Seismic anisotropy near source region in subduction zones around Japan

Seismic anisotropy near source region in subduction zones around Japan

PHYSICS O F T H E EARTH ANDPLANETARY INTERIORS ELSEVIER Physics of the Earth and Planetary Inl~riors95 (1996) 237-250 Seismic anisotropy near sourc...

997KB Sizes 0 Downloads 59 Views

PHYSICS O F T H E EARTH ANDPLANETARY INTERIORS

ELSEVIER

Physics of the Earth and Planetary Inl~riors95 (1996) 237-250

Seismic anisotropy near source region in subduction zones around Japan Yoshihiro Hiramatsu *, Masataka Ando Disaster Prevention Research Institute, Kyoto University, Ufi, Kyoto, 611, Japan

Received 2 October 1994; accepted 31 August 1995

Abstract

Broad-hand seismographs have been distributed widely over the Japanese islands and have provided us with high-quality digital waveform data in recent years. We have investigated splitting of ScS waves from deep earthquakes in the Kuril and Izu-Bonin subduction zones recorded on the STS seismograms. We used ScS waves from 13 events observed in Japan and recognized shear-wave splitting which is mainly caused by anisotropy in the lithospheric slab. The anisotropy was confirmed around the source region by comparing ScS-wave splitting from deep earthquakes in different regions and at different depths with S-wave splitting from deep events just beneath central Japan. We have found that the anisotropic regions exist within the subducting slab, although they are distributed locally in and around the source, possibly as patches with a diameter of 100km or less. Discrepancy in the direction of the fast polarized shear waves between two nearby events is observed in the Izu-Bonin subduction zone: the splitting of the shallower event (289 km) shows its fast polarization direction to be parallel to the fossil motion of the Pacific plate, but that of the deeper event (361 km) immediately beneath is parallel to the current motion. These observations suggest that the change in shear-wave splitting with depth comes from the modification of preferred orientation of minerals depending on depth. We propose in this study that the reorientation of minerals (olivine) occurs owing to the change in physical conditions associated with the phase transformation of olivine ( a phase) to modified spinel ( fl phase).

1. I n t r o d u c t i o n The Earth is a thermally active planet and as a result plastic flow occurs in the Earth's interior, Preferred orientation of minerals owing to the plastic flow causes anisotropy in the lithosphere and asthenosphere. Shear-wave splitting is a powerful tool to investigate the seismic anisotropy in the Earth. Since Ando et ai. (1980) first found clear shear-wave splitting beneath a volcanic area in central Japan,

* Corresponding author,

many researchers have reported shear-wave splitting using S waves (Bowman and Ando, 1987; Xie, 1992) and ScS waves (Ando, 1984; Fukao, 1984) in subduction zones, and SKS and SKKS waves in continental regions (Silver and Chan, 1988, 1991; Vinnik et al., 1989, 1992; Savage et al., 1990). Fukao (1984) analyzed ScS waves from a deep event beneath the Sea of Okhotsk and obtained a uniform orientation of fast polarized shear waves parallel to the fossil Pacific plate motion near Japan. Recently, by combining the results for SKS, SKKS and direct S waves, it has become possible to investigate anisotropy not only beneath the receiver but also

0031-9201/96/$15.00 © 1996 Elsevier Science B.V. All rights reserved SSDI 0031-9201(95)03 119-7

238

Y. Hiramatsu, M. Ando / Physics q/the Earth and Phmetary Interiors 95 (1996) 237-250

around the source. Silver and Chart (1988, 1991) analyzed S-wave splitting using SKS and SKKS waves with steep incident angles beneath receivers in North America. On the basis of the results of Silver and Chan (1991), Kaneshima and Silver (1992) reported anisotropy occupying the source side in the subduction zones of the Nazca plate beneath South America and the Pacific plate beneath Kamchatka. In each region, the time delays between the two split waves are found to be 1.0-1.5 s where the polarization direction of the fast wave is perpendicular to the trench. Vinnik and Kind (1993) showed source side anisotropy using teleseismic S waves recorded at GRF from events in the North Pacific from the Kuril Islands to the Aleutians. The direction of the fast polarized waves approximately coincides with the strike of the island arc. The anisotropy of a mantle wedge of the Kuril-Kamchatka subduction zone was investigated by Fischer and Yang (1994)using teleseismic S and sS phases from nearby deep events, They obtained a fast polarization direction parallel to the plate convergence. Russo and Silver (1994) observed trench-parallel fast polarization in South America and proposed trench-parallel flow attributed to retrograde motion of the subducting slab beneath the Nazca plate. To distinguish source side anisotropy from receiver side anisotropy, the splitting of SKS or SKKS waves is useful because it is caused only by the receiver side anisotropy lying along the path in the mantle between the P-to-S conversion point at the core-mantle boundary and the receiver point, and has no contribution from anisotropy of the source region. However, no clear SKS and SKKS waves can be recorded in Japan. Although these phases can be observed at epicentral distances of 80 ° to 110°, such earthquakes are not well recorded, Anisotropy near the source region and that beneath the receiver region are indistinguishable from ScS record data alone. However, we have overcome this obstacle by using as many waves with different combinations of sources and receivers as possible. In this paper, we report the ScS-wave splitting from deep events in the northwest Pacific, Kuril and I z u Bonin subduction zones. Comparing the results for ScS with those for direct S waves from deep events recorded in central Japan, we confirm the existence of local anisotropic regions within the mantle wedge

beneath Japan as well as within the slab around the source.

2. Data The broad-band digital three-component seismometer is a powerful tool in global seismology. STS seismometers have been widely distributed over Japan since 1988. We use ScS waves at seven stations distributed over Honshu. These stations are AOB (Aoba) of Tohoku University, TSK (Tsukuba) and SHK (Shiraki) of the Earthquake Research Institute of the University of Tokyo, INU ( I n u y a m a ) o f Nagoya University, KTJ (Kamitakara), HKJ (Hokuriku) and TTT (Tottori)of the Disaster Prevention Research Institute, Kyoto University (Fig. 1). Shear-wave splitting is usually studied using S waves which are incident vertically, to avoid the interaction of P - S V conversion at the free surface, such as direct S from deep events near receivers, ScS waves from deep events with an epicentral distance less than 30 °, and SKS waves from teleseismic events with an epicentral distance of 80° to 110 °. For the restricted location of receivers and events, direct S and ScS waves are analyzed for the subduction zones and SKS waves for continental regions. A large number of records from deep events along the subduction zone of the Pacific plate in and around Japan are now available. The observed S and ScS waves are appropriate for the study of shear-wave splitting. To analyze ScS-wave splitting we select events for a period of 1988-1993 satisfying the following conditions: epicentral distance less than 20°; body wave magnitudes greater than 5.5; hypocentral depths greater than 100kin; signal-tonoise ratio sufficiently high. Thirteen events are finally selected, for most of which we have only a few available records. In this study, we mainly use digital records of ScS waves at KTJ, HKJ, T I T and INU where many ScS-wave polarizations can be observed. We analyze four events in the Kuril subduction zone, eight in the Izu-Bonin subduction zones and one beneath central Honshu, Japan. All rays incident at the stations from these events are steep, with an angle less than 10°. We therefore can neglect the effect of the free surface on the S-wave particle motion. Fig. 1 shows the location of the

239

Y. Hiramatsu, M. Ando / Physics of the Earth and Planetary Interiors 95 (1996) 237-250

45" N

~

g

37 °

i'i •

1

N

55"E

Fig. 1. The station ( • ) and the epicenter (O)distribution used in this study. The number in each circle corresponds to the event

36"N

numberof Table 1. 137" E

stations and e v e n t s used in this study. W e also analyze S w a v e s o f deep events beneath central Honshu, Japan, r e c o r d e d at K T J (Fig. 2). A n d o et al. (1983), hereafter referred to as A N 8 3 , o b s e r v e d s h e a r - w a v e splitting in this area using short period s e i s m o g r a m s and c o n c l u d e d that anisotropic bodies exist beneath K T J and I N U . The area a n a l y z e d in this study is smaller than that o f A N 8 3 . W e restrict our use o f S w a v e s to those w h o s e incident angles are less than 20 ° at the free surface. T o separate the anisotropy into the source side and the r e c e i v e r side, w e use S c S - w a v e splitting f r o m different sources and

Fig. 2. Station KTJ (,t) and epicenters (O) used for splitting study beneath central Japan. S - w a v e splitting i m m e d i a t e l y b e n e a t h receivers• If an anisotropic region exists locally beneath a receiver, we w o u l d e x p e c t a u n i f o r m orientation o f fast polarized w a v e s and a u n i f o r m arrival time difference b e t w e e n the two polarized w a v e s for both ScS and S waves• O n the contrary, if an anisotropic b o d y is located in the source area, a u n i f o r m polarization w o u l d be o b s e r v e d at all the stations. A c o m p a r i s o n o f the S c S - w a v e splitting obtained in this study with

Table 1 The source parameters of the events used for ScS-wave splitting in this study No.

Year

Month

Day

Hour

Minute

Second

Latitude (°N)

Longitude (°E)

Depth (km)

Mb

Region

1 2 3 4 5 6 7 8 9 10 11 12 13

1988 1989 1990 1990 1991 1991 1992 1992 1992 1992 1992 1992 1993

09 06 05 08 10 12 01 06 06 08 08 10 I0

07 16 12 05 08 17 20 01 16 07 29 30 11

11 23 04 01 03 06 13 18 05 11 19 02 15

53 42 50 34 31 38 37 29 51 I1 19 49 54

24.13 35.12 08.70 55.83 15.60 17.32 03.08 20.27 03.74 41.69 05.59 48.17 22.58

30.245 31.807 49.040 29.551 45.587 47.393 27.983 29.739 45.704 35.728 33.190 29.941 32.005

137.431 137.982 141.850 137.630 149.049 151.499 139.405 140.699 142.263 135.152 137.975 138.975 137.953

485 360 606 496 146 157 499 134 317 358 289 393 361

6.1 5.9 6.5 6.0 6.0 5.8 5.8 5.5 5.7 5.6 6.0 6.0 6.1

South of Honshu, Japan South of Honshu, Japan Sakhalin Island South of Honshu, Japan Kuril Islands Kuril Islands Bonin Islands South of honshu, Japan Hokkaido, Japan Western Honshu, Japan South of Honshu, Japan South of Honshu, Japan South of Honshu, Japan

240

Y. Hiramatsu.M. Ando/ Physicsof the Earthand PlanetaryInteriors95 (1996)237-250

45 N

~ ~ ,

3. Methodof estimationof splittingparameter

t

The shear wave splits into two orthogonal components with different velocities when it passes through an anisotropic region. In anisotropic homogeneous medium an S waveform amving at time T O may be described by

~ ,/-

~rc'f . ~ ~

35=Erl LF..,~" 4~~121J12~ ~.J~:~ ~,,.,,,~/ f--'j~~5 ~

us(t°) = ws(t°)exp[-it°To]F(qb,St)"Ps

the % displacement as afunction functionwhich of freqwhere u e n c y , usto,isand is the wavelet is

TM

25°N25

(1)

°E

Fig. 3. The best double couple of CMT solutions of deep earthquakes analyzed in this study. All mechanisms are represented by

lower-hemisphere equal-area projection,

the S-wave splitting in AN83 may include some uncertainties owing to the difference in instrumental response. The seismometers used in AN83 are of 1 Hz velocity type. However, the STS-1 used in this study is flat in velocity over 1 s to 360 s. To avoid the difference in the effect of instrumental response on splitting parameters we first compare the S-wave splitting of deep events beneath KTJ observed in this study with those observed by AN83. The source parameters are summarized in Table 1 for ScS waves and Table 2 for S waves. Fig. 3 shows centroid moment tensor (CMT)solutions of the events used for ScS-wave splitting analyses, but no CMT solution is available for the events of the S-wave study because of its small magnitude,

the product of scalar amplitude, the Fourier transformed source time function, attenuation operator, and instrument response (Silver and Chan, 1991). The vector Ps describes the orientation of the displacement vector in an isotropic medium. The tensor F(d,,St) describes shear-wave splitting as a projection of Ps onto fast and slow horizontal directions, at azimuths of q~ and q6 + ~r/2 from north, with a time-shifting by - S t / 2 and ~t/2, respectively. In this formulation we neglect phase shifts in the radial component owing to the free surface or the c o r e mantle boundary because of the nearly vertical incidence of S waves used in this study. We measure these parameters (~b,~t) using the two horizontal components of the seismogram. The orientation of the initial motion of an ScS or S wave is measured directly from the particle motion diagram of the two horizontal components clockwise from the north. This orientation is taken as that of the fast direction. After rotating the two horizontal components by the measured angle, the arrival time difference is estimated by taking cross-correlation coefficients for time shift + 3 s to - 3 s with an increment of 0.05 s. The lag time is taken when an absolute cross-correlation coefficient becomes maximum. Fig. 4 shows an

Table 2 The source parameters of the events used for S-wave splitting in this study (reported in the Seismological Bulletin of the Japan Meteorological Agency) No. Year Month Day Hour Minute Latitude Longitude Depth Magnitude (°N) (°E) (km) 1 2 3 4 5

1989 1989 1991 1991 1992

03 09 03 I1 07

07 19 04 18 22

03 06 22 09 05

30 42 39 18 22

36.28 37.07 35.86 37.20 36.96

137.01 137.15 137.59 137.53 137.45

280 269 244 259 260

5.0 5.4 4.9 4.9 5.0

Y. Hiramatsu, M. Ando / Physics' of the Earth and Planetary Interiors 95 (1996) 237-250

example o f original and rotated ScS waveforms and their particle motions. However, the cross-correlation coefficients occasionally have a maximum absolute 1903110111 SOUTH OF HONSHU, JAPAN

-

South of Honshu, Jal~n

9vlz117 KURILISLANDS

./~':~

! !

~

........... E-W_l

/i

-

HKJ

a9/06/16

241

] !

....~./]

_-

......•

41-

20S

-

N

E

-

Fig. 5. The predicted polarization of soS waves (thin arrows) by CMT solution and corrected polarization obtained by removing

a = 0.50 s

" /\/1'-'..........,I,-;+,:-_

f,

r , il

,V I , _

294

(b)

I~aa~WaaSOUmOFHONSHU.aAP~ I A [ ' - - 'N - s . . . . . . . . . .-.E. -. wE-w ~,.,.J I _ ~ - / , 't I , I VI , I , - 20, - 1989/03/07 TOYAMA GIFU BORDER

///' J,

'i

HKJ I"

'

"

-

"

"

N E

KTJ

-"

......... [ - W

I\

:

4. Local anisotropy beaneath KTJ, central Hon-

N

i] [ ' ' ' ' ' '!]' ' ' -

"

length. To confirm the validity o f splitting parameters we compare the two directions o f ScS-wave polarization: the observed direction and the theoretical one predicted from the C M T solutions of the corresponding event. The observed direction is corrected for the anisotropy by shifting and rotating the two seismograms backward. If the corrected polarization direction significantly differs from the C M T solution, we choose a value at the next peak of the correlogram as the time lag. In this case, a best-fit lag time takes a non-maximum absolute value. Fig. 5 shows the corrected polarization directions o f ScS waves and predicted ones from the C M T solutions.

los ~

J , ~ ,~ !, , t , r ~ ]1 I!........ +o6-28s~ i ~- - l []I ' ~= "~ ~ ~ "' ~\ ~t ~t

E ~t = 0 . 0 0 s

:

~

16

...................... 106 Fig. 4. Examples of waveform and particle motion of ScS and S waves. (a) Upper: original ScS waveform and particle motion; lower, rotated waveform and time-shifted particle motion. (b) Example of linear particle motion. (c) Record of S waves at KTJ. The arrows show the time window of particle motion.

First we show the results o f S-wave splitting of deep earthquakes beneath KTJ (Fig. 6; see Table 3). The black bars show the direction of the fast polarized waves ~b and the size of the black circle corresponds to the arrival time difference ~t. All circles are plotted at the epicenters. The directions o f fast polarized waves 4' fall in the range 5 - 2 4 ° from north, but arrival time differences vary from 0.35 to 0.95 s. Only within a limited area around KTJ did AN83 observe S-wave splitting, with a dominant orientation of 15 ° and time delay o f 0.7-1.1 s (Fig. 7). Two horizontal components were rotated every 15 ° and the two rotated components were shifted

242

Y. Hiramatsu, M. Ando / Physics of the Earth and Planetary Interiors 95 (1996) 237-250

i/i:

f

f-

"

\

, 36° N

;,J

137° E

137° E

Fig. 6. The results of the direction and scale o f anisotropy of S waves beneath central Japan observed at KTJ. The solid bars show the direction of initial motion of S waves, which is taken as the anisotropy direction, and the radius o f the solid circles represents the arrival time difference between the two polarized waves.

Fig. 7. The results of S-wave splitting reported by Ando et al. (1983). All symbols arc the same as for Fig. 6.

These symbols arc plotted at epicenters,

between the two but to the difference in ray path. Our observations clearly show an incident angle dependence upon arrival time difference between the two polarized waves. Event 1, with the smallest incident angle of 4.2 °, shows the largest f t of 0.95 s, whereas Event 4, with incident angle of 15.9 °, shows the smallest fit of 0.35 s. The event with the largest f t lies just beneath KTJ, and the events with smaller f t are further away from the receiver. A similar tendency was also recognized by AN83. This fact implies that an anisotropic body exists just beneath KTJ. The average incident angle to the free surface for S waves in the present study is 10.9 °, which is close to the ScS incidence to KTJ and close to the result given by AN83 for local events immediately beneath KTJ (5°). We therefore conclude that no significant difference in S-wave splitting exists between the present study and AN83 as a result of the differences in instrumental response. W e adopt the results obtained in this study and those reported by AN83 as the reference anisotropy on the receiver side beneath KTJ and INU. McKenzie (1979) and Ribe (1989) showed that strong anisotropy of olivine with the a-axis parallel to the dip of the slab is formed just above a subducting slab. However, we cannot consider that the S-wave splitting observed in this study and by AN83 is

every 0.1 s to find th and St in AN83. The direction of fast polarized waves of this study is consistent with that found by AN83. Nevertheless, the lag time differs somewhat from that of AN83; the lag time of the present study is smaller than that of AN83. We believe the data used in this study are more reliable than those of AN83 because our data were recorded digitally. As AN83 used hand-digitized data, some errors were probably included. Ando et al. estimated the digitization error to be 0.15 s at most. This value is too small to explain the discrepancies between the two estimates. We consider that this discrepancy is attributed not to difference in instrumental response

Table 3 The splitting parameter of S waves estimated in this study at KTJ, central

No. 1 2 3 4 5

Japan 4, 16 5 11 10 24

St 0.95 0.40 0.40 0.35 0.50

Y. Hiramatsu, M. Ando / Physics of the Earth and Planetary Interiors 95 (1996) 237-250

caused by anisotropy just above the subducting slab because the dip direction of the subducting slab beneath KTJ is east-west. Our results are also consistent with the model proposed by Ando (1986), in which anisotropy is caused by a mantle diapir with magma-filled crack alignment, based on the finding that the location of concentration of the ray paths with clear S-wave splitting corresponds to a lowvelocity body uncovered by a P-wave three-dimensional inversion by Hirahara et al. (1989).

5. Anisotropy observed in SeS waves

243

Table 4 The splitting parameters of ScS waves estimated in this study

No. 1 2

STC TIT INU TrT

3

KTJ

q~ 48 125 118

~t 1 1.5 0.70

thc 110 114

tbcmt 24 107 106

183

0.85

148

160

HKJ INU TIT HKJ

141 148 310 340

1 I 0.95 0.80

5 6

KTJ KTJ INU

14 181 175

1.00 1.75 0.8

7

KTJ

177

INU

67

4

5 39

160 160 30 28

332 153 135

314 149 150

0.55

50

43

0.4

39

43

All ScS waves used in this study are incident at the surface nearly vertically, or at most at an angle of

8

KTJ TrT

179 232

0.40 I

217

252 252

10°, and the apparent ellipsoidal motion induced by the free surface or core-mantle boundary is negligible. The observed splitting reflects anisotropic properties along the entire ray paths of ScS waves from source to receivers. First, we assume there is one anisotropic region along the ScS-wave as well as the S-wave ray paths. Shear-wave splitting parameters 4, and 8t from ScS waves thus obtained are summarized in Table 4. Only a few seismograms are available for each event, except Event 13, for which six receivers are available. The values of 4, vary over 0 ° to 180 ° and 81 over 0.3s to 1.75s. The observed arrival time differences cannot be attributed entirely to anisotropy in the crust because the crustal anisotropy generates 0.2-0.3 s at most (Kaneshima, 1990). We can easily recognize various cases as follows: (1) 4' and 8t are different among receivers; (2) they are almost the same at all the receivers for a single event; (3) they are different among events even at the same receiver. Fig. 8 shows 4' and 8t of each event plotted at receivers. For Kuril events, we observe clear ScS-wave splitting for Events 5 and 6. Both events show the same 4' (nearly north-south), which is consistent with Fukao (1984) although 81 of Event 6 is twice as large as that of Event 5 at KTJ. However, Event 3 shows clear splitting only at KTJ, and no splitting is observed for Event 9. For Izu-Bonin events, Events 7 and 8 show different 4' values between KTJ and INU but these are similar to 4' values of S waves caused by the local anisotropy beneath KTJ and INU (AN83), although the values of 8t are slightly smaller than those of AN83. Event

9 10 11

HKJ KTJ HKJ INU HKJ

INU

241

0.5

226

252

12 13

48 14 138 120 65

1 0.45 0.55 0.7 1

63 84 87

45 54 94 94 43

AOB TSK

111 110

0.60 1.15

83 78

88 88

INU KTJ HKJ SHK

112 102 114 107

1.1 0.15 0.50 1.60

85 74 79 85

88 88 88 88

13 shows a uniform polarization direction 4, of 105 ° to 120 ° with scattered arrival time differences among stations distributed widely over Japan. This nearly uniform distribution of 4' is also found in the ScSwave splitting reported by Fukao (1984), although there is a difference in 4' between the two. We also observe less splitting, or none, at HKJ and T I T for Events 1, 9 and 12. These observations may be consistent with the idea that anisotropic regions lie nonuniformly beneath receivers only. However, ScS-wave splitting is observed even at HKJ a n d / o r T I T for Events 2, 4, 11 and 13. The difference in splitting parameters among events can be explained consistently in terms of source side anisotropy because the ray paths incident to the receiver are nearly vertical for all ScS waves at a receiver and always pass through the anisotropic region beneath it. We discuss this problem below.

244

Y. Hiramatsu, M. Ando / Physics of the Earth and Planetary Interiors 95 (1996) 237-250

6. Location of anisotropic region implicated by ScS-wave splitting

solve this problem by comparing the results of different source-receiver combinations of ScS waves with local anisotropy beneath KTJ and INU by Swave splitting reported in both the present study and by AN83. Following the formula of Kaneshima and Silver (1992) for an S waveform traveling through two

We observe clear ScS-wave splitting in this study, but the location of the anisotropic region along the ScS-wave propagation path--beneath the receivers, near the source, or both--remains unknown. We can

3S'N

35"N~

35"N, O1.5 s

el.0s .0.5s

S '" 25"N

125" E

135" E

145" E

q '" 155" E

25"N

125" E

]

el.0s .0.5s 135" E

(~

,

"

145" E

.

.

e1.5 s

2S'N 11:5" E .

q '"

125"

"

el.0s .0.5s 135"

145"

155"

.

~J

3.q'N

35"N

*

el.0s • 0~5s

"" 2S" N S" E

155" E

145" E

el.0s • 0.5s

'" 156" E

25" 11125. E

. . . . . . . . .

38"N

135" E

(~)

145" E

155" E

*

155"E

25"

135" E

145' E

@/'~/~"

155" E

"

35"N

*

.o.5s 145"E

25' N 2 5 ' E

~p,~,, ,

el.0s

135"E

el.0s • O.5s

'"

35"N

'" S " N2S* E

3S'N

I '" %' ;-~:

01.5 s el.0s

e1.5 s el.0s

.o.~s 135"E

145"E

! 155"E

25" N2S" E

'"

.o.~s 135'E

145"E

155'E

Fig. 8. The results of SeS-wave splitting observed in this study. ,k, Epicenter. The solid bars and circles show the direction of fast polarized ScS waves and arrival time difference between the two split waves, respectively. The open circles show the linear particle motion representing no splitting or less splitting of SeS waves. These symbols are plotted at receivers.

Y. Hiramatsu, M. Ando / Physics of the Earth and Planetary Interiors 95 (1996) 237-250

3S'N

245

35"N

f 25" N25" E

el0s .0.5s

'"

"135"E

145"e

t 155"E

el0s .05s

'"

25" N125" E

135"E

t45"E

lit # 155'E

25" N25" E

"

135"E

el0s .O5s 145"E

155"E

~'N

ol.0s • 05 s

¢ '" 25"N

E

135" E

145" E

155" I=

Fig. 8 (continued).

anisotropic regions, we estimate the splitting parameters of the source region. Assuming that S and ScS undergo identical splitting beneath the receiver region, ScS waveforms can be written Uses(to)

=Wscs(to)exp[-itoTo]F(~Pr,Str)F((bs,Sts) "Ps~s (2) Here F(~br,St r) describes a splitting operator beneath a receiver estimated locally by S-wave splitring of deep events. F(qbs,Sts) represents splitting of the ScS phase within the near-source region. We apply F-'(~br,St r) to the horizontal components of ScS to estimate differential splitting in ScS relative to S, and parameterize any residual splitting in ScS as F(q~s,Sts) in the same way as mentioned above. From the results of ((h,St) for the ScS wave at KTJ and INU we can judge whether an ScS wave passes through an anisotropic region on the source side or not. If the splitting parameter (4,,8t) of ScS

waves is consistent with that of S waves, we cannot recognize anisotropy near the source region. The observed elliptical particle motion of ScS waves is explained simply by anisotropy beneath receivers. On the other hand, if the splitting parameter ((h,St) of the ScS wave differs clearly from that of S waves - - f o r example, we have a large 8t of 1.75s for Event 6, with 4) nearly east-west, a small 8t for Event 13 for KTJ, and a large 8t of 1.5 s for Event 2 at INU--then anisotropy near the source region, as well as beneath receivers, should be considered. From the above criteria, we suggest the existence of source side anisotropy for Events 2, 6 and 13. To represent local anisotropy beneath KTJ and INU we adopt the typical pair of ((hr,6to reported in this study and by AN83, i.e. (15 °, 0.8 s) for KTJ and (110 °, 0.6s) for INU. The source side splitting parameters (4,s,Sts) of Events 2, 6 and 13 are listed in Table 5. Fig. 9 shows the waveform and its particle motion in the original form, corrected after removing the anisotropic effect beneath the receiver F(~r,~tr),

246

Y. Hiramatsu, M. Ando / Physies of the Earth and Planetary Interiors 95 (1996) 237-250

Table 5 The splitting parameters in the source region estimated by removing the splitting beneath the receiver, for KTJ and INU No. STC 4) 8t

1,4 6

KTJ INU

11

13

INU KTJ INU

1 4

126 117 105

1.,

(a)1989/06/16 SOUTH OF HONSHU, JAPAN !ORIGINAL

+s --

i

/\

INU

I!

1.15 1.t

~

0.2 0.9 0.6

~

/-~,

/ ~ }' ( \

/"J ~

~

~ 1

N

~ a s (s)

CORRECTED N- - -

and reconstructed after removing the source side anisotropy F(~bs,~ts). The splitting parameters beneath TVI" and HKJ are unknown. However, using the splitting operator on the source side, F-=(~bs,8 t s), estimated from KTJ and INU, we can obtain the splitting operator F(qbr,St r) beneath these receivers. Then applying this /"-l(t~r,~t r) to other events, F(4,s,Sts) can be estimated for TYr or HKJ. The splitting parameters of the source side (~bs,~ts) are consistent with (~b,$t) of T I T or HKJ for Events 2, 11 and 13 in Table 4. This fact leads us to conclude that a small anisotropic region exists beneath TYI" or HKJ because applying /-'-=((~s,~ts) shows a null result for -F(~br,+tr) of T I T and HKJ. We obtain the null splitting parameter beneath HKJ and TTT from Events 3 and 8. Comparing splitting parameters obtained from Event 8 with those obtained by AN83, the values th at KTJ and INU are close to those of the previous study. How-

E

f~

' • ~=110

':sw--

m=0.6s N

~0 ,, +o uTIME ~00 (s)~o +,, +~°+..... .~co~mue~D /\ ~'.4' ; N-S-I ~ w -/ \ ~ 5t=~.ls // t ,_.~ _ ~ 1 ~ ~ 0 N

~.~t,~

E

_

E

.o

(b)1993/lO/11 SOUTH OF HONSHU, JAPAN

~,.at

KTJ

[~ ~'" /

~-w-I -- ~ /

~ N

ever, ~t is smaller. The signal-to-noise ratio o f ScS waves is lower than that o f S waves in general. We

suspect that contamination by noise results in the discrepancy in St. In fact, a large discrepancy remains between the estimated F(~bs,St s) of KTJ and INU even though we assume that source side anisotropy causes little difference in ~t between ScS-wave splitting and S-wave splitting, and the estimated (~bs,Sts) cannot explain the null result of TTT. We therefore consider there is no anisotropy

, ,, +~ t, +=0+,~ +,.0+~,+,, o0 "nine (s) :o..~emD /~ ++=~5" N-S--I

E

'

W

,/~t-

-I

8t=0.8s

]~ ~

E

I

~

~--~. N

~o ~, ~0 ~0 ~0~+=o,o ,~0 ~o =+ •rmE (s)

RECONSTRUCTED

Fig. 9. Examples of waveform and particle motion of ScS w a v e s : upper, original waveform and panicle motion; middle, corrected waveform and particle motion after removing splitting caused by anisotropy beneath receiver;, lower, reconstructed waveform and particle motion after removing anisotropic effect in source region. The hatched area shows the time window of the particle motion.

E

/\

a

I = - /i/,

~

f

1

'- 117"

t

J N

"rIMs (=)

E

J

Y. Hiramatsu, M. Ando / Physics of the Earth and Planetary Interiors 95 (1996) 237-250

near the source region of Event 8, and as a result F(qbr,Btr) beneath TTT is also null. We also obtain the null splitting parameter beneath HKJ for Event 3 because applying F-l(~br,Btr) for KTJ shows a null result for F(qbs,Bts). The linear particle motion of some ScS phases (Events l, 3, 7, 8, 9 and 12) also supports the null splitting parameter beneath TTF and HKJ. Therefore we can directly take the pair (~b,Bt) for Events 4 and 8 listed in Table 4 as the source side splitting parameter (~bs,Sts) for TTT and HKJ. In summary, we recognize the existence of source side anisotropy for Events 2, 4, 6, 11 and 13, and less anisotropy, or none, at Events 1, 3, 5, 7, 8, 9, l0 and 12. The consistent size of (~b,Bt) of S waves with that of ScS waves for Events 3, 5, 7, 8 and l0 also implies that the anisotropic region beneath KTJ and INU is restricted within the mantle wedge above the subducting slab.

7. Origin of source side anisotropy Fig. l0 shows the source side anisotropy plotted at epicenters together with fossil and current directions of the Pacific plate motion. We recognize some events showing principal axes of anisotropy parallel

~ ~ , ~ ~ ~ • 4S'N-

• aSON

o

2S°N

125"[

~

135°E

145"E

155°E

Fig. 10. The direction and scale of source side anisotropy. Solid arrows represent the direction of the current plate motion predicted by Minster and Jordan (1978) and dashed arrows the direction of the fossil motion of the Pacific plate. Other symbols are the same as for Fig. 8.

247

to the direction of the fossil plate motion, but others are parallel to the current direction. The arrival time difference between the two polarized waves is 0.51.0s for all the cases. Assuming ScS-wave splitting is mainly caused by the preferred orientation of olivine, with a 5% velocity difference between fast and slow shear waves, the anisotropic region is required to be 50-100kin in size. Source side anisotropy is recognized for five of the 13 events, which implies a nonuniform distribution of the anisotropic region, similarly to receiver side anisotropy. Two locations can be considered for anisotropy near the source: (1) within the subducting slab; (2) in the mantle beneath the slab. Previous researchers considered the source side anisotropy to be located in the asthenosphere beneath the subducting slab (e.g. Kaneshima and Silver, 1992). In particular, Russo and Silver (1994) interpreted the observed trench-parallel 4) beneath South America as being caused by trench-parallel flow beneath the subducting Nazca plate. The average qb of the present study in the Izu-Bonin subduction zone also implies the polarization to be parallel to the trench rather than perpendicular to it. A difference exists between the strike of the trench axis and the direction of plate motion in both the Nazca and the Izu-Bonin subduction zones. In the Izu-Bonin region the direction of plate motion does not coincide with the trench-normal or trench-parallel direction. The lineations of magnetic anomaly are also NNW-SSE in the IzuBonin region (Nakanishi et al., 1992), and the strike of the trench is parallel to the fossil plate motion. If the ScS-wave splitting is due to anisotropy produced by large-scale mantle flow related to the subduction of a slab, such spatial variation in ~b would not occur in the asthenosphere. We cannot expect any abrupt change in direction of the mantle flow beneath the subducting slab within such small areas if the mantle flow is controlled by the subducting slab. Therefore we reject the former possibility and propose that anisotropy possibly exists in the subducting slab. We can recognize the difference in the direction of anisotropy depending on source depth, N W - S E

fast for shallower events and WNW-ESE fast for deeper o n e s . In oceanic lithosphere the fast direction of Pn waves is parallel to the spreading direction (e.g.

248

Y. Hiramatsu, M. A ndo/Physics of the Earth and Planetary Interiors 95 (1996) 237-250

I.v

~_._/~-v.___~'v-

~ONI( /'~ ~ ~ 30oN

28°N

\ T"-__

~/~'--~ \ •, /j~~-----

'~

/':'~k ~

28°N \ I~*E l~*E 14(10E '14~E 144*E Fig. 11. The horizontalprojection of P-axes of deep earthquakes in the Izu-Bonin subcluetion zone predicted by CMT solutions. Solid lines show the plate boundaries between the Eurasian plate and the Pacific and the Philippine Sea plates.

Hess, 1964). In the northwest Pacific the olivine a-axis is oriented parallel to the fossil plate motion, in a NNW-SSE direction, which is perpendicular to the lineations of the magnetic anomaly (Shimamura et al., 1983). This preferred orientation of olivine in the oceanic lithosphere can be preserved under the pressure and temperature conditions above 400km in the subducting slab (McKenzie, 1979; Babuska and Cara, 1991). Therefore we consider that fossil anisotropy is preserved in the subducting slab at shallow depths as shown for Event 6 of 157km depth and Event 11 of 289km depth. However, the 4) of Event 13 of 361 km is close to the direction of the current plate motion. This difference of 4) between Event 11 and 13 cannot be explained if the source side anisotropy is caused by the preserved preferred orientation within the slab. We can conclude that the change in the preferred orientation occurs within the subducting slab. Around the 400kin discontinuity the focal mechanisms of deep events exhibit dominantly down-dip compression (Burbach and Frohlich, 1986). Fig. 11 shows the P-axes from the CMT solutions projected at the surface. This direction is similar to the fast direction of anisotropy. Assuming this anisotropy is caused by the preferred orientation of olivine, this relation between the P-axis and anisotropy suggests that the olivine a-axis is oriented along the compres-

sion direction. The relation of preferred orientation with foliation and lineation of xenoliths shows that olivine has three types of preferred orientation: (1) b-axis perpendicular to foliation; (2) c-axis perpendicular to foliation; (3) a-axis perpendicular to foliation. The preferred orientation with a-axis perpendicular to foliation is observed especially in xenoliths originating at about 300 km depth. This depth is greater than for xenoliths with b-axis or c-axis perpendicular to foliation (Fujimura, 1986). However, many experimental and numerical studies have reported that the preferred orientation of olivine is controlled by the finite strain, and [100] and [010] axes can become aligned with the longest and shortest axes of the finite strain ellipsoid, respectively (McKenzie, 1979; Toriumi, 1984; Ribe and Yu, 1991; W e n k et al., 1991). Experimental studies can be performed only under restricted conditions. Therefore the mechanism of preferred orientation of olivine with the a-axis perpendicular to foliation is a further problem. We suggest that local shear stress in the subducting slab causes the reorientation of olivine under the reliable temperature and pressure conditions. Sugi et al. (1989) pointed out that deep events in the south of Honshu, which occurred near Events 11 and 13 and also had a source mechanism simi!~ to them, have a horizontal fault plane. Deep:events generally have a tendency in the preferred orientation of fault planes depending on focal depth, indicating the existence of weak planes within a slab (Sugi et al., 1989). The strain field in the slab is such that the tensional axis is horizontal and the compressional axis vertical. Assuming the fault plane is horizontal in this region, olivine crystals can have a preferred orientation with the a-axis parallel to the slip direction. Another possible mechanism for the change in preferred orientation of minerals is as follows. The phase change of olivine to modified spinel occurs at depths near 400 km, but this process may destroy otherwise preserved fossil orientation and leads to reorientation of the minerals. The source depth of 361 km for Event 13 is a little shallower than the depth of the phase transition, 400km. However, within a cold slab the phase transformation of olivine can occur at 70-80 km above the 400 km discontinuity (Ito and Katsura, 1989) because of the positive

Y. Hiramatsu, M. Ando / Physics of the Earth and Planetary Interiors 95 (1996) 237-250

slope of d P / d T in the a to /3 transformation, Modified spinel (/3) possesses weak anisotropy compared with a olivine, but is anisotropic enough to account for the observed splitting. Spinel possesses less anisotropy than both olivine and modified spinel. However, Fujimura (1984) considered that preferred orientation occurs in spinel on the basis of experiments using a material analogous to spinel, The expected shear-wave splitting derived from his study of 3% fits the observed splitting of ScS waves in the present study. He also found that spinel has bimodal preferred orientation of the a-axis parallel to both maximum and minimum pressure axes. The orthogonal direction of the first polarization to the pressure axis for Event 4 at 496 km depth is probably caused by the bimodal preferred orientation of spinel crystals.

8. Conclusion Shear waves with nearly vertical incident angles are useful to investigate seismic anisotropy in the upper mantle. In subduction zones the direct S and ScS waves are often used in studies of shear-wave splitting. Since 1988, broad-band digital data have been available in Japan and sufficient amounts of high-quality data have been stored for splitting study, We analyzed ScS waves of 13 deep earthquakes in the Kuril and Izu-Bonin subduction zones. We recognized discrepancies in splitting parameters, and in the orientation and the degree of anisotropy, among ScS waves of the 13 events observed at six stations. This discrepancy is caused by the ScS wave sampling anisotropy of the mantle twice along its path: first going down from the source and second coming up from the core-mantle boundary. To restrict its anisotropic region, we compared each ScS-wave splitting from different ray paths. We find a local anisotropic region defined by S-wave splitting just beneath central Honshu, Japan. The local anisotropy is detected on seismograms of all events. The change in the direction and degree of anisotropy is found in ScS waveforms, which indicates that this change results from anisotropy lying not on the receiver side but on the source side. We also recognized the change in anisotropy on the source side varying with depth. Shallow events above the or-to-~3 transition

249

show the first polarized wave direction parallel to the fossil spreading direction of the Pacific plate. In contrast, deep events show the direction of fast polarized waves parallel to the current plate motion direction and the pressure axes from CMT solutions. Two possibilities are likely for the location of the anisotropy: one within the subducting slab and the other in the asthenosphere just beneath the slab. In the latter case, however, the observed first polarization of anisotropy being parallel to the fossil plate motion cannot be explained if the induced flow by the current plate motion forms the preferred orientation of olivine in the asthenosphere. Therefore the observed anisotropy is mainly distributed within the subducting slab. Above the depth of a-to-~3 transition the fossil anisotropy is preserved within the slab. Below the transition depth the reorientation of olivine crystal accompanied by the phase transformation can alter the direction of preferred orientation under the stress within the slab. A description of the detailed mechanism for the preferred orientation associated with the phase change remains a future problem. We have clarified the location of the anisotropic region in the present study. The anisotropic regions of the subduction zone are distributed as patches with a possible dimension of 100km or so around the sources, and in addition some exist in the mantle wedge beneath the receivers.

Acknowledgements The authors thank the staff members of the Center of Earthquake Prediction and Volcanic Eruption Prediction of Tohoku University, the Earthquake Research Institute of the University of Tokyo, Nagoya University, the Building Research Institute and the National Research Institute for Earth Science and Disaster Prevention for allowing us to use the STS data for the present study. This study was partly supported by the JSPS Fellowship for Japanese Junior Scientists.

References Anao, M., 1984. ScS polarization anisotropy around the Pacific Ocean. J. Phys. Earth, 32: 179-195.

250

Y. Hirarnatsu. M. A ndo/Physics of the Earth and Planetary Interiors 95 (1996) 237-250

Ando, M., 1986. Mantle diapirs observed in the seismic window (in Japanese). Bull. Volcanol. Soc. Jpn., 31: 45-53. Ando, M., Ishikawa, Y. and Wada, H., 1980. S-Wave anisotropy in the upper mantle under a volcanic area in Japan. Nature, 268: 43-46. Ando, M., Ishikawa, Y. and Yamazaki, F., 1983. Shear-wave polarization anisotropy in the upper mantle beneath Honshu, Japan. J. Geophys. Res., 88: 5850-5864. Babuska, V. and Cara, M., 1991. Seismic Anisotropy in the Earth. Kluwer, Dordrecht. Bowman, J.R. and Ando, M., 1987. Shear-wave splitting in the upper-mantle wedge above the Tonga subduction zone. Geophys. J. R. Astron. Soc., 88: 25-41. Burbach, G.V. and Frohlich, C., 1986. Intermediate and deep seismicity and lateral structure of subducted lithosphere in the circum-Pacific region. Rev. Geophys., 24: 833-874. Fischer, K. and Yang, X., 1994. Anisotropy in Kuril-Kamchatka subduction zone structure. Geophys. Res. Lett., 21: 5-8. Fujimura, A., 1984. Preferred orientation of silicate spinel inferred from experimentally deformed aggregates of trevorite. J. Phys. Earth, 32: 273-297. Fujimura, A., 1986. Preferred orientation of mantle minerals. In: S. Karato and M. Toriumi (Editors), Rheology of Solids and of the Earth. Tokai University Press, Tokai, pp. 202-2t8. Fukao, Y., 1984. Evidence from core-reflected shear waves for anisotropy in the earth's mantle. Nature, 309: 695-698. Hess, H.H., 1964. Seismic anisotropy of the uppermost mantle under oceans. Nature, 203: 629-631. Hirahara, K., lkami, A., Ishida, M. and Mikumo, T., 1989. Three-dimensional P-wave structure beneath central Japan: low-velocity bodies in the wedge portion of the upper mantle above high-velocity subducting plates. Tectonophysics, 163: 63-73. lto, E. and Katsura, T., 1989. A temperature profile of the mantle transition zone. Geophys. Res. Lett., 16: 425-428. Kaneshima, S., t990. Origin of crustal anisotropy: shear wave splitting studies in Japan. J. Geophys. Res., 95:11127-11133. Kaneshima, S. and Silver, P.G., 1992. A search for source side mantle anisotropy. Geophys. Res. Lett., 19: 1049-1052. McKenzie, D., 1979. Finite deformation during fluid flow. Geophys. J. R. Astron. Soc., 58: 689-715. Minster, J.B. and Jordan, T.H., 1978. Present-day plate motions. J. Geophys. Res., 83: 5331-5354. Nakanishi, M., Tamaki, K. and Kobayashi, K., 1992. Magnetic anomaly lineations from late Jurassic to early Cretaceous in

the west-central Pacific Ocean. Geophys. J. Int., 109: 70l719. Ribe, N.M., 1989. Seismic anisotropy and mantle flow. J. Geophys. Res., 94: 4213-4223. Ribe, N.M. and Yu, Y., 1991. A theory for the evolution of orientation textures in deformed olivine polycrystals. J. Geophys. Res., 96: 8325-8335. Russo, M. and Silver, P.G., 1994. Trench-parallel flow beneath the Nazca plate from seismic anisotropy. Science, 263:11051111. Savage, M.K., Silver, P.G. and Meyer, P.P., 1990. Observations of teleseismic shear-wave splitting in the Basin and Range from portable and permanent stations. Geophys. Res. Lett., 17: 21-24. Shimamura, H., Inatani, H., Asada, T., Suyehiro, K. and Yamada, T., 1983. Longshot experiments to study velocity anisotropy in the oceanic lithosphere of the Northwestern Pacific. Phys. Earth Planet. Inter., 31: 348-362. Silver, P.G. and Chan, W.W., 1988. Implications for continental structure and evolution from seismic anisotropy. Nature, 335: 34-39. Silver, P.G. and Chart, W.W., 1991. Shear wave splitting and subcontinental mantle deformation. J. Geophys. Res., 96: 16429-16454. Sugi, N., Kikuchi, M. and Fukao, Y., 1989. Mode of stress release within a subducting slab of lithosphere: implication of source mechanism of deep and intermediate-depth earthquakes. Phys. Earth Planet. Inter., 55: 106-125. Toriumi, M., 1984. Preferred orientation of olivine in mantle-derived peridotites and stress in the lithosphere. J. Phys. Earth, 32: 259-271. Vinnik, L.P. and Kind, R., 1993. Ellipticity of teleseismic S-particle motion. Geophys. J. Int., 113: 165-174. Vinnik, L.P., Kind, R., Kosarev, G.L. and Makeyeva, L.I., 1989. Azimuthal anisotropy in the lithosphere from the observations of long-period S waves. Geophys. J. Int., 99: 549-559. Vinnik, L.P., Makeyeva, L.I., Milev, A. and Usenko, A.Yu., 1992. Global patterns of azimuthal anisotropy and deformations in the continental mantle. Geophys. J. Int., 111: 433-447. Wenk, H.R., Bennett, K., Canova, G.R. and Molinari, A., 1991. Modeling plastic deformation of peridotite with the self-consistency theory. J. Geophys. Res., 96: 8337-8349. Xie, J., 1992. Shear-wave splitting near Guam. Phys. Earth Planet. Inter., 72:211-219.