• Aucun résultat trouvé

Systematic variation in anisotropy beneath the mantle wedge in the Java-Sumatra subduction system from shear-wave splitting

N/A
N/A
Protected

Academic year: 2021

Partager "Systematic variation in anisotropy beneath the mantle wedge in the Java-Sumatra subduction system from shear-wave splitting"

Copied!
49
0
0

Texte intégral

(1)

HAL Id: hal-00610669

https://hal.archives-ouvertes.fr/hal-00610669

Submitted on 23 Jul 2011

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de

Systematic variation in anisotropy beneath the mantle wedge in the Java-Sumatra subduction system from

shear-wave splitting

J.O.S. Hammond, J. Wookey, S. Kaneshima, H. Inoue, T. Yamashina, P.

Harjadi

To cite this version:

J.O.S. Hammond, J. Wookey, S. Kaneshima, H. Inoue, T. Yamashina, et al.. Systematic varia- tion in anisotropy beneath the mantle wedge in the Java-Sumatra subduction system from shear- wave splitting. Physics of the Earth and Planetary Interiors, Elsevier, 2010, 178 (3-4), pp.189.

�10.1016/j.pepi.2009.10.003�. �hal-00610669�

(2)

Accepted Manuscript

Title: Systematic variation in anisotropy beneath the mantle wedge in the Java-Sumatra subduction system from

shear-wave splitting

Authors: J.O.S. Hammond, J. Wookey, S. Kaneshima, H.

Inoue, T. Yamashina, P. Harjadi

PII: S0031-9201(09)00213-1

DOI: doi:10.1016/j.pepi.2009.10.003

Reference: PEPI 5211

To appear in: Physics of the Earth and Planetary Interiors Received date: 6-7-2009

Revised date: 3-10-2009 Accepted date: 13-10-2009

Please cite this article as: Hammond, J.O.S., Wookey, J., Kaneshima, S., Inoue, H., Yamashina, T., Harjadi, P., Systematic variation in anisotropy beneath the mantle wedge in the Java-Sumatra subduction system from shear-wave splitting, Physics of the Earth and Planetary Interiors (2008), doi:10.1016/j.pepi.2009.10.003

This is a PDF file of an unedited manuscript that has been accepted for publication.

As a service to our customers we are providing this early version of the manuscript.

The manuscript will undergo copyediting, typesetting, and review of the resulting proof

before it is published in its final form. Please note that during the production process

errors may be discovered which could affect the content, and all legal disclaimers that

apply to the journal pertain.

(3)

Accepted Manuscript

Systematic variation in anisotropy beneath the mantle

1

wedge in the Java-Sumatra subduction system from

2

shear-wave splitting

3

J. O. S. Hammond ∗,a,b , J. Wookey a,c , S. Kaneshima b , H. Inoue d , T.

4

Yamashina d , P. Harjadi e

5

a Department of Earth Sciences, University of Bristol, Bristol, UK.

6

b Department of Earth and Planetary Sciences, Kyushu University, Hakozaki 6-10-1,

7

Fukuoka 812-8581, Japan

8

c Department of Earth Sciences, University College London, Gower Street, London, UK

9

d Badan Meteorologi, Klimatologi dan Geofisika, Jakarta, Indonesia

10

e National Research Institute for Earth Science and Disaster Prevention, Tsukuba, Japan

11

Abstract

12

The tectonic context of south-east Asia is dominated by subduction. One such major convergent boundary is the Java-Sunda trench, where the Australian- Indian plates are being subducted beneath the Eurasian plate. We measure shear-wave splitting in local and teleseismic data from 12 broadband sta- tions across Sumatra and Java to study the anisotropic characteristics of this subduction system, which can provide important constraints on dynam- ical processes involved. Splitting in S-waves from local earthquakes between 75-300 km deep show roughly trench parallel fast directions, and with time- lags 0.1-1.3 s (92% ≤0.6 s). Splitting from deeper local events and SKS, however, shows larger time-lags (0.8-2.0 s) and significant variation in fast direction. In order to infer patterns of deformation in the slab we apply a hybrid modelling scheme. We raytrace through an isotropic subduction zone velocity model, obtaining event to station raypaths in the upper man-

*Manuscript

(4)

Accepted Manuscript

tle. We then apply appropriately rotated olivine elastic constants to various parts of the subduction zone, and predict the shear-wave splitting accrued along the raypath. Finally, we perform grid searches for orientation of de- formation, and attempt to minimise the misfit between predicted and ob- served shear-wave splitting. Splitting from the shallow local events is best explained by anisotropy confined to a 40 km over-riding plate with horizon- tal, trench parallel deformation. However, in order to explain the larger lag times from SKS and deeper events, we must consider an additional region of seismic anisotropy in or around the slab. The slab geometry in the model is constrained by seismicity and regional tomography models, and many SKS raypaths travel large distances within the slab. Models placing anisotropy in the slab produce smaller misfits than outside for most stations. There is a strong indication that inferred flow directions are different for sub-Sumatran stations than for sub-Javanese, with >60 change over ∼375 km. The former appear aligned with the subduction plate motion, whereas the latter are closer to perpendicular, parallel to the trench direction. There are significant differences between the slab being subducted beneath Sumatra, and that beneath Java: age of seafloor, maximum depth of seismicity, relative strength of the bulk sound and shear-wave velocity anomaly and location of volcanic front all vary along the trench. We speculate, therefore, that the anisotropy may be a fossilised signature rather than due to contemporary dynamics.

Key words: Seismic anisotropy; Subduction zones; Indonesia; S-wave

13

splitting

14

(5)

Accepted Manuscript

1. Introduction

15

At the Java-Sunda trench the Indian-Australian plates are being sub-

16

ducted beneath the Eurasian plate. This is one of the most seismically ac-

17

tive regions in the world, with earthquakes occurring from the surface to the

18

base of the transition zone. This makes the region an ideal natural labora-

19

tory to study subduction dynamics using seismic imaging techniques. Here,

20

we present a study of seismic anisotropy (the variation of seismic wavespeed

21

with direction) beneath the Java-Sunda subduction system, providing some

22

insights into the deformation history and active processes in a subduction

23

setting.

24

As a seismic shear-wave propagates through an anisotropic medium it is

25

split into two quasi shear-waves, one with particle motion in the direction of

26

fast seismic wavespeed, and another perpendicular to this. This results in

27

the two shear-waves travelling at different speeds and arriving at a seismic

28

station separately; a phenomenon called shear-wave splitting (see Crampin

29

and Booth, 1985; Silver and Chan, 1991). At a 3-component seismic station

30

it is possible to measure these separate shear-waves and determine the fast

31

direction of wave polarisation and the time between the shear-waves (see

32

Silver, 1996; Savage, 1999; Kendall, 2000, for reviews).

33

In many cases anisotropy in the upper mantle is assumed to be caused

34

by the lattice preferred orientation (LPO) of olivine. Typically, the olivine

35

fast axis (a-axis) aligns in the direction of upper-mantle flow (e.g., Babuska

36

and Cara, 1991; Mainprice et al., 2000; Mainprice, 2007). As a result shear-

37

wave splitting can provide direct information about dynamic processes, such

38

as mantle flow, and accumulated strain due to previous deformation events

39

(6)

Accepted Manuscript

which have ‘frozen’ a source of anisotropy into the lithosphere beneath a

40

seismic station.

41

This technique has been applied to many subduction zone settings around

42

the world, both using local slab related events, and teleseismic events (e.g.,

43

Fouch and Fischer, 1996; Long and Silver, 2008). The results are far from sim-

44

ple with fast directions appearing trench parallel, perpendicular and oblique,

45

often at the same subduction zone [e.g., New Zealand, (Morley et al., 2006),

46

Japan (Nakajima and Hasegawa, 2004), South America (Helffrich et al., 2002)

47

and the Carribean (Pinero-Felicangeli and Kendall, 2008)]. This suggests a

48

complex region of anisotropy and many different mechanisms have been in-

49

voked to explain these results.

50

Deformation in the overriding plate and return flow in the mantle wedge

51

are discussed as causing trench parallel and perpendicular fast directions

52

for New Zealand (Morley et al., 2006). The presence of water has been

53

suggested to change the alignment mechanism of olivine, thus producing

54

fast directions perpendicular to the direction of maximum stress (Jung and

55

Karato, 2001) and this has been invoked to explain trench parallel splitting

56

beneath Japan (Nakajima and Hasegawa, 2004), although the authors also

57

acknowledge that trench parallel flow, caused by along strike dip variations

58

might also explain these results. Evidence for deeper sources of anisotropy

59

has also been described to explain splitting seen in teleseismic shear-wave ar-

60

rivals, for example slab anisotropy in New Zealand (Brisbourne et al., 1999),

61

sub-slab anisotropy due to toroidal flow beneath the slab in the Carribean

62

(Pinero-Felicangeli and Kendall, 2008) and basal drag beneath the slab in

63

South America (Helffrich et al., 2002).

64

(7)

Accepted Manuscript

Some attempts have been made to explain subduction zone shear-wave

65

splitting observations with uniform global models. Long and Silver (2008)

66

compiled global observations of shear-wave splitting at subduction

67

zones. They postulated that the largely trench parallel splitting

68

observations which arise from sub-mantle wedge regions (i.e. slab

69

or sub-slab) are dominated by lateral flow in the sub-slab region.

70

This is induced by return flow due to trench migration. Long and

71

Silver (2008) explain mantle wedge anisotropy (above the slab) as

72

being due to the interaction of lateral flow due to trench migration

73

and corner flow driven by viscous coupling between the slab and

74

the wedge. Such corner flow would induce a trench perpendicular fast

75

direction in the splitting. The authors argue that the observed supra-slab

76

trench parallel or trench perpendicular splitting results observed globally are

77

due to the dominance of one or the other of these mechanisms.

78

Another model invoked to explain global observations of subduction zone

79

anisotropy suggests preferentially-oriented hydrated faults in the subducting

80

slab (Faccenda et al., 2008). During the subduction process, faulting in the

81

downgoing plate allows water to migrate into the crust and mantle. This

82

results in the formation of sheets of silicates in the top few kms of the slab,

83

inducing a shape preferred orientation, resulting in high degrees of anisotropy.

84

Superimposed on this, an alignment of serpentinized rocks will cause an

85

additional crystallographic preferred orientation, thus giving the potential

86

for large degrees of shear-wave splitting from a thin layer at the top of the

87

slab (Faccenda et al., 2008).

88

The variability in shear-wave splitting results, and the wide variety of

89

(8)

Accepted Manuscript

mechanisms described to explain them shows that subduction zone settings

90

are very complicated. Possible regions of anisotropy exist in the crust, litho-

91

sphere and mantle wedge, and in or below the slab. In this study we try to

92

understand the problem in more detail by forward modelling our shear-wave

93

splitting results with realistic subduction zone models.

94

2. Data and methodology

95

We use data from 9 stations of the JISNET (Japan Indonesia Seismic

96

Network) array (Ohtaki et al., 2000) to determine the anisotropic character-

97

istics beneath Sumatra and Java. The stations were deployed in 1998 and

98

some stations are still recording. We use data up until the end of

99

2003. We also use 2 stations from the GEOFON array (UGM, 9 years, GSI,

100

3 years) and 1 from the Ocean hemisphere network project (PSI, 13 years),

101

a station used by Long and Silver (2008) to investigate subduction zone

102

shear-wave splitting. Station locations are shown in Figure 1.

103

We employ earthquakes with distances between 85 -140 , events ideal for

104

SKS/SKKS-wave splitting analysis. A total of 47 earthquake-station paths

105

provided SKS/SKKS-phases of sufficient quality to enable splitting to be

106

estimated (2% of data used). All SKS/SKKS data were filtered

107

between 0.05-0.3 Hz. We also measure splitting in S-waves from local

108

earthquakes where the incidence angle is less than 45 [i.e., within the shear-

109

wave window (Evans, 1984)]. Earthquake locations from the catalogue of

110

Engdahl et al. (1998) are used where possible, others are taken from the

111

USGS catalogue, which could introduce some significant mislocation

112

errors into our results. This results in 127 measurements of splitting

113

(9)

Accepted Manuscript

from local earthquakes (49% of data used). Individual results are shown

114

in Tables S1-S4. All local earthquake data were filtered between 0.1-

115

1 Hz. The overlap between SKS/SKKS- and local S-wave filter bands helps

116

to minimise frequency dependent effects between the two data sets.

117

We estimate shear-wave splitting in both the core phases and S-phases

118

from local earthquakes using the method of Teanby et al. (2004), which is

119

based on the methodology of Silver and Chan (1991). It uses a grid search

120

over fast direction (φ) and time-lag (δt), minimising the second eigenvalue

121

of the covariance matrix for the particle motion around a given time win-

122

dow. The advantage of the Teanby et al. (2004) method is that it automates

123

window selection, making 100 measurements around the relevant phase, and

124

applying cluster analysis to identify the most stable result. Figure 2 displays

125

a typical result of a local S- and teleseismic SKS-wave splitting result at PSI.

126

3. Results

127

From our initial analysis we reject results with 1σ errors greater than

128

0.4 s in δt and 30 in φ. Furthermore, the polarisation of the shear-wave af-

129

ter the anisotropy correction is applied (source polarisation, hereafter spol)

130

is compared with the back azimuth for SKS arrivals. For radially polarised

131

phases (i.e a P-S conversion at the core/mantle boundary) these should be

132

similar, thus we reject results with differences >30 . Local earthquakes are

133

not radially polarised as they have not been converted to a P-wave at any

134

point along their ray path. However some earthquakes are large enough to

135

have a previously determined focal mechanism (we use the Global CMT

136

and USGS catalogue), from which the initial polarisation can be calculated.

137

(10)

Accepted Manuscript

For these earthquakes the spol measurement and initial polarisation are com-

138

pared to provide an extra quality assessment (10% of local splitting mea-

139

surements included this check). Again spol and the initial polarisation

140

must agree within 30 to be included. Those results coming from earthquakes

141

without an available focal mechanism are assumed to be correct. Results are

142

shown in Figure 3 and Tables S1-S4. We discuss the results for Sumatra and

143

Java separately in the following sections.

144

3.1. Sumatra

145

3.1.1. Local S-wave splitting

146

Seismic stations BSI, PSI, PPI, KSI and KOTA are located approximately

147

above the 150 km contour of the subducting Indian plate (based on the slab

148

contours of Gudmundsson and Sambridge (1998)). As a result splitting is

149

estimated in shear-waves generated from earthquakes in the ∼100 km-200 km

150

depth range. One station, GSI, is located closer to the trench, and thus

151

provides estimates of splitting from earthquakes in the depth range ∼30-

152

80 km. All these splitting results show little variation with depth (Figure 4).

153

Time-lags for all stations, including GSI, are generally between 0.2-0.4 s, and

154

fast directions roughly trench, parallel, with some scatter, and parallel

155

to the strike of the Sumatran fault (Figures 3 and 4). This indicates that

156

the source of this anisotropy is in the over-riding plate and that the mantle

157

wedge above the slab is largely isotropic.

158

3.1.2. SKS/SKKS-wave splitting

159

The stations located above the 150 km slab contour (BSI, PSI, PPI, KSI

160

and KOTA) show ∼1 s of splitting, with a general north-south trend. How-

161

(11)

Accepted Manuscript

ever some variation in the fast direction is evident, particularly at PSI (Figure

162

3). The station close to the trench, GSI, is markedly different with a large

163

time-lag of 1.8 s and a trench perpendicular fast direction of 52.5 . The ob-

164

vious discrepancy between the time-lags for local and teleseismic splitting

165

suggests that a deeper source of anisotropy, within or beneath the slab, must

166

be present.

167

3.2. Java

168

3.2.1. Local S-wave splitting

169

Local earthquakes, within the shear-wave window, have been recorded

170

down to depths of 700 km at most stations in Java. Splitting results from

171

these events display considerable variation in both fast direction and time lag

172

on individual stations as well as from different stations (Figure 3). Most sta-

173

tions show both trench parallel and trench perpendicular fast directions. The

174

trench perpendicular results are in general from intermediate depth earth-

175

quakes (250-350 km), with both shallower and deeper earthquakes having

176

more trench parallel fast directions. Some deep events produce large delay

177

times and fast directions oblique to the strike of the trench (UGM, BRB).

178

These seem isolated to one region of the subduction zone, with adjacent

179

splitting measurements from nearby earthquakes showing much smaller delay

180

times, albeit with similar fast directions. Splitting at BMI is very compli-

181

cated, possibly because it lies above a part of the slab which bends beneath

182

Sulawesi.

183

(12)

Accepted Manuscript

3.2.2. SKS/SKKS-wave splitting

184

In general SKS-wave results have an east-west fast direction, with ∼1.6 s

185

of splitting (Figure 3). There does seem to be some subtle variation in the

186

splitting results with the estimated fast directions occurring in two groups,

187

∼75 and ∼105 . The magnitude of splitting is again larger than the splitting

188

seen in local events, except for a few deep earthquakes at UGM, and BRB.

189

Again, similarly to the results at Sumatra, this suggests a slab or sub-slab

190

source of anisotropy.

191

4. Modelling

192

It is evident that more than one region of anisotropy exists beneath Suma-

193

tra and Java. Shallow earthquakes (<300 km deep) beneath Sumatra and

194

Java show little variation in time-lag with depth, and show consistent trench

195

parallel fast directions. This suggests that the mantle wedge is isotropic, and

196

that the splitting observed from these earthquakes is accrued in the over-

197

riding plate. However larger time-lags observed in SKS/SKKS splitting at

198

Sumatra and Java, and splitting from deeper (>300 km deep) earthquakes

199

beneath Java suggest that anisotropy must exist in the slab, or sub-slab as

200

well.

201

To understand the anisotropy in more detail we have developed a tech-

202

nique based on ray-tracing through an inhomogeneous subduction zone model

203

(Guest and Kendall, 1993). We build a model consisting of four regions:

204

layer 1, the over-riding plate and mantle wedge; layer 2, a thin

205

10 km layer above the slab; layer 3, the subducting slab; and layer

206

4, the sub-slab mantle (Figure 5). The surface of the slab is calculated

207

(13)

Accepted Manuscript

using the regionalized upper mantle (RUM) seismic model slab contours

208

(Gudmundsson and Sambridge, 1998) which is consistent with a relocated

209

local seismicity study (Fauzi et al., 1996) and deep events from the USGS

210

earthquake catalogue. The model is also consistent with regional tomogra-

211

phy models for Indonesia (Gorbatov and Kennett, 2003). Uncertainty in

212

these slab models could introduce error into these results, how-

213

ever recently published slab models of Syracuse and Abers (2006)

214

show only minor changes (∼20 km shift east) in Sumatra-Java sub-

215

duction zone to those calculated by Gudmundsson and Sambridge

216

(1998).

217

We use ATRAK, a ray-tracer capable of tracking seismic rays in multi-

218

layered 3-D media with general anisotropy (Guest and Kendall, 1993). This

219

technique has been utilised to investigate anisotropy in other subduction

220

zones [e.g., Kurils (Kendall and Thomson, 1992) and New Zealand

221

(Brisbourne et al., 1999)]. Our model differs from previous studies by the

222

fact that we trace rays through an isotropic model, assuming that anisotropy

223

only minimally distorts the raypath (a reasonable assumption for relatively

224

weakly anisotropic media). We trace rays for each station separately (except

225

UGM and SWH which are very close together, and so are incorporated into

226

one model). We appeal to reciprocity and shoot rays from the station, satu-

227

rating the model (>12,000 rays per station, Figure 5). We then match each

228

event to its nearest ray, thus defining the raypath taken from the earthquake

229

to the station. For local earthquakes we use the latitude, longitude and

230

depth to match the ray, and for SKS/SKKS phases we calculate the piercing

231

point of the phase at a depth of 800 km beneath the station [through the

232

(14)

Accepted Manuscript

AK135 1-D velocity model (Kennett et al., 1995)], and use this information

233

as the event location (Figure 5). This ray-path for each station-event pair

234

is then broken down into a length, time and angle spent in each part of

235

the model. With this information we can then impose an anisotropy on a

236

specified region of the model (layers 1-4), represented by an appropriately

237

rotated elastic tensor (we use olivine elastic constants from Abramson et al.

238

(1997)). Anisotropy is confined to the olivine stability field (i.e. <410 km),

239

and can be restricted to a particular depth range within a layer.

240

For each contiguous section of the raypath we can use the Christoffel equa-

241

tion ( ˇ Cerven` y, 2001) to calculate the fast and slow shear-wave velocities and

242

particle motion, thus, along with the distance in the layer, we can calculate

243

a time-lag and fast direction through the layer.

244

We then perform a grid search over orientation (γ, ∆γ=5 ), and

245

anisotropic strength (by Voigt-Reuss-Hill dilution of the elastic

246

tensor; denoted volume fraction aligned - VFA, ∆VFA=0.01–0.05).

247

The angle γ is the orientation of the olivine a-axis in a plane which

248

is either horizontal (shallow models) or inclined at the average dip

249

of the subducting slab (deep models). The b-axis is held normal to

250

this plane. For each node in the grid we calculate δt and φ for each

251

event-station raypath in the data. The model is then evaluated

252

as the summed misfit between these and the real measurements

253

(for δt and φ separately, and for a combination). Typically, VFA

254

is poorly constrained, so is not interpreted in great detail, instead

255

we primarily analyse the best fitting orientation γ, illustrated by

256

the density rotorgram in Figure 6.

257

(15)

Accepted Manuscript

This can be performed for each layer in the model. Four models

258

are run in total, the first constraining the layer 1 anisotropy, and

259

the last three incorporate this result with layer 2, 3 and 4 respec-

260

tively. Multiple layers of anisotropy can be included, though only

261

one may vary, as a result we can not formely asses trade-off between

262

these layers. Where two layers of anisotropy are included we com-

263

bine the splitting operators using the n-layer splitting equations of

264

Silver and Savage (1994), and the initial source polarisation from

265

the splitting measurements. An example result is shown in Figure

266

6 for station GSI with anisotropy in layer 1 only.

267

The locations of stations and earthquakes used in this study are ideal to

268

image the various parts of the subduction system. We image paths in the

269

sub-slab, slab, supra-slab, mantle wedge and over-riding plate. This enables

270

us to go one step further than typical shear-wave splitting studies and place

271

constraints on the location of the anisotropy, and thus better understand the

272

dynamics of Java-Sunda subduction.

273

The modelling scheme outlined here allows us to extract information on

274

which part of the subduction system each event is sampling. Therefore, we

275

are able to investigate further the observation made in section 3, that the

276

mantle wedge appears largely isotropic. Figure 7 shows the variation in delay

277

time as a function of path length through various parts of the model. It is

278

evident that there is no clear relationship between splitting delay times and

279

path length in the mantle wedge. In fact, those data with paths almost

280

exclusively in the wedge display the smallest delay times (<0.4 s, except two

281

SKS events at PSI which, in our model, never sample the region below the

282

(16)

Accepted Manuscript

wedge). An estimate can be placed on the amount of anisotropy in a 40 km

283

thick over-riding plate and the mantle wedge below. Figure 7a shows a linear

284

fit to the data with paths exclusively in the wedge (removing the two SKS-

285

wave arrivals at PSI). The line has equation δt = 0.0007l + 0.1306, where l

286

is the path length in the mantle wedge and over-riding plate. Assuming an

287

average S-wave velocity of 4.5 km/s, this equates to an anisotropy of 1.8%

288

in a 40 km thick plate, and 0.3% anisotropy in the mantle wedge below this

289

plate. This shows that the anisotropy in the over-riding plate is dominating

290

the shear-wave splitting. In comparison, some correlation is evident between

291

splitting delay time and path length spent in the region beneath the mantle

292

wedge, but above 410 km (layers 2-4 in our model) (Figure 7b). The splitting

293

results can be explained by a model which has 20% aligned olivine crystals

294

along this path length.

295

This observation supports the idea that the mantle wedge is largely

296

isotropic, and that anisotropy is largely confined to the over-riding plate,

297

slab and sub-slab region. To further constrain the location of the anisotropy

298

we split the modelling into two steps, shallow and deep anisotropy. To con-

299

strain the characteristics of the shallow anisotropy we use events with depths

300

<300 km, and to model deeper anisotropy we use events with depths >300 km

301

and SKS-events.

302

4.1. Shallow anisotropy

303

We impose anisotropy in the uppermost 40 km of layer 1 of the

304

model, simulating the overriding plate lithosphere. The choice of this

305

thickness is fairly arbitrary (i.e. a layer half as thick with twice the anisotropy

306

strength will fit the data just as well), but we base our model on the limit

307

(17)

Accepted Manuscript

of our splitting results (the shallowest events at GSI are ∼30 km). In these

308

cases the olivine a-axis is horizontal and the misfit is calculated

309

only using events shallower than 300 km.

310

Table 1 and Figure 8 show that the best fitting anisotropy orientations at

311

all stations are well constrained and show trench parallel orientations. BRB

312

is not modelled as we have too few splitting measurements. Errors in depth

313

locations could have an effect on our estimated VFA, however we

314

expect this effect to be minimal due to the fact that we impose

315

a 40 km thick layer of anisotropy at the top of the model. Events

316

which are mislocated from below 40 km to above 40 km, and events

317

which are mislocated to below 40 km from above 40 km will cause

318

the VFA to be over and underestimated respectively. This is only

319

a problem at GSI, where event depths range from 30-80 km. At

320

other stations only one event occurs shallower than 70 km (SWH,

321

33 km).

322

4.2. Deep anisotropy

323

To fit local slab events with depths >300 km, and SKS/SKKS phases

324

we test 3 classes of model where anisotropy is confined to a thin

325

layer above the slab (layer 2), the slab (layer 3) or the sub-slab (layer

326

4). We impose an olivine anisotropy with its a-axis oriented (γ)

327

in a plane inclined at the average dip of the subduction zone (b-

328

axis is normal to the plane). In all deep models we also impose

329

the best fitting shallow anisotropy for each station (determined

330

above). Thus, each model has two anisotropic regions, one fixed in

331

orientation and strength (shallow; layer 1) and one varying (deep;

332

(18)

Accepted Manuscript

layers 2–4).

333

Results are shown in Figures 8 and 9 and Table 2, and it is evident

334

that it is hard to confidently discriminate between the slab and sub-slab

335

models (layer 3 and 4), with both giving compatible anisotropy orientations.

336

Anisotropy in a thin layer at the top of the slab (layer 2) consistently has the

337

largest misfit, and requires a very high degree of alignment, suggesting this is

338

a less likely cause of our observations (Figure 9 and Table 2). For both Java

339

and Sumatra the misfit is considerably lower for a model where anisotropy is

340

confined to the slab (assuming a thickness of 100 km). A robust feature in all

341

models (with the exception of GSI) is that the deeper anisotropy orientation

342

beneath Sumatra is more north-south, and rapidly changes at the Sunda

343

strait becoming more east-west beneath Java (Figures 8 and 9). GSI shows

344

a trench perpendicular anisotropy orientation, suggesting that a different

345

mechanism is causing anisotropy at the trench compared to that beneath the

346

volcanic/back-arc region (Figure 8). GSI has the longest path in the

347

sub-slab region. The anomolous trench perpendicular anisotropy

348

orientation, and large time delays at this station may suggest that

349

sub-slab anisotropy is dominant in the forearc. BRB and KHK do

350

not have enough observations to model. Mislocated events for deep and

351

SKS/SKKS-wave events will have minimal effects on their raypaths.

352

5. Discussion

353

From the modelling it is evident that anisotropy in the over-riding plate

354

is required to explain the splitting observations from events <300 km deep

355

at Java and Sumatra. The modelled anisotropy orientation is near paral-

356

(19)

Accepted Manuscript

lel to the strike of the trench. The simplest explanation for this anisotropy

357

is vertically aligned cracks associated with deformation in the over-riding

358

Eurasian plate, with a largely isotropic mantle wedge below. This observa-

359

tion is supported by a recent local earthquake tomography study for central

360

Java, where high P-wave anisotropy (7-10%) is observed in the crust and

361

lithospheric mantle (Koulakov et al., 2009). Similar observations of over-

362

riding plate anisotropy and isotropic mantle wedges have been seen at New

363

Zealand (Morley et al., 2006), the Carribean (Pinero-Felicangeli and Kendall,

364

2008) and South America (Polet. et al., 2000). The fact that the wedge is

365

isotropic suggests that either 2D corner flow caused by viscous coupling of

366

the slab and wedge is of similar magnitude as lateral flow from trench mi-

367

gration, or that they are both weak in this region, thus meaning no coherent

368

LPO can be developed (Long and Silver, 2008). This observation is in drastic

369

contrast to other subduction zones such as Japan, where Hiramatsu et al.

370

(1998) observe two 75-175 km deep anisotropic zones that are caused by melt

371

filled cracks, Tonga where Smith et al. (2001) suggest that along strike flow

372

in the mantle related to slab rollback and the Samoan plume is inducing an

373

anisotropy and the Aleutians where Long and Silver (2008) suggest that 2D

374

corner flow is causing a large anisotropy. In all these regions they observe

375

large δt ∼1-2 s, much larger than we observe beneath Java and Sumatra.

376

Models of deeper anisotropy, constrained by deep subduction events, and

377

SKS/SKKS show that a slab source of anisotropy is most likely, although a

378

sub-slab source can not be ruled out and gives similar best fitting anisotropy

379

orientations. However, in all these models a sharp change in anisotropic

380

characteristics is observed between Sumatra and Java (Figure 8). This change

381

(20)

Accepted Manuscript

is also evident in the raw data (Figure 10). SKS fast direction estimates for

382

the same event recorded at KOTA, the southernmost Sumatran station, and

383

LEM the westernmost Javan station vary by ∼60 over ∼375km (Figure 10).

384

This change in anisotropic characteristics is mirrored by changes in other

385

geophysical parameters close to the Sunda Strait. The most obvious change

386

near this location is in the age of the downgoing slab (Figure 1). Magnetic

387

anomalies show a change in age from ∼60Ma to ∼100Ma from southern

388

Sumatra to Western Java (Sdrolias and Muller, 2006). The sharpness, and

389

exact location of this boundary is unclear, especially at depth in the sub-

390

ducted slab. Syracuse and Abers (2006) observed that this age boundary

391

co-incides with a change in volcanic characteristics. They found that the

392

volcanic front in Sumatra is located above the 90 km slab contour, with an

393

abrupt change close to the Sunda Strait where volcanism is present above

394

the 150 km slab contour. This change occurs over <150 km, similarly to the

395

abrupt change in anisotropy, and also correlates with the abrupt reduction

396

in deep (>500 km) seismicity beneath Sumatra (e.g., Engdahl et al., 1998),

397

although no change in the shape of the slab is observed. Syracuse and Abers

398

(2006) also correlate this change in the location of the volcanic front with

399

variation in the ratio of potassium to silica (Wheller et al., 1987), but the

400

reasons for this are unclear (Syracuse and Abers, 2006). Regional tomogra-

401

phy profiles of Gorbatov and Kennett (2003) show an abrupt change in the

402

bulk sound speed in the slab close to the Sunda Strait further suggesting a

403

sharp change in the slab characteristics in this region. The possible causes of

404

the observed rapid change in anisotropy, with reference to the other changes

405

in geophysical characteristics are discussed further below.

406

(21)

Accepted Manuscript

5.1. Change in geometry

407

The Indian and Australian plates are subducting northwards beneath the

408

Java-Sunda arc. As a result subduction moves from normal to the trench

409

beneath Java to oblique subduction beneath Sumatra (most obviously char-

410

acterised by the Great Sumatran transform fault). This change in the geom-

411

etry of subduction across the region could potentially explain the change in

412

anisotropy observed in the splitting results.

413

The alignment of the deep anisotropy orientations at all stations beneath

414

Sumatra, except GSI, roughly matches with the APM of the subducting

415

slab and oblique to the orientation of the trench. This would suggest that

416

in this case shear associated with the subducting plate moving through the

417

mantle is aligning sub-slab olivine, and causing APM aligned anisotropy.

418

However, beneath Java, the opposite is true. The deep anisotropy orientation

419

is perpendicular to the APM of the subducting slab. The slab between

420

Sumatra and Java appears continuous (Syracuse and Abers, 2006), so it

421

seems unlikely that basal drag would create an anisotropy beneath southern

422

Sumatra, and not western Java <375 km away. Thus, basal drag does not

423

seem a likely cause of anisotropy.

424

Long and Silver (2008) suggest that lateral flow beneath the slab, pro-

425

duced by trench migration, can induce a trench parallel anisotropy orienta-

426

tion, and this is observed beneath Java. However, this would predict trench

427

parallel anisotropy orientations beneath Sumatra, and we observe trench

428

oblique orientations. Again, this change from lateral flow beneath western

429

Java, to an oblique flow beneath southern Sumatra seems unlikely over such

430

a short distance.

431

(22)

Accepted Manuscript

5.2. Water in the mantle

432

Jung and Karato (2001) have suggested that the presence of water can

433

make the b-axis of olivine align with the flow direction, manifesting in a

434

anisotropy orientation perpendicular to the direction of flow. Thus a change

435

in the amount of water beneath Java and Sumatra could induce a change

436

from APM parallel anisotropy orientations to APM normal anisotropy ori-

437

entations. The most likely region of water rich mantle is above the slab, in

438

the mantle wedge. We observe that the mantle wedge in this region is largely

439

isotropic, with little variation in time-lags with depth, suggesting that this

440

is an unlikely scenario to cause the sharp change in anisotropy.

441

5.3. Fossil anisotropy

442

Figure 5 shows that the deep and teleseismic events travel some distance

443

in the subducting slab, and thus anisotropy in the slab will manifest itself in

444

the observed splitting results. Interestingly, the sharp change in anisotropic

445

characteristics between Java and Sumatra co-incides with major changes in

446

slab characteristics.

447

The age of the subducting plate beneath Java and Sumatra varies between

448

100 Ma in the north to 43 Ma at 2 S and up to 160 Ma in the east (Sdrolias

449

and Muller, 2006). This transition in ages occurs near the region where we see

450

an abrupt change in anisotropy with the plate subducting beneath Sumatra

451

40-80 Ma and beneath Java 80-140 Ma. Other features of the slab are also

452

different between these regions. Seismicity beneath Sumatra only extends to

453

a depth of ∼300 km and as the slab bends beneath Java it extends to the base

454

of the transition zone and this change in seismicity is located approximately

455

at the same location as an abrupt change in the bulk sound speed (Gorbatov

456

(23)

Accepted Manuscript

and Kennett, 2003) and change in the location of the volcanic front (Syracuse

457

and Abers, 2006). The abrupt changes in age and seismic properties suggests

458

that differences between the Javan and Sumatran slabs may be the cause for

459

the abrupt change in anisotropic characteristics.

460

Faccenda et al. (2008) show that preferentially oriented hydrated faults in

461

the uppermost part of the slab can produce a high degree of shear-wave split-

462

ting. A change in age between the slab beneath western Java and southern

463

Sumatra may effect the efficacy of fracture generation in the slab, and thus

464

hydrated faults may change in style or quantity across this region. However,

465

our models indicate that a thin highly anisotropic layer at the top of the slab

466

poorly fits our results, mainly due to the fact that rays spend such a small

467

amount of time in this region. If this layer behaves as a strong waveguide, a

468

feature not modelled by our simple subduction models, then this may explain

469

the observed shear-wave splitting.

470

Another possible location for a fossil anisotropy is in the slab lithosphere.

471

As oceanic lithosphere is formed and moves over the asthenosphere it cre-

472

ates a large anisotropic signature (e.g. at the East Pacific Rise Wolfe and

473

Solomon, 1998; Harmon et al., 2004), in the direction of plate motion. These

474

anisotropic characteristics can be frozen into the lithosphere, and may re-

475

main present in the subducted slab beneath Java and Sumatra. This mech-

476

anism has been suggested to be responsible for at least part of the observed

477

shear-wave splitting seen at the Marianas (Fouch and Fischer, 1998). Also,

478

Brisbourne et al. (1999) show that significant anisotropy, with trench par-

479

allel anisotropy orientations, exists in the Hikurangi subduction zone, New

480

Zealand. The Pacific plate in this region has a trench parallel age progres-

481

(24)

Accepted Manuscript

sion [i.e. the age of the Pacific oceanic lithosphere changes in a direction

482

parallel to the trench (Sdrolias and Muller, 2006)]. This is similar to that

483

seen beneath Java, suggesting that a fossilised anisotropic signature, frozen

484

in as the plate formed, could explain anisotropy in this region. The oceanic

485

lithosphere subducting beneath Java is 80-140 Ma, and Lithgow-Bertelloni

486

and Richards (1998) show that plate motions for the Australian plate be-

487

tween 74-119Ma are generally east-west. The oceanic lithosphere subducting

488

beneath Sumatra is 40-80 Ma, and Lithgow-Bertelloni and Richards (1998)

489

show that the plate motions for the Indian plate at this time were north-

490

south. The shear-wave splitting change observed between Java and Sumatra

491

could, therefore, be an effect of sampling these differing frozen anisotropic

492

signatures.

493

6. Conclusion

494

Shear-wave splitting has been measured in SKS-wave arrivals and S-waves

495

from local subduction zone earthquakes beneath the Java-Sunda subduc-

496

tion zone. Results show that the mantle wedge is mainly isotropic, with

497

anisotropy largely confined to the over-riding plate. This has been fur-

498

ther emphasised with modelling, which shows a best-fitting trench-parallel

499

anisotropy orientation, suggesting anisotropy seen in splitting from local

500

earthquakes is derived from anisotropy caused by vertically aligned cracks as-

501

sociated with deformation in the over-riding Eurasian plate. Splitting from

502

deeper earthquakes, and SKS-wave arrivals have much larger delay times,

503

suggesting a deeper source of anisotropy is present.

504

Modelling constrains the location of this anisotropy, with the most likely

505

(25)

Accepted Manuscript

anisotropic region being the slab. The best fitting anisotropy orientations

506

in the slab were predominantly north-south beneath Sumatra (except GSI

507

which lies above the trench, and shows trench perpendicular anisotropy ori-

508

entationss, suggesting a different mechanism to that seen at the volcanic

509

arc/back arc). This change in orientation occurs over less than 375 km and

510

at the same location as changes in other geophysical characteristics such as

511

slab age (Sdrolias and Muller, 2006), bulk sound speed (Gorbatov and Ken-

512

nett, 2003) and location of the volcanic front (Syracuse and Abers, 2006),

513

suggesting that a change in the slab characteristics is the reason for the

514

change in anisotropy.

515

We speculate that the anisotropy has been frozen into the subducting

516

lithosphere as it was formed, or moved over the asthenosphere. The subduct-

517

ing lithosphere beneath Java is older, and would have formed with eastward

518

plate motions (Lithgow-Bertelloni and Richards, 1998), resulting in east-west

519

anisotropy orientations, compared with the younger lithosphere subducting

520

beneath Sumatra which formed when the Indian-Australian plate was mov-

521

ing north (Lithgow-Bertelloni and Richards, 1998), resulting in north-south

522

anisotropy orientations.

523

7. Acknowledgments

524

We would like to thank Mike Kendall and George Helffrich for valuable

525

discussion and help improving this paper. Also, we would like to thank Alexei

526

Gorbatov for the use of his tomographic models. All the data used in this

527

paper was provided by the Japan Indonesian Seismic Network (JISNET),

528

IRIS, GEOFON and the Ocean Hemisphere Network Project, courtesy of

529

(26)

Accepted Manuscript

the Earthquake Research Institute, University of Tokyo. This research was

530

funded by a Japan Society for the Promotion of Science (JSPS) postdoctoral

531

fellowship (short term), JSPS/FF1/367

532

References

533

Abramson, E., Brown, J., Slutsky, L., Zaug, J., 1997. The elastic constants

534

of San Carlos olivine to 17 GPa. J. Geophys. Res 102 (6), 12253–12263.

535

Babuska, V., Cara, M., 1991. Seismic Anisotropy in the Earth. Kluwer Acad.,

536

Norwell, Mass.

537

Brisbourne, A., Stuart, G., Kendall, J.-M., 1999. Anisotropic structure of

538

the Hikurangi subduction zone, New Zealand-integrated interpretation of

539

surface-wave andbody-wave observations. Geophys. J. Int. 137 (1), 214–

540

230.

541

Cerven` ˇ y, V., 2001. Seismic Ray Theory. Cambridge University Press.

542

Crampin, S., Booth, D. C., 1985. Shear-wave polarizations near the North

543

Anatolian Fault, II, Interpretation in terms of crack-induced anisotropy.

544

Geophys. J. R. Astr. Soc. 83, 75–92.

545

Engdahl, E., van der Hilst, R., Buland, R., 1998. Global teleseismic earth-

546

quake relocation with improved travel times and procedures for depth de-

547

termination. Bull. Seism. Soc. Am 88 (3), 722–743.

548

Evans, R., 1984. Effects of the free surface on shear wavetrains. Geophys. J.

549

R. Astr. Soc. 76 (1), 165–172.

550

(27)

Accepted Manuscript

Faccenda, M., Burlini, L., Gerya, T., Mainprice, D., 2008. Fault-induced

551

seismic anisotropy by hydration in subducting oceanic plates. Nature 455,

552

1097–1100.

553

Fauzi, McCaffery, R., Wark, D., Sunaryo, Haryadi, P. Y. P., 1996. Lateral

554

variation in slab orientation beneath Toba Caldera, northern Sumatra.

555

Geophys. Res. Lett. 23 (5), 443–446.

556

Fouch, M., Fischer, K., 1996. Mantle anisotropy beneath northwest Pacific

557

subduction zones. J. Geophys. Res. 101 (B7), 15987–16002.

558

Fouch, M., Fischer, K., 1998. Shear wave anisotropy in the Mariana subduc-

559

tion zone. Geophys. Res. Lett. 25 (8), 1221–1224.

560

Gorbatov, A., Kennett, B. L. N., 2003. Joint bulk-sound and shear tomogra-

561

phy for Western Pacific subduction zones. Earth Planet. Sci. Lett. 210 (3),

562

527–543.

563

Gripp, A. E., Gordon, R. G., 1990. Current plate velocities relative to the

564

hotspots incorporating the Nuvel–1 global plate motion model. Geophys.

565

Res. Lett. 17, 1109–1112.

566

Gudmundsson, O., Sambridge, M., 1998. A regionalized upper mantle(RUM)

567

seismic model. J. Geophys. Res. 103 (B4), 7121–7136.

568

Guest, W. S., Kendall, J.-M., 1993. Modelling seismic waveforms in

569

anisotropic inhomogeneous media using ray and maslov asymptotic the-

570

ory: Applications to exploration seismology. J. Can. Soc. Expl. Geophys,

571

78–92.

572

(28)

Accepted Manuscript

Harmon, N., Forsyth, D., Fischer, K., Webb, S., 2004. Variations in shear-

573

wave splitting in young Pacific seafloor. Geophys. Res. Lett. 31 (15).

574

Helffrich, G., Wiens, D. A., Vera, E., Barrientos, S., Shore, P., Robertson,

575

S., Adaros, R., 2002. A teleseismic shear-wave splitting study to investi-

576

gate mantle flow around South America and implications for plate-driving

577

forces. Geophys. J. Int. 149, F1–F7.

578

Hiramatsu, Y., Ando, M., Tsukuda, T., Ooida, T., 1998. Three-dimensional

579

image of the anisotropic bodies beneath central Honshu, Japan. Geophys.

580

J. Int. 135 (3), 801–816.

581

Jung, H., Karato, S., 2001. Water-induced fabric transitions in olivine. Sci-

582

ence 293, 1460–1463.

583

Kendall, J.-M., 2000. Seismic anisotropy in the boundary layers of the mantle.

584

In: Karato, S., Forte, A. M., Liebermann, R. C., Masters, G., Stixrude,

585

L. (Eds.), Earth’s Deep Interior: Mineral Physics and Tomography From

586

the Atomic to the Global Scale. No. 117. Geophysical Monograph, pp.

587

133–159.

588

Kendall, J.-M., Thomson, C. J., 1992. Seismic modelling of subduction zones

589

with inhomogeneity and anisotropy–I. Teleseismic P-wavefront tracking.

590

Geophys. J. Int. 112, 39–66.

591

Kennett, B. L. N., Engdahl, E. R., Buland, R., 1995. Constraints on seismic

592

velocities in the Earth from traveltimes. Geophys. J. Int. 122 (1), 108–124.

593

(29)

Accepted Manuscript

Koulakov, I., Jakovlev, A., Luehr, B. G., 2009. Anisotropic structure be-

594

neath central Java from local earthquake tomography. Geochem. Geophys.

595

Geosyst. 10 (2).

596

Lithgow-Bertelloni, C., Richards, M., 1998. The dynamics of Cenozoic and

597

Mesozoic plate motions. Rev. Geophys 36 (1), 27–78.

598

Long, M. D., Silver, P. G., 2008. The subduction zone flow field from seismic

599

anisotropy: A global view. Science 319, 315–318.

600

Mainprice, D., 2007. Seismic anisotropy of the deep Earth from a mineral

601

and rock physics perspective. Treatise in Geophysics 2, 437–492.

602

Mainprice, D., Barruol, G., Ben Ismail, W., 2000. The seismic anisotropy

603

of the Earth’s mantle: from single crystal to polycrystal. In: Karato, S.,

604

Forte, A. M., Liebermann, R. C., Masters, G., Stixrude, L. (Eds.), Earth’s

605

Deep Interior: Mineral Physics and Tomography From the Atomic to the

606

Global Scale. No. 117. Geophysical Monograph, AGU, Washington D.C.,

607

pp. 237–264.

608

Morley, A. M., Stuart, G. W., Kendall, J.-M., Reyners, M., 2006. Mantle

609

wedge anisotropy in the Hikurangi subduction zone, central North Island,

610

New Zealand. Geophys. Res. Lett. 33.

611

M¨ uller, R. D., Roest, W. R., Royer, J. Y., Gahagan, L. M., Sclater,

612

J. G., 1997. Digital isochrons of the world’s ocean floor. J. Geophys. Res.

613

102 (B2), 3211–3214.

614

Nakajima, J., Hasegawa, A., 2004. Shear-wave polarization anisotropy and

615

(30)

Accepted Manuscript

subduction-induced flow in the mantle wedge of northeastern Japan. Earth

616

Planet. Sci. Lett. 225, 365–377.

617

Ohtaki, T., Kanjo, K., Kaneshima, S., Nishimura, T., Ishihara, Y., Yoshida,

618

Y., Harada, S., Kamiya, S., 2000. Sunarjo, Broadband Seismic Network in

619

Indonesia-JISNET. Bull. Geol. Surv. Japan 51, 189–203.

620

Pinero-Felicangeli, L., Kendall, J.-M., 2008. Sub-slab mantle flow parallel to

621

the Caribbean plate boundaries: Inferences from SKS splitting. Tectono-

622

physics 462, 22–34.

623

Polet., J., Silver, P. G., Beck, S., Wallace, T., Zandt, G., Ruppert, S., Kind,

624

R., Rudloff, A., 2000. Shear wave anisotropy beneath the Andes from the

625

BANJO, SEDA and PISCO experiments. J. Geophys. Res. 105 (B3), 6287–

626

6304.

627

Savage, M. K., 1999. Seismic anisotropy and mantle deformation: What have

628

we learned from shear wave splitting? Rev. Geophys. 37 (1), 65–106.

629

Sdrolias, M., Muller, R. D., 2006. Controls on back-arc basin formation.

630

Geochem. Geophys. Geosyst. 7 (4).

631

Silver, P. G., 1996. Seismic anisotropy beneath the continents: Probing the

632

depths of geology. Annu. Rev. Earth Planet. Sci 24, 385–432.

633

Silver, P. G., Chan, W. W. J., 1991. Shear–wave splitting and subcontinental

634

mantle deformation. J. Geophys. Res. 96, 16429–16454.

635

Silver, P. G., Savage, M. K., 1994. The interpretation of shear wave splitting

636

(31)

Accepted Manuscript

parameters in the presence of two anisotropic layers. Geophys. J. Int. 119,

637

949–963.

638

Smith, G., Wiens, D., Fischer, K., Dorman, L., Webb, S., Hildebrand,

639

J., 2001. A complex pattern of mantle flow in the Lau backarc. Science

640

292 (5517), 713–716.

641

Syracuse, E., Abers, G., 2006. Global compilation of variations in slab depth

642

beneath arc volcanoes and implications. Geochem. Geophys. Geosyst. 7.

643

Teanby, N. A., Kendall, J.-M., Van der Baan, M., April 2004. Automation of

644

shear–wave splitting measurements using cluster analysis. Bull. Seis. Soc.

645

Am. 94 (2), 453–463.

646

Wheller, G., Varne, R., Foden, J., Abbott, M., 1987. Geochemistry of Quater-

647

nary volcanism in the Sunda-Banda arc, Indonesia, and three-component

648

genesis of island-arc basaltic magmas 32 (1-3), 137–160.

649

Wolfe, C. J., Solomon, S. C., 1998. Shear–wave splitting and implications

650

for mantle flow beneath the MELT region of the East Pacific Rise. Science

651

280, 1230–1232.

652

Références

Documents relatifs

Consecutive reactions of adducts, resulting from OH radicals and aromatics, with the tropospheric scavenger molecules O 2 , NO and NO 2 have been studied for benzene, toluene, m-

Identifying productive zones of the Sarvak formation by integrating outputs of different classification methods.. Pedram Masoudi, Behzad Tokhmechi, Alireza Bashari,

standardized steady state heavy-duty diesel engine emission testing.. Duty Cycle for nonroad applications. This and heavier loading conditions than. )iesel Duty

Mais à peine Porsenna et Galerite eurent-ils le loisir de connaître leur bonheur, qu'ils eurent une douleur extrême ; car la sage et prudente Nicétale mourut peu de temps après

indicators consistently characterize flow channeling in porous and fractured media, with D

For power amplifier (PA) systems, digital predistortion has been a popular way to enhance the linearity of the system without sacrificing power efficiency by power

The experience of watching Train of Thought v2.0 is one of exploration. The viewer sits back and follows the main narrative threads created by the system. The story

The influence of manufactured errors was neglected in this analysis. This corresponds to the case of heavily loaded gears as defined by J.. The non-linear