Earth and Planetary Science Letters 203 (2002) 25^33 www.elsevier.com/locate/epsl

Seismic evidence for stepwise thickening of the crust across the NE Tibetan plateau Je¤rome Vergne a , Ge¤rard Wittlinger a; , Qiang Hui b , Paul Tapponnier c , Georges Poupinet d , Jiang Mei b , Georges Herquel a , Anne Paul d a

Institut de Physique du Globe de Strasbourg, Centre National de la Recherche Scienti¢que, Universite¤ Louis Pasteur, 5 rue Descartes, 67084 Strasbourg, France b Institute of Mineral Deposits, Chinese Academy of Geological Sciences, Baiwanzhuang Road, 100037 Beijing, PR China c Laboratoire de Tectonique, Institut de Physique du globe de Paris, 4 place Jussieu, 75205 Paris, France d Laboratoire de Ge¤ophysique Interne et Tectonophysique, Centre National de la Recherche Scienti¢que, Universite¤ Joseph Fourier, Boite Postale 53, 38041 Grenoble, France Received 25 February 2002; received in revised form 4 July 2002; accepted 23 July 2002

Abstract We performed a receiver function analysis on teleseismic data recorded along two 550 km-long profiles crossing the northeastern Tibetan plateau. Results from time to depth migration, grid-search Vp /Vs determination and simulated annealing inversion of waveforms, reveal that the crust thickens from V50 km near the northern edge of the plateau to V80 km south of the Jinsha suture in the Qiang Tang block. Crustal thickening occurs in staircase fashion with steps located beneath the main, reactivated sutures. The Vp /Vs ratio, close to the global continental average does not suggest widespread partial melting but rather a more usual separation between an upper felsic and a lower mafic part within the northeastern Tibetan crust. 4 2002 Elsevier Science B.V. All rights reserved. Keywords: teleseismic signals; crust; velocity; S-waves; P-waves; Qinghai-Xizang Plateau

1. Introduction Recent seismic investigations [1^4] have con¢rmed the existence of signi¢cant variations in the thickness and structure of the crust beneath Tibet [5]. Mapping and understanding such variations is critical to test con£icting models of for-

* Corresponding author. Tel.: +33-3 90 24 00 73; Fax: +33-3 90 24 01 25. E-mail address: [email protected] (G. Wittlinger).

mation of the plateau [4,6^8]. Receiver function (RF) analysis [9], waveform modeling [10], or Pn arrival time analysis [11,12] yield controversial insights. For instance there are clear low-velocity zones (LVZ) in the middle^lower crust [1^4] but whether they are widespread or correspond to partial melt is debatable. Here we show, using teleseismic receiver functions along two V550 km-long pro¢les crossing the northeastern Tibetan plateau (Fig. 1) that the crust thickens southward, from V50 km in the Qaidam Basin to V80 km beneath the Qiang Tang block. The Moho deepens in staircase fashion by 10^12 km steps,

0012-821X / 02 / $ ^ see front matter 4 2002 Elsevier Science B.V. All rights reserved. PII: S 0 0 1 2 - 8 2 1 X ( 0 2 ) 0 0 8 5 3 - 1

EPSL 6372 24-9-02 Cyaan Magenta Geel Zwart

26

J. Vergne et al. / Earth and Planetary Science Letters 203 (2002) 25^33

Fig. 1. Target area map of the northeastern Tibetan plateau. Yushu-Gonghe and Lhasa-Golmud pro¢les where Sino-French passive seismic experiments in 1998 and 1993 were performed, are indicated. For clarity, only three-component stations (short period 0.2 Hz and PASSCAL broad-band stations) are represented. Faults and sutures are from [6].

one near the topographic edge of the plateau, and the others near reactivated Mesozoic sutures. The Vp /Vs ratio, which remains close to the global continental average, and the absence of a thick, continuous low-velocity zone do not suggest widespread partial melting within the northeastern Tibetan crust.

2. Receiver functions and crustal structure The RF technique [13] enhances structural response near the receiver to incident P-waves by isolating P to S conversions produced at crust or mantle interfaces (Fig. 2). Careful time-domain iterative deconvolution [14] brings down the magnitude threshold of usable earthquakes to MV5.0 thus improving azimuthal coverage. We use three methods to interpret the RFs. First, at each station, a robust grid-search stacking [15] of the PmS

and PPmS phases of the radial receiver functions (RRFs) is performed, to recover mean Moho depth and mean crustal Q = Vp /Vs ratio (Fig. 3). This stacking algorithm sums the amplitudes of the RRF at the predicted arrival times of the PmS and PPmS phases for di¡erent crustal thickness and Q ratios. The best estimates of crustal thickness and Q ratio are obtained when the two phases are stacked coherently (Fig. 3b). Second, simulated annealing inversion [16] of stacked RRFs having common Moho conversion points is used to recover crustal layer thicknesses and S-wave velocities (Fig. 4). Finally, a time-to-depth migration [17,18] of all the individual RRFs (about 1000 along each of the two pro¢les) is performed to recover a realistic and spatially continuous image of the P to S conversion interfaces (Fig. 5a and c). Migrated RRFs, ordered along section, help assess the continuity of converted phases, particularly the strong P to S conversion

EPSL 6372 24-9-02 Cyaan Magenta Geel Zwart

J. Vergne et al. / Earth and Planetary Science Letters 203 (2002) 25^33

Fig. 2. Schematic ray diagram for the direct and converted P- and S-waves at the Moho. Also shown the ¢rst multiple PPmS and SPmP.

at the Moho. The crust^mantle transition is characterized everywhere by a sharp S-velocity increase (from 0.35 km/s up to 0.6 km/s) except just south of the Kunlun fault, along the YushuGonghe pro¢le, where the lower crust/upper mantle transition is smoother (Fig. 5a). In contrast with the average sharp Moho velocity increase on our pro¢les, other studies in southern Tibet [9] show more gradual crust^mantle transitions.

27

The higher frequency content (0.1^0.8 Hz) of our data may account for the higher resolution. From northeast to southwest, the crust appears to thicken from V50 km beneath the Qaidam and Gonghe basins to V80 km south of the Jinsha suture. The Moho deepens in steps of up to 10 km near ¢rst-order geological features. Along both pro¢les, the mean elevation rises from V2.7 km to V4.8 km above sea level. The steps that interrupt the concurrent, gentle deepening of the southwest-dipping Moho may be interpreted to re£ect successive ramps beneath which the lower crust was underthrust southwestwards. Along the Lhasa-Golmud transect, where such steps have already been detected [4,10,12,19], the RF analysis performed on combined short period (SinoFrench experiment [19]) and ¢ltered broad-band data (PASSCAL experiment [4]), provides a Moho interface image with two major steps. The northern one coincides roughly with the north edge of the Kunlun range, where the Qaidam crust subducts beneath the plateau [6], and the southern one with the Jinsha suture and Fenguo

Fig. 3. (a) Stack over 10‡ epicentral distance intervals of RRFs at station 105, Yushu-Gonghe pro¢le, arranged in increasing distance order. Numbers on right side are numbers of RRF stacked for each distance interval. Triangles indicate computed arrival times of phases PmS and PPmS corresponding to Moho depth and Vp /Vs ratio as read in (b) and mean crustal velocity of 6.2 km/s. (b) Grid-search stacking of the RRFs at station 105 (a). Stack becomes constructive and reaches maximum (arbitrarily set to 1) for the right values of Moho depth and Vp /Vs ratio; for station 105 these values are respectively V56 km and V1.65.

EPSL 6372 24-9-02 Cyaan Magenta Geel Zwart

28

J. Vergne et al. / Earth and Planetary Science Letters 203 (2002) 25^33

Shan. On the Yushu-Gonghe pro¢le, north of the Kunlun fault, the crustal structure seems to be complex so much so that Moho identi¢cation becomes unsure. However the Moho depth increases from V55 km near the Gonghe Basin rim to V65 km just south of the Kunlun fault. The proximity (60 km) between that fault and the plateau rim south of Golmud might account for the single, larger step there. The 10 km Moho step south of Gonghe corresponds to the V2 km elevation increase that marks the north edge of the plateau, and to two large south-dipping thrusts, one of which slipped during the MS = 6.9, 26/04/ 1990 earthquake [20]. This step thus re£ects the ramp along which the crust of the Gonghe Basin is underthrust beneath the plateau [6], whose elevation is supported by thick crust. Despite their similar amplitude, the steps farther south are not associated with a signi¢cant increase in elevation. They roughly correspond, however, to the passage of the great strike-slip fault zones that partly follow the main Mesozoic sutures of northern Tibet: the Kunlun fault and Anyemaqin suture to the north, and the Xianshuihe fault and Jinsha suture to the south. Such zones, which separate radically di¡erent crustal blocks, appear to have played a major role in the growth of Tibet since the Eocene [6,8].

3. Vp /Vs ratio and crustal composition The crustal low-velocity zones found thus far in Tibet have fostered much interest, even though there is little evidence that they extend beneath the entire plateau. Based mostly on re£ection bright spots [1] and low resistivity [21], the LVZ found beneath the Yangbajin graben north of Lhasa [9] was inferred to re£ect partially molten crust [1^3]. In northern Tibet, a similar inference was based on high values of the Vp /Vs ratio (Q) derived from direct shear-waveform modeling [4], implying low shear and normal compressional velocities. To assess whether low velocities re£ect temperature increase or compositional change, both the P- and S-wave velocities, hence their ratio Q must be determined. The value of Q helps discriminate

Fig. 4. Simulated annealing (SA) inversion results for stacks of the clearest RRFs at stations located in the three main tectonic units crossed by the Yushu-Gonghe pro¢le. At left, mean shear velocity (bold line) versus depth for the 50 best solutions (in light gray). Arrows indicate Moho interface. The mean crustal Q ratio obtained by SA inversion is also indicated. At right, comparison of computed using velocity model given at left (dashed line) and observed stacked RF (solid line). Arrows indicate the P^S conversion corresponding to the Moho. Simulated annealing is a fast grid-search method that explores entire model space, and there is no need to introduce an initial model. We ¢nd that using only four crustal layers with layer thickness steps of 1 km and S-wave and P-wave velocity steps of 0.1 km/s avoids adjustment of noisy RRFs by ad hoc thin layers at the expense of a less good ¢t of cost function (here root mean square of di¡erence between observed and computed RRF). Final crustal model is obtained by averaging 50 solutions having best ¢t to data. Span of parameter variation for best solutions (gray zones) is an indication of the existence of a unique class of ‘best’ solutions and yields a rough measure of error on model parameters. Even if absolute velocities are not well constrained due to velocity depth trade-o¡, relative velocities and thus velocity contrasts are much better de¢ned.

between compositional e¡ects or partial melting [22,23]. Grid-search stacking (Fig. 3b) yields normal to low Q ranging from 1.65 to 1.84 (meanV1.74) along the Yushu-Gonghe pro¢le (Fig. 5b), and from 1.72 to 1.8 (meanV1.74) along the Lhasa-Golmud pro¢le (Fig. 5d). SA inversion (Fig. 4) yields very similar Q ratios and

EPSL 6372 24-9-02 Cyaan Magenta Geel Zwart

J. Vergne et al. / Earth and Planetary Science Letters 203 (2002) 25^33

29

Fig. 5. (a) Plot of time-to-depth migrated RRFs along Yushu-Gonghe pro¢le (mean pro¢le azimuth VN48‡E). For migration we used reference earth P velocity models with Moho at 65 km depth and Vp /Vs ratio determined by grid-search stacking (Fig. 3). Origin on pro¢le is at Kunlun fault and sea level. Blue to red and gray to black colors indicate increasing positive, respectively negative, P-to-S conversion amplitudes. Arrows underline Moho steps. (b) Moho depth and Vp /Vs ratio, deduced from gridsearch stacking as in Fig. 3b, plotted along the pro¢le. The single, sharp V10 km step a few kilometers north of the Jinsha suture is also clear on this diagram. (c) Same as (a) along Lhasa-Golmud pro¢le (Az = N40‡E), origin is still at Kunlun fault. PASSCAL broad-band stations also used in this study are indicated by red triangles, short period 3C stations from Sino-French pro¢le are black triangles. (d) Same as (b). Only broad-band PASSCAL stations have, individually, enough data to constrain Vp /Vs well. In Bayan Har Vp /Vs is obtained with a stack of RRFs from four short period stations.

EPSL 6372 24-9-02 Cyaan Magenta Geel Zwart

30

J. Vergne et al. / Earth and Planetary Science Letters 203 (2002) 25^33

Fig. 6. Wadati diagram for local earthquakes, the di¡erence between S and P arrival times is plotted versus the P arrival time at each station. The linear ¢t slope approximates the Q = Vp /Vs ratio in the upper and middle crust, here 1.63 6 Q 6 1.80. Note that source location and velocity model are not needed for this plot.

Moho depths. Moreover, robust Wadati diagrams [24] from local earthquakes occurring near the Yushu-Gonghe pro¢le con¢rm the same range of Q values (Fig. 6). On the Yushu-Gonghe pro¢le, north of the Kunlun fault, Q is unusually low (1.65). In the vicinity of the fault it is higher (1.77, 1.84 and 1.89 at PASSCAL station MAQI). Southwards throughout Bayan Har, and across the Jinsha suture, a nearly constant normal Q value (1.73 to 1.78) prevails. The very low values observed north of the Kunlun fault are most probably associated with an average felsic composition, and ma¢c lower crust is likely absent [23,25,26]. The absence of high-velocity lower crust is also obvious from the inversions (Fig. 4a). This 55 km thickness of predominantly felsic rocks may indicate that the crust of the North Kunlun block thickens mostly in its upper part, probably by successive intra-crustal thrusts [6,8]. This rapid thickening and shortening (V15 mm/ yr, [6]) may have pushed the initial ma¢c lower part of the crust to deeper levels. Such sinking might have caused phase transition of the ma¢c

rocks to eclogite facies, transforming the lower crust into mantle-like materials. Such a scenario, based on very similar Vp /Vs observations has already been suggested for the Southern Altiplano crust in Bolivia [27] and by an explosion seismology study [26]. As a corollary to these low values obtained north of the Kunlun fault, the classical Q values obtained for the Qiang Tang and Bayan Har terranes may indicate that in these regions, the crust should be divided into a felsic upper part and a ma¢c lower part underlined by a weak Conrad interface. The V50 km-wide strip with higher Q values south of the Kunlun fault, near the Anyemaqin suture, may result from the existence of more ultrama¢c and ma¢c rocks [28]. Alternatively, the origin of such localized high Q values might be rising melts due to continental subduction [6,8,29]. On the Lhasa-Golmud transect, short period and broad-band RRFs also yield normal Q values (1.74) that contrast with the high values (Q up to 2.1) obtained with waveform modeling of shear waves [4]. The main reason for this discrepancy might originate from the use of di¡erent seismic phases. Instead of being deduced from PmS and PPmS waves, Q can alternatively be deduced (as in [4]) from the travel time di¡erences between an S-wave, the corresponding S to P conversion at the Moho (SmP) and the SPmP multiple phase corresponding to an S-wave re£ected and converted in a P-wave at the free surface, and then re£ected back from the Moho (Fig. 2). The synthetic S-waveforms computed in the one-layer model inferred from our RF analysis (Fig. 7b) are not compatible with the observed S-waveforms. On the other hand, the synthetic RFs computed in the model proposed by [4] are again not compatible with our RF observations (Fig. 7a). However, synthetic S-waveforms and RFs computed in a two-layer model, with an upper felsic and lower ma¢c part as in Fig. 7c, are both compatible with the S-waveforms and RF observations. The presence of a ma¢c lower crust characterized by a high Q and a high Vp , might result in a greater contrast for Vp at the interface i between the upper felsic crust and the lower ma¢c crust

EPSL 6372 24-9-02 Cyaan Magenta Geel Zwart

J. Vergne et al. / Earth and Planetary Science Letters 203 (2002) 25^33

31

Fig. 7. Comparison of synthetic S-waveforms on vertical component (left) and P-wave receiver functions on radial component (right) for di¡erent crustal models (middle). Synthetic seismograms are computed with a re£ectivity algorithm and a ray parameter of 0.11 s/km for the incident S-wave and 0.07 s/km for the incident P-wave. Ray paths of the di¡erent labelled phases are shown in Fig. 2. In the phase labelling, lowercase m and i stand for the Moho and the intra-crustal layer. Models represent (a) a thin and almost partially melted crust, as deduced from S-waveform observations near the Jinsha suture (see [4]), (b) a uniform thick crust with no melt deduced from grid-search analysis on RFs in the same region and (c) a thick crust divided into an upper felsic and a lower ma¢c part. Moho depth and mean crustal Vp /Vs ratio of model (c) are same as for model (b). Vp /Vs ratio in each layer is also indicated. Note nearly identical arrival time of SmP (on S-waveforms) and PmS (on P-wave RFs) phases in the three models. In model (c) phase SPmP is of very low amplitude and phase SPiP has identical arrival time as phase SPmP in model (a).

than at the Moho. In this case the SPmP phase would be less energetic than the SPiP phase (Fig. 7c) and confusion between those two phases may result in abnormally high Q and thin crust. On the other hand, the same high Q restricted to the ma¢c lower crust implies that the contrast in Vs at interface i is less than at the Moho. Thus the use of PPmS and PPiS phases (as in this study) instead of SPmP and SPiP phases (as in [4]) avoids misidenti¢cation of the phases. The S-wave observations reported by Owens and Zandt [4] ¢t as well the single-layer crustal model preferred by these authors (Fig. 7a) as the two-layer model we present here (Fig. 7c), which is also in agreement

with the thick crust and normal Q value we observe using P-waves. Our results thus do not support the existence of widespread partial melting in northeast Tibet but show instead that despite its important thickness, the crust can still be separated into an upper felsic and a lower ma¢c part. Moreover, the relatively small thickness of this ma¢c lower crust, that vanishes in the North Kunlun block along the YushuGonghe pro¢le and does not exceed one third of the total crustal thickness elsewhere, supports the idea that the thickening and shortening of the northeastern Tibetan crust is preferentially accommodated in its weak upper felsic part.

EPSL 6372 24-9-02 Cyaan Magenta Geel Zwart

32

J. Vergne et al. / Earth and Planetary Science Letters 203 (2002) 25^33

4. Conclusion In summary, our data along the two pro¢les clearly show low to normal Vp /Vs ratio. Our results also yield an image of the crust of northeast Tibet that shows southwestward deepening of the crust^mantle interface, consistent with the results obtained by seismic refraction [26], with Moho ramps located roughly beneath ¢rst-order topographic or geological boundaries. Such an image is compatible with the successive stacking of large crustal thrust wedges that would have grown northeastwards. Both the low^normal Q and the Moho steps, which would probably not survive if rapid ductile £ow prevailed, are not in favor of signi¢cant partial melting in the middle^lower crust of the northern Tibet plateau. So, why is the plateau £at and not the Moho if there is no weak middle layer that decouples the deformation from the surface ? The alternative would be that the Moho topography re£ects lateral density variation in a weak upper mantle.

[3]

[4]

[5]

[6]

[7]

[8]

[9]

Acknowledgements This study is part of a long-term cooperation between Institut National des Sciences de l’Univers (French CNRS) and the Institute of Geology, Chinese Academy of Geological Sciences. Su Heping skillfully set up the ¢eld logistic. Seismic stations where provided by French ‘Lithoscope’ organization. We are grateful to Jean Virieux for his remarks and to George Zandt for his thoughtful and constructive review.[AC]

[10]

[11]

[12]

[13]

[14]

References [1] K.D. Nelson, W. Zhao, L.D. Brown, J. Kuo, J. Che, X. Liu, S.L. Klemperer, Y. Makovsky, R. Meissner, J. Mechie, R. Kind, F. Wenzel, J. Ni, J. Nabelek, L. Chen, H. Tan, W. Wei, A.G. Jones, J. Booker, M. Unsworth, W.S.F. Kidd, M. Hauck, D. Alsdorf, A. Ross, M. Cogan, C. Wu, E.A. Sandvol, M. Edwards, Partially molten middle crust beneath southern Tibet: synthesis of project INDEPTH results, Science 274 (1996) 1684^1688. [2] Y. Makovsky, S.L. Klemperer, L. Ratschbacher, L.D.

[15]

[16]

[17]

Brown, M. Li, W. Zhao, F. Meng, INDEPTH wide-angle re£ection observation of P-wave-to-S-wave conversion from crustal bright spot in Tibet, Science 274 (1996) 1690^1691. R. Kind, J.F. Ni, W. Zhao, J. Wu, X. Yuan, L. Zhao, E.A. Sandvol, C. Reese, J. Nabelek, T. Hearn, Evidence from earthquake data for a partially molten crustal layer in southern Tibet, Science 274 (1996) 1692^1694. T.J. Owens, G. Zandt, Implication of crustal property variations for models of Tibetan plateau evolution, Nature 387 (1997) 37^43. A. Hirn, M. Jiang, M. Sapin, J. Diaz, A. Nercessian, Q.T. Lu, J.C. Lepine, D.N. Shi, M. Sachpazi, M.R. Pandey, K. Ma, J. Gallart, Crustal structure and variability of the Himalayan border of Tibet, Nature 307 (1984) 23^25. B. Meyer, P. Tapponnier, L. Bourjot, F. Me¤tivier, Y. Gaudemer, G. Pelzer, G. Shumin, C. Zhitai, Crustal thickening in Gansu-Qinghai, lithospheric mantle subduction, and oblique, strike-slip controlled growth of the Tibet plateau, Geophys. J. Int. 135 (1998) 1^47. G.A. Houseman, P.C. England, Finite strain calculations of continental deformation. 1. Method and general results for convergent zones, J. Geophys. Res. 91 (1986) 3651^ 3663. P. Tapponnier, X. Zhiqin, F. Groger, B. Meyer, N. Arnaud, G. Wittlinger, Y. Jingsui, Oblique stepwise rise and growth of the Tibet Plateau, Science 294 (2001) 1671^ 1677. Y. Yuan, J.F. Ni, R. Kind, J. Mechie, E.A. Sandvol, Lithospheric and upper mantle structure of southern Tibet from a seismological passive source experiment, J. Geophys. Res. 102 (1997) 27491^27500. L. Zhu, D.V. Helmberger, Moho o¡set across the northern margin of the Tibetan plateau, Science 281 (1998) 1170^1172. D.E. McNamara, T.J. Owens, W.R. Walter, Observation of regional phase propagation across the Tibetan plateau, J. Geophys. Res. 100 (1995) 22215^22229. G. Herquel, G. Wittlinger, J. Guilbert, Anisotropy and crustal thickness of northern Tibet. New constraints for tectonic modelling, Geophys. Res. Lett. 22 (1995) 1925^ 3228. C.A. Langston, Structure under Mount Rainer, Washington, inferred from teleseismic body waves, J. Geophys. Res. 84 (1979) 4749^4762. J.P. Ligoria, C.J. Ammon, Iterative deconvolution and receiver-function estimation, Bull. Seism. Soc. Am. 89 (1999) 1395^1400. L. Zhu, H. Kanamori, Moho depth variation in southern California from teleseismic receiver functions, J. Geophys. Res. 105 (2000) 2969^2980. T. Shibutani, M. Sambridge, B. Kennett, Genetic algorithm inversion for receiver functions with application to crust and uppermost mantle structure beneath Eastern Australia, Geophys. Res. Lett. 23 (1996) 1829^1832. C.H. Jones, R.A. Phinney, Seismic structure of the lithosphere from teleseismic converted arrivals observed at

EPSL 6372 24-9-02 Cyaan Magenta Geel Zwart

J. Vergne et al. / Earth and Planetary Science Letters 203 (2002) 25^33

[18]

[19]

[20]

[21]

[22]

small arrays in the Southern Sierra Nevada and vicinity, California, J. Geophys. Res. 103 (1998) 10065^10090. G. Kosarev, R. Kind, S. Sobolev, X. Yuan, W. Hanka, S. Oreshin, Seismic evidence for a detached Indian lithospheric mantle beneath Tibet, Science 283 (1999) 1306^ 1309. G. Wittlinger, P. Tapponnier, G. Poupinet, J. Mei, S. Danian, G. Herquel, F. Masson, Seismic tomography of northern Tibet and Kunlun: evidence for crustal blocks and mantle velocity contrast, Earth Planet. Sci. Lett. 139 (1996) 263^279. Y.T. Chen, L.S. Xu, X. Li, M. Zhao, Source process of the 1990 Gonghe, China, earthquake and tectonic stress ¢eld in the northeastern Qinghai-Xizang (Tibetan) plateau, Pageophysics 146 (1996) 697^715. L. Chen, J. Booker, A. Jones, N. Wu, M. Unsworth, W. Wei, H. Tan, Electrically conductive crust in southern Tibet from INDEPTH magnetotelluric surveying, Science 274 (1996) 1694^1696. G. Zandt, C.J. Ammon, Continental crust composition constrained by measurements of crustal Poisson’s ratio, Nature 374 (1995) 152^154.

33

[23] N.I. Christiansen, Poisson’s ratio and crustal seismology, J. Geophys. Res. 102 (1996) 3139^3156. [24] C. Kisslinger, E.R. Engdahl, The interpretation of the Wadati diagram with relaxed assumptions, Bull. Seism. Soc. Am. 63 (1973) 1723^1736. [25] R.L. Rudnick, D.M. Fountain, Nature and composition of the continental crust: a lower crustal perspective, Rev. Geophys. 33 (1995) 267^309. [26] A. Galve¤, A. Hirn, M. Jiang, J. Gallart, B. de Voogd, J.C. Lepine, J. Diaz, Y. Wang, H. Qian, Mode of raising northeastern Tibet probed by explosion seismology, Earth Planet. Sci. Lett., S0012-821X(02)00863-4. [27] G. Zandt, A.A. Velasco, S.L. Beck, Composition and thickness of the southern Altiplano crust, Bolivia, Geology 22 (1994) 1003^1006. [28] Geological structure map of Qinghai province PRC, Geol. Publishing House, Beijing, 1991. [29] N.O. Arnaud, P. Vidal, P. Tapponnier, P. Matte, W. Deng, The high KO volcanism of northwestern Tibet: geochemistry and tectonic implications, Earth Planet. Sci. Lett. 111 (1992) 351^367.

EPSL 6372 24-9-02 Cyaan Magenta Geel Zwart