Modelling of Yingxiu–Beichuan fault zone based on refined DInSAR data of 2008 Wenchuan earthquake

    Modelling of Yingxiu-Beichuan fault zone based on refined DInSAR data of 2008 Wenchuan earthquake Meng-Che Wu, Jianguo Liu, John Cosgrove, Philippa Jane Mason, Hongshi Yan, Wen-Yen Chang PII: DOI: Reference:

S0040-1951(14)00288-1 doi: 10.1016/j.tecto.2014.05.024 TECTO 126326

To appear in:

Tectonophysics

Received date: Revised date: Accepted date:

30 August 2013 18 May 2014 20 May 2014

Please cite this article as: Wu, Meng-Che, Liu, Jianguo, Cosgrove, John, Mason, Philippa Jane, Yan, Hongshi, Chang, Wen-Yen, Modelling of Yingxiu-Beichuan fault zone based on refined DInSAR data of 2008 Wenchuan earthquake, Tectonophysics (2014), doi: 10.1016/j.tecto.2014.05.024

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.

ACCEPTED MANUSCRIPT MODELLING OF YINGXIU-BEICHUAN FAULT ZONE BASED ON REFINED DINSAR DATA OF 2008 WENCHUAN EARTHQUAKE

RI

SC MA

NU

Corresponding author. Tel.: +886 38633310 E-mail address: [email protected]

D

*

College of Environmental Studies, National Dong Hwa University, No. 1, Sec. 2, Da Hsueh Rd., Shoufeng, Hualien 97401, Taiwan

TE

b

Department of Earth Science and Engineering, Imperial College London, South Kensington Campus, London SW7 2AZ, UK

AC CE P

a

PT

Meng-Che Wua,b *, Jianguo Liua, John Cosgrovea , Philippa Jane Masona, Hongshi Yana and Wen-Yen Changb

1

ACCEPTED MANUSCRIPT Abstract Differential interferometric synthetic aperture radar (DInSAR) provides an essential constraint for

PT

forward modelling of seismic faulting. However, the crucial evidence of deformation along the very

RI

near vicinity of a seismic fault is often missing because chaotic ruptures destroy the InSAR signal

SC

coherence; such as the case of the Yingxiu-Beichuan fault zone of the 2008 Wenchuan earthquake. We developed an Adaptive Local Kriging (ALK) interpolation method to produce a refined DInSAR

NU

dataset which continues without the interruption of decoherence gaps. Thus the 3D configuration of the seismic faulting can be better constrained for forward modelling using the Poly3D software. Through

MA

parameter fine tuning in a sequence of trial and error simulation, the best fitting between the simulated interferogram and the ALK refined DInSAR interferograms is achieved using a structural model of the

TE

D

Yingxiu-Beichuan fault zone which comprises a total of 14 segments with dip angles varying from 20º60º, in the south-west, to 40º-70º, in the north-east; some major fault segments are composed of multiple

AC CE P

planes with varying dip angles. The model reveals that in the much disputed Beichuan–Pengguan area, the Yingxiu-Beichuan fault links with the Pengguan fault (which dips to the NW at ~30º) at a depth of ~13km. These two structures are interpreted as representing the foreland propagation of the YingxiuBeichuan fault by footwall collapse, a process that, in the vicinity of the Pengguan fault, involves the transfer of displacement from the older, Yingxiu-Beichuan fault to the younger Pengguan fault. Keywords: Wenchuan earthquake DInSAR (Differential interferometric synthetic aperture radar) Adaptive Local Kriging Yingxiu-Beichuan fault zone

2

ACCEPTED MANUSCRIPT Forward modelling

Introduction

PT

1.

A devastating earthquake of magnitude Mw 7.9 hit the Longmenshan region of Sichuan Province,

RI

China on 12th May 2008 and caused a great many casualties and economic damage. The epicenter area

SC

is located on the eastern edge of the Tibetan Plateau, where it is cut by the NE-SW trending trace of the Longmenshan thrust belt which has a length of ~300 km (Burchfiel et al. 1995; Densmore et al., 2007;

NU

Burchfiel et al. 2008). Fig. 1 shows the location of the area affected by the Wenchuan earthquake and

MA

the two major seismic faults in the Yingxiu-Beichuan fault zone, e.g. the Yingxiu-Beichuan fault (YBF) and the Pengguan fault (PF), as presented in many publications (Burchfiel et al. 2008; Chen and

D

Wilson, 1996; Dong et al., 2008; Hubbard and Shaw, 2009). There is a considerable dispute regarding

TE

the relationship between these two faults. Some publications indicate that the YBF is the main fault, in

AC CE P

the section where the two faults are parallel and that it merges with the PF at a depth of about 10 km and extends to depth of about 15 km below the land surface (Li et al., 2010; Shen et al., 2009; Xu et al., 2010; Tong et al., 2010). Other authors (Yu et al., 2010; Furuya et al., 2010; Li et al., 2009) suggest that in the section of the fault zone where the two faults exist (Fig. 1), the PF is the main fault and it merges with the YBF at a depth of about 13 km and then extends to a depth of about 15 km below the surface. Cross-event PALSAR (Phased Array type L-band Synthetic Aperture Radar) data acquired by ALOS (Advanced Land Observing Satellite) were supplied by JAXA (Japan Aerospace Exploration Agency), which enabled the study of co-seismic deformation using the DInSAR technique (Tong et al., 2010; Furuya et al., 2010; Ge et al., 2008; Hashimoto et al., 2009). Forward modelling using DInSAR as a constraint is an effective technique for determining the geometry of the seismic faulting. However, it is often hindered by the fact that the crucial data relating to the deformation in the vicinity of the seismic 3

ACCEPTED MANUSCRIPT fault are lost because chaotic ruptures destroyed the InSAR signal coherence in these areas. In this paper, we present the forward modelling of the Yingxiu-Beichuan fault zone for the 2008 Wenchuan

Adaptive Local Kriging (ALK) for DInSAR data refinement and

RI

2.

PT

earthquake constrained by the refined DInSAR data.

SC

interpretation of radar Line-Of-Sight (LOS) motion

NU

A total of six paths of PALSAR data (Path 471–Path 476) covering the entire Longmenshan thrust belt were processed using the Repeat Orbit Interferometry Package (ROI_PAC) software (Rosen et al.,

MA

2004) developed at the Jet Propulsion Laboratory (Caltech/JPL). The details of PALSAR data are shown in Table 1. For each path, the pre-event image is the master image while the post-event the slave

TE

D

in the DInSAR processing.

As shown in Fig. 2, in the area along the seismic fault zone, the coherence between cross-event SAR

AC CE P

images is completely lost because of the earthquake induced violent and chaotic destruction of the land surface. Consequently no surface displacement can be measured using the DInSAR technique. In our previous work (Wu et al., 2013), an Adaptive Local Kriging (ALK) technique incorporated in a multistep processing procedure was developed to retrieve the interferometric fringe patterns in the decoherence zone with high fidelity. The ALK is a modification of Ordinary Kriging, which predicts unknown values from data observed at known (sampled) locations based on a semivariogram model that expresses the spatial variation of the data. The interpolation is made by minimizing the ‘deviation’ in the predicted values (i.e. variance about the mean) which are estimated from known values weighted by their spatial distribution (Wang, 1990; Johnston et al, 2003; Gudmundsson et al., 2002). Base on the principle of the Ordinary Kriging and in consideration that deformation trend across the seismic fault zone should intensify in the opposite direction according to basic faulting mechanism, we divided the

4

ACCEPTED MANUSCRIPT DInSAR unwrapped interferogram images into two parts along the interpreted YBF line (Burchfiel et al., 2008; Chen and Wilson, 1996; Dong et al., 2008; Hubbard and Shaw, 2009), as the longer red line

PT

shown in Fig. 2. Large part of the fault line is within the decoherence zone and separating the original

RI

DInSAR data based on this fault line acts as a priori to constrain the data interpolation. However, the Ordinary Kriging based on a global semivariograms (Johnston et al., 2003) derived from the

SC

interferograms partitioned at the fault line cannot be representative to the strong directional trends of

NU

increasing gradients towards the seismic fault (Wu et al., 2013). The ALK method was then designed, which uses an adaptive scanning window to calculate the local semivariogram at every DInSAR data

MA

pixel. It thus follows the deformation trend of the original DInSAR data with a series of local linear semivariograms to interpolate and recover the DInSAR data in the decoherence zone. Further, a multi-

D

step processing procedure was developed to first carry out ALK separately on the two parts (hanging

TE

wall and footwall) of the DInSAR data and then apply ALK again to reprocess the merged ALK

AC CE P

refined DInSAR data of the two parts. Thus the artificial fault line between the hang wall and footwall datasets is smoothly removed.

There are, however, block shifts and discontinuities between the ALK data stripes of different paths, which are in general caused by satellite orbit drifts and differing acquisition times between imaging paths of interferometric pairs of the SAR data. Empirical linear adjustments were applied to shift and tilt the adjacent stripes of ALK data in the range and azimuth directions (Chini et al., 2010). This post processing stitches the ALK data of all the paths, with minimized inter-path displacement and discontinuity. There is no interpolation between tracks to maintain the fidelity of the ALK data. A final ALK refined DInSAR data is then produced, as shown in Fig. 3, which matches the original interferogram precisely in the areas of good quality fringe patterns while in decoherence gaps, continuous fringe patterns are filled in based on the local trend. This final product of the whole ALK

5

ACCEPTED MANUSCRIPT interpolation procedure shows the deformation trend across the seismic fault zone without artificial discontinuity and with gradients in opposing directions away from the fault. In fact, the independent

PT

operation of the ALK for hanging wall and footwall datasets ensures the retrieval of the correct

RI

deformation trend on opposing sides of the fault line while the final ALK applied to the merged data makes the initially interpreted fault line nearly irrelevant in the final data retrieval step as shown by the

SC

profiles in Fig. 4 in which the LOS minima generally follows the fault line but not always especially in

NU

the area where the YBF and PF are parallel.

For comparison, in the original DInSAR interferograms, the seismic rupture zone is largely

MA

composed of decoherence pixels (Fig. 2(a)) from which estimation of the surface deformation is

D

impossible and thus have to be masked off as gaps for unwrapping process (Fig. 2(b)). In contrast, the

TE

ALK DInSAR interferograms in Fig. 3(a) is smooth and continuous and shows very dense fringe patterns in the decoherence gaps indicating that the magnitude of the surface displacement in the

AC CE P

immediate vicinity of the Yingxiu-Beichuan fault zone is significantly greater than in the adjacent area. The ALK DInSAR data maintains high fidelity to the original dataset. This can be seen directly by noting that in the region outside decoherence zones, the fringe patterns presented in both datasets are nearly identical. The matching between the original and ALK DInSAR data along profiles across every path of PALSAR data is nearly perfect as shown in Fig. 4 and the correlation between the two datasets is higher than 0.99. Moreover, it is obvious to notice that there are some spikes of the red lines in Fig. 4(g), which is caused by the phase unwrapping errors during the DInSAR processing. The robust ALK processing is able to remove these spikes to ensure that the interpolation data follows the actual deformation trends from the original dataset, as the green thin line shows in Fig. 4(g). The high fidelity is also proved by rigorous tests in the data retrieval of the dense fringe patterns which are artificially masked off (Wu et al., 2013).

6

ACCEPTED MANUSCRIPT The ALK DInSAR data as a refined map of coseismic deformation in the radar Line Of Sight (LOS) are not adequate to fully characterize the 3D deformation in term of unit vector in E/N/U (Easting,

PT

Northing and Upward) along the seismic fault zone of Wenchuan earthquake. There is an ambiguity

RI

between the vertical motion and horizontal motion in ground range direction. The ALOS PALSAR images used in this study were taken in ascending orbits looking east and therefore, as illustrated in

SC

Fig. 5, the eastward component of horizontal terrain motion along the NE trending Yingxiu-Beichuan

NU

fault will produce a positive phase difference (LOS displacement away from the satellite) while an uplift of the hanging wall of the fault will produce a negative phase difference (LOS displacement

MA

toward the satellite) (Agustan, 2010). Thus the section of uplift thrust to SE direction may produce either positive, negative or zero LOS displacement depending on the balance between the eastward

TE

D

motion and uplifting.

As illustrated in Fig. 6, if the first observation, O1, is moved eastward to the second observation, O2,

AC CE P

it will produce a positive LOS. The magnitude of the eastward motion is equal to the ground range displacement in this case. However, to keep the magnitude of the positive LOS fixed with some uplift motion (

) involved (observation O3), the magnitude of the eastward motion has to be increased to

the position O3’, the projection from O3 to the surface. Obviously the actual eastward motion magnitude (

) is significantly greater then and ground range displacement. Now, keep the

and let the LOS approach 0, the magnitude of

has to increase to

fixed

as,

, then

(1)

Thus, according to Fig. 6, given the nominal look angle of the SAR, θ, we have (2) while,

7

ACCEPTED MANUSCRIPT (3) The formula (3) indicates that to produce a positive LOS, the uplift motion must be less than times of eastward motion. For L-band PALSAR, ≈39°,

PT

. This means that under the

RI

condition of uplifting, any positive LOS must involve significant eastward motion.

SC

We therefore can be sure that any positive LOS on the uplifting hanging wall of the NE trending YBF must involve considerable eastward motion while on the vice versa for the negative LOS on the

NU

subsiding footwall. On the other hand, the negative LOS on the hanging wall definitely indicates

MA

uplifting just like the positive LOS on the footwall indicates subsiding. Observing the ALK DInSAR data in Fig. 3(b), the hanging wall as whole involved significant eastward motion toward the YBF fault

D

line presented in positive LOS and near the fault line, in the region from the epicenter Yingxiu to old

TE

Beichuan town, this motion is transformed to significant uplift shown as negative LOS. Field photo

AC CE P

(Fig. 7(a)) took in Hongkou town very near the epicenter confirm the motion of the seismic faulting in this area is mainly thrust uplift with minor right lateral slip. Besides ALK refinement of the DInSAR data, the cross-event PALSAR amplitude images covering the entire Longmenshan thrust belt were processed using the Phase Correlation based Image Analysis System (PCIAS) to measure the pixel offset field induced by the earthquake in the ground range and azimuth directions. The PCIAS is developed by our research group and it is capable of estimating pixel-wise disparity (pixel offset) at accuracies of up to 1/50th of a pixel (Liu and Yan, 2006 and 2008). The ground range and azimuth offset maps were then combined to produce the horizontal offset vector fields, as illustrated in Fig. 8 for a few key areas of the Wenchuan earthquake (Wu et al., 2012). Bearing in mind the analysis before based on Fig. 5 and 6, the ground range offset is subject to the same ambiguity as the LOS and we can the appreciate that Fig. 8(b) indicates strong thrust toward SE direction almost perpendicular to the YBF in the Beichuan area, as shown by the large magnitude offset 8

ACCEPTED MANUSCRIPT vectors on the hanging wall. While on the footwall, the horizontal offset vectors are almost parallel to the YBF, pointing from NE to SW, indicating dextral slip motion. The YBF in the old Beichuan town

PT

area is therefore dominated by thrusting with some right-lateral slip motion. In Yingxiu area including

RI

the epicentre of the main shock shown in Fig. 8(c), hanging wall moved significantly in SE direction, perpendicular to the YBF, while the footwall moved little roughly in the same direction. The net

SC

motion in this section of the YBF is therefore dominantly thrust as confirmed by field observation (Fig.

NU

7(a)). Fig. 8(d) reveals the motion in the area of the oblique-slip XF (Xiaoyudong fault), the horizontal motion on the SW side of the fault is mainly from NW to SE and is roughly parallel to the fault line

MA

resulting in a left-lateral slip (Fig. 7(b)). However, as the SW side of the XF is in the footwall of the YBF, the eastward motion of the offset vectors could also be resulted from the subsidence of the area.

D

In contrast, the NE side of the XF is in the hanging wall of the PF and the relatively low values of

TE

ground rang pixel offset could be the result of significant uplift (ref Fig. 6) that agrees with the field

AC CE P

measurement of up to 3.4m uplift (Xu, et al., 2009). For further understanding the seismic faulting of the Wenchuan earthquake, a sequence of forward modelling constrained by the ALK DInSAR and sub-pixel offset data has been carried out.

3.

Fault modelling

The ALK DInSAR data present continuous coseismic deformation in LOS and the horizontal offset data indicate the faulting motions along the seismic faults, thus both data provide refined constraint for forward modelling of the fault geometry and slip motion of the Yingxiu-Beichuan fault zone. We used the Poly3D software for the modelling in a sequence of trial and error experiments. The Poly3D calculates the displacements induced in an elastic whole or half-space using planar, polygonal-shaped elements of displacement discontinuity in a forward modelling simulation (Thamos, 1993).

9

ACCEPTED MANUSCRIPT For DInSAR simulation, Poly3D provides the simulated surface displacement in E/N/U directions. Accordingly, the E/N/U components of surface displacement can then be converted to the LOS

PT

displacement based on the unit vector (Fialko et al., 2001), as following:

RI

sin( ) sin( )  u E

SC

sin( ) cos( )  u N

NU

 cos( )  u U

(5) (6) (7)

MA

LOS  E  u E  N  u N  U  u U

(4)

where  is the radar look angle and  is the azimuth angle. Based on (7), all the modelling results will

D

be converted to the LOS displacement in order to compare with the ALK DInSAR data.

TE

Based on fault line interpreted from the ALK DInSAR data (Fig. 3(b)) and with reference to the

AC CE P

GDEM (Global Digital Elevation Model) (Tachikawa et al., 2011; Fujisada et al., 2012), the forward modelling was initiated by dividing the YBF, PF and XF into 14 major fault segments with varying strike angles, as shown in Fig. 9(a) and (b). The maximum bottom depth was assumed in a range between ~17 km to ~10 km from SW to NE based on relocated aftershocks data from the CDSN and USGS (Li et al., 2010; Huang et al., 2008). The slip magnitude and direction for each segment is based on the horizontal pixel offsets and the averaged field observation data (Xu et al., 2009; Yu et al., 2010) and previous inversion models (Shen et al., 2009; Xu et al., 2010; Hashimoto et al., 2009; Tong et al., 2010; Furuya et al., 2010). Controversy exists regarding the fault dip angles of the YBF and PF. The seismic reflection profiles presented in Li et al. (2010) indicate that the dip angle of the YBF is ~45° in the SW and becomes shallower toward the NE (~30°). However, the published simulation results, which are based on variety

10

ACCEPTED MANUSCRIPT of models, suggest that the fault dip steepens along strike from SW towards the NE reaching ~70° in the northeast-most segments of the YBF (Shen et al., 2009; Xu et al., 2010; Hashimoto et al., 2009; Tong

PT

et al., 2010; Furuya et al., 2010). In the numerical modelling, we considered both scenarios of dip

RI

angles for the NE segments of the YBF. The forward modelling was initiated with dip angles based on the seismic reflection profiles and then in a trial and error process, the shapes of the fault segments

SC

were modified with their 3D geometry and faulting parameters fine turned to achieve an optimal fit

NU

between the simulated fringe patterns generated by the Poly3D and the ALK DInSAR interferograms. In the following sections, three representative models are presented together with our interpretation

Model 1: Yingxiu-Beichuan fault is the main fault with a single dipping plane for each

D

3.1

MA

of the Yingxiu-Beichuan fault zone.

TE

fault segment

AC CE P

The most complicated segment of the Yingxiu-Beichuan fault zone is the area where the YBF and PF are parallel. Most publications indicate that the NE-SW trace of the YBF locally takes a sharp kink into an E-W direction at the position where the north-eastern tip of the PF emerges (Fig. 1). The PF extends to the SW, parallel to the YBF and its trace takes the position that the YBF would occupy if the kink did not exist (Burchfiel et al. 2008; Chen and Wilson, 1996; Dong et al., 2008; Hubbard and Shaw, 2009). We take this interpretation as the starting point for the forward modelling using the Poly3D. As shown in Fig. 9(a) and (b), the fault zone is divided into 14 segments. Ten segments represent the YBF, 3 the PF and 1 the XF. The dip angle for each fault segment is based on the seismic reflection profiles which indicate gentle dip angles for the entire YBF (Li et al., 2010). The details of the fault model settings are given in Table 2. The key configuration is that the PF segments P1-P3 are intercepted by the YBF segment B3 at a depth of about 10 km, as illustrated in the inset in Fig. 9(a). After fine tuning the fault parameters, the simulated differential interferograms are shown in Fig. 10(b). 11

ACCEPTED MANUSCRIPT In general, the fringe patterns are similar to the ALK DInSAR interferograms but there are clear discrepancies. The key observations from SW to NE along the fault zone are:

PT

1) As shown in the ALK DInSAR interferograms (Fig. 10(a)), the fringe patterns in the section of

RI

the hanging wall of B1–B3, in the SW section of the YBF, comprise an elongated two-ring

SC

pattern on the NW side of the hanging wall of the B1 segment (the area NW to the dense fringe patterns) and a circular one-ring pattern on the hanging wall of the B3 segment; these two

NU

patterns are enclosed by a larger ring extending from the SW end of the YBF to B3 segment. While the simulated fringe patterns in the corresponding area comprise two elongated multiple-

MA

ring patterns which are not enclosed by a larger ring, as shown in Fig. 10(b).

D

2) The dense fringe pattern belt corresponding to B1, at the SW end of the YBF in the ALK

AC CE P

interferogram.

TE

DInSAR interferograms, is too narrow and does not extend to the fault line in the simulated

3) The dense fringe patterns between the parallel segments of YBF (B3) and PF (P1-P3) shown in the ALK DInSAR interferograms (Fig. 10(a)) are largely absent from the simulated interferogram (Fig. 10(b)).

4) The gentler dip angles used for segments B8 and B9, in the NE section of the YBF, produced rectangular fringe patterns in the hanging wall and as whole, there is only one fringe appearing in the footwall in the simulated interferogram. These present a poor match to the ALK DInSAR interferograms. 5) The dense multi-ring circular fringe pattern shown in the ALK DInSAR interferograms, at the NE end of the YBF and corresponding to B10 in the model, is not shown in the simulated interferogram.

12

ACCEPTED MANUSCRIPT 3.2

Model 2: Yingxiu-Beichuan fault is the main fault with variable dip angles at depth for the major fault segments and with steepening dip angle toward the NE segments of the YBF

PT

To reduce the major discrepancies between the predicted and observed interferograms observed in

RI

Model 1, this model is modified as follows (Fig. 9(c) and (d)):

SC

 The major segments of YBF (B1-B7) and PF (P1-P3) are re-configured as multiple planes with different dip angles to simulate a curved fault plane i.e. steeply inclined near the surface

NU

and more gently inclines at depth. The details of the 26 segments and sub-segments are listed

MA

in Table 3.

 The dip angles in the NE segments of the YBF (B8-B10) were set steeper according to the

D

models proposed by some researchers (Shen et al., 2009, Xu et al., 2010, Hashimoto et al.,

TE

2009, Tong et al., 2010 and Furuya et al., 2010).

AC CE P

 Segment B10 at the NE end of the YBF (Qingchuan area) is treated as a blind thrust at a depth of 5 km rather than an apparent strike slip fault. The simulated interferograms for this model in Fig. 10(c) show a considerable improvement in fitting between the ALK DInSAR interferograms (Fig. 10(a)) and that generated by the model. In particular, the dense circular fringe pattern in the hanging wall of the YBF at its NE end, shown in the ALK DInSAR interferograms, is well simulated by the model. The setting of B10 as a blind thrust is the only fault configuration that can produce a multi-ring circular fringe pattern to match such a pattern in the ALK DInSAR interferograms in this area. The dense fringe pattern belt corresponding to B1 is widened to reach the YBF fault line as a result of variable dip angle of this segment. However, the new configuration of variable dip angles for segments B2–B3 and P1–P3 used in this model is ineffective to recover the dense fringe patterns between the YBF and the PF. Instead, dense 13

ACCEPTED MANUSCRIPT fringe patterns are introduced in the B3 segment area on the hanging wall of the YBF along the NW side of the fault line, making the fringe patterns in this area less like the ALK DInSAR interferograms.

PT

The discrepancy of the ring patterns described in observation 1 of Model 1 is improved in Model 2 but

RI

extra fringe rings are introduced in the hanging wall of segments B1–B3 i.e. in the SW section of the YBF. Our experiments with various modifications of parameters under the basic model in which the

SC

YBF is the main fault are all unable to recover the dense fringe patterns between the YBF and the PF.

3.3

NU

We therefore have to consider changing this basic setting.

Model 3: Pengguan fault is the main fault and that the major fault segments have variable

MA

dips and with steepening dip angle toward the NE segments of the YBF

D

Although in the majority of publications on this fault zone, the YBF is taken to be the main fault,

TE

other researchers, for example (Li et al., 2009; Furuya et al., 2010; Yu et al., 2010), argued that the PF is currently the main fault rupture. Our observation and measurements from the ALK interferograms

AC CE P

and ALK DInSAR data, in particular the trace of local negative maxima of the LOS displacement in the ALK DInSAR data in Fig. 3(b) and Fig. 4, all support this interpretation, i.e. the PF is the main rupture fault. We have therefore further modified Model 2 so that the YBF is intercepted by the PF at a depth about 13 km as illustrated in the inset C2 in Fig. 9(e). In this model, segment B3 is much less steeply inclined than before and it is thus configured as a single dip angle plane. Other parts of the model are largely unchanged and in total there are 25 segments and sub-segments, as shown in Fig. 9(e), (f) and listed in Table 4. The fitting between the Model 3 simulation shown in Fig. 10(d) and the ALK DInSAR interferograms in Fig. 10(a) is a clear improvement on Model 2 as outlined below. 1) The dense fringe patterns between the parallel segments of the YBF (B2–B3) and PF (P1–P3) shown in the ALK DInSAR interferograms are very well simulated. 14

ACCEPTED MANUSCRIPT 2) The ALK DInSAR fringe patterns (Fig. 10(a)) in the section of the hanging wall of B1–B3 in the SW section of the YBF, as described in the observation 1 of Model 1, are much better

PT

simulated in Model 3 which generates an oval fringe pattern composed of two rings in the

RI

hanging wall of segment B1 (the area NW of the fault trace and the dense fringe patterns) and an elongated ring pattern on the hanging wall of segment B3. These two patterns are

SC

enclosed by a larger ring which, as shown in Fig. 10(d), extends from the SW end of the

NU

YBF to segment B3.

3) The low density fringe patterns along the SE side of the Yingxiu-Beichuan fault zone on the

MA

footwall are also better simulated and they closely resemble the ALK DInSAR

D

interferograms.

Discussion

4.1

Assessment of the modelling results

AC CE P

TE

4.

The comparison between the simulation results of the three models and the DInSAR fringe patterns indicates that the best fitting is the Model 3 when the PF is set as the main fault that intercepts the YBF at a depth of about 13 km. However, the depth of the interception between the PF and the YBF in Model 2 (segment P2_1 and P2_2 in Fig. 9(c), (d)) and in Model 3 (segment P2_1, P2_2 and P2_3 in Fig. 9(e), (f)) is inconsistent. The depth for Model 2 is set as about 10 km below the land surface; while that in Model 3, it is set as about 13 km below the land surface. Therefore, the comparison between the two models may not be entirely fair. For a more fair comparison, an extra model, Model 4, was carried out that makes the segment P2_3 in the Model 3 to become part of the YBF, segment B3_2 and other setting remain the same. As such, the YBF in the Model 4 is the main rupture fault which intercepted

15

ACCEPTED MANUSCRIPT the PF at a depth of about 13 km below the land surface. The fault geometry of the Model 4 is shown in Table 5 and Fig. 9(g), (h).

PT

Fig. 10(e) shows the modeled interferogram of the Model 4 which is generally similar to that of the

RI

Model 3 (Fig. 10(d)) however, with the basic setting of the YBF being the main fault, the Model 4 is

SC

unable to recover the dense fringe patterns in the area between the YBF and the PF. There are also discrepancies in the section of the hanging wall of B1–B3, in the SW section of the YBF where the

NU

fringe patterns presented in the ALK DInSAR interferograms (Fig. 10(a)) are less well simulated in the

MA

Model 4 than in the Model 3. The ALK DInSAR interferograms (Fig. 10(a)) and the Model 3 (Fig. 10(d)) presents an oval fringe pattern composed of two rings in the hanging wall of segment B1 (the

D

area NW of the fault trace and the dense fringe patterns) and a ring pattern on the hanging wall of

TE

segment B3. These two patterns are enclosed by a larger ring shown in both Fig. 10(a) and Fig. 10(d), extending from the SW end of the YBF to segment B3. While the fringe patterns in the same area in

AC CE P

Model 4 present only one oval ring pattern in the hanging wall of the segment B1 and enclosed by a larger ring on the hanging wall of segment B3 (Fig. 10(e)). The simulation result of Model 3, in comparison with those of Model 1, 2 and 4 confirms that the DInSAR fringe patterns can be better simulated when the PF is set as the main rupture fault in the areas where the YBF and PF are parallel. In this interpretation, the PF links directly with the YBF, at both its NE and SW tips. It forms part of the main rupture fault zone of the Wenchuan earthquake, as denoted by the solid line in Fig. 10(d). This means that the PF is actually part of the YBF while segment B3 of the YBF, which is parallel to the PF is possibly a secondary fault splay that joins the YBF at 13 km depth. In the original DInSAR interferograms (Fig. 2(a)), the area in between the two parallel faults lost coherence completely indicating intense and chaotic surface rupturing that is evident in field observations (Xu et al., 2009; Yu et al., 2010; Lin et al., 2009). The dense fringes recovered by ALK in 16

ACCEPTED MANUSCRIPT this zone (Fig. 3(a)) reveal the large magnitude of coseismic deformation which is significantly greater than the deformation in the areas SE of the PF and NW of the YBF. If we consider this zone is the

PT

footwall of the YBF and hanging wall of the PF, and if the YBF is the main fault which goes to a

RI

greater depth than the PF, then the deformation in the hanging wall of the YBF would be greater than in its footwall i.e., the fringe density should be higher on NW side of the YBF than on its SE side (the

SC

area in between the two parallel faults) as indicated by the simulation result of Model 2 (Fig. 10(c)).

NU

This is obviously in contradiction to the ALK DInSAR data in this area. Table 6 shows the Root Mean Square Errors (RMSEs) and the Correlation Coefficient (Cor-Coef)

MA

between the ALK data and each of the simulation models. Model 3 presents the lowest RMSE

D

(0.299945 m) and highest Cor-Coef (0.68909) while Model 1, with single planar fault segments and

TE

shallower dip angles, has highest RMSE (0.324574 m) and lowest Cor-Coef (0.648461) to the ALK data. The RMSE and the Cor-Coef of Model 4 are very close to that for Model 3. This is no surprising

AC CE P

because the setting of the two models are different only in the section where the PF and YBF are parallel and the two models (Fig. 10(d)-(e)) are nearly identical except in this section of the fault zone where the Model 3 patterns are more resemble to the ALK interferograms. The RMSEs indicate there are about two and half fringes difference between the ALK interferograms and modeled fringe patterns. These fringe differences are due to the spatial resolution limitation in the Poly3D software and consequently, the LOS displacement has been over smoothed and the fault structures are simplified in the models. With this point in mind the 0.68909 correlation coefficient achieved by the Model 3 is impressive and the simulated fringe patterns in the key locations in this best fitting model reasonably coincide with the ALK interferograms and better fit than the other models. Both the YBF and PF have relatively clear topographic expression in the satellite image and DEM but the linkage between the two at the NE and SW ends of the PF is not at all clear. While the local 17

ACCEPTED MANUSCRIPT negative maxima of LOS displacement (Fig. 3(b)) shows a continuous extension of the rupture zone without any obvious bends or kinks, the widely accepted interpretation of the YBF shown as the traced

PT

line in the satellite image in Fig. 1 illustrates an clear eastward kink north of Hanwang. Although these

RI

observations do not favour either interpretation, the Poly3D simulation, as constrained by the ALK DInSAR data, strongly support the model in which the PF is the main fault rupture in the area where

SC

the YBF and PF are parallel.

NU

Seismic reflection profiles across the Yingxiu-Beichuan fault zone could provide concrete evidences of the fault geometry and the relationship between the YBF and the PF. Unfortunately, none of the

MA

published seismic reflection profiles across the Longmenshan fault zone (e.g. Li et al., 2010) covers

D

sufficient scope and depth to reveal the relationship between the YBF and the PF. The generally

TE

accepted opinion, that the YBF intercepts the PF, is based on interpretation and extrapolation (Li et al., 2010). However, the evidence presented in this article indicates that this may not be the case.

AC CE P

The ALK DInSAR fringe patterns in the NE segments (B8-B10) of the YBF can be better simulated only when the fault is given a rather steep dip angle (60°-70°). This is in agreement with most published models but is in confliction with the seismic reflection profiles. The discontinuity features in the seismic reflection profiles at Nanba and further to the NE become quite complicated and the interpretations are not conclusive (Li et al., 2010). However, extending the fault from the traceable surface rupture and displacement features to the statistical centres of the aftershock clusters at depth seems to imply a rather steeply angled fault. For instance, the clusters of aftershocks in the Nanba area (B8) are almost directly below the displacement maxima, at depths of 12 km to 20 km. Another dispute is in the Qingchuan area, the NE end segment of the YBF fault (B10). Here, the majority of focal mechanism solutions from the USGS and CDSN indicate a right-lateral strike slip fault. However, the characteristic circular fringe pattern with multiple rings in the DInSAR 18

ACCEPTED MANUSCRIPT interferograms, corresponding to B10 segment, can only be simulated by setting the fault segment B10 as a blind thrust 5 km below the terrain surface. On satellite images, the track of the YBF

A tectonic interpretation

RI

4.2

PT

corresponding to segment B10 is rather vague and not traceable.

SC

The geometry of the two main faults, the YBF and the PF, active during the 2008 Wenchuan earthquake and the distribution of movement along them determined from this study should be

NU

considered in the context of their tectonic setting. The earthquake was the result of slip along a major

MA

transpressional boundary and displacement was concentrated along the YBF and the much shorter PF, Fig. 11. The faults run parallel to each other and in close proximity, and the study described in this

D

paper demonstrates that the amount of thrust displacement on the YBF drops dramatically in the

TE

portion where it is adjacent to the PF. In this region the displacement is transferred to the PF. This transfer of displacement from one thrust to another, located further into the foreland, is known as

AC CE P

foreland propagation by footwall collapse and is a mechanism of deformation that characterizes many fold/thrust belts (Boyer and Elliott, 1982; Price and Cosgrove, 1990; McClay(ed.), 1992). As illustrated in Fig. 11, the early thrust, the YBF as a splay thrust rising from a floor thrust (Fig. 11(a)), being abandoned as the floor thrust propagates to the SE and transfers its displacement onto a new splay, the PF (Fig. 11(b) and (d)). There are more field observation data available along the PF than the YBF in this region and measurements show 2.4 ~ 3.5m uplift of the PF hanging wall from Bailu to Hanwang (see Fig. 1) (Xu, et al. 2009). Our observation in the Bailu school indicates at least 1.5m uplift at the location (Fig. 7(b)). These confirm the major thrust rupture occurred along the PF during the Wenchuan earthquake. The linkage between the PF and the YBF is different at its two ends. At its north-eastern end the PF merges with the YBF as the displacement along it dies out from a maximum in the middle of the fault 19

ACCEPTED MANUSCRIPT to zero at its end. In contrast the south-western end of the PF is marked by the Xiaoyudong Fault (XF) (see field photo Fig. 7(c)) which is steeply inclined and strikes at right angles to the PF. It is possible

PT

that this fault has acted as a lateral ramp which allows the transfer of displacement from the YBF to the

RI

PF as suggested by Xu, et al. (2009), as shown in Fig. 11(c).

SC

The tectonic model that we presented here is extremely simplified to show the concepts that capture the forward modelling results. The seismic faulting is far more complicated than this simple model and

NU

our interpretation is far from conclusive, further work is being carried out to refine this initial

5.

MA

understanding.

Conclusions

TE

D

In this study, we first derived a refined dataset from ALOS PALSAR DInSAR data to characterize the displacement induced by the Wenchuan earthquake. The data processing includes refining the

AC CE P

DInSAR data to continuous LOS displacement maps without decoherence gaps using the ALK technique. Then a forward modelling of Yingxiu-Beichuan fault zone was carried out using the Poly3D with the LOS displacements as “ground truth” constraints. The forward modelling of the Yingxiu-Beichuan fault zone indicates that the fault motion was mostly thrusting in the southern segments of the Yingxiu-Beichan fault (YBF) and the Pengguan fault (PF), and progressively changing to the right-lateral strike slip as the fault is traced NE direction. An analysis of the data indicates that in the region where the PF runs parallel to the YBF, displacement has been transferred to the former from the latter. This process of forward propagation of a thrust by footwall collapse characterizes many fold/thrust belts and is compatible with the general tectonic setting of the region.

20

ACCEPTED MANUSCRIPT The model experiments produced results which, although at variance with previous interpretations, in the light of the evidence presented, are difficult to refute. They are summarized below.

PT

1. In the section of the fault zone where the YBF and the PF are parallel, the dominant opinion is

RI

that the YBF is the main fault and that it intersects the PF at a depth of about 10 km. However,

SC

the dense fringe patterns in the ALK DInSAR interferograms in the area between the two faults can only be simulated in the model when the PF is set as the main fault that intercepts the YBF at

NU

a depth of approximately 13 km (Fig. 9(e) and Fig. 10(d)). This geometric relationship between the two faults and the distribution of slip is compatible with them being two adjacent splay faults

MA

on a propagating thrust. Thrusting is currently concentrated along the YBF except in the region

D

where it is parallel to the PF. Here the PF represents the propagation of the thrust towards the

TE

foreland and the transfer of the majority of the displacement from the YBF to the PF, Fig. 11(d). 2. To produce simulated fringe patterns which are a reasonable fit to the ALK DInSAR

AC CE P

interferograms, the dip angle of the YBF has to increase (up to 70 °) as the fault is traced from the SW to NE. This agrees with most simulation results by other researchers but contradicts the interpretation based on limited seismic reflection profiles which suggests much gentler dip angles along the section from Nanba (Fig. 1) toward the NE. 3. In the Qingchuan area, the NE segment of the YBF fault (B10) is interpreted as a blind thrust 5 km below the terrain surface according to the model that best simulates the characteristic circular, multiple ring pattern in the ALK DInSAR interferograms in the area. However, the majority of focal mechanism solutions from the USGS and CDSN indicate a dextral strike slip fault. Poly3D provides a simplified simulation using discrete rectangular planes. Consequently the curved fault planes and the diminishing fault displacement in the strike direction can only be roughly approximated by a sequence of discrete planes. Indeed, any such simulation is fundamentally limited 21

ACCEPTED MANUSCRIPT by the software functionality and the key algorithms. However, consistency between the simulation results and the DInSAR data verifies the credibility of the software.

PT

Acknowledgment

RI

This work is supported in part by the State Key Laboratory of Geohazard Prevention, Chengdu

SC

University of Technology, who provided emergency funding to support the field investigation of this

NU

research. JAXA is acknowledged for providing ALOS PALSAR data, as is Caltech/JPL for providing the ROI_PAC software. In particular, we would like to express our great appreciation to Dr. E. Fielding,

MA

Caltech/JPL and Dr. Z. Li, University of Glasgow, for their advice on DInSAR data processing and forward 3D modelling.

TE

D

References

Agustan (2010), Ground deformation detection based on ALOS-PALSAR data utilizing DInSAR technique in Indonesia,

AC CE P

PhD. Thesis, Dept. of Earth and Environmental Sciences, Nagoya University. Boyer, S.E. & Elliott, D., 1982. Thrust systems. Bulletin of the American Association of Petroleum Geologists 66 (9), 11961230.

Burchfiel, B. C., C. Z. Zhilianf, L. Yuping, and L. H. Royden (1995), Tectonics of the Longmen Shan and adjacent regions, central China, Int. Geol. Rev. 37, 661-735. Burchfiel, B. C., L. H. Royden, R. D. van der Hilst, B. H. Hager, Z. Chen, R. W. Li, C. Lu, J. Yao and E. Kirby (2008), A geological and geophysical context for the Wenchuan earthquake of 12 May 2008, Sichuan, People’s Republic of China, GAS Today, 18(7), pp. 4-11. Chen, S. F. and C. J. L. Wilson (1996), Emplacement of the Longmenshan Thrust-Nappe belt along the eastern margin of the Tibetan plateau, Journal of Structural Geology, 18(4), pp. 413-430. Chini, M., S. Atzori, E. Trasatti, C. Bignami, C. Kyriakopoulos, C. Tolomei, and S. Stramondo (2010), The May 12, 2008, (Mw 7.9) Sichuan Earthquake (China): Multiframe ALOS-PALSAR DInSAR Analysis of Coseismic Deformation, IEEE Geoscience and Remote Sensing Letters, VOL. 7, No. 2.

22

ACCEPTED MANUSCRIPT Densmore, A. L., M. A. Ellis, Y. Li, R. Zhou, G. S. Hancock, and N. Richardson (2007), Active tectonics of the Beichuan and Pengguan faults at the eastern margin of the Tibetan Plateau, Tectonics 26, TC4005, doi 10.1029/2006TC001987.

Yingxiu Town of Sichuan Province, China, Sciences, 51(2), pp. 154-163.

PT

Dong, S., Z. Han and Y. An (2008), Surface deformation at the epicenter of the May 12, 2008 Wenchuan M8 earthquake at

RI

Fialko, Y., M. Simons, and D. Agnew (2001), The complete (3-D) surface displacement field in the epicentral area of the 1999 Mw 7.1 Hector Mine earthquake, California, from space geodetic observations, Geophys. Res. Lett. 28, no. 16,

SC

3063-3066.

Geoscience and Remote Sensing, 49, pp.5080-5091.

NU

Fujisada, H., M.Urai and A.Iwasaki (2012) Advanced Methodology for ASTERDEM Generation, IEEE Transactions on

MA

Furuya, M., T. Kobayashi, T. Takada and M. Murakami (2010), Fault Source Modeling of the 2008 Wenchuan Earthquake Based on ALOS/PALSAR Data, Bulletin of the Seismological Society of America, Vol. 100, No. 5B, pp. 2750-2766. Ge, L., K. Zhang, A. Ng, Y. Dong, H. C. Chang and C. Rizos (2008), Preliminary results of satellite radar differential

D

interferometry for the co-seismic deformation of the 12 May 2008 Ms8.0 Wenchuan earthquake, The International

TE

Association of Chinese Professionals in Geographic Information Science (CPGIS).

707-709.

AC CE P

Goldstein,R. M. and H. Z. Zebker (1987), Interferometric radar measurement of ocean surface currents, Nature, 328, pp.

Goldstein, R. M., T. P. Barnett and H. Z. Zebker (1989), Remote sensing of ocean currents, Science, 246, pp. 1282-1285. Gudmundsson, S., Sigmundsson, F., and Carstensen, J. (2002), Three-dimensional surface motion maps estimated from combined interferometric synthetic aperture radar and GPS data, J. Geophys. Res., 107, no. B10, pp.ETG13.1ETG13.14.

Hashimoto, M., M. Enomoto and Y. Fukushima (2009), Coseismic deformation from the 2008 Wenchuan, China, earthquake derived from ALOS/PALSAR images, Tectonophysics, doi:10.1016/j.tecto.2009.08.034. Huang, Y., J. P. Wu, T. Z. Zhang and D. N. Zhang (2008), Relocation of the M w 8.0 Wenchuan earthquake and its aftershock sequence, Sci. China Ser. D-Earth Sci., Vol. 51, No. 12, pp. 1703-1711. Hubbard, J. and J. H. Shaw (2009), Uplift of the Longmenshan and Tibetan plateau and the 2008 Wenchuan (M=7.9) earthquake, Nature, 458(7235), pp. 194-197. Johnston, K., Ver Hoef, J. M., Krivoruchko, K. and Lucas (2003), N. ArcGIS 9: Using ArcGIS Geostatistical Analyst, ESRI, United States of America.

23

ACCEPTED MANUSCRIPT Kobayashi, T., Y. Takada, M. Furuya, and M. Murakami (2009), Locations and types of ruptures involved in 2008 Sichuan earthquake inferred from SAR image matching, Geophys. Res. Lett. 36, L07302, doi 10.1029/2008GL036907. Li, Q., R. Gao, H. Wang, J. Zhang, Z. Lu, P. Li, Y. Guan and R. He (2009), Deep background of Wenchuan earthquake and

PT

the upper crust structure beneath the Longmen Shan adjacent areas, ACTA GEOLOGICA SINICA, Vol. 83, No. 4, pp.

RI

733-739.

Li, Y., D. Jia, J. H. Shaw, J. Hubbard, A. Lin, M. Wang, L. Luo, H. Li, and L. Wu (2010), Structural interpretation of the

SC

coseismic faults of the Wenchuan earthquake: Three-dimensional modeling of the Longmen Shan fold-and-thrust belt,

NU

Journal of Geophysical Research, vol. 115, B04317, doi: 10.1029/2009JB006824. Lin, A., Z. Ren, D. Jia, and X. Wu (2009), Co-seismic thrusting rupture and slip distribution produced by the 2008 M w 7.9

MA

Wenchuan earthquake, China, Tectonophysics 471, 203-215.

Liu, J. G. and H. Yan (2008), Phase correlation pixel-to-pixel image co-registration based on optical flow and median shift propagation, International Journal of Remote Sensing.

D

Liu, J. G. and H. Yan (2006), Robust phase correlation methods for sub-pixel feature matching, Proc. 1st annual Conference

TE

of Systems Engineering for Autonomous Systems, Defence Technology Centre, Edinburgh, A13.

V.36.

AC CE P

Massonnet, D. and K. L. Feigl (1998), Radar interferometry and its application to changes in earth’s surface, Rev. Geophys.

McClay, K.R. (ed.) 1992, Thrust Tectonics, Chapman & Hall, London, pp 419-433. Price, N. J. & Cosgrove, J. W. 1990. Analysis of geological structures. Cambridge University Press, pp. 161–185. Rosen, P. A., S. Henley, G. Peltzer, and M. Simons (2004), Updated Repeat Orbit Interferometry Package Released, EOS Trans. AGU, 85(5), 47.

Shen, Z. K., J. Sun, P. Zhang, Y. Wan, M. Wang, R. Burgmann, Y. Zeng, W. Gan, H. Liao, and Q. Wang (2009), Slip maxima at fault junctions and rupturing of barriers during the 2008 Wenchuan earthquake, Nature Geoscience, Vol. 2. Shimada, M., T. Ozawa, Y. Fukushima, M. Furuya, and A. Rosenqvist (2008), Japanese L-band radar improves surface deformation monitoring, Eos Trans. AGU 89, 277-278. Tachikawa, T., M. Hato, M. Kaku and A. Iwasaki (2011), The characteristics of ASTER GDEM version 2, IGARSS. Thomas, A. L. (1993), Poly3D: a three-dimensional, polygonal element, displacement discontinuity boundary element computer program with application to fractures, faults and cavities in the Earth’s crust, M. S. thesis, Stanford University, Stanford, CA.

24

ACCEPTED MANUSCRIPT Tong, X., D. T. Sandwell and Y. Fialko (2010), Coseismic slip model of the 2008 Wenchuan earthquake derived from joint inversion of interferometric synthetic aperture radar, GPS, and field data, Journal of Geophysical Research, Vol. 115, B04314, doi: 10.1029/2009JB006625.

PT

Wang, Q., X. Qiao, Q. Lan, F. Jeffrey, S. Yang, C. Xu, Y. Yang, X. You, K. Tan and G. Chen (2010), Rupture of deep

RI

faults in the 2008 Wenchuan earthquake and uplift of the Longmen Shan, Nature Geoscience, Vol. 4, doi: 10.1038/NGEO1210.

SC

Wang, Q., Z. Chen, and S. Zheng (2009), Spatial segmentation characteristic of focal mechanism of aftershock sequence of

NU

Wenchuan earthquake, Chinese Science Bulletin, 54: 2263-2270, doi: 10.1007/s11434-009-0367-0. Wang, Z., 1990, Principle of Photogrammetry (with Remote Sensing), Publishing House of Surveying and Mapping,

MA

Beijing, China: Press of Wuhan Technical University of Surveying and Mapping. Wu, M. C., J. G. Liu and P. J. Mason (2013), Adaptive Local Kriging (ALK) to retrieve the slant range surface motion maps of Wenchuan earthquake, International Journal of Remote Sensing, Vol. 34, No. 21, 7589-7606.

D

Wu, M. C., J. G. Liu and P. J. Mason (2012), Three-Dimensional Surface Displacement Map of the 2008 Wenchuan

TE

Earthquake Derived From Phase Correlation (PC) Sub-Pixel Offset Method and Adaptive Local Kriging (ALK)

AC CE P

DInSAR Data, Geoscience and Remote Sensing Symposium (IGARSS). Xu, C., Y. Liu, Y. Wen and R. Wang (2010), Coseismic slip distribution of the 2008 M w 7.9 Wenchuan earthquake from joint inversion of GPS and InSAR data, Bulletion of the Seismological Society of America, Vol. 100, No. 5B, pp. 27362749.

Xu, X. W., X. Z. Wen, G. H. Yu, G. H. Chen, Y. Klinger, J. Hubbard, and J. Shaw (2009), Coseismic reverse‐ and oblique‐slip surface faulting generated by the 2008 Mw 7.9 Wenchuan earthquake, China, Geology, 37(6), 515–518, doi:10.1130/G25462A.1. Yu, G., X. Xu, Y. Klinger, G. Diao, G. Chen, X. Feng, C. Li, A. Zhu, R. Yuan, T. Guo, X. Sun, X. Tan, and Y. An (2010), Fault-scarp features and cascasing-rupture model for the Mw 7.9 Wenchuan earthquake, eastern Tibetan Plateau, China, Bulletion of the Seismological Society of America, Vol. 100, No. 5B, pp. 2590-2614. Zebker, H. A., C. Werner, P. A. Rosen and S. Hensley (1994b), Accuracy of topographic maps derived from ERS-1 interferometric radar, IEEE Trans. Geosci. Rem. Sensing, 32, pp. 823-836.

25

ACCEPTED MANUSCRIPT Zebker, H. A., P. A. Rosen, R. M. Goldstein, A. K. Gabriel and C. Werner (1994a), On the derivation of coseismic displacement fields using differential radar interferometry: the Landers earthquake, Journal of Geophysics Research, 99, pp. 19617-19634.

PT

Zhang, P. Z., X. Z. Wen, Z. K. Shen and J. H. Chen (2010), Oblique, high-angle, listric-reverse faulting and associated

SC

RI

development of strain: the Wenchuan earthquake of May 12, 2008, Sichuan, China, Annu. Rev. Earth Planet. Sci..

MA

NU

Table 1 The Path information of the ALOS PALSAR of the 2008 Wenchuan earthquake.

Table 2 Fault parameters for the Model 1 (Fig. 9(a) and (b)).

TE

Positive value means the thrust and right-lateral motion respectively. The depth is measured from the terrain surface downwards.

**

AC CE P

*

D

B represents the Yingxiu-Beichuan fault, P the Pengguan fault and X the Xiayudong fault. The latitude and longitude are the coordinates of the mid-point of the surface trace of each fault segment.

Table 3 Fault parameters for the Model 2 (Fig. 9(c) and (d)). B represents the Yingxiu-Beichuan fault, P the Pengguan fault and X the Xiayudong fault. The latitude and longitude are the coordinates of the midpoint of the surface trace of each fault segment. *

Positive value means the thrust and right-lateral motion respectively. The depth is measured from the terrain surface downwards. *** B10 is a blind thrust fault 5000m below the terrain surface. **

Table 4 Fault parameters for the Model 3 (Fig. 9(e) and (f)). B represents the Yingxiu-Beichuan fault, P the Pengguan fault and X the Xiayudong fault. The latitude and longitude are the coordinates of the mid-point of the surface trace of each fault segment. *

Positive value means the thrust and right-lateral motion respectively. The depth is measured from the terrain surface downwards. *** B10 is a blind thrust fault 5000m below the terrain surface. **

26

ACCEPTED MANUSCRIPT Table 5 Fault parameters for the Model 4 (Fig. 9(g) and (h)). B represents the Yingxiu-Beichuan fault, P the Pengguan fault and X the Xiayudong fault. The latitude and longitude are the coordinates of the mid-point of the surface trace of each fault segment. *

Positive value means the thrust and right-lateral motion respectively. The depth is measured from the terrain surface downwards. *** B10 is a blind thrust fault 5000m below the terrain surface.

RI

PT

**

NU

SC

Table 6 Root Mean Square Error (RMSE) and the Correlation Coefficient (Cor-Coef) between the ALK DInSAR data and each of simulation models.

MA

Fig. 1. The location of the study area in Sichuan Province, China. The colour composite of ETM+ bands 5, 3 and 1 (RGB) shows the Wenchuan earthquake affected area. The red star indicates the epicenter. The blue square boxes are cities and towns. Red lines represent interpreted faults of the Yingxiu-Beichuan fault zone (Burchfiel

TE

D

et al., 2008; Chen and Wilson, 1996; Dong et al., 2008; Hubbard and Shaw, 2009): the longer red line is the

AC CE P

Yingxiu-Beichuan fault (YBF), the shorter line is the Pengguan fault (PF).

Fig. 2. (a) ALOS PALSAR differential interferograms. (b) Unwrapped PALSAR differential interferograms. The red lines represent the Yingxiu-Beichuan fault (YBF) and the Pengguan fault (PF) as shown in Fig. 1. The red star indicates the epicenter.

Fig. 3. (a) ALK DInSAR interferograms. (b) ALK DInSAR data, the black line denotes the traced local minima (the negative maxima) of the LOS displacement. The red star indicates the epicenter.

Fig. 4. (a) ALK refined DInSAR data marked with profile lines (black), mapped Yingxiu-Beichuan fault (red dotted line) and the trace of LOS minima along the fault line (white dotted line). (b)-(g) profiles of Paths 471 to 476 denoted as A - F in (a) respectively, where the mapped fault and the profile minima are marked only when there is discrepancy between them. 27

ACCEPTED MANUSCRIPT

Fig. 5. The InSAR imaging geometry of the relationship between eastward motion and uplifting. (a) When the

PT

ground position A is moved eastward to A’ by deformation, a positive LOS displacement is induced between the across-event InSAR pair. (b) When the ground position A is uplifted to A’ by deformation, a negative LOS

RI

displacement is induced. (c) When the ground position A is moved to A’ by eastward and uplifting deformation,

SC

the opposite LOS displacements cancel each other resulting in a very small net LOS.

NU

Fig. 6. The imaging geometry of the relationship between horizontal motion and vertical motion for a positive

MA

DInSAR LOS under the condition of eastward and uplift motion.

D

Fig. 7. (a) Thrust fault plane exposed in organic mud deposits near Hongkou town. The high pressure and

TE

temperature produced by the thrust make the surface of the thrust plane relatively hardened forming a carbonate skin. The slickenside on the plane indicate the major motion is thrust uplift of the western block but the

AC CE P

secondary overlaid scratches across the major scratch in about 20° from horizontal indicate right lateral movement. (Field photo taken by JG Liu in July 2008, looking NW). (b) The PF goes through between the two buildings in the Bailu School resulting in >1.5 m vertical displacement. (Field photo taken by JG Liu in July 2009, looking SW). (c) The XF cut through and destroyed the Xiaoyudong bridge in a tear strike slip motion. (Field photo taken by JG Liu in July 2009, looking NW).

Fig. 8. (a) Ground range displacement map. The red star denotes the epicentre of the main shock of the Wenchuan earthquake. Black solid circles are towns. The red boxes correspond to enlarged areas in (b)-(d). (b) Horizontal offset vector field overlaid on (a) in Beichuan region. (c) Horizontal offset vector field overlaid on (a) in the epicentre area of the Wenchuan earthquake. (d) Horizontal offset vector field overlaid on (a) in the area of the XF near Bailu. The brown line in (b), (c) and (d) is the YBF, the red line the PF and the blue line the XF.

28

ACCEPTED MANUSCRIPT Fig. 9. (a) The geometric configuration of the 3D fault structure of Model 1 and the labels for each fault segment. The inset diagram illustrates geometric relationship between the YBF and the PF along the cross-section denoted

PT

by the red line. (b) A different view of the modeled 3D fault geometry of Model 1. (c) The geometric configuration of the 3D fault structure of Model 2 and the labels for each fault segment. (d) A different view of

RI

the modeled 3D fault geometry of Model 2. (e) The geometric configuration of the 3D fault structure of Model 3

SC

and the labels for each fault segment. (f) A different view of the modeled 3D fault geometry of Model 3. (g) The geometric configuration of the 3D fault structure of Model 4 and the labels for each fault segment. (h) A

NU

different view of the 3D fault geometry of Model 4. The inset diagrams in (c), (e) and (g) illustrate geometry of

MA

the YBF along the cross-section denoted by the red line C1 and the relationship between the YBF and the PF along the cross-section denoted by the red line C2. B represents the YBF, P the PF and X the XF. The green

TE

segments of the fault used in simulation.

D

arrow points to the north. The color of each fault segment is for illustration purpose to distinguish different

AC CE P

Fig. 10. (a) ALK DInSAR interferograms; the red line represents the YBF, the yellow line the PF and the green line the XF. (b) Simulated interferogram of Model 1. (c) Simulated interferogram of Model 2. (d) Simulated interferogram of Model 3. (e) Simulated interferogram of Model 4. The black dotted lines in (b)-(e) indicate the setting as the secondary fault of the relationship between the YBF and the PF.

Fig. 11. The spatial relationship and faulting geometry of the YBF, PF and XF. (a) The profile assumed in Model 1 and 2, see text. (b) Profile of Model 3. The YBF is shown as a splay fault rising from the floor thrust. With the propagation of the floor thrust further towards the foreland to the SE, the majority of the displacement was transferred from the YBF to the PF. (c) Plane view of the relationship between the YBF, PF and XF. (d) The 3D faulting geometry of the YBF, PF and XF; the PF is currently the main thrust taking the majority of the motion.

29

AC CE P

TE

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

Figure 1

30

ACCEPTED MANUSCRIPT

PT

Path 474

RI

Path 475

SC

Path 476

Path 471

NU

Path 472

MA

Path 473

TE

D

(a)

AC CE P

Path 474

Path 475

Path 476

Path 471 Path 472

Path 473

(b)

Figure 2

31

ACCEPTED MANUSCRIPT

PT

Path 474

RI

Path 475

SC

Path 476

Path 471

NU

Path 472

MA

Path 473

AC CE P

TE

D

(a)

Path 474

Path 475

Path 476

Path 471 Path 472

Path 473

(b)

Figure 3

32

ACCEPTED MANUSCRIPT Magnitude (Meters) 1.0

A

B

0.5 0

C A ’ A’

F

-1.0

B ’

C’

0

(b)

SC

1.0 0.5

-0.5

NU

F’ E’

-1.0

(c)

MA

(a) Magnitude (Meters)

Profile minima

0.5

C’

D

Mapped fault

-1.0

Magnitude (Meters) 1.0 0.5 0

E

-0.5 -1.0

(f)

0

75000

150000 Distance (Meters)

AC CE P

0

TE

-0.5

(d)

B’

B

0

C

150000 Distance (Meters)

Magnitude (Meters)

D’

0

75000

RI

E

PT

-0.5

D

1.0

A’

A

75000

0

75000

150000 Distance (Meters)

Magnitude (Meters) 1.0 0.5 0

D

Mapped fault

Profile minima

D’

-0.5 -1.0 0

(e)

75000

150000 Distance (Meters)

Magnitude (Meters) 1.0 0.5

E’

150000 Distance (Meters)

Original unwrapped interferogram profile

0

F’

F

-0.5 -1.0

(g)

0

75000

150000 Distance (Meters)

ALK data profile

Figure 4

33

ACCEPTED MANUSCRIPT

SA R

SA R

SA R

θ

θ

A’

E

A (b)

A

LOS E

(c)

MA

(a)

A’

E

NU

A

SC

A’ LO S

LOS

RI

PT

θ

AC CE P

TE

D

Figure 5

34

ACCEPTED MANUSCRIPT

RI

PT

θ

O3

NU

θ

O1

SC

Ground range displacement

O2

ΔZ2 ΔZ1 O3’

Δx

MA

LOS

AC CE P

TE

D

Figure 6

35

RI

PT

ACCEPTED MANUSCRIPT

MA

NU

SC

(b)

(c)

Figure 7

AC CE P

TE

D

(a)

36

ACCEPTED MANUSCRIPT

Qingchuan

PT

Shiba Nanba

RI

Beichuan

SC

≈5 m Hanwang Bailu ≈-5 m

Azimuth Range

(b)

(c)

AC CE P

TE

D

MA

(a)

NU

Yingxiu

(d)

The length of the horizontal offset vector for 2 meters displacement. Figure 8

37

ACCEPTED MANUSCRIPT

(a)

(d)

AC CE P

(c)

TE

D

MA

NU

(b)

SC

RI

PT

Figure 9

(e)

(f)

(g)

(h) 38

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

(b)

(d)

AC CE P

TE

D

(a)

(c)

(e)

Figure 10 39

ACCEPTED MANUSCRIPT

PF

YBF

PT

10 km Floor thrust

RI

(a)

PF

YBF

13 km

MA

NW

NU

PF

Floor thrust

(b)

SC

YBF

(c)

SE Splay

D

The intersection between the YBF and the PF

PF

TE AC CE P

XF

YBF XF

Floor thrust

Figure 11

(d)

40

ACCEPTED MANUSCRIPT Table 1 Acquisition

Days from event

Perpendicular baseline ( m)

471

2008/02/29

-73

≈ 70

2008/05/31

19

2007/01/28

-470

≈ 160

506

2008/06/17

36

2008/02/17

-85

2008/05/19

7

2008/03/05

-68

2008/06/05

24

2007/06/20

-143

2008/06/22

41

2008/04/08 2008/05/24

-34

RI

SC

476

92

≈ 190

92

≈ 53

184

≈ -300

46

NU

≈ 343

MA

475

D

474

TE

473

92

12

AC CE P

472

Temporal baseline (days)

PT

Path

41

ACCEPTED MANUSCRIPT Table 2

B1 B2 B3 B4 B5 B6 B7 B8 B9 B10 P1 P2 P3 X1

30° 57' 31° 8' 31° 23' 31° 37' 31° 43' 31° 51' 32° 1' 32° 16' 32° 28' 32° 35' 31° 9' 31° 20' 31° 32' 31° 11'

103° 28' 103° 41' 103° 55' 104° 13' 104° 22' 104° 29' 104° 40' 104° 56' 105° 13' 105° 25' 103° 46' 104° 2' 104° 12' 103° 45'

227.87 209.03 220.61 259.56 210.96 229.76 217.75 227.68 237.24 225.93 248.20 222.91 195.95 124.22

**

Dip (°) 35 35 45 40 40 40 40 30 30 30 25 25 25 90

Depth (m) 17510 16000 15000 13000 11000 11000 8000 8000 8000 8000 10000 10000 10000 15000

Length (m)

Width (m)

52316.91 9135.99 61298.62 17128.86 23263.78 15098.28 34758.84 40167.06 28346.72 17720.02 21485.23 42972.73 16136.40 13402.06

30527.75 27895.15 21213.20 20224.41 17112.96 17112.96 12445.79 16000.00 16000.00 16000.00 23662.02 23662.02 23662.02 15000.00

*

Dip slip (m) 3.50 3.00 0.50 1.50 2.00 3.00 2.00 0.50 0.20 0.20 0.50 1.50 1.50 0.00

PT

Strike (°)

RI

Longitude

SC

Latitude

*

Strike slip (m) 0.30 0.30 0.80 1.50 2.00 2.50 2.50 1.50 1.30 1.25 0.50 0.20 0.50 -1.00

AC CE P

TE

D

MA

NU

Segment

42

ACCEPTED MANUSCRIPT Table 3

B6 B7 B8 B9 *** B10 P1 P2 P3 X1

P1_1 P1_2 P2_1 P2_2 P3_1 P3_2

Length (m) 52316.91 52316.91 52316.91 9135.99 9135.99 61298.62 61298.62 61298.62 17128.86 17128.86 23263.78 23263.78 15098.28 15098.28 34758.84 34758.84 40167.06 28346.72 17720.02 21485.23 21485.23 42972.73 42972.73 16136.40 16136.40 14470.15

Width (m) 2309.40 24774.38 3800.95 2309.40 24408.26 2309.40 13054.07 6000.00 2610.81 17112.96 2610.81 14001.51 2610.81 14001.51 2610.81 9334.34 8513.42 8513.42 8513.42 2309.40 18929.61 2309.40 18929.61 2309.40 18929.61 15000.00

PT

Depth (m) 2000 16210 17510 2000 16000 2000 12000 15000 2000 13000 2000 11000 2000 11000 2000 8000 8000 8000 13000 2000 10000 2000 10000 2000 10000 15000

RI

**

SC

B5

103° 28' 103° 28' 103° 28' 103° 41' 103° 41' 103° 55' 103° 55' 103° 55' 104° 13' 104° 13' 104° 22' 104° 22' 104° 29' 104° 29' 104° 40' 104° 40' 104° 56' 105° 13' 105° 25' 103° 46' 103° 46' 104° 2' 104° 2' 104° 12' 104° 12' 103° 45'

Dip (°) 60 35 20 60 35 60 50 30 50 40 50 40 50 40 50 40 70 70 70 60 25 60 25 60 25 90

NU

B4

30° 57' 30° 57' 30° 57' 31° 8' 31° 8' 31° 23' 31° 23' 31° 23' 31° 37' 31° 37' 31° 43' 31° 43' 31° 51' 31° 51' 32° 1' 32° 1' 32° 16' 32° 28' 32° 35' 31° 9' 31° 9' 31° 20' 31° 20' 31° 32' 31° 32' 31° 11'

Strike (°) 227.87 227.87 227.87 209.03 209.03 220.61 220.61 220.61 259.56 259.56 210.96 210.96 229.76 229.76 217.75 217.75 227.68 237.24 225.93 248.20 248.20 222.91 222.91 195.95 195.95 134.22

MA

B3

Longitude

D

B2

Latitud

*

Dip slip (m) 3.00 4.00 2.00 2.50 3.00 0.50 1.00 0.50 1.50 1.25 2.00 1.80 2.50 3.00 2.00 1.50 1.00 0.50 2.00 0.50 1.00 1.00 2.00 0.80 1.20 0.00

*

Strike slip (m) 0.30 0.30 0.00 0.30 0.30 0.80 0.50 0.00 1.50 1.00 2.00 1.50 2.50 2.25 2.50 2.00 2.50 2.00 0.00 0.20 0.20 0.50 0.50 0.80 0.80 -1.00

TE

B1

Subsegment B1_1 B1_2 B1_3 B2_1 B2_2 B3_1 B3_2 B3_3 B4_1 B4_2 B5_1 B5_2 B6_1 B6_2 B7_1 B7_2

AC CE P

Segment

43

ACCEPTED MANUSCRIPT Table 4

B7 B8 B9 *** B10 P1 P2

P3 X1

P1_1 P1_2 P2_1 P2_2 P2_3 P3_1 P3_2

Length (m) 52316.91 52316.91 52316.91 9135.99 9135.99 61298.62 17128.86 17128.86 23263.78 23263.78 15098.28 15098.28 34758.84 34758.84 40167.06 28346.72 17720.02 21485.23 21485.23 42972.73 42972.73 42972.73 16136.40 16136.40 14470.15

Width (m) 2309.40 24774.38 3800.95 2309.40 24408.26 17623.00 2610.81 17112.96 2610.81 14001.51 2610.81 14001.51 2610.81 9334.34 8513.42 8513.42 8513.42 2309.40 15000.00 2309.40 15000.00 13014.11 2309.40 15000.00 15000.00

PT

Depth (m) 2000 16210 17510 2000 16000 13500 2000 13000 2000 11000 2000 11000 2000 8000 8000 8000 13000 2000 9500 2000 9500 15000 2000 9500 15000

RI

**

SC

B6

103° 28' 103° 28' 103° 28' 103° 41' 103° 41' 103° 55' 104° 13' 104° 13' 104° 22' 104° 22' 104° 29' 104° 29' 104° 40' 104° 40' 104° 56' 105° 13' 105° 25' 103° 46' 103° 46' 104° 2' 104° 2' 104° 2' 104° 12' 104° 12' 103° 45'

Dip (°) 60 35 20 60 35 50 50 40 50 40 50 40 50 40 70 70 70 60 30 60 30 25 60 30 90

NU

B5

B4_1 B4_2 B5_1 B5_2 B6_1 B6_2 B7_1 B7_2

30° 57' 30° 57' 30° 57' 31° 8' 31° 8' 31° 23' 31° 37' 31° 37' 31° 43' 31° 43' 31° 51' 31° 51' 32° 1' 32° 1' 32° 16' 32° 28' 32° 35' 31° 9' 31° 9' 31° 20' 31° 20' 31° 20' 31° 32' 31° 32' 31° 11'

Strike (°) 227.87 227.87 227.87 209.03 209.03 220.61 259.56 259.56 210.96 210.96 229.76 229.76 217.75 217.75 227.68 237.24 225.93 248.20 248.20 222.91 222.91 222.91 195.95 195.95 124.22

MA

B3 B4

Longitude

*

Dip slip (m) 3.00 4.00 2.00 2.50 3.00 0.50 1.50 1.25 2.00 1.80 2.50 3.00 2.00 1.50 1.00 0.50 2.00 0.50 1.00 1.00 2.00 1.00 0.80 1.20 0.00

*

Strike slip (m) 0.30 0.30 0.00 0.30 0.30 0.80 1.50 1.00 2.00 1.50 2.50 2.25 2.50 2.00 2.50 2.00 0.00 0.20 0.20 0.50 0.50 0.00 0.80 0.80 -1.00

D

B2

Latitude

TE

B1

Subsegment B1_1 B1_2 B1_3 B2_1 B2_2

AC CE P

Segment

44

ACCEPTED MANUSCRIPT Table 5

B6 B7 B8 B9 *** B10 P1 P2 P3 X1

P1_1 P1_2 P2_1 P2_2 P3_1 P3_2

Length (m) 52316.91 52316.91 52316.91 9135.99 9135.99 61298.62 61298.62 17128.86 17128.86 23263.78 23263.78 15098.28 15098.28 34758.84 34758.84 40167.06 28346.72 17720.02 21485.23 21485.23 42972.73 42972.73 16136.40 16136.40 14470.15

Width (m) 2309.40 24774.38 3800.95 2309.40 24408.26 17623.00 3549.30 2610.81 17112.96 2610.81 14001.51 2610.81 14001.51 2610.81 9334.34 8513.42 8513.42 8513.42 2309.40 15000.00 2309.40 13500.00 2309.40 15000.00 15000.00

PT

Depth (m) 2000 16210 17510 2000 16000 13500 1500 2000 13000 2000 11000 2000 11000 2000 8000 8000 8000 13000 2000 9500 2000 27000 2000 9500 15000

RI

**

SC

B5

103° 28' 103° 28' 103° 28' 103° 41' 103° 41' 103° 55' 103° 55' 104° 13' 104° 13' 104° 22' 104° 22' 104° 29' 104° 29' 104° 40' 104° 40' 104° 56' 105° 13' 105° 25' 103° 46' 103° 46' 104° 2' 104° 2' 104° 12' 104° 12' 103° 45'

Dip (°) 60 35 20 60 35 50 25 50 40 50 40 50 40 50 40 70 70 70 60 30 60 30 60 30 90

NU

B4

30° 57' 30° 57' 30° 57' 31° 8' 31° 8' 31° 23' 31° 23' 31° 37' 31° 37' 31° 43' 31° 43' 31° 51' 31° 51' 32° 1' 32° 1' 32° 16' 32° 28' 32° 35' 31° 9' 31° 9' 31° 20' 31° 20' 31° 32' 31° 32' 31° 11'

Strike (°) 227.87 227.87 227.87 209.03 209.03 220.61 220.61 259.56 259.56 210.96 210.96 229.76 229.76 217.75 217.75 227.68 237.24 225.93 248.20 248.20 222.91 222.91 195.95 195.95 124.22

MA

B3

Longitude

*

Dip slip (m) 3.00 4.00 2.00 2.50 3.00 0.50 1.00 1.50 1.25 2.00 1.80 2.50 3.00 2.00 1.50 1.00 0.50 2.00 0.50 1.00 1.00 2.00 0.80 1.20 0.00

*

Strike slip (m) 0.30 0.30 0.00 0.30 0.30 0.80 0.50 1.50 1.00 2.00 1.50 2.50 2.25 2.50 2.00 2.50 2.00 0.00 0.20 0.20 0.50 0.50 0.80 0.80 -1.00

D

B2

Latitude

TE

B1

Subsegment B1_1 B1_2 B1_3 B2_1 B2_2 B3_1 B3_2 B4_1 B4_2 B5_1 B5_2 B6_1 B6_2 B7_1 B7_2

AC CE P

Segment

45

ACCEPTED MANUSCRIPT Table 6 RMSE (m)

Cor-Coef

Model 1

0.324574

0.648461

Model 2

0.305761

Model 3

0.299945

Model 4

0.300965

PT

Models

0.689090 0.687643

AC CE P

TE

D

MA

NU

SC

RI

0.671285

46

ACCEPTED MANUSCRIPT Highlights

TE

D

MA

NU

SC

RI

PT

Fault slip was predominantly thrust in the SW and changed to dextral slip in NE. The dip angles of the Longmenshan fault system are increasing from the SW to NE. In the parallel fault zone, displacement has been transferred to the PF from YBF. PF is the main fault that intercepts the YBF at a depth of 13 km. Qingchuan fault segement is set to a blind thrust to match the actual fringe pattern.

AC CE P

    

47