SCIENCE CHINA Earth Sciences. Influence of fault geometry and fault interaction on strain partitioning within western Sichuan and its adjacent region

Transcription

1 SCIENCE CHINA Earth Sciences RESEARCH PAPER January 2010 Vol.53 No.1: 1 15 doi: /s Influence of fault geometry and fault interaction on strain partitioning within western Sichuan and its adjacent region WANG Hui 1,2*, LIU Jie 3, SHEN XuHui 1, LIU Mian 2, LI QingSong 4, SHI YaoLin 5 & ZHANG GuoMin 1 1 Institute of Earthquake Science, China Earthquake Administration, Beijing , China; 2 Department of Geological Sciences, University of Missouri, Columbia, MO 65211, USA; 3 China Earthquake Network Center, China Earthquake Administration, Beijing , China; 4 Lunar and Planetary Institute, Houston, TX 77058, USA; 5 Laboratory of Computational Geodynamics, Graduate University of Chinese Academy of Sciences, Beijing , China; Received March 31, 2009; accepted November 9, 2009 There are several major active fault zones in the western Sichuan and its vicinity. Slip rates and seismicity vary on different fault zones. For example, slip rates on the Xianshuihe fault zone are higher than 10 mm/a. Its seismicity is also intense. Slip rates on the Longmenshan fault zone are low. However, Wenchuan M s 8.0 earthquake occurred on this fault zone in Here we study the impact of fault geometry on strain partitioning in the western Sichuan region using a three-dimensional viscoelastoplastic model. We conclude that the slip partitioning on the Xianshuihe-Xiaojiang fault presents as segmented, and it is related to fault geometry and fault structure. Slip rate is high on fault segment with simple geometry and structure, and vice versa. Strain rate outside the fault is localized around the fault segment with complex geometry and fault structure. Strain partitioning on the central section of the Xianshuihe-Xiaojiang fault zone is influenced by the interaction between the Anninghe-Zemuhe fault and the Daliangshan fault zone. Striking of the Longmenshan fault zone is nearly orthogonal to the direction of eastward extrusion in the Tibetan Plateau. It leads to low slip rate on the fault zone. Xianshuihe-Xiaojiang fault zone, Longmenshan fault zone, fault geometry, strain partitioning, 3D viscoelastoplastic model Citation: Wang H, Liu J, Shen X H, et al. Influence of fault geometry and fault interaction on strain partitioning within western Sichuan and its adjacent region. Sci China Earth Sci, 2010, doi: /s *Corresponding author ( wanghui500@gmail.com) Located in the southeast borderland of the Tibetan Plateau, the western Sichuan region and its vicinity is a transition zone between the active Tibetan Plateau and the stable South China. Tectonics in the region is active. There are many northwest, northeast and near north-south striking faults cutting through the crust in the region. The northwest-west striking faults are mainly lateral strike-slip faults, and the near north-south striking faults are mainly convergence faults. The crustal deformation pattern in the region shows obvious deformation localization [1]. Two of the most important active fault zones in the western Sichuan area are the Xianshuihe-Xiaojiang fault zone and the Longmenshan fault zone. These two faults form a Y -shaped fault system. They divide this region into three tectonic blocks: the Sichuan-Yunnan Block, the Bayan Har Block, and the South China Block (Figure 1). The Xianshuihe-Xiaojiang fault zone is also one of the most active fault zones in Chinese mainland. Its northwestern section extends in northwest direction, while its southern Science China Press and Springer-Verlag Berlin Heidelberg 2010 earth.scichina.com

2 2 WANG Hui, et al. Sci China Earth Sci January (2010) Vol.53 No.1 Figure 1 Simplified tectonic map of the western Sichuan and its adjacent region and geological fault slip rates (mm/a). section extends approximately from north to south. Its length is more than 1000 km. Slip rates on the fault zone are higher than 10 mm/a [2, 3]. Seismicity in the fault zone is also intense. In the 20th century, several earthquakes with magnitude over M7.0 occurred on the fault zone. Distribution of major earthquakes indicates the segmentation of the Xianshuihe-Xiaojiang fault zone [3]. The Longmenshan fault zone strikes in northeast direction. The fault zone accomodates crust shortening between the Tibetan Plateau and the South China [4]. Although present-day activity on the Longmenshan fault zone is weak, the seismogenic fault sustained occurrence of M s 8.0 Wenchuan earthquake on May 12th, Fault slip rates are the basis for describing regional strain partitioning and seismicity. Both geology and GPS survey provide slip rates on each segment of the Xianshuihe- Xiaojiang and the Longmenshan fault zone [3, 5 7]. However, the observations are limited. Geological study provides long-term motions of several given sites on fault zone. Uncertainty of the dating of samples always introduces errors for long-term motion. GPS survey establishes slip rates according to velocity profiles across faults. However, GPS results depend on seismic cycle on the fault because the fault is locked when seismicity is quiet [8]. The geological suvey and GPS results are not enouth to explain the dynamics of fault activity, because they only reflect the kinematic characteristics. There are many factors influencing the distribution of fault slip rates, such as evolution history of fault system [9], fault mechanical parameters [10], fault geometry [11, 12], and interaction between faults [9, 13] etc. Crustal motion in the western Sichuan area presents a clockwise rotation around the Eastern Himalayan Syntax (EHS for short) [6, 14 16]. The most significant feature of fault system in the study region is along-strike variation of the Xianshuihe- Xiaojiang fault zone. The northwest striking Xianshuihe fault is the western section of the Xianshuihe-Xiaojiang fault system, and the north-south striking Xiaojiang fault is the southern section of the fault system. The trace of entire Xianshuihe-Xiaojiang fault system looks like an arc section around the EHS. Fault geometry of the Xianshui-Xiaojiang fault zone may play a significant role in strain partitioning

3 WANG Hui, et al. Sci China Earth Sci January (2010) Vol.53 No.1 3 in the western Sichuan and its adjacent area. In this paper, we construct a three-dimensional geodynamic model to study the strain partitioning in the western Sichuan and its vicinity. We employ 3D viscoelastoplastic finite element model to model slip rates on the Xianshuihe-Xiaojiang and the Longmenshan fault zone. Based on our modeling results, the impact of fault geometry on slip rates and regional strain partitioning is discussed in detail. 1 Tectonic background of the western Sichuan and its adjacent area 1.1 Tectonics in the western Sichuan and its adjacent area Both shallow and deep structures are complex in the western Sichuan and its adjacent area. The upper crust is thin [17, 18] and the middle-lower crust is weak [19, 20]. The study region is divided into several tectonic units by major faults. Tectonic motion of each unit is accounted to be a complex or superimposition of three basic types of motions: sliding, rotation, and uplift [5]. The regional motion patterns lend support to a model with a mechanically weak lower crust experiencing deformation beneath a stronger, highly fragmented upper crust [6]. Contemporary GPS velocity field with respect to the stable South China presents a clockwise rotation around the EHS in the Sichuan-Yunnan region (Figure 2). The crust moves southward in the interior of Sichuan-Yunnan Block, and moves southwestward in the southwest Yunnan region. This motion pattern might be the reflection of the eastward extrusion in the Tibetan Plateau during the Indo-Asian Figure 2 GPS velocity field with respect to South China (after ref. [6]). collision [6, 14 16]. The crust of the eastern Tibet moves about 13 mm/a toward east with respect to the stable South China Block. 1.2 Major fault systems and their kinematics in the western Sichuan area The western Sichuan region is divided into the South China block, the Sichuan-Yunnan block, and the Bayan Har block by the Xianshuihe-Xiaojiang and the Longmenshan fault systems. Different movement between two blocks is centralized mainly on these two fault systems. Kinematics and seismicity on the two fault systems are different. The Xianshuihe-Xiaojiang fault zone formed in Cenozoic is active in Late Quaternary. Dozens of historical earthquakes with magnitude over M7.0 were recorded on the fault zone in the past 300 years [3]. The most significant feature of the fault system is along-strike variation [21]. The Longmenshan fault system was formed in the Indosinian and in the Yanshanian. It is a convergence fault with striking in northeast direction. The M s 8.0 Wenchuan earthquake occurred on this fault on May 12th, The Xianshuihe-Xiaojiang fault system defines the northern and eastern boundaries of the Sichuan-Yunnan Block. The fault system consists of several active fault zones. Its northern section is the Xianshuihe fault zone, which is a narrow linear tectonic zone. The Xianshuihe fault zone consists of the Luhuo fault in the northwest and the Moxi fault in the southeast, and conjoins the Anninghe-Zemuhe fault and the Daliangshan fault near Shimian in Sichuan Province. The central section of the Xianshuihe-Xiaojiang fault system consists of the Anninghe-Zemuhe fault zone, the Daliangshan fault zone, and Xiaoxiangling fragment between the two fault zones. The Anninghe-Zemuhe fault zone consists of the north-south striking Anninghe fault zone and the northwest striking Zemuhe fault. The Daliangshan fault zone consists of the Haitang-Yuexi fault, the Puduhe fault, the Butuo fault, and the Jiaoji fault etc. [22]. The Anninghe-Zemuhe fault zone and the Daliangshan fault zone conjoin near Ningnan-Qiaojia in Yunnan province. The Xiaojiang fault cuts through the crosspoint. The southern section of the Xianshuihe-Xiaojiang fault system is the Xiaojiang fault zone. The Xiaojiang fault zone includes two parallel branch faults with a distance less than 20 km between them. Motions of each fault segment on the Xianshuihe- Xiaojiang fault zone are different based on geological survey. The left-lateral slip rate on the Luhuo fault is 15±5 mm/a [3]. Total left-lateral slip rate in Holocene on southeastern section of the Xianshuihe fault is 9.6±1.7 mm/a [23]. Therefore, horizontal slip rate on the Xianshuihe fault zone striking NW45 is 14±2 mm/a. The left-lateral slip rate on the Anninghe fault is only 6.5±1 mm/a [5]. Slip rate on the Zemuhe fault zone is about 4.9±0.6 mm/a [24]. Mean left-lateral slip rate on the Daliangshan fault is about 3

4 4 WANG Hui, et al. Sci China Earth Sci January (2010) Vol.53 No.1 mm/a in Late Quaternary [5, 25, 26]. Left-lateral slip rate on the Xiaojiang fault is about 10±2 mm/a [27]. Comtemporary fault slip rate can be measured using GPS velocity profile across fault. The slip rate on the northwestern segment of the Xianshuihe fault is 10±2 mm/a. It is 10±2 mm/a on the central section, and is 11±2 mm/a on the eastern section. Left-lateral slip rates on the Anninghe fault zone and on the Daliangshan fault zone are both 4±2 mm/a. Therefore, total slip rates on the central section of the Xianshuihe-Xiaojiang fault zone is about 8 mm/a. Leftlateral slip rate on the Zemuhe fault zone is 7±2 mm/a, and the same value is true on the Xiaojiang fault zone [6]. The Longmenshan fault zone consists of the Wenchuan-Maowen thrust, the Yingxiu-Beichuan thrust, the Penxian-Guanxian fault, and many buried piedmont thrust faults [4]. Geological crust shortening on the Longmenshan fault zone is about 4 6 mm/a [28]. Contemporary shortening derived from GPS survey is less than 3 mm/a. Rightlateral slip rate on the fault zone is indetectable [6, 29]. Slip rates on the Xianshuihe-Xiaojiang fault zone and the Longmenshan fault zone given by previous studies are listed in Table 1. The data in Table 1 show difference between the geological and the geodetic results. The geological slip rates include fault creeping and coseismic dislocation. They not only indicate the long-term motion of the fault, but also include information about fault evolution [9]. On the other hand, the GPS results mainly present interseismic crustal motion [8]. The difference between the two kinds of data is reasonable [30, 31]. The geological slip rates reflect long-term fault activity and the geodetic slip rates show short-term fault activity. Dispite the different value, they have the same slip pattern. Slip rate on the Xianshuihe-Xiaojiang fault system is higher than that on the Longmenshan fault system. Seismicity on the Xianshuihe-Xiaojiang fault has a higher frequency than that on the Longmenshan fault. Slip rates on each fault segment are different. Structures of the Xianshuihe fault and the Xiaojiang fault are both simple. It leads to high slip rates on these two faults. The central section of the Xianshuihe- Xiaojiang fault zone consists of the Anninghe-Zemuhe fault and the Daliangshan fault. The complex fault structure leads to low slip rates on the fault section. Previous observations provided basic kinematic patterns of the Xianshuihe-Xiaojiang fault zone and the Longmenshan fault. Using a three-dimensional model, we further investigate the geodynamic factors contributing to the fault slip rates. 2 Three-dimensional viscoelastoplastic model in the western Sichuan and its adjacent region 2.1 Introduction of viscoelastoplastic model We employ a three-dimensional viscoelastoplastic model to simulate long-term motion and deformation of elastoplastic crust underlying viscous lower crust and upper mantle. Details of the viscoelastoplastic model are given in the appendix. Plastic deformation occurs when stress reaches the plastic yield criterion (yield envelope), described here by the Drucker-Prager yield function [32]: F = α I + J k 1 2, where I 1 is first invariant of the stress tensor and J 2 is second invariant of the deviatoric stress tensor. The parameters α and k are related to cohesion and inner effective frictional coefficient, respectively. A paralleled software package of three-dimensional viscoelastoplastic finite element (FE) model was developed by Li et al. [32]. It has been used to model long-term motion and evolution of the San Andreas fault system, which is the boundary between the Pacific Plate and the North America Plate [11, 13]. We use the software package in this study. The fault zone in our 3D FE model is assumed to be a viscoelastoplastic layer with finite thickness. Calculated results include fault slip rate and effective strain rate outside the fault. We take one point on fault and one point outside the fault to see the evolution process of slip rate and effective strain rate, respectively (Figure 3). Positions of these two points are shown in Figure 4(b). In a viscoelastoplastic model with constant loading, the effective strain rate is related to deformation time. If defor- Table 1 Kinematics of the Xianshuihe-Xiaojiang fault zone and the Longmenshan fault zone Active fault zone Striking Motion mode Geological slip rate (mm/a) GPS slip rate (mm/a) Xianshuihe NW40 Left-lateral strike slip Western Xianshuihe NW40 Left-lateral strike- slip 15±5 10±2 Eastern Xianshuihe NW20 Left-lateral strike slip with thrust 9.6±1.7 11±2 Anninghe NS Left-lateral strike slip 6.5±1 ~4 Zemuhe NW25 Left-lateral strike slip with normal 4.9±0.6 7±2 Daliangshan NS Left-lateral strike slip ~3 ~4 Xiaojiang NS Left-lateral strike slip 10±2 7±2 Longmenshan NE45 Reverse with right-lateral strike slip Shortening rate, 4 6 Shortening rate <3 with little right-lateral strike slip

5 WANG Hui, et al. Sci China Earth Sci January (2010) Vol.53 No.1 5 Figure 3 Sketch map for evolution of the predicted fault slip rate and effective strain rate outside the fault. (a) Evolution of slip rate on point A shown in Figure 4(b); (b) grey line shows evolution of effective strain rate on point A shown in Figure 4(b). Dark line shows evolution of effective strain rate on point B shown in Figure 4(b). Figure 4 Numerical mesh for the finite element model. The entire Xianshuihe-Xiaojiang fault and the Longmenshan fault (white line) are explicitly included in the model. (a) Span of the finite element model and boundary constraints; (b) three-dimensional numerical mesh for the finite element model. mation time is much shorter than relaxation time, the effective strain rate is mainly elastic strain rate when stress does not reach the plastic yield criterion, and is mainly plastic strain rate when stress reaches the plastic yield criterion. If deformation time is long enough, the effective strain rate is mainly viscous strain rate when stress does not reach the plastic yield criterion, and is mainly plastic strain rate when stress reaches the plastic yield criterion. If deformation time is much longer than relaxation time, then stress in rock will reach litho-static pressure without shear stress. The effective strain rate is mainly viscous strain rate (no shear deformation) at that time. The crustal effective strain rate mentioned in the paper consists of plastic and viscous strain rate. 2.2 Three-dimensional finite element model of the western Sichuan area The study area spans from 98 E to 106 E and from 24 N to 32 N. It is located in the southeast borderland of the Tibetan Plateau. We construct a three-dimensional finite element model for it. The crust thickness shows large lateral variation in the study area. It reaches km in the eastern Tibet [33, 34]. The focal depths indicate a shallow elastic upper crust in the western Sichuan region. According to the results of earthquake relocation, the focal depths in the western Sichuan plateau are centralized mainly in 0 15 km [18, 35]. To study the long-term motion and deformation of the upper crust in the western Sichuan area, we assume the mean thickness of regional upper crust to be 15 km and ignore the influence of up and down of the Moho discontinuity. Although lateral and vertical variations of the crust have effects on the deformation in the western Sichuan area [36], they are also ignored to simplify the model. Such processing highlights the influence of fault geometry on regional strain partitioning. The viscosities for the finite element model are Pa s for the upper crust and Pa s for the lower crust and upper mantle. Mean Young s modulus is established using three-dimensional velocity structure in the

6 6 WANG Hui, et al. Sci China Earth Sci January (2010) Vol.53 No.1 western Sichuan area [36]. We focus on the Xianshuihe-Xiaojiang fault zone and the Longmenshan fault zone. These two fault zones are simplified as connective fault zones though they are composed of many disconnected faults. They are assumed to cut through the whole model according to their tectonic background. All faults are modeled as viscoelastoplatic layers with 400 m in width in the three-dimensional finite element model. The Longmenshan fault is defined as a thrust fault zone trending northwest. Its dip angle is about 30. The Xianshuihe- Xiaojiang fault zone is simplified as a vertical fault. Previous studies show that the effective friction coefficient on the Xianshuihe-Xiaojiang fault zone is less than [10], we assume 0 as the fault effective friction coefficient. Rock cohesions are based on previous results [11, 13, 32]. Material parameters for the three-dimensional model are listed in Table 2. Figure 4 shows the scale and the mesh of the FE model. The total area is about km 2. The threedimensional finite element model is divided into 2591 elements in each layer. Mean size for the elements is about 16.4 km. We construct four models with the same boundary constraints to investigate the influence of the Xianshuihe- Xiaojiang fault zone and the Longmenshan fault zone on the regional strain partitioning. The simple descriptions for the four models are shown in Table 3. The details of different models are introduced in section Boundary constraints Crustal motion within the western Sichuan area is controlled by the eastward extrusion of the Tibetan Plateau, material heterogeneity, and deep process etc. The first factor might be the most important [1, 37]. Based on GPS results, the crust of the eastern Tibet moves eastward respect to the stable South China, while the crust of the southwest Yunnan moves southwestward [6]. Previous studies show that the present-day crustal motion pattern derived from GPS survey is consistent with that deduced by studying on the late Qua- ternary tectonics [7, 37 39]. It is reasonable to use the GPS results as the boundary constraints for our models. In order to study how the strain distributes along the Xianshuihe-Xiaojiang fault zone, we constrain boundary conditions using the GPS results respect to the stable South China Block (Figure 4(a)). The eastern boundary of our model is located in the interior of South China. It is assumed to be fixed since the South China Block is stable and deforms little. The western boundary is located in the eastern Tibet and the southeast Yunnan block. Its boundary constraints are the interpolation of the observed GPS velocity nearby. The southern and northern boundaries are 24 N and 32 N respectively. The southern and northern boundaries not only cut across the channel for the lateral extrusion of the Tibetan Plateau [40, 41], but also cut across the Xianshuihe-Xiaojiang fault zone and the Longmenshan fault zone respectively. To reduce the influence of boundary constraints on the results, we assume that the northern and southern boundaries are free. The surface of our model is also set free. The bottom (at 60 km depth) is fixed in vertical direction, and free in horizontal direction. The three-dimensional boundary constraints are assumed to be invariant with depth, for the study region is small enough to be simplified as a planar thin shell model. The GPS survey results in the study area represent crustal motion and deformation well. Although previous studies discussed mechanical coupling between the crust and the upper mantle using the observations of the directions of principal stress [42, 43], the constraints on deep motion are still scarce. Since our model is a simplified physics model to discuss the influence of fault geometry on long-term strain partitioning, using the GPS results as the approximate boundary constraints for our model is reasonable. 3 Impact of fault geometry on the long-term strain partitioning We use a paralleled finite element package to simulate the Table 2 Material parameters for the three-dimensional viscoelastoplastic model Young s modulus Viscosity Effective friction coefficient (MPa) Cohesion Poisson s rate (MPa) (Pa s) Upper crust Lower crust and upper mantle Fault in upper crust Fault in lower crust and upper mantle Table 3 Simple descriptions for the four finite element models in the western Sichuan and its adjacent region Model No. Fault zones included in the FE model and their cohesions 1 The Xianshuihe-Anninghe-Xiaojiang fault zone (10 MPa) 2 The Xianshuihe-Anninghe-Xiaojiang fault zone (10 MPa), the Daliangshan fault zone (10 MPa, 20 MPa) 3 The Xianshuihe-Daliangshan-Xiaojiang fault zone (10 MPa) 4 The Xianshuihe-Anninghe-Xiaojiang fault zone (10 MPa), the Daliangshan fault zone (20 MPa), the Longmenshan fault zone (10 MPa)

7 WANG Hui, et al. Sci China Earth Sci January (2010) Vol.53 No.1 7 lithospheric deformation including faults. We adopt 5 years as one time step, and calculate steps. Boundary displacement constraints increase linearly with time step. If calculating were longer enough, both fault slip rates and effective strain rate outside the faults would reach the steady state (Figure 3). The steady state represents the long-term crustal motion. The constant slip rates represent long-term fault slip rate. 3.1 Impact of geometry of the Xianshuihe-Anninghe- Xiaojiang fault In the first model, only the Xianshuihe-Anninghe- Zemuhe-Xiaojiang fault zone is assumed active. We compare the predicted crustal velocity and the GPS velocity field in the western Sichuan and its adjacent region (Figure 5(a)). Although two velocity fields differ obviously, they show the same pattern and clockwise rotation around the EHS. The differences are small at the points in the vicinity of the faults. The crustal motion patterns in the study area are different at two sides of major faults. The Xianshuihe-Anninghe- Zemuhe-Xiaojiang fault zone consumes most of the different motion between the Sichuan-Yunnan Block and the South China Block. The displacement across the fault system shows gradient variation. On the east side of the Xianshuihe-Xiaojiang fault zone are two blocks: the Bayan Har block and the South China Block. The Sichuan-Yunnan Block and the southwest Yunnan Block are located at its west side. The blocks move faster on the west side than on the east side of the Xianshuihe-Xiaojiang fault zone. The crustal motion pattern of the Bayan Har Block differs from that of the South China Block. The entire Bayan Har Block moves northeastward. The crustal velocity of the South China Block is nearly zero. On the west side of the Xianshuihe-Xiaojiang fault zone, the Sichuan-Yunnan Block rotates clockwise around the EHS. The crust motion patterns of the northwestern Sichuan sub-block, the middle Yunnan sub-block, and the southwest Yunnan sub-block are different from each other. The northwestern Sichuan sub-block moves southeastward. The middle Yunnan sub-block moves nearly southward. The southwest Yunnan sub-block moves southwestward. The predicted slip rates show that the left-lateral strike-slip dominates the entire Xianshuihe-Xiaojiang fault zone (Figure 5(b)). However, slip rates on each fault segments are different. Slip rates on the northwestern segment of the Xianshuihe fault are the highest. The mean value is 12.4 mm/a with extension component. Mean slip rate on the southeastern segments of the Xianshuihe fault is 8.4 mm/a. Slip rate on the Anninghe-Zemuhe fault is the smallest, 4.4 mm/a on the Anninghe fault and 4.7 mm/a on the Zemuhe fault, respectively. Slip rate on the Xiaojiang fault varies from 8.0 to 8.7 mm/a. Predicted fault slip rates are consistent with geological observations. Slip rates are high on the northern and the southern sections of the Xianshuihe- Xiaojiang fault zone since they are straight. The Xianshuihe-Xiaojiang fault zone changes its orientation sharply on its central section. It leads to low slip rates on this section. Effective strain rates outside the faults are localized in the regions where the fault striking changes sharply, such as the region near Moxi, the region to the east of the Anninghe-Zemuhe fault, and the region near Dongchuan etc. Figure 5 Results produced by the first model. (a) Predicted velocity field (red arrows show observed GPS velocity, black arrows show predicted velocity); (b) predicted fault slip rates and effective strain rate outside the fault.

8 8 WANG Hui, et al. Sci China Earth Sci January (2010) Vol.53 No.1 (Figure 5(b)). As a transfer segment connecting the Xianshuihe fault zone and the Anninghe fault zone, the Moxi fault segment rotates clockwise about 15 than the Luhuo fault segment. Strike altering of fault restrains the activity on the Moxi fault segment. Slip rate is lower on the Moxi fault segment than on the Luhuo fault segment. Strike of the Anninghe-Zemuhe fault changes from north-south direction to northwest direction. Since the Anninghe fault strikes in north-south and prevents the eastward motion, crustal motion changes from southeastward in the west Sichuan plateau to southward in the central Yunnan region. The Anninghe-Zemuhe fault zone has the smallest slip rates on the whole fault system. Effective strain rates around the Anninghe-Zemuhe fault zone are located at the Xiaoxiangling fragment, which is located at the east side of the Anninghe-Zemuhe fault. It leads to a large area with high strain rate in the region. Toward south, fault strike changes from northwest to near north-south near Dongchuan, where the Zemuhe fault conjoins the Xiaojiang fault. It results in high effective strain rate near Dongchuan. High effective strain around the western boundary of the model may be related to the boundary effect. High effective strain rate outside the fault zone correlates to large crustal deformation around the Xianshuihe- Xiaojiang fault zone. The highest mountain in the western Sichuan region, the Gongga Mountain, is located on the west side of the Moxi fault. The Gongga region was uplifted in Late Cenozoic. Vertical movement in the region absorbs part of strain off faults. The active Daliangshan and the Mabian fault zones are located on the east side of the Anninghe-Zemuhe fault zone. The three fault zones are nearly parallel. Field investigations reveal that the newly generated Daliangshan fault zone consists of several discontinuous fault zones. Slip rate on the fault zone is low [25], so is the seismicity [44]. Initiation of the Daliangshan fault and the Mabian fault is later than that of the Anninghe-Zemuhe fault zone [26]. Distribution of predicted effective strain rates indicates that the initiation of the Daliangshan faul zone and the Mabian fault zone might be related to the strain localization on the east side of the Anninghe-Zemuhe fault zone. 3.2 Impact of the Daliangshan fault zone The Xianshuihe fault zone is separated into two branches near Shimian town. The west branch is the Anninghe- Zemuhe fault, and the east branch the newly generated Daliangshan fault zone. The Daliangshan fault zone consists of four fault segments, and its striking is about [5]. The Anninghe-Zemuhe fault zone and the Daliangshan fault zone conjoin with the Xianshuihe fault near Shimian and with the Xiaojiang fault zone near Qiaojia in Yunnan. The Daliangshan fault is a newly generated fault zone. Its maturity is lower than the Anninghe-Zemuhe fault zone [26]. The first model reveals that the geometry of the Xianshuihe-Anninghe-Zemuhe-Xiaojiang fault zone influences the strain partitioning in the western Sichuan area. The region with high effective strain rate is located on the east side of the Anninghe-Zemuhe fault. It extends to the neighborhood of the Daliangshan fault zone. High effective strain rate in the region is in favor of initiation of the Daliangshan fault zone. To investigate the impact of the Daliangshan fault zone on regional strain partitioning, we modify the first model and construct the second model. In the second model, the Daliangshan fault zone is assumed active. Assuming the cohesion of the Daliangshan fault zone to be the same as that of the Anninghe-Zemuhe fault zone, we calculate slip rates on each of the fault segments and effective strain rate outside the fault (Figure 6(a)). The initiation of the Daliangshan fault makes the whole Xianshuihe- Xiaojiang fault smooth. It makes the striking of the fault system more consistent with the crustal motion pattern. Since the Daliangshan fault initiates, the whole Xianshuihe- Xiaojiang fault system slips more easily. In the second model, slip rates on the northwestern segment of the Xianshuihe fault increase to 12.9 mm/a. Slip rates on the Xiaojiang fault zone increase to mm/a. Slip rates on the central section of the Xianshuihe-Xiaojiang fault zone are distributed on the Annhehe-Zemuhe and the Daliangshan faults. Because the geometry of the Daliangshan fault is relatively simple and is consistent with trending of regional crustal motion, slip rates on the Daliangshan fault are higher than that on the Anninghe-Zemuhe fault. Slip rates on the Daliangshan fault are about 6.9 mm/a, and only mm/a on the Anninghe-Zemuhe fault. The complex fault geometry may contribute to lower slip rate on the Anninghe-Zemuhe fault. The sum of slip rates on the Anninghe-Zemuhe fault zone and the Daliangshan fault zones is nearly 10 mm/a. High effective strain rates outside the Xianshuihe- Xiaojiang fault system are concentrated in the region near the Moxi fault, the Xiaoxiangling fragment, and the region near Dongchuan, respectively. High effective strain rate that centralized near the Moxi fault and near Dongchuan is almost the same as that derived from the first model. However, the initiation of the Daliangshan fault zone changes the strain partitioning around the central section of the Xianshuihe-Xiaojiang fault zone. It makes strain rate localized on the Anninghe-Zemuhe fault zone and the Daliangshan fault zone. It also makes strain rate decline in the Xiaoxiangling fragment. Results produced by the second model indicate that the Daliangshan fault zone plays a significant role in strain partitioning in the western Sichuan area. However, there is a big discrepancy between the predicted and observed slip rates on the Daliangshan and the Anninghe-Zemuhhe fault zones. As the Daliangshan fault zone is a newly generated fault zone, the cohesion on the Anninghe-Zemuhe fault zones and the Daliangshanfault zone should be different.

9 WANG Hui, et al. Sci China Earth Sci January (2010) Vol.53 No.1 9 Figure 6 Predicted slip rates and effective strain rate outside the fault produced by the second model. (a) Predicted results when cohesion of the Daliangshan fault zone is the same as that of the Anninghe-Zemuhe fault zone; (b) predicted results when cohesion of the Daliangshan fault zone is twice of that of the Anninghe-Zemuhe fault zone. Therefore, we adjust the cohesion on the Daliangshan fault to fit the observed slip rates. The effective strain rate outside the fault changes with the adjustment of cohesion on the Daliangshan fault. If the cohesion on the Daliangshan fault zone doubles and reaches 20 MPa, the mean slip rate on the Daliangshan fault is about 3.3 mm/a. At the same time, slip rate on the parallel Aninghe-Zemuhe fault is mm/a. Slip rates on the Xianshuihe fault and the Xiaojiang fault are still higher than those predicted from the first model (Figure 6(b)). As a newly-generated fault zone, initiation of the Daliangshan fault zone not only makes the whole Xianshuihe- Xiaojiang fault system slip easier, but also adjusts strain partitioning around the central section of the whole fault system. The interaction between the parallel Daliangshan fault zone and the Anninghe-Zemuhe fault zone also affects strain partitioning around the central section of the whole Xianshuihe-Xiaojiang fault system. With its development, the Daliangshan fault might replace the role performed by the Anninghe-Zemuhe fault zone in the whole Xianshuihe- Xiaojiang fault [26]. In the future, the Xianshuihe-Xiaojiang fault zone may finally develop to a smooth arc section around the EHS. The strain partitioning processes produced by the evolution of parallel fault zones are also observed in other area [9, 13]. To further investigate the effect of the Daliangshan fault zone on strain partitioning in the study region, we construct the third model. In the third model, the Daliangshan fault zone replaces the Anninghe-Zemuhe fault zone. It indicates that the whole Xianshuihe-Xiaojiang fault zone consists of the Xianshuihe fault zone, the Daliangshan fault zone, and the Xiaojiang fault zone. The whole fault system is much smoother than it is in the first model. Striking altering of the fault system is more consistent with the crustal motion pattern within the study area. Results of the third model further show that the significant influence of the Daliangshan fault on strain partitioning in the western Sichuan area. The geometry of the Xianshuihe-Xiaojiang fault influences the regional strain partitioning. The motion between the South China Block and the Sichuan-Yunnan Block is mainly absorbed by the Xianshuihe-Daliangshan-Xiaojiang fault zone (Figure 7). Long-term slip rates predicted by the third model are much smoother than those produced by the first and the second models. Slip rates on the fault system are attenuated toward the south in the third model. Predicted slip rates on the Daliangshan section in the third model are higher than those on the Anninghe-Zemuhe segment from the first model. They are also higher than the sums of slip rates on the Daliangshan fault and the Anninghe-Zemuhe fault in the second model. At the same time, there is little effective strain rate located on the east of the Daliangshan fault zone. Previous studies mainly focused on the Anninghe- Zemuhe fault zone and ignored the activity of the parallel Daliangshan fault zone that also absorbs much crust shortening. Previous geological studies illustrate the imbalance

10 10 WANG Hui, et al. Sci China Earth Sci January (2010) Vol.53 No.1 titioning. Our models demonstrate the impact of the Daliangshan fault zone on the strain partitioning in the western Sichuan region. The results are also useful to explain the occurrence of historical devastating earthquakes in the region extending from the east of the Anninghe fault zone to the Mabian fault zone. 4 Influence of the Longmenshan fault zone Figure 7 Predicted slip rates and effective strain rate outside the fault produced by the third model. between the total displacements and slip rates on each fault segment of the Xianshuihe-Anninghe-Xiaojiang fault [27, 45]. It is found that there are large displacement deficits on the Anninghe-Zemuhe fault zone. Most of previous studies presumed qualitatively that the southeastward slip rates on the Xianshuihe fault may be distributed on the tectonics at east of the Anninghe fault zone [22, 46]. They ignored the impact of the Daliangshan fault zone on regional strain par- The Longmenshan fault strikes in northeast direction. It is also an important block boundary in the western Sichuan area. It connects the Xianshuihe-Xiaojiang fault zone in the study region. The two fault zones divided the western Sichuan region into three blocks. To discuss the impact of the Longmenshan fault, we construct the fourth model on the basis of our second model. We get the regional crustal motion pattern and the distribution of crustal strain in the fourth model (Figure 8). The activity of the Longmenshan fault makes the crust move like block motion in the western Sichuan region. The Bayan Har block moves northeastward. The Longmenshan fault zone absorbs the movement between the Bayan Har block and the South China Block. It is a dextral strike-slip fault with thrust component. Results obtained in the fourth model show that the impact of the active Longmenshan fault on the regional strain partitioning is less than that of the Xianshuihe-Xiaojiang fault. The slip rate on the Longmenshan fault is about 1.3 mm/a, and the thrust component is less than 1 mm/a (Figure 8(b)). The active Longmenshan fault reduces the effective strain rate around the Xianshuihe-Xiaojiang fault zone. In Figure 8 Results produced by the fourth model. (a) Predicted velocity field (red arrows show observed GPS velocity, black arrows show predicted velocity); (b) predicted fault slip rates and effective strain rate outside the fault.

11 WANG Hui, et al. Sci China Earth Sci January (2010) Vol.53 No.1 11 the fourth model, slip rate on each fault segment of the Xianshuihe-Xiaojiang fault zone decreases about 0.5 mm/a than that produced by the first and the second models. However, the distribution of effective strain shows that the Xianshuihe-Xiaojiang fault is more active than the Longmenshan fault. The insignificant slip on the Longmenshan fault is consistent with the results from historical seismicity and geodetic survey [6, 29, 47]. Difference between the activity of the Longmenshan fault and the Xianshuihe-Xiaojiang fault is caused by the geometry differences of these two fault systems. During the uplift of the Tibetan Plateau, huge gravity potential energy makes extension in the east-west direction in the plateau [48]. The eastward extrusion in the Bayan Har region is blocked by the steady Sichuan basin. Crust shortening occurred in the Longmenshan region. The Longmenshan thrust fault zone was formed during the shortening process. Its strike is orthogonal to the direction of regional crustal motion. At the same time, no old craton like the Sichuan basin exists in the southwest Yunnan region to block the crust motion. The lateral extrusion of the Tibetan Plateau causes the initiation of the sinistral Xianshuihe-Xiaojiang fault system. The different geodynamic environments and different fault strikes lead to the different activities on the Xianshuihe-Xiaojiang fault system and the Longmenshan fault system. Slip rates on the Longmenshan fault zone are much lower than that on the Xianshuihe-Xiaojiang fault system. Thus, the recurrence interval of major earthquakes on the Longmenshan fault zone is longer than that on the Xianshuihe- Xiaojiang fault system [49]. Historical earthquake records and paleoseismicty reveal millennium quiescent seismicity on the Longmenshan fault [50]. The accumulated strain in past millenniums should be enough to sustain the Wenchuan M s 8.0 earthquake on May 12th, 2008 on the Longmenshan fault zone. The vertical movement shows difference in southern and northern parts of the western Sichuan area (Figure 9). Crust uplift dominates the deformation in northwestern Sichuan region. The crust uplift rate around the Gongga Mountain is larger than 1.5 mm/a. The mean elevation is about m in the northwestern Sichuan region, whereas it is only 2500m in the central Yunnan [41]. Geological uplift rate of the Gongga Mountain is about 3.2 mm/a [5]. The predicted vertical movement pattern is consistent with the pattern obtained from the geological results. 5 Discussions Three-dimensional finite element model in the western Sichuan region in the paper is in fact a simplified mechanical model. The model describes the first order characteristics of the real geological unit. Although limited by the computational complexity and computing capability, the model is Figure 9 Predicted crustal uplift produced by the fourth model. significant for understanding regional strain partitioning process. 5.1 The effect of material parameters Many material parameters are used in our models. The important parameters are viscosity, effective friction coefficient, and cohesion etc. Li et al. [32] discussed the influence of different material parameters on predicted results in the three-dimensional viscoelastoplastic model in detail. Based on Li et al s results [32], we briefly introduce the influence of viscosity and effective friction coefficient on the modeling results. We take Pa s as the viscosity of upper crust in our three-dimensional model. Deformation time of material in upper crust is shorter than its relaxation time. The effective strain rate in upper crust is mainly elastic strain rate and plastic strain rate. The viscosity of the lower crust and upper mantle beneath the study area is about Pa s [20]. Previous study [11] shows that the less the viscosity difference is, the stronger the coupling between the upper and lower layer is. The effective friction coefficient of fault also affects the distribution of the effective strain rate outside the faults. A low friction coefficient was assumed on fault when simulating the long-term fault slip in many papers. In that case, the long-term fault activity is simplified as steady creeping. For example, He et al. [10] thought that the effective friction coefficient is lower than on the Xianshuihe- Xiaojiang fault zone. Zhang et al. [51] adopted as the effective friction coefficient on the Bangongcuo-Nujiang-Red River fault and the Zemuhe fault zone when simu-

12 12 WANG Hui, et al. Sci China Earth Sci January (2010) Vol.53 No.1 lating the recent crust movement and the fault activities in the Tibetan Plateau. The effective friction coefficient in our model is assumed as zero. The assumed low value is not the true rock inner friction coefficient. It is used to simulate the long-term slip rate on major fault zone, which may be equivalent to the mean dislocation of historical earthquakes in a given time span. Numerical experiments show that a small shift (0 0.1) of the effective friction is insignificant for modeling results, especially for predicted fault slip rates and distribution of effective strain rate outside the fault [11]. The constitutive model in the paper is nonlinear. The numerical solution therefore depends strongly on the initial stress field. The stress field produced by gravity is part of the initial stress field. Gravity does not affect neotectonic displacement field, such as fault slip rate, but it affects the plastic yield of material. It is necessary to study the impact of gravity when we study the distribution of the plastic strain rate. Different initial stress would produce different distribution of plastic strain. Gravity is ignored since we mainly focus on fault slip rate. Stress produced by the regional elevation difference and undulation of the Moho discontinuity may also be insignificant. Numerical experiments show that although material parameter variations would produce different fault slip rate and effective strain rate outside the fault, the spatial distribution of strain rate changes little in the case of small deformation. On a fault zone, slip rates are high on the segment with simple structure, and vice versa. Regional strain is localized around the fault segment with complex orientation and structure. Such deformation distribution makes regional crustal move more harmoniously. 5.2 The effect of other major faults in the western Sichuan area There are still other fault zones affecting regional strain partitioning in the western Sichuan area, including the Jinshajiang fault zone, Lijiang-Xiaojinhe fault zone, and the Red River fault zone etc. The structures of the Jinshajiang fault zone and the Lijiang-Xiaojinhe fault are complex. Activities on these two fault zones may be controlled by the motion of the deep strucutres. The Jinshajiang fault zone is a thrust fault system with dozen kilometers width. It consists of several arc-shape thrust faults protruding eastward [52]. The fault system not only changes the eastward extrusion of the Tibetan Plateau into crustal shortening and uplift in the western Sichuan region, but also transfers the remaining horizontal motion to the northwest Sichuan block. The geological right-lateral slip rate with reverse component on the Jinshajiang fault zone is about 5 mm/a [5]. The Lijiang-Xiaojinhe fault is a northeast striking fault that was formed on the Jinpingshan-Yulongxueshan thrust zone. The eastward extrusion of the Tibetan Plateau has been confined mainly in the west Sichuan area before Quaternary, which is located in the west of the Jinpingshan-Yulongxueshan thrust zone. The Jinpingshan-Yulongxueshan thrust zone is the largest topography transform zone in the southwestern China mainland, and is the sharp gradient zone of crustal depth between the central Yunnan region and the Western Sichuan Plateau [27]. The eastward extrusion of the Tibetan Plateau extends to the east of the Jinpingshan-Yulongxueshan thrust zone in Quaternary. The Lijiang-Xiaojinhe fault zone absorbs part of southeastward motion of the northwestern Sichuan region. Geological left-lateral slip rate with thrust component is 3.8±0.7 mm/a [5]. Although the Lijiang-Xiaojinhe fault is regarded as a sub-block boundary zone in the Sichuan-Yunnan Block based on the studies of surface active tectonics [53]. The earthquake relocation results show the connection of the Litang fault zone and western segment of the Lijiang- Xiaojinhe fault [18]. Therefore, the northern segments of the Lijiang-Xiaojinhe fault system (north of Muli) may play an insignificant role in strain partitioning around the eastern boundaries of the Sichuan-Yunnan Block. The Jinshajiang fault zone and the Lijiang-Xiaojinhe fault zone are ignored in our model for their tectonic complexity. Numerical experiments show that slip rates on these two fault zones are less than 0.5 mm/a. The influences of these two fault zones on strain partitioning on the Xianshuihe- Xiaojiang fault zone are also insignificant. The impact of the Red River fault is also ignored because we focus on strain partitioning between the Sichuan- Yunnan block and the South China Block. The influences of the Jinshajiang fault zone, the Lijiang-Xiaojinhe fault, and the Red River fault need further study. 6 Conclusions The western Sichuan and its vicinity is an important region to study the evolution and geodynamics of the Tibetan Plateau. The crustal motion presents altering from lateral shear on faults striking northwest-west to convergence with shear components on faults striking near north-south. The same crustal motion pattern is also found in other areas in the eastern borderland of the Tibetan Plateau, such as in the Haiyuan-Liupanshan fault zone and in the east Kunlun- Minshan fault zone etc. We provide a case study for the special deformation pattern in the eastern borderland of the Tibetan Plateau. The research is helpful to understand the geodynamics of the eastern Asia, especially the evolution of the Tibetan Plateau. Major conclusions of this study are: (1) Fault geometry obviously influences regional strain partitioning. Slip rates on each segment of the Xianshuihe- Xiaojiang fault are influenced by the fault geometry and fault structure. Slip rates are high on fault segment with simple geometry and structure, and vice versa. Regional strain is concentrated around the fault segment with complex geometry and structure.

13 WANG Hui, et al. Sci China Earth Sci January (2010) Vol.53 No.1 13 (2) Strain partitioning around the central section of the Xianshuihe-Xiaojiang fault zone is controlled by interaction between the parallel Anninghe-Zemuhe fault zone and the Daliangshan fault zone. The initiation of the Daliangshan fault makes the whole Xianshuihe-Xiaojiang fault zone slip much easier. (3) Striking of the Longmenshan fault zone is nearly orthogonal to the direction of lateral extrusion in the Tibetan Plateau. It leads to low slip rates on the thrust fault. The activity of the Longmenshan fault decreases the strain rate on and around the Xianshuihe-Xiaojiang fault zone. (4) The impact of fault geometry on strain partitioning is significant in the western Sichuan and its adjacent region. Eastward motion about 13 mm/a in the eastern Tibet is absorbed mainly by the deformation in the Xianshuihe- Xiaojiang fault system and the Longmenshan fault zone. Slip rates on the Longmenshan fault zone are about 1 mm/a. The remnant motion is absorbed by slip on the Xianshuihe- Xiaojiang fault zone and strain outside the fault. The whole Xianshuihe-Xiaojiang fault can be divided into three sections based on their different motion pattern. The north section consists of the Xianshuihe fault zone. The left-lateral slip rate is about 12 mm/a. The uplifting of the Gonggashan region near the Xianshuihe fault zone absorbs part of crustal strain. The central section of the Xianshuihe-Xiaojiang fault system consists of the parallel Anninghe-Zemuhe fault and the Daliangshan fault. Slip rates on these two fault zones change with the change of fault cohesions. The southern section of the fault system is the Xiaojiang fault. Slip rate on the fault is about 8.0 mm/a. Appendix: control equation for three-dimensional viscoelastoplastic model Three-dimensional viscoelastoplastic model used in the paper is developed based on classical elastoplastic model. Viscosity term is added to classical elastoplastic model. Physical equations that control the model include static equilibrium equation: σ ij + fi = 0, (A1) x j where σ ij is stress tensor, f i is body force. Stress increment consists of viscous component, elastic component, and plastic component: v e p { dε} = { dε } + { dε } + { dε }, (A2) where ε v, ε e and ε p are viscous stress increment, elastic stress increment, and plastic stress increment, respectively. {} represents tensor form of stress increment. When stress is beyond the yield, relationship between stress and strain follows the rule that controls the Maxwell material. Viscous strain component is related with stress. Elastic strain component is related with stress increment: v 1 { dε } = [ Q] { σ } dt, e 1 { dε } = [ D] { dσ}, t (A3) where, {σ t } stress tensor at t moment, dt is time increment, {dσ} is increment of stress tensor, [Q] is matrix for viscous material parameter, [D] is matrix for elastic material parameter [ Q] = 1 1 1, q (A4) E [ D] = (1 + ν)(1 2 ν) 1 ν ν ν ν 1 ν ν ν ν 1 ν 0 0 0, ν ν ν (A5) where q is viscosity, E is Young s module and v is Poisson s rate. {σ t } is stress at t moment: t t dt { σ } = { dσ} + { σ }. (A6) According to eqs. (A2), (A3), and (A6), we get p { dσ} = [ D% ]({ dε} { dε }) + { d % σ }, (A7) where [ D% ] = ([ D] + [ Q] dt), (A8) 1 t dt { d % σ} = [ D% ][ Q] dt{ σ }. Plastic deformation occurs when stress reaches the plastic yield criterion (yield envelope), described here by the Drucker-Prager yield function F = α I + J k (A9) 1 2, where I 1 is first invariant of the stress tensor and J 2 is second invariant of the deviatoric stress tensor. The parameters α and k are related to cohesion and effective frictional coef-