude, seismic anisotropy shows nearly an N-S fast direction with a small average delay time of ∼0.7 s (Figure 1). The region of N-S anisotropy is approximately inline with the flat part of the Indian continental lithosphere imaged from previous studies [e.g., Tilmann et al., 2003; Schulte-Pelkum et al., 2005]. More importantly, null measurements are observed at all stations to the north of the IYS (Figure 1) except at the northernmost station (T038), where seismic anisotropy shows a NE-SW fast direction with a delay time of ∼1 s.

4. Discussion

[9] Built on previous studies, results of this study allow us to investigate the variation of seismic anisotropy across the Tibetan Plateau (Figure 4). Previous shear wave splitting analyses have indicated that in northern Tibet, there was a progressive and coherent change of the fast polarization directions from NE-SW to EW [McNamara et al., 1994]. In central Tibet, Huang et al. [2000] reported substantial splitting with fast directions oriented E-W and an average delay time of 1.5 s north of the BNS but essentially null measurements south of the BNS. Using the INDEPTH II's portable seismic arrays in southern Tibet and the Himalayas, Sandvol et al. [1997] showed no coherent azimuthal shear wave splitting in the Himalayas-Tibet collision zone except for two northern stations located about 100 km north of the IYS. Our study adds new data and shear wave splitting results both in the region north of the High Himalayas and in the vicinity of the IYS.

Figure 4.

Map of shear wave splitting results in Tibet and northern India (modified from Huang et al. [2000]). Results are from McNamara et al. [1994], Sandvol et al. [1997], Chen and Özalaybey [1998], Huang et al. [2000], and this study. The dark shaded region outlines the high Sn attenuation zone in north and central Tibet. Black circles without lines are stations with only null measurements. The light shaded rectangle shows the zone of null splitting results, which is roughly centered along the IYS.

[10] The N-S fast orientation at 28°N near the southern boundary of Tibet (Figure 4) is observed for the first time in this area. In southern Tibet, the India lithosphere moves northward relative to the Eurasia predicted by global plate tectonic models. GPS measurements also confirm the dominant northward surface motion in southern Tibet [Wang et al., 2001]. In addition, the Indian lithosphere itself is mostly isotropic as suggested by Chen and Özalaybey [1998] and therefore, it makes little contribution to the observed anisotropy. Given that the region of N-S anisotropy observed in this study (Figure 4) is approximately located above the flat segment of the subducting Indian lithosphere, this anisotropy is more likely due to the northward moving mantle flow in the asthenosphere and induced shear at the base of the Indian lithosphere (Figure 5).

Figure 5.

Cartoon of the Himalaya-Tibet continental collision zone (modified from Tilmann et al. [2003]). MFT, Main Frontal Thrust; MCT, Main Central Thrust; STD, Southern Tibetan detachment; IYS, Indus-Yarlung suture; BNS, Bangong-Nujiang suture; JRS, Jinsha-River suture. Assuming an isotropic Indian lithosphere the N-S shear wave splitting in southern Tibet is caused by shear between Indian lithosphere and asthenosphere. The nulls near IYS could be explained by a corner flow induced by the subvertical subduction of the Indian lithosphere south of the BNS.

[11] The dominant null measurements near the IYS in this study are consistent with previous shear wave splitting observations between the IYS and the BNS sutures in southern Tibet (Figure 4). One possibility is a multilayered anisotropic model. This model can be identified if splitting measurements depend on azimuth. Since there are at least 4 stations with nulls that are measured from a broad azimuth range in our study area, we can rule out the multilayered anisotropy as a cause of the observed nulls near the IYS.

[12] An alternative explanation is that these nulls reflect anisotropy that has a vertical fast direction. Since the nulls in our study area are located south of the imaged subvertical Indian lithosphere [Tilmann et al., 2003], they are more likely associated with a vertical asthenospheric flow, a corner mantle flow resulted from the subvertical subduction of the Indian lithosphere (Figure 5).

[13] Our results indicate that there is a transition at 29°N from a weak, N-S aligned anisotropy in the south to an isotropic region near and to the north of the IYS. This latitude probably marks the change of the corner flow field from a northward to a downward direction (Figure 5). This simple model can well explain the observed N-S anisotropy at the southern edge of Tibet and nulls near the IYS. It supports the current notion that the BNS acts as the boundary between the Indian and Eurasian mantle, where the Indian lithosphere subducts nearly vertically to ∼400 km depth [Tilmann et al., 2003; Li et al., 2006] and anisotropy changes sharply from null in the south to EW in the north [Huang et al., 2000]. Our results also partially support the hypothesis that the Indian plate subhorizontally underthrust beneath the Himalayas and southern Tibet to the south of the BNS. Because seismic anisotropy inferred from SKS/SKKS splitting measurements is mainly due to the deformation in the mantle, our observation can not well constrain the flow in the lower crust in Tibet. The uncertainty in our model is largely due to the lack of azimuth coverage of seismic events. Long-term stations in southern Tibet would improve the data coverage and help to refine our simple Indian mantle corner flow model.


[14] Authors would like to thank John Nábelek and his HI-CLIMB field team for their hard work at the southern Tibet to collect the data used in this study. Eric Sandvol provided valuable help in data processing and SKS signal analysis. Constructive comments from two anonymous reviewers help improve the manuscript. Figures are produced by the GMT tool [Smith and Sandwell, 1997]. This study was supported by Chinese NSF grants (40520120222; 40125011; 40521002) and Peking University.