Elsevier

Tectonophysics

Volume 584, 22 January 2013, Pages 209-220
Tectonophysics

Crustal and uppermost mantle S velocity structure under Hi-CLIMB seismic array in central Tibetan Plateau from joint inversion of surface wave dispersion and receiver function data

https://doi.org/10.1016/j.tecto.2012.08.024Get rights and content

Abstract

Trade-off between the depth of an interface and the average velocity above the interface is well known in receiver function inversion. To resolve model ambiguity, surface wave dispersion data, which are sensitive to vertical shear velocity average, are commonly combined with P receiver functions in the inversion. In this paper, we design a joint inversion scheme using the neighborhood searching algorithm to solve shear velocity structure of the crust and uppermost mantle with receiver function and surface wave dispersion data. Two sets of stretched splines are implemented in the inversion to obtain smooth velocity profiles, separated by a specifically defined Moho discontinuity. We apply this method to the dense Hi-CLIMB seismic array in the central Tibetan Plateau (TP). The Indian crust is observed to subduct and underplate the Tibetan crust between the Yarlung-Zangbo suture and the Bangong-Nujiang suture, at the latitude of about 31.5°N. The crust thickens from about 50 km below the Indian foreland to over 75 km south of Bangong-Nujiang suture and turns shallower at around 65 km under the Qiangtang block. We identify two zones with complicated Moho structure beneath the northern Lhasa block and southern Qiangtang block, indicating highly deformed Moho morphology, and beneath the Himalaya block associated with the underthrusting of the Indian Plate, respectively. A low velocity zone is present in the mid-crust in most parts of the profile under the TP with up to more than 10% velocity reduction. The low velocity structure is not continuous, forming discontinuous patches rather than channels, calling for a revision of the channel flow model.

Highlights

► Designed a joint inversion scheme for shear velocity structure of the crust and uppermost mantle ► Search models are parameterized in stretched splines separated by a Moho of varying depths. ► Data include receiver functions and surface wave dispersions under the Hi‐CLIMB array. ► Results show a prominent low velocity zone in midcrust and a complex morphology of the Moho. ► The low velocity structure is not continuous, forming patches rather than channels.

Introduction

In seismic inversion, trade-offs between model parameters are well known. To resolve the ambiguity and to improve resolution, a combination of different data sets that have sensitivities to different parameters is required or a priori constraints have to be imposed. For example, receiver functions are an efficient tool to study the structure of the earth's interior including Moho depth (e.g., Zhu and Kanamori, 2000), crustal velocity models (e.g., Owens et al., 1984), and upper mantle discontinuities (e.g., Shen et al., 1998). However, it has long been recognized that teleseismic P receiver functions are sensitive to shear velocity contrast and depth–velocity product, instead of velocity alone. Therefore substantial trade-offs exist between the average wave velocity above the interface and the depth to the interface (Ammon et al., 1990). On the other hand, surface waves are primarily sensitive to vertical shear-velocity averages but insensitive to velocity discontinuities. Thus, combining these two complementary data by jointly inverting receiver functions and surface wave dispersions would greatly reduce depth–velocity ambiguity and has been commonly used to resolve depth resolution of S-velocity structure (e.g., Julia et al., 2000, Ozalaybey et al., 1997).

Different research groups have implemented different joint inversion schemes, among which the traditional least square based linearized inversions are widely used (Julia et al., 2000, Ozalaybey et al., 1997). However, due to the nonlinearity of the inversion problem and the dependence of the starting model based on a priori knowledge, a linearized inversion may not be the best approach in such an inversion. With the improvement of computational power, several nonlinear searching algorithms have been developed and applied in geophysical inversion problems. Commonly used algorithms include but are not limited to the genetic algorithm, the simulated annealing, and the neighborhood algorithm. The advantages of using searching algorithms include the following. 1) They can deal with nonlinear problems easier than traditional linear inversions. 2) They are less likely trapped in local minimums. 3) The location of Moho can be specifically defined. In a linearized inversion, the Moho is typically inferred from the resulting velocity profile. In a search method, the Moho can be specified as a search parameter. 4) Error analysis can be performed in a relatively straightforward way. However, a major drawback of search algorithms is that they are computationally intensive. Nevertheless, with the increase of computer power and the utilization of parallel computing, the computation time becomes less of a concern. In this paper, we adopt the neighborhood algorithm (NA) developed by Sambridge, 1999a, Sambridge, 1999b. The NA is an optimized direct searching algorithm that converges faster than traditional Monte Carlo searches and has been proven successful in various geophysics inversions. It makes use of geometrical constructs known as Voronoi cells in the search and appraisal stages. Recent NA development has also implemented message passing interface (MPI) for parallel processing that can greatly reduce the computational time.

Western China, marked by spectacular topographic relief, complex geology, and active deformations, is one of the most active regions of continental earthquakes in the world. Major basins, high plateaus, and great mountains all co-exist in the region, which are largely resulted from the Indian–Eurasian collision. The geological diversity and the active deformation make western China a hot spot for geological and geophysical studies. Mapping accurate crustal and upper mantle velocity structures will help answer many fundamental geological questions, such as the deformation of the Tibetan Plateau (TP) during continental collision, the relation between N–S shortening and E–W extension, mid crust channel flow in the TP, subduction of Indian mantle lithosphere, and underplating and delamination of the Indian crust. With the increase of new permanent and temporary seismic stations deployed in the TP during recent decades, various seismic studies have been carried out in this region. An integrated study of receiver function inversion, two-station Rayleigh wave phase velocity inversion and teleseismic P-waveform modeling by Kind et al. (1996) using INDEPTH-II seismic stations revealed prominent low velocity zone in the mid-crust north of the Yarlung-Zangbo suture, beneath the southern Lhasa block, while such feature is absent south of the suture under Tethyan Himalaya. Yuan et al. (1997) reported a Moho depth at 70–80 km and a second discontinuity at 50–60 km across entire INDEPTH-II profile south and north of the Yarlung-Zangbo suture. They also observed pronounced low-velocity zone north of the Yarlung-Zangbo suture at about 10–20 km depth. An independent receiver function study by Nabelek et al. (2009) using Hi-CLIMB data revealed weak Moho Ps conversion from 31° N to 33° N. Their study also suggested that the crust/mantle interface beneath Tibet is anisotropic. Similar Moho variation is also reported by Tseng et al. (2009) using SsPmp phase and by Nowack et al. (2010) using Gaussian beam migration. Hetenyi et al. (2006) estimated the effective elastic thickness (EET) of the Indian Plate from receiver functions, gravity anomalies, and thermomechanical modeling. They suggested the EET of the Indian Plate drops from ~ 60–70 km under the Indian Continent to ~ 20–30 km north of the Main Frontal Thrust due to the loss of the crustal strength. They also suggested a strong mantle in the study region between the Ganges basin and the central TP in order to be consistent with modeled flexural geometry. Wittlinger et al. (2009) reported the detection of an eclogite layer in the lower crust under the southern Lhasa block from grid search stacking of receiver functions. Hung et al. (2010) presented a first multi-scale finite-frequency P and S velocity model in the southern and central TP, revealing a sub-horizontally advancing Indian lithospheric mantle beyond Bangong-Nujiang suture as well as regions of low velocity structures in the crust that correlate with low electric resistivities (Unsworth et al., 2005). Sun et al. (2010) performed Rayleigh wave tomography on the Chinese continent using ambient noise correlation data. They reported widespread low velocity zones in the TP in the mid crust. By comparing the Rayleigh wave and Love wave dispersion, Shapiro et al. (2004) found strong S-velocity radial anisotropy with VSH > VSV in the mid to lower crust. The anisotropy was found to be the largest in the western TP, corresponding to the thinning of the mid crust. Although some first order common features are shared among these aforementioned studies, discrepancies and controversies exist, e.g. channel flow vs. isolated low velocity zone, strong mantle vs. weak mantle. In this paper, we jointly invert surface wave group and phase velocities and receiver functions from the Ganges basin in the Indian Plate to the central TP using seismic data recorded by Hi-CLIMB experiment stations. The near north–south dense linear array provides ideal geometry to construct high resolution 2D shear velocity profile.

Section snippets

Data and method

The NA algorithm developed by Sambridge, 1999a, Sambridge, 1999b has been widely used in geophysical inversions (e.g. Yao et al., 2008). It searches the model space to find “good” data fitting region. It employs the Voronoi cell to drive the search toward the best fit model while at the same time explores a relatively large variety of different models. As such with a good choice of control parameters, one can search the complete model space without being trapped in a local minimum. Another

S velocity from surface wave dispersion data

Since our data are composed of a large amount of surface wave dispersions, it is worthwhile to do tomography using surface wave dispersion alone. Fig. 3 shows the shear velocity profile along Hi-CLIMB linear array using the dispersion data only. The tomography yields a smooth image due to the nature of surface wave inversions. Nevertheless, a pronounced low velocity zone at a depth of around 25 km in the mid-crust continuously exists under the TP. The thickness and the strength of the low

Acknowledgments

The Hi-CLIMB data were retrieved from the IRIS Data Management Center. The ambient noise extraction of Green functions and the measurement of surface wave dispersions were based on modifications of codes from Mike Ritzwoller's group. Most figures were made using GMT software (Wessel and Smith, 1991). We benefited from discussions with Bob Nowack and Wang-Ping Chen and constructive comments from two anonymous reviewers. This research was supported by NSF 0838188 and AFRL FA9453-10-C-0216.

References (33)

  • C.J. Ammon et al.

    On the nonuniqueness of receiver function inversions

    Journal of Geophysical Research

    (1990)
  • G.D. Bensen et al.

    Processing seismic ambient noise data to obtain reliable broad-band surface wave dispersion measurements

    Geophysical Journal International

    (2007)
  • M.K. Clark et al.

    Topographic ooze: building the eastern margin of Tibet by lower crustal flow

    Geology

    (2000)
  • R.B. Herrmann

    Computer Programs in Seismology, Version 3.30

    (1991)
  • G. Hetenyi et al.

    The effective elastic thickness of the India Plate from receiver function imaging, gravity anomalies and thermomechanical modeling

    Geophysical Journal International

    (2006)
  • S.-H. Hung et al.

    First multi-scale, finite-frequency tomography illuminates 3-D anatomy of the Tibetan Plateau

    Geophysical Research Letters

    (2010)
  • J. Julia et al.

    Joint inversion of receiver function and surface wave dispersion observations

    Geophysical Journal International

    (2000)
  • R. Kind

    Seismic images of crust and upper mantle beneath Tibet: evidence for Eurasia plate subduction

    Science

    (2002)
  • J.M. Lees et al.

    Tomographic inversion for three-dimensional velocity structure at Mount St. Helens using earthquake data

    Journal of Geophysical Research

    (1989)
  • A.L. Levshin et al.

    Automated detection, extraction, and measurement of regional surface waves

    Pure and Applied Geophysics

    (2001)
  • J. Nabelek et al.

    Underplating in the Himalaya–Tibet collision zone revealed by the Hi-CLIMB experiment

    Science

    (2009)
  • J.E. Nafe et al.

    Physical properties of marine sediments

  • R.L. Nowack et al.

    Application of Gaussian beam migration to multi-scale imaging of the lithosphere beneath the Hi-CLIMB array in Tibet

    Bulletin of the Seismological Society of America

    (2010)
  • T.J. Owens et al.

    Implications of crustal property variations for models of Tibetan plateau evolution

    Nature

    (1997)
  • T.J. Owens et al.

    Seismic evidence of an ancient rift beneath the Cumberland Plateau, Tennessee: a detailed analysis of broadband teleseismic P waveforms

    Journal of Geophysical Research

    (1984)
  • S. Ozalaybey et al.

    Shear-wave velocity structure in north basin and range province from the combined analysis of receiver functions and surface waves

    Bulletin of the Seismological Society of America

    (1997)
  • Cited by (0)

    View full text