Formation of the Great Bend and Enhanced Quaternary Incision of the Upper Yangtze River: New Insights from Low-Temperature Thermochronology and Tributary Morphology Open Access

Stretching from east to west, the Yangtze, Mekong, and Salween rivers flow for hundreds of kilometers in close parallel along the southeastern margin of the Tibetan Plateau (Figure 1(a) and (b)). The erosional forces exerted by these rivers are pivotal in revealing the evolution of the regional landscape and patterns of crustal deformation (e.g., [1]). In contrast to the Salween and Mekong rivers, which generally flow southward, the Yangtze River flows southward in its upper regions, from which it makes an abrupt turn and flows northeastward, forming a pronounced bend in the Shigu area (Figure 1(b)). Accordingly, a large-scale capture of the upper Yangtze River has been proposed by most of the previous studies on the assumption that the southward ancestral flow from the Shigu area drained further south into the Red River [1-5]. Although this reorganization of upper Yangtze River drainage has generally been attributed to crustal deformation or topographic uplift of the southeastern margin of the Tibetan Plateau (SETP), there is still no consensus as to when and why this rearrangement occurred. Based on provenance studies of sediments from the present-day upper and middle sections of the Yangtze River, which are considered as residues left by the paleo-Yangtze River, the reconfiguration of the modern Yangtze River has been constrained to Early Pliocene–Middle Pleistocene (e.g., [2, 3, 6-8 ]). In contrast, low-temperature thermochronological data from the area to the southwest of the Great Bend have provided evidence for widespread rapid river-valley incision during the Oligocene–Early Miocene (e.g., [9-12]), corresponding to exhumation/basement cooling of the Tibetan Plateau, but no late Cenozoic incision has been documented. In addition, studies of the stratigraphy of basins along the paleo-southward river course have suggested that the eastward flow of the Yangtze River was most likely initiated before the Miocene (e.g., [4, 5, 13, 14]). However, this conclusion is not consistent with the provenance analysis of these basin-fill sediments, which indicates that these materials were most likely derived from local basement sources and were instead transported by a large through-flowing river (e.g., [15-17]).

After flowing southeastward and passing through the V-shaped bend in the vicinity of Shigu, the Yangtze River heads abruptly northeast (Figures 1(b) and 2(a)) and flows through a narrow (30 m, 80 m wide) and deeply incised canyon between the Yulong and Haba mountains (Figure 3(a)). This deeply incised canyon is called “Tiger Leaping Gorge,” and fluvial incision along this part of the Yangtze River valley is associated with the formation of the Great Bend. The cooling and exhumation histories of the Tiger Leaping Gorge are poorly constrained, hindering the understanding of its formation. The only previous attempt to date the uplift age of the Yulong Mountains was based on the ⁴⁰Ar/³⁹Ar dating of K-feldspar, which indicated an onset of uplift at ca. 17 Ma, corresponding to a change in deformation mechanism associated with the Indochina’s extrusion [18]. However, this timing is inconsistent with geomorphologic evidence that links the uplift to normal faulting bounding the Haba and Yulong mountains during the Quaternary (e.g., [19-21]).

This study presents apatite (U–Th)/He (AHe) and fission track (AFT) thermochronological data for rocks obtained from the metamorphic core of the Yulong Mountains cropping out in the Tiger Leaping Gorge. Similar analyses along a vertical profile of the Chongjiang River valley adjacent to the Great Bend are also conducted. The AFT and AHe systems document cooling and exhumation of the upper ~5 km of the crust through the thermal ranges of 60–120 and 40–80°C, respectively [22], and can therefore be used in the present study to recover the exhumation and cooling histories of the Great Bend area. Comparing our new low-temperature thermochronometric data with results from the Jianchuan Basin to the southwest of the Great Bend, we reveal a younger two-phase uplift history of the Great Bend during the Miocene and the Quaternary. In addition, we conduct quantitative analyses of tributary morphology and channel morphometric parameters for the Tiger Leaping Gorge to reveal river-incision and crustal-deformation histories.

The SETP underwent complex and multiple stages of deformation (Figure 1(a)) associated with the indentation of the Indian plate into the Eurasia plate. Since ca. 50 Ma, plate motion reconstructions have shown ~3000 km of relative motion from India to Eurasia [23]. Although much of this convergence has been accommodated by shortening both in the Himalaya orogen and Asia, a considerable amount is considered to have been accommodated by material migration around the Eastern Himalaya Syntaxis (EHS) of the Tibetan Plateau [23]. Such a material migration has resulted in long-wavelength uplift and associated river-valley incision in the SETP [24-26]. A typical proxy for this material movement is the lateral extrusion of the Chuandian Block (Figure 1(a)). The Chuandian Block is a lens-shaped crustal fragment, bounded by the right-lateral Red River fault (RRF) and left-lateral Dali fault to the west and by the left-lateral Xianshuihe–Xiaojiang fault to the east (Figure 1(b)).

To the southwest of the Great Bend lies the Jianchuan Basin, characterized by its triangular shape. It is delineated by the Jianchuan normal fault, exhibiting a left-lateral component to the east, and the Qiaohou dextral strike-slip fault to the west (Figure 2(a) and (b)). The Qiaohou fault is related to the RRF and is characterized by thrusting with a right-lateral component, which caused the folding of the basin interior during the early Cenozoic (Figure 2(c)). The Jianchuan fault has shown a predominantly normal sense of motion under an extensional regime since the Late Miocene [10, 13, 26]. Paleogene fluvial–lacustrine sedimentary rocks have formed a thick (>5.5 km) fill in the Jianchuan Basin (Figure 2(b)). Analysis of sedimentary facies and provenance has indicated that these sedimentary rocks, comprising mudstones, siltstones, fluvial sandstones, and conglomerates, are remnants of the ancient paleo-Yangtze River [5, 13], flowing southward during the late Eocene. These deposits were contemporaneously laid down alongside the widespread occurrence of high-K magmatism in the Jianchuan Basin [5, 10, 13]. The ultrapotassic plutons in the Jianchuan Basin are mainly composed of quartz monzonite, granite, and syenite. The emplacements of these plutons are regarded as indicative of lithospheric delamination or continental subduction during the Eocene [27]. Therefore, recent studies of the Jianchuan Basin have suggested that fluvial incision associated with the generation of topographic relief in SETP occurred during the late Eocene–Early Miocene [5, 10-12], implying that the formation of the Great Bend and regional drainage rearrangement took place during the same period.

The study area of this investigation is located in the northwestern corner of the Dali fault system (DFS) (Figure 2(a)). The Haba and Yulong mountains constitute the highest range in the study area, and the flow of the Yangtze River through these mountains has formed the Tiger Leaping Gorge (Figures 2(b) and 3(a)). A fairly complete section of the elongated Yulong–Haba range is revealed by this deeply incised gorge, including a broad N–S-trending anticline comprising folded Lower Devonian to Permian strata of the Yangtze Block (Figure 2(c); [18, 20]). The juxtaposition of metamorphosed Permian basalt and Devonian–Permian marbles defines a normal fault zone along the eastern boundary of the Yulong anticline (Figures 2(b)-(c) and 3(b)-(c)). The metamorphic core of the anticline mainly comprises mica schists, gray schists, and intercalated graywacke and quartzite layers (Figure 3(c)–(e)). The predominant minerals are deformed chlorite, epidote, muscovite, biotite, quartz, albite, K-feldspar, and accessory minerals of zircon, apatite, and siderite, indicating a low-temperature, greenschist metamorphic grade [18]. The protolith of these schists comprises volcanic clastic rocks and finely laminated sandstones [18]. In the western and eastern limbs of the anticline, very similar lithological units to those described above for the metamorphic core are found, but with a lower level of deformation. The folding of the Yulong Mountains is related to E–W shortening associated with Indochina extrusion since ca. 17 Ma, which has also been interpreted as the main cause of the regional uplift [18]. Nevertheless, a growing body of studies has suggested that much of the uplift of antiformal structures was related to normal displacement along boundary faults of the Yulong Mountains during late Cenozoic extension (e.g., [19-21]). Towards the southwest of the Great Bend, a comparable assemblage of metamorphic rock series, similar to the one previously mentioned, is present, primarily comprising mica schists and plagioclase gneiss. These metamorphic rocks are widely developed along the upper course of the Yangtze River (Figure 2(b)) and are referred to as the Shigu Complex.

The temperature-sensitive zone of AHe and AFT of 40‒80 and 60‒120°C [28, 29], respectively, result in their thermal sensitivity to cooling of the shallow crust. Synthesizing both analyses therefore allows constraining on the thermal evolution involved in very young change in relief.

To better understand the incision history of the Tiger Leaping Gorge, three bedrock samples (YN02, YN04, and YN07) were collected from the metamorphic core of the Yulong anticline (Figure 2(b) and (c)). The samples were distributed with limited elevation variation owing to the limited exposure of the metamorphic core (Figure 2(c)). To complement these data, we also collected 5 samples along a vertical profile of ~700 m to the southwest of Shigu along a tributary of the upper Yangtze River, the Chongjiang River, within the Shigu Complex (Figure 2(b) and (c)). To compare the thermal history of the Great Bend, which experienced episodic incision events, with an area where thermal history remained unaffected by river network reorganization, we reaffirmed the previously documented thermal history [10, 11]. This confirmation was achieved through the analysis of AFT and zircon fission track (ZFT) data from samples collected along a vertical transect, spanning elevations from 2550 to 3870 m, within the Laojun Shan syenitic plutons situated in the Jianchuan Basin (online supplementary Figure S1).

More than 1000 apatite and zircon grains were separated by conventional magnetic and heavy-liquid separation procedures and then handpicked under a binocular. The AFT analysis of the samples collected along the Tiger Leaping Gorge and Chongjiang River was performed at the School of Earth Sciences, University of Melbourne. A Zeiss Axio Imager microscope was used to capture fission track images, and the data were further processed through the softwares Trackworks and FastTracks [30]. An Agilent 7700 LA-ICP-MS coupled with a New Wave UP-213 laser was used to determine uranium concentrations of corresponding grains. Detailed procedures are described in [30]. AHe analysis was performed at the Institute of Geology, China Earthquake Administration. Only euhedral crystals without the presence of potential small inclusions were selected and loaded in platinum capsules based on observation under a polarized light binocular microscope. The isotope ⁴He was determined using an alphachron instrument, and the ²³⁸U–²³²Th analyses were conducted by an Agilent 7500× quadrupole mass spectrometer. Accuracy of the data was verified by measuring the Durango apatite grain after every five measurements. Apparent AHe ages were corrected for alpha emission following the method of [29].

The QTQt inverse-modeling software [28] was used to perform thermal-history modeling by inverting ages determined from different low-temperature thermochronological systems based on a Markov Chain Monte Carlo method. Owing to the young cooling ages, a few confined track lengths, which are the most commonly used kinetic indicator in thermal-history modeling, were obtained for the samples collected from the Tiger Leaping Gorge. Therefore, we first performed the modeling including only AFT and AHe data of the Chongjiang profile. Then, we also synthesized our AFT and AHe data for the Chongjiang profile and Tiger Leaping Gorge to perform a complete thermal-history modeling of the Great Bend on the basis that (1) these samples were collected from the same metamorphic rock series that occurs extensively along the upper reaches of the Yangtze River (Figure 2(b)) corresponding to a previous underthrust into the Yangtze Block; and (2) all of these samples were collected from the northeastern and southwestern flanks of the Great Bend (Figure 2(b)), located within ~60 km of the river and should therefore have undergone similar tectonic and thermal evolutions. For comparison, the thermal history of the Jianchuan Basin was inferred using AFT and ZFT data from this study in conjunction with AFT and AHe data previously reported by Cao et al. [10] and Shen et al. [11].

Generally, channel erosion rate (E) for both alluvial rivers and bedrock can be described by a stream-power incision model as a function of local gradient (S) and contributing drainage area (A):

where n and m are positive exponents related to erosion process, basin hydrology, and hydraulic geometry and K is an erosion coefficient dependent on sediment load, climate, and rock properties [31]. Given that the competition between erosion rate and rock uplift governs the evolution of a typical river longitudinal profile, the elevation change of channel bed within certain time (dz/dt) can be expressed as:

where x is horizontal distance along a channel and U is rock uplift rate. Once a steady state is achieved, dz/dt = 0, the equilibrium slope (S_e) can be described by a power law function:

where k_s = (U/K)^1/n was usually normalized to the contributing drainage area, representing the channel steepness index k_sn; θ = m/n equaling to the concavity index and referring to a proxy for discharge. Empirical studies suggest that the channel steepness (k_sn) has a positive correlation with the uplift rate under steady-state conditions [32], whereas the concavity index (θ) commonly shows the independence of rock uplift/erosion rate. Thus, equation (3) implies that decrease in erosivity and/or increase in uplift rate will result in a steeper slope corresponding to a new equilibrium channel starting from the basin outlet [32, 33]. Owing to the new equilibration, the knickpoint will propagate upstream, which will result in an increase in the steepness index (k_s). However, the section upstream of the knickpoint will be unaffected by an old equilibrium profile under conditions of constant erosion coefficient and uplift rate [32, 34]. Therefore, the paleo-base level can be projected from the knickpoint to the tributary outlet [33, 34] and a total river incision can be calculated by comparing the upper segment and modern river elevations.

Our analysis centers on tributaries to the Tiger Leaping Gorge given that its formation is a key indication of the eastward flow of the Yangtze River (Figure 4(a)). The 30 m ASTER digital elevation model (DEM) was used for channel morphology analyses. ChiProfiler [35], TopoToolbox [36], and Transient-Profiler [33] were adopted to perform tributary profile analyses and to produce a map of river network k_sn. We smoothed the elevation data using a window size of 250 pixels. Thus, the initial form of the profile was retained as much as possible. A cutoff value of 1 km² was used as the critical drainage area to exclude hillslope-dominated channels from fluvial scaling [37]. Both projected profiles (Figure 4(b)) and steepness indices were calculated by a log scaling relationship between slope and drainage area (Figure 4(c)), and the slope-break knickpoint was identified based on the persistent break in the linear pattern. These results were further confirmed by χ-elevation profiles based on an integral approach [38]. The k_sn was calculated by a reference concavity of 0.45 [15, 37, 39], which was then interpolated onto a regular grid through an ordinary kriging method. Owing to 2σ errors in the estimation of normalized k_sn, both the lower and upper boundaries can be calculated for the reconstructed paleo-base level.

We obtained 3 AFT ages (Table 1) and 10 single-grain AHe ages (Table 2) from the metamorphic core of the anticline in the Tiger Leaping Gorge. Sample BN04, collected from the central part of the metamorphic core, yielded an AFT age of 3.9 ± 1.3 Ma (1σ), with an age dispersion of 0 and P(χ²) of 0.93. In addition, AHe ages obtained from grains range from 1.9 ± 0.1 to 0.3 ± 0.1 Ma, with a mean of 0.9 ± 0.1 Ma. In the limbs of the anticline, samples BN02 and BN07 show more scattered AFT ages of 1.6 ± 1.8 and 0.8 ± 0.4 Ma, respectively, with the same age dispersion of 31% and P(χ²) ranging between 23% and 34%. The wider variation in single-grain ages of these 2 samples compared with that of BN04 is attributed to their low U contents and young ages (Table 2). Sample BN02 yielded three apatite grains suitable for AHe dating, with ages between 2.39 ± 0.10 and 6.71 ± 2.34 Ma, and sample BN07 provided 2 suitable apatite grains with AHe ages of 0.22 ± 1.79 and 0.61 ± 0.09 Ma.

Along the transect close to the Chongjiang River southwest of the Great Bend (Figure 2(b) and (c)), 5 samples ranging in elevation from 2341 to 3046 m yielded both AFT and AHe results (tables 1 and 3). AFT ages range from 6.8 to 11.0 Ma and yield a slope of 0.15 ± 0.03 mm/year, showing a positive relationship with elevation (Figure 5(a)), and their confined track lengths show a unimodal distribution with similar mean track lengths (MTLs) ranging between 13.7 and 14.2 µm (Table 1). Of these 5 samples, two low-elevation samples show significant dispersion of >20% and fail the P(χ²) test (<5%), probably owing to a prolonged period of residence within the partial annealing zone. Additionally, the 5 samples yielded 11 single-grain AHe ages. The lowest-elevation sample SG04 shows homogeneous single-grain ages of 9.6 ± 0.6 and 8.7 ± 0.5 Ma. Sample SG05 provided a single suitable apatite grain with an AHe age of 11.5 ± 0.7 Ma. Sample SG07 yielded three concordant single-grain ages between 13.2 ± 0.8 and 9.5 ± 0.6 Ma. The higher-elevation samples (SG03 and SG14) exhibit more scattered single-grain ages ranging from 17.4 ± 1.1 to 6.5 ± 0.4 Ma. There are no clear relationships between the AHe ages and the eU (effective uranium contents) and between the AHe ages and the grain sizes (Figure 5(b) and (c)). Therefore, even though the eU (5, 50 ppm) and grain sizes vary in relatively large ranges, it is difficult to explain the dispersion of single-grain AHe ages. Given that 8 of the 11 single-grain ages are younger than the AFT ages, we attribute the abnormally old ages to the additional sources of ⁴He that derive either from the U-rich neighboring minerals or U-rich mineral inclusions in apatites which commonly occur in high-grade metamorphic rocks like gneiss and schist [40]. However, all the AHe ages cluster between 17 Ma and 6 Ma, suggesting a rapid cooling around 6−17 Ma cooling.

Thermal-history modeling synthesizing all AFT, AHe, and ZFT data in this study (online supplementary Tables S1 and S2) and previous studies [10, 11] further confirms a three-phase cooling history model of the Jianchuan Basin (Figure 6(a)): (1) initial 40–35 Ma rapid cooling associated with thermal equilibration since pluton emplacement at ca. 36 Ma; (2) a second phase of rapid cooling during 35–23 Ma corresponding to rapid exhumation; and (3) a phase of thermal stability since ca. 23 Ma. Single-grain ages predicted by the modeling fit the observed data well for the high-elevation samples but deviate from those for the low-elevation samples (Figure 6(b)). Therefore, compared with the previous model based only on AFT data [10], modeling results taking into account multiple low-temperature thermochronometers show no subsequent phase of thermal stability between 24 and 20 Ma.

A different three-phase cooling history from that of the Jianchuan Basin is suggested by our modeling results for samples from the Great Bend area (Figure 6(c)–(h)): (1) a phase of rapid cooling during ca. 11–8 Ma; this phase of rapid cooling were recorded both by the Chongjiang (Figure 6(c)) and Tiger Leaping Gorge profiles, with a maximum rate of 24 ± 6°C/Ma for the highest sample (SG14 in Figure 6(f)) and a minimum rate of 13 ± 10°C/Ma for the lowest sample (BN02 in Figure 6(g)); (2) a phase of thermal stability during the Late Miocene to early Quaternary, with a mean cooling rate of 0 ± 3°C/Ma; and (3) very rapid cooling since the early Quaternary; this phase of rapid cooling was only recorded by the samples collected from the bottom of the Tiger Leaping Gorge, with a maximum rate of 38 ± 5°C/Ma for the lowest sample (BN02 in Figure 6(g)). Figure 6(d) and (h) shows the fitting levels between the modeled data and observed data. The modeled AFT ages show a break-in-slope point at ca. 8 Ma, most likely indicating the age at which the exhumation of a fossil partial annealing zone ended (Figure 6(h)). The thermal-history modeling also suggests that the temperature offset between different samples was most likely constant before the Quaternary (Figure 6(c) and (e)), allowing a geothermal gradient of ~45°C/km to be estimated for the period prior to the Quaternary. This gradient is consistent with the value for the adjacent Mekong River estimated by [41]. Therefore, the exhumation rate is estimated at 0.3–0.5 mm/year during the first phase of cooling from 11 to 8 Ma, corresponding to a total exhumation amount of >1–2 km. Since the early Quaternary, the temperature offset has decreased sharply to 0°C at present owing to the occurrence of differential erosion among different samples. Thus, based on the present-day thermal gradient of ~25°C/km in the Shigu area [42], the exhumation rate is estimated at 1.5 ± 0.2 mm/year, corresponding to ~3 km of exhumation during the Quaternary.

In general, variation in k_sn along the Tiger Leaping Gorge shows minimal correlation with mapped lithological changes (Figure 7(a)), indicating only a limited influence of variable lithology on the channel profile gradient. The mean annual precipitation derived from the Tropical Rainfall Measurement Mission (TRMM) [43] decreases substantially from west to east in the vicinity of the Tiger Leaping Gorge (Figure 7(b)) under the influence of the East Asian monsoon [44]. However, the steepest channel gradients are located in the area of weaker precipitation (Figure 7(b) and (c)) to the southeast of the gorge, generally corresponding to the front of the shallow low-angle thrust fault along the metamorphic core of the anticline (Figure 7(a)).

We analyzed 43 tributary longitudinal profiles to produce sufficient data for a meaningful regression model (online supplementary Table S3). Two major phases of preincision are identified base on the reconstructed channel profiles, prominent knickpoint positions, and variation in θ and k_sn values (Figures S2 and S3). Of the 43 tributaries, 25 join the Tiger Leaping Gorge from the northeast draining the Haba Mountain, and 18 join from the southeast draining the Yulong range (Figure 4).

Of the 43 profiles, phase I channel segments were identified in 10 tributaries for each side (NW and SE) of the Tiger Leaping Gorge (online supplementary Table S3), with mean θ values of 0.29 and 0.43 (online supplementary Table S3 and Figure 8(a)), respectively. Phase II segments were identified in 20 tributaries for the northwestern side and in 10 tributaries for the southeastern side, with mean θ values of 0.50 and 0.91, respectively. With increasing distance downstream along the Tiger Leaping Gorge, there are progressive increases in channel steepness and total river-valley incision of the upper phase I profiles, whereas these indices are essentially constant for the phase II profiles (Figure 8(b) and (c)). In addition, there is no apparent relationship between variation in these indices and topographic relief (Figure 8(d)), in particular, the two peaks of the Haba and Yulong Mountains. In general, the upper phase I segments are characterized by low steepness gradients increasing from 26 to 181 with mean values of 91 and 98 for the northwestern and southeastern sides, respectively (Figure 8(b)). Reconstructed longitudinal profiles of phase I incision reveal that total river-valley incision increases from 1207 to 1919 m with increasing distance downstream (Figure 8(c)). Values of k_sn of phase II segments range from 68 to 289 with mean values of 153 and 187 for the northwestern and southeastern sides, respectively. The mean total incision of this phase is ca. 722 m.

Our thermal-history modeling of the Jianchuan Basin (Figure 6(a)) confirmed an extensive pluton emplacement and rapid cooling before ca. 35 Ma [10, 45], corresponding to widespread rapid exhumation in the SETP since the late Paleogene [10, 30, 46-50]. An increasing number of studies have suggested an early stage of Tibetan Plateau uplift since the late Paleogene based on paleo-altimetric, low-temperature thermochronological, and sedimentological data [12, 13, 19, 51, 52], accompanied by the widespread intrusion of plutons [5, 13, 14]. The Oligocene or earlier uplift is also recorded by the boundary faulting of the metamorphic rocks in the Great Bend area, which was characterized by the development of a large-scale décollement at ca. 36 Ma [18], corresponding to extrusion of the Indochina Block [53]. The phase characterized by bedding-parallel ductile décollement has been recognized as the primary mechanism responsible for the rapid exhumation of the Jianchuan Basin [10]. This process included the exhumation of significant portions of the Proterozoic crystalline basement from the western margin of the Chuandian Block within the Yangtze Block. The exhumation process occurred in areas such as the Tiger Leaping Gorge and the Ailaoshan–Red River shear zone [49, 54].

However, compared with samples of the low-temperature data of samples collected from the Jianchuan Basin with elevations between 2029 and 4045 m, the results in this study derived from the adjacent Yangtze River and its tributary valleys also cover a similar elevation range but show a pronounced jump in both AFT and AHe ages, representing an initial phase of rapid cooling before ca. 11 Ma (Figure 6(c) and (e)). These results demonstrate that the Middle–Late Miocene rapid cooling recorded by rocks of the Yangtze River in the Great Bend area was not related to extensive Oligocene or earlier regional uplift but corresponds to rapid river-valley incision. Based on cosmogenic nuclide burial ages of sediments collected from caves, the incision of the Yangtze River gorge at the Great Bend was also dated between 18 and 9 Ma [45]. During this period of rapid cooling, the deformation mechanism in the vicinity of the Tiger Leaping Gorge shifted from ~N–S-directed drift on a single large-scale décollement to ~E–W-directed shortening on the décollement [18]. Results of ⁴⁰Ar/³⁹Ar feldspar dating have suggested that the associated anticlinal folding in the Tiger Leaping Gorge started at ca. 17 Ma (Early Miocene) [18]. Approximately 80 km northwest of the Great Bend (Figure 1(b)), the eastern flanks of the Mekong River valley also show fast exhumation during this phase of cooling. Replumaz et al. [41] attributed this phase of cooling to the reactivation of the Zhongdian and Parlung right-lateral strike-slip faults which results in the formation of a large-scale restraining left-stepping overstep. The Zhongdian fault also extends southeastward to the Great Bend and probably connected with the Jianchuan fault (Figures 1(b) and 2(a)).

In fact, this Middle–Late Miocene rapid cooling has been widely identified from rocks obtained from most of the major river valleys of southeastern Tibet: from Early Miocene until at least ca. 11 Ma in the Salween River valley [12, 39, 55], starting during the Middle–Late Miocene in the Mekong River valley [41, 48, 56], 11–8 Ma for the Yangtze River valley (this study), 14–10 Ma in the Red River valley [15, 49], 13–9 Ma in the Dadu River valley [9, 57], and starting at ca. 20–14 Ma in the Yalong River valley [57, 58]. The northeastern end of the branch of the Jianchuan fault that extends along the eastern flank of the Tiger Leaping Gorge exhibited strong thrusting activity during the middle–late Cenozoic (Figure 2(b)), forming steep channel gradients along its front (Figure 7(c)). Wang et al. [49] proposed that the high rate of exhumation in the Red River valley since the Middle–Late Miocene likely resulted from kinematic reversal of the RRF from a single strike-slip faulting to strike-slip faulting with a dip-slip component. Based on low-temperature thermochronological analyses of the Salween River, the rapid cooling during the Middle–Late Miocene is considered to have resulted from a shift into diffuse deformation in the SETP [39]. Furthermore, in the area to the north of the Great Bend, rapid incision of rivers during the Late Miocene is directly considered to have resulted from the uplift of the Tibetan Plateau [9]. It has also been argued that the incision of the large-scale rivers along the SETP is climatically driven [44, 56]. If there is a global signal, the samples covering a similar elevation in the adjacent Jianchuan Basin should also record such a Miocene rapid cooling. Therefore, compared with the earlier (pre-Miocene) uplift, the rapid river-valley incision since the Early Miocene most likely occurred during a period of geodynamic transition. During this period, most faults of the DFS started to reactivate mainly as a series of strike-slip faults with dip-slip components [10, 26]. Thus, the lateral and/or vertical movements (transpressional) of the Zhongdian and/or Jianchuan faults triggered tectonic uplift, which resulted in the enhanced river incision of the Yangtze River valley system in the Great Bend area during the Miocene.

The multiple phases of rapid cooling recorded in most parts of the Jianchuan Basin (Figure 6(a)) are substantially older than that of the Great Bend for similar elevations (Figure 6(b) and (c)), suggesting that the capture of the Yangtze River in the Great Bend area was not caused by the widespread uplift prior to the Miocene [10, 51]. In the Tiger Leaping Gorge, sharp knickpoints identified on most tributaries generally separate their longitudinal profiles into 3 segments (Figure 4(b)), suggesting two major pulses of river-valley incision. Our low-temperature thermochronological age versus elevation plot indicates that the exhumation located above ca. 3 km asl. occurred before 11–8 Ma and that between 3–2 km asl. occurred after ca. 1.9 Ma (Figure 5(a)). Therefore, the corresponding height of exhumation is generally consistent with the reconstructed paleo-base level (Figure 4(b)), suggesting that the first phase of rapid cooling before 11–8 Ma corresponded to the initial phase of river-valley incision and that the second phase of rapid cooling since the early Quaternary corresponded to the subsequent pulse of incision. Compared with the later incision, the k_sn values and total incision amounts of the upper-channel profiles display a progressive increase with increasing distance downstream (Figure 8(b) and (c)). This trend shows no apparent relationship with topographic relief (Figure 8(d)) but most likely reflects headward erosion associated with river capture, meaning that the Great Bend formed during the initial phase of rapid cooling during the Miocene. The lower-channel profiles show reasonably uniform k_sn values and incision amounts (Figure 8(b) and (c)), indicating that during the period of later incision, the Tiger Leaping Gorge has been cut through, and that the formation of the knickpoints was not related to river capture. Thus, our new results contrast with previous interpretations of the evolution of the modern Yangtze drainage system starting since the early Quaternary [2, 3, 6].

Results of thermal-history inversion of the Tiger Leaping Gorge show rapid Quaternary cooling (Figure 6(e)) with a mean AFT age of ca. 1.9 Ma (Table 1) and a total exhumation amount of ~3 km. The ~700 m total incision since the early Quaternary (Table S3 and Figure 8(c)) is far less than the ~3 km total exhumation amount, indicating that the longitudinal profiles are still undergoing a transient response to a change in climatic and/or tectonic conditions [59, 60]. In contrast, the initial phase of rapid cooling during the Miocene corresponds to a total exhumation amount of >1–2 km, which is generally consistent with the total river-valley incision amounts (Table S3 and Figure 8(c)), suggesting that a steady-state channel profile has essentially been achieved.

The enhanced Quaternary incision in the SETP has previously been attributed to either local uplift or river reorganization [41, 61-63]. However, an increasing number of cases of rapid Quaternary exhumation associated with river incision, mainly in large-scale river valleys, have been identified (Figure 9(a)). Adjacent to the EHS, enhanced incision since ca. 2 Ma is recorded by both the Yarlung and Yigong rivers [62, 63]. In the Three Rivers region, low-temperature thermochronological data derived from valley bottoms have revealed rapid river-valley incision since the early Quaternary [59, 64, this study]. In the western margin of the Sichuan Basin (WSB), rapid Quaternary incision has been identified in the Dadu, Yalong, and Anning river valleys [57, 61, 65]. Therefore, the widespread occurrence of this rapid incision from the WSB to the EHS and comparable exhumation rates of 1–5 mm/year [41, 57, 62, this study] most likely suggest a widespread uplift in most of the SETP region.

Increased climatic fluctuations, including global cooling and intensified monsoon precipitation, have been identified as starting during the early Quaternary [66], which could also have induced enhanced exhumation and river-valley incision. Consequently, surface uplift of the Tibetan Plateau accompanied by valley incision in rivers of the SETP remains at the center of debate regarding the driving mechanisms of tectonic activity and climate change [67]. However, the influence of an intensified Asian monsoon would not have been restricted only to the region between the WSB and the EHS, and Quaternary glaciation has been identified in only a few locations [39, 64]. Thus, we favor tectonic deformation as the first-order control on the occurrence and distribution of enhanced Quaternary exhumation. Since the early Quaternary, the DFS has been characterized by extensional deformation accompanied by clockwise rotation [19]. A typical case is the substantial normal displacement along the eastern boundary fault of the Yulong and Haba mountains (Figures 2(a), (b), 3(b), (c)) [20, 21]. The extensional deformation since the early Quaternary has also been recorded by other river valleys in the SETP. A previous study related the enhanced exhumation of the Yalung River near the EHS to local N–S-trending extension [62]. In the Three Rivers region, the widespread occurrence of shear zones in crystalline-rock massifs was associated with later rapid cooling during late Cenozoic in response to transtensional deformation [15, 39, 49, 54, 65].

In addition, one notable characteristic of this widespread rapid Quaternary incision is its narrow geographic occurrence between 26°N and 30°N in the SETP (Figure 9(a)). It corresponds to a narrow pass that is bounded by the WSB and the EHS (Figure 9(a)). Using SIO V15.1 topographic data and EIGEN6C4 Bouguer gravity anomaly data, Hu et al. [68] showed that both the Sichuan Basin and the EHS have strong lithospheres, with a thick elastic thickness (T_e) of 50–100 km, whereas the lithosphere between them is weak, with a generally thin T_e of <15 km (Figure 9(b)), corresponding to high-conductivity, low-velocity layers [69]. Accordingly, the Sichuan Basin and EHS are inferred to have hindered or prevented the clockwise movement of crustal materials sourced from the hinterland of Tibet, which thereafter were forced to be extruded through the narrow pass between these two blocks. Inversion of Rayleigh wave dispersion and magnetotelluric imaging have together revealed the existence of two major crustal channels with low-velocity anomalies and high electrical conductivities under this narrow region (Figure 9(c)) [70, 71]. Therefore, owing to the obstruction of the stable Sichuan Basin and the EHS, the clockwise movement of crustal materials will cause contraction between the WSB and the EHS. Above the material migration in depth, the thin T_e started to fold in the form of flexural deformation (Figure 9(b)). Various geophysical data, including isostatic compensation data [72], focal mechanism data [73], GPS measurements [74], seismic anisotropy [75], and low-velocity anomalies [71] have demonstrated the existence of a geodynamic regime of E–W extension superimposed above crustal flexural deformation in the transition region between the WSB and the EHS. Therefore, the enhanced Quaternary exhumation widely recorded by large river valleys between 26°N and 30°N (Figure 9(a)) can be interpreted as resulting from middle- to upper-crustal flexural deformation. This flexural deformation is characterized by overall contractive flexure between the WSB and the EHS, with extensional faulting and rifting above it (Figure 9(b)). The flexural deformation can well explain the two pulses of deformations that widely occur along the Three Rivers fault system, the Dali fault system, the RRF system, and the Xianshuihe–Xiaojiang fault system [41, 59, 61, 63-65]: the early compressional deformation occurred in the depth, and the late extensional deformation mainly occurred in the subsurface.

We used AHe and AFT thermochronological analyses of rock samples from the deep Yangtze River valley in the Great Bend area of the SETP to infer the cooling and exhumation histories of this area. Compared with the adjacent Jianchuan Basin, our new results show a marked jump in the low-temperature thermochronological age, indicating an initial phase of rapid cooling during 11–8 Ma, the Middle–Late Miocene. Samples from the bottom of the Tiger Leaping Gorge also record a phase of rapid Quaternary cooling since ca. 1.9 Ma.

Tributaries of the Tiger Leaping Gorge record two phases of river-valley incision. The initial phase of this incision is expressed by a progressive increase in tributary channel steepness and total river-valley incision with increasing distance downstream, which is consistent with headward erosion driven by river capture. This pulse of incision corresponds to the initial phase of rapid cooling that started at ca. 17 Ma and ended at 11–8 Ma. Thus, the Great Bend was formed during this period. The second pulse of incision corresponds to rapid exhumation since the early Quaternary, and the relatively uniform k_sn and incision amounts along tributary channels confirm that the Tiger Leaping Gorge has been cut through during this period.

River capture in the area of the Great Bend during the Miocene was a result of the reactivation of the DFS, which is characterized by a series of strike-slip faults with dip-slip components. Since the early Quaternary, enhanced exhumation has occurred in the narrow region between the WSB and the EHS, implying a kinematic reversal from contractional to extensional deformation. We attribute this kinematic reversal to middle- to upper-crustal flexural deformation above the narrow pass.

The data of this study are available in the manuscript and supplementary materials.

The authors declare that they have no conflicts of interest.

This work was financially supported by the Natural Science Foundation of China (42072240; 41941016; 41830217), the Second Tibetan Plateau Scientific Expedition and Research Program (2019QZKK0901), the Key Special Project for Introduced Talents Team of the Southern Marine Science and Engineering Guangdong Laboratory (GML2019ZD0201), China Geological Survey (DD20221630), and the Basic Fund from the Key Laboratory of Deep-Earth Dynamics of Ministry of Natural Resources (JKYQN202314).

Table S1: geological information and apatite fission track data for the Jianchuan Basin. Table S2: geological information and zircon fission track data for the Jianchuan Basin. Supplementary Table S3: concavity (θ), normalized steepness (k_sn), and river incision values of 43 tributary longitudinal profiles along the Tiger Leaping Gorge. Supplementary Figure S1: geological maps of the Dali fault system and Jianchuan Basin, and geological cross sections through the Jianchuan Basin. Figure S2: slope-area data for longitudinal profiles of 43 tributaries to the Tiger Leaping Gorge. Linear regressions of slope-area data are used to determine concavity (θ) which are shown as light blue lines. A normalized steepness (k_sn) was determined over the same drainage area with a fixed concavity of 0.45 which are shown in dark blue lines. Figure S3: longitudinal profiles with projected channel segments for upper and/or middle segments indicated with a solid line, and uncertainty indicated with dashed line using Transient-Profiler [67]. Elevations of projected intersections with the Yangtze River are indicated by δz. Supplementary AFT and ZFT analytical methods.