Investigation on the kinematics and deformation rates about active fault interior of the Tian Shan can provide significant information for strengthening our understanding on the present tectonic evolution of this range. The Baoertu Fault (BETF) is a major E-W striking active structure within the eastern Tian Shan and separates the south and central Tian Shan. But its kinematics and slip rates in the late Quaternary have never been systematically reported before. Based on interpretations of remote sensing images, drone photography, and detailed field investigations, we propose that the BETF is characterized by left-lateral strike-slip faulting with a thrust component and provides the first late Pleistocene slip rate for this fault. At the northern margin of the Kumishi Basin, combining offset reconstructions of displaced alluvial fan surfaces with the terrestrial cosmogenic nuclide (TCN) exposure age dating, we calculate an average sinistral slip rate of mm/yr and average vertical slip rate of mm/yr for the BETF since 95-106 ka. The differential movement eastward between the central Tian Shan block and Yanqi-Kumishi Basin block is likely the dominant driver of the left-lateral slip of the BETF. Synthesizing other quantitative data in eastern Tian Shan, we suggest that the hinterland active faults or folds, including the BEFT, roughly accommodate ~28-45% of the total N-S convergence across the eastern Tian Shan.
The Tian Shan is one of the most intensely deformed and seismically active intracontinental mountain belts in the world, which extends for nearly 2500 km from Uzbekistan in the west to NW China in the east (Figure 1(b)). This intracontinental range was reactivated due to the ongoing collision between the India and Eurasia plates during the Cenozoic [1–4]. Although the Tian Shan has experienced a complex geological evolution since the Paleozoic , the more recent uplift of this range initiated in the late Oligocene-Miocene . The overall deformation regime of the Tian Shan range is N-S crustal shortening. Global Positioning System (GPS) measurements indicate that the western Tian Shan has accommodated ~20 mm/yr north-south crustal shortening [7–9], nearly half amount of the northward indentation of the India Plate. Previous studies have shown that the active shortening in the Tian Shan is accommodated not only by foreland thin-skinned folding and thrust belts [4, 10, 11] but also by deformation in the interior of the range [12–14], which is mainly distributed along the thrust structures with late Quaternary slip rates of 0.1-3 mm/yr in the western Tian Shan .
GPS measurements (e.g., [8, 9]) and surface ruptures of several large historical earthquakes indicate that the interior of the Tian Shan also involves strike-slip faulting [15, 16]. These large intermontane strike-slip faults, which were identified active due to obvious landform offset mapped from satellite images, play an important role in accommodating the deformation of the Tian Shan . Large continental right-lateral strike-slip faults, usually trending NW to WNW with a slip rate of >2 mm/yr, such as the Talas-Fergana Fault and Dzhungarian Fault, are thought to accommodate crustal shortening by anticlockwise block rotation in the western Tian Shan ([17–19]. Figure 1(b)). Left-lateral strike-slip faults are also identified within the Tian Shan [9, 20]. At present, few studies provide detailed geometric and quantitative kinematic descriptions for these left-lateral strike-slip faults in contrast with thrust and right-lateral slip faulting, especially in the eastern Tian Shan, which hampers our understanding of the tectonic deformation of this range.
This paper is focused on the approximately E-W striking strike-slip Baoertu Fault (BETF) interior of the eastern Tian Shan (Figure 1(b)). This fault extends ~300 km from the southern margin of the Yultuz Basin in the west to northeastern tip of the Kumishi Basin in the east. The BETF has experienced complex geologic process since the Permian [21, 22] and delineates a boundary between the central and the southern Tian Shan , the central Tian Shan marks the region between the north and south Tian Shan. The fault vergence varies with dip to the south in the west part and dip to the north in the east part [24, 25]. The east part of this fault dominates the northern margin of the Kumishi Basin and produces significant topographic relief between the mountains and basin, indicating active faulting and uplift. Some researchers proposed that this fault is characterize by dextral faulting based on the interpretations of satellite images [2, 26]. However, offset landforms, including inflected rivers, ranges, and beheaded channels (Figures 2 and 3), along the fault all indicate sinistral strike-slip movement  despite the fact that this fault is subparallel to several WNW trending dextral strike-slip faults in the eastern Tian Shan. Along the west part of the BETF (west of Benbutu village), a series of rivers have been left-laterally displaced with offsets of ~1-4 km (Figure 2(b)). These drainages are thought to have formed since late Middle Pleistocene, which indicates the BETF has been active since that time . Up to now, the sense of movement and estimates of geological slip rate of this strike-slip fault have not been reported systematically.
In this paper, we first describe the geometry and characteristics of displaced geomorphic features along the BEFT based on high-resolution satellite images and detailed field-based reconnaissance along the north margin of the Kumishi Basin. Then, we constrain the late Quaternary slip rate for this fault using in situ cosmogenic 10Be exposure ages and measurements of displaced alluvial fans. Finally, we discuss the regional kinematics pattern and its role in accommodating deformation across the Tian Shan, as well as the seismic hazards along this fault within the eastern Tian Shan.
2. Tectonic Setting and Seismology in the Eastern Tian Shan
The eastern Tian Shan range forms an elongated region of deformation between the stable Kazakh Platform to the north and the rigid Tarim Basin to the south (Figure 1(b)). Crustal shortening rate of 2-5 mm/yr is estimated across the eastern Tian Shan by geodetic measurements [8, 9, 28], much less than the rate in the west. Shortening difference may imply different deformation regimes along the trend, which could be related to the clockwise rotation of the Tarim Block . The eastern Tian Shan is composed of east-west trending Paleozoic ranges and separating basins with Cenozoic sediments  (Figure 1(c)). Most of the intermontane and foreland basins are bounded by distributed reverse faults, and nearly every intermontane basin covered by Neogene and Quaternary syntectonic strata is deformed. The Kumishi Basin and Yanqi Basin both are intermontane basins in the eastern Tian Shan (Figure 1(c)). Our study area is located along the north margin of the Kumishi Basin, a NW trending elongated basin with a length of 230 km and width of ~12-40 km. This basin is controlled by the BETF on the north margin and the Kumishi Fault on the south margin. The Kumishi Fault (KMSF) is a thrust fault and displays linear fault scarps on piedmont late Quaternary alluvial fans . The Yanqi Basin is located at the south of the Kumishi Basin; the Hejing fold and thrust belt and Yanqi north fault are active structures located along north margin of the Yanqi Basin . The Kaiduhe Fault (KDHF) is a large right-lateral strike-slip fault, which cuts through the Yanqi Basin and shows clear linear fault scarps . Late Quaternary right-lateral slip rate of 1.2-1.6 mm/yr is estimated based on active channel offsets and the 10Be exposure dating of corresponding geomorphic surfaces . The Bolokenu-Aqikekuduke Fault (BAF) is another right-lateral strike-slip fault in the eastern Tian Shan, which originates from Kazakhstan to south of Turpan Basin and divides the northern and central domains of the Tian Shan. The slip rates of the BAF vary along its strike, with ~5 mm/yr slip rate on its northwest strand and 1.0-1.4 mm/yr on its southeastern strand .
Several large historical earthquakes that occurred in the eastern Tian Shan have been recorded , including the 1842 Ms 7.5 and the 1914 Ms 7.5 earthquake near the Barkol County. These two earthquakes both produced ~20 km long surface ruptures distributed across different fault segments  and are thought to be related to the reactivation of the Jianquzi-Barkol Fault . However, there are no instrumental or historical records of large earthquakes along the BETF, despite the clear surface break on the bedrock (Figure 3) and geomorphic expression of this fault along the north margin of the Kumishi Basin. In 1995, the Ms 5.0 Heshuo earthquake (Figure 1(c)), which caused moderate damage to Heshuo and Hejing County, is located along the BETF and is reported as a sinistral-thrust type event . The occurrence of this event and clear surface rupture indicates that the BETF has been active in the late Quaternary.
3.1. Geomorphic Surface Mapping and Offset Measurements
Due to the semiarid climate in the eastern Tian Shan, a sequence of late Quaternary alluvial fans are well preserved on the piedmont along the northwestern margin of the Kumishi Basin. To better characterize and interpret the different alluvial fan surfaces and the fault traces in our study area, Landsat satellite images, Google Earth images, and high-resolution digital elevation models (DEMs) with a resolution of <0.2 m/pix constructed for specific sites were used. Several field visits were conducted to verify our interpretations of mapping and to choose suitable sites for further study.
Structure from motion (SfM) is now a widely used photogrammetric method and has valuable applications to active tectonics community [35, 36]. This technique combines sufficiently overlapping optical images of a survey target from various view angles and distances . High-resolution DEMs are created by photo-based 3D reconstructions based on the structure-from-motion algorithm and can be used to illustrate more subtle morphology and make more accurate measurements. Photos were acquired with a digital camera carried on a DJI Phantom 4 Pro UAV. The DEMs were created using Agisoft Photoscan software. Several ground controlling points (GCP) were evenly placed throughout the survey area and measured using differential GPS (dGPS) to scale and position the 3D models.
The late Quaternary alluvial surfaces are mapped and divided into different geomorphic units mainly based on variations in surficial characteristics, including degrees of surface incision, relative height above the modern riverbed, and deposited material. Alluvial fan surfaces are also distinguishable on Google Earth images because of tonal difference which may be the result of various surface development over time, and the older surfaces appear lighter in color than the younger surfaces on the Google Earth images.
To obtain vertical offsets of deformed alluvial fans along the BETF, topographic profiles across the fault trace were extracted from the DEMs. We first fit trend lines to both the hanging wall and footwall surfaces by least squares linear regressions. Then, we used the mean value of vertical distance between the hanging wall and the footwall to determine the absolute height of scarps and the standard error. The optimal left lateral offsets of displaced geomorphic features are obtained using the LaDiCaoz_V2.1 Matlab software developed by Zielke et al.  through back slipping the hillshade plots. First, we use hillshade plots in the LaDiCaoz_V2.1 platform as well as field observations to identify the fault trace and seek for potentially displaced markers (e.g., the thalwegs and edges of channels). Next, the ideal fault plane is marked and the channel portions of upfault and downfault are considered to represent the original morphology of channels and then extract topographic profiles parallel to the fault trace several meters upstream and downstream from the fault (Figure S2). Finally, taking the trace of those markers and their distance to the fault trace into account, we project those sections onto the fault plane and back slip the topographic hillshade plots to determine the offset . The lateral offsets of stream channels crossing the fault can be varied because of their prefault drainage geometry. Uncertainties of such offsets thus primarily come from the identification of the piercing lines for the displaced markers. To minimize the epistemic uncertainty of the offset of a stream channel, we principally select the channels normal to the fault and inspect the plane geometry carefully to assess likelihood of channel deflection, then define its general orientation to choose the optimal piercing lines (usually thalwegs).
3.2. Sampling Strategy and 10Be Exposure Dating
In situ-produced cosmogenic exposure dating has become an important way to determine the timing of alluvial fan abandonment in an arid environment where organic material or fine-grained sediment is unavailable for radiocarbon or Optically Stimulated Luminescence (OSL) dating. Sourced from granitic bedrock (Figure S1), sediments exposed on these fan surfaces along north margin of Kumishi basin are mainly comprised of quartz-rich angular to subangular clasts or gravels and therefore, it is suited to use an in situ-produced cosmogenic 10Be exposure dating method to constrain the abandonment timing of alluvial fan surfaces, which is extensively used in central Asia (e.g., [39–41]).
The materials on fan surfaces may have been exposed to cosmic rays during the exhumation and subsequent transport prior to deposition, which will lead to overestimation of the abandonment age if the inherited component is neglected . There are two methods that are usually applied to estimate the inheritance. First, quartz samples were taken from the bed of active streams, whose 10Be concentration is assumed to derive entirely from predepositional exposure. These measurements can be used to correct the concentrations of individual samples taken from surfaces of alluvial fans . Second, provided there is a uniform inherited 10Be concentration over time, concentrations of samples taken beneath the surface should be decreased exponentially with depth. And then, the average inherited concentration can be determined from the concentration-depth profile. We use the second method to correct inheritance in this paper.
We have collected three samples on the deformed fan surfaces at BET02 and BET03 sites and six different samples within one depth profile at the BET01 site. Each surface sample contains ~100 round cobbles or gravels with a diameter of 2-4 cm (e.g., Figure 4(b)). The T1 terrace is on the east flank of Fan1 at the BET01 site, and this terrace is also deformed by the BETF. Quartz samples were collected at this terrace surface to confine the upper limit on time of the latest activity of the BETF at the BET01 site. Sample preparation and chemical extraction of 10Be from quartz were carried out at the Cosmogenic Nuclide Chronology Laboratory in the Institute of Crustal Dynamics, China Earthquake Administration (CEA). Detailed chemical processing is provided in the Supplement (Text 1). The beryllium target making and analysis were processed at the French National Research Center (CNRS). Detailed 10Be analytical results are listed in Table 1. The exposure age of the 10Be depth profile was calculated using the Hidy et al.  Matlab script. The other single surface samples presented in Table 1 were calculated using the CRONUS-Earth version 2.3 online calculator  using the time-dependent medal of Lal  and Stone . We assumed a constant sea-level and high-latitude 10Be production rate of 4.3 atoms/g/a . The density of alluvial fan gravels is limited to 2.0 g/cm3. Erosion is usually a large source of uncertainties in exposure dating . Despite the fact that all the surfaces are incised by several channels, the surface tops we targeted are flat, and most of the boulders or clasts are coated with desert varnish and remain in the condition in which they deposited with no apparent rolling. These observations show minor surface erosion of the surface. Hence, erosion was not considered in the age calculation as other researchers have done in central Asia that also use 10Be or 26Al cosmogenic nuclides (e.g., [39, 50]). Due to the arid climate in our study area, snow cover is also neglected.
T1 is next to Fan1 at the BET01 site; we assume that the concentration of inheritance of the T1 terrace is closed to the modeled inheritance from the 10Be depth profile on the Fan1 surface. The inherited 10Be concentration extracted from the depth profile is estimated about ~20% of the concentration of the first sample (-30 cm) under the Fan1 surface at the BET01 site (see Section 4.2), which suggests that the inheritance accounts for a large proportion and therefore cannot be neglected. Unfortunately, we failed to get samples from active streams corresponding to alluvial fans at the BET02 and BET03 site. Because of similar climatic and geological conditions over a 20-kilometer-wide area along the north margin of the Kumishi Basin and these deposits may experience similar exhumation and transport history, it is very likely that the amount of inheritance contained in samples from fan surfaces at the BET02 and BET03 site is similar to that from the BET01 site. Therefore, the inherited 10Be concentration from the Fan1 depth profile is used to correct the surface samples at the BET02 and BET03 site to get a rough abandonment age. All the age results are shown in Table 1.
4.1. Mapping and Exposure Dating of Alluvial Fan Surfaces
Alluvial fans are commonly formed at the mountain fronts where a laden river flows from the mountain to a large flat basin area . The formation and evolution of alluvial fan surfaces are generally considered to have been governed by climate fluctuations or tectonics (e.g., rock uplift) . Tao et al.  propose that the formation of fluvial terraces across the Tibetan Plateau is controlled by cyclical climate changes. In the Yanqi Basin, Huang et al.  identifies eight Quaternary alluvial fan surfaces since the middle-late Quaternary period and suggests that the abandonment ages are well correlated with every transition between glacial and interglacial. Also, in the Kumishi Basin, Wang  proposes that the formation of alluvial fan surfaces distributed along south of the basin is correlated with climate change from cold to warm since 60 ka. Therefore, it is likely that the formation of alluvial fans distributed along the north margin of the Kumishi Basin is also mainly associated with climate change.
Based on the superposition, relative height above the present riverbed, degrees of erosion, and channel development, we mapped four distinct alluvial fan surfaces, which are named Fan1, Fan2, Fan3, and Fan4 from the oldest to the youngest. Fan1 is the oldest surface and is preserved only at sparse locations. They are ~15-20 m above the present riverbed and usually deeply dissected by channels. In addition, this surface is the most prominently deformed fan surface from which we are able to constrain the offset and estimate the slip rate. The Fan2 surface is no more than ~15 m above the present riverbed and exhibits lesser dissection and flat surface because only some small gullies have incised it. Fan4 is the youngest geomorphic surface with undissected and washed surface which is most extensive in the study area. Fan4 is considered to be Holocene because of active streams on its surface. Fan3 is an intermediate surface between Fan2 and Fan4 with rough surface. We have chosen three sites for further study after field observation. The locations of these fault segments are shown in Figure 4.
Site BET01 is located ~10 km east of Baoertu village (Figure 4). At this site, we mapped two distinct alluvial fans (Fan1 and Fan4) and a river terrace (T1) (Figure 5(b)). The average height of the Fan1 surface above the present riverbed is ~20 m. Fan1 is incised by several channels with a width of ~10-50 m and a depth of ~1-10 m, but the undissected surface remains planar without extra erosion and aggradation. This surface is predominately composed of granitic and metamorphic gravels and clasts, with scales ranging from several millimeters to ~10 cm, although there are some larger boulders up to ~80 cm. Fan4 is incised into Fan1, which indicates that its age should be younger than the age of Fan1. Fan4 may be an active alluvial fan because of fresh boulders and scattered gullies. Additionally, a river terrace (T1) is located along the east edge of Fan1. This terrace extends for ~650 m from north to south with an average width of ~40 m (Figure 5(b)).
To determine an accurate time of Fan1 abandonment at the BET01 site, we collected six mixed gravel samples in total from a 2 m deep hand-excavated pit on its hanging wall. The sampling site is shown in Figure 5(a). The 10Be concentration of the uppermost sample (KMS-01) was not acquired because the signal of beryllium is too low to be detected, which resulted from excess Al(OH)3, instead of Be(OH)2 which was precipitated during the process of chemical purification. So this sample is not used to fit the age calculation and we use only the last five samples (depth of ~30 cm, ~60 cm, ~90 cm, ~150 cm, and ~180 cm, respectively). The average 10Be concentrations exponentially decrease from atoms/g at the depth of 30 cm to atoms/g at a depth of 180 cm (Figure 6). The best fitting exposure age and model average inheritance concentration values we obtained in this profile are /-11.6 ka and atoms/g, respectively. We also collected 80~100 gravels (Figure 7(b)) from the T1 terrace surface for 10Be surface dating. The concentration value of this sample is atoms/g. We assume that most gravels or clasts deposited on the T1 surface have a similar inheritance of 10Be concentration as those on the Fan1 surface; therefore, we get a corrected 10Be concentration for sample BET01-1 of atoms/g, and we obtain an abandonment age of the T1 terrace of ka (Table 1).
The site BET02 is located~5 km east of the Baoertu village (Figure 4). Four alluvial fans (from Fan1 to Fan4) were identified at this site (Figure 7(b)). The distribution of Fan1 is limited due to incision by other younger alluvial fans. And the surface of Fan1 is flat although intensively incised by a series of channels. The surface of Fan2 is flat and is not intensively incised with only a few channels on it. Fan3 is distributed at southwest and northeast of the BET02 site and has a higher altitude than Fan4. To constrain the abandonment age of Fan1 and Fan2 at the BET02 site, we collected an amalgamated surface sample from the hanging wall of Fan1 (BET02-1) and another amalgamated surface sample at the footwall of Fan2 (BET02-7), respectively. The two samples from the Fan1 and Fan2 surfaces are mainly composed of 2-4 cm subangular granite and quartz gravels. The results have shown that 10Be concentrations of BET02-1 and BET02-7 are 2 and 1 atoms/g, respectively. The maximum exposure time of ka and ka to the two surfaces can be obtained directly. After correction for the inherited 10Be component inferred from the 10Be depth profile at the BET01 site, we obtained exposure ages of these two samples of ka (BET02-7) and ka (BET02-1), respectively (Table 1).
The BET03 site is located ~6 km west of the Baoertu village (Figure 4). At the BET03 site, we also mapped four different geomorphic units—from Fan1 to Fan4. Clasts on the surface of the Fan1 alluvial fan at the BET03 site display moderate desert varnish development. To estimate the exposure age of Fan1 at this site, we collected an amalgamated surface sample (BET03-1) on the hanging surface (Figure 2). This amalgamated sample contains ~80 subangular clasts or gravels of 2-4 cm. As shown in Table 1, the 10Be concentration of BET03-1 is atoms/g, and the maximum exposure time is ka. Likewise, after eliminating the inheritance concentration inferred from the 10Be depth profile, the corrected exposure age we yield is ka.
4.2. Fault Geomorphology and Late Quaternary Activity
At the BET01 site, the BETF forms a pattern of sinistral slip with a reverse component. The fault traverses the fan surface and forms an EW trending fault scarp (Figure 7(a)). The fault scarp faces south and extends for ~600 m, which is clearly shown on the SfM-derived DEM image (Figure 5(a)). We extracted two profiles perpendicular to the fault scarp from the DEM and measured the vertical displacements of m (P1a) and m (P1b) (Figure 5(c)). The height of the scarp in the east is a bit higher than that in the west, which may be a result of heterogeneous denudation on the hanging wall or aggradation on the footwall near the scarp. The west edge of this Fan1 is clearly left-laterally offset. Both upstream and downstream parts of the west edge trend N35°E and are nearly linear (Figures 5(a) and 5(b)). This offset is more precise and reliable than the offset of stream channels on the fan surface. Therefore, the west edge of Fan1 is used as the piercing line to realign the cumulative displacement of this surface across the fault. Restoration of m of offset is produced for the present geometry of this edge of Fan1 (Figure 5(c)). However, subsequent aggradation may cause erosion to the upstream riser, so we propose that this offset may be less than true offset. Several shallow channels incised into the alluvial fan are left laterally displaced from ~10 m to ~70 m (realignment can be seen in Figure S2) by the fault (Figures 7(c) and 7(d)). In the east of the Fan1 surface, the topographic profile on the T1 surface has shown that the vertical offset is m (P1c) (Figure 5(c)). The horizontal offset of the T1 terrace is highly likely eroded because the downstream terrace is displaced into the path of river (Figure 5(a)) . Additionally, we did not find any evidence of faulting on the Fan4 surface at the BET01 site, which indicates that the fault is not active since the formation of geomorphic surfaces in this site.
Similar to BET01, the BETF at the BET02 site also shows a sinistral-reverse fault pattern. The late Quaternary surface traces of the BETF at this site are shown in Figure 8(a). The fault traces cut through Fan1 and form a right-stepping en echelon pattern of south-facing fault scraps. The fault strand strikes near EW and extends approximately 560 m. We extracted a topographic profile from the high-resolution SfM model and obtain a maximum cumulative vertical displacement of m, and this scarp shows an apparently two-stepped pattern (Figure 8(c)). Fan1 is the oldest morphologic unit in this fault segment and is incised by several different-sized channels (Figure 8(a)). Seven stream channels crossing the fault are left-laterally deflected with lateral displacements ranging from 10.7 m to 26.5 m (Figure 8(d)).
Compared with the displacement on the Fan1 surface at the BET01 site, the vertical offset on the Fan1 at BET02 is larger, but the maximum horizontal displacement is less, which may be a result of changes in fault geometry. In the east of Fan1, the BETF cuts through the surfaces of Fan2 and forms a linear south-facing continuous fault scrap (Figure 8(f)). The fault scarp on the Fan2 surface strikes EW and extends ~700 m (Figure 8(a)). We extracted a topographic profile from the high-resolution SfM model and obtained a vertical displacement of m here (Figure 8(c)). We propose that this vertical displacement may represent the coseismic vertical offset caused by the most recent earthquake event. Figure 8(g) shows a small gully perpendicular to the scarp to the east of Fan2, which is left laterally deflected m by the BETF. Fan3 is distributed to the southwest and northeast of the BET02 site. Numerous small braided streams are developed on the Fan3 surface, and this surface is not deformed by the fault. Fan4 between Fan2 and remnant of Fan3 is the youngest morphologic surface. Similarly, we did not find any trace of active faulting here.
The traces of the EW-striking BETF at the BET03 site are shown in Figure 9(a). Signs of late Quaternary fault activity consist of left-lateral displacement of several stream channels and two EW-striking, south-facing fault scarps at this site. These two subparallel fault strands are ~140 m apart, and the scarps in the north and middle on the Fan1 surface extend ~350 m and~180 m, respectively (Figures 9(a) and 9(d)). A topographic profile extracted from the DEM data shown in Figure 9(c) shows a vertical displacement of m and m for the two strands. Therefore, the total vertical displacement of the fault across Fan1 is m. Fan2 is located at the east of the large stream channel, and the surface is planar with limited stream incision. The fault deformed Fan2, forming a ~250 m long fault scarp across the middle of the fan surface. This scarp aligns with the scarp in the middle of Fan1 in a straight line. However, there exists no sign of deformation in the north part of Fan2. A topographic profile extracted from the DEM data shown in Figure 9(c) indicates a vertical displacement of m, which is nearly one-third of the vertical displacement on the Fan1 surface. Two channels crossing the upper fault scarp are clearly sinistrally beheaded. Realignment of one channel indicates a displacement of m (c-6) (Figure 9(e)). The larger stream channel shows a significant left-lateral offset at the lower portion of the alluvial fan. The horizontal displacement along the upper strand is estimated to be m. Overall, we obtain a cumulative horizontal displacement of 5 m and a maximum cumulative vertical displacement of m at the BET03 site.
5.1. Late Quaternary Activity and Slip Rate of the BETF
5.1.1. Characteristics of Fault Activity
Our field investigations and interpretations from Google Earth and high-resolution SfM model images show that the BETF has displaced a series of geomorphic units along its strike. In the west part of this fault, several draining rivers are left-laterally displaced at the kilometer level (Figure 2), and a range is offset clearly with a left-lateral displacement of ~10 m (Figure 3(c)). At the northern margin of the Kumishi Basin, numerous left-laterally displaced channels on late Quaternary alluvial fans are observed at the BET01, BET02, and BET03 sites, and the disproportionate ratios (approximately 2-13) of horizontal to vertical displacement suggest that BEFT is a left-lateral strike-slip fault with a reverse component. The fact that the topographic direction of the fault changes from north down (Figure 2(c)) in the western part to south down in the Kumishi Basin obviates the possibility of this being a primarily reverse fault. The fault motion interpreted from the geomorphic and geologic evidence is also consistent with the focal mechanism of the Ms 5.0 Heshuo earthquake in 1995 that occurred in the BETF . The fresh surface ruptures we have observed along the bedrock (Figure 3) suggest that the BETF may be recently reactivated during an earthquake.
5.1.2. Estimates for Slip Rates
Offsets of stream channels and fan edges along with the abandonment ages of geomorphic surface can be used to estimate fault slip rates. At the BET01 site, combining with the 10Be age constraint of 95.5 (+8.8/-11.6) ka for the Fan1 surface, the left-lateral offset of m of the west edge of Fan1 indicates a left-lateral slip rate of mm/yr. And we measured a vertical offset of m resulting in a slip rate of mm/yr. Likewise, at the BET03 site, combing the total lateral offset of m and vertical offset of m with the corrected abandonment age of ka, we obtain the left-lateral slip rate of mm/yr and vertical slip rate of mm/yr since ka. Because there are only three sites where we can study in detail, the number of gullies used to restore offset is limited and we may misestimate displacement due to oblique slip. We assume that the offsets of channels or fan edge have accumulated after overall abandonment of the alluvial fan surface and the age of channel should be later than the age of fan abandonment. Therefore, the slip rates we estimate are minimum from this assumption. To sum up, the average left-lateral slip of the BETF is mm/yr and average vertical rate is mm/yr since 95-106 ka. Compared to other strike-slip fault within the eastern Tian Shan, the slip rate of the BETF is lower [29, 31] (Figure 10).
More strikingly, at the BET01 site, the T1 terrace displays minor deformation (Figure 5(a)) but with an unmatched 10Be age of 65 ka. As described above, the deposition and incision of fluvial terraces in the eastern Tian Shan are well correlated with climatic fluctuation and the ages can be comparable (e.g., [53, 56, 57]). Some studies indicate the age of youngest terrace at the eastern Tian Shan is not more than 10 ka . The sampling site for T1 (BET01-1) is close to the east side slope of Fan1, and the characteristics of surface materials are similar with those on the Fan1 surface. Therefore, we are inclined to the assumption that these clasts we sampled on the T1 surface are more likely partially originated from the east slope of Fan1, and thus, this 65 ka cannot represent the real age of the T1 terrace.
5.2. Kinematics of BETF and Implications to Regional Deformation
Our results show that the BETF is predominately dominated by left-lateral slip faulting. The BETF is E-W trending and intersects with the NNE-directed convergence direction at a high angle (nearly perpendicular), which implies that this fault is not optimally oriented to the lateral-left slip. In Kyrgyz Tian Shan, Zubovich et al.  associated the left-lateral strike-slip component parallel to the belt with oblique indentation of the Tarim Basin to the Kazakh Platform. But the slip rates (~2 mm/yr in the west and ~4 mm/yr in the east) are much higher than rate of the BETF, which seems that the left-lateral slip of the BETF is not directly associated with the indentation of the Tarim Basin.
In the north Tian Shan, Campbell et al.  suggest that the Dzhungarian Fault accommodates strain by counterclockwise rotation about a vertical axis; another example of a similar kinematical model is the Talas-Fergana Fault in the western Tian Shan (; Figure 1(b)). Wu  assumes that there exists lateral extrusion to the east and west through the large strike-slip faults. In sum, because of distributed range-cutting faults, we favor the hypothesis that the Tian Shan range can be divided into several rigid fault-bounded mountain range blocks or basin blocks by these large faults [13, 29], and deformation of the Tian Shan can be absorbed or accommodated by the relative motion of several small blocks. In the eastern Tian Shan, the WNW-ESE trending BAF and KDHF are both large inherited dextral slip structures with steep dip angles, which control the boundary of the Turpan Basin and Yanqi Basin, respectively (Figure 11). These strike-slip faults including the reactivated structure of the BETF separate the eastern range into several fault-bounded blocks (Figure 11). From the north to the south, they are the central Tian Shan block and rigid Yanqi-Kumishi block, respectively.
In the eastern Tian Shan, Cunningham et al.  suggest that the upper crust of the eastern Tian Shan along with southwest Mongolia is being displaced eastward. GPS data show that GPS points nearly all slip toward NNE with respect to stable Eurasia and the velocities decrease from south to north (Figure 1(c)). But we can still qualitatively propose that the eastward component in the south of range is slightly larger than that in the north although there are significant uncertainties about the slip rates due to sparse distribution of GPS stations [8, 28], which indicates that there is discrepancy in E-W slide between the north and south Tian Shan. Based on the above analysis, the central Tian Shan block and Yanqi-Kumishi block move eastward collectively and the Yanqi-Kumishi block slides a little faster than the central Tian Shan block. Our result of ~0.65 mm/yr sinistral slip rate for the BETF is less than the rates of the KDHF and the BAF, and this low slip rate may indicate that the BETF is passive and can be regarded as a structure to accommodate the difference of movement between the central Tian Shan block and Yanqi-Kumishi block in the eastern Tian Shan. Although the exact time of the Cenozoic reactivation of these three strike-slip faults is unclear, we can create a simple model (Figure 12) based on the geometry and kinematics of these active faults, which shows that the difference in movement toward the east between the central Tian Shan block and Yanqi-Kumishi block is likely the dominant driver of the left-lateral strike slip on the BETF.
Analog models for continental compression in central Asia [61, 62] suggest that large-scale conjugate strike-slip faulting is significant in accommodating India-Eurasian convergence by lateral extrusion. The eastern Tian Shan consists of several sinistral slip faults, the BETF and Balikun-Jianquanzi Fault  north of Barkol Tagh. The existence of sinistral slip along with the dextral slip faults in the eastern Tian Shan interacts together to respond to the lateral extrusion to the east. Our study indicates that the present-day deformation in the Tian Shan related to the ongoing India-Eurasia collision has been accommodated by strike-slip faulting and block rotation besides relative motion of blocks.
5.3. Slip Distribution across Interior of the Eastern Tian Shan
Strain partitioning is crucial for understanding the tectonic deformation and geodynamics of convergent orogens. As presented in Figure 10, there exist several active faults within the eastern Tian Shan at the longitude between longitude 86° and 89°E. The north foreland and south foreland have accommodated ~2-4 mm/yr and ~4-5 mm/yr crustal shortening rate, respectively . The east strand of the BAF strikes WNW with a right-lateral slip rate of 1-1.4 mm/yr [31, 64], which will contribute to ~0.31 mm/yr N-S shortening rate if assuming a crossing angle of 75° between the fault strike and shortening direction. In the north of the Kumishi Basin, the BETF strikes nearly normal to the shortening direction and its low vertical rate contributes little to crustal shortening. The KMSF at the south of the Kumishi Basin has accommodated 0.31 mm/yr crustal shortening . In the Yanqi Basin, crustal shortening is accommodated by HJNF on the north margin of the basin with a shortening rate of 0.4 mm/yr  and the thrust-fold belts with a shortening rate of ~0.3 mm/yr . On the south margin of the Yanqi Basin, the KDHF strikes 300° with an average late Quaternary right-lateral slip rate of ~1.4 mm/yr , which gives a N-S shortening component of ~0.35 mm/yr. Farther east, although lack of precise deformation rates to the Turfan Basin and the easternmost Tian Shan, low relief and narrow range compared with the western Tian Shan indicate a small amount of crustal shortening. Considering that the Kumishi Fault and the thrust-fold zone at the north of the Yanqi Basin are located along strike, a total shortening rate of ~1.4 mm/yr is accommodated by the hinterland active structures (Figure 10). GPS measurements reveal that the crustal shortening rate across the eastern Tian Shan is ~3-5 mm/yr [9, 66]. Therefore, active structures within the range roughly accommodate ~28-45% of the total convergence across the segment of the eastern Tian Shan (86°-89°E).
The cumulative compressive slip within the western Tian Shan is ~11 mm/yr (Figure 10), mainly along thrust faults with late Quaternary slip rates ranging from ~0.1 to 3 mm/yr . In the central Tian Shan, Charreau et al.  quantified the crustal shortening rate of mm/yr accommodated by the Bayanbulak and Nalati intermontane basins. The KSHF is a major E-W active hinterland fault at the north Yili Basin, and the late Quaternary vertical slip rate of this fault is ~3.8 mm/yr [34, 68] with a dip of 56°-62°, which contributes to an ~2.2 mm/yr N-S shortening rate. In the north of the Zhaosu Basin, the KBHF has a late Quaternary slip rate of 4.2 mm/yr and dip ~70° ; thus, this fault can accommodate ~1 mm/yr shortening rate. The summation of the shortening rate is estimated to be >4.6 mm/yr across the internal central Tian Shan. From the synthesis above, we can conclude that the active structures within the entire Tian Shan play a major role in accommodating crustal shortening. Both geodetic and geological evidences have shown that crustal shortening rates decrease from west to east, and this shortening difference may be a result of different compressive forces between the west and east Tian Shan. Zhang et al.  speculated that the northward push of the Pamir block is the dominant force causing the deformation of the Tian Shan. However, Avouac et al.  and Molnar  emphasized that the clockwise rotation of the Tarim Block is the main cause of the deformation pattern and N-S crustal shortening. Niu et al.  propose that the northward push of Pamir and clockwise rotation of Tarim cocause the clear difference of deformation pattern. In the western Tian Shan (west of 75°E), the tectonic deformation is due to the north-northwest thrust of the Pamir Plateau. In the region east of 75°E, the cause of the difference in tectonic deformation is mainly due to the clockwise rotation of the Tarim Block.
5.4. Seismic Hazards along Faults with Low Slip Rate within the Eastern Tian Shan
The Tian Shan is one of the most seismically active intracontinental mountain belts around the world, and the majority of earthquakes are confined along the foreland piedmont regions where most of the crustal shortening is concentrated on. Historical earthquakes (such as Ms 8.0 Nilka earthquake and Ms 8.0 Tex earthquake) and obvious geomorphic evidence (fault scarps or surface ruptures) indicate strong activity on faults interior of this range.
Recent studies show that long-term slip rates on some faults within the Tian Shan are relatively low compared with those in interpolate areas and the recurrence intervals of great earthquakes on these active faults can be long which is always several thousand years . Our interpretation about BETF is consistent with this result. Our results show that the late Quaternary slip rate of the BETF is 0.65 mm/yr. The 0.7 m high fault scarp on Fan2 at the BET02 site is a result of a single event (Figure 8(f), Section 3). If we assume, the inflected stream with a displacement of 2.1 m across this scarp on the Fan2 surface is also caused by one event and the displacement could represent the coseismic displacement and strain accumulation time is estimated to be ~3000 years. In addition, we found no clear deformation evidence on the younger Fan3 and Fan4 surfaces. The latest alluvial fan is considered to exposure at the onset of Holocene due to a general change to a warmer and wetter climate . Huang  and Wang  both reported a period of fan accumulation of 5 ka ago because of regional climate change at the Yanqi Basin and south of the Kumishi Basin, respectively. The abandonment of Fan4 may therefore date from 5000 years to 10000 years, indicating that there may be no slip that occurs since at least 5000 years. Our estimate of a recurrence time of over 3000 years for BETF seems not unreasonable. The T1 terrace at the BET01 site is likely to develop at the early Holocene but with a small vertical offset of 0.4 m, which also indicates the seismic period of the BETF is long. Within the eastern Tian Shan, active faults usually have low slip rates, but low strain rate does not imply an absence of seismic hazards. Clear fault scarps and surface ruptures (e.g., Figure 3) are strong evidences which indicate that these structures still have the potential to generate large and destructive earthquakes.
The Baoertu Fault is one of the major structures interior of the eastern Tian Shan which has been recognized as active in the late Quaternary. This fault cuts pre-Cenozoic bedrock at its western segment and separates the Tian Shan from the Kumishi Basin at its eastern segment forming a straight linear feature in the satellite image.
Observations of the geomorphology have shown that the Baoertu Fault is characterized by left-lateral strike slip with a reverse component during the late Quaternary. Based on the 10Be exposure age and realignment of displaced stream channels, we obtain a left-lateral slip rate of mm/yr and vertical slip rate of mm/yr since 95-106 ka
The differential motion between the central Tian Shan and the Yanqi-Kumishi Basin subblocks is the dominant driver for the left-lateral strike-slip motion of the Baoertu Fault, which demonstrates that the BETF is regarded as an accommodation structure within the eastern Tian Shan
Active structures within the eastern Tian Shan roughly accommodate ~28-45% of the total crustal shortening across the eastern Tian Shan
Active faults with low strain rate do not imply an absence of seismic hazards and still have the potential to generate large earthquakes interior of the eastern Tian Shan
The data used to support the findings of this study are included within the article.
Conflicts of Interest
We wish to confirm that we have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
We are grateful to Weibin Zhang, Peng Su, and Zongkai Hu for their advice on our preliminary results. We like to thank Yanwu Lv at the Institute of Crustal Dynamics, China Earthquake Administration, for his help in preparing the 10Be sample. And we thank Haibo Yang for his guidance in calculating the exposure age. This work was financially supported by funds from the Fundamental Research Funds in the Institute of Geology, China Earthquake Administration (IGCEA1607) and the National Natural Science Foundation of China (42072250, 41672208). We greatly thank Andrea Billi and Aaron Bufe for their constructive and detailed comments to improve the manuscript.