Intramontane basins in actively deforming regions contain significant information about the evolution of orogenic belts. We explored the tectonic characteristics and evolution of an intramontane basin between the Qilian Shan and Yumu Shan mountains on the NE Tibetan Plateau. We utilized the deformation of fluvial terraces along the Dahe River to constrain the rate and pattern of Quaternary deformation across the basin. Fluvial landforms include a widespread Mid-Pleistocene alluvial-fluvial fan surface and five terrace levels inset below this surface. We dated deposits associated with these landforms by optically stimulated luminescence (OSL) analyses on eolian loess and fluvial sediments. Our results yield ages of 142.8 ± 11.3 ka (Fs), 103–125 ka (T5), 96–115 ka (T4), 87–104 ka (T3), and 13–15 ka (T2), which we interpret to reflect the abandonment of fluvial terrace surfaces. Elevation surveys indicate that the terrace surfaces are folded along the Dahe anticline and are gently tilted northward across the basin. Analysis of terrace deformation suggests that the Dahe anticline grew by limb rotation and accommodated upper-crustal shortening at a rate of 0.14 +0.14/−0.03 mm/a. We determined the onset of deformation to have occurred between 0.3 and 0.9 Ma, based on the rotation rate of the southern limb of the anticline. This age is significantly younger than the onset ages of the thrust faults along the Qilian Shan (ca. 10 Ma) and the Yumu Shan (ca. 3.7 Ma). The character of the terrace deformation suggests that the Dahe anticline was growing from a south-dipping décollement, which also induced surface tilting across the entire basin. The Late Quaternary activation of the Dahe anticline folding and the detachment slipping provide evidence that the deformation of the intramontane basin occurred much later than that of the surrounding mountain ranges and accommodated part of the crustal shortening through basin narrowing.

Deformation within orogenic belts is intimately connected with slip along thrust-fold systems, e.g., the Himalayas (e.g., Lave and Avouac, 2000), the Andes (e.g., Gubbels et al., 1993), and the Tien Shan (e.g., Thompson et al., 2002). In thrust-fold belts, crustal shortening is often accommodated by displacement along range-bounding thrust faults and by growth of fault-related folds distributed within intramontane basins (e.g., Burbank et al., 1999). In addition to recent studies focusing on fault slip along the thrusts bordering mountain ranges (e.g., Hetzel, 2013; Zheng et al., 2013b), the rates and wavelengths of surface deformation within intramontane basins can provide additional insight into crustal deformation in active orogenic belts (Goode et al., 2014). Located between the Qaidam Basin and the Hexi Corridor, the NE Tibetan Plateau is characterized by a series of parallel mountain ranges and narrow intramontane basins (Fig. 1) and spans a width of ∼350 km from south to north. The deformation structures in the orogenic belt were interpreted to record a gradual northward growth since the Early Miocene (Burchfiel et al., 1989; Tapponnier et al., 1990; Yuan et al., 2013). In interpreting the crustal shortening across this region, most studies have relied on fault slip rates for the thrust faults along the mountain edge (e.g., Hetzel et al., 2004, 2006; Palumbo et al., 2009; Zheng et al., 2013a, 2013b, 2013c). According to modern GPS velocities, the shortening across the Qilian Shan is estimated to occur at a rate of 5–7 mm/a (Hetzel et al., 2004; Zheng et al., 2013c), although how this shortening is accommodated across the wide mountain-basin region is far from fully understood. Despite significant recent work on surface-breaking thrust faults (Hetzel et al., 2004; Palumbo et al., 2009; Champagnac et al., 2010), we know little about deformation in intramontane basin regions that lie between parallel mountain ranges.

In this study, we provide new constraints on surface deformation in a recently deformed basin, which is located between the Qilian Shan (“Shan” means mountain in Chinese) and the Yumu Shan on the northeastern margin of the NE Tibetan Plateau (Fig. 1). In this region, evidence of thick Quaternary deposits prior to river incision and fold growth suggests a relatively young deformation history within this intramontane basin. We surveyed fluvial terraces along the Dahe River, which transects the basin, to characterize the rate and wavelength of the active deformation. Based on the geometry of the deformed terrace surfaces and tilted strata, we infer the presence of an underlying fold. Our field observations combined with OSL dating on the overlying loess and fluvial sands on the terraces allow us to calculate the rate of shortening accommodated in this region. We discuss how the shortening in this region and the topographic evolution of the basin fits into the ongoing uplift and growth of the NE Tibetan plateau.

2.1 The Growth of the Qilian Shan

The NW-SE–trending Qilian Shan lies along the northeastern margin of the Tibetan Plateau (Fig. 1) and was part of a Middle Ordovician–Early Devonian orogenic belt (Yang et al., 2007). Related to the India-Asia collision (Tapponnier et al., 2001), the Qilian Shan was reactivated in the late Cenozoic, as evidenced by rapid cooling of rocks (George et al., 2001; Pan et al., 2013b), initiation of slip along thrust faults (Zheng et al., 2010), and an increase in sedimentation rate and grain size along at least some of the mountain fronts (Fang et al., 2005). The Qilian Shan has a peak elevation ∼5500 m and is bordered by an active thrust fault (Hetzel et al., 2004, 2006; Champagnac et al., 2010) of the North Qilian Shan fault (NQF), which forms significant topographic relief that reaches 1500–3000 m in the Hexi Corridor (a mountain front basin with average elevation of ∼1500 m) to its north (Fig. 1). Due to the continuing crustal shortening between the Qilian Terrain and the Gobi-Alashan Block (Zhang et al., 2004; Zheng et al., 2013a), a series of new fault-fold systems grew in the mountain front (Hetzel et al., 2002; Zheng et al., 2013a) during the Quaternary, indicating topographic growth to the northeast. North of the central Qilian Shan (Fig. 1), the Yumu Shan is a typical example of recent mountain growth on the NE Tibetan Plateau (Tapponnier et al., 1990; Palumbo et al., 2009), mostly related to the thrusting of the Yumu Shan fault since the Late Pliocene (Palumbo et al., 2009).

2.2 Geology and Fault Activity near the Dahe Region

The studied region (named the Dahe region in this study) is confined by the Qilian Shan in the south and the Yumu Shan in the north (Fig. 2). Stratigraphic studies (Liu et al., 2010; Fang et al., 2012) on the Tertiary–Quaternary sediments reveal that the western section of the Yumu Shan fault was probably active since the Mid–Late Pleistocene in the western Yumu Shan. Palumbo et al. (2009) inferred that the growth of the Yumu Shan initiated at ca. 3.7 Ma based on millennial fault slip rates and catchment-wide erosion rates. Little work has been done on the activity of the North Qilian Shan fault in this region, though the high river incision rate (∼0.8 mm/a) at the upper reach of the Bailang River (Zhou et al., 2002) indicates that the fault has been quite active during the Quaternary. Two larger perennial rivers, the Liyuan and Bailang, flow to the east and west of the Dahe region, respectively, while between these two rivers, only some small seasonal rivers have developed, such as the Dahe River (Fig. 2). Between the Qilian Shan and the Yumu Shan, the Dahe region reaches elevations of 2000–3000 m asl with a maximum relief of ∼300 m in the upper reach of the Dahe River. Close to the Qilian Shan and Yumu Shan, Paleogene and Neogene sedimentary rocks, which are unconformably overlying Paleozoic rocks, are exposed (Fig. 2B). Unlike the Qilian and Yumu mountains, which have cores formed by Paleozoic rocks, the Dahe region’s surface is made up of Mid-Pleistocene conglomerates (Fig. 2B). In the central of the Dahe region between the Yumu Shan and the Qilian Shan, we found an active anticline that exposed the old granite rocks (Fig. 2B) at its core and deformed Late Pleistocene fluvial terraces. This anticline is transected by the Dahe River, and we have named it the Dahe anticline. We will mainly focus on this newly found anticline in the following discussion.

3.1. Terrace Sequence and Distribution

Along the Dahe River, five fluvial terraces are developed, and above the terrace system, a Mid-Pleistocene alluvial fan surface is widely distributed forming the basin top (Fig. 3). This alluvial fan is composed of thick (>150 m) conglomerates, which are referred to as the Q2 conglomerates on the geologic map (Fig. 2B). Below the fan surface, we name the terrace levels by increasing terrace numbers as T1 to T5 from youngest to oldest. Near the village of Panjiawan (Figs. 3 and 4A), terrace sequence is clear, and one transect of the river valley illustrates the different terrace levels and lithology (Fig. 4B). Along this transect, the T5 terrace is 20–40 m lower than the old alluvial fan surface, and the bedrock of T5 is the Q2 conglomerates; however, we did not find the exact contact between the T5 gravels and the Q2 conglomerates. The top of the fluvial gravels of T5 on the east side of the river is 93 m above the modern river bed and is overlain by 5–16 m loess (Fig. 4B). T4 and T3 differ in height by less than 15 m, and both terraces are covered by up to 15 m loess. However, the terrace scarp between them is clear. The fluvial gravels of T4 and T3 are larger than the clasts contained in the Q2 conglomerates. The top of the fluvial gravels of T2 is 10 m above the river bed, and it is a typical strath terrace with a clear contact between the fluvial gravels and its bedrock of Q2 conglomerates. The loess on the T2 surface is ∼3–5 m thick. The top of the T1 gravel layer is 3 m higher than the river bed and is also covered by 1–2 m loess.

All five terraces are relatively well preserved on both limbs of the Dahe anticline. On the south limb, T5 forms a paired terrace, while T4 and T2 are only preserved on the west side of the valley (Fig. 3B). T3 is preserved on the east side of the valley upstream of Panjiawan and on the west side downstream of Panjiawan. On the north limb, T5 is also preserved on both sides of the river valley, while T3 and T2 are only preserved on the east side of the valley. Downstream of Zhoujiazhuang, only T5 and T1 are preserved, and they are distributed in a fan shape (Fig. 3B). Downstream of Nuanquanbu, all terraces except T1 have been eroded or are undeveloped (Fig. 3). Near the axis of the Dahe anticline, except for T1, the terraces are poorly preserved, probably due to the rapid rock uplift and subsequent intense river erosion in the anticline.

In this study, we base our terrace correlation primarily on field observations of terrace distribution and deposition properties. On the north and south side of the Dahe anticline (Fig. 3), each terrace surface is continuous, which allows for a straightforward terrace correlation. For the discontinuity that exists near the anticline axis, we base our correlation on relations between terrace sequence and distribution. For example, we first identified the T2 terrace, which is easy to recognize from its low height and better preservation of terrace surface, and then we identified the higher terraces T3 and/or T4. T5 is a relatively well-developed terrace (Fig. 3B) and easier to identify. Above T5, the Late Pleistocene alluvial fan surface (Fs) is widely distributed as the basin top, and T5 is usually in contact with the Fs. After identifying T2 and T5, we can correlate the lower terraces, T4 and T3. Furthermore, most of the terrace surfaces are covered by loess, and the degree of erosion of the loess cover can also be an identifying feature of the different terrace levels.

3.2. Terrace Deformation

In order to characterize terrace deformation along the Dahe anticline, terrace heights were surveyed by differential GPS with a vertical error of less than 5 cm. Because most of the terrace surfaces are covered by eolian loess, we surveyed both the heights of the fluvial gravel tops and the loess surfaces. Figure 5 shows our terrace correlation: T2, T3, and T4 were correlated by gravel tops; T5 was correlated by loess surfaces in the north (where loess thickness is less than 3 m) and by gravel tops in the south (where loess is relatively thick, up to 16 m). Most of the points on the Fs were correlated by loess caps. Longitudinal profiles of the terraces indicate active folding of the Dahe anticline (Fig. 5). It is obvious that the T3, T4, T5, and Fs are tilted south in the south limb of the anticline (Fig. 5), opposite to the dip of the river bed. Along the north limb, the terrace surfaces are tilted to the north, in good agreement with anticline deformation. Compared to higher terraces (T3, T4, and T5), the magnitude of T2 deformation is smaller, perhaps showing the progressive growth of the anticline. The anticline deformation is concentrated from downstream of Nuanquanbu to Zhoujiazhuang (Fig. 5) to a projected distance of ∼3 km.

For a more detailed analysis, we calculated the slopes of the terrace surfaces with respect to different sections along the river profiles (Fig. 5B). In these calculations, we attempted to remove the influence of the initial terrace slope by tilting every terrace profile so that it corresponded with the slope of the modern river (e.g., Lave and Avouac, 2000), which is completed by subtracting the riverbed elevation from the terrace elevation. For the beveled strata of the Q2 conglomerates, we simply subtracted the tilted angle values of modern river slope. The results (Fig. 5B) show that the surfaces of T3, T4, T5, and Fs had a similar slope of ∼3° in the south limb of the Dahe anticline and had slopes a little smaller in the north limb. T2 is also tilted to the south and to the north on the north and south limbs, respectively, but it has angles of less than 1°. In regions A and C (Fig. 5B) without anticline deformation, the terrace profiles also exhibit a slightly northward tilt with low angles. From the highest surface, Fs to T2, the slopes of the longitudinal profiles systematically decrease from ∼1.1° to 0°. The slopes of each terrace surface are in good agreement, with similar values on both sides (regions A and C in Fig. 5B) of the Dahe anticline. Thus, other than the anticline deformation, the longitudinal profiles of the terrace surfaces are not parallel to the modern river bed; instead, all terrace levels are convergent in a downstream direction.

3.3. Strath Deformation

The bedding of the Q2 conglomerates was also tilted by the growth of the anticline on both limbs of the Dahe anticline. Near the village of Zhoujiazhuang (Fig. 3A) on the north limb of the anticline, T1 was recently exposed by the erosion of a ∼7 m cliff on the west bank of the river (Fig. 6). The T1 gravel layer, buried by ∼1.5 m of eolian loess, is 2–2.5 m thick with approximately horizontal bedding parallel to the modern river bed. The T1 gravels are moderately sorted with diameters of 5–20 cm, and the largest clast is 40 cm. Beneath the T1 gravel, the Q2 conglomerates have a smaller mean grain size than the terrace deposits and are also moderately sorted. Sand lenses are sandwiched in the conglomerates (Fig. 6). The bedding of the Q2 conglomerates dips north at an angle of 12°–17°.

On the south limb of the Dahe anticline, in the Q2 conglomerates below the T5 gravels, an erosion-resistant, well-cemented, interbedded layer is present as a marker bed and shows the tilting of the Q2 conglomerates (Fig. 7). The interbedded layer exhibits a dip to the south with an angle of −2.8°, which is in contrast to the flow direction of the modern river (with a slope of ∼2.1° to the north). The T5 surface also shows a southward dip with an angle of −0.6° in contrast to the northward gradient of the modern river. These observed bed dips of the Q2 conglomerates on both limbs are in agreement with the anticline deformation, with the bedding dipping north in the north limb and dipping south in the south limb.

4.1. Method

The loess overlying on the Dahe terraces has a maximum thickness of ∼16 m and thus is much thicker than the Holocene loess deposits studied by Stokes et al. (2003) in the eastern Yumu Shan and reported by Küster et al. (2006) in the western Qilian Shan. Recent studies of loess deposits near the Yumu Shan (Nottebaum et al., 2015; Zhang et al., 2015) also show that loess deposits are usually thinner than 3 m, and OSL dating on the loess sections mostly yields Holocene ages. Based on this evidence, it is argued that the loess accumulation in the western and central Qilian Shan was initiated ca. 13 ka (Küster et al., 2006). However, some old loess deposits in the western and central Qilian Shan were also discovered, such as the loess section with OSL ages up to 43.1 ± 3.5 ka in the central Qilian Shan (Zhang et al., 2015) and the loess section with OSL age of ca. 81 ka in the western Qilian Shan (Zhang et al., 2015). The old loess deposit found in the central and western Qilian Shan probably suggests that the loess can be preserved given suitable surfaces and environments (Zhang et al., 2015), such as the fluvial terrace surface in the river valley (Fig. 4). According to the relatively thick loess accumulation on the Dahe terraces and the previous loess studies in the Qilian Shan (Pan et al., 2013a; Zhang et al., 2015), we choose the method of OSL dating of the loess and fluvial deposits to constrain the age of the fluvial terrace.

We collected 14 OSL samples in nine sections (Fig. 3) from four terrace levels (T2, T3, T4, and T5) and the Fs to determine their ages. Most of the loess samples were collected from the bottom of loess deposits overlying the fluvial gravels; therefore, the OSL date of the loess provides a minimum age for the time of the river incision, and the terrace surface was abandoned (Pan et al., 2013a). In sampling, a 25-cm-long stainless steel tube was hammered horizontally into the original loess and/or fluvial sediments across the newly excavated vertical section to collect samples. Then, immediately after the tube was removed, the tube was sealed with tinfoil tape and plastic tape at both ends. In the OSL Chronology Laboratory in the Key Laboratory of Western China’s Environmental Systems (Ministry of Education), Lanzhou University, quartz grains of different sizes (38–63, 63–90, and 90–125 μm), which are chosen according to different components of the grain size in the sediment, were extracted and purified following the procedure of Zhao and Li (2002) and Fan et al. (2010). Luminescence signals were measured for 2-mm-diameter small aliquots using an automated Risø TL/OSL-DA-20 reader.

The OSL signal was measured using the modified single-aliquot regenerative protocol (Banerjee et al., 2001) to eliminate the potential contribution of any infrared stimulated luminescence (IRSL) signals from feldspar inclusions within the quartz crystals, and the post-infrared (IR) OSL signal was used to obtain the De values of the quartz fractions. The detailed protocol is described in Fan et al. (2016). The decay curves of the natural OSL and IRSL signals and the growth curve of the OSL signal of a representative sample (DH-611-01) are shown in Figure 8A. The D0 values of 220 Gy for the representative sample in this area support the assumption that accurate OSL ages can be conservatively obtained for the samples with De values close to 440 Gy (Wintle and Murray, 2006). The environmental dose rate was calculated from the concentrations of U, Th, and K in the samples and from the contribution of cosmic rays. The water content was estimated according to the natural water content and saturated water content. Age calculation uses the central age model (Fig. 8B) and represents a 2σ uncertainty. All OSL sample information and measurement results are presented in Table 1.

4.2. T2 Ages

We obtained three OSL ages (Table 1) for the loess overlying on T2 in two different sections located along the north and south limbs of the Dahe anticline (Figs. 3 and 9, respectively). On the north limb, the loess cover overlying T2 is 3–5 m thick (Fig. 9A). From the bottom of the loess 30 cm above the fluvial gravels, an OSL sample (DH625-02) of the loess yielded an age of 13.6 ± 0.9 ka. The loess cover on T2 on the south limb is ∼4 m thick, and fluvial sediment is mixed into the lower part of loess deposits. Optically stimulated luminescence sample DH610-04, collected from the loess 90 cm above the fluvial gravels (Fig. 9B), yielded an age of 12.8 ± 1.5 ka. Sample DH610-03, collected from the position of 10 cm above the fluvial gravels, was loess mixed with fluvial silts and sands and yielded an age of 14.5 ± 1.5 ka. The two ages in succession agree with the depositional sequence. The OSL ages from the bottom of the loess deposit indicate that the T2 surface was abandoned before 12.8 ± 1.5 and 13.6 ± 0.9 ka, while the age of 14.5 ± 1.5 ka probably supplies the age at which the fluvial deposition ended. Thus, we suggest that the T2 surface was abandoned between 13 ka and 15 ka.

4.3. T3 Ages

Four OSL samples (Table 1) were taken from the loess deposits overlying the T3 gravels in two sections located on the north and south limbs of the Dahe anticline (Fig. 3). On the north limb section, the loess on the T3 gravels is 5–8 m thick (Fig. 10A). At the base of the eolian loess, an OSL sample located at 10 cm above the fluvial gravels was collected. In this sample, the eolian loess is mixed with fluvial silts and sands. Optically stimulated luminescence analysis yielded an age of 87.6 ± 9.1 ka (DH625-04). In the south section, three OSL samples from 10 cm (DH624-01), 50 cm (DH624-03), and 100 cm (DH624-04) above the fluvial gravels (Fig. 10B) yielded ages of 104.0 ± 10.1 ka, 87.0 ± 6.7 ka, and 83.0 ± 10.1 ka, respectively. The succession of these three ages is in relatively good agreement with the inferred depositional sequence. Sample DH-624-01 is composed of loess and fluvial sands and silts and supplies the age of the end of fluvial deposition. Among the obtained ages, the loess sample (DH624-03) provides the youngest age for the terrace abandonment, and samples DH625-04 and DH624-01 provide the oldest age control. Thus, the T3 terrace was abandoned between 87 ka and 104 ka.

4.4. T4 Ages

For the T4 age, we collected four samples from sediments overlying the T4 gravels in two sections (Fig. 3), which are located on the south limb of the Dahe anticline. On the south limb, the loess deposition on T4 is 10–15 m thick. We obtained two OSL ages from the bottom of the loess in the first section (Fig. 11). Sample DH611-01, collected from 10 cm above the fluvial gravels, yielded an age of 97.4 ± 8.4 ka. This sample is composed of loess and fluvial silts and sands and probably represents the age of the end of fluvial deposition. Sample DH-611-03 from loess 30 cm above the fluvial gravels yielded an age of 82.3 ± 9.1 ka. In another section, the OSL sample DH610-05 was collected from loess 30 cm above the fluvial gravels and yielded an age of 93.6 ± 9.4 ka. These two OSL ages from the overlying loess could supply the youngest age limit for terrace abandonment. Thus, the OSL result suggests that the T4 was abandoned 82–98 ka. This age range is relatively young and cannot be distinguished from the abandonment age of T3. This confusing result may be derived from the uncertainty of the OSL data and the closure formation age for the T3 and T4.

4.5. T5 Ages

We collected three OSL samples (Fig. 12) from T5 along two sections that are located on the north and south limb of the Dahe anticline (Fig. 3). On the north limb (Fig. 8A), the loess on the T5 surface is relatively thin (∼2 m), and the OSL sample (DH613-03) from the loess 10 cm above the fluvial gravels yielded an age of 102.9 ± 11.8 ka. While on the south limb, the loess on T5 is relatively thick, up to 15 m. On the south limb section (Fig. 12B), one sample (DH612-02) from the bottom of the overlying loess yielded an age of 125.4 ± 16.4 ka, and another sample (DH612-01) from the top of the fluvial gravel layer yielded an age of 116.3 ± 9.2 ka. Although these three samples are in good agreement with their age ranges and are not showing saturation, considering the three ages are all close to the gravel top, we propose that the T5 terrace was abandoned 103–125 ka.

4.6. Alluvial Fan Surface Age

On the north limb of the Dahe anticline, the loess overlying the Fs (Fig. 13) is ∼2 m thick, similar to the loess deposits on T5. Sample DH613-01 on the north limb was collected from the fluvial sediments under the surface of the Fs (Fig. 13). The OSL result yielded an age of 142.8 ± 11.3 ka. Because this sample is close to saturation (De = 446.9 ± 22.6 Gy), the dating result can be conservatively interpreted as the minimal age for the Fs sediment. Since we obtained relatively better age control for T3 and T5, we can evaluate this interpretation by comparing it to an estimated age for abandonment of the Fs calculated by assuming constant incision. The extrapolated age for the Fs is ca. 110–145 ka (Fig. 14), which is in a good agreement with the OSL age for the top of the Q2 conglomerates. Thus, we suggest that the obtained OSL age of 142.8 ± 11.3 ka provides a reasonable constraint for the abandonment age of the Fs or the age at which Q2 conglomerates deposition ended.

In summary, the OSL dating results provide relatively good constraints on the ages of abandonment for T2, T3, T5, and Fs, which are 12.8–14.5 ka, 87–104 ka, 103–125 ka, and 142.8 ± 11.3 ka, respectively. Assuming a constant river incision rate (Fig. 4C), the age for T4 abandonment can be extrapolated as 96–115 ka.

5.1. Fold Models to Explain the Development of the Dahe Anticline

North and south (regions A and C in Fig. 5B) of the Dahe anticline, the longitudinal profiles of the terrace surfaces are not parallel with the modern river bed. Instead, all of the terrace levels and the current floodplain are convergent in a downstream direction. This phenomenon may suggest that the initial longitudinal profile of the older terrace before folding deformation was not like the modern river’s longitudinal profile. Instead, it may have had a steeper river profile than the modern river profile, perhaps reflecting a change in climate factors (e.g., Poisson and Avouac, 2004). Alternatively, the Dahe region has also been deformed on a broader scale from the Qilian Shan mountain front to the Yumu Shan, which may be related to a deeper level of detachment. We discuss this possibility in detail in the next section.

In order to obtain the tilt angle of the terrace directly induced by the growth of the Dahe anticline, we need to exclude the influence of the dips of the initial terrace profiles. We set the longitudinal profiles in regions A and C (Fig. 5B) as the undisturbed profile (by the Dahe anticline) for each terrace. By assuming this, the tilted slopes of the southern and northern limbs would be close to the slope that results from the anticline. We rotate each longitudinal profile of the terrace to align the profiles in regions A and C with a horizontal line. For example, Fs is rotated clockwise by 1.1°, so that the tilting slopes of Fs will become −4.1° and 1.1° on the south and north limbs of the anticline, respectively (Fig. 15). Meanwhile, the observed strata are also rotated by assuming they were formed in a profile similar to Fs.

These adjustments of the limbs (Fig. 14) show that the tilting of the terraces in the south limb of the Dahe anticline decrease gradually from older terrace surfaces to younger terrace surfaces, indicating that the anticline grew through progressive limb rotation. The rotating forelimb would have been the result of either deformation by a detachment fold (Hardy and Poblet, 1994) or by a displacement-gradient fold (Wickham, 1995). Limited strata bedding data preclude a more specific model at present. However, we have reasonably good constraints on the changes in the dip slope (Fig. 14), and these constraints can help us calculate the limb rotation rate and the age of the onset of folding.

5.2. Initiation of the Dahe Anticline Activation

With either a detachment or a displacement-gradient fold, the forelimb can be assumed to have a constant length with rotation (Poblet and McClay, 1996). With a constant forelimb length, the dip of the forelimb will tilt at a rate corresponding to the shortening rate (Poblet and McClay, 1996). If we assume that the forelimb was rotated without bedding-parallel shear, the limb angle θ since initial rotation can be calculated as (Fig. 15A):
where R is the shortening rate of the forelimb (m/ka), L is the forelimb length (m), and t is the time elapsed since the onset of the anticline’s initiation (ka). In this relationship, cosθ would decrease (increasing of θ) linearly with t at a rate of R/L (Fig. 15A), if the shortening rate R remains constant. As the limb rotates, the terrace surface would also rotate at a rate described by Equation (1). Thus, we can use the forelimb slope of the terraces and their ages to determine the unknown factor of R/L (Fig. 15B). We can then use this relation to calculate the strata tilting time elapsed since the Dahe anticline’s onset.

On the south limb (with a length of ∼1200 m, Fig. 14) of the Dahe anticline, the adjusted slope of the bedding dip of the Q2 strata is 6.0°. Using the determined R/L value (1.7 ± 0.3 10−5 ka−1) in combination with Equation (1), the calculated t (time elapsed since initial limb rotation) for the Q2 strata in the south limb would be 320 ± 52 ka (Fig. 15C). Another simple way commonly chosen to calculate the onset age for an anticline is using the linear relationship between terrace tilting and terrace ages (Goode et al., 2014). According to this linear relationship (Fig. 15C), the onset age for the south limb rotation would be 187 ± 35 ka, which is much younger than the age calculated with Equation (1). We know that for a constant shortening rate, the dip of the forelimb would not change at a constant rate (Poblet and McClay, 1996), although the dip may have a nearly linear relationship with age for short periods of time (Fig. 15C). Thus, we choose the age of 320 ± 52 ka calculated by Equation (1) as the starting time of the forelimb rotation. This age would be the youngest age limit for the onset of the Dahe anticline. The strath on the north limb (which is below the strath measured in the south limb) has a steeper dip than that on the south limb, which may be related to an earlier growth initiation for the anticline. In the western Yumu Shan, Liu et al. (2010) found that the greatest angular disconformity in the Cenozoic strata was formed 0.8–0.9 Ma, which probably supplies the oldest age for the onset of the slip on the décollement in the west Yumu Shan. Thus, we suggest that the onset of the Dahe anticline occurred 0.3–0.9 Ma.

5.3. Shortening Rate along the Dahe Anticline

Regardless of the geometry of the underlying fault, we are able to calculate the shortening rate accommodated by the Dahe anticline using the method of area conservation of Lave and Avouac (2000). In the calculation of the shortening rate for each terrace, we use (1) the folded area of each terrace as the shortening area; (2) the longitudinal profile of the terrace surface as its initial profile before deforming; and (3) the age of terrace abandonment as the duration of folding. Hence, in 2D, the amount of shortening length (Ls) can be calculated by:
where At is the folded area (m2), and D is the décollement depth (m). In this region, no available décollement depth data from seismic reflection could be used. However, a previous seismic study along a long transect (Gao et al., 2001) to the west of the study area showed that the décollement depth beneath the Qilian Shan is 10–12 km, and in the north, it is shallower (6–8 km) in the Hexi Corridor area. A new published seismic profile to the west of the Dahe River (Zuza et al., 2016) also shows a depth of 5–8 km for the décollement north of the Qilian Shan. Our study region is located north of the Qilian Shan; therefore, we use the décollement depth of 7 ± 1 km to calculate the shortening rate. The folded area is calculated as a simple triangular area, i.e., the anticline width (we assume a constant width of 3000 m) is multiplied by half of the summit height at the anticline crest (Fig. 14). Most of the terrace heights of the anticline crest are interpolated using the surveyed heights on both limbs, and we assign a 10% uncertainty to these heights. In the results, from Fs to T2, the shortening length gradually decreases (Fig. 14), indicating a progressive shortening since ca. 140 ka. The calculated shortening rate since ca. 140 ka is ∼0.14 +0.14/−0.03 mm/a along the Dahe anticline (Fig. 15D).

6.1. Tectonics, Climate Change, and Geomorphic Evolution

Chronology results indicate that the river incisions along the Dahe River mainly happened in two periods, late MIS 2 (13–15 ka) and MIS 5 (125–82 ka). In general, river incision could be induced by reduced sediment supply (mostly relating to climate change) and/or relative regional uplift (Poisson and Avouac, 2004; Stokes et al., 2012). Studies on the NE Tibetan Plateau suggest that summer monsoon enhancement was synchronous with worldwide deglaciation at the MIS6/5 transition (Wu et al., 2001). The enhancement of the summer monsoon at ca. 140–125 ka (Wu et al., 2001) may have induced an increase in the discharge/sediment ratio and triggered river incision in the Dahe region. A loess study (Wu et al., 2005) also suggests that the climate has been cycling over the past 140 ka, which may have driven the switch from fan deposition to incision and may be the controlling factor for the age at when the river terrace surface was abandoned (Pan et al., 2009). From the terrace composition (Fig. 4B), we can see that the terrace gravels are beveled by the older fan deposits, forming strath terraces (Bull, 1991). This phenomenon may suggest that the river did not experience cyclical transition from deposition to incision and form fill terraces (Bull, 1991) since ca. 140 ka. And thus, climate is not necessarily the sole driving mechanism creating the strath terrace without regional rock uplift (Pan et al., 2001, 2009; Stokes et al., 2012).

The regional uplift and gradual lowering gradients of the terraces from the highest level to the lowest level (Fig. 5) are probably caused by tectonic deformation. The Dahe region is located between the Yumu Shan and Qilian Shan, and its deformation appears to be related to the activation of the North Qilian Shan fault and the Yumu Shan fault (Figs. 1 and 2), which control the uplift of the Qilian Shan and the Yumu Shan, respectively. Previous research (Tapponnier et al., 1990) suggests that a shallow crustal décollement grew from the North Qilian Shan fault, and the crustal slip on the décollement caused the uplift of the Yumu Shan since the Late Pliocene (Palumbo et al., 2009). From the Qilian Shan to the western Yumu Shan (Fig. 2), the Dahe River flows through an incised region upstream of Nuanquanbu (Fig. 2), to a depression between Nuanquanbu and the western Yumu Shan; it then flows to another incised region of the Yumu Shan. The transition between different geomorphic regions may be connected with the geometry of the underlying décollement. To the south of the Yumu Shan, a depression without a terrace indicates that the décollement is flat or only slightly dipping (Fig. 16). To the south, in the incised region, the décollement may have a steeper southward dip, which would result in a higher vertical rock uplift rate. Considering the changes in the slopes of the profiles along the terraces, we propose a curved décollement with a dip decreasing toward the surface (Fig. 16). This geometry would easily explain the systematic tilt of the terraces from south to north.

Contrary to the explanation that tectonics cause the upstream tilt of terraces, climate change can also induce the downstream convergence of fluvial terraces (e.g., Bull, 1991). The modeling result (Wobus et al., 2010) showed that it is hard to distinguish between tectonics and climate as the cause for the downstream convergent geometry of terrace surfaces; however, the channel gradient will decrease caused by a wetter trend in climate, while the channel gradient will increase due to a tilting structure (Wobus et al., 2010). In the Qilian Shan area, the climate record (Wu et al., 2001, 2005) did not suggest the climate is gradually wetter since ca. 140 ka; instead, there seemed to be a dryer trend in Late Pleistocene (Wu et al., 2001). Furthermore, there is no modern glacier and no typical U-shape valley upstream of the Dahe River (Zhou et al., 2002); therefore, we can ignore the influence from melting of the glacier. Thus, it is hard to explain the gradual decrease of the channel gradient along the Dahe River by climate. To more explicitly interpret the role of climate and tectonics in shaping the landscape in the Dahe Region, we need more work on the Q2 deposition and fluvial processes in Late Pleistocene.

6.2. Crustal Deformation of the NE Tibetan Plateau

The deformation of the Dahe region is associated with the fault activity along the west tip of the Yumu Shan fault. Compared to the central part of the Yumu Shan, the western Yumu Shan has a core of exhumed Neogene sediment (Figs. 2 and 16) rather than Paleozoic rocks. The Middle Pleistocene conglomerate covering the Neogene strata suggests the new growth of the thrust-fold belt started in the Quaternary. The newly deformed western part of the Yumu Shan (Liu et al., 2010), together with the evidence for deformation along the Dahe region of the intramontane basin, indicates that the Yumu Shan fault extended laterally in association with the lateral extension of the décollement surface underlying the western Yumu Shan and the Dahe region. Since 0.3–0.9 Ma, slip on the shallow décollement surface induced the uplift of both the western Yumu Shan and the Dahe region.

Our estimate of the shortening rate across the Dahe anticline is ∼0.14 mm/a, which is lower than vertical slip rates of ∼0.4 to ∼0.8 mm/a for the range-bounding thrust fault along strike of the Yumu Shan and the Qilian Shan (Hetzel et al., 2004; Palumbo et al., 2009; Seong et al., 2011). The difference between these rates illustrates that on timescales of thousands of years, strain is not distributed homogeneously across the margin of the Qilian Shan (Hu et al., 2015). The uplifting and folding of the Dahe region (since 0.3–0.9 Ma) are much younger than the initial uplift of the Qilian Shan (ca. 10 Ma, Zheng et al., 2010) and the Yumu Shan (ca. 3.7 Ma, Palumbo et al., 2009). The most recent (Mid-Late Quaternary in this study) lateral extension of the Yumu Shan and the deformation of the Dahe region reveal the lateral growth of the décollement on the NE edge of the Tibetan Plateau, which also suggests that from the north edge of the Qaidam to the Yumu Shan, the mountain ranges and intramontane basins have been gradually added to the crustal accretionary wedge since ca. 20 Ma (Yuan et al., 2013; Zheng et al., 2013b). The uplifting and folding of the Dahe region provide evidence that between the parallel mountain ranges on the NE Tibetan Plateau, intramontane basins (such as the Dahe region, the Heihe valley, the Changma He valley, and the Danghe valley) are also experiencing a certain amount of shortening and uplift above a south-dipping décollement. Uplift and folding of the intramontane basins may have started later than that of the adjacent mountain ranges (Goode et al., 2014) and would accommodate part of the crustal shortening by folding and basin narrowing.

On the NE Tibetan Plateau, we found anticline deformation in the intramontane basin between the Qilian and the Yumu Shan. The intramontane basin of the Dahe region began to deform at 0.3–0.9 Ma ago, which is much younger than the onset time of the bounding thrust faults of the Yumu Shan fault (ca. 3.7 Ma) and the North Qilian Shan fault (ca. 10 Ma). According to continuous regional uplift due to detachment slipping on a curved décollement surface, fluvial terraces were gradually tilted to the north, and the growth of the Dahe anticline on the décollement induced gradual folding of the terrace surfaces. The Dahe anticline grew with limb rotation and accommodated crustal shortening at a rate of 0.14 +0.14/−0.03 mm/a. The late Quaternary activation of the Dahe anticline folding and the detachment slipping provide evidence that the deformation of the intramontane basin occurred much later than that of the surrounding mountain ranges and accommodated part of the crustal shortening through basin narrowing.

The authors would like to thank Yaoyang Lu, Shaofei Jiang, Xilin Cao, and Junwei Mao for their assistances with the field work. The OSL experiments were completed by Defen Mu and Tianlai Fan; we acknowledge their careful work. We thank Eric Kirby, Andrew Meigs, and Peizheng Zhang for their suggestions in preparing the manuscript. We also thank the anonymous reviewers for their excellent suggestions to improve the structure and language for the previous and present versions of the manuscript. Eduardo Guerrero is specially acknowledged for his hard work in improving the language. We would also like to thank LetPub for its linguistic assistance during the preparation of this manuscript. This work was financially supported by the National Natural Science Foundation of China (NSFC) projects (41471009, 41571003, and 41371033) and the Fundamental Research Funds for Central Universities (lzujbky-2016-168).