Mantle flow under the western United States from shear wave splitting

https://doi.org/10.1016/j.epsl.2006.05.010Get rights and content

Abstract

We show that SKS splitting in the westernmost United States (polarization of the fastest shear waves and splitting times, including their back-azimuthal dependence) can be explained by a geodynamic model that includes a continuum-mechanics description of plate motions and underlying asthenospheric circulation. Models that include a counterflow at depths of ∼ 300 km are preferred, which may indicate a far-field effect of the Farallon slab anomaly sinking underneath the central continental United States. This finding is broadly consistent with earlier suggestions, and we demonstrate that a mechanically coupled system, though with a strong viscosity contrast with depth, is consistent with the data. We explore the depth dependence of predicted anisotropy by means of computing seismogram synthetics and comparing synthetic splits with observations. Some patterns in the data, including null observations, are matched well. Linked models of geodynamic flow and mineral alignment in the mantle provide a means to test the relationship between strain and the saturation of texturing. Lower fabric saturation strains than for global models are preferred by the data, which may reflect the relatively active tectonic setting and thin asthenosphere of the study region. In general, our results show that seismic anisotropy, when interpreted jointly with mineral physics theories, may be used to quantitatively constrain the spatial character of flow, and the degree of force coupling, at depth.

Introduction

Parts of the upper mantle are seismically anisotropic and particularly shear wave splitting from SKS waves has often been used to infer deformation and flow at depth [1], [2]. Especially in continental regions such studies may be used to constrain the degree of lithosphere-mantle coupling [3], [4], an issue that is still debated. We wish to contribute to this discussion and test if a geodynamic model of mantle flow and inferred lattice preferred orientation (LPO) of olivine can explain the observed anisotropy. We focus on splitting observations in the westernmost United States (Fig. 1), for several reasons: the data coverage there is better than in geologically more simple regions such as oceanic plates and can be expected to improve significantly through the EarthScope program. For a continental region, the setting is relatively simple as the region is tectonically young and underlain by slow velocity anomalies, indicating a thin lithosphere [e.g. [5]]. Lastly, previous interpretations of the data have invoked a variation of the polarization orientation of the fastest shear wave (fast azimuth) with depth from parallel to the San Andreas Fault in the shallow mantle [6], [7], [8] to close to E–W deeper [6], [9], [10]. The splitting signal along the San Andreas has recently also been compared with the tectonic setting in New Zealand using a model that includes lateral viscosity contrasts in the lithosphere across the plate boundary [4].

Of particular interest here is the model by Silver and Holt [13] (hereafter: SH02), who used surface strain-rates from geodesy and SKS splitting to infer an EW oriented return flow underneath North America (in a hotspot reference frame), whereas the surface motion of the plate is NESW (in both hotspot and no-net-rotation reference frames). In the kinematic description chosen by SH02, a complete decoupling of the lithosphere from the underlying asthenospheric flow is suggested. Here, we substantiate our earlier findings that an alternative model with a continuum mechanics treatment of plate motion and underlying mantle flow fits the data just as well for typical radial viscosity profiles, though with a low viscosity, asthenospheric channel [14]. We use a global circulation model similar to Becker et al. [15], with recent improvements due to inclusion of the Kaminski and Ribe [16] method of predicting LPO development and treatment of lateral variations in viscosity [17], [18], though we do not fully explore the latter at this point. To generate synthetic splitting that is as close as possible to the published seismological results, we improve on Hall et al.'s [19] and Fouch et al.'s [48] studies and compute full waveforms. We show that not only fast azimuths but also the amplitudes of delay times can be explained by mantle-based LPO anisotropy, allowing inferences on the accumulated finite strain. A large degree of waveform complexity is predicted, which casts doubt on simple interpretations of anisotropy as being due to one or two layers with hexagonal symmetry in the horizontal plane. Models that include density anomalies due to geodynamic subduction models [20], [21] lead to reduced misfits compared to those based on plate-motion driven flow only, or based on density inferred from global shear wave tomography. We find that the inferred density structure of the mantle appears to have the strongest effect on anisotropy predictions and flow, compared to other uncertainties such as lateral viscosity variations. It may thus be possible to use anisotropy for tests of models of the tectonophysical history of the region [cf. [22]].

Section snippets

Data

Fig. 1 shows available shear wave splitting data for the study region, and the subset selected for our study from [6], [9]. Moving east from the Pacific to the North American plate, fast propagation directions trend roughly NW–SE and turn in the plate boundary region to a more W–E orientation. Data from different studies agree overall on regional scales, but there are large variations in splitting at individual stations depending on event back-azimuth. Such variations require a departure from

Modeling anisotropy

We are mainly considering sub-lithospheric anisotropy, but recognize the potentially complicating nature of the crust and lithospheric mantle in the sense of carrying frozen-in anisotropy from past tectonics. There may also be brittle damage anisotropy with a more complicated relationship to large-scale viscous deformation than mantle LPO. As we will show, the match of mantle-based LPO predictions to observations is good, implying either coherence of mantle and lithospheric deformation, or a

Mantle flow

Global mantle flow in a spherical shell is estimated based on prescribing plate motions on the surface while the core–mantle boundary is mechanically free-slip [15]. The buoyancy driven component of flow, and lateral viscosity variations for some models, are inferred from seismic velocity anomalies as imaged and inferred from a range of tomography and subduction models, analyzed in detail in [34]. We have verified that the buoyancy forces in such models would lead to realistic plate velocities

LPO fabrics

There is field and laboratory evidence on how the crystallographic axes of olivine, an inherently anisotropic mineral, align with shear [e.g. [44], [45]], and mineral physics theories allow quantitative estimates of LPO development [46], [16]. Anisotropy has been modeled on regional and global scales [e.g. [47], [19], [48], [13], [49], [15], [50]], and such linked geodynamic and seismic models successfully reproduce part of the observed structure. Recent work has focused on evaluating some of

SKS splitting

It is well known that fast azimuths of shear wave splitting measurements depend on back-azimuth for single layers whose anisotropy is not purely hexagonal with axis oriented in the horizontal, and splitting measurements suffer from non-linear superposition effects if anisotropy varies spatially [63], [64], [65]. However, such important geometries are rarely taken into account for geodynamic interpretation. Hall et al. [19] and Fouch et al. [48] perhaps achieved the greatest seismological

Results

Fig. 2 compares modeled velocities at different layers with predicted fast seismic azimuth, as visualized by hexagonal axes, at depth. (We only show TI axes for reference, the full tensor is used for the seismological modeling below.) For this particular region and flow model at ξc = 1, the depth averaged mean anisotropy is 7.6% for tensor norms. Out of this, the hexagonal symmetry component makes up ∼ 79% [cf. [62], [18]], the largest remainder, ∼ 17% of the total, is orthorhombic. The strength of

Discussion

Fig. 5 shows model performance in terms of 〈Δα〉 and 〈Δδ〉 for all experiments performed (see Table 1), separated by saturation target strains ξc and assumptions on the depth extent of LPO anisotropy. For a “good” model, we will consider low Δα and Δδ as criteria, but discuss those separately. A wide range of models with different assumptions on density or viscosity structures are able to fit the data quite well. We show results where all the computed LPO upper mantle anisotropy is used above

Conclusions

Models such as ours indicate that regional seismic anisotropy in the western US is consistent with simplified mantle flow models, substantiating earlier conclusions based on global datasets. Splitting can be explained both in terms of amplitudes and fast azimuths using kinematically based methods for computing the development of LPO during mantle flow [16], where preferred geodynamic models exhibit a return current at depth driven by the Farallon anomaly. This is similar to what was suggested

Acknowledgments

This work was inspired by extensive discussions with P. Silver, for which we are grateful. We also thank R.D. van der Hilst, M. Savage, and an anonymous reviewer for their constructive and helpful comments, E. Kaminski for providing D-REX, H. Schmeling for comments, A. McNamara, J. van Hunen, and E. Tan for the help with Citcom, S. Zhong, L. Moresi, M. Gurnis and others involved in geoframework.org for sharing this code, W. Menke for providing his splitting routines, S. Chevrot for the help

References (80)

  • P.G. Silver

    Seismic anisotropy beneath the continents: probing the depths of geology

    Annu. Rev. Earth Planet. Sci.

    (1996)
  • M.K. Savage

    Seismic anisotropy and mantle deformation: what have we learned from shear wave splitting?

    Rev. Geophys.

    (1999)
  • W.E. Holt

    Correlated crust and mantle strain fields in Tibet

    Geology

    (2000)
  • M.K. Savage et al.

    Strain modelling, seismic anisotropy and coupling at strike-slip boundaries: applications in New Zealand and the San Andreas fault

  • S. van der Lee et al.

    Upper mantle S velocity structure of North America

    J. Geophys. Res.

    (1997)
  • S. Özalaybey et al.

    Shear-wave splitting beneath western United States in relation to plate tectonics

    J. Geophys. Res.

    (1995)
  • G.P. Smith et al.

    A global study of Pn anisotropy beneath continents

    J. Geophys. Res.

    (1999)
  • V. Schulte-Pelkum et al.

    Upper mantle anistropy from long-period P polarization

    J. Geophys. Res.

    (2001)
  • R. Hartog et al.

    Depth-dependent mantle anisotropy below the San Andreas fault system: apparent splitting parameters and waveforms

    J. Geophys. Res.

    (2001)
  • P.M. Davis

    Azimuthal variation in seismic anisotropy of the southern California uppermost mantle

    J. Geophys. Res.

    (2003)
  • M. Fouch, Upper mantle anisotropy database. available online, (2004), accessed in 12/2004,...
  • J. Polet et al.

    Anisotropy beneath California: shear wave splitting measurements using a dense broadband array

    Geophys. J. Int.

    (2002)
  • P.G. Silver et al.

    The mantle flow field beneath Western North America

    Science

    (2002)
  • T.W. Becker, Lithosphere.Mantle Interactions. PhD thesis, Harvard University, Cambridge MA, (2002)....
  • T.W. Becker et al.

    Comparison of azimuthal seismic anisotropy from surface waves and finite-strain from global mantle-circulation models

    Geophys. J. Int.

    (2003)
  • T.W. Becker et al.

    Seismic anisotropy in the western US as a testbed for advancing combined models of upper mantle geodynamics and texturing (abstract)

    EOS Trans. AGU

    (2004)
  • T.W. Becker, S. Chevrot, V. Schulte-Pelkum, D.K. Blackman, Statistical properties of seismic anisotropy predicted by...
  • C.E. Hall et al.

    The influence of plate motions on three-dimensional back arc mantle flow and shear wave splitting

    J. Geophys. Res.

    (2000)
  • C. Lithgow-Bertelloni et al.

    The dynamics of Cenozoic and Mesozoic plate motions

    Rev. Geophys.

    (1998)
  • H.-P. Bunge et al.

    Mesozoic plate-motion history below the northeast Pacific Ocean from seismic images of the subducted Farallon slab

    Nature

    (2000)
  • R. Hartog et al.

    Subduction-induced strain in the upper mantle east of the Mendocino triple junction, California

    J. Geophys. Res.

    (2000)
  • M.K. Savage

    Seismic anisotropy and mantle deformation in the western United States and southwestern Canada

    Int. Geol. Rev.

    (2002)
  • S. Chevrot

    Multichannel analysis of shear wave splitting

    J. Geophys. Res.

    (2000)
  • A. Nicolas et al.

    Formation of anisotropy in upper mantle peridotites; a review

  • D. Mainprice et al.

    The seismic anisotropy of the Earth's mantle: from single crystal to polycrystal

  • S.-I. Karato

    On the Lehmann discontinuity

    Geophys. Res. Lett.

    (1992)
  • J. Wookey et al.

    Mid-mantle deformation inferred from seismic anisotropy

    Nature

    (2002)
  • E.J. Garnero et al.

    Variable azimuthal anisotropy in Earth's lowermost mantle

    Science

    (2004)
  • G. Hirth et al.

    Rheology of the upper mantle and the mantle wedge: a view from the experimentalists

  • T.W. Becker et al.

    A comparison of tomographic and geodynamic mantle models

    Geochem. Geophys. Geosyst.

    (2002)
  • Cited by (78)

    • Uppermantle shear-wave splitting measurements in Mainland China: A review

      2021, Earth-Science Reviews
      Citation Excerpt :

      Since Hess (1964) first observed seismic anisotropy and linked its origin to mantle flow, mantle flow model has played an important role in probing asthenospheric anisotropy and kept improving with the rapid increase of SWS measurements all over the world. Three kinds of models, simple asthenospheric flow (SAF) model (e.g., Vinnik et al., 1992; Leven et al., 1981), steady-state mantle flow (SMF) model (e.g, Forte et al., 2010; Behn et al., 2004; Conrad and Behn, 2010; Conrad et al., 2007; Becker et al., 2006a, b; Ribe, 1992; Zhu, 2014, 2018a, 2018b; Gaboret et al., 2003; Long and Becker, 2010) and time-dependent mantle flow model (e.g., Hu et al., 2017; Becker et al., 2014; Li et al., 2014b; Faccenda and Capitanio, 2013), have been proposed. SAF model indicates that the flow is driven by a moving plate overlying the stationary asthenosphere, which leads to the directions of shear, and hence φ in alignment with APM directions (Silver, 1996; Vinnik et al., 1992; Leven et al., 1981).

    • Complex upper mantle anisotropy in the Pacific Northwest: Evidence from SKS splitting

      2020, Earth and Planetary Science Letters
      Citation Excerpt :

      In contrast to anisotropy in the relatively high-velocity mantle, asthenospheric anisotropy is thought to indicate young mantle flow. Geodynamic modeling studies (e.g., Becker et al., 2006; Liu and Stegman, 2011; Wang and Becker, 2019; Zhou et al., 2018) have had success matching splitting-derived anisotropy in the western U.S. with coupled models of mantle flow and fabric development. In general, geodynamic models agree that ENE-WSW oriented anisotropy in the PNW is due to ENE-directed mantle flow relative to the North American plate.

    • Upper mantle seismic anisotropy as a constraint for mantle flow and continental dynamics of the North American plate

      2019, Earth and Planetary Science Letters
      Citation Excerpt :

      LVV2 is taken from the reference craton model of Miller and Becker (2012) where keel geometry is simpler, and keel depth constant at 300 km (Fig. 2d). Based on the mantle circulation models, we then use particle tracking and the D-Rex mineral physics approximation (Kaminski et al., 2004) to compute LPO as the tracers are advected until a logarithmic saturation strain of 0.75 is reached (Becker et al., 2006b; Miller and Becker, 2012). We assume that mantle circulation is stationary over the few Myr that it takes to achieve this strain (cf. Becker et al., 2003, 2006a).

    • Subduction-controlled mantle flow and seismic anisotropy in South America

      2017, Earth and Planetary Science Letters
      Citation Excerpt :

      By comparing the prediction with the observed anisotropy, we provide an explanation for the origin of the seismic anisotropy at various places across the study region. Various proxies have been used to represent the orientation of seismic anisotropy, including mantle velocity (e.g. Zandt and Humphreys, 2008; Assumpção et al., 2011), infinite strain axis (ISA) (e.g. Conrad et al., 2007; Conrad and Behn, 2010), the longest axis of finite strain ellipsoid (FSE) (e.g. Becker et al., 2003), and LPO predicted by fabric development models (Becker et al., 2006a, 2006b; Faccenda et al., 2008; Faccenda and Capitanio, 2013; Becker et al., 2014). However, these proxies use different assumptions and may be inconsistent with each other.

    View all citing articles on Scopus
    View full text