You are on page 1of 22

Geophysical Journal International

Geophys. J. Int. (2017) 210, 1410–1431 doi: 10.1093/gji/ggx247


Advance Access publication 2017 June 8
GJI Seismology

P-wave anisotropy, mantle wedge flow and olivine fabrics beneath


Japan

Xin Liu1,2 and Dapeng Zhao1


1 Departmentof Geophysics, Tohoku University, Sendai 980–8578, Japan. E-mail: zhao@tohoku.ac.jp
2 Key
Lab of Submarine Geosciences and Prospecting Techniques, Ministry of Education, and College of Marine Geosciences, Ocean University of China,
Qingdao 266100, China. E-mail: liuxin@ouc.edu.cn

Downloaded from https://academic.oup.com/gji/article/210/3/1410/3865122 by guest on 09 September 2021


Accepted 2017 June 6. Received 2017 June 5; in original form 2016 December 31

SUMMARY
We present a new 3-D anisotropic P-wave velocity (Vp) model for the crust and upper
mantle of the Japan subduction zone obtained by inverting a large number of high-quality
P-wave traveltime data of local earthquakes and teleseismic events. By assuming orthorhom-
bic anisotropy with a vertical symmetry axis existing in the modeling space, isotropic Vp
tomography and 3-D Vp azimuthal and radial anisotropies are determined simultaneously.
According to a simple flow field and the obtained Vp anisotropic tomography, we estimate the
distribution of olivine fabrics in the mantle wedge. Our results show that the forearc mantle
wedge above the subducting Pacific slab beneath NE Japan exhibits an azimuthal anisotropy
with trench-parallel fast velocity directions (FVDs) and Vhf > Vv > Vhs (here Vv is Vp in
the vertical direction, Vhf and Vhs are P-wave velocities in the fast and slow directions in the
horizontal plane), where B-type olivine fabric with vertical trench-parallel flow may dominate.
Such an anisotropic feature is not obvious in the forearc mantle wedge above the Philippine
Sea (PHS) slab under SW Japan, probably due to higher temperatures and more fluids there
associated with the young and warm PHS slab subduction. Trench-normal FVDs and Vhf >
Vv > Vhs are generally revealed in the mantle wedge beneath the arc and backarc in Japan,
where E-type olivine fabric with FVD-parallel horizontal flow may dominate. Beneath western
Honshu, however, the mantle wedge exhibits an anisotropy of Vv > Vhf > Vhs and so C-type
olivine fabric may dominate, suggesting that the water content is the highest there, because
both the PHS and Pacific slabs exist there and their dehydration reactions release abundant
fluids to the overlying mantle wedge.
Key words: Body waves; Seismic anisotropy; Seismic tomography; Subduction zone
processes.

The subduction processes, in particular, those occurring in


1 I N T RO D U C T I O N
the mantle wedge, can be well constrained by studying seismic
Strong interactions among four lithospheric plates are taking place anisotropy (e.g. Anderson 1989; Savage 1999; Maupin & Park
in and around the Japan Islands (Fig. 1). In NE Japan, the Pacific 2007, 2015; Long 2013; Skemer & Hansen 2016; Zhao et al.
plate is subducting beneath the Okhotsk plate at the Kuril-Japan 2016). Many previous studies have investigated seismic azimuthal
Trench. In SW Japan, the Philippine Sea (PHS) plate is descending anisotropy of the Japan subduction zone by making shear wave
beneath the Eurasian plate at the Nankai Trough and the Ryukyu splitting measurements (e.g. Ando et al. 1983; Okada et al. 1995;
Trench, whereas the Pacific plate is subducting beneath the PHS Hiramatsu et al. 1997; Nakajima et al. 2006; Tono et al. 2009;
and Eurasian plates. More than 100 active arc volcanoes exist on Wirth & Long 2010; Huang et al. 2011a,b; Savage et al. 2016),
the Japan Islands along the volcanic front and in the backarc areas anisotropic receiver-function analyses (e.g. Wirth & Long 2012;
(Fig. 1), which are produced by joint effects of fluids from dehy- Watanabe & Oda 2014) and P-wave velocity (Vp) azimuthal
dration reactions of the subducting slabs and convection (corner anisotropy tomography (e.g. Hirahara & Ishikawa 1984; Ishise &
flow) in the mantle wedge above the slabs (e.g. Tatsumi 1989; Zhao Oda 2005; Wang & Zhao 2008, 2013; Huang et al. 2011c, 2015;
et al. 1992; Iwamori & Zhao 2000; Tamura et al. 2002; Hasegawa Liu et al. 2013; Koulakov et al. 2015; Wei et al. 2015). A recent
et al. 2013). The arc magmas and fluids also affect the nucleation high-resolution 3-D model of azimuthal anisotropy obtained by a
of various earthquakes in subduction zones (see Zhao 2015 for a joint inversion of P- and S-wave traveltimes clearly reveals trench-
recent review). normal fast velocity directions (FVDs) in the mantle wedge beneath

1410 
C The Authors 2017. Published by Oxford University Press on behalf of The Royal Astronomical Society.
Anisotropic tomography beneath Japan 1411

Downloaded from https://academic.oup.com/gji/article/210/3/1410/3865122 by guest on 09 September 2021


Figure 1. Tectonic settings of the study region (modified from Liu et al. 2013). The red triangles denote the active and Quaternary volcanoes. The solid
sawtooth lines and the black dashed line denote the plate boundaries. The blue and red lines denote the depth contours of the upper boundaries of the subducting
Pacific and Philippine Sea slabs, respectively, which are constructed based on previous studies (e.g. Zhao et al. 1994, 2012; Hayes et al. 2012; Nakajima et al.
2009; Hasegawa et al. 2013).

the arc and backarc areas (Liu & Zhao 2016c). This feature is well understanding of subduction dynamics beneath Japan. To resolve
consistent with the previous results of body-wave (P-wave) tomog- this problem, in this work we develop a new method to simulta-
raphy(see Zhao et al. 2016 for a detailed review) and surface wave neously determine Vp azimuthal and radial anisotropies, as well
tomography (Liu & Zhao 2016a). The radial anisotropy beneath the as isotropic Vp tomography in the crust and upper mantle. We
Japan Islands has been also investigated by Vp radial anisotropy to- use a large number of high-quality P-wave traveltimes of local
mography (Wang & Zhao 2013; Niu et al. 2016; Wang et al. 2017a) earthquakes and teleseismic events recorded by the dense seismic
and surface wave tomography (Yoshizawa et al. 2010). However, networks on the Japan Islands. Our present results shed important
there is a discrepancy in the mantle wedge radial anisotropy be- new light on seismic anisotropy, structural heterogeneity, as well as
tween the body-wave and surface wave models (see Zhao et al. flow pattern and olivine fabrics in the mantle wedge of the Japan
2016 for details). subduction zone.
The previous studies of Vp radial anisotropy tomography
(Wang & Zhao 2013; Niu et al. 2016; Wang et al. 2017a) assumed a
vertical axis of hexagonal symmetry existing in the modeling space,
which is different from the Vp azimuthal anisotropy tomography 2 D ATA
which assumes a horizontal axis of hexagonal symmetry in the mod- We use 1852 seismic stations belonging to the Kiban seismic net-
eling space (e.g. Wang & Zhao 2008, 2013). The different assump- work installed on the Japan Islands (Fig. 2a, Okada et al. 2004),
tions make it difficult to simultaneously interpret the azimuthal and which include 796 stations of the High-sensitivity seismic network
radial anisotropic results in the same study region, which hinders our (Hi-net; Fig. 2b). Two sets of high-quality traveltime data are used to
1412 X. Liu and D. Zhao

Downloaded from https://academic.oup.com/gji/article/210/3/1410/3865122 by guest on 09 September 2021


Figure 2. Distribution of seismic stations (red squares) of (a) the Kiban seismic network and (b) the High-sensitivity seismic network (Hi-net). Epicentral
distribution of (c) the local earthquakes (colour dots) and (d) the teleseismic events (blue dots) used in this study. The colours in (c) denote the focal depths
whose scale is shown in the map.

conduct tomographic inversions. The first data set contains 501 571 study region was divided into many cubic blocks with a size of
P-wave arrival times of 2528 local shallow and deep earthquakes 50 × 50 × 10 km3 . In each block, a local earthquake was selected
which occurred in and around the Japan Islands (Fig. 2c). The by a specific scheme of selection according to the maximum number
second data set contains 505 452 P-wave relative traveltime resid- of recording stations and the minimum formal uncertainty of the
uals of 747 teleseismic events (blue circles in Fig. 2d), which are hypocentral parameters. As a result, 2528 local events were selected
measured precisely from digital seismograms using a waveform (Fig. 2c), which generated 501 571 P-wave arrival times whose pick-
cross-correlation method. The two sets of data are derived from Liu ing accuracy is estimated to be 0.05–0.10 s.This data set contains
& Zhao (2016b) who determined only isotropic Vp tomography of more P-wave arrivals than those released by the Japan Meteoro-
the Japan subduction zone. logical Agency (JMA) Unified Earthquake Catalog, because all the
Among the great number of local earthquakes (∼1598 000) clear P arrivals were picked up from the original vertical-component
recorded by the dense seismic networks on the Japan Islands seismograms by the seismological staffs of Tohoku University. The
(Fig. 2a) during 2002 June to 2015 April, a best set of events was uncertainty of hypocentral locations is <5 km for all the events and
selected for tomographic inversions (Liu & Zhao 2016b). The shal- <3 km for the events beneath the seismic network.
low (<100 km depth) offshore events (∼302 000) were removed Our second data set contains 505 452 P-wave relative travel-
from the data set because of their poor hypocentral locations. The time residuals from 747 teleseismic events (Fig. 2d), which were
Anisotropic tomography beneath Japan 1413

So far, the parameter C in eq. (1) has been treated in different ways
by the previous studies of anisotropic tomography. Some researchers
(e.g. Wang & Zhao 2008, 2013) assumed that C = −M, thus the
vertical velocity is equal to the fast velocity in the horizontal plane,
that is, a horizontal hexagonal symmetry axis is assumed to exist
in the modeling space. Liu & Zhao (2016c) focused on azimuthal
anisotropy and so assumed that C = 0, thus the slowness in the
vertical direction is S0 in their study.
In this study, C is taken as an unknown parameter to be determined
by traveltime inversion. Eq. (1) can be rewritten as (e.g. Hearn 1996;
Barclay et al. 1998; Eberhart-Phillips & Henderson 2004; Wang &
Zhao 2008, 2013; Liu & Zhao 2016c):
S P = S0 + (Acos2φ + Bsin2φ) · sin2 i + C · cos2 i, (2)
where φ is the ray-path azimuth, A = −Mcos(2ψ) and B =

Downloaded from https://academic.oup.com/gji/article/210/3/1410/3865122 by guest on 09 September 2021


−Msin(2ψ) are two azimuthal anisotropy parameters, and ψ is
the horizontal FVD (Fig. 3). Using eq. (2), P-wave traveltime Tj of
the jth ray segment with a length d can be expressed as

Figure 3. The coordinate system specifying a P-wave ray (the blue line) and Tj = d · S j
the orthorhombic anisotropy with three mutually orthogonal binary symme- = d[S0 j + (A j cos2φ j + B j sin2φ j ) · sin2 i j + C j · cos2 i j ], (3)
try axes (the red lines). Two binary symmetry axes are in the horizontal
symmetry plane a. The plane e is a vertical symmetry plane. The P-wave where Sj is the total P-wave slowness of the jth ray segment, S0j is the
ray is located in the vertical plane c. i is the ray incident angle, ψ represents isotropic P-wave slowness at the middle point of the ray segment,
the horizontal fast-velocity direction (FVD), φ is the ray path azimuth and i j and φ j are the incident angle and azimuth of the ray segment,
θ is the angle between the ray propagation vector in the plane a and the axis respectively, A j = −M j cos(2ψ j ) and B j = −M j sin(2ψ j ) are two
of the intersection of two mutually orthogonal symmetry planes a and e. azimuthal anisotropy parameters at the middle point, Mj and Cj
are parameters for anisotropy at the middle point, and ψ j is the
measured precisely by applying the multichannel cross-correlation horizontal FVD at the middle point which is defined as
technique (VanDecar & Crosson 1990) to high-quality seismo- ⎧ π
⎪   , Aj > 0
grams recorded at the Hi-net seismic stations (Fig. 2b) during 2004 ⎪
⎪ −1 B j
⎪ 1
⎨ 2 tan + 2
0, A j < 0 .
Aj
April–2014 December (for details, see Liu & Zhao 2016b). The
ψj = π (4)
P-wave relative traveltime residuals were measured from the ⎪ −
⎪ 4 , B > 0


j
, Aj = 0
vertical-component seismograms filtered in a frequency band of ⎩ π,B < 0
0.1–0.2 Hz. The teleseismic events used are all ≥ M 6.0, and their 4 j

epicentral distances are in a range of ∼30◦ –90◦ from central Japan By defining A P j =
Aj
, BP j =
Bj
and C P j =
Cj
, eq. (3) can be
S0 j S0 j S0 j
(Fig. 2d). Hypocentral parameters of the teleseismic events are ob- rewritten as
tained from the Bulletins of the International Seismological Center.
The azimuthal coverage of these events is very good in all direc- T j = d[1 + (A P j cos2φ j + B P j sin2φ j ) · sin2 i j
tions except for the Pacific Ocean, and most of the events occurred +C P j · cos2 i j ]/V0 j , (5)
in subduction zones (Fig. 2d).
where V0j is the isotropic Vp at the middle point of the jth ray
segment. Then, the amplitude α j of Vp azimuthal anisotropy at the
3 METHOD middle point can be expressed as
We modify the azimuthal anisotropy tomography method of Liu V h f j − V hs j Mj
αj = =

& Zhao (2016c) to invert the P-wave traveltime data for 3-D Vp 2V0 j S0 j − M 2j S0 j
azimuthal and radial anisotropies, as well as Vp isotropic tomog-
raphy beneath the Japan Islands. Following Liu & Zhao (2016c), A P j 2 + BP j 2
= , (6)
we assume that the modeling space is a weakly anisotropic medium 1 − (A P j 2 + B P2 j )
where orthorhombic anisotropy with a vertical symmetry axis ex-
ists (Fig. 3). For a general P-wave with an incident angle i, P-wave where Vhfj and Vhsj are P-wave velocities in the horizontal fast and
slowness Sp can be simplified as (e.g. Eberhart-Phillips & Hender- slow directions at the middle point, respectively. We define β j to
son 2004; Wang & Zhao 2008, 2013; Koulakov et al. 2009; Liu & express Vp radial anisotropy at the middle point, which is written
Zhao 2016c): as
V h j − V vj
S P = S0 + (Mcos2θ ) · sin2 i + C · cos2 i, (1) βj =
2V0 j
where S0 is the isotropic P-wave slowness (i.e. the isotropic com- C P j − (A P j cos2φ j + B P j sin2φ j )
ponent), M is a parameter for azimuthal anisotropy, θ is the angle = , (7)
2(1 + A P j cos2φ j + B P j sin2φ j )(1 + C P j )
between the P-wave propagation vector in a horizontal symmetry
plane a and the axis of the intersection of two mutually orthogonal where V h j = S0 j ·(1+A P j cos2φ
1
j +B P j sin2φ j )
and V v j = S0 j ·(1+C
1
Pj )
are
symmetry planes a and e, and C defines the amount of anisotropy P-wave velocities in the horizontal and vertical directions at
in the vertical direction (i.e. radial anisotropy, Fig. 3). the middle point, respectively. The partial derivatives of P-wave
1414 X. Liu and D. Zhao

Downloaded from https://academic.oup.com/gji/article/210/3/1410/3865122 by guest on 09 September 2021

Figure 4. Distributions of P-wave ray azimuths and incident angles at nine depths (10–400 km) beneath an area (38◦ ∼40◦ N, 139◦ ∼141◦ E) in NE Japan. The
P-wave ray segments at each depth are plotted in an equal area (Schmidt) stereonet at the lower hemisphere projection using the ray path azimuths and incident
angles. The rays with large incident angles (i.e. near-horizontal rays) are close to the edge of the stereonet, whereas the rays with smaller incident angles (i.e.
near-vertical rays) are close to the centre of the stereonet. Note that we equate the numbers of ray segments with azimuths of φ and φ + π in the stereonet
projections, because the effect of a P-wave ray segment with an azimuth of φ is the same as that of a ray segment with an azimuth of φ + π in the tomographic
inversion. The total number of ray segments at each depth is normalized. In each stereonet, the blue colour shows the areas with more ray hit counts, whereas
the yellow colour shows the areas with fewer rays. The blank areas in each stereonet have no rays. The scale for the frequency of rays in each stereonet is shown
at the bottom.
Anisotropic tomography beneath Japan 1415

Downloaded from https://academic.oup.com/gji/article/210/3/1410/3865122 by guest on 09 September 2021


Figure 5. Map views of isotropic Vp tomography and Vp azimuthal anisotropy obtained by a joint inversion for 3-D isotropic Vp and 3-D Vp azimuthal
and radial anisotropies (the VAR inversion). The layer depth is shown at the upper left corner of each map. The blue thick line shows the location of the
upper boundary of the subducting Pacific slab at each depth. The red and blue colours denote low and high isotropic velocities, respectively. The isotropic Vp
perturbation (in per cent) scale is shown in (a). The orientation and length of the short black bars represent the horizontal FVD and the azimuthal anisotropic
amplitude (α P ), respectively. The α P scale is shown in (d). The other labeling is the same as that in Fig. 1.

traveltime Tj with respect to the isotropic P-wave velocity (V0j ) and anisotropic parameters δ A P j , δ B P j and δC P j can be expressed as
the three anisotropic parameters (APj , BPj and CPj ) are
∂ Tj δV0 j
 
∂ Tj

r = V0 j + δ APj
∂ Tj ∂ V0 j V0 j ∂ APj
= −d[1 + (A P j cos2φ j + B P j sin2φ j ) · sin2 i j j

j

∂ V0 j ∂ Tj ∂ Tj
+ δ BP j + δC P j + E, (12)
+ C P j · cos2 i j ]/V0 j 2 , (8) ∂ BP j ∂C P j
j j

where E represents higher order terms of the perturbations, pick-


∂ Tj d ing errors of the P-wave traveltime data and uncertainties of the
= sin2 i j cos2φ j , (9)
∂ APj V0 j hypocentral parameters.
Two 3-D grids are set up in the modeling space. One grid is
for expressing the 3-D isotropic Vp whose perturbations from a
∂ Tj d starting 1-D velocity model at the grid nodes are taken as un-
= sin2 i j sin2φ j , (10)
∂ BP j V0 j known parameters. The starting 1-D velocity model (Fig. S1, Sup-
porting Information) is derived from the CRUST1.0 model (Laske
et al. 2013) and the IASP91 Earth model (Kennett & Engdahl
∂ Tj d
= cos2 i j . (11) 1991). In addition, the initial isotropic Vp of the subducting Pacific
∂C P j V0 j slab is assigned to be 2 per cent faster than the mantle isotropic
Vp at the same depth, according to the previous tomographic re-
For a well-located local earthquake (or a teleseismic event), the sults of the Japan subduction zone (e.g. Zhao et al. 1994, 2012;
observation equation relating observed local-earthquake P-wave Liu & Zhao 2016b,c). The other 3-D grid is for expressing the
traveltime residuals (or teleseismic P-wave relative traveltime resid- 3-D Vp anisotropic structure. Perturbations of the three anisotropic
δV
uals) r to the isotropic Vp perturbation V00jj and the perturbations of parameters (AP , BP and CP in eq. 12) at each grid node are also
1416 X. Liu and D. Zhao

Downloaded from https://academic.oup.com/gji/article/210/3/1410/3865122 by guest on 09 September 2021


f −V v
Figure 6. Map views of Vp radial anisotropy (β f = V h2V ) obtained by the VAR inversion, whose scale (in per cent) is shown in (a). V0 , Vhf and Vv are
0
isotropic Vp, the fastest Vp in the horizontal direction and Vp in the vertical direction, respectively. The blue colour shows Vhf > Vv, whereas the red colour
shows Vhf < Vv. The layer depth is shown at the upper left corner of each map. The other labeling is the same as that in Fig. 5.

taken as unknown parameters. The isotropic Vp perturbation at any each grid node is calculated from the obtained AP , BP and CP using
point in the modeling space is computed by linearly interpolating eq. (7).
its perturbations at the eight grid nodes surrounding that point. To obtain a robust result, we conducted many tomographic in-
This linear interpolation scheme is also adopted for computing per- versions to find the optimal values of the damping and smoothing
turbations of the three anisotropic parameters. An efficient 3-D parameters and the number of iterations. The optimal values of these
ray-tracing technique (Zhao et al. 1992) is used to compute the- parameters are shown in Table S1 in the Supporting Information.
oretical traveltimes and ray paths. The depths of crustal layers in The final root-mean-square (rms) traveltime residual obtained by the
the CRUST1.0 model (Laske et al. 2013) and station elevations joint inversion for Vp azimuthal and radial anisotropies and isotropic
are taken into account in the 3-D ray tracing. The LSQR algorithm Vp tomography is smaller than those obtained by the separate tomo-
(Paige & Saunders 1982) is used to solve the observation eq. (12). graphic inversions for isotropic Vp, Vp azimuthal anisotropy and Vp
Smoothing and damping regularizations are adopted to suppress radial anisotropy (Table S1, Supporting Information). To evaluate
the dramatic short-scale variations of V0 , AP , BP and CP during the whether the reductions of the rms residual are statistically signifi-
tomographic inversion. The final tomographic results are obtained cant or not, we conducted F-test (Draper & Smith 1966) following
after ten iterations. We relocate the local earthquakes using the the approach of Zhao et al. (1994, 2012). The test results indicate
obtained 3-D anisotropic Vp model after each iteration. Note that that the rms residual reductions are indeed statistically significant.
S-wave velocity heterogeneities and anisotropies are not considered
here, which may slightly affect the hypocentral locations of the lo-
cal events. Relative traveltime residuals of the teleseismic events 4 A N A LY S I S A N D R E S U LT S
are also re-determined after each iteration. To minimize the effect Fig. S2 in the Supporting Information shows ray density distribu-
of uncertainty of the initial isotropic Vp model, the final isotropic tions of the P-wave data at different depths. The azimuthal distri-
Vp perturbation at each grid node is calculated from the average bution of the ray paths in each 2◦ × 2◦ square represented by a rose
of the obtained isotropic Vp perturbations at each depth. After the diagram is also shown in Fig. S2 in the Supporting Information,
tomographic inversion, the FVD and amplitude of Vp azimuthal indicating that the rays crisscross very well in and around the Japan
anisotropy at each grid node are calculated from the obtained AP Islands. To obtain a robust model of azimuthal anisotropy tomog-
and BP using eqs (4) and (6), respectively. The radial anisotropy at raphy, good azimuthal coverage of rays is necessary, whereas radial
Anisotropic tomography beneath Japan 1417

Downloaded from https://academic.oup.com/gji/article/210/3/1410/3865122 by guest on 09 September 2021


v
Figure 7. Map views of Vp radial anisotropy (βs = V hs−V 2V0 ) obtained by the VAR inversion, whose scale (in per cent) is shown in (a). V0 , Vhs and Vv are
isotropic Vp, the slowest Vp in the horizontal direction and Vp in the vertical direction, respectively. The blue colour denotes Vhs > Vv, whereas the red colour
denotes Vhs < Vv. The layer depth is shown at the upper left corner of each map. The other labeling is the same as that in Fig. 5.

anisotropy tomography requires good ray coverage in the incident parameter (CP ) at the adjacent grid nodes in the second grid. Then,
angle (e.g. Wang & Zhao 2010; Huang et al. 2015; Zhao et al. we calculated synthetic traveltimes for the checkerboard model
2016). In this study for a joint inversion for both azimuthal and with the same numbers of seismic stations, events and ray paths
radial anisotropies, good ray coverages in both azimuth and inci- as those in the real data set, and then inverted the synthetic data
dent angle are necessary. Fig. 4 shows the azimuthal and incident- to see whether the input checkerboard model could be recovered
angle coverages of P-wave rays at different depths beneath an area or not. The obtained CRT results (Figs S14–S19, Supporting In-
(38◦ ∼40◦ N, 139◦ ∼141◦ E) in NE Japan, indicating that the good az- formation) indicate that our anisotropic tomography has a spatial
imuthal coverage of rays extends down to ∼400 km depth, whereas resolution of ∼30–100 km in the lateral direction and ∼10–50 km
the good incident-angle coverage of rays extends down to a depth in depth beneath most of the study region in and around the Japan
of <200 km. The azimuthal and incident-angle coverages of rays at Islands.
different depths under the entire study region (Figs S3–S13, Sup- For comparing the tomographic models obtained with the differ-
porting Information) show that the ray paths crisscross well down ent inversion methods, hereafter the joint inversion for Vp azimuthal
to ∼200 km depth for most of the study region in and around the and radial anisotropies and isotropic Vp tomography is called the
Japan Islands. VAR inversion for short. Using the same traveltime data set, the spa-
To further assess the adequacy of ray coverage and to evaluate tial resolution of the VAR tomography (Figs S14–S19, Supporting
the spatial resolution of the tomographic images, we conducted Information) is slightly lower than those of Vp azimuthal anisotropy
extensive checkerboard resolution tests (CRTs) and restoring res- tomography obtained using the method of Liu & Zhao (2016c)
olution tests following the previous studies (e.g. Zhao et al. 1992; and Vp radial anisotropy tomography obtained with the method of
Wang & Zhao 2008, 2013; Liu & Zhao 2016c). To perform a CRT Wang & Zhao (2013) (Figs S20–S31, Supporting Information),
for the anisotropic tomography, we first assigned positive and neg- partly because there are more unknown parameters in the VAR
ative isotropic Vp perturbations of 3 per cent to the 3-D grid nodes inversion.
in the first grid. We also assigned anomalies of ±0.02 to the two Figs 5–10 show the optimal tomographic model obtained by our
azimuthal anisotropy parameters (AP and BP ) at the adjacent grid VAR inversion. In this model, grid meshes are located at depths
nodes in the second grid, which represent horizontal FVDs of 22.5◦ of 10, 25, 40, 70 km and 100–700 km with a vertical grid interval
and 112.5◦ with azimuthal anisotropic amplitudes of ∼3 per cent. In of 50 km, and the lateral grid interval is 0.33◦ (∼33 km) in the
addition, anomalies of ±0.03 are assigned to the radial anisotropy first grid for expressing the 3-D isotropic Vp, whereas it is 1.0◦
1418 X. Liu and D. Zhao

Downloaded from https://academic.oup.com/gji/article/210/3/1410/3865122 by guest on 09 September 2021

Figure 8. Vertical cross-sections of the isotropic Vp tomography along the profiles (blue lines) shown on the inset map. This 3-D isotropic Vp model is obtained
by conducting the VAR inversion. The red and blue colours denote low and high Vp perturbations, respectively, whose scale (in per cent) is shown at the upper
right corner. The surface topography and the land area (the black horizontal bar) along each profile are shown atop each cross-section. The red triangles denote
active arc volcanoes within a 20 km width of each profile. The background seismicity and low-frequency microearthquakes which occurred within a 20 km
width of each profile are shown in white and red circles, respectively. The curved black line shows the Moho discontinuity.
Anisotropic tomography beneath Japan 1419

B D

Downloaded from https://academic.oup.com/gji/article/210/3/1410/3865122 by guest on 09 September 2021


C E

f −V v
Figure 9. Vertical cross-sections of Vp radial anisotropy (β f = V h2V ) along the profiles (blue lines) shown on the inset map. This 3-D anisotropic Vp model
0
is obtained by conducting the VAR inversion. The blue colour shows Vhf > Vv, whereas the red colour shows Vhf < Vv. The other labeling is the same as that
in Fig. 8.
1420 X. Liu and D. Zhao

B D

Downloaded from https://academic.oup.com/gji/article/210/3/1410/3865122 by guest on 09 September 2021


C E

v
Figure 10. Vertical cross-sections of Vp radial anisotropy (βs = V hs−V
2V0 ) along the profiles (blue lines) shown on the inset map. This 3-D anisotropic Vp
model is obtained by conducting the VAR inversion. The blue colour shows Vhs > Vv, whereas the red colour shows Vhs < Vv. The other labeling is the same
as that in Fig. 8.
Anisotropic tomography beneath Japan 1421

Downloaded from https://academic.oup.com/gji/article/210/3/1410/3865122 by guest on 09 September 2021

Figure 11. Map views of isotropic Vp images obtained by conducting (a)–(c) isotropic Vp tomography alone, (d)–(f) Vp azimuthal anisotropy tomography,
(g)–(i) Vp radial anisotropy tomography and (j)–(l) the VAR inversion (see the text for details). In these tomographic models, the lateral grid interval is 0.33◦
for the 3-D isotropic Vp, whereas it is 1.0◦ for the Vp anisotropy. The layer depth is shown at the upper left corner of each map. The other labeling is the same
as that in Fig. 5.

(∼100 km) in the second grid for expressing the Vp anisotropy. The Using our P-wave data set, we also determined isotropic Vp
tomographic results show that significant lateral heterogeneities and tomography applying the method of Zhao et al. (1994, 2012)
anisotropies exist in the crust and upper mantle beneath the study (Figs 11 and Figs S32 and S33, Supporting Information), Vp az-
region (Figs 5–10). imuthal anisotropy tomography applying the method of Liu & Zhao
1422 X. Liu and D. Zhao

Downloaded from https://academic.oup.com/gji/article/210/3/1410/3865122 by guest on 09 September 2021


Figure 12. Map views of isotropic Vp image and Vp azimuthal anisotropy obtained by conducting (a)–(c) Vp azimuthal anisotropy tomography and (d)–(f) the
VAR inversion (see the text for details). In the two tomographic models, the lateral grid interval is 0.33◦ for the 3-D isotropic Vp, whereas it is 1.0◦ for the Vp
anisotropy. The layer depth is shown at the upper left corner of each map. The other labeling is the same as that in Fig. 5.

(2016c) (Figs 11 and 12, and Figs S34 and S35, Supporting Informa- in the Supporting Information, except for the subducting Pacific
tion) and Vp radial anisotropy tomography applying the method of slab. We also conducted a VAR inversion with a lateral grid interval
Wang & Zhao (2013) (Figs 11 and 13, Figs S36–S41, Supporting of 0.33◦ for expressing the isotropic Vp tomography and 0.5◦ for
Information). The isotropic Vp images obtained with the different the Vp anisotropy. The obtained results (Figs S54–S57, Support-
methods are similar to each other for most of the study region ing Information) are also similar to the optimal tomographic model
(Fig. 11). The Vp azimuthal anisotropy obtained by the VAR inver- (Figs S42–S45, Supporting Information).
sion is very similar to that obtained with the method of Liu & Zhao
(2016c) (Figs 12 and 14).There are some differences in the images of
Vp radial anisotropy obtained by the VAR inversion and the method
5 DISCUSSION
of Wang & Zhao (2013) (Fig. 13), which may be mainly caused
by the different assumptions of anisotropy adopted in the two ap- As the best-studied subduction zone in the world, the 3-D seis-
proaches, that is, hexagonal symmetry is assumed in Wang & Zhao mic structure and anisotropy beneath Japan have been investigated
(2013), whereas orthorhombic anisotropy is assumed in our VAR by a great number of researchers (e.g. Hirahara & Ishikawa 1984;
inversion. Zhao et al. 1992, 2012; Ishise & Oda 2005; Nakajima et al. 2006;
To confirm the obtained tomographic results (Figs 5–10 and Wang & Zhao 2005, 2013; Xia et al. 2007, 2008; Yoshizawa et al.
Figs S42–S45, Supporting Information), we conducted a VAR in- 2010; Huang et al. 2011c, 2013, 2015; Cheng et al. 2011; Tong et al.
version with a different lateral grid interval. Figs S46–S49 in the 2012; Hasegawa et al. 2013; Tian & Liu 2013; Liu et al. 2013, 2014;
Supporting Information show the obtained results with a lateral Asamori & Zhao 2015; Koulakov et al. 2015; Wei et al. 2015; Kita
grid interval of 0.33◦ for both the isotropic Vp tomography and the & Matsubara 2016; Niu et al. 2016; Liu & Zhao 2015, 2016c; Wang
Vp anisotropy, which are similar to the optimal tomographic re- et al. 2017a,b,c; Zhao et al. 2017). These studies revealed clearly
sults (Figs S42–S45, Supporting Information). In addition, we also the subducting Pacific and PHS slabs which exhibit high-velocity
conducted a VAR inversion using a starting Vp model that is a ho- (high-V), high-Q (low attenuation), low Poisson’s ratio and trench-
mogeneous 1-D isotropic model and does not include the high-Vp parallel FVDs which may reflect fossil fabrics in the slabs (e.g.
subducting Pacific slab. The obtained results (Figs S50–S53, Sup- Zhao 2015; Zhao et al. 2016). Note that not all the subducting slabs
porting Information) are similar to those shown in Figs S46–S49 exhibit trench-parallel FVDs. For example, trench-normal FVDs
Anisotropic tomography beneath Japan 1423

Downloaded from https://academic.oup.com/gji/article/210/3/1410/3865122 by guest on 09 September 2021


Figure 13. Map views of Vp radial anisotropy (β0 = V0 V−V v ) obtained by conducting (a)–(c) Vp radial anisotropy tomography and (d)–(f) the VAR inversion
0
(see the text for details). V0 is isotropic Vp; Vv is Vp in the vertical direction. In the two tomographic models, the lateral grid interval is 0.33◦ for the 3-D

isotropic Vp, whereas it is 1.0 for the Vp anisotropy. The layer depth is shown at the upper left corner of each map. The other labeling is the same as that in
Fig. 5.

have been revealed in the subducting Cocos slab beneath Costa [100]{0kl}, [100](001) slip systems for C-, D-, E-type fabrics, re-
Rica (Rabbel et al. 2011) and in the subducting Pacific slab beneath spectively (Karato et al. 2008). According to the simple relation
Alaska (Tian & Zhao 2012). In addition, significant low-velocity between the dominant slip system and flow geometry, the slip direc-
(low-V), low-Q (high attenuation), high Poisson’s ratio and trench- tion is parallel to the macroscopic flow direction and the slip plane is
normal FVDs are revealed in the mantle wedge beneath the volcanic parallel to the macroscopic flow plane (Karato et al. 2008). In such a
front and backarc areas in Japan, which may be related to slab de- case, for horizontal mantle flow with a horizontal flow direction and
hydration, corner flow, arc magmatism and earthquake generation a horizontal flow plane, the flow-induced seismic anisotropy should
(e.g. Hasegawa et al. 2013; Zhao 2015). These features are well be characteristic of Vhf > Vhs > Vv and b-parallel horizontal FVD
recovered by our present results (Figs 5 and 8), which have been for A-type fabric, Vhf > Vhs > Vv and b-normal horizontal FVD
confirmed by our extensive resolution tests (Figs S14–S57, Sup- for B-type fabric, Vv > Vhf > Vhs and b-parallel horizontal FVD
porting Information). Most of these features have been discussed for C-type fabric, and Vhf > Vv > Vhs and b-parallel horizontal
thoroughly in the previous studies. In the following, we focus our FVD for E-type fabric (Fig. 15), where Vv is Vp in the vertical di-
discussion on the processes occurring in the mantle wedge, and we rection, whereas Vhf and Vhs are P-wave velocities in the fast and
attempt to relate our new anisotropic results to the flow pattern and slow directions in a horizontal plane. In addition, for vertical mantle
olivine fabrics in the mantle wedge of the Japan subduction zone. flow with a vertical flow direction and a vertical flow plane, its Vp
Seismic anisotropy in the mantle wedge is not only related to anisotropy should be Vv > Vhf > Vhs and n-normal horizontal FVD
mantle flow driven by the plate subduction, but also caused by dif- for A-type fabric, Vhf > Vv > Vhs and n-normal horizontal FVD for
ferent types of olivine fabrics due to the variation in water content, B-type fabric, Vhf > Vv > Vhs and n-parallel horizontal FVD for
stress and temperature in the mantle wedge (e.g. Jung & Karato C-type fabric, and Vv > Vhf > Vhs and n-parallel horizontal FVD
2001; Kneller et al. 2005, 2008; Jung et al. 2006; Karato et al. for E-type fabric (Fig. 15). Such a simple feature could not be well
2008). Each olivine fabric has its own dominant slip system charac- identified for D-type fabric due to its complex dominant slip system.
terized by slip direction b and slip plane normal (n), such as A-type Hence, in this study, we ignore D-type fabric which only develops in
fabric corresponding to the [100](010) slip system, B-type fabric some localized areas where high-stress deformation occurs (Karato
corresponding to the [001](010) slip system and the [001](100), et al. 2008).
1424 X. Liu and D. Zhao

Downloaded from https://academic.oup.com/gji/article/210/3/1410/3865122 by guest on 09 September 2021


Figure 14. Map views of isotropic Vp image and Vp azimuthal anisotropy along three slices shown in the sketch beside (c). The 3-D anisotropic Vp model
shown in (a)–(c) was obtained by a joint inversion of P- and S-wave traveltime data of both local and teleseismic events, with a lateral grid interval of 0.33◦ for
both the isotropic Vp tomography and Vp azimuthal anisotropy (Liu & Zhao 2016c). The 3-D anisotropic Vp model shown in (d)–(f) is obtained by conducting
the VAR inversion (see the text for details). (a) and (d) show the images in a plane located at the middle between the Moho discontinuity and the upper boundary
of the subducting Pacific slab (UBPS). (b) and (e) show the images in a plane located 50 km below the UBPS, whereas (c) and (f) show the images in a plane
located 200 km below the UBPS. The white lines to the east of the Japan Trench denote isochrones of the Pacific Ocean floor (Müller et al. 2008) which are
parallel to the palaeomagnetic lineaments in the Pacific plate. The white arrows in each map show the moving direction of the Pacific plate relative to NE Japan
(Kreemer et al. 2003). The purple dashed lines in each map denote the depth contours of the UBPS. The other labeling is the same as that in Fig. 5.

Shear wave splitting measurements are often used to constrain the arc and backarc areas, the mantle wedge generally exhibits an
the flow pattern and olivine fabrics in the mantle wedge (e.g. anisotropy of Vhf > Vhs > Vv or Vhf > Vv > Vhs, whereas an
Long & Wirth 2013; Nagaya et al. 2016; Lynner et al. 2017), but anisotropy of Vhf > Vv > Vhs dominates in the forearc mantle
their depth resolution is poor. Azimuthal anisotropy tomography can wedge (Figs 16b and c). In addition, a distinct feature Vv > Vhf
only constrain the relation between Vhf and Vhs in the 3-D mod- > Vhs exists near the critical boundary in the mantle wedge under
eling space, whereas radial anisotropy tomography can only reveal NE Japan (Figs 16b and c). These anisotropic features (Figs 16a–c)
the relation between Vh and Vv (e.g. Yoshizawa et al. 2010; Wang have been confirmed by our extensive resolution tests (Figs S58–
& Zhao 2013; Zhao et al. 2016; Wang et al. 2017a). In contrast, S72, Supporting Information).
our present VAR tomography is able to clarify a complete relation We assume that simple corner flow exists in the mantle wedge
among Vhf, Vhs and Vv. Hence, we can better constrain the flow above the Pacific slab (see Fig. 16e). In other words, we assume that
pattern and olivine fabrics in the mantle wedge using our new VAR horizontal mantle flow with a horizontal flow plane and horizontal
tomography and the simple relations as shown in Fig. 15. FVD-parallel flow directions exists in the mantle wedge beneath
Fig. 16 shows our VAR tomography along a slice in the the arc and backarc areas (i.e. above and below the green line in
middle of the mantle wedge beneath NE Japan where the Fig. 16e). In contrast, vertical flow in a vertical plane parallel to the
old Pacific plate is subducting beneath the Okhotsk plate. trench is assumed to exist in the mantle wedge beneath the forearc
Fig. 16(a) shows the isotropic Vp image and Vp azimuthal area (i.e. above and below the blue line in Fig. 16e). Under such
anisotropy. Beneath the volcanic front and backarc areas, the assumptions and adopting the simple relations shown in Fig. 15,
mantle wedge mainly exhibits low-V and trench-normal FVDs, we estimate the distribution of olivine fabrics in the mantle wedge
whereas high-V anomalies and trench-parallel FVDs exist in the beneath NE Japan (Fig. 16d), which shows that E-type fabric dom-
forearc mantle wedge (Fig. 16a). The boundary of the two different inates in the arc and backarc mantle wedge, except for some local
domains is generally located above the 65 km depth contour of the areas where A-type fabric is dominant. By contrast, B-type fabric
upper interface of the Pacific slab (Fig. 16). Hereafter, this boundary dominates in the forearc mantle wedge under NE Japan (Fig. 16d).
in the mantle wedge is called the critical boundary for short. In the same way, we also estimate the distribution of olivine fab-
Figs 16(b) and (c) show images of Vp radial anisotropy, which are rics in the mantle wedge beneath Hokkaido using our Vp anisotropic
f −V v v
expressed as β f = V h2V 0
and βs = V hs−V
2V0
, respectively. Beneath results (Figs 17a–c) which are confirmed by our detailed resolution
Anisotropic tomography beneath Japan 1425

Downloaded from https://academic.oup.com/gji/article/210/3/1410/3865122 by guest on 09 September 2021

Figure 15. Schematic diagrams illustrating different olivine fabrics, the associated dominant slip systems and the related seismic anisotropy corresponding to
two types of flow geometries. HFVD, the horizontal fast velocity direction. See the text for details.

tests (Figs S73–S87, Supporting Information). The obtained result Fig. 18 shows our VAR tomography along a slice in the mid-
(Fig. 17d) shows that E-type fabric also dominates in the mantle dle of the mantle wedge under SW Japan, where the young PHS
wedge beneath the arc and backarc, except for a few small areas, plate is subducting beneath the Eurasian plate at the Nankai Trough
where A- and C-type fabrics exist. The same as Tohoku, B-type and the Ryukyu Trench. The mantle wedge mainly exhibits low-
fabric dominates in the forearc mantle wedge beneath Hokkaido V and trench-normal horizontal FVDs (Fig. 18a). Different from
(Fig. 17d). NE Japan, no obvious horizontal trench-parallel FVDs exist in the
1426 X. Liu and D. Zhao

Downloaded from https://academic.oup.com/gji/article/210/3/1410/3865122 by guest on 09 September 2021


f −V v v
Figure 16. Map views of (a) isotropic Vp tomography and Vp azimuthal anisotropy, (b) and (c) Vp radial anisotropy (β f = V h2V and βs = V hs−V
2V0 ) and
0
(d) the distribution of olivine fabrics, in a plane in the central mantle wedge beneath NE Japan as shown in (e). The plane (shown in green and blue lines in
(e)) is located at the middle between the Moho discontinuity and the upper boundary of the subducting Pacific slab (UBPS). The red dashed lines in (e) denote
a simple corner-flow pattern assumed in this study. This 3-D anisotropic Vp model is obtained by conducting the VAR inversion (see the text for details). The
scales for panels (a)–(d) are shown at the lower right corner. The other labeling is the same as that in Fig. 14.

forearc mantle wedge, except for some local areas (Fig. 18a). Be- pattern should be complex, and even for simple 2-D corner flow,
neath Kyushu, the mantle wedge generally exhibits an anisotropy of the distribution of finite strain ellipses may be substantially more
Vhf > Vhs > Vv or Vhf > Vv > Vhs, whereas an anisotropy of Vv > complicated than the simple configuration used in this study (e.g.
Vhf > Vhs dominates in the mantle wedge beneath western Honshu Hall et al. 2000; Long et al. 2007; Honda 2011; Wirth & Korenaga
(Figs 18b and c). These tomographic results (Figs 18a–c) are con- 2012). Further detailed anisotropic studies may address a complex
firmed by our extensive resolution tests (Figs S88–S102, Supporting flow pattern, such as the inclined flow in the mantle wedge which
Information). Similar to that in NE Japan, we also assume that sim- may result in an olivine fabric with an inclined slip direction and
ple corner flow exists in the mantle wedge above the subducting slip plane. To constrain the olivine fabric, we prefer using the simple
PHS slab as shown in Fig. 18(e). The critical boundary between the relations of Vhf, Vhs and Vv as shown in Fig. 15 to adopting the
horizontal FVD-parallel flow and the vertical trench-parallel flow absolute amplitudes of anisotropic Vp, because the amplitude of the
is set to be above the 50 km depth contour of the upper interface obtained velocity anomalies strongly depends on the damping and
of the PHS slab (Fig. 18). Then, we estimate the distribution of the smoothing parameters adopted in the tomographic inversion. The
olivine fabrics in the mantle wedge above the PHS slab under SW pattern of the tomographic images obtained in this work is much
Japan, using the relations shown in Fig. 15. Under Kyushu, E-type more robust than the absolute amplitudes of velocity anomalies.
fabric dominates in the mantle wedge beneath the arc and backarc, In addition, for simplicity, we only attribute the Vp anisotropy in
besides A-type fabric in some local areas, whereas C-type fabric the mantle wedge to different olivine fabrics with the assumed
develops in the forearc mantle wedge (Fig. 18d). Beneath western flow pattern, because olivine is the major anisotropic mineral in the
Honshu, C-type fabric dominates in the mantle wedge under the arc upper mantle. Note that, besides olivine, other anisotropic minerals,
and backarc, whereas different fabrics develop in the forearc mantle such as serpentine and antigorite, should also contribute to the
wedge (Fig. 18d). seismic anisotropy in the mantle wedge (e.g. Katayama et al. 2009;
Note that the estimated olivine fabric distribution in the mantle Brownlee et al. 2013). With our present tomographic results, we
wedge depends on the flow pattern assumed. In this study, we adopt cannot estimate the olivine fabrics in the subducting Pacific and
a simple flow pattern in the mantle wedge (Fig. 16e). In fact, the flow PHS slabs formed at the mid-ocean ridges, because the slab fossil
Anisotropic tomography beneath Japan 1427

Downloaded from https://academic.oup.com/gji/article/210/3/1410/3865122 by guest on 09 September 2021


Figure 17. The same as Fig. 16 but for the images beneath Hokkaido.

fabric may be overprinted by subduction-related structures, such where obvious C-type olivine fabric exists. When water content
as plate bending leading to an olivine fabric with an inclined slip increases at high temperatures and at low stresses, A-type olivine
plane, normal faults produced at the outer rise area near the trench fabric first changes to E-type and then to C-type and the A- to
axis (e.g. Faccenda et al. 2008) and the slab internal deformation, E-type fabric boundary, as well as the E- to C-type fabric boundary,
etc. (e.g. Eakin et al. 2016). are sensitive to water content (Karato et al. 2008). The E-type
Our present results (Figs 16–18) suggest that, when assuming dominant fabric in the arc and backarc mantle wedge beneath Japan
vertical trench-parallel flow in the forearc mantle wedge, B-type may reflect a hydrated mantle wedge due to the slab dehydration
olivine fabric seems to dominate in the forearc mantle wedge only and corner flow (e.g. Zhao 2015). In addition, the C-type dominant
beneath NE Japan where the old and cold Pacific slab is subducting. fabric in the arc and backarc mantle wedge beneath western Honshu
Such a feature is consistent with a mineral physics result that B-type may reflect a high water content there, because both the PHS and
fabric develops only at low temperatures, whereas subduction of a Pacific slabs exist there and so more fluids can be released from
young and hence warm lithosphere, such as the PHS slab under SW the two slabs to the overlying mantle wedge (e.g. Asamori & Zhao
Japan, would not cause B-type fabric to a large extent (Karato et al. 2015; Liu & Zhao 2016b).
2008). Recent attenuation tomography has revealed obvious high-Q
and low-Q anomalies in the forearc mantle wedge under NE Japan
and SW Japan, respectively (e.g. Liu et al. 2014; Liu & Zhao 2015; 6 C O N C LU S I O N S
Wang et al. 2017b,c). In addition, non-volcanic deep low-frequency To clarify seismic anisotropy, flow pattern and olivine fabrics in the
earthquakes occur actively below the forearc mantle wedge under mantle wedge, we conduct a joint inversion for 3-D Vp azimuthal
SW Japan (Fig. 18), whereas none of them has been detected under and radial anisotropies, as well as isotropic Vp tomography, in the
NE Japan, though a so-called water wall may exist in the forearc crust and upper mantle of the Japan subduction zone using a large
mantle wedge under NE Japan (Zhao et al. 2015). All these features number of high-quality P-wave traveltime data of local earthquakes
indicate that the forearc mantle wedge in NE Japan is much cooler and teleseismic events. Main findings of this work are summarized
than that in SW Japan. as follows.
Our present results (Figs 16–18) also suggest that, if horizontal
FVD-parallel flow exists in the mantle wedge under the arc and (1) An anisotropic feature with horizontal trench-parallel FVDs
backarc, then E-type olivine fabric may dominate in the arc and and Vhf > Vv > Vhs is revealed in the forearc mantle wedge above
backarc mantle wedge beneath Japan, except for western Honshu the Pacific slab under NE Japan, which may indicate B-type olivine
1428 X. Liu and D. Zhao

Downloaded from https://academic.oup.com/gji/article/210/3/1410/3865122 by guest on 09 September 2021


Figure 18. The same as Fig. 16 but for the images beneath SW Japan. The red dots denote low-frequency earthquakes. UBPHS, the upper boundary of the
subducting Philippine Sea slab.

fabric with vertical trench-parallel flow existing there. Such a feature AC K N OW L E D G E M E N T S


is not obvious in the forearc mantle wedge above the PHS slab under
We thank the data centres of the Japanese seismic networks and
SW Japan, probably due to the higher temperature and more fluids
the JMA Unified Earthquake Catalog (www.hinet.bosai.go.jp) for
there associated with the young and warm PHS slab subduction.
providing the high-quality waveform and arrival-time data used in
(2) The mantle wedge beneath the arc and backarc in Japan
this study. Some of the arrival-time data were measured from the
generally exhibits horizontal trench-normal FVDs and Vhf > Vv
original seismograms by the staffs of Research Center for Predic-
> Vhs, which may indicate E-type olivine fabric with horizontal
tion of Earthquakes and Volcanic Eruptions, Tohoku University.
FVD-parallel flow there.
The free software GMT (Wessel & Smith 1998) is used for plot-
(3) The mantle wedge beneath western Honshu exhibits an
ting the figures. We appreciate the helpful discussions with Drs J.
anisotropy of Vv > Vhf > Vhs, where C-type olivine fabric may
Wang, Z. Huang, X. Wang and Y. Hua. We are very grateful to Prof
dominate. This result may suggest that the water content is the
M. Ritzwoller (the editor), Prof I. Koulakov and an anonymous re-
highest there, because both the PHS and Pacific slabs exist there
viewer who provided thoughtful review comments and suggestions
and their dehydration reactions release abundant fluids to the over-
which have improved this paper. This work was supported by grants
lying mantle wedge.
from the JSPS (Kiban-S 23224012) and the MEXT (26106005)
Anisotropic tomography beneath Japan 1429

to D. Zhao, as well as grants from the Chinese NSFC (41602207, Iwamori, H. & Zhao, D., 2000. Melting and seismic structure beneath the
41190072 and 41325009) to X. Liu. northeast Japan arc, Geophys. Res. Lett., 27, 425–428.
Jung, H. & Karato, S., 2001. Water-induced fabric transitions in olivine,
Science, 293, 1460–1463.
Jung, H., Katayama, I., Jiang, Z., Hiraga, T. & Karato, S., 2006. Effect of
REFERENCES water and stress on the lattice-preferred orientation of olivine, Tectono-
Anderson, D., 1989. Theory of the Earth, Blackwell Science Publication, physics, 421, 1–22.
p. 366. Karato, S, Jung, H, Katayama, I. & Skemer, P., 2008. Geodynamic
Ando, M., Ishikawa, Y. & Yamazaki, F., 1983. Shear wave polarization significance of seismic anisotropy of the upper mantle: new in-
anisotropy in the upper mantle beneath Honshu, Japan, J. geophys. Res., sights from laboratory studies, Annu. Rev. Earth Planet. Sci., 36,
88, 5850–5864. 59–95.
Asamori, K. & Zhao, D., 2015. Teleseismic shear wave tomography of the Katayama, I., Hirauchi, K., Michibayashi, K. & Ando, J., 2009. Trench-
Japan subduction zone, Geophys. J. Int., 203, 1752–1772. parallel anisotropy produced by serpentine deformation in the hydrated
Barclay, A., Toomey, D. & Solomon, S., 1998. Seismic structure and crustal mantle wedge, Nature, 461, 1114–1117.
magmatism at the Mid-Atlantic Ridge, 35◦ N, J. geophys. Res., 103, Kennett, B. & Engdahl, E., 1991. Traveltimes for global earthquake location
17 827–17 844. and phase identification, Geophys. J. Int., 105, 429–465.

Downloaded from https://academic.oup.com/gji/article/210/3/1410/3865122 by guest on 09 September 2021


Brownlee, S., Hacker, B., Harlow, G. & Seward, G., 2013. Seismic signatures Kita, S. & Matsubara, M., 2016. Seismic attenuation structure associated
of a hydrated mantle wedge from antigorite crystal-preferred orientation with episodic tremor and slip zone beneath Shikoku and the Kii peninsula,
(CPO), Earth planet. Sci. Lett., 375, 395–407. southwestern Japan, in the Nankai subduction zone, J. geophys. Res., 121,
Cheng, B., Zhao, D. & Zhang, G., 2011. Seismic tomography and anisotropy 1962–1982.
in the source area of the 2008 Iwate-Miyagi earthquake (M 7.2), Phys. Kneller, E., van Keken, P., Karato, S. & Park, J., 2005. B-type olivine
Earth planet. Inter., 184, 172–185. fabric in the mantle wedge: insights from high-resolution non-Newtonian
Draper, N. & Smith, H., 1966. Applied Regression Analysis, John Wiley. subduction zone models, Earth planet. Sci. Lett., 237, 781–797.
Eakin, C., Long, M., Scire, A., Beck, S., Wagner, L., Zandt, G. & Tavera, Kneller, E., Long, M. & van Keken, P., 2008. Olivine fabric transitions and
H., 2016. Internal deformation of the subducted Nazca slab inferred from shear wave anisotropy in the Ryukyu subduction system, Earth planet.
seismic anisotropy, Nat. Geosci., 9, 56–59. Sci. Lett., 268, 268–282.
Eberhart-Phillips, D. & Henderson, C., 2004. Including anisotropy in 3- Koulakov, I., Jakovlev, A. & Luehr, B., 2009. Anisotropic structure be-
D velocity inversion and application to Marlborough, New Zealand, neath central Java from local earthquake tomography, Geochem. Geophys.
Geophys. J. Int., 156, 237–254. Geosyst., 10, Q02011, doi:10.1029/2008GC002109.
Faccenda, M., Burlini, L., Gerya, T. & Mainprice, D., 2008. Fault-induced Koulakov, I., Kukarina, E., Fathi, I., El Khrepy, S. & Al-Arifi, N., 2015.
seismic anisotropy by hydration in subducting oceanic plates, Nature, Anisotropic tomography of Hokkaido reveals delamination-induced flow
455, 1097–1100. above a subducting slab, J. geophys. Res., 120, 3219–3239.
Hall, C., Fischer, K., Parmentier, E. & Blackman, D., 2000. The influence of Kreemer, C., Holt, W. & Haines, A., 2003. An integrated global model of
plate motions on three-dimensional back arc mantle flow and shear wave present-day plate motions and plate boundary deformation, Geophys. J.
splitting, J. geophys. Res., 105, 28 009–28 033. Int., 154, 8–34.
Hasegawa, A., Nakajima, J., Yanada, T., Uchida, N., Okada, T., Zhao, Laske, G., Masters, G., Ma, Z. & Pasyanos, M., 2013. Update on
D., Matsuzawa, T. & Umino, N., 2013. Complex slab structure and CRUST1.0—A 1-degree Global Model of Earth’s Crust, Geophys. Res.
arc magmatism beneath the Japanese Islands, J. Asian Earth Sci., 78, Abst., Vol. 15, Abstract EGU2013-2658.
277–290. Liu, X. & Zhao, D., 2015. Seismic attenuation tomography of the Southwest
Hayes, G., Wald, D. & Johnson, R., 2012. Slab1.0: a three-dimensional Japan arc: new insight into subduction dynamics, Geophys. J. Int., 201,
model of global subduction zone geometries, J. geophys. Res., 117, 135–156.
B01302, doi:10.1029/2011JB008524. Liu, X. & Zhao, D., 2016a. Backarc spreading and mantle wedge flow be-
Hearn, T., 1996. Anisotropic Pn tomography in the western United States, neath the Japan Sea: insight from Rayleigh-wave anisotropic tomography,
J. geophys. Res., 101, 8403–8414. Geophys. J. Int., 207, 357–373.
Hirahara, K. & Ishikawa, Y., 1984. Travel time inversion for three- Liu, X. & Zhao, D., 2016b. P and S wave tomography of Japan subduc-
dimensional P-wave velocity anisotropy, J. Phys. Earth, 32, 197–218. tion zone from joint inversions of local and teleseismic travel times and
Hiramatsu, Y., Ando, M. & Ishikawa, Y., 1997. ScS wave splitting of deep surface-wave data, Phys. Earth planet. Inter., 252, 1–22.
earthquakes around Japan, Geophys. J. Int., 128, 409–424. Liu, X. & Zhao, D., 2016c. Seismic velocity azimuthal anisotropy of the
Honda, S., 2011. Planform of small-scale convection under the island arc, Japan subduction zone: constraints from P and S wave traveltimes, J.
Geochem. Geophys. Geosyst., 12, Q11005, doi:10.1029/2011GC003827. geophys. Res., 121, 5086–5115.
Huang, Z., Zhao, D. & Wang, L., 2011a. Shear wave anisotropy in the Liu, X., Zhao, D. & Li, S., 2013. Seismic heterogeneity and anisotropy of
crust, mantle wedge, and subducting Pacific slab under northeast Japan, the southern Kuril arc: insight into megathrust earthquakes, Geophys. J.
Geochem. Geophys. Geosyst., 12, Q01002, doi:10.1029/2010gc003343. Int., 194, 1069–1090.
Huang, Z., Zhao, D. & Wang, L., 2011b. Frequency-dependent shear-wave Liu, X., Zhao, D. & Li, S., 2014. Seismic attenuation tomography
splitting and multilayer anisotropy in Northeast Japan, Geophys. Res. of the Northeast Japan arc: insight into the 2011 Tohoku earth-
Lett., 38, L08302, doi:10.1029/2011GL046804. quake (Mw 9.0) and subduction dynamics, J. geophys. Res., 119,
Huang, Z., Zhao, D. & Wang, L., 2011c. Seismic heterogeneity and 1094–1118.
anisotropy of the Honshu arc from the Japan Trench to the Japan Sea, Long, M., 2013. Constraints on subduction geodynamics from seismic
Geophys. J. Int., 184, 1428–1444. anisotropy, Rev. Geophys., 51, 76–112.
Huang, Z., Zhao, D. & Liu, X., 2015. On the trade-off between seismic Long, M. & Wirth, E., 2013. Mantle flow in subduction systems: the mantle
anisotropy and heterogeneity: numerical simulations and application to wedge flow field and implications for wedge processes, J. geophys. Res.,
Northeast Japan, J. geophys. Res., 120, 3255–3277. 118, 583–606.
Huang, Z., Zhao, D., Hasegawa, A., Umino, N., Park, J. & Kang, I., 2013. Long, M., Hager, B., De Hoop, M. & Van Der Hilst, R., 2007. Two-
Aseismic deep subduction of the Philippine Sea plate and slab window, dimensional modelling of subduction zone anisotropy with application
J. Asian Earth Sci., 75, 82–94. to southwestern Japan, Geophys. J. Int., 170, 839–856.
Ishise, M. & Oda, H., 2005. Three-dimensional structure of P-wave Lynner, C., Long, M., Thissen, C., Paczkowski, K. & Montési, L., 2017.
anisotropy beneath the Tohoku district, northeast Japan, J. geophys. Res., Evaluating geodynamic models for sub-slab anisotropy: effects of olivine
110, B07304, doi:10.1029/2004jb003599. fabric type, Geosphere, 13, doi:10.1130/GES01395.1.
1430 X. Liu and D. Zhao

Maupin, V. & Park, J., 2007. Theory and observations—wave propagation Wang, J., Zhao, D. & Yao, Z., 2017a. Seismic anisotropy evidence for de-
in anisotropic media, in Treatise on Geophysics, pp. 289–321, Elsevier. hydration embrittlement triggering intermediate-depth earthquakes, Sci.
Maupin, V. & Park, J., 2015. Theory and observations—seismic anisotropy, Rep., 7, 2613, doi:10.1038/s41598-017-02563-w.
in Treatise on Geophysics 2nd edn, pp. 277–305, Elsevier. Wang, Z.W., Zhao, D., Liu, X. & Li, X., 2017b. Seismic attenuation tomog-
Müller, R., Sdrolias, M., Gaina, C. & Roest, W., 2008. Age, spreading rates, raphy of the source zone of the 2016 Kumamoto earthquake (M 7.3), J.
and spreading asymmetry of the world’s ocean crust, Geochem. Geophys. geophys. Res., 122, 2988–3007.
Geosyst., 9, Q04006, doi:10.1029/2007gc001743. Wang, Z.W., Zhao, D., Liu, X., Chen, C. & Li, X., 2017c. P and S wave at-
Nagaya, T., Walker, A., Wookey, J., Wallis, S., Ishii, K. & Kendall, J., 2016. tenuation tomography of the Japan subduction zone, Geochem. Geophys.
Seismic evidence for flow in the hydrated mantle wedge of the Ryukyu Geosyst., 18, 1688–1710.
subduction zone, Sci. Rep., 6, 29981, doi:10.1038/srep29981. Watanabe, M. & Oda, H., 2014. Regional variations of the shear-wave
Nakajima, J., Shimizu, J., Hori, S. & Hasegawa, A., 2006. Shear-wave split- polarization anisotropy in the crust and mantle wedge beneath the Tohoku
ting beneath the southwestern Kurile arc and northeastern Japan arc: a district, Phys. Earth planet. Inter., 235, 49–65.
new insight into mantle return flow, Geophys. Res. Lett., 33, L05305, Wei, W., Zhao, D., Xu, J., Wei, F. & Liu, G., 2015. P and S wave tomography
doi:10.1029/2005gl025053. and anisotropy in Northwest Pacific and East Asia: constraints on stagnant
Nakajima, J., Hirose, F. & Hasegawa, A., 2009. Seismotectonics beneath the slab and intraplate volcanism, J. geophys. Res., 120, 1642–1666.
Tokyo metropolitan area, Japan: effect of slab-slab contact and overlap on Wessel, P. & Smith, W.H.F., 1998. New, improved version of Generic

Downloaded from https://academic.oup.com/gji/article/210/3/1410/3865122 by guest on 09 September 2021


seismicity, J. geophys. Res., 114, B08309, doi:10.1029/2008JB006101. Mapping Tools released, EOS, Trans. Am. geophys. Un., 79, 579,
Niu, X., Zhao, D., Li, J. & Ruan, A., 2016. P wave azimuthal and radial doi:10.1029/98EO00426.
anisotropy of the Hokkaido subduction zone, J. geophys. Res., 121, 2636– Wirth, E. & Korenaga, J., 2012. Small-scale convection in the subduction
2660. zone mantle wedge, Earth planet. Sci. Lett., 357–358, 111–118.
Okada, T., Matsuzawa, T. & Hasegawa, A., 1995. Shear-wave polarization Wirth, E. & Long, M., 2010. Frequency-dependent shear wave splitting
anisotropy beneath the north-eastern part of Honshu, Japan, Geophys. J. beneath the Japan and Izu-Bonin subduction zones, Phys. Earth planet.
Int., 123, 781–797. Inter., 181, 141–154.
Okada, Y., Kasahara, K., Hori, S., Obara, K., Sekiguchi, S., Fujiwara, H. & Wirth, E. & Long, M., 2012. Multiple layers of seismic anisotropy and a
Yamamoto, A., 2004. Recent progress of seismic observation networks low-velocity region in the mantle wedge beneath Japan: evidence from
in Japan: Hi-net, F-net, K-net and KiK-net, Earth Planet Space, 56, teleseismic receiver functions, Geochem. Geophys. Geosyst., 13, Q08005,
15–28. doi:10.1029/2012gc004180.
Paige, C. & Saunders, M., 1982. LSQR: an algorithm for sparse lin- Xia, S., Zhao, D., Qiu, X., Nakajima, J., Matsuzawa, T. & Hasegawa, A.,
ear equations and sparse least squares, ACM Trans. Math. Soft., 8, 2007. Mapping the crustal structure under active volcanoes in central
43–71. Tohoku, Japan using P and PmP data, Geophys. Res. Lett., 34, L10309,
Rabbel, W., Koulakov, I., Dinc, A. & Jakovlev, A., 2011. Arc-parallel shear doi:10.1029/2007gl030026.
deformation and escape flow in the mantle wedge of the Central America Xia, S., Zhao, D. & Qiu, X., 2008. Tomographic evidence for the subducting
subduction zone: evidence from P wave anisotropy, Geochem. Geophys. oceanic crust and forearc mantle serpentinization under Kyushu, Japan,
Geosyst., 12, Q05S31, doi:10.1029/2010GC003325. Tectonophysics, 449, 85–96.
Savage, M., 1999. Seismic anisotropy and mantle deformation: what have Yoshizawa, K., Miyake, K. & Yomogida, K., 2010. 3D upper mantle struc-
we learned from shear wave splitting?, Rev. Geophys., 37, 65–106. ture beneath Japan and its surrounding region from inter-station disper-
Savage, M. et al., 2016. Stress, strain rate and anisotropy in Kyushu, Japan, sion measurements of surface waves, Phys. Earth planet. Inter., 183,
Earth planet. Sci. Lett., 439, 129–142. 4–19.
Skemer, P. & Hansen, L., 2016. Inferring upper-mantle flow from seismic Zhao, D., 2015. Multiscale Seismic Tomography, Springer, p. 304.
anisotropy: an experimental perspective, Tectonophysics, 668–669, 1–14. Zhao, D., Fujisawa, M. & Toyokuni, G., 2017. Tomography of the subducting
Tamura, Y., Tatsumi, Y., Zhao, D., Kido, Y. & Shukuno, H., 2002. Hot fingers Pacific slab and the 2015 Bonin deepest earthquake (Mw 7.9), Sci. Rep.,
in the mantle wedge: new insights into magma genesis in subduction 7, 44487, doi:10.1038/srep44487.
zones, Earth planet. Sci. Lett., 197, 105–116. Zhao, D., Hasegawa, A. & Horiuchi, S., 1992. Tomographic imaging of P
Tatsumi, Y., 1989. Migration of fluid phases and genesis of basalt magmas and S wave velocity structure beneath northeastern Japan, J. geophys.
in subduction zones, J. geophys. Res., 94, 4697–4707. Res., 97, 19 909–19 928.
Tian, Y. & Zhao, D., 2012. Seismic anisotropy and heterogeneity in the Zhao, D., Hasegawa, A. & Kanamori, H., 1994. Deep structure of Japan
Alaska subduction zone, Geophys. J. Int., 190, 629–649. subduction zone as derived from local, regional, and teleseismic events,
Tian, Y. & Liu, L., 2013. Geophysical properties and seismotectonics of the J. geophys. Res., 99, 22 313–22 329.
Tohoku forearc region, J. Asian Earth Sci., 64, 235–244. Zhao, D., Yanada, T., Hasegawa, A., Umino, N. & Wei, W., 2012. Imaging the
Tong, P., Zhao, D. & Yang, D., 2012. Tomography of the 2011 Iwaki earth- subducting slabs and mantle upwelling under the Japan Islands, Geophys.
quake (M 7.0) and Fukushima nuclear power plant area, Solid Earth, 3, J. Int., 190, 816–828.
43–51. Zhao, D., Kitagawa, H. & Toyokuni, G., 2015. A water wall in the
Tono, Y., Fukao, Y., Kunugi, T. & Tsuboi, S., 2009. Seismic anisotropy of the Tohoku forearc causing large crustal earthquakes, Geophys. J. Int., 200,
Pacific slab and mantle wedge beneath the Japanese islands, J. geophys. 149–172.
Res., 114, B07307, doi:10.1029/2009JB006290. Zhao, D., Yu, S. & Liu, X., 2016. Seismic anisotropy tomography: new
VanDecar, J. & Crosson, R., 1990. Determination of teleseismic rela- insight into subduction dynamics, Gondwana Res., 33, 24–43.
tive phase arrival times using multi-channel cross-correlation and least
squares, Bull. seism. Soc. Am., 80, 150–169.
Wang, Z. & Zhao, D., 2005. Seismic imaging of the entire arc of Tohoku S U P P O RT I N G I N F O R M AT I O N
and Hokkaido in Japan using P-wave, S-wave and sP depth-phase data,
Phys. Earth planet. Inter., 152, 144–162. Supplementary data are available at GJI online.
Wang, J. & Zhao, D., 2008. P-wave anisotropic tomography beneath North-
east Japan, Phys. Earth planet. Inter., 170, 115–133. Table S1 shows values of some parameters and other related infor-
Wang, J. & Zhao, D., 2010. Mapping P-wave anisotropy of the Honshu arc mation on the tomographic inversions.
from Japan Trench to the back-arc, J. Asian Earth Sci., 39, 396–407. Figure S1 shows the 1-D Vp model used in this study.
Wang, J. & Zhao, D., 2013. P-wave tomography for 3-D radial and azimuthal Figures S2–S13 show the coverages of P-wave rays in our data set.
anisotropy of Tohoku and Kyushu subduction zones, Geophys. J. Int., 193, Figures S14–S31 show the results of checkerboard resolution tests
1166–1181. (CRTs) we performed.
Anisotropic tomography beneath Japan 1431

Figures S32–S57 show the tomographic models obtained by using Figures S88–S102 show obtained tomographic images and related
different parametrization strategies and inversion methods. synthetic test results in a plane located in the central mantle wedge
Figures S58–S72 show the obtained tomographic images and re- beneath SW Japan.
lated synthetic test results in a plane located in the central mantle
wedge beneath NE Japan. Please note: Oxford University Press is not responsible for the con-
Figures S73–S87 show obtained tomographic images and related tent or functionality of any supporting materials supplied by the
synthetic test results in a plane located in the central mantle wedge authors. Any queries (other than missing material) should be di-
beneath Hokkaido. rected to the corresponding author for the paper.

Downloaded from https://academic.oup.com/gji/article/210/3/1410/3865122 by guest on 09 September 2021

You might also like