The Yabrai range-front fault accommodates deformation within the middle Gobi Alashan block between the Tibetan Plateau and the Ordos block. As such, it provides the opportunity to examine the transition between contractional deformation associated with the growth of the Tibetan Plateau and extensional deformation across North China. Geomorphic mapping of the active fault trace and trench investigations reveal that the Yabrai range-front fault is composed of three segments of varying fault strike, but for which the sense of motion, scarp height, and slip history appear to be kinematically compatible along the fault. Displaced Holocene and late Pleistocene alluvial deposits indicate that the southwestern segment is characterized by oblique-normal displacement with a minor sinistral component, whereas the middle segment appears to exhibit nearly dip-slip normal displacement. In contrast, slip along the northeastern segment appears to be primarily sinistral strike-slip with a minor reverse component. Geomorphically fresh fault scarps are developed within late Pleistocene–Holocene alluvial fans and terraces along the southwestern and northeastern segments, whereas the middle segment of the fault defines the bedrock-alluvial contact along the range front. The 10Be exposure ages of displaced alluvial fans along the southwestern segment yield a throw rate of ∼0.1 mm/yr over late Pleistocene time. Lateral slip rates along the northeastern fault segment range between 0.23 ± 0.02 and 0.78 ± 0.12 mm/yr. Regionally, the orientation and sense of motion along the Yabrai range-front fault are consistent with NE-SW shortening, and we suggest that recent activity along this fault system reflects incipient deformation of the foreland at the northeastern margin of the Tibetan Plateau.

The Gobi Alashan block is located in the transitional zone between the northeastern Tibetan Plateau, which was dominated by compressional deformation in late Cenozoic time (e.g., Yin and Harrison, 2000), and the Ordos block, which was bounded by extensional fault systems (Fig. 1; e.g., Xu et al., 1993; Zhang et al., 1998). The low topographic relief, limited seismicity, and few active faults in the Gobi Alashan region have led most workers to consider this to be a relatively stable block in the foreland of the Tibetan Plateau (Tapponnier and Molnar, 1977; Xu et al., 2010). More recently, workers have begun to recognize that fault systems within the Gobi Alashan block may indeed be active (Tapponnier et al., 1982, 2001; Yue and Liou, 1999; Meyer et al., 1998; Yue et al., 2001a; Yin et al., 2002; Darby et al., 2005), albeit at low strain rates. However, there is little information with which to assess the kinematics of active deformation within the Gobi Alashan block (e.g., Webb and Johnson, 2006; Xu et al., 2010). Because the region lies between the North China block, a relatively coherent block within the Asian tectonic collage (e.g., Yin and Nie, 1996; Zhao et al., 2005), but one that experienced NW-SE extension during the late Mesozoic and early Cenozoic (Ma and Wu, 1987; Ye et al., 1987; Northrup et al., 1995), and the Tibetan Plateau, a block experiencing northeastward extrusion at the northeastern margin during the late Cenozoic, assessing the kinematics of active faulting will provide insight into the interaction of these two strain regimes.

The collision between India and Eurasia generated widespread Cenozoic deformation throughout the continental interior of Eurasia, including both the Himalaya-Tibetan orogen and its peripheral regions (e.g., Molnar and Tapponnier, 1975; Tapponnier and Molnar, 1979; Avouac and Tapponnier, 1993; Meyer et al., 1998; Tapponnier et al., 2001; Thompson et al., 2002; Zhang et al., 2004; Darby et al., 2005; Yuan et al., 2013; Zheng et al., 2013a). In northeastern Tibet, shortening and crustal thickening across the Qilian Shan (Burchfiel et al., 1989; Tapponnier et al., 1990; Meyer et al., 1998) have been linked to sinistral slip along the Altyn Tagh fault since at least Oligocene time (Yue et al., 2001b). Although the modern slip rates along the Altyn Tagh fault have been the subject of much study (e.g., Peltzer et al., 1989; Van der Woerd et al., 2002; Ryerson et al., 2003; Mériaux et al., 2004, 2005; England and Molnar, 2005; Cowgill, 2007; Cowgill et al., 2009; Gold et al., 2009, 2011; Zhang et al., 2007), with most data now suggesting rates of ∼10 mm/yr (e.g., Zhang et al., 2007; Cowgill et al., 2009), the history of slip through time is less well known (Ritts and Biffi, 2000; Yue et al., 2001b). In particular, some workers have suggested that the Altyn Tagh fault was once connected with faults in the Gobi Alashan block during early Cenozoic time (Yue and Liou, 1999; Darby et al., 2005). These suggestions are based largely on interpretations of satellite images and regional geological maps (Lamb et al., 1999; Yue and Liou, 1999; Yue et al., 2001a, 2001b; Darby et al., 2005; Webb and Johnson, 2006). Thus, whether or not faults within the Gobi Alashan block are associated with shortening outboard of the Tibetan Plateau is an important, yet unanswered, question.

In this paper, we present geomorphologic and geologic evidence for active faults along the Yabrai range front, one of several topographically prominent ranges in the Gobi Alashan region. We combined interpretation of high-resolution satellite imagery, geomorphic mapping along the fault, and shallow trench investigations to determine fault geometry. Kinematics and rates of recent fault slip were derived using displaced alluvial-fan surfaces as markers of fault slip and cosmogenic exposure age dating to estimate the ages of alluvial surfaces. Combining our data and previous studies on the Qilian Shan and the Hexi Corridor, we discuss the significance of the deformation pattern in the Gobi Alashan block.

The Gobi Alashan block, with an average elevation of ∼1300 m, is composed of a basement of Precambrian metamorphic rocks intruded by Paleozoic plutons, overlain by Jurassic and Cretaceous sedimentary rocks (Bureau of Geology and Mineral Resources of Gansu Province, 1989; Vincent and Allen, 1999). The region is largely devoid of Cenozoic sediment, with the exception of Quaternary eolian and alluvial deposits. Regionally extensive series of extensional basins, metamorphic core complexes, and rift-related volcanic deposits attest to regional extension during the late Mesozoic (e.g., Kimura et al., 1990; Zorin et al., 1995; Yin and Nie, 1996; Zorin, 1999; Graham et al., 2001; Ren et al., 2002; Meng, 2003). Although the cause of this deformation is still debated (Meng, 2003), it may have been related to crustal extension following the collision between the amalgamated North China and Mongolia-Siberia plate in early Mesozoic time (e.g., Graham et al., 2001; Meng, 2003).

Within the Gobi Alashan block proper, several active faults occur along a series of isolated ranges north of, and parallel to, the Qilian Shan, including the Heli Shan fault, Beida Shan fault, Ayouqi fault, and Longshou Shan fault (Fig. 2); all of these were developed along isolated high-standing ranges in the Hexi Corridor. To the northeast, however, active faults are recognized along the Bayanwula Shan and the Yabrai range front (Fig. 2), but the kinematics of these systems are not well known, with workers alternatively suggesting an association with deformation in the Qilian Shan (Tapponnier and Molnar, 1977; Yue and Liou, 1999; Deng et al., 2003; Darby et al., 2005) or with extension around the Ordos block (Northrup et al., 1995; Meng, 2003). Global positioning system (GPS) velocities indicate that the southern Gobi Alashan block is undergoing relatively coherent eastward movement at rates of ∼5 mm/yr with respect to stable Eurasia (Wang et al., 2001; Zhang et al., 2004; Gan et al., 2007); however, deformation within the block itself is difficult to resolve in these data. Historic seismicity is also sparse, and the few large earthquakes (∼Ms 7.0 Minqin earthquake 1954; Liu et al., 2000) do not provide much insight into the regional tectonics (Tapponnier and Molnar, 1977; Xu et al., 2010).

The Yabrai range-front fault, with a length of ∼80 km, marks the boundary between the Yabrai Shan and Yabrai Basin (Fig. 3). Tapponnier et al. (2001) and Meyer et al. (1998) considered the Yabrai range-front fault as a normal fault in their morpho-tectonic map (their fig. 1 and fig. 15a, respectively), an interpretation that is supported by industry seismic profiles (Wu et al., 2007; Zhong et al., 2010). Later workers extended this interpretation to suggest that the Altyn Tagh fault zone could be continued eastward as a series of active faults across the Gobi Alashan block to the Yabrai Shan (Meyer et al., 1998; Tapponnier et al., 2001; Darby et al., 2005). Darby et al. (2005) developed this idea most fully, suggesting that the northeastern segment of the Yabrai range-front fault developed as a sinistral strike-slip fault prior to Miocene time, while the southwestern and middle segments slipped as normal faults throughout the Cenozoic. These authors envisioned a regionally extensive fault system that linked slip along the Altyn Tagh fault to extensional faults north of the Ordos block (see figs. 3, 4D, and 4E inDarby et al., 2005). However, most of these interpretations were based on satellite image analysis with only limited field investigations. Here, we investigated the geomorphology of alluvial deposits along the range front, documented the geometry and slip along the Yabrai range-front fault where it has displaced the alluvial deposits, and determined late Quaternary slip rates from cosmogenic dating of alluvial fans.

Southwestern Segment

Beginning west of Yabrai town, the southwestern segment of the Yabrai range-front fault strikes ∼N60°E and extends for ∼35 km along the base of the southeastern flank of the Yabrai Shan (Fig. 3). An industry seismic profile across the southwestern segment of the Yabrai range-front fault near the town of Yabrai displays typical half-graben geometry of the Yabrai range-front fault at a depth of 1–4 km, with an asymmetric basin floored by tilted Mesozoic strata and with the depocenter of the Yabrai Basin located adjacent to the range-bounding fault (Wu et al., 2007; Zhong et al., 2010; Fig. S11). These geometries suggest that the southwestern segment of the Yabrai range-front fault has had a significant component of normal slip since the late Mesozoic to early Cenozoic. Presently, the fault is defined at the surface as a linear contact between the bedrock and Quaternary alluvium and decorated by fault scarps that displace alluvial-fan surfaces (Fig. 4). Along much of the fault trace, well-preserved alluvial free faces and bedrock fault planes are exposed; fault striae observed on some of these planes range from downdip to oblique (Darby et al., 2005). As documented later herein, our observations confirm the interpretation that slip along the southwestern segment of the Yabrai range-front fault was largely normal with a component of sinistral strike slip.

The southeastern margin of the Yabrai Shan is marked by an abrupt, steep range front of granite or metamorphic rocks produced by dip slip along the Yabrai range-front fault (Fig. 3). The difference in elevation between the mountain and basin increases from tens of meters at the southwestern tip of the range to more than 400 m along the strike of the fault. Coarse clastic debris from the interior of the range has been deposited as alluvial fans along the southeastern pediment of the Yabrai Shan (Figs. 4A and 4B). Three regionally extensive alluvial-fan surfaces (labeled Q1, Q2, and Q3) are recognized across the piedmont (Figs. 4A and 4B). The youngest alluvial fans (Q1) are the most extensive and represent alluvium that is present in active washes or on young fans; these surfaces do not appear to have been offset by faulting along the range front. Q2 alluvial fans are preserved at elevations ranging from ∼1 to 3 m above the Q1 fans. Surface clasts are largely unweathered and lack desert varnish, suggesting relatively young age. Fault scarps with heights of 1–1.5 m are preserved on this group of fans. Q3 alluvial surfaces are only sporadically preserved along the fault trace; these surfaces exhibit well-developed pavement, and thin, dark brown varnish is present on surface clasts. Fault scarps that displace the surface of Q3 fans are typically in the range of 4–6 m. Older alluvial fans are preserved locally with fault scarps in height up to ∼20 m, but these surfaces are usually found as isolated remnants of alluvial fans uplifted on the footwall block of the fault and cannot be correlated to deposits on the hanging wall.

The linear to curvilinear fault scarps in the southwestern segment of the Yabrai range-front fault are continuous and truncate distinct alluvial fans (Fig. 4B). The fault trace in Quaternary deposits consists of a single strand of fault scarps along most parts of this segment, and it marks the main contact between the range-front alluvium and the uplifted mountain (Fig. 4C). The fault scarps consistently face toward the southeast, with heights ranging from 1 to 20 m (Fig. 4D). The youngest fault scarps truncate the Q2 alluvial fan, with heights ranging from 1 to 1.5 m, and can be recognized along the entire southwestern segment. Free faces are preserved at the top of these young scarps (Fig. 4E), which might suggest that the most recent earthquake that occurred in this area is relatively young. Other than scarps developed in alluvium, the fault trace lies between the bedrock surfaces and undeformed alluvium, marking the sharp contact between the uplifted mountain and the basin (Fig. 4F). Bedrock surfaces are exposed at the base of the mountain and dip to the southeast at 40° to 60°. The base of these fault facets exhibits polish up to 5 m up dip from the bedrock-alluvium contact. Streams developed in the bedrock perpendicular to the bedrock surfaces were uplifted by 0.5 m to several meters in the footwall of the fault (Fig. 4F), which is inferred to reflect dip-slip motion in the most recent earthquakes.

To investigate the subsurface geometry of the fault along this segment, we excavated a trench across a fault scarp on an alluvial-fan surface where the fault trace faulted the root of the alluvial surface near the base of the range front (Fig. 5A). In the trench, offset strata, tilted beds, and the dip of faults are all consistent with normal slip (Figs. 5B and 5C). Strata exposed in this trench are composed of metamorphic bedrock, gravels, and colluvium, which are mainly displaced by three synthetic normal faults (F1, F2, F3) that dip 40° to 50° to southeast. Major vertical offset occurred on faults F1 and F3. The deposit units on each side of these two faults are completely different. Unit 1 is metamorphic bedrock that only lies under fault F1 and is overlain by gravel units (units 2–9). Units between faults F1 and F3 (units 2–9) are gravel and coarse to fine sand layers; a number of these layers (units 2–6) can be correlated across F2. Bedding direction of gravels suggests that these layers have been titled toward fault F1. This geometry is consistent with block rotation during normal faulting (e.g., Jackson and McKenzie, 1983). Units 10–16 are colluvial and alluvial deposits in front of the scarp developed above F3. The lack of bedding fabrics makes it difficult to correlate these beds to the titled units (units 2–9). We infer that the original sediments of these beds on the footwall of F3 have been eroded, suggesting a vertical offset greater than 6 m on fault F3. This trench is located on the main stand of the fault scarps between the uplifted bedrock and range-front alluvial fans, which is the most common faulted geomorphic expression in the southwestern segment. Therefore, the geometry revealed in the trench may represent a general geometry of faults near shallow surfaces.

Limited evidence exists for sinistral slip in the southwestern segment (Fig. 6). The most definitive evidence for lateral slip was found along a relatively short 5 km section (Fig. 6A), where a series of gullies appears displaced across fault scarps in a sinistral sense. As shown in Figure 6A, the southeast-facing linear fault scarp truncates alluvial-fan surfaces along the base of range front. The base of the alluvial fan is usually preserved on the footwall of the fault with vertical separation ranging from 3 to 6 m from the downthrown block. Gullies developed on the alluvial surfaces in this section are usually displaced by the fault trace in sinistral sense by ∼1–3 m. Some alluvial risers between surfaces are also displaced in sinistral sense where they appear to have been protected by the upstream scarp (Fig. 6A). We conducted a topographic survey at one of these alluvial surface (Fig. 6B), where the surface was displaced by ∼1.5–2 m. The bottom of a small gully on the surface was offset by ∼2 m (Figs. 6B and 6C), and the riser of the alluvium was offset in sinistral sense by ∼8 m (Fig. 6B). Although potential lateral erosion of the riser makes it difficult to quantify whether the apparent offset is entirely related to fault slip, the consistency of sinistral offset for numerous gullies along this section of the fault implies a non-negligible component of sinistral slip.

Middle Segment

The middle segment of the Yabrai range-front fault strikes ∼N40°E and is ∼30 km in length, extending northeastward to the mouth of the Shazaogou Canyon along the range front (Fig. 3). In addition to the fault orientation, the most conspicuous difference between the middle and southwestern segments lies in the geomorphological expression of the fault trace. The fault trace along the middle segment defines the contact between the bedrock and relatively undeformed late Quaternary deposits (Fig. 7A). High escarpments generated by repeated dip-slip faulting in this segment dominate the landscape (Fig. 7A), with up to 500 m of relief, and polished fault surfaces along the bases of the faceted range front extend a few meters above the ground surface (Figs. 7B, 7C, and 7D). The fault surfaces dip to the southeast at 40° to 70°, and well-preserved facets exhibit dark varnish (Figs. 7B and 7D). Hanging valleys and “wineglass” canyons are common along the bedrock fault plane (Fig. 7E). Several knick points were developed at the bottoms of these hanging valleys, suggesting that the fault has repeatedly occupied the same trace along the bedrock plane. The fault surfaces are considered to have formed in repeat earthquakes as normal slip planes (e.g., Hancock and Barka, 1987; Jackson and McKenzie, 1999) and offer opportunities for a variety of structural investigations, including descriptions of slip vector azimuth, surface roughness, and exposure ages related to paleoearthquake history (e.g., Jackson and McKenzie, 1999; Schlagenhauf et al., 2010). Unfortunately, we did not find any linear striae or other kinematic indicators on the polished fault surfaces because the original granite surfaces have either weathered, or did not develop strong striae.

To the northeast of Aguimiao, the main fault trace continues along the contact between the bedrock and late Quaternary sediments, and bedrock fault planes can be observed (Fig. 3). A broad zone of discontinuous scarps, with heights ranging from <1 m to 2–3 m, is developed across the piedmont fans within a zone of ∼2 km to the southeast from the main range-front fault. These fault scarps merge along strike to the northeast with the main range-front fault trace and then continue as a single bedrock fault plane to the mouth of Shazaogou Canyon. Most of these discontinuous scarps face southeast and appear to represent a series of normal faults stepping out from the base of the range into the basin.

Northeastern Segment

The northeastern segment of the Yabrai range-front fault strikes ∼N75°–85°E and extends ∼26 km from the mouth of Shazaogou Canyon along the range front (Fig. 3). Darby et al. (2005) observed subhorizontal fault striae on outcrops near the western end of Shazaogou Canyon, within the Yabrai Range (Fig. 3). As noted previously, these authors regarded the Shazaogou Canyon fault and its continuation along the northeastern segment of the Yabrai range-front fault as the northeastern extension of the Altyn Tagh fault during early Cenozoic time (Darby et al., 2005). On the basis of satellite imagery, this segment of the Yabrai range-front fault was regarded as an active normal fault during the Quaternary (Meyer et al., 1998; Tapponnier et al., 2001). Oil exploration in the Yabrai Basin suggests that the depocenter of the basin is located in the southwest of the basin, with ∼5000 m of sedimentary rocks, while the northeastern part of the basin has been considered as an uplifted region (Wu et al., 2007, 2015). However, no field investigation on late Quaternary active faults was carried out on the northeastern segment of the Yabrai range-front fault.

In contrast to the southwestern and middle segments of the Yabrai range-front fault, the fault trace along the northeastern segment is not defined as a single strand along the base of the mountain, but it consists of a distributed network of faults that lie ∼500–1000 m south of the range front (Fig. 8). Along the western 4–5 km of the segment, the fault is marked by a single scarp on the alluvial fans (Fig. 8B). Farther east, however, the fault trace bifurcates into two main strands characterized by fault scarps that face opposite directions; scarps developed along the northern strand face south, while those developed along the southern strand face north (Fig. 8B). These appear to define a graben system in a left step from the southern to northern strands (Fig. 8B), suggestive of sinistral slip. Near site 2 (Fig. 8B), several subparallel fault scarps developed within alluvium suggest a complex fault array (Fig. 8B). In order determine the sense of motion along the northeastern segment of the Yabrai range-front fault, we mapped the fault zone on WorldView satellite images with 0.5 m resolution, and we excavated five trenches across fault scarps. Each of these sites is discussed separately next.

Site 1 (39°48′49.93″N, 103°26′38.82″E)

The first site is located near the eastern tip of the northeastern segment (Fig. 8). Three remnant alluvial terraces (T1–T3) are present at this site and have been displaced across the fault (Fig. 9). Our mapping and interpretation of satellite imagery reveal that the fault is characterized by a single linear trace that displaces terrace risers and streams in a sinistral sense (Fig. 9A). The terrace riser T1/T3 is displaced by 40 ± 5 m (Fig. 9B); a stream on terrace T2 is displaced by 20 ± 3 m (Fig. 9C); and two subrisers on the T1 surface are displaced by 32 ± 2 m (Fig. 9D). An offset tiny stream and beheaded channel on terrace T1 also suggest a sinistral sense of slip (Fig. 9D). Systematic differences in height of the alluvial fans on terraces T2 and T3 suggest a minor vertical component of slip; vertical separation of 3 m and 6–8 m is observed (Figs. 9E and 9F). However, there is no significant vertical offset (<0.5 m) of terrace T1 (Fig. 9G). Notably, vertical separation of these surfaces is small relative to lateral displacement. To determine whether the vertical separation results from normal or reverse faulting, we excavated a trench (trench A) across the slope of the fault scarp on T2 (Figs. 9B, 9E, and 10).

The strata in this trench are composed of Neogene sandstone and late Quaternary alluvial gravel units (Fig. 10A; Fig. S2 [see footnote 1]). Unit 1 represents Neogene sandstones that were thrust over the late Quaternary gravels of unit 3. Clast alignment of gravels and the geometry of the interbedded Neogene units involved in the fault zone suggest that the dip of the fault is 30°–40° to the north (Fig. 10B). The clasts near the fault zone within unit 3 were curved upward, suggesting a reverse component on the fault zone. Field observations suggest that there is another minor scarp ∼2–3 m to the north of the north wall of the trench, which offset the fan surface slightly. However, large boulders (>2 m) in the trench made it difficult to dig further north. Considering the fault geometry revealed by our excavation, and sinistral displaced geomorphological features at this site, we infer that a primary vertical fault strand is located farther north than our trench. We consider our excavation to have revealed a portion of a positive flower structure fault geometry (Fig. 10C). Thus, the vertical separations on T2 and T3 appear to reflect a reverse-slip component during sinistral displacement.

Site 2 (39°47′58.73″N, 103°24′14.88″E)

A second site, ∼3 km farther west of site 1, also shows evidence for sinistral strike slip (Fig. 11). The fault is marked by a continuous scarp within alluvium (Fig. 11A). Two alluvial surfaces are preserved at this site (Figs. 11B and 11C); the older one has darker varnish on the clasts on the surface and sits above the younger one. Both of the two fans are displaced by the fault. However, the fault scarp on the younger alluvial surface faces northwest with a height of 0.5–1 m, whereas the fault scarp on the higher and older surface shows no relative vertical displacement (Figs. 11B and 11C). To the east of site 2, the fault trace defines a boundary between white sandstones and black metamorphic rocks with no apparent vertical separation (Figs. 11A and 11D), and the stream channels across the fault trace are deflected by ∼2 m in a sinistral sense (Fig. 11E).

Trench B was excavated ∼12 m long, ∼3 m deep across a fault scarp facing north, the height of which was ∼1 m (Fig. 11A). This trench exposed Neogene sandstones (units 1 and 2) and late Quaternary alluvial gravels (units 3–7) displaced by the fault (Fig. 12A; Fig. S3 [see footnote 1]). The Neogene sandstones are massive with no bedding, and the alluvial gravels show subhorizontal bedding. Except for the cover layer of unit 7, all the alluvial layers are deposited in the downthrown block. The main fault zone is composed of two fault splays with variable dip angles; one is near vertical, and the other curves upward with a dip of 50°–70° to the south. The two fault splays appear to merge near the base of the excavation (Fig. 12B). A secondary fault between unit 1 and unit 2 terminates against the base of the alluvial layers. Units 3–6 were faulted by the two faults at the main fault zone, with vertical separation of ∼1 m at the vertical one and ∼0.5 m at the curved one based on the displacement of the top of unit 1. Sandstones between these two splays appear to have been extruded upward as a triangle shape, potentially consistent with a positive flower structure related to transpressional deformation (Sylvester, 1988).

Site 3 (39°47′10.92″N, 103°20′18.27″E)

The third site is located on the southern strand of the northeastern segment (Fig. 8B). A north-facing fault scarp, with height of ∼20–50 cm, is present within the youngest alluvial fan and may represent the last coseismic slip. Trench C was excavated across this fault scarp to ∼2.5 m in depth through loosely consolidated eolian and alluvial deposits (Fig. 13). At the bottom of the trench, a fine eolian deposit with no gravels is present (unit 1) and exhibits cross-bedding. This unit is prone to collapse and prevented further excavation. Unit 2 is a white sand layer with thickness of ∼5 cm, which is calcified and more compacted than unit 1. This unit developed discontinuously on top of unit 1. Unit 3 is an alluvial gravel interlayer with different sizes of clasts and cobbles. It is not continuous across the exposure. A vertical fault displaced these deposits to the surface with vertical displacement of ∼0.4 m (Fig. 13B), suggestive of only one single event. Several tiny faults displace units 1 and 2 and disappear upward.

Site 4 (39°47′0.24″N, 103°18′33.76″E)

From the mouth of Shazaogou Canyon to the location of site 4, the fault trace of the Yabrai range-front fault is primarily concentrated along a single south-facing strand (Fig. 8B). The fault scarp is continuous near the base of the mountain. Near site D, two alluvial fans are preserved (Fig. 14). The higher one (T2) is preserved along a gully and the range front. The surface is flat, and clasts exhibit dark varnish on clasts. T1 is younger, and only thin varnish is present on surficial clasts. This surface has been heavily eroded by active channels, and it exhibits bar-and-swale topography. The T2 surface is displaced by a sinistral sense to form an uphill (north)–facing shutter ridge (Fig. 14). Streams to the north of the shutter ridge are deflected to the east. To the east of the shutter ridge, an uphill-facing scarp truncated alluvial surfaces, with different facing direction from the scarp on T1 further to the east. To the west of the shutter ridge, a small inset surface is preserved adjacent to T2. It is difficult to identify whether this surface is equivalent to T1 or represents a partially eroded remnant of T2, but geomorphic characteristics of the degree of surface varnish suggest the former. This interpretation suggests a sinistral offset of the riser of T2 with displacement at least of 58 ± 5 m.

A trench (trench D) ∼14 m long and 3–4 m deep was cleared from a gully wall that is perpendicular to the fault scarp with height of ∼3.5 m (Fig. 14). Strata exposed in the trench are metamorphic bedrock and alluvial-fan gravels derived from the Yabrai Shan (Figs. 15A and 15B; Fig. S4 [see footnote 1]). Unit 1 is metamorphic bedrock interbedded with red sandstones. Units 2–4 are composed of alluvial gravels, and the bedding of gravels suggests that these three layers are tilted and dip to the south at ∼30°. Units 5 and 6 are alluvial gravels with horizontal bedding. Units 7 and 8 are alluvial gravels covered at the top of the trench, and these appear to not have been displaced. Vertical separation occurred on the faults in the main fault zone between meters 3 and 6. Red sandstones in the main fault zone are extruded upward upon the gravels. Another fault zone, between meters 7 and 9, consists of three low-angle (∼45°) faults. The orientation of the gravel layer and aligned clasts of unit 4 near the left-most fault (Fig. S5 [see footnote 1]) illustrate that reverse motion occurred on this fault. We infer that the low-angle faults merge with the high-angle faults at depth (Fig. 15C). The geometry revealed in this trench is similar to that in trench B, with a system of vertical faults and low-angle reverse faults that collectively suggest positive flower structures.

Site 5 (39°46′41.77″N, 103°16′35.64″E)

At this site, we recognized a push-up ridge (Fig. 16A), which is generally a feature associated with transpressional deformation along strike-slip faults (e.g., Mann et al., 1983). The push-up ridge is ∼700 m in length and 200 m in width and developed along the major fault trace of the northeastern segment of the Yabrai range-front fault. The surfaces of the alluvial fans exhibit bar-and-swale topography, suggestive of relatively young age (Fig. 16A). Two strands of fault scarps are present on either side of the push-up ridge and face away from the ridge itself (Fig. 16A). A locally higher remnant alluvial fan is preserved at the west tip of the push-up ridge and has been offset by the northern fault with a sinistral separation of ∼3–5 m (Fig. 16A, inset, and 16B). Farther east, the fault scarp is continuous and faces opposite directions, changing from facing north to south, suggesting a subvertical geometry of the fault.

Site 6 (39°47′42.92″N, 103°20′54.99″E)

Parallel to the south strand of fault scarps, another strand is developed along the range front on alluvial fans close to the bedrock (Fig. 8). We did not find definitive geomorphic evidence for strike-slip motion on this strand. To characterize the sense of motion of the fault, we excavated a trench (trench E) across the fault scarp. The trench was cleared from a gully wall where the fault scarps was obvious on the surface with height of ∼90 cm (Figs. 17A and 17B). The trench was ∼12 m long and 3–4 m deep and exposed alluvial-fan gravels in which it was difficult to parse the units (Fig. 17C; Fig. S5 [see footnote 1]). We interpreted two reverse faults dipping ∼35°N from the oriented clasts and offset boundary between the coarse and thin gravel strata. The fault scarp on the alluvial-fan surface appears between these two reverse faults at meters 2–4. We did not find any other vertical or low-angle faults on the natural exposure along the gully wall further north. Thus, the trench exposure suggests that the northern section is a reverse fault with low dip angle (∼35°N). Notably, this geometry implies that the apparent left step in the fault system between sites 4 and 6 (Fig. 8) is not a simple releasing step in a sinistral fault system, but rather reflects a complicated array of reverse and strike-slip fault strands.

In summary, along the southern strand of the Yabrai range-front fault, geomorphic features indicate a sinistral sense of slip; displacement of the T1/T3 terrace riser is ∼40 m. The trench exposures of trenches A, B, and D reveal positive flower structures, which are consistent with local transpressional deformation along a strike-slip fault. Along the northern fault strand, the geometry of a north-dipping fault suggests a significant component of reverse separation. Collectively, these observations point to sinistral strike slip with a component of reverse slip along the northeastern segment of the Yabrai range-front fault during late Quaternary time.

Our field investigations demonstrate that the southwestern segment of the Yabrai range-front fault is primarily a normal fault with a minor component of oblique, sinistral displacement, whereas the northeastern segment is primarily a sinistral strike-slip fault with a subordinate reverse component of motion. These faults displace numerous alluvial-fan surfaces and present an opportunity to place bounds on the rates of fault slip. We focus on two sites along the southwestern segment where preservation of multiple alluvial surfaces allows estimation of the vertical slip rates and on one site along the northeastern segment where we can reconstruct lateral strike-slip rates.


Determination of Displacement along the Yabrai Range-Front Fault

We surveyed displaced alluvial surfaces with a global navigation satellite system (GNSS, Trimble R8) differential positioning system. Across normal fault scarps, we measured the vertical separation of alluvial-fan surfaces along topographic profiles perpendicular to the fault scarp trace. In the field, we selected preserved depositional surfaces on both sides of the fault scarps. However, it is common that an alluvial surface may have continued to accumulate sediment on the downthrown side of the fault and degrade on the upthrown surface. In either case, it will result in a minimum estimate of fault offset. We estimated vertical offset across fault scarps by fitting linear trends through upthrown and downthrown alluvial surfaces and measured the vertical separation between them at the scarp (e.g., Hanks, 2000; Thompson et al., 2002). We selected the linear trends that approximated the original surfaces, not eroded or disrupted areas near the scarp that were identified in field. Data collected during each survey were projected into the direction perpendicular to fault scarps, providing a more realistic estimate for profiles that are perpendicular to fault scarps. Vertical separation was measured at the fault scarp zone to derive an average vertical displacement.

Terrace risers are generally used in the determination of lateral displacement along strike-slip faults (e.g., Sieh and Jahns, 1984; Weldon and Sieh, 1985; Van der Woerd et al., 1998, 2002; Zhang et al., 2007). We used the intersection of the riser crest to estimate the lateral offset, but not the base of the risers, because the accumulation of colluvial materials obscured the precise intersection of the riser and lower terrace tread. Horizontal displacements were estimated from both high-resolution imagery (nominal pixel size of 0.5 m) and detailed topographic surveys with the GNSS positions. Horizontal slip rates that rely on terrace risers are subject to uncertainties in the interpretation of the beginning of terrace riser displacement (e.g., Lensen, 1964; Van der Woerd et al., 2002; Cowgill, 2007; Kirby et al., 2007; Zhang et al., 2007). Two end-member possibilities exist. The first considers that lateral erosion of the riser occurred continuously during fluvial occupation of the lower terrace surface, and, thus, displacement did not begin to accrue until abandonment of the lower terrace (e.g., Van der Woerd et al., 2002; Kirby et al., 2007). In this scenario, the age of abandonment of the lower terrace approximates the age of the offset terrace riser. In the second scenario, the terrace riser is protected by topography at the fault upstream, and therefore displacement can begin to accrue while the lower terrace tread is still occupied by the channel (Cowgill, 2007). In this situation, the age of abandonment of the upper terrace is more appropriate to estimate the average slip rate (e.g., Kirby et al., 2007; Harkins and Kirby, 2008). Along the northeastern segment of the Yabrai range-front fault, we used the ages of both the upper and lower terraces to derive a range of slip rates that bound the average rate along the fault. As we explain in the following, however, the geomorphic context of the site is similar to the second scenario described here, and we regard the slip rate from the age of the upper terrace as a more likely value.

Ages of the Alluvial-Fan Surfaces

Cosmogenic nuclides may be used to determine exposure ages of alluvial fans and terrace surfaces if the predepositional nuclide component can be accounted for and postdepositional surface modification can be evaluated (e.g., Gosse and Phillips, 2001; Hetzel, 2013). The predepositional component can be quantified by the depth profile method (e.g., Anderson et al., 1996; Wolkowinsky and Granger, 2004), or by analyzing samples from active channels (e.g., Brown et al., 1998; Schmidt et al., 2011; Hetzel, 2013). At the Yabrai range front, the Yabrai Shan is the only source of sediments supplied to the range-front alluvial deposits, and the headwaters of drainages are less than 3 km (generally <1 km) away from the site of deposition. This suggests that inheritance related to the source of materials and exposure during transport are likely to be small (e.g., Brown et al., 1998; Van der Woerd et al., 1998). In order to estimate the inherited component nuclide concentration, we collected 1–2 kg samples (pebbles) from active channels near the abandoned alluvial fans where we collected surface samples.

Postdepositional reworking or erosion of the surface are typical sources of uncertainties in surface exposure dating (e.g., Bierman, 1994; Matmon et al., 2005). In this study, we selected well-preserved alluvial surfaces that were flat and smooth with no incised rills, where there was no evidence for postdeposition reworking or erosion. The alluvial surfaces bore no vegetation and were devoid of a silt cover that would indicate eolian deposition. Clasts on the surfaces, ranging in size from 0.5 to 5 cm in diameter, were angular to subangular. We collected at least 1 kg vein quartz and/or granite clasts 2–3 cm in diameter from the most pristine part of each surface where negligible erosion occurred. Located in the low topographic area, topographic shielding is less than 4° in any direction. Therefore, we did not consider the shielding impact on nuclide production rates.

We dated a total of six quartz-rich granite samples from alluvial surfaces and three from the active channels near each site (Table 1). The quartz was separated and purified using the chemical isolation method of Kohl and Nishiizumi (1992). First, all samples were crushed and ground to 250∼500 μm, and HCl and HF/HNO3 leaches were used to dissolve carbonates, iron oxides, and potential surface contaminate 10Be produced in the atmosphere. Then, the cleaned quartz was completely dissolved in acid, and a 9Be carrier of ∼0.3 mg was added to the sample solution. Beryllium was purified and separated from the dissolved quartz by anion and cation exchange chromatography. Next, ammonium hydroxide was added to convert chlorine beryllium to hydroxides, and the beryllium hydroxide was oxidized by ignition at 750 °C in quartz crucibles. Two of the samples (Yu-14-Be-038, 039) were processed and tested at the Xi’an Accelerator Mass Spectrometry Center (XAAMS). The remaining samples were prepared at the Chinese State Key Laboratory of Earthquake Dynamics and measured on the accelerator mass spectrometer (AMS) at the French Centre National de la Recherche Scientifique (CNRS). To calculate exposure ages from the 10Be concentration, production rates of 10Be were scaled to the location of the sampling sites using the procedure of Lal (1991). All of the age calculations were performed using the online CRONUS calculator (Balco et al., 2008). The isotopic concentrations and calculated ages appear in Table 1.

Slip Rates along the Yabrai Range-Front Fault

Vertical Slip Rates along the Southwestern Yabrai Range-Front Fault

The first site was chosen because four alluvial fans truncated by a single, continuous fault scarp have been preserved (Figs. 18A and 18B). Exposure dating constrains the age of abandonment of each alluvial surface, which, combined with measured vertical offset surfaces, allowed us to calculate the vertical slip rates across the fault scarps. From field observation and satellite images, only the highest (T4) and lowest (T1) alluvial surfaces can be found on both sides of the fault scarp; profiles from the intermediate age surfaces, T2 and T3, appear to have been buried by deposition of T1 in the downthrown block of the fault (Fig. 13B). Therefore, the slip rate determined using T1 and T4 is likely the closest to the true vertical slip rate of the fault, and the intermediate surfaces represent minimum vertical slip rates because of aggradation on the downthrown block. To determine the abandonment ages of these alluvial surfaces, we collected four samples from the upthrown block of the fault (Fig. 13B), where the alluvial surfaces are well preserved and level with no incised rills and bar-and-swale topography, showing no evidence of erosion and aggradation. One sample (YBL-05-05) from the active channel nearby was collected to estimate inherited component nuclide concentration. The five samples from four alluvial surfaces yielded ages ranging from 24 to 135 ka (Table 1). Combined with the vertical separations derived from topographic profiles (Fig. 18C), we can obtain a minimum bound on the averaged vertical slip rate of 0.13 ± 0.03 mm/yr (Fig. 18D). Despite the potential of burial of the T2 and T3 surfaces in the downthrown block, all four estimates of the slip rate yield estimates within uncertainty of each other (Fig. 18D). It appears that these rates are likely close to the true slip rate, suggesting that aggradation must have been minimal during deposition of T1.

At the second site, 4 km to the northeast, alluvial fans on the piedmont of the mountain are faulted (Figs. 4, 19A, and 19B) across a sharp linear trace on the alluvial-fan surface both in the satellite imagery and in the field (Figs. 19A and 19C). Based on color, degree of incision, and varnish accumulation on clasts, the surfaces on both sides of the fault appear to be of the same age, with no degradation observed on the downthrown block of the fault. One cosmogenic sample (YBL-06-1) was collected from the most pristine part of the upthrown block of the fault to determine the abandoned age (Fig. 19B; Table 1). A sample (YBL-06-2) from an active channel was used to estimate the inherited concentration in this catchment (Fig. 19A). These results suggest that the alluvial surface was abandoned at 115 ± 11 ka. Topographic profiles of the alluvial-fan surfaces suggest vertical separation of ∼7–10 m (Fig. 19D) and an average slip rate of 0.08 ± 0.04 mm/yr at this site. The average vertical slip rate of the southwestern segment of the Yabrai range-front fault derived from these two site is ∼0.11 ± 0.05 mm/yr.

Sinistral Strike-Slip Rates along the Northeastern Yabrai Range-Front Fault

As mentioned already, site 1 displays evidence for sinistral strike slip along the northeastern segment of the Yabrai range-front fault. The riser crest that marks the T3/T1 boundary is displaced by 40 ± 5 m (Fig. 9). Lateral offset of gullies on both T1 and T2 attest to sinistral displacement (Figs. 9C and 9D) but are of only limited utility to estimate the slip rate because there is no direct estimate of when the gullies began to accumulate slip. We collected cosmogenic samples (Yu-14-Be-038, Yu-14-Be-039) from two alluvial surfaces (T1, T3; Fig. 9B; Table 1). We collected ∼2 kg of quartz pebbles from each of the surfaces (Fig. 9H) where there was no evidence of erosion or surface degradation. A sample (YBL-09-5) from a nearby active channel was used to calculate the inherited concentration of nuclides from this catchment. Exposure dating yielded abandoned age of 53 ± 8 ka for T1, and 175 ± 19 ka for T3. Combining these ages with the displacement of the riser (40 ± 5 m) yields maximum and minimum bounds on the slip rate of 0.78 ± 0.12 and 0.23 ± 0.02 mm/yr, respectively. The position of the riser relative to the sinistral strike-slip fault suggests the possibility that it was protected during progressive fault slip, and thus we regard the lower slip rate (0.23 ± 0.02 mm/yr) from the age of the upper terrace (T3) as a more likely value. Notably, this rate is similar to, albeit a bit higher than, the extension inferred across the southern range-front fault.

Kinematics and Geometry of the Yabrai Range-Front Fault

The southern and middle segments of Yabrai range-front fault are clearly a normal fault system, and the basin stratigraphy suggests a clear association with extension across the Gobi Alashan block during late Mesozoic and early Cenozoic time (Wu et al., 2007; Zhong et al., 2010). Our results refine the history of slip along this basin-bounding fault during the latest Cenozoic. We observed differences in the kinematics of slip along the three various segments of the fault, but these appear to be compatible with an overall regime that involves NW-SE–directed extension and NE-SW–directed shortening (Fig. 20).

The differences in strike direction for each of the segments of the Yabrai range-front fault are consistent with the different kinematics inferred from displacement of the Pleistocene alluvial fans. The southwestern and middle segments of the Yabrai range-front fault, which strike N40°–60°E, show a simple geometry with only one fault trace curving along the base of the range. The prominent footwall topography of the range itself, fault scarps on alluvial fans, fault exposures in trenches, and bedrock facets along the range front imply normal faulting along these two segments. In contrast, the northeastern segment strikes N75°–85°E and exhibits largely strike slip, with a reverse component. Moreover, the northeastern ∼20 km of the fault trace consists of two primary fault strands, arranged in a left-stepping en echelon pattern (Fig. 8). The northern of these, at the foot of the range, appears to be primarily reverse slip, whereas the southern strand appears to be largely strike slip. Thus, it seems that displacement along the northeastern segment of the Yabrai range-front fault is partitioned, at least in the shallowest portions of the crust. The lateral slip rate along the northeastern segment is relatively low (<1 mm/yr), suggesting the fault has experienced small magnitudes of strain relative to the structures in the inner Tibetan Plateau.

Interestingly, the topographic relief between the range and basin is similar along most of the Yabrai Shan, ranging from 400 to 500 m. The topography itself is asymmetric across the range, suggesting that the whole Yabrai range has been back tilted during normal faulting. However, the results of this study show that the northeastern segment of the Yabrai range-front fault presently accommodates sinistral strike slip with a reverse component. Thus, we infer that the northeastern segment of the Yabrai range-front fault must have been reactivated at some time in the Cenozoic. Unfortunately, there is no definitive constraint on the timing of initiation of transpressional motion. Ranges farther south, in the Hexi Corridor, have been suggested to have uplifted at ca. 2 Ma (Palumbo et al., 2009; Zheng et al., 2013b), and it may be that the present-day deformation field is a relatively young (i.e., Quaternary) stage of deformation within the Gobi Alashan block (e.g., Meyer et al., 1998; Tapponnier et al., 2001).

Possible Modes of Deformation in the Gobi Alashan Block

The Gobi Alashan block is adjacent to the Tibetan Plateau and the Ordos block (Fig. 1). This location suggests that there are three possible deformation scenarios that may explain the geometry and kinematics of the Yabrai range-front fault in the Gobi Alashan block. As discussed previously, the Altyn Tagh fault could extend into the Gobi Alashan block (e.g., Yue and Liou, 1999; Darby et al., 2005; Tapponnier et al., 2001; Webb and Johnson, 2006). Sinistral strike-slip faults in the southern Gobi Alashan block could accommodate the extrusion of crustal blocks beyond the Tibetan Plateau, and this model would be consistent with plate-like lateral extrusion (e.g., Avouac and Tapponnier, 1993). Second, as several faults were found to be normal faults in the Mesozoic from seismic reflections, the deformation could be caused, not by driving forces from the Tibetan Plateau, but as a result of rifting around the Ordos block associated with the development of the Hetao-Yinchuan rift system and the Shanxi graben system (Xu et al., 1993; Zhang et al., 1998). Third, the deformation in the southern Gobi Alashan block could be linked to the northeastward propagating extrusion of the Tibetan Plateau as a far-field dynamic effect of the India-Eurasia collision (Tapponnier and Molnar, 1979; Tapponnier et al., 1982; Kimura and Tamaki, 1986; Liu and Yang, 2000). Such a model has recently been proposed for the growth of the Tibetan Plateau proper (Clark, 2012), and it may be that the active deformation in the foreland region is a modern manifestation of this process. Of course, some combination of the aforementioned tectonic regimes may be present.

We argue against the first model for several reasons. First, there is no a linkage fault system between the northeastern tip of the Altyn Tagh fault near the north of Yumen (Fig. 3) and the Yabrai range-front fault. Active faults north of the Hexi Corridor, such as the Heli Shan fault (HLSF, Fig. 2), are largely reverse faults with no significant lateral-slip motion (Zheng et al., 2013b). Likewise, although the Beidashan fault (BDSF, Fig. 2) remains poorly studied, interpretation of satellite imagery reveals no significant lateral displacement. Second, long-term geological slip rates and present-day geodetic velocity suggest that deformation along the Altyn Tagh fault decreases from ∼10 mm/yr at ∼90°E to ∼1–2 mm/yr east of 97°E (e.g., Washburn et al., 2001; Cowgill, 2007; Zhang et al., 2007; Cowgill et al., 2009). This pattern is consistent with transfer of lateral displacement to thrust faults within the western Qilian Shan region (e.g., Yin and Harrison, 2000; Yin et al., 2002; Zheng et al., 2013a). Third, recent magnetotelluric profiles across the northeastern part of the Altyn Tagh fault and the Hexi Corridor image high-resistivity crust of the western Hexi Corridor that may be unfavorable for the Altyn Tagh fault passing through toward the northeast, and faults in the north Hexi Corridor seem to be reverse faults with shallow dips (Xiao et al., 2013, 2015). Finally, in the Yabrai Shan, as mentioned before, Darby et al. (2005) inferred that the northeastern segment of the Yabrai range-front fault and the Shazaogou Canyon represent the continuation of the Altyn Tagh fault based on satellite images and striae found in the western tip of Shazaogou Canyon. However, we did not find evidence that the Shazaogou Canyon fault has been active in late Quaternary time. Rather, the active deformation appears to be restricted to the portion of the fault that defines the topographic range front.

Our data also do not support the second tectonic model, that the deformation in the Gobi Alashan block primarily reflects an extensional regime similar to that around the Ordos block in the late Cenozoic (Kimura et al., 1990; Zorin et al., 1995; Yin and Nie, 1996; Zorin, 1999; Graham et al., 2001; Ren et al., 2002; Meng, 2003). Although it seems clear from the seismic reflection data (Wu et al., 2007; Zhong et al., 2010) that the Yabrai range-front fault was a normal fault in early Cenozoic time, our data suggest that the northeastern segment of the Yabrai range-front fault is presently active as a sinistral strike-slip fault with a reverse component. Although we cannot rule out the possibility that the normal faulting in the southwestern and middle segments of the Yabrai range-front fault were once linked to extension around the Ordos block, it seems unlikely that an oblique sinistral strike-slip fault would be compatible with a tectonic regime dominated by northwest-southeast extension.

Our results appear most compatible with the third tectonic regime, that deformation in the Gobi Alashan block is the result of oblique extrusion of the northeastern Tibetan Plateau. An east-west–striking sinistral fault, named the Ayouqi fault, is present along the south edge of the Gobi Alashan block (Figs. 2 and 3). The fault is expressed as a straight linear trace on high-resolution satellite imagery, and numerous small gullies along the fault trace are displaced in a sinistral sense. We suggest that the Yabrai range-front fault and the Ayouqi fault formed a releasing stepover to absorb the deformation in the southern Gobi Alashan block (Fig. 20). The Ayouqi fault and the northeastern segment of the Yabrai range-front fault represent strike-slip faults bounding a left step in a sinistral system (Fig. 20). The southwestern and central segments of the Yabrai range-front fault thus represent dominantly extensional faults accommodating extension and subsidence of the hanging wall within the stepover zone (Fig. 20). As we mentioned before, the topographic relief between the Yabrai Shan and the basin is similar along the whole section of the Yabrai range-front fault. This is incompatible with different senses of motion occurring at each segment of the Yabrai range-front fault consistently. Thus, we infer that this kind of topographic relief might be the result of back tilting during normal faulting previously, and the releasing stepover structure represents reactivation or superposition of old structures by the northeastward growth of the Tibetan Plateau in the late Cenozoic (Fig. 20).

Our air-photo interpretations and field geomorphology investigations suggest that the Yabrai range-front fault has been active during the late Quaternary within the Gobi Alashan block. The observed geomorphology of the fault system, the geometry of the faults exposed in shallow trenches, and displaced alluvial surfaces indicate that the southwestern segment acts as a normal fault with minor sinistral strike slip, the middle segment acts like a dip-slip normal fault, and the northeastern segment as a sinistral strike-slip fault with a reverse component of motion. Cosmogenic exposure data in combination with measured heights of fault scarps that truncated abandoned fans imply an average vertical slip rate of ∼0.11 mm/yr for the southwestern segment, and strike-slip rates between 0.23 ± 0.02 and 0.78 ± 0.12 mm/yr for the northeastern segment derived from displaced alluvium risers and their exposure ages. We infer that the different senses of motion at each of the segments of the Yabrai range-front fault could be driven from the northeastward oblique extrusion of the Tibetan Plateau.

This work was jointly supported by the Public Service Funds for Earthquake Studies (201408023, 201308012), the Fundamental Research Funds in the Institute of Geology, China Earthquake Administration (IGCEA1220), and the National Science Foundation of China (41590861, 41372220, 41172194). We thank Dewen Zheng, Huiping Zhang, and Steven Wesnousky for comments that improved a previous version of the manuscript, and Tim Middleton and Lalo Eduardo Guerrero for their help on the language of the manuscript. Rongzhang Zheng and Li Zhang are thanked for their help in processing cosmogenic dating samples. We also acknowledge Daoyang Yuan, an anonymous reviewer, and the Science Editor Arlo B. Weil, whose comments improved this manuscript. The supporting information includes a seismic reflection profile modified from a published Chinese paper and high-pixel photography of the trenches showing more details. For other data requests, please contact the corresponding author in China (

1GSA Data Repository Item 2016058, Figures S1–S5, is available at, or on request from