Late Quaternary slip rate of the Aksay segment and its rapidly decreasing gradient along the Altyn Tagh fault

Large active faults in the interior and boundary of a plateau can provide kinematic constraints for fault interactions and rupture dynamics, which is of great significance to understanding the strain partitioning and seismic hazard assessment (Meade, 2007; Molnar and Dayem, 2010; Royden et al., 2008; Tapponnier et al., 2001). Meanwhile, the fault slip rate pattern along strike is also helpful for understanding the strain accumulation pattern, fault slip history, and the strain adjustment and its influence on adjacent faults or structures (Resor et al., 2018; Zechar and Frankel, 2009). A decreasing pattern of slip rates has been widely observed along the typical strike-slip faults around the world, such as the Altyn Tagh (Gold et al., 2009; Zhang et al., 2007; Zheng et al., 2013), Kunlun (Kirby et al., 2007; Lin and Guo, 2008), and North Anatolian fault zones (Walters et al., 2014). For instance, a rapid decrease in the slip rate of the Haiyuan fault occurs at ∼106°E in the Madongshan-Liupanshan region, where the deformation is dominated by reverse faulting, folding, and the uplift of regional structures (Li et al., 2009; Zheng et al., 2013). Similarly, the uplift of the Anyemaqen Shan and the clockwise rotation of eastern Kunlun faults accommodated the slip rate along the Kunlun fault during structural thickening (Kirby and Harkins, 2013).

The Altyn Tagh fault, as a large-scale boundary fault between the Tarim block to the north and the Tibetan Plateau to the south, plays an important role in affecting, and even controlling, the evolution of the Tibetan Plateau during its NE growth (Molnar and Rajagopalan, 2012; Tapponnier et al., 2001; Yuan et al., 2013). Studies on temporal and spatial distributions of the slip rate along the Altyn Tagh fault are essential for understanding the tectonic evolution of the northern Tibetan Plateau. With a series of mountains oblique to the fault (Fig. 1), the eastern segment of the Altyn Tagh fault is a very important strain partitioning zone on the northeastern Tibetan Plateau. The strike-slip rate gradients along the fault and deformation pattern are critical to understanding the mechanics of intracontinental deformation. Several researchers have suggested that a decreasing trend in the slip rate can be roughly obtained in the east segment of the Altyn Tagh fault from these constraints (Zhang et al., 2007; Zheng et al., 2013). Although previous studies have shown that the slip rate begins to decrease from the Aksay segment (Xu et al., 2005; Zhang et al., 2007), it is still unclear whether the slip rate decreases with a uniform gradient.

Therefore, we focused on the ∼60-km-long Aksay segment in order to systematically constrain the late Quaternary slip rates on the eastern segment of the Altyn Tagh fault based on multilevel displaced river terraces at four sites. The initial decrease in the slip rate on the eastern Altyn Tagh fault and multilevel offset geomorphic markers could be used to constrain more reliable slip rates on the Aksay segment. Along this segment, the West of Aksay, Old Aksay, Jiaerwuzongcun, and Yandantu sites (Fig. 2), from west to east, were investigated both in the field and by applying a Monte Carlo analysis method. Combined with local geodetic results, we determined the kinematic model of the Aksay segment, and we discuss its role in the strain partitioning of the eastern Altyn Tagh fault here.

The Qilian Shan (Shan = Mountains), located on the northeastern margin of the Tibetan Plateau (Fig. 1), has experienced intensive deformation since the late Cenozoic related to the northeastward growth of the plateau (Meyer et al., 1998; Zhang et al., 2004; Duvall et al., 2013; Yuan et al., 2013; Allen et al., 2017). Due to the NE-SW regional stress induced by plate motion, the strain in the Qilian area is partitioned into dip-slip and strike-slip components (Allen et al., 2017). The strike-slip structures are mainly large-scale strike-slip faults, such as the Altyn Tagh fault and Haiyuan fault. In contrast, the dip-slip structures are dominated by thrusts in the Qilian Shan, causing series of NW-trending mountain ranges, e.g., the Qilian Shan, Daxue Shan, Danghe Nan Shan, and Qaidam Shan from NE to SW (Meyer et al., 1998; Xu et al., 2005; Allen et al., 2017), and late Cenozoic uplift parallel to the Altyn Tagh fault, observed in the Sanwei Shan and Nanjie Shan to the north (Yang et al., 2020). There is a close interaction between the thrusts in the Qilian Shan and the Altyn Tagh fault. Previous research proposed that the subparallel Cenozoic thrust branches southeastward from the Altyn Tagh fault absorbed the crustal shortening along these structures (Meyer et al., 1998; Van der Woerd et al., 2001). However, Allen et al. (2017) suggested that the Qilian thrust was not a secondary feature caused by the activity of Altyn Tagh fault, but a major part of the whole convergence of India-Eurasia. In any case, the Qilian Shan is an ideal region to study strain partitioning.

The Altyn Tagh fault, marking the northern boundary of the Tibetan Plateau for nearly ∼2000 km, is a large left-lateral strike-slip fault at a lithosphere scale in the Eurasian plate (Fig. 1; Molnar and Tapponnier, 1975; Tapponnier et al., 1982, 2001; Yin et al., 2002; Xu et al., 2005; Zhang et al., 2007). Studying its temporal and spatial distributions of slip rates is essential to reveal the pattern of northeastward growth of the Tibetan Plateau (Meyer et al., 1998; Cunningham et al., 2016).

The long-term average slip rates over millions of years are constrained by offset geological markers and the time of initiation. The total left-lateral displacement of the Altyn Tagh fault was thought to be up to 350–400 km (Ritts et al., 2004; Yin et al., 2002; Yue et al., 2001, 2004). Yin and Harrison (2000) estimated the displacement of 280 ± 30 km since the late Pleistocene (ca. 30 Ma) and calculated an average slip rate of 7–9 mm/yr. Yin et al. (2002) suggested that the slip rate was 9 ± 2 mm/yr since 49 Ma. However, Yue et al. (2001) determined the displacement of the eastern and central segments of the Altyn Tagh fault to be 375 ± 25 km and showed that the slip rate was 12–16 mm/yr since the late Oligocene, according to a field geological survey. They further obtained displacement of 165 km since 16.4 Ma and suggested that the slip rate of Altyn Tagh fault was at least 10 mm/yr since the late Miocene (Yue et al., 2004).

For a long time, the activity of the Altyn Tagh fault has been regarded as a key index to test deformation models of the Tibetan Plateau. The “eastward extrusion model” is supported by the very high slip rates on the main boundary faults of the rigid blocks (Avouac and Tapponnier, 1993; Peltzer and Tapponnier, 1988; Peltzer et al., 1989), and the “crust thickening model” is characterized by continuous deformation distributed within the blocks as well as on the boundary faults, i.e., a relatively low slip rate on the Altyn Tagh fault (England and Houseman, 1986; Houseman and England, 1993).

Study of the late Quaternary slip rate of the Altyn Tagh fault began in the 1970s (Molnar and Tapponnier, 1975; Tapponnier et al., 1982; Table 1). Molnar et al. (1987) and Peltzer et al. (1989) interpreted 100–400 m of displacement of the river system along the Altyn Tagh fault using SPOT satellite images and estimated a high Holocene slip rate of 20–30 mm/yr. Tapponnier et al. (2001) summarized the reported Quaternary slip rates of the whole Altyn Tagh fault and obtained the slip rate in the central segment of Altyn Tagh fault between 83°E and 94°E reaching 20–30 mm/yr. Mériaux et al. (2004, 2005) determined an average slip rate of ∼20.3 ± 1.1 mm/yr by dating the alluvial fans and river terraces of the Altyn Tagh fault. Combining the high-resolution SPOT satellite images with field investigation, Xu et al. (2005) used various dating methods, such as radiocarbon dating (¹⁴C), in situ cosmogenic dating (¹⁰Be, ²⁶Al), and thermoluminescence dating (TL), and found that the slip rates of the central and western segments were up to 17.5 ± 2 mm/yr. These high rates largely support the eastward extrusion model.

However, the results of geological mapping and geomorphic studies carried out by the “Altyn Tagh Active Fault” research group of the Chinese State Bureau of Seismology along the Altyn Tagh fault suggested that the minimum Quaternary slip rate of the Altyn Tagh fault was ∼5 mm/yr (Chinese State Bureau of Seismology, 1992). Xiang et al. (2000) classified the systemic displacement of the river system in the eastern Altyn Tagh fault and constrained the slip rate of the Altyn Tagh fault in this area at 4.7–6.7 mm/yr. Wang et al. (2003, 2004) obtained a Holocene slip rate on the Altyn Tagh fault of 11.4 ± 2.5 mm/yr by studying river terraces and alluvial fans. Cowgill (2007) and Zhang et al. (2007) reinterpreted the results of Xu et al. (2005) and Mériaux et al. (2005) and suggested that the slip rate on the main segment of the Altyn Tagh fault was 10 ± 2 mm/yr. Gold et al. (2009) and Cowgill et al. (2009) constrained the Quaternary slip rates at 8–17 mm/yr and 9–14 mm/yr at 86.7°E to 88.5°E, respectively. Mériaux et al. (2012) obtained a slip rate of 13.9 ± 1.1 mm/yr in the Pingdingshan area on the western segment of the Altyn Tagh fault using the in situ cosmogenic dating methods.

Relatively low slip rates obtained along the Altyn Tagh fault based on global positioning system (GPS) and interferometric synthetic aperture radar (InSAR) measurements have been reported during the past decades. For instance, Bendick et al. (2000) used GPS data to obtain a left-lateral slip rate of 9 ± 5 mm/yr at 89°E–91°E. A similar result was also reported by Wallace et al. (2004). The present-day average geodetic slip rate between 85°E and 90°E was estimated to be 9 ± 2 mm/yr using GPS data (Shen et al., 2001). Zhang et al. (2007) showed a slip rate of 11.9 ± 3.3 mm/yr at 89°E–91°E, 4 mm/yr at 94°E–96°E, and 3.9 ± 2.3 mm/yr at ∼96°E, respectively. However, at ∼86.2°E, He et al. (2013) got a current slip rate of 9.0 ± 4 mm/yr based on a GPS array. From west to east, Zheng et al. (2017) estimated a left-lateral rate of 8.1 ± 0.7 mm/yr at ∼86°E, 8.6 ± 1.5 mm/yr at ∼90.4°E, and 4.5 ± 0.8 mm/yr at ∼94.6°E, respectively. Jolivet et al. (2008) used European Remote Sensing (ERS) radar data and Envisat radar data to determine a slip rate of 8–10 mm/yr at 94°E. For the segment between 91.5°E and 95°E, the slip rate derived from InSAR was 6.4 mm/yr (Liu et al., 2018a).

Recently, the widely accepted slip rate of the Altyn Tagh fault is ∼10 mm/yr, which implies that the previous estimation of ∼20–30 mm/yr may be overestimated. The results of recent research with more viable analysis, consistent with GPS and InSAR measurements, as the important proofs of the “crust thickening model,” imply that the Altyn Tagh fault is not the main fault allowing the northeastward extrusion of the Tibetan Plateau. The large-scale strike-slip movement is mainly confined into the interior of the plateau, which is manifested by the crustal thickening of the Qilian Shan (Zhang et al., 2007).

The Aksay segment is located at the eastern part of the Altyn Tagh fault (Fig. 1). The surface elevation of the alluvial fans varies from 2700–2800 m to 1650 m in the range of ∼25 km (Fig. 2). The Altyn Tagh fault cuts through most of the fans (Mériaux et al., 2005), and many studies have focused on this area. Based on the exposure ages of terraces and the displacement of terrace scarps, Mériaux et al. (2005) obtained a slip rate of 17.8 ± 3.6 mm/yr along the Aksay segment. Based on the ¹⁴C age and displacements of the river system, Xu et al. (2005) determined that the slip rate was 16.4 ± 2 mm/yr near the Old Aksay town. By reanalyzing the research results of Mériaux et al. (2005) and Xu et al. (2005), Zhang et al. (2007) considered the abandoned age of the upper terrace as the accumulated age of fault displacement, and the result was a slip rate of 8–12 mm/yr. Chen et al. (2012, 2013) studied the slip rate near the Old Aksay town by using the method of luminescence dating and the scarp erosion model. Through the scarp evolution, the age closest to the actual displacement of the gully was selected, and the slip rate of the Aksay segment since 6 ka was estimated at 11 ± 2 mm/yr, which was basically consistent with the results of Zhang et al. (2007) and Cowgill (2007). How is the strain partitioned on the eastern segment of the Altyn Tagh fault? In this study, we focused on the Aksay segment to obtain the fault slip rates and the strain partitioning pattern of the Altyn Tagh fault.

The fault slip rate represents the average velocity of fault movement in a certain period of time and reflects the rate of strain accumulation on a fault zone, so it is widely used in tectonic reconstructions and seismic hazard assessments. At present, the Holocene fault slip rate is mainly determined by dividing the total displacement of the geomorphic indicators (such as terrace risers of rivers and gullies, etc.) by cumulative time (the formation age of the geomorphic surface). In recent years, with light detection and ranging (LiDAR), stereo pairs of remote-sensing images, and aerial photographs, photo-based three-dimensional (3-D) reconstruction techniques have been applied to active tectonics, and it is more and more convenient to acquire high-precision 3-D topographic data (Ren et al., 2018). In this study, in order to constrain the slip rate in the study area, high-resolution terrain data at specific sites were obtained using aerial photographs collected by an unmanned aerial vehicle (UAV), according to the high-resolution structure-from-motion models (Westoby et al., 2012). The offsets were measured from the digital elevation model (DEM) using the LaDiCaoz_v2 code (Zielke et al., 2010, 2015), which is a widely used tool in strike-slip fault offset measurements. The basis for offset measurement is fitting the topographic profiles across linear markers such as gullies, terrace risers, and mountain ridges, etc. However, due to the natural evolution of channels or terrace risers, both sides of the fault might not fit with each other. The solution in this code is to horizontally or vertically stretch or compress the topographic profiles to find the best-fitting results. Meanwhile, the minimum and maximum offsets are also provided regarding the uncertainties, as well as the goodness-of-fit (GOF) values. Furthermore, the corresponding back-slip maps are provided in the results to check the reliability of the offsets. All the collected samples from the four sites were dated by optically stimulated luminescence (OSL) and situ cosmogenic nuclide (¹⁰Be) methods at the State Key Laboratory of Earthquake Dynamics, Institute of Geology, China Earthquake Administration and the French National Centre for Scientific Research, respectively.

For slip rate calculation, the Monte Carlo analysis method proposed by Gold and Cowgill (2011) was applied in order to obtain a paleoslip history. In this study, we tried to obtain a reliable fault slip rate with a synthetic data set in this way (Liu et al., 2018b). Each pair of offsets and dating results (including the uncertainties) was used to construct the input envelope, the edges of which were bounded by the minimum and maximum limits of displacement and the corresponding accumulation time for a given marker. The offsets of the younger terraces should not exceed those of the older terraces, and the ages should follow a similar rule. Hence, there was no negative slope for the gradient of the fitting curve from the results of offsets and dating, i.e., the slip rate curve. Furthermore, the average slip rates were constrained only by the rest of the displacement and age data. By defining the method points of each slip history, we could perform a linear regression to yield an average slip rate with its associated uncertainty. The fault slip rate was calculated repeatedly for 1000 times until testing showed that the solution did not significantly change for larger numbers of repetitions (Gold and Cowgill, 2011). We calculated the median value and nonparametric 68% and 95% confidence envelopes, which included the uncertainties in both the displacement and age.

The West of Aksay site was the westernmost site in the study area. The terraces could be classified into three levels, T1, T3, and T4 from young to old (Fig. 3). According to the regional landforms, the T2 terrace was not developed at this site. There were three different levels of channels developed on T4, i.e., R1, R2, and R3, respectively (Fig. 3). As channels are inherently younger than the landform they incise, this common practice may cause underestimation of slip rates, as well as a mismatch in slip rates between channels that are offset by different magnitude but incise in the same surface (Shelef et al., 2019). To obtain the accumulated displacement closest to the formation age of the T4 terrace, we measured R3 with the largest displacement, which can also be evidenced by the deepest erosion downward on the terrace profile of R3. The offset was estimated to be ∼98 ± 4 m (Fig. 4). Consequently, at the West of Aksay site, we only got one offset value to be used in fault-slip-rate determination.

The Old Aksay site was located in the east of Old Aksay county, where a linear structure could be observed clearly from the satellite image (Fig. 2). According to previous studies, Mériaux et al. (2005) identified three groups of terraces based on satellite images. Combining the relationships between active faults, offset channels, and loess on banks, Chen et al. (2013) determined the slip rate of the Altyn Tagh fault within 6 k.y. Based on the summary, it was found that only the lower terrace was studied (Chen et al., 2013; Mériaux et al., 2005; Zhang et al., 2007). Through the field investigation, three different levels of old terraces of Changcaogou (Ta, Tb, and Tc) were found in the Old Aksay site, which provided an opportunity to determine the slip rate on a larger time scale (Fig. 5). These old terraces are relatively high, and they represent residual terraces. Unfortunately, the corresponding geomorphic markers on the other side of the fault have not been found for displacement restoration. Therefore, the displacements were obtained by matching the terrace risers upstream of Changcaogou (Fig. 5), which were closer to the accumulated offset since the formation of these geomorphic markers. The displacements of these old terraces are very large, and the corresponding errors will not exceed the distance between the two terraces. We took the distances between these old terraces as the measurement error and obtained the displacement sequences, which were ∼1360 ± 110 m, ∼1790 ± 130 m, and ∼2290 ± 220 m, respectively (Fig. 6).

The Jiaerwuzongcun site was located in the middle part of the Aksay segment. There are four levels of terraces and multilevel secondary terraces (Fig. 7). The terraces are mainly base terraces, with the bottom composed of Tertiary red bedrock, the middle layer composed of river gravel accumulation, and the upper part covered with thick loess. No detailed work has been carried out on the slip rate there. There are also multilevel terraces displaced by the Altyn Tagh fault at the Jiaerwuzongcun site (Fig. 7), providing an ideal place to study the fault slip rates. As before, we used a UAV with a differential GPS (DGPS) system to obtain high-resolution DEM data. T2 on the western bank is eroded and cut, which makes the adjacent T2ʹ only partially remain. According to the high-resolution DEM data, the scarp riser was identified, and the measured displacement was found to be ∼25.4 +1.6/–1.4 m (Fig. S1¹). The scarp of T3ʹ was relatively obvious, and the offset was estimated to be ∼46.8 +1.2/–1.3 m (Fig. S2). The T3 riser was eroded by temporary flow, especially at the fault position. According to the obvious scarp away from the fault position, the displacement was estimated to be ∼49.3 +1.7/–1.8 m (Fig. S3). The T2ʹ terrace on the eastern bank was also eroded near the gully bank, and the measured result was ∼18.7 +1.3/–1.2 m using the marker at the back of T2ʹ terrace (Fig. S4). The processes of displacement measurement are given in the Supplemental Material. At the Jiaerwuzongcun site (Fig. 7), we finally got four offset values that could be used in fault slip rate determination (Table 2).

The Yandantu site was the easternmost site in this study, lying ∼15 km east of the Jiaerwuzongcun site. For the displacement measurement at Yandantu, we also obtained high-resolution DEM data by photogrammetry technology (Fig. 8). The T4 terrace, with obvious offset, is well preserved on the northern side of the fault. Several gullies have developed on the T4 terrace, while it is partly eroded on the southern side of the fault. The measured displacement of ∼36.2 ± 1.5 m is the lower limit of T4 (Fig. S5).

Along the Aksay fault, most of the terraces are covered by eolian loess. The eolian loess has a fine grain size, which could ensure the reset of the OSL signal in quartz due to the exposure to light for enough time during its transport by the wind. Such eolian loess samples are suitable for OSL dating (Küster et al., 2006). The terraces were formed by incision, fluvial erosion, and uplift due to the motions of active faults. The terraces were uplifted above the fluvial channel and provided ideal flat surfaces for the deposition of eolian loess. Hence, the age at the bottom of the loess represents the lower limit to the age of the terrace. To constrain the ages of geomorphic markers, we collected 13 OSL samples and 16 situ cosmogenic nuclide (¹⁰Be) samples.

Through a comprehensive analysis of regional landforms mainly covered by loess along the Aksay segment, four levels of terraces were identified, accompanied by some secondary terraces. The results of OSL age dating in different research sites are consistent, with ages of ca. 1–2 ka, 2–4 ka, 5–8 ka, and 8–10 ka for the T1–T4 terraces, respectively. Among them, the ages of the T2 terrace and T3 terrace are close to the ages of the corresponding secondary terraces, which are consistent with the results interpreted from satellite images.

Three OSL samples were collected at the West of Aksay site on T3 and T4. The age of T3 is ca. 7.6 ± 0.7 ka, from a depth of ∼60 cm under the ground surface. We infer that the offset terrace riser has been seriously eroded, so there is great uncertainty in displacement, which is far less than the accumulated offset since the formation of the terrace. This age was not used in the calculation of the slip rate. The ages of the other two samples from T4 are ca. 8.3 ± 1.1 ka and ca. 9.0 ± 0.6 ka, respectively (Table 3). The oldest age of the geomorphic marker is interpreted to be closer to the formation of the geomorphic surface, so the age of T4 is ca. 9.0 ± 0.6 ka.

Due to the high terrain and long retention time of these old landforms, the remaining terrace surfaces are mainly covered by gravels, so the age constraint depends on the in situ cosmogenic nuclide (¹⁰Be) dating method at the Old Aksay site. For the Tc terrace, the age was determined by age fitting based on the cosmic nuclide concentration in different depth layers in the profile, using the calculation tool developed by Hidy et al. (2010). For the Aksay area, the offset terraces are almost uncovered by cosmic-ray irradiation, so the shading effect was assumed to be 1. The arid or semiarid climate conditions lead to a very small degree of erosion, so the erosion rate can be 0. The fitting age of Tc was 226.7 +21.9/–34.6 ka, and the inherited concentration was ∼12.99 × 10⁴ atoms·g^-1 (Fig. 9). For the calculation of the exposure ages of the Ta and Tb terraces, the inherited concentration obtained from Tc needs to be eliminated based on the ¹⁰Be-²⁶Al exposure age calculator, version 2.3 (https://hess.ess.washington.edu/). The exposure age of Ta is ca. 125.7 ± 11.3 ka, and that of Tb is ca. 172.9 ± 15.7 ka. The formation ages of these three old terraces have a good sequence.

The loess covering the terraces was well collected and exposed thoroughly, which is very suitable for OSL dating (Fig. 10A). On the western bank of Jiaerwuzongcun, T2ʹ is only developed along a small part upstream of the fault, but it is well preserved in the downstream area. We dug a sampling section of 1 m depth in the upstream T2ʹ terrace, which was covered by loess with a thickness of 0.93 m, and some grassroots were mixed in the top of the loess. The OSL samples collected at the position of 0.78 m depth yielded an age of ca. 2.8 ± 0.4 ka. The thickness of the loess on T2 is close to that of T2ʹon the western bank, indicating that its age is close to that of T2ʹ, with an age of ca. 3.2 ± 0.3 ka. The loess thickness of T3ʹ was 1.45 m. Two samples were collected at 0.9 m and 1.32 m depth, which were dated at ca. 4.9 ± 0.4 ka and ca. 6.0 ± 0.6 ka (Fig. 10C), respectively. The thickness of the loess deposited on T3 can be up to 4.42 m, which is much thicker than that of T3ʹ. Age samples were collected at the location of 3.47 m depth, with a measured age of ca. 6.6 ± 0.9 ka. T4 terrace is older and eroded more seriously. It is estimated that the loess cover thickness can be up to 5.26 m. Age samples collected at a depth of 5.18 m yielded an age of ca. 9.0 ± 1.2 ka. On the eastern bank of Jiaerwuzongcun, T2 is covered by loess with a thickness of 1.31 m, and the age of samples collected at 1.23 m depth is 3.2 ± 0.5 ka (Fig. 10B). The measured age of T3ʹis relatively old, ca. 6.0 ± 0.7 ka. The ages of T2 and T3ʹ on the eastern bank are consistent with those on the western bank, which further proves the reliability of the ages.

Although the Tertiary red bedrock has appeared in Yandantu, the surface of T4 on the north side of the fault is covered by loess, with a thickness of 0.58 m. We collected the OSL samples at the lowest layer of the loess, with an age of ca. 7.2 ± 0.6 ka. We also collected an age sample at T2, but the T2 is covered by thin loess. The age obtained was only 2.1 ± 0.5 ka, so it was not used in the calculation of the slip rate.

The formation age of T4 is ca. 9.0 ± 0.6 ka at the West of Aksay site, and the maximum displacement of the gully is ∼98 ± 4 m; these were are used to calculate the slip rate of 10.9 ± 1.2 mm/yr. Because the formation of the gully is younger than the age of T4, the obtained result is the lower limit of the slip rate at this location.

In this study, for multilevel offset geomorphic markers, we calculated the fault slip rate using the Monte Carlo analysis method proposed by Gold and Cowgill (2011). Assuming consistent geological evolution of the Altyn Tagh fault averaged over the late Pleistocene, the fault has not reversed its slip sense. Based on these assumptions, the data show that the reversal of fault motion should not be used in the final slip rate calculation. Based on the above preliminary work, we obtained the displacement and age sequences of Ta, Tb, and Tc at the Old Aksay site, which are ∼1360 ± 150 m, ∼1790 ± 150 m, and ∼2290 ± 200 m, and ca. 125.7 ± 11.3 ka, 172.9 ± 15.7 ka, and 226.7 +21.9/–34.6 ka, respectively. When calculating the slip rate, we considered the errors of offset and age comprehensively and got a slip rate of 10.2 +1.2/–1.1 mm/yr since 220 ka at the Old Aksay site (Fig. 11). Meanwhile, we also summarized the results of previous studies (Table 4). Using the Monte Carlo analysis method, we obtained a slip rate of 9.9 ± 1.1 mm/yr since 15 ka at the Old Aksay site and Huermo Bulak site (Fig. 12; Table 4), which shows that the slip rate of the Altyn Tagh fault was stable over different time scales.

At the Jiaerwuzongcun site, the displacements of the four terraces are 18.7 +1.3/–1.2 m, 25.4 +1.6/–1.4 m, 46.8 +1.2/–1.3 m, and 49.3+1.7/–1.8 m, and the age sequences are 2.8 ± 0.4 ka, 3.2 ± 0.3 ka, 6.0 ± 0.6 ka, and 6.6 ± 0.9 ka. We used the same method—the Monte Carlo analysis method—to fit the slip rate, and the result is 7.5 +1.2/–0.6 mm/yr since 7 ka (Fig. 13). At the Yandantu site, combining the measured displacement of 36.2 ± 1 m and the age of 7.2 ± 0.6 ka, the calculated slip rate is 5.1 ± 0.8 mm/yr. Because T4 is eroded to some extent, the result is the lower limit of the slip rate.

There are many factors that affect the results of the fault slip rate. In practice, the measurement error in displacement, the error in dating, and the methods used to match the two parameters will affect the final results. In this study, displacements were measured based on the high-resolution DEM collected by photogrammetry technology, which was also validated during the field survey. High-resolution DEMs are helpful for us to classify the geomorphic surfaces more reliably and reduce the offset errors at the same time. The semiautomatic measurement code LaDiCaoz_v2 was used in displacement measurements, which could also partially reduce the error of human measurement.

The ages of the geomorphic surfaces were constrained by OSL and ¹⁰Be dating methods, combined with the depth profile method. For geomorphic surfaces that have been eroded, the age of the geomorphic surface obtained by depth profile fitting is more reliable. The inherited concentration obtained by using the depth profile approach is more representative of the inherited concentration of the regional geomorphic surface.

Previously, for calculation of slip rate, the results mostly depended on the offset and age of a single gully or terrace. However, due to the influence of special terrain and human factors, the slip rate of a single point may have great uncertainty. The Monte Carlo analysis method was introduced to comprehensively consider the errors in the two parameters. The method assumes consistent geological evolution of the Altyn Tagh fault averaged over the late Pleistocene; that is to say, the offsets of the younger terraces should not exceed those of the older terraces, and the ages should follow a similar rule, which can narrow the data range and help to reduce the uncertainty in the calculation of average slip rate. Therefore, in this study, offsets and dating results of multilevel geomorphic markers were used to constrain the slip rates at the Old Aksay site and the Jiaerwuzongcun site.

The slip rate since 220 ka was determined to be 10.2 +1.2/–1.1 mm/yr at the Old Aksay site. By comparison with the results of ca. 15 ka obtained by previous studies at this site (Zhang et al., 2007; Chen et al., 2012, 2013), the slip rate of Altyn Tagh fault was determined to be stable over different time scales since 220 ka. Wu et al. (2013) reconstructed the mid-Miocene strike-slip history of the middle segment of the Altyn Tagh fault and concluded that the Altyn Tagh fault had a stage of structural adjustment in the Miocene (ca. 15 Ma), which was also the time of initiation of large-scale sinistral slip on the Altyn Tagh fault. After that, the present geometry and kinematics patterns finally formed. Our results of the study on the slip rate at different time scales in the Old Aksay site are consistent with those of Wu et al. (2013).

Zheng et al. (2017) analyzed the rates of major faults throughout mainland China by using the GPS observation results from 1991–2015. The slip rate of the main segment of the Altyn Tagh fault was ∼8–9 mm/yr, which is basically consistent with the previous GPS observation results of 8–11 mm/yr at 84°E–94°E (Bendick et al., 2000; Shen et al., 2001; Zhang et al., 2007; He et al., 2013). Further eastward, i.e., ∼95°E, the obtained velocity was 4.5 ± 0.9 mm/yr, which shows a good agreement with the GPS profile velocity of 3.9 ± 1.8 mm/yr of Zheng et al. (2017) at 96°E. Combined with the results of this study, four research sites in the Aksay segment are all located in the range of 94°E–95°E, and the slip rate results are also in the range of 8–11 mm/yr and 4.5 ± 0.9 mm/yr. The slip rates gradually decrease starting at the Aksay segment from west to east, which has a good consistency with the existing eastward-decreasing trend from GPS observation rates (Zhang et al., 2007; Zheng et al., 2017).

By analyzing the spatial variation in the slip rate of the Aksay segment, a slip rate reduction has occurred at the middle position of the Aksay segment, from ∼10 ± 1 mm/yr at the Old Aksay site to 7.5 +1.2/–0.6 mm/yr at the Jiaerwuzongcun site to 5.1 ± 0.8 mm/yr near the Yandantu site. We also synthesized geological rates (Kang et al., 2019; Zhang, 2016; Zhang et al., 2007), GPS rates (Li et al., 2018; Zhang et al., 2007; Zheng et al., 2017), and InSAR rates (Liu et al., 2018a) to determine the spatial distributions of slip rate on the eastern Altyn Tagh fault (Fig. 14). From the Aksay segment to the easternmost extent, the slip rate of the Altyn Tagh fault decreases faster in the Aksay-Subei segment, from ∼10 ± 1 mm/yr to 5.1 ± 0.8 mm/yr, which takes up ∼50% of the slip rate within a distance of ∼50 km, with a slip gradient of ∼9.8 mm/yr/100 km. However, the slip rate decreases slowly from Subei to the east, where the other 50% of the slip rate decreases to almost zero over a distance of ∼200 km, with a much lower gradient of 2.5 mm/yr/100 km. The slip rate gradient at the Aksay segment is ∼4 times that of Subei to the eastward termination of the Altyn Tagh fault (Fig. 14). Our results indicate that the slip rate decrease along the Altyn Tagh fault is not uniform at its termination segment, and the gradient can be also variable.

We think that there are several possible causes for the phenomenon of fast-decaying slip rates associated with the Aksay segment in the Altyn Tagh fault, such as the lithological differences along the fault zone and/or the heterogeneity. Li et al. (2018) estimated the fault-locking depth using the screw dislocation model along the Altyn Tagh fault, which showed a heterogeneous distribution. The Aksay segment has a deeper fault-locking depth of 18.9 ± 8.4 km compared to the western and eastern segments of the Altyn Tagh fault. The eastward rapid decay of the slip rate may be caused by the heterogeneity in the physical properties of the deep crust (Liu et al., 2018a). The rapid decrease of slip rate means there is a slip rate deficit, which is also an important sign of potential seismic hazard. The Altyn Tagh fault zone near Aksay County is located at the junction of Cenozoic strata in the north and Sinian metamorphic rocks in the south, with a fracture width of 1000 m. The rocks on both sides of the fault are strongly deformed and broken to form a wide mylonitized and fractured zone (Chinese State Bureau of Seismology, 1992). There is no obvious evidence that the rapid decrease of slip rate is related to the lithology along the fault zone.

The recent research results with more viable analysis, consistent with GPS and InSAR measurements, report a relatively uniform slip rate of ∼10 mm/yr west of 94°E and a decease eastward to only ∼0–1 mm/yr near 97.5°E. If “block extrusion” were occurring, the Altyn Tagh fault would represent a transform fault accommodating the movement between rigid blocks. Its slip rate should be fast enough and not decrease to allow the material on the south side of the fault to extrude eastward, so the low slip rate and the pattern of eastward decrease are not compatible with the block extrusion model. In addition, the syngenetic crustal thickening and strike-slip faults, which span almost the whole Qaidam Basin and Qilian Shan, confirm the wide-ranging deformation to the south of the Altyn Tagh fault. The decreased rates from 94°E to 97.5°E are accommodated by the thrusts and crustal shortening across the Qilian Shan, which support the “continuous deformation model.” Therefore, the Altyn Tagh fault is not the main fault allowing the northeastward extrusion of the Tibetan Plateau. Crustal shortening and thrust faults occur in almost the whole extent of the Qilian Shan, which prove the widespread deformation to the south of Altyn Tagh fault (Meyer et al., 1998). The slip rate of Altyn Tagh fault decreases sharply from the Aksay segment to zero gradually eastward, accompanied by the uplift of the Danghe Nan Shan and Qilian Shan. Therefore, the high decay slip rate on the Aksay segment is related to the uplift of Danghe Nan Shan, which indicates that the extrusion of the Tibetan Plateau has been gradually blocked at this position, resulting in the uplift of these mountain systems. The slip pattern of eastward-decreasing values reflects the limited nature of the extrusion of the northeastern Tibetan Plateau. The large-scale strike-slip movement is mainly confined into the interior of the plateau, which is manifested by crustal thickening and compressional imbricate fault fans of the Qilian Shan (Zhang et al., 2007).

The spatial distribution of slip rate along a particular strike-slip fault and the transformation or dissipation of strike-slip displacement near its termination are of great significance, not only for continental deformation, but also for the behavior of strike-slip faults (Zheng et al., 2013). Zhang et al. (2007) found that the Altyn Tagh fault did not play the role of transition fault in crustal strike-slip extrusion in the collision between the Indian plate and the Eurasian plate, but only acted to redistribute crustal thickening. On the basis of geological slip rates, Zhang et al. (2007) and Zheng et al. (2013) observed that the shortening rate of the whole Qilian Shan is 7.5 ± 2.0 mm/yr. Combined with the shortening amount of the Qilian Shan obtained by Zhang et al. (2014), it is considered that the total shortening amount parallel to the Altyn Tagh fault of the Qilian Shan in the western section is equivalent to the shortening on thrust faults (Zhang et al., 2014; Zheng et al., 2013). For the end of the Altyn Tagh fault, Zheng et al. (2013) and Zhang (2016) analyzed the thrust fault and Cenozoic fold in Jiuxi basin at the western end of Hexi Corridor (Fig. 1) and concluded that the shortening rate of thrust faults on the boundary and interior of the basin is 0.9–1.43 mm/yr in the direction parallel to the Altyn Tagh fault, while the total shortening rate of Cenozoic folds is 0.5–1.0 mm/yr in the direction parallel to the Altyn Tagh fault. The sum of the shortening rates is 1.4–2.4 mm/yr, consistent with 1–2 mm/yr at the end of the Altyn Tagh fault.

The easternmost part of the Altyn Tagh fault, including the NW-trending Danghe Nan Shan, Daxue Shan, and Qilian Shan ranges, is absorbing the sinistral displacement through crustal thickening (Meyer et al., 1998; Xu et al., 2005). Xu et al. (2005) and Shao (2010) inferred that the decrease of the slip rate mainly starts at the position of Subei (Fig. 14), and the decrease is mainly distributed by shortening of the Danghe Nan Shan, with secondary slip vectors in the direction of N70–80°E and N280–290°E, which were obtained through reanalysis of the shortening rate of Danghe Nan Shan, with results of 1.0–1.2 mm/yr. Based on the results of previous studies on the distribution of slip rates, we find that most of the previous studies focused on the position of Subei (Shao, 2010), while the reduction of slip rate in the Aksay segment was not considered. According to the results of Shao (2010) and Van der Woerd et al. (2001) on the shortening rate of the western Danghe Nan Shan, the shortening rate parallel to the direction of the Altyn Tagh fault is only 0.8–1.0 mm/yr. At the same time, Shao et al. (2016) studied the slip rate of the southern edge of Danghe Nan Shan (Houtang fault; Fig. 2) and obtained a slip rate of 2.7 ± 0.9 mm/yr. It is considered that even after a part of the strike-slip rate of the Altyn Tagh fault is allocated, the eastward extent of the Houtang fault will be transformed into the uplift of mountains and shortening of Cenozoic basins and thrust faults. Therefore, the uplift of the Danghe Nan Shan, the crustal shortening of Cenozoic basins, and the thrust faults have absorbed the main reduction in the slip rate, which was ∼3.9–4.1 mm/yr. This absorption is very large in the distribution of slip rate in the whole eastern Altyn Tagh fault, which shows that the Aksay segment plays a crucial role in strain partitioning in the eastern part of Altyn Tagh fault (Fig. 15).

The thrust faults on the western side of Danghe Nan Shan may be growing as an oblique branch fault from the Altyn Tagh fault. The geometry and kinematic characteristics of the fault indicate that the thrust fault and crustal shortening of Danghe Nan Shan may be related to the strike-slip transition of the Altyn Tagh fault (Meyer et al., 1998; Van der Woerd et al., 2001; Yin et al., 2002). Based on the vertical uplift and carbon 14 dating of alluvial fans and terraces, Van der Woerd et al. (2001) estimated the uplift rate of the Danghe Nan Shan thrust at 4–7 mm/yr. Considering a thrust ramp dip of 45°, this result can provide a shortening rate of 5.5 ± 1.5 mm/yr. Assuming that the shortening rate is stable in time, minimum cumulative shortening of 10–12 km can imply that thrusting may have started at least 4 ± 2 m.y. ago. Yu et al. (2019) studied the thermochronological data of the elevation transects of the western Danghe Nan Shan and revealed that there has been rapid exhumation since the middle Miocene (ca. 15 Ma). Generally considered as a branch of the Altyn Tagh fault, the uplift of the Danghe Nan Shan is closely related to the strike-slip transition of the Altyn Tagh fault, so the age may represent the origin of the transition from strike-slip movement on Altyn Tagh fault to crustal shortening of the Danghe Nan Shan. At the same time, previous studies have shown that the Danghe Nanshan has been active at least since the middle Miocene. Our results also suggest that at the eastern termination of the Altyn Tagh fault, the strain is absorbed by shortening of the crust within the Tibetan Plateau, forming mountains and basins, consistent with the continuous deformation model (Molnar et al., 1987; Zhang et al., 2004). This could also be the reason why the slip rate of the Altyn Tagh fault rapidly decreases in the Aksay segment.

The movement and deformation of the Altyn Tagh fault are mainly caused by NE-SW regional compression induced by plate motion. Wu et al. (2013) concluded that the Altyn Tagh fault had a stage of structural adjustment in the mid-Miocene, which may also have been the time of initiation of large-scale sinistral slip on the Altyn Tagh fault. We mainly targeted the activity of the Altyn Tagh fault since the late Pleistocene. Field investigations showed that the Altyn Tagh fault inclines southward with a dip angle of 50°–70°. The main movement mode is strike slip, and no obvious vertical component was found. In addition, there are many thrust faults parallel to the Altyn Tagh fault, south of the fault. Even though the Altyn Tagh fault, as a pure strike-slip fault since the Miocene, will only partially contribute to the uplift along the fault, the thrust faults to the south of Altyn Tagh fault may also be responsible for the uplift on the south side of the fault.

The fault slip rate, which represents the cumulative rate of strain, is of great significance for seismic hazard assessment. Such a high decay rate in the slip rate probably indicates that the movement of the Altyn Tagh fault has been sharply decelerated in the Aksay segment. A part of the slip is absorbed and accommodated by adjacent NW-trending thrust structures, but the NE extrusion pressure on the Altyn Tagh fault still exists, so there is still a high strain accumulation in the Aksay segment. Some studies have shown that there are many obvious seismic surface rupture zones along with the eastern Altyn Tagh fault (Shao et al., 2018; Xu et al., 2017). A relatively complete seismic catalog began in the twentieth century, though some earlier earthquakes may be missing in the historical seismic catalog. The most recent paleo-earthquake near the Old Aksay town occurred 665 ± 40 yr ago, with a recurrence interval of ∼600 yr (Xu et al., 2015), so it is believed that the Aksay segment of the Altyn Tagh fault has not broken for ∼700 yr and may be a seismic gap. Li et al. (2018) calculated the accumulated seismic moments for different segments, which implied that the Aksay bend can be considered as a barrier to earthquake rupture and has a high strain rate concentration, yet it remains unruptured. By calculating the moment accumulation, Liu et al. (2018a) suggested that there was a slip deficit at the Aksay segment, which is equivalent to an Mw 7.9 earthquake based on the elapsed time since the latest M 7 event. Thus, further field investigations will be necessary to determine future cascade seismic ruptures.

In addition, the observed pattern wherein the most slip rate is absorbed by the reverse and thrust faults along the eastern Altyn Tagh fault suggests that the NW-trending thrust structures may be assigned larger strain. For example, the 1932 Ms 7.6 Changma earthquake produced a N70°W-striking coseismic surface rupture along the Changma fault (Peltzer and Tapponnier, 1988). In contrast, the seismic hazard of the Danghe Nan Shan is also worthy of attention.

The slip rate on the Aksay segment provides key information for understanding the evolution of the Altyn Tagh fault and the deformation of the northeastern Tibetan Plateau. We constrained the slip rate of the Aksay segment along the Altyn Tagh fault. Combining high-resolution DEM data and OSL and ¹⁰Be dating methods, the slip rate of the West of Aksay segment is 10.9 ± 1.2 mm/yr since 9 ka. At the Old Aksay site, the slip rate is 10.2 +1.2/–1.1 mm/yr since 220 ka, consistent with a previous study of the Old Aksay site since 15 ka, showing that the slip rate of the Altyn Tagh fault is stable over different time scales. The slip rates at the Jiaerwuzongcun and Yandantu sites are 7.5 +1.2/–0.6 mm/yr and 5.1 ± 0.8 mm/yr since 7 ka. The reason for and mechanism of the spatial change in slip rates were discussed herein. The slip rate decreases ∼5.0 mm/yr within the 50 km Aksay segment, which is the fastest slip gradient along the Altyn Tagh fault. However, the slip rate decreases to almost zero from Subei to the eastern termination of the Altyn Tagh fault, with a much lower gradient of 2.5 mm/yr/100 km over a distance of ∼200 km. The slip rate gradient on the Aksay segment is ∼4 times that of the section from Subei to the eastward termination of the Altyn Tagh fault. It is believed that the remaining 3.9–4.1 mm/yr is absorbed mainly by uplift of the Danghe Nan Shan and crustal shortening of the Cenozoic basins. Considering a thrust ramp dip of 45°, uplift of the Danghe Nan Shan may have been ∼10–12 km since 4 ± 2 Ma (Van der Woerd et al., 2001).

This work was funded by the National Natural Science Foundation of China (41590861 and 41761144071), the National Key R&D Program of China (2017YFC1500401), and the National Nonprofit Fundamental Research Grant of China (IGCEA1901, IGCEA1803). We are grateful to the editors and the two reviewers, Rodolfo Carosi and Maomao Wang, for their constructive comments and suggestions. We thank Jianguo Xiong, Gan Chen, and Ming Ai for fieldwork, and Huili Yang and Yanwu Lv for their assistance with the ¹⁰Be and optically stimulated luminescence dating. We also gratefully acknowledge Olaf Zielke and Ryan Gold for sharing their Matlab codes for displacement measurement and slip rate determination. We especially appreciate Zhiliang Zhang’s valuable suggestions, which greatly improved the manuscript. The authors declare no competing financial interests.