Estimating earthquake source parameters€¦ · Web viewFile name = DuboisFeiglKomatitschEPSL20061003b.doc. Application name Microsoft Word ... seismic slip by joint inversion

Three-dimensional mechanical models for the June 2000 earthquake sequence in the South Icelandic Seismic Zone

Loïc Dubois a, Kurt L. Feigl a, *, Dimitri Komatitsch b, Thóra Árnadóttir c, and Freysteinn Sigmundsson c

a Dynamique Terrestre et Planétaire CNRS UMR 5562, Université Paul Sabatier, Observatoire Midi-Pyrénées, 14 av. Édouard Belin, 31400 Toulouse, France

b Laboratoire de Modélisation et d'Imagerie en Géosciences CNRS UMR 5212, Université de Pau et des Pays de l'Adour, BP 1155, 64013 Pau Cedex, France

c Nordic Volcanological Center, Institute of Earth Sciences, University of Iceland, Reykjavik, Iceland

Tuesday, May 09, 2023

Submitted to Earth and Planetary Science Letters

Number of characters = 0Number of words = 0Number of words in text only = 6455 Max is 6500 (not abstract, captions or references)Supplementary Material? YesNumber of pages (including figures) = 22Number of figures = 6Number of tables = 2Total number of display items = 8 Max is 10.Number of references = 79 Max is 80.

File name = document.doc

Application nameMicrosoft Word 2004 for MacEndNote Version 10.0 for MacMathType Version 5 for Mac.

*Corresponding author, now atDepartment of Geology and GeophysicsUniversity of Wisconsin-Madison1215 West Dayton StreetMadison, WI53706 [email protected]. +1 608 262-8960Fax. +1 608 262-0693

1

1

2

3

4

5

6

7

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

3233

3435

36

37

38

39

40

41

42

43

44

45

46

47

48

2

mailto:[email protected]

Dubois et al. submitted 2023-05-09

TABLE OF CONTENTS

Table of contents 2

Abstract 3

1. Introduction 3

2. Data and Methods 6

2.1 Data 62.2 Numerical Solution using the Finite Element Method 72.3 Modeling poroelastic deformation with an elastic approximation 82.4 Viscoelastic modeling 92.5 Stress transfer 10

3. Results 10

3.1 Estimating the distribution of co-seismic fault slip 103.2 Modeling postseismic deformation with the elastic approximation for poroelastic effects 113.3 Viscoelastic model for post-seismic deformation 123.4 Stress transfer 13

4. Discussion and Conclusions 14

4.1 Co-seismic effects 144.2 Viscoelastic effects 154.3 Poroelastic effects 154.4 Migrating seismicity 16

Acknowledgments 16

Supplementary Material 17

SM.1 Testing against other solutions 17SM.2 Estimating co-seismic slip by joint inversion of GPS and INSAR measurements 17SM.3 Approximating complete poroelastic relaxation by an elastic solution 18

References 20

Tables 26

Last page 26

Page 2

34

49

50

51

52

53

54

55

56

57

58

59

60

61

62

63

64

65

66

67

68

69

70

71

72

73

74

75

76

77

56


ABSTRACT

We analyze a sequence of earthquakes that began in June 2000 in the South Iceland Seismic Zone (SISZ)

in light of observations of crustal deformation, seismicity, and water level, focusing on the tendency for

seismicity to migrate from east to west. At least five processes may be involved in transferring stress from one

fault to another: (1) propagation of seismic waves, (2) changes of static stress caused by co-seismic slip, (3)

inter-seismic strain accumulation, (4) fluctuations in hydrological conditions, and (5) ductile flow of subcrustal

rocks. All five of these phenomena occurred in the SISZ during the post-seismic time interval following the

June 2000 earthquakes. By using three-dimensional finite element models with realistic geometric and rheologic

configurations to match the observations, we test hypotheses for the underlying geophysical processes. The

coseismic slip distribution for the Mw 6.6 June 21 mainshock is deeper when estimated in a realistic layered

configuration with varying crustal thickness than in a homogenous half-space. Processes (1) and (2) explain

only the aftershocks within a minute of the mainshocks. A quasi-static approximation for modeling poroelastic

effects in process (4) by changing Poisson’s ratio from undrained to drained conditions does not provide an

adequate fit to post-seismic deformation recorded by interferometric analysis of synthetic aperture radar images.

The same approximation fails to explain the location of subsequent seismicity, including the June 21 mainshock.

Although a viscoelastic model of subcrustal ductile flow (5) can increase Coulomb failure stress by more than

10 kPa, it does so over time scales longer than a year. Accumulating interseismic strain (3) in an elastic upper

crust that thickens eastward breaks the symmetry in the lobes where co-seismic stress exceeds the Coulomb

failure criterion. This process may explain the tendency of subsequent earthquakes to migrate from east to west

across the South Iceland Seismic Zone within a single sequence. It does not, however, explain their timing.

1. INTRODUCTION

The South Iceland Seismic Zone (SISZ) is aptly named. Located on the boundary between the North

America and Eurasia plates, it accommodates their relative motion as earthquakes. The plate boundary follows

the spreading mid-Atlantic Ridge, comes ashore on the Reykjanes Peninsula, where it becomes a shear zone to

reach the Western Volcanic Zone at a triple point near the Hengill volcanic center (Fig 1). There, most of the

present-day deformation continues across the SISZ to the Eastern Volcanic Zone. Overall, the SISZ

accommodates sinistral shear along its east-to-west trend, but most of its large (M 6) earthquakes rupture as

right-lateral strike slip on vertical faults striking north. These parallel faults are 10 – 20 km long and ~5 km

apart, analogous to books on a shelf [1]. Indeed, the SISZ deforms as a shear zone, accommodating

18.9 ± 0.5 mm/yr of relative plate motion [2, 3] in a band less than 20 km wide [4]. Consequently, the

interseismic strain rate is ~10–6 per year. Its kinematics can be described as a locked fault zone slipping 16 to

20 mm/yr below 8 to 12 km depth with a dislocation in an elastic half-space [5]. If accumulated elastically over

Page 3

78

78

79

80

81

82

83

84

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

100

101

102

103

104

105

106

107

108

109

910


a century, this strain could be released as ~2 m of co-seismic slip. Fortunately, earthquakes in the SISZ do not

rupture its entire 100-km length from east to west; the 20-km length of the north-south faults seems to limit their

magnitude to 6.5 or 7.

The seismic sequence of events began on 17 June 2000, with a main shock of magnitude Mw = 6.6 [6]. It

produced co-seismic deformation that was measured geodetically by both GPS [7] and INSAR [8]. Three and a

half days (81 hours) later, a second earthquake ruptured a distinct fault some 18 km to the west of the first event.

Both earthquakes have strike-slip focal mechanisms on faults dipping nearly vertically [9]. Both earthquakes

ruptured the ground surface in a right-lateral, en echelon pattern with an overall strike within a few degrees of

North [10]. The slip distributions estimated from geodetic data for both earthquakes have maxima over 2 m,

centroid depths around 6 km, down-dip widths of 10 – 12 km, and along-strike lengths of 16 – 18 km [7, 11].

In this study, we build on previous estimates of the distribution of slip that occurred on the rupture

surfaces of the two mainshocks in June 2000 from GPS measurements of displacements [7], INSAR recordings

of range change [8], and a combination of both in a joint inversion [11]. All three of these studies assume an

elastic half space with uniform values of shear modulus (or rigidity) and Poisson’s ratio . Yet other work

STUDY/STUDIES used four times in the last two sentences outside of Iceland suggest that this simple

assumption can bias estimates of slip distribution [12, 13]. For example, the aftershock hypocenters located by

seismological procedures are deeper than the bottom of the fault rupture inferred from geodetic measurements in

the case of the 1994 Northridge earthquake in California [14]. In other words, the seismological and geodetic

procedures for locating co-seismic slip will yield different estimates if they assume different elastic models. A

similar discrepancy between the deepest aftershocks and the bottom of the slip distribution also appears to apply

to the June 2000 sequence in the SISZ. Although the slip diminishes to less than 1 m below 8 km depth [15], the

relocated aftershock hypocenters reach down to 10 km, with a few as deep as 12 km [16]. The hypocentral

depths estimated from seismology using a layered velocity model are 6.3 and 5.0 km for the June 17 and 21

mainshocks, respectively [6]. By using layered models that account for increasing rigidity with depth, we expect

deeper slip distributions [17, 18]. An accurate estimate of the slip distribution is required to test the hypothesis

that the M 6 earthquakes in the SISZ rupture into the lower crust [19]. Accurate estimates of slip distribution are

also important for calculating maps of earthquakes triggered by stress transfer. The latter is sensitive to the

former because the threshold value of 10 kPa (0.1 bar) usually used to evaluate the Coulomb failure criterion is

several orders of magnitude smaller than the stress drop for a typical earthquake [20, 21].

The earthquakes in the 2000 SISZ sequence seem to be causally related. Just after the June 17 mainshock,

additional earthquakes struck the Reykanes Peninsula, as far as 80 km to the west of the June 17 epicenter.

Three of these generated coseismic displacements large enough to measure by INSAR and GPS and to model

with dislocation sources with moments equivalent to Mw = 5.3, 5.5 and 5.9 [22, 23]. These secondary

earthquakes ruptured the surface, broke rocks, and coincided with a drop of several meters in the water level of

Page 4

1112

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

1314


Lake Klevervatn [24]. Three of these events, at 8 s, 26 s, and 30 s after the June 17 mainshock at 15:40:41 UTC

were dynamically triggered by the seismic wave train as it propagated westward [6]. A fourth event occurred 5

minutes after the June 17 mainshock at a distance of almost 80 km [6]. The June 17 earthquake appears to have

triggered the June 21 earthquake, based on the increase in static Coulomb stress calculated at the June 21

hypocenter on a vertical, north-striking fault [25]. The migration of M ≥ 6 earthquakes from east to west across

the SISZ has been observed in previous earthquake sequences in 1732-1734, 1784, and 1896 [1]. Another M ≥ 6

event occurred in 1912 at the east end of the SISZ [26, 27].

Understanding the processes driving the seismicity from east to west is the second objective of this study.

Although static Coulomb stress changes can explain the location of the triggered earthquakes, this simple theory

cannot explain the time delay between the “source” (triggering) and “receiver” (triggered) events, as pointed out

by Scholz [21, 28] and recently reviewed by Brodsky and Prejean [29]. At least five processes may be involved

in transferring stress from one fault to another: (1) propagation of seismic waves, (2) changes of static stress

caused by co-seismic slip, (3) inter-seismic strain accumulation, (4) fluctuations in hydrological conditions, and

(5) ductile flow of subcrustal rocks. All five of these phenomena occurred in the SISZ during the post-seismic

time interval following the June 2000 earthquakes.

(1) The dynamic stresses caused by propagation of seismic waves apparently triggered the earthquakes on

the Reykanes Peninsula in the first minute following the June 17 mainshock, as described above [6].

(2) In the “cascading seismicity” hypothesis, one earthquake triggers the next in a kind of seismic

“domino effect”. Changes in static stress could alter rate- and state-dependent friction, thus producing and

(?????) a finite time delay between successive earthquakes[30].

(3) In terms of moment, the main events in the June 2000 sequence released only about a quarter of the

strain accumulated since 1912, the date of the last major earthquake in the SISZ, assuming a constant strain rate

over the intervening 88 years [4, 15]. Such interseismic strain accumulation could conceivably contribute to

stress transfer. If strain accumulates at the rate of 10-6/yr, it could increase stress by 10 kPa in less than a year,

assuming a Young’s modulus of 50 GPa.

(4) Fluctuations in hydrological conditions occurred in the weeks to months after the June 17 mainshock

[31, 32]. Poroelastic effects can explain about half the post-seismic signal in an interferogram spanning June 19

through July 24 in the SISZ [32]. The same effects might also explain stress transfer in the 2000 SISZ sequence.

Under this hypothesis, the co-seismic perturbation to the hydrologic pressure field also alters the stress

conditions on faults near the mainshock, triggering aftershocks in areas where increasing pressure unclamps

faults near failure [33, 34].

(5) Ductile flow of subcrustal rocks occurs over longer time scales of several years. For example post-

seismic deformation has been recorded by GPS around the June 2000 faults as late as May 2004 [35]. These

centimeter-sized displacements have been modeled using a viscoelastic Burger’s rheology in a semi-analytic

Page 5

1516

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

1718


spherical harmonic formulation [35]. Here, we use a linear Maxwell viscoelastic rheology in a Cartesian finite

element formulation to investigate the effect of geometry, viscosity and other parameters on surface

displacements. Although which one ??? you talk about geometry, viscosity and other parameters in the previous

sentence ductile flow is much too slow to explain the 86-hour time delay between the June 17 and June 21

events, it may contribute to the time delays (~100 years) between earthquake sequences in the SISZ.

All these considerations motivate us to move beyond the first-order in which sense ??? better to suppress

that word I guess approximation of a half-space with uniform elastic properties. In this study, we explore

models with more realistic geometric and rheologic configurations. To do so, we calculate stress and

displacement fields using a finite element method [e.g., 36]. In particular, we use the TECTON code, renovated

by Charles Williams [37-39], as described below. This approach allows us to account for the variations in

crustal thickness (Fig 1 and Fig 2) inferred from earthquake hypocenter locations, seismic tomography, and

gravity modeling [40-44].

2. DATA AND METHODS

2.1 Data

In this study, we consider four types of data, all of which have been analyzed and described previously.

GPS

The complete GPS data set, including the post-seismic time series, has been described elsewhere [5, 35].

The co-seismic displacement vectors are the same as used previously to estimate the distribution of slip on the

faults that ruptured on June 17 and 21 [7, 11].

INSAR

Interferometric analysis of synthetic aperture radar images (INSAR) has recorded the crustal deformation

that occurred between June and September, 2000, as described previously [8, 15, 22, 32]. The data set used here

is the same as analyzed by Pedersen et al. [11]The characteristic relaxation time of the post-seismic deformation

near the June 17 fault is of the order of a month because another interferogram spanning the second post-seismic

month (12 August to 16 September) shows less than 15 mm of differential range change across the fault (blue

curve in Figure 2e of Jónsson et al. [32]). About half of the post-seismic deformation observed within 5 km of

the June 17 fault recorded by INSAR in the first month can be explained by poroelastic effects involving fluid

flowing in the epicentral area [32].

Page 6

Call Fig 1.

1920

178

179

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

2122


Water level in wells and boreholes

Fluctuations in hydrological conditions have been observed in periodic measurements of pressure and

water level in wells and boreholes around the SISZ [31, 32]. The polarity of these changes can be well described

by the two mainshock focal mechanisms [31, 32]. In the post-seismic interval, however, the wells recover over

time scales varying from hours to weeks, indicating large variations in the rock-mechanical properties over

distances as short as a few kilometers.

Earthquake locations and focal mechanisms

The earthquakes recorded by the SIL seismological network [19, 45] operated by the Icelandic

Meteorological Office have been relocated using multiplet algorithms [46-48] to establish a catalog of focal

mechanisms and locations with a precision better than 1 km in all three dimensions [16, 49, 50] as part of the

PREPARED project [51].

2.2 Numerical Solution using the Finite Element Method

TECTON is a software package that implements a finite element formulation based on three-dimensional

hexahedral elements [37-39]. We have tested TECTON for accuracy by comparing its results to semi-analytic

solutions, as described in the doctoral thesis of the first author [52] and in the Supplemental Material. We use

the word “configuration” to describe the combination of the geometry of the mesh (topology) and distribution of

mechanical properties (rheology) within it. These configurations appear in Fig 1, Fig 2, Table 1, and Table 2.

Mesh

We have developed two meshes specifically for this problem to reproduce the main interfaces in our

problem: upper/lower crust, mantle/crust, as well as the fault rupture surfaces for the June 17 and June 21

mainshocks. The boundary conditions on the bottom and four sides are fixed while the top surface is free. The

region under study is 180 km by 100 km by 20 km (blue box in Fig 1). To avoid edge effects, we extend the

meshes to 300 km by 300 km by 100 km. The elastic solution achieves numerical convergence with 70 000

nodes, but we use 140 000 nodes to handle the time-dependence in the visc-elastic case [52]. The fault slip is

described as relative, horizontal displacements at 690 “split nodes” [38].

The simplest configuration is a homogeneous half-space labeled HOM. The mesh for HOM includes

horizontal layers with constant thickness. The same mesh defines the geometry of the heterogeneous HET

configuration. A different mesh, used in the TOM and ISL configurations, allows lateral variations in crustal

thickness, and thus gradients in rheologic properties, as sketched in Fig 1B.

Page 7

2324

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

225

226

227

228

229

230

231

232

233

234

2526


Rheology

After defining the meshes, we now define the rheology for each configuration. All the rheologic

configurations obey an elastic constitutive relation (Hooke’s law), with stress linearly proportional to strain. For

each element in the mesh, we specify the Poisson’s ratio and a shear modulus , as summarized in Fig 2 and

Table 1.

In the HOM and TOM configurations, we assume 0.28 and 30 GPa throughout the half-

space [11]. In the layered HET and ISL configurations, Poisson’s ratio varies from 0.28 in the upper crust

to 0.25 in the lower crust, and then to 0.30 in the mantle, in accordance with previous studies of

Iceland [40, 53, 54]. The thickness of these layers is constant in the HET configuration, with the base of the

upper crust set to Hu = 5 km depth and the base of the lower crust set to Hc = 23 km depth. In the ISL

configuration, the upper and lower crustal layers vary in thickness, as sketched in Fig 2B and Fig 2C. These two

crustal layers both thicken and dip toward the east)ity I don’t like this sentence very much, we should maybe

rephrase i?],) as evident in previous three-dimensio by studies [40, 41, 43, 44, 54]. Given the density, we then

use the P-wave and S-wave velocities [52] to calculate and (under undrained conditions) for each element

(Fig 2).

The shear modulus is set to 30 GPa in the HOM and TOM configurations. In the HET and ISL

configurations, it varies from 30 GPa in the upper crust to 60 GPa in the mantle. The HOF and

HEF configurations add weak fault zones to the HOM and HET configurations, respectively. In the elements

located within a few kilometers of the faults, the HOF and HET configurations include values as low as

3 GPa, i.e. four times smaller than in the surrounding crustal rocks [55]. Such low values can be justified by

previous co-seismic studies that have assumed values as low as 10 GPa [56] and as high as 50

GPa [57]. Following Cocco and Rice [55, 58], we assume that the bulk modulus has the same value in the

damage zone as in the surrounding rock. Hence we set Poisson’s ratio as high as 0.43 within the weak

fault zones in the HEF and HOF configurations.

A thorough sensitivity analysis suggests that the co-seismic displacement field at the surface is most

sensitive to the value of the shear modulus (rigidity) [52]. For this parameter, a perturbation of 10 percent can

increase the misfit to coseismic displacements measured by GPS and INSAR by as much as 50 percent. The

shape of the displacement field is also fairly sensitive to the geometric location of the fault, particularly its

depth. The displacement field, however, is relatively insensitive to the geometry of the subcrustal layers.

Similarly, the absolute value of Poisson’s ratio makes little difference. A displacement field calculated with

= 0.25 is almost (??????)extremely similar to one with = 0.28, in agreement with a previous study of the

June 2000 earthquakes [11].

To describe the post-seismic time interval, we introduce viscosity in the lower crust below a depth of Hu,

and in the upper mantle, below a depth of Hc.

Page 8

2728

235

236

237

238

239

240

241

242

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

2930


2.3 Modeling poroelastic deformation with an elastic approximation

To explain the post-seismic interferogram (Fig 5), Jónsson et al. invoke poroelastic effects in the first

month following the June 17 mainshock [32]. To describe the observed signal, they employ a modeling

approach involving changes in Poisson’s ratio . this quasi-static approximation, which we will call the Poisson

Recipe in the rest of this study, originally introduced by Rice and Cleary [59] and described in the

Supplementary Material, uses only two elastic coefficients to solve a problem in a two-phase, poroelastic

medium that requires at least two more coefficients to describe completely. This quasi-static model assumes a

change in elastic properties, specifically Poisson’s ratio, from undrained to drained conditions, as for several

other earthquakes in different tectonic settings [60, 61]. The validity of the assumptions underlying

this simple, static and elastic approximation to what is actually a complex, time-dependent,

poroelastic problem [62] will enter into our discussion below. Using all these assumptions, this single-parameter

description approximates the complete post-seismic displacement fields generated by poroelastic rebound as the

difference between two co-seismic fields calculated with different values of Poisson’s ratio. Consequently, only

the change in Poisson’s ratio undrained – drained controls the shape of the displacement field.

The previous study of the 2000 SISZ sequence [32] makes strong assumptions about the spatial

distribution of the Poisson’s ratio change . As a first approximation, it uses a half-space formulation to

calculate the Green’s functions G such that = 0.04 everywhere [63]. As a second approximation, it develops

a different formulation with horizontal layers, such that the drainage occurs only in the uppermost 1.5 km of the

crust, where = 0.08.

Making a third approximation is one of the goals of this study. Here we allow the Poisson’s ratio change

to vary horizontally as well as vertically. Indeed, the finite element method can calculate the elastic Green’s

function G from an arbitrary geometric distribution of change in Poisson’s ratio

In the Supplementary Material, we conclude that = 0.04 is a reasonable average value for the change in

Poisson’s ratio. We note, however, that this parameter trades off with the thickness of the drainage layer.

Accordingly, we experiment with different values and geometric configurations, as listed in Table 2. To

summarize, we consider three categories of poroelastic configurations, depending on spatial distribution of .

In the first category, is uniform (INV1, FOR1 and FOR2). In the second category is a function of depth

only (FOR3, TRY1, TRY2, TRY3, INV2, and INV3). In the third category, varies with depth and horizontal

position (INV4).

2.4 Viscoelastic modeling

To describe post-seismic deformation, we introduce a linear Maxwell viscoelastic rheology in the lower

crust and upper mantle layers of the geometric configurations described above. Varying the viscosity in a series

of forward models, we seek the value that minimizes the misfit to the horizontal components of the

Page 9

Call Fig 5

3132

269

270

271

272

273

274

275

276

277

278

279

280

281

282

283

284

285

286

287

288

289

290

291

292

293

294

295

296

297

298

299

300

301

3334


displacement field measured by GPS in the post-seismic interval from the time of the earthquakes through May

2004 [5, 35].

2.5 Stress transfer

Using the slip distributions estimated as described above, we calculate the Coulomb stress on faults in the

SISZ using the same notation and values as Árnadóttir et al. [25]

σC = Δσ t + μ f (Δσ n − B3

Δσ kk ) (1)

where B = 0.5 is Skempton’s coefficient, f = 0.75 is the effective coefficient of friction on the fault, σt is the

increase in shear stress projected onto the fault plane (taken positive in the sense of the slip vector), σn is the

increase in normal stress (taken positive in tension), and σkk is the trace of the stress tensor.

To evaluate the Coulomb failure criterion, we project the stress tensor onto the horizontal strike slip

vector on a north-striking vertical “receiver” fault that typifies the SISZ focal mechanisms [25]. The result is a

scalar Coulomb stress σC. At a given location, if the value of σC exceeds a critical (threshold) value σC(critical),

then we VERB MISSING HEREconclude that the stressing process favors the occurrence of earthquakes there.

If, in fact, an aftershock occurs at the given location, then we count it as a “triggered” event. In this study, we

assume a threshold value of σC(critical) = 10 kPa, equivalent to the stress induced by a column of water 1 m high.

3. RESULTS

3.1 Estimating the distribution of co-seismic fault slip

Using the method described in the Supplementary material, we estimate the distribution of co-seismic

slip for the two mainshocks in a joint inversion of the GPS and INSAR measurements. The shape and depth of

the slip distribution depend on the geometric and rheologic configurations used in the calculation.

Homogenous configurations

Using the homogeneous geometric configuration (HOM), we estimate the distribution of co-seismic slip

on the June 17 and June 21 fault planesFig 3. It is essentially similar to that estimated using the dislocation

formulation [15], further validating our finite-element approach. The largest differences are in the upper-most

part of the fault, where the dislocation and finite-element WORD MISSING HERE: “methods” maybe?methods

parameterize the slip differently, as described above. The uncertainty in the slip estimates reaches 1 m near the

surface, particularly in the section of the June 17 fault south of the hypocenter.. This result may reflect

Page 10

Call Fig 3

3536

302

303

304

305

306

307

308

309

310

311

312

313

314

315

316

317

318

319

320

321

322

323

324

325

326

327

328

3738


discrepancies between the INSAR and GPS measurements for points near the fault. Elsewhere on both faults,

however, the standard deviation of the slip is less than 0.5 m. Most of the slip is shallow, above 6 km depth. The

slip decays to less than 0.5 m below 9 km depth for the June 17 fault and 11 km depth for the June 21 fault. The

centroid depths are 3.7 and 3.8 km, respectively. Some of the aftershock hypocenters are located outside the

high-slip areas: between 8 and 10 km depth in the June 17 fault plane and between easting coordinates -8 and -

6 km in the June 21 fault plane.

Layered configurations

The upper two panels of Fig 3 show the slip distribution estimated using the HET configuration. The

difference between the HET and HOM configurations appears in Fig 4. The slip extends slightly deeper in HET

than in HOM. The centroid is 0.3 km deeper for the June 17 fault and 0.5 km deeper for the June 21 fault. For

the June 21 fault, there is over 1 m more slip at 10 km depth in HET than in HOM. The increase in slip

concentrates around the hypocenter relocated at 5.0 km depth from seismology using a layered velocity model

[6]. This difference is statistically significant with over 95 percent confidence because it is more than two times

larger than the standard deviation of the HET estimate (Fig 4). The surface slip is also significantly larger in the

HET configuration than in the HOM estimate. The slip distributions using the ISL configuration are essentially

the same as those found with HET [52].

Configurations with fault damage zones

The HOF configuration adds a weak damage zone around the mainshock faults to the HOM

configuration. The slip estimated in HOF concentrates in the upper crust, significantly increasing the maximum

slip value around the June 21 hypocenter at 5 km depth (Fig 4). The slip is significantly shallower in HOF than

in HOM. Of all the configurations, HOF shows the steepest slip gradient and thus the highest residual stress.

The aftershock locations seem to correlate with these areas of high slip gradient, particularly for the southern

edge of the June 21 rupture. The aftershocks also concentrate at the edges of the slip distribution where the slip

estimates become smaller than their typical 50-cm uncertainties. Adding a damage zone to the HET

configuration to specify the HEF configuration also changes the slip distribution. The HEF result has

significantly more slip in the uppermost mesh elements and around the hypocenters in both fault planes than

does the HET result.

3.2 Modeling postseismic deformation with the elastic approximation for poroelastic effects

Fluids flowing in the fractured and fissured rock around the June 17 fault seem to be the most likely

explanation for the observed time delays: 86 hours between the two mainshocks and the month-long post-

seismic signature observed in the interferogram (Fig 5). To explore this possibility, we consider poroelastic

Page 11

3940

329

330

331

332

333

334

335

336

337

338

339

340

341

342

343

344

345

346

347

348

349

350

351

352

353

354

355

356

357

358

359

4142


effects by extending the work of Jónsson et al. [32] with a series of model calculations summarized in Table 2.

To begin, we reproduce their results by using the homogeneous geometric configuration HOM and a Poisson’s

ratio contrast = 0.04. This model calculation (FOR1) has a variance reduction of σ2 = 39 percent for the

observed one-month interferogram with respect to a null model. Another such calculation, FOR2, concentrates

the drainage in the uppermost 2 km of crust, similar to the layered model of Jónsson et al. [32]. It increases the

Poisson’s contrast to = 0.06 to achieve a variance reduction of 48 percent. Both FOR1 and FOR2 appear to

fit the data well in a profile across strike (A-A’ in Fig 5c), in agreement with the black and green dashed curves,

respectively, in Figure 2 of Jónsson et al. [32]. In map view and a profile parallel to strike (B-B’ in Fig 5b),

however, the deformation calculated with the finite element approach exhibits small features near the fault that

are not present in the dislocation calculation [32]. We interpret these features as a consequence of the increased

surface slip allowed in the finite element formulation. Another attempt, FOR3, using = 0.06 in the HET

configuration, yields a slightly different result (σ2 = 40 percent), highlighting the importance of layering in

poroelastic models.

We also estimate the Poisson’s ratio change in a trial-and-error procedure. The TRY1 configuration

fits the data (σ2 = 44 percent) as well as any other solution we could find by assigning a high value of = 0.10

to the shallow layer between 1 and 2 km. The TRY2 configuration achieves the same reduction with lower

values of The TRY3 solution is equivalent to FOR3 (Table 2).

Next, we allow horizontal variations in the Poisson’s contrast in an attempt to reproduce the

diminished southwest lobe of the post-seismic fringe pattern. This set of geometric configurations includes a

damage zone extending 2-3 km around the fault. The damage zone can be further subdivided into four quadrants

in two layers, adding as many as eight free parameters that we estimate using a non-negative least squares

method [64]. One such attempt, the INV1 solution with only one free parameter, seems reasonable because it

resembles FOR1 in and σ2. The INV2 solution with two free parameters and INV3 solution with five free

parameters, resemble TRY2 and TRY1 respectively (Table 2). An extreme attempt with 8 free parameters,

INV4, achieves a variance reduction of σ2 = 65 percent by forcing the Poisson’s ratio contrast to the physically

unreasonable value of = 0.244 in the southwest quadrant of the damage zone between 2 and 5 km depth.

3.3 Viscoelastic model for post-seismic deformation

Of the parameters in the viscoelastic model, viscosity has the most influence on the post-seismic

displacement field, based on a sensitivity study that varies each parameter individually [52]. By optimizing the

fit to the horizontal components of the GPS measurements, we estimate the viscosities in the lower crust and

upper mantle, albeit with large uncertainties. For the lower crust, we find the 95 percent confidence interval for

viscosity to fall between 2 and 20 EPa.s during the first post-seismic year, and 9 – 90 EPa.s for the interval from

July 2001 to May 2004. (1 exapascal-second = 1018 Pa.s). The different geometric configurations HOM, HET

Page 12

4344

360

361

362

363

364

365

366

367

368

369

370

371

372

373

374

375

376

377

378

379

380

381

382

383

384

385

386

387

388

389

390

391

392

4546


and ISL yield similar results. For the upper mantle, the best-fitting viscosity is 1 – 30 EPa.s for the early post-

seismic interval and the ISL configuration. The other configurations yield lower viscosity estimates with wider

confidence intervals, suggesting that using a realistic geometry is particularly important when continuous GPS

observations are sparse. For the upper mantle, a viscosity of 2 – 100 EPa.s, irrespective of geometric

configuration, fits the 2001-2004 data. These values are compatible with a previous study that used a different

modeling approach to describe the same post-seismic observations [35].

3.4 Stress transfer

Next, we evaluate the consequences of various sources of stress in terms of their ability to trigger

earthquakes.

June 17 and 21 co-seismic stress

Does the June 17 mainshock increase the Coulomb stress on a north-striking, vertical strike-slip fault at

the June 21 hypocenter? Yes, the increase is 100 to 150 kPa, in the three layered configurations (HEF, HET and

ISL), confirming the result of a previous study [25] that assumes a homogeneous half-space equivalent to our

HOM configuration. In all three cases, the stress increase calculated at the June 21 hypocenter exceeds the

conventional threshold of 10 kPa. In map view, the area where earthquakes are favored, within the contour of

10 kPa, is somewhat smaller when calculated with the layered HET configuration than with the uniform HOM

configuration (Fig 6a).

Do the two mainshocks increase the Coulomb stress at the location of the aftershocks? Yes, about half the

aftershocks are located within the 10-kPa threshold contour. Thus we conclude that only half the earthquakes

after the June 17 mainshock, including the June 21 mainshock, can be explained by the co-seismic stress

produced by the June 17 mainshock under the Coulomb failure criterion. In other words, only half the seismicity

is triggered by increases in static stress. The other half must be triggered by some other, presumably dynamic,

process, as considered in the following paragraphs.

“Domino effect” of cascading aftershocks

We generalize the Coulomb stress calculation to include all the 14 388 earthquakes between June 17 and

December 31 for which reliable focal mechanisms have been estimated, except the June 21 mainshock. This

calculation determines the time-dependent changes in the Coulomb stress field on vertical, north-striking strike-

slip faults. (To save calculation time, we use the layered configuration HET and the EDGRN/EDCMP

implementation to calculate the Green’s functions [65]). Fig 6c shows the changes in Coulomb stress at 4 km

depth. This stress increase occurs within less than 10 minutes of the June 17 mainshock and remains roughly

constant until the end of 2000. At the June 21 hypocenter, however, the increase is only 5 kPa. The primary

Page 13

4748

393

394

395

396

397

398

399

400

401

402

403

404

405

406

407

408

409

410

411

412

413

414

415

416

417

418

419

420

421

422

423

4950


contributor to the static stress increase is a magnitude 5 aftershock that occurred some 130 seconds after the

June 17 mainshock at 15:42:50 midway between the two mainshock faults [16, 49]. Accounting for this large

event enlarges the area of increased Coulomb stress on the June 21 fault plane.

Modeling stress transfer with the elastic approximation for poroelastic effects

The elastic approximation for modeling poroelastic effects explains the location of only the aftershocks

nearest to the mainshock faults. For example, even the extreme HET configuration with the INV4 values

does not increase the Coulomb failure stress on the June 21 fault (Fig 6b).

Viscoelastic processes

Viscoelastic processes are also capable of triggering seismicity. After four years of post-seismic

relaxation, the Coulomb failure stress at 4 km depth exceeds the threshold of 10 kPa over most of the SISZ (Fig

6d). The threshold contour for the ISL configurationsis less symmetric than the one for the HET configuration,

highlighting the importance of a realistic geometric configuration in stress calculations.

Interseismic strain accumulation

Stress also builds up during the interseismic phase of the earthquake cycle as elastic strain accumulates

between large earthquakes. If the interseismic strain rate is constant in time and uniform in space, but the rigid

upper crust thickens from west to east, then the stress rate field will exhibit a gradient. At a given depth, say

5 km (on the upper/lower crust boundary near the June 21 mainshock hypocenter), the rigidity increases

westward because the stiffer lower crust is shallower there. Consequently, stress will be higher on the west side

of a fault than on the east side, thereby explaining the tendency of seismicity to migrate from east to west across

the SISZ. To quantify this purely geometric effect, we run a simple finite element model to mimic the constant

interseismic strain rate. This elastic interseismic model applies the far-field boundary conditions as 90 cm of

displacement at N110o on the east and west faces of the mesh to mimic 100 years of plate motion. Since the

strain depends on the dimension of the mesh, we cannot calibrate the absolute stress values. To isolate the stress

contributed by the geometric variation in crustal thickness, we run this model twice, once in the HET geometric

configuration with horizontal layers, and then again in the ISL configuration with the crust thickening toward

the east. After projecting the resultant stresses onto North-striking vertical strike-slip faults, we subtract the two

fields to obtain the geometric contribution to the change in Coulomb stress σC. This stress field exhibits a

strong gradient. It is 40 kPa higher on the June 21 fault than on the June 17 fault (Fig 6e). Combined with the

co-seismic stress imposed by the June 17 mainshock alone, this geometric interseismic effect produces a stress

field that is distinctly asymmetric (Fig 6f). Its threshold contour (10 kPa) includes more of the aftershocks

located between 3 and 5 km depth than for the other models.

Page 14

5152

424

425

426

427

428

429

430

431

432

433

434

435

436

437

438

439

440

441

442

443

444

445

446

447

448

449

450

451

452

453

454

5354


4. DISCUSSION AND CONCLUSIONS

4.1 Co-seismic effects

The slip distributions are significantly deeper in realistic three-dimensional geometric configurations than

in the conventional uniform half-space configuration. The slip diminishes to negligible values less than 50 cm

below 12 km depth, supporting the notion that large earthquakes in the SISZ can penetrate into the lower crust

[19]. The layered HET configuration is well suited for calculating coseismic deformation and stress. The

coseismic stress increase generated by the June 17 mainshock can explain the location of the June 21

mainshock, confirming a previous study [25].

The domino effect of cascading aftershocks following the June 17 aftershock also increases the stress on

the (future) location of the June 21 mainshock. Since static stress migrates at the speed of an elastic wave,

standard Coulomb calculations based on the June 17 mainshock or its aftershocks cannot explain the 86-hour

time delay between the two mainshocks.

4.2 Viscoelastic effects

The simple Maxwell viscoelastic model can explain the postseismic deformation between June 2001 and

May 2004 with viscosities of the order of ~10 EPa.s for both the lower crust and the upper mantle. Although the

uncertainties are large “plus or minus a factor” sounds weird, should be “times a factor” (an order of magnitude

at 68 percent confidence), the acceptable ranges are slightly higher and narrower using realistic geometric

configurations. For the first year, the GPS measurements of post-seismic displacements can be fit with a lower

viscosity, of the order of ~1 EPa.s, albeit with large uncertainties. We conclude that the available data provide

only weak constraints on subcrustal viscosity, partly because the poroelastic and viscoelastic effects have similar

amplitude.

4.3 Poroelastic effects

None of the poroelastic models fit the geodetic data very well, despite drastic increases in the complexity

of the models. It seems that the elastic approximation described in the Supplementary Material is insufficient to

describe the post-seismic observations in this case. Even with drastic increases in the number of free parameters,

this approach achieves only modest reductions in variance. Nor does it account for the location of many

aftershocks.

The explanation for this shortcoming probably involves time-dependent effects. The elastic

approximation described above is valid only after a very long time has elapsed and the fluid flow field has

returned to equilibrium (drained) conditions. The elastic approximation does not reproduce the map view of the

post-seismic interferogram very well, particularly in the fringe lobe southwest of the June 17 rupture. There, the

Page 15

5556

455

456

457

458

459

460

461

462

463

464

465

466

467

468

469

470

471

472

473

474

475

476

477

478

479

480

481

482

483

484

485

5758


half-space model with = 0.04 everywhere predicts about two times more uplift than is observed. To explain

this discrepancy, Jónsson et al. [32] suggest that the poroelastic rebound was faster in this area than elsewhere.

Indeed, if the hydrological state had returned to “drained” conditions in the two days following the June 17

mainshock, then the poroelastic signal would not appear in the interferogram beginning June 19. The water level

in the well at Hallstun (HR-19), just south of the June 17 rupture, lends some support to this interpretation. It

recovers faster than other wells, suggesting locally high permeability (and thus a large change in Poisson’s ratio

in this area. Yet our attempts to account for such a local heterogeneity do not fit the observed fringe pattern

much better than the uniform half-space model. We infer that time dependence, rather than spatial

heterogeneity, is contributing to the misfit. We conclude that the conditions for the static elastic approximation

to be a valid are not fulfilled. A complete time-dependent treatment, however, would require describing the

material parameters with more parameters to account for permeability, porosity, etc. [62, 66, 67].

4.4 Migrating seismicity

We have found that a combination of co-seismic and accumulated interseismic stress can explain the

tendency of seismicity to migrate from east to west across the SISZ during a single sequence of earthquakes.

This new result indicates that realistic geometry, including crustal thickening, is important for modeling

interseismic processes.

This mechanism does not, however, explain the 86-hour delay observed between the June 17 and June 21

mainshocks. The interseismic process is quite slow, increasing Coulomb failure stress at a given point VERB

MISSING HERE by only ~1 Pa per day. To explain the delay, we imagine a two-step process like the “coupled

poroelastic effect” that Scholz suggested to explain two successive earthquakes in the Imperial Valley,

California [68]. In the first step, the source shock instantaneously changes the pressure of pore fluids in the rock

around the receiver fault. In the second (slower) step, the hydrological flow field gradually returns to

equilibrium, increasing the pore pressure, unclamping the receiver fault and triggering the second earthquake.

Other examples of this two-step “unclogging” process have been invoked in California [29, 69, 70], Oregon

[71], and elsewhere [72, 73].

Over longer time scales, in the ~100 year intervals between sequences of earthquakes in the SISZ (e.g.,

1912 to 2000), inter- and post-seismic processes seem the most plausible explanation. Indeed, we have shown

that a linear viscoelastic rheology can increase the Coulomb failure stress by 10 kPa at distances longer than

10 km in only 4 years, as to be expected from the characteristic Maxwell time of 1 to 10 years approximated by

the ratio of a viscosity 1 to 10 EPa.s to a rigidity of 30 GPa. Longer time scales presumably involve more

complicated rheologies.

Page 16

5960

486

487

488

489

490

491

492

493

494

495

496

497

498

499

500

501

502

503

504

505

506

507

508

509

510

511

512

513

514

515

516

6162


ACKNOWLEDGMENTS

We thank Charles Williams for generously providing the source code for his revised version of

TECTON, as well as Frank Roth and his colleagues for the EDGRN/EDCMP and PSGRN/PSCMP codes. We

also thank Eric Hetland, Tim Masterlark, Rikke Pedersen, Kristin Vogfjord, Fred Pollitz, Sandra Richwalski,

Pall Einarsson and Herb Wang for helpful discussions. Grimur Bjornsson kindly explained the well data

graciously provided by Iceland Geosurvey and Reykjavik Energy. Kristin Vogfjord and Ragnar Stefansson

provided the relocated earthquake catalog estimated by the Icelandic Meteorological Office as part of the

PREPARED project. This work was financed by EU projects RETINA, PREPARED and FORESIGHT as well

as French national programs GDR STRAINSAR and ANR HYDRO-SEISMO.

SUPPLEMENTARY MATERIAL

SM.1 Testing against other solutions

To evaluate the accuracy of our finite-element modeling procedure, we compare its predictions to a

standard analytic solution for a rectangular dislocation buried in an elastic half-space, as formulated by Okada

[74] and implemented in the RNGCHN code [75]. Using the coseismic slip distribution estimated in a uniform

half space [15], we calculate the displacement field at the surface by both methods. The average difference is

less than 1 mm in all three components, thus validating our finite-element approach [52].

To “CONDITIONING” could be confusing here because in the finite element literature it always refers to

the conditioning of the matrix system to solve, so let us use another word increase our confidence in the

geometrically complex ISL configuration, we test its mesh in the TOM configuration. Like HOM, the TOM

configuration has a uniform rheology: Poisson’s ratio 0.28 and shear modulus 30 GPa throughout

the half-space. Like ISL, the TOM configuration uses a mesh that accounts for dipping layers with variable

thickness. The co-seismic surface displacement fields calculated using the HOM and TOM configurations differ

by less than 1 mm for all three components.

For the post-seismic viscoelastic models, we compare the surface displacement fields calculated using

two different methods. The first is our finite-element approach using the TECTON software, as described above.

The second is a semi-analytic Green’s function approach using the PSGRN/PSCMP software [76]. The average

difference in surface displacement is less than 3 mm for all three components after complete relaxation [52].

SM.2 Estimating co-seismic slip by joint inversion of GPS and INSAR measurements

To estimate the distribution of slip on the June 17 and June 21 fault planes, we use the same data set and

inversion algorithm as Pedersen et al. [15]. In this study, we allow more realistic geometric and rheologic

configurations by using the three-dimensional finite-element formulation to calculate the elastic Green’s

Page 17

6364

517

518

519

520

521

522

523

524

525

526

527

528

529

530

531

532

533

534

535

536

537

538

539

540

541

542

543

544

545

546

547

6566


functions, i.e. the partial derivatives of the measurable quantities (displacements) with respect to the model

parameters (slip values in discrete fault patches). This difference in approach entails a difference in the fault

parameterization. The dislocation formulation specifies the slip as a constant value inside each rectangular

rupture “patch” on the fault plane. The parameterization in the finite-element approach assigns a slip value to

each “split” node. In other words, the slip distribution is discontinuous in the dislocation formulation, but

continuous (piece-wise planar) in the finite-element formulation. As a result, the latter leads to smoother slip

distributions and smaller stress concentrations than the former. The finite element formulation also allows non-

zero slip at the uppermost node located at the air-rock interface, thus avoiding a singularity in the case of an

earthquake with surface rupture.

Another technical improvement involves choosing the appropriate amount of smoothing to optimize the

trade-off between fitting the data and finding a reasonable slip distribution. Here we choose the value of the

smoothing parameter that minimizes the sum of the absolute values of the slip parameters [67].

To evaluate the uncertainty of the slip estimates, we use a bootstrap algorithm [77]. Accordingly, we

resample Nb = 1000 times the data vector randomly according to its (diagonal) covariance. For each realization

of the data vector, we solve the inverse problem to estimate a population of the estimates m̂ for the model

parameters m. Then the vector of the standard deviations of the model parameter estimates is

σ (m) =(m̂i − E(m̂)[ ]

2

i=1

Nb

∑Nb − 1

(2)

where the E operator denotes the expected value.

SM.3 Approximating complete poroelastic relaxation by an elastic solution

Just after the co-seismic rupture, at time t+q, “undrained” conditions apply because the fluid in a

representative elementary volume has not yet reached equilibrium with an external boundary [62]. Presumably,

the same undrained conditions also apply during the brief time interval when a seismic wave propagates through

a representative elementary volume. According to previous studies [32, 60, 78], the Poisson’s ratio under

undrained conditions is

(tq+ ) = ν undrained B 0.31 (3)

At a much later time t∞, the fluids have reached equilibrium with an external boundary, deformation is

controlled by the solid matrix alone, and the rocks can be considered “drained”. Under these conditions,

Poisson’s ratio is [32, 60, 78]

(t∞ ) = ν drained B 0.27 (4)

Page 18

6768

548

549

550

551

552

553

554

555

556

557

558

559

560

561

562

563

564

565

566

567

568

569

570

571

572

573

574

575

576

6970


The differential post-seismic displacement accumulated in the interval of time between t+q and t∞ is thus

w u(t∞ −u(tq+ G raie

−G uraie⎡⎣ ⎤⎦u (5)

where G is the Green’s function relating co-seismic slip u on the fault plane to displacement u at the free

surface. We assume the geometric configuration and the slip u to be identical in both terms. Furthermore, we

assume that the rigidity is the same under drained and undrained conditions [55, 58, 62].

In the following, we attempt to determine reasonable values for . Under drained conditions, the value

of Poisson’s ratio is [62]

drained = 3ν − α B(1+ ν undrained )3 − 2α B(1+ ν undrained ) (6)

where we set the Biot-Willis parameter a = 0.28 and the Skempton’s coefficient B = 0.5 to find drained = 0.25.

Under undrained conditions, the value of Poisson’s ratio can be calculated from the ratio of P-wave to S-

wave velocities Vp/Vs estimated from seismological studies

uundrained =12

1−1

Vp

Vs

⎛⎝⎜

⎞⎠⎟

2

−1

⎛

⎝

⎜⎜⎜⎜⎜

⎞

⎠

⎟⎟⎟⎟⎟

(7)

In and around the SISZ, the Vp/Vs ratio is typically between 1.70 and 1.85, as measured by seismic tomography

[40, 79]. Bjarnason et al. [54] suggest a value of 0.29 for the undrained value for Poisson’s ratio in the SISZ.

REFERENCES

Page 19

7172

577

578

579

580

581

582

583

584

585

586

587

588

589

590

591

592

7374


Captions

Fig 1. Map showing (above) the main tectonic features of the study area from Árnadóttir et al. [23]. The

Reykjanes Peninsula (RP), the Hengill triple junction (He), the Western Volcanic Zone (WVZ), the South

Iceland Seismic Zone (SISZ), and the Eastern Volcanic Zone (EVZ) are indicated. The light shaded areas are

individual fissure swarms with associated central volcanoes. Mapped surface faults of Holocene age are shown

with black lines [80]. The epicenter locations of earthquakes on June 17, 2000 are shown with black stars. The 17

June main shock is labeled J17. The Reykjanes Peninsula events are 26 s afterwards on the Hvalhnúkur fault (H),

30 s afterwards near Lake Kleifarvatn (K), and 5 minutes afterwards at Núpshlídarháls (N) as located by geodesy

[23] and timed by seismology [6]. Location of the 21 June 2000 main shock is shown with a white star labeled

J21. The location of Reykjavik, the capital of Iceland, is noted. Inset shows a simplified map of the plate

boundary across Iceland, with the location of the main part of the figure indicated by the black rectangle. The

black arrows show the relative motion between the North America and Eurasian plates from NUVEL-1A [2, 3].

Block diagram (below) showing thickening of the crust from east to west across the SISZ as inferred from the

literature [43, 44]. The dimensions of the mesh are 300 x 300 x 100 km (whole box). Color code shows Iceland

(gray), the SISZ (blue rectangle), water (blue), upper crust (green), lower crust (orange), the upper mantle (red),

fault traces (red) for the June 17 (east) and June 21 (west) faults, and the reference point (gray dot) at the center

of the modeled June 21 fault trace (63.99oN, 20.71oW) for the local easting and northing coordinate system.

Fig 2. Geometric configuration of rheologic parameters. A) Cross section showing, as a function of depth, the

following SENTENCE NOT FINISHED HERE.rheologic parameters: density r (thick dashed line), rigidity

(thick solid line), and Poisson’s ratio (thin solid line). The thicknesses of the rheologic layers depend on the

depths Hu and Hc to the bases of the upper and lower crust, respectively. The configurations are composed of: a

uniform half space (HOM and TOM), horizontal layers with constant thickness (HET), dipping layers with

variable thickness (ISL), a half-space with a damage zone (HOF), and horizontal layers with a damage zone

(HEF). (B) Thickness of the upper crust Hu as parameterized in the TOM and ISL configurations. Contour

interval is 1 km. (C) Depth to the interface between lower crust and upper mantle Hc in the TOM and ISL

configuration. Contour interval is 5 km. Solid lines represent traces of modeled faults for June 17 (east) and June

21 (west). Dot indicates the reference point of the local cartographic coordinate system, centered on the June 21

epicenter.

Fig 3. View of fault planes showing the slip distribution estimated for the June 17 main shock (left column) and

June 21 main shock (right column) in the horizontally-layered HET configuration (upper row). White stars

indicate the hypocenters of the mainshocks. Black dots indicate aftershocks within 1 km of each fault and 24

hours of the June 17 and June 21 mainshocks, respectively, as relocated by the Icelandic Meteorological Office,

Page 20

7576

593

594

595

596

597

598

599

600

601

602

603

604

605

606

607

608

609

610

611

612

613

614

615

616

617

618

619

620

621

622

623

624

7778


as described in the section on data and methods. The lower panels show the 1-σ uncertainty of the slip estimates

for the June 17 (left) and June 21 (right) events.

Fig 4. View of fault planes of the June 17 main shock (left column) and June 20 main shock (right column)

showing the differences in estimated slip distribution for the HET configuration with respect to HOM (top row),

HOF with respect to HOM (middle row), and HEF with respect to HET (bottom row). The black lines denote the

significance ratio, defined as the slip estimate divided by its 1-σ uncertainty. The contour where this ratio equals

2 thus includes the area where the differences are statistically significant at the level of 95% confidence. For each

panel, the white circle indicates the centroid of the first slip distribution in the pair, while the black circle

indicates the centroid of the second, e.g., white circle is the HET centroid in the first row.

Fig 5. Poroelastic models for a postseismic interferogram, showing range change (the component of

displacement along the line of sight from the ground to satellite) in map view as observed (a) and calculated from

the FOR1 configuration (d). Solid black lines outline the co-seismic portion of the fault plane that ruptured

during the June 17 mainshock. Dashed lines indicate locations of profiles. Panel (b) shows negative range change

as a function of distance along east-west profile A-A’ as observed (points), and calculated from various

configurations: FOR3 (black line), TRY1 (green line), INV2 (blue line), and INV4 (red lines). Panel (c) shows

negative range change as a function of distance along north-south profile B-B’ with the same plotting

conventions. All range changes pertain to the first month of the post-seismic interval, from June 19 to July 24.

Negative range change corresponds to uplift; positive range change corresponds to subsidence.

Fig 6. Change in Coulomb failure stress σC as a result of four candidate processes for stress transfer. The

contours outline the areas where the change in Coulomb failure stress σC resolved on north-striking vertical

faults at 4 km depth exceeds the conventional threshold value of σC(critical) =10 kPa. The sources of stress

include: (A) the co-seismic stresses generated by the June 17 mainshock calculated using configurations HOM

(black line) and HET (blue line); (B) poroelastic effects calculated using the elastic approximation and two

different configurations for the change in Poisson’s ratio : INV2 (green line) and INV4 (blue line); (C) the

domino effect of cascading aftershocks, (D) viscoelastic processes accumulated in mid 2004 calculated from the

HET (blue line) and ISL (green line) configurations (E) geometric contribution to 100 years of interseismic strain

accumulation obtained by subtracting the interseismic stress field calculated using the HOM geometric

configuration from that obtained with the ISL configuration; (F) combination of the co-seismic and the geometric

contribution to interseismic stress integrated over 100 years, calculated as the sum of the HET field (A) and panel

(E). Red lines indicate faults for the June 17 (east) and June 21 (west) mainshocks. Dots in panels (A, B, C, E, F)

indicate the epicenters of earthquakes between June 17 and 21, 2000 [16, 49].

Page 21

7980

625

626

627

628

629

630

631

632

633

634

635

636

637

638

639

640

641

642

643

644

645

646

647

648

649

650

651

652

653

654

655

8182


TABLES

Table 1. Geometric and rheologic parameters for configurations used to estimate distributions of co-seismic slip.

Table 2. Parameter values for poroelastic modeling of post-seismic deformation and the variance reduction. In

the FOR and TRY series, the parameter values are input to forward calculations. In the INV series, the

parameters are estimated from the data shown in Fig 5.

LAST PAGE

Page 22

8384

656

657

658

659

660

661

662

8586

Documents

Estimating earthquake source parameters€¦ · Web viewFile name = DuboisFeiglKomatitschEPSL20061003b.doc. Application name Microsoft Word ... seismic slip by joint inversion