Westward Migration of the Chenghai–Jinsha Drainage Divide and Its Implication for the Initiation of the Chenghai Fault

Abstract

The Chenghai Fault in the Chuan–Dian block terminates at the northwestern segment of the Red River Fault, and is a significant seismogenic structure. The kinematic evolution of this fault should be closely related to the regional tectonic deformation. However, it is difficult to obtain information on structural deformation of the Chenghai Fault due to the large amount of precipitation and well-developed vegetation. The Chenghai normal faulting may drive drainage reorganization in this region, which provides a new perspective for reconstructing and evaluating the tectonic history. High-resolution digital elevation models (DEM) obtained by remote sensing greatly facilitate the study of drainage evolution and active tectonics. We use two methods (χ-plot and Gilbert metrics) to measure the drainage divide stability based on the ALOS DEM (12.5 m resolution) and further reproduce the drainage evolution process in response to the asymmetric uplift by numerical modeling. The results show that the Chenghai–Jinsha drainage divide, hosted by the footwall block of the Chenghai Fault, is migrating westward (away from the Chenghai Fault) and will continue moving ~2.2–3.5 km to reach a steady state. Its migration is controlled by the Chenghai normal faulting. The Chenghai–Jinsha drainage divide formed close to the Chenghai Fault’s surface trace and continues to migrate westward in response to the asymmetric uplift. It only took a few million years for the Chenghai–Jinsha drainage divide to migrate to its current location based on the numerical modeling. The restoration of the drainage reorganization implies that the Chenghai Fault initiated in the Pliocene, which probably results from kinematic reversal along the Red River Fault.

1. Introduction

The India–Eurasia collision built the Tibetan Plateau and produced extensive intracontinental deformation in the Eurasian continent [1,2,3,4]. Although the mechanisms for plateau growth are debated [4,5,6,7,8], it is widely accepted that the southeastern margin of the Tibetan Plateau experienced southeastward motion and clockwise rotation (Figure 1) [5,9,10,11,12,13]. As consequences of this extrusion phase, the Red River Fault yielded a kinematic reversal from sinistral to dextral slip [9,14,15,16,17], and rifting occurred in the Chuan–Dian block [5,9,13,18,19]. Among them, the Chenghai Fault, located in the southwestern Chuan–Dian block, terminates at the northwestern segment of the Red River Fault. It is dominated by normal faulting with a sinistral strike–slip component. The kinematic evolution of the fault should be significantly influenced by the Red River Fault. Meanwhile, the Chenghai Fault is a significant seismogenic structure, generating the 1515 Yongsheng M 7³/₄ earthquake [20,21,22,23] and the 2001 Qina Mw 5.6 earthquake [24]. Therefore, the Chenghai Fault plays a crucial role in regional tectonic evolution and bears implications for regional deformation.

However, it is difficult to obtain the structural deformation of the Chenghai Fault, because the large amount of precipitation and well-developed vegetation hinder the maintenance of geomorphic markers and the collection of dating samples. A way of drainage divides analysis provides the potential to study the tectonic evolution of the Chenghai Fault. The high-resolution DEM obtained by remote sensing greatly facilitates the study of drainage evolution and active tectonics [25,26,27,28,29,30,31]. Across the Chenghai Fault, a large number of gullies have developed. Most of these gullies are transverse and approximately perpendicular to the fault. The east-flowing rivers in these gullies that drain into the Chenghai Lake are separated from the west-flowing tributaries of the Jinsha River by the Chenghai–Jinsha drainage divide (Figure 2A,B). A drainage divide’s stability is controlled by tectonic uplift, climatic perturbations, and lithology [32,33,34,35,36]. In general, a mountain belt has uniform rock erodibility and precipitation in a short distance. In this regard, the migration of the Chenghai–Jinsha drainage divide can potentially record the evolution process of river network systems from the onset of normal faulting, and, therefore, can be used to reconstruct the Chenghai Fault’s history [30,37,38,39,40,41,42,43].

Figure 1. (A) Schematic tectonic map of the Tibetan Plateau region (compiled from [4]). (B) Major faults in the eastern Tibetan Plateau (compiled from [44]). The light blue lines are the river systems. RRF, Red River Fault; CHF: Chenghai Fault; LMF: Longmenshan Fault; SGF: Sagaing Fault; XJF: Xiangjiang Fault; XSHF: Xianshuihe Fault.

Figure 1. (A) Schematic tectonic map of the Tibetan Plateau region (compiled from [4]). (B) Major faults in the eastern Tibetan Plateau (compiled from [44]). The light blue lines are the river systems. RRF, Red River Fault; CHF: Chenghai Fault; LMF: Longmenshan Fault; SGF: Sagaing Fault; XJF: Xiangjiang Fault; XSHF: Xianshuihe Fault.

Figure 2. Perspective views and χ map of channels for the Chenghai–Jinsha drainage divide. The location is shown in Figure 1B. (A) Perspective views of channels mapped with k_sn. Arrows indicate the migration results based on the Gilbert metrics method. (B) Map of χ and geology. Blue filling represents sedimentary rocks, yellow filling represents Quaternary sediments, and transparent filling represents igneous rocks. Arrows show the divide migration directions and cross-divide difference in normalized k_sn. (C) χ-plots for nine paired rivers across the divide. Numbers in the χ-plots are the average k_sn values. The results show that the Chenghai–Jinsha drainage divide is moving west.

Figure 2. Perspective views and χ map of channels for the Chenghai–Jinsha drainage divide. The location is shown in Figure 1B. (A) Perspective views of channels mapped with k_sn. Arrows indicate the migration results based on the Gilbert metrics method. (B) Map of χ and geology. Blue filling represents sedimentary rocks, yellow filling represents Quaternary sediments, and transparent filling represents igneous rocks. Arrows show the divide migration directions and cross-divide difference in normalized k_sn. (C) χ-plots for nine paired rivers across the divide. Numbers in the χ-plots are the average k_sn values. The results show that the Chenghai–Jinsha drainage divide is moving west.

Here, we first analyze the stability of the Chenghai–Jinsha drainage divide using two methods (χ-plot and Gilbert metrics), and calculate its future stable location. Then we use numerical modeling to reproduce the drainage evolution process in response to the normal faulting and asymmetric uplift, which constrains the initiation time of the Chenghai Fault. This study aims to provide new, independent evidence for the fluvial geomorphic evolution in response to the Chenghai Fault in southeastern Tibetan Plateau.

2. Background

The southeastern margin of the Tibetan Plateau consists mainly of the Burma block, the Indochina block, and the Chuan–Dian block of South China (Figure 1). The Chuan–Dian block is bounded to the north and east by the Xianshuihe–Xiaojiang Fault, separating it from the Bayankhara block and the rest of the South China block, respectively. The South China block, including Chuan–Dian, is separated from the Indochina block by the Red River Fault [10,45], which extends a total length of ~1000 km from the eastern Himalayan syntaxis to the South China Sea. The regional topography is characterized by a high-elevation, low-relief landscape [46,47] and is incised by the Nu, Lancang, and Jinsha Rivers and their tributaries (Figure 1).

The Chenghai Fault in the southwest Chuan–Dian block is one of the active, major faults regulating the intracontinental deformation [17,23,48]. It strikes in a nearly N–S direction, extends for more than 170 km, and terminates at the Red River Fault (Figure 1). The Chenghai Fault has strong seismic activity in the late Quaternary. According to paleoseismic and modern seismological records, it generated eight earthquakes with M ≥ 5 since the 1515 Yongsheng M 7³/₄ earthquake [20,21,22,23,49]. The 2001 Yongsheng M 6.0 earthquake was the largest earthquake along the Chenghai Fault in the past 200 years [24]. The Chenghai Fault has a composite activity manner of vertical and horizontal left-slip movement [23,49,50,51,52]. Its Quaternary activity has obvious segmentation. One proposed explanation is that the Chenghai Fault is a normal fault, only characterized by sinistral strike–slip movement along the middle segment [49,50,51,52]. However, recent studies have argued that the northern segment of the Chenghai Fault has obvious normal slip movement, while other fault segments are sinistral strike–slip, with small dip–slip components [23]. This study focuses on the northern segment of the Chenghai Fault, where normal faulting prevails in the late Cenozoic. The estimated vertical slip rate ranges from 0.5 to 4 mm/yr in the Quaternary based on geomorphological features and ¹⁴C dates [49,50,51,52].

In the northern Chenghai Fault, normal faulting resulted in the subsidence of the Chenghai Basin, and formed a sharp basin–mountain boundary [9,23,48]. The Chenghai–Jinsha drainage divide separates the rivers that flow eastward to the Chenghai Basin from the west-flowing tributaries of the Jinsha River on the west flank (Figure 2). Lithologies in the drainage basins are broadly composed of igneous rock and sedimentary rock formed in the Ordovician to the Quaternary (Figure 2B). Moreover, Pliocene and Early Pleistocene sediments, including lacustrine, fluvial, and pluvial deposits (i.e., the Xigeda Formation or Sanying Formation in this region), are distributed in the Chenghai Basin [51,52].

3. Methods

3.1. Measuring Drainage Divide Stability

Drainage divide migration is essentially driven by the cross-divide difference in erosion rate [32,33,34,35,36]. Therefore, metrics that represents such a contrast, such as the top-most slope of the χ-plot [32,36] and the Gilbert metrics [33], can be used to assess the stability of the drainage divide.

According to the detachment-limited stream power model [53], the channel gradient (S) is expressed as [54]:

$$S={({\displaystyle \frac{E}{K}})}^{{\displaystyle \frac{1}{n}}}{A}^{{\displaystyle \frac{-m}{n}}}$$

(1)

where E is erosion rate, K is erosion coefficient, A is upstream area, m is area exponent, and n is slope exponent [55,56].

The river longitudinal elevation (z) profile is given by integrating upstream ($x$) from the base level ($x_b$):

$$z\left(x\right)=z_b+{\int }_{x_b}^{x}S\left(x\right)dx$$

(2)

By combining Equations (1) and (2), a transformed river profile can be expressed as the following equations:

$$z(x)=z_b+k_{sn}{\left(A_0\right)}^{-{\displaystyle \frac{m}{n}}}\chi$$

(3)

with

$$k_{sn}={\left({\displaystyle \frac{E}{K}}\right)}^{{\displaystyle \frac{1}{n}}}=S{A}^{{\displaystyle \frac{m}{n}}}$$

(4)

and

$$\chi ={\int }_{x_b}^x{\left({\displaystyle \frac{A_0}{A\left(x\right)}}\right)}^{{\displaystyle \frac{m}{n}}}dx$$

(5)

where χ is a spatial integral of the drainage area, k_sn is steepness index, and A₀ is an arbitrary scaling area [32,57].

The χ-plot depicts the relationship between the river longitudinal elevation (z) and χ. The slope value of the χ-plot is proportional to the k_sn if A₀ is set to unity (Equation (3)), and the k_sn is proportional to the erosion rate if K and n are set to unity (Equation (4)). Therefore, comparing the top-most k_sn value (i.e., the slope of the linear or quasi-linear segments of the χ-plots) has been used to measure the drainage divide stability, assuming similar lithology and precipitation [32,33,40,58,59]. A steeper slope of the χ-plot (i.e., higher k_sn value) indicates that the drainage divide tends to migrate towards the other side [58,60].

We also adopt the “Gilbert metrics” method [33] to determine the migration direction of divide (Figure 3). This method is proposed based on Gilbert’s (1877) [61] law of unequal declivities. It assumes that divides will migrate from the side with faster erosion towards the slow erosion side. It evaluates three topographic indexes, including headwater elevation, mean hillslope relief, and mean hillslope gradient, which are topographic proxies for erosion rate in close proximity to the drainage divide. A drainage divide would migrate towards the side with a lower slope, lower relief, or higher channel head elevation [33]. Procedures to perform the calculations are implemented in the Topographic Analysis Kit [62] and DivideTools [33] based on TopoToolbox in Matlab program [63].

3.2. Estimation of the Stable Location of Drainage Divides

The future steady location of a drainage divide can be calculated by the steady-state equation of a drainage divide [36]:

$$C={\displaystyle \frac{{\int }_0^{D_{\mathsf{\beta }}-x_c^{\prime }}{\left[U_{\mathsf{\beta }}+\frac{x\prime }{D_{\mathsf{\alpha }}+D_{\mathsf{\beta }}}\left(U_{\mathsf{\alpha }}-U_{\mathsf{\beta }}\right)\right]}^{\frac{1}{n}}{\left(D_{\mathsf{\beta }}-x\prime \right)}^{\frac{-bm}{n}}dx\prime }{{\int }_0^{D_{\mathsf{\alpha }}-x_c^{\prime }}{\left[U_{\mathsf{\alpha }}-\frac{x\prime }{D_{\mathsf{\alpha }}+D_{\mathsf{\beta }}}\left(U_{\mathsf{\alpha }}-U_{\mathsf{\beta }}\right)\right]}^{\frac{1}{n}}{\left(D_{\mathsf{\alpha }}-x\prime \right)}^{\frac{-bm}{n}}dx\prime }}$$

(6)

with

$$C={({\displaystyle \frac{K_{\mathsf{\beta }}}{K_{\alpha }}})}^{{\displaystyle \frac{1}{n}}}({\displaystyle \frac{H_{\mathsf{\beta }}}{H_{\alpha }}}){({\displaystyle \frac{T_{\mathsf{\beta }}}{T_{\alpha }}})}^{{\displaystyle \frac{mb}{n}}-1}{({\displaystyle \frac{k_{\mathsf{\beta }}}{k_{\alpha }}})}^{{\displaystyle \frac{m}{n}}}$$

(7)

where C index (called the cross-divide contrast index) quantifies the relationship between the erosion coefficient (K), channel height (H), tortuosity coefficient (T), and Hack’s coefficient and exponent (k and b) [36]. D is the horizontal distance from the base level to the divide along the divide normal direction,$x_c^{\prime }$ is the headwater hillslope length, $x\prime$ is the horizontal upstream distance of a channel along the divide normal direction, and subscripts α and β represent the two sides across the divide. In this study, we designate α as the eastern side, and β the western side. We predict the final stable location of the divide based on the relationship between the uplift-rate ratio and normalized drainage divide location at steady state.

3.3. Numerical Modeling of Landscape Evolution

To reproduce the drainage evolution in response to asymmetric uplift, we employ the TopoToolbox Landscape Evolution Model [64], a MATLAB-based landscape evolution model contained in TopoToolbox 2 [63]. It only simulates the process of river incision and hillslope processes without the deposition of materials. The designed model extends 20 km long in the E–W direction and 50 km wide in the N–S direction, with a spatial resolution of 50 m. The length of 20 km corresponds to the distance from the downstream of the Chenghai River to the confluence of the tributary and main stream of the Jinsha River (Figure 2). The initial elevation is set as 100 m along the eastern edge, whereas the elevations at the other edges are fixed to 0 m. In addition, a Gaussian noise of 0–50 m is applied to initiate a random river network. The uplift rate linearly decreases from the eastern to the western edge to model the asymmetric uplift associated with fault activity. We conducted five numerical experiments with different uplift rates and rate ratios (i.e., the ratio of the uplift rates at the western and eastern edges).

Other model parameters are set as follows: erosion coefficient (K), 3 × 10⁻⁶/yr; area exponent (m), 0.5; slope exponent (n), 1; hillslope diffusivity, 0.03 m²/yr; and drainage area threshold, 0.1 km². These input parameters are guided by values used in similar studies [34,35,41,64,65]. Each model is run for 50 Myr, with a time step of 0.5 Myr.

In this study, we performed a series of numerical models to elucidate the evolution of the Chenghai–Jinsha drainage divide under an asymmetrical uplift. Due to the uncertainty of vertical slip rate (0.5–4 mm/yr [49,50,51], and the unknown uplift rate ratio across the Chenghai Fault, we performed five models with different uplift rates and rate ratios. We kept a constant uplift rate on the western edge (0 mm/yr), and assigned the uplift rate on the eastern edge with 0.5 (Model A), 1 (Model C), and 2 mm/yr (Model E). In addition, Model B and Model D were designed with an uplift rate ratio of 0.5, which were achieved by linearly decreasing the uplift rate from 0.5 and 1 mm/yr on the eastern edge to 0.25 and 0.5 mm/yr on the western edge, respectively.

4. Results

4.1. Stability of the Chenghai–Jinsha Drainage Divide

Because the lithology and precipitation are nearly identical across the Chenghai–Jinsha drainage divide over short distances (several kilometers), the results of χ-plots and Gilbert metrics both reflect the cross-divide difference in erosion rates.

We first conducted a regional comparison of the cross-divide differences in topographic features, k_sn, and χ. The satellite imagery and χ map show that the Chenghai Drainage exhibits a steeper topography, higher k_sn, and lower χ value than those in the Jinsha Drainage (Figure 2A,B). Then, we selected nine pairs of river channels across the Chenghai–Jinsha drainage divide to create the χ-plots. All the top-most reaches of the Chenghai River have greater slopes of χ-plot (i.e., k_sn) than those of the Jinsha River (Figure 2C), which shows that the Chenghai drainage divide is migrating westward. The normalized difference in k_sn (i.e., $∆k_{sn}/\overline{k_{sn}}$) is shown in Figure 2B. Here, we showed a characteristic example to demonstrate the migration pattern. We identified a tributary to the east of the divide as barbed tributary, as the west-flowing channel turns by approximately 180° and flows into the Chenghai Drainage (Figure 4A). It is a direct diagnostic character for river capture [66,67,68], which means the Chenghai Drainage must have just captured a small area of the headwaters of the Jinsha Drainage. Geometric features of the captured area have not adjusted to the Chenghai Drainage, but are still similar to those in the Jinsha Drainage (Figure 4A). The χ-plot of a river pair close to the capture point also signifies a drainage area gain by river capture (Figure 4B) [32]. Although the Jinsha River and Chenghai River have similar and lower slope (k_sn) at the top reaches, the Chenghai River shows a high slope below the top reach. It implies a drainage area gain for the Chenghai River (Figure 4B).

We further adopt Gilbert metrics method to analyze the stability of the Chenghai–Jinsha drainage divide. This drainage divide is divided into three segments (ab, bc, and cd in Figure 2A), mainly based on the morphological features of the drainage basin. In this study, we used a reference drainage area of 10⁵ m² to calculate the Gilbert metrics. The channel head relief, elevation, and gradient values on either side of a divide were visualized as individual divide sections by histograms, which shows the Chenghai Drainage has a greater relief, lower elevation of the channel, and steeper hillslope gradient than that of the Jinsha Drainage (Figure 5A). Meanwhile, the standardized deviation of those metrics for the sub-segment along the divide reveal a consistent, uniform westward movement of the entire drainage divide (Figure 5B).

4.2. Future Stable Location of the Chenghai–Jinsha Drainage Divide

We applied the steady-state equation of drainage divides [36] to predict the future steady location of the Chenghai–Jinsha drainage divide. We selected a pair of representative rivers across the drainage divide flowing separately into the Jinsha and Chenghai Drainage (Figure 6A). Along the river channel in the Jinsha Drainage, a significant knickpoint separates the low channel gradient upstream from the steep downstream (Figure 2C). In addition, the Chenghai Fault cuts the river channel in the Chenghai Drainage. We thus chose the knickpoint and normal fault as the base level of the river channels on the west and east side, respectively. To the first order, we assumed that the lithology between the two base points is uniform, and the uplift rate varies along a linear gradient.

We first measured the elevation difference from the headwater to the base level, where H_α is 708 m and H_β is 474 m. Then, the channel tortuosity coefficients on the east side (T_α = 1.05) and west side (T_β = 1.23) were calculated based on the ratio between the horizontal distance along the channel (dx_α = 2145 m, dx_β = 11,537 m) and that along the divide normal direction (dx′_α =2045 m, dx′_β = 9367 m). By fitting the drainage area and channel length, the Hack’s coefficient (k_α = 0.35, k_β = 1.01) and exponent (b = 1.7) were determined (Figure 6B) [36,69]. We further calculated the C value (C = 1.04) assuming a uniform erosion coefficient ($K_{\beta }/K_{\alpha }=1$) because of the similar lithology and precipitation across the divide. The parameters are detailed in

Table 1. The parameters of typical rivers across the Chenghai–Jinsha drainage divide used for calculating the future stable location. It should be noted that all the parameters are from present topography, which may different from those in the final steady state.

Hα (m)	Hβ (m)	xα (m)	xβ (m)	$$\begin{gathered} {\mathit{D}}_{\mathit{\alpha }}-{\mathit{x}}_{\mathit{c}}^{\prime } \\ \left(m\right) \end{gathered}$$	$$\begin{gathered} {\mathit{D}}_{\mathit{\beta }}-{\mathit{x}}_{\mathit{c}}^{\prime } \\ \left(m\right) \end{gathered}$$	Tα	Tβ	kα	kβ	b	C
708	474	2145	11,537	2045	9367	1.05	1.23	0.35	1.01	1.7	1.04

. Accordingly, we plotted the relationship diagram between the normalized drainage divide location ($D_{\beta } ⁄ \left(D_{\alpha }+{ D}_{\beta }\right)$) and uplift rate ratio ($U_{\beta }/U_{\alpha }$) (Figure 6C). Considering a relatively short distance from the selected base level in the Jinsha Drainage to the fault, the uplift rate ratio ($U_{\beta }/U_{\alpha }$) may fall in a wide range of 0.5–1.0. With that, the normalized drainage divide location ($D_{\mathsf{\beta }} ⁄ (D_{\alpha }+{ D}_{\beta }\left)\right)$ value was determined as 64%–52%, where the drainage divide attains steady state. Based on the present normalized location of the Chenghai–Jinsha drainage divide (~84%), we predict that the divide will achieve steady state after migrating for ~2.2–3.5 km to the west, given all geological conditions remain unchanged.

4.3. Numerical Modeling on Landscape Evolution

Figure 7 shows selected representative snapshots, which provides an overview of the evolution process. When the asymmetric uplift began, the divides in all models occurred at the eastern edge and started to move westward until reaching a steady state. The detailed evolution process of an example (Model C) is recorded in Supplementary Movie S1. The difference lies in the rate of divide migration. The present Chenghai–Jinsha drainage divide is ~2–3 km away from the Chenghai Fault. To obtain the timescale for the divide to reach its current location, we plotted the position of the divide at each moment and recorded the time (Figure 7). In the cases of an uplift rate ratio of 0, the timescale decreases as the uplift rate on the eastern edge increases. When the uplift rate is 2 mm/yr (Model e), it takes 3.5 Myr for the divide to migrate 2.5 km (a simple average estimate) to the west (Figure 7E). Even at a low uplift rate of 0.5 mm/yr (Model A), this timescale is only 9 Myr (Figure 7A). Moreover, a higher uplift rate ratio accelerates divide migration. For instance, when the uplift rates on the eastern edge remain constant, the timescale of the divide to its current location is shortened in both Model B and Model D (Figure 7B,D), as compared to Model A and Model C (Figure 7A,C), respectively. After 23 Myr of asymmetric uplift, the drainage divide in each model has migrated westward by more than 5 km (Figure 7), which is much greater than the migration distance of the present Chenghai–Jinsha drainage divide.

5. Discussion

5.1. Drainage Reorganization around the Chenghai Fault

Based on a positive, monotonic relationship between the k_sn and erosion rate [55], the distribution of k_sn indicates an asymmetric uplift pattern around the Chenghai Fault (Figure 2A). The uplift rate (and k_sn) is higher east of the divide, and then gradually decreases westward from the fault. Accordingly, the east side of the Chenghai–Jinsha divide has a steeper topography than that in the west side (Figure 6D). Detailed field investigations and ¹⁴C dates suggested that the northern segment has the highest vertical slip rate along the Chenghai Fault, which is ~0.6 mm/yr and possibly as high as 4 mm/yr [49,50,51,52]. Meanwhile, the largest earthquake (the 1515 Yongsheng M 7³/₄ earthquake) in the region occurred the norther segment of the Chenghai Fault. Therefore, the Chenghai Fault results in the regional asymmetric uplift.

Drainage divide stability is fundamentally controlled by tectonics, lithology, and climate [32,34,39,70]. When rock erodibility and precipitation are homogeneous across the divide, tectonic activities will lead to drainage reorganization and divide migration [39,46,71]. In many natural cases where a mountain range undergoes asymmetric uplift, the drainage divide that it hosts migrates towards the side with a high uplift rate [30,32,33,38,40,72]. Nevertheless, the Chenghai–Jinsha drainage divide is migrating away from the high-uplift-rate side and towards the geometric center. This suggests that the initial topography and drainage network before the onset of the Chenghai normal fault should be considered to understand the modern-day geomorphic patterns [41].

We propose a hypothesis for the drainage reorganization and divide migration associated with the development of the northern Chenghai Fault (Figure 8). Before the initiation of the Chenghai Fault, the topography may be tended gently towards the west, facilitating the development of the west-flowing rivers (Figure 8A). This is supported by several geomorphic features. First, the topography width profile along the drainage divide has revealed multiple wind gaps, which imply the occurrence of river captures (Figure 8D) [41,67]. Also, the satellite imagery and χ-plot show that some west-flowing rivers are being reversed to flow eastward (Figure 4). Furthermore, a wide basin is distributed at the source of the tributary of Jinsha River (i.e., the beheaded river in Figure 4), demonstrating that a river capture event must have occurred here [46,73,74].

Following the initiation of the Chenghai normal faulting, E–W-directed extension resulted in the subsidence of the Chenghai Basin (Figure 8B) [23,48]. The Chenghai Fault is the boundary between the basin and mountain in this region. Due to the normal faulting, the Chenghai–Jinsha drainage divide emerged in the footwall block of the Chenghai Fault, and the topography to the east of the drainage divide was characterized by high relief and steep slope. The Chenghai Basin in the hanging-wall block of the Chenghai Fault has been filled by ~1800 m of sediment since the Neogene [51]. As the river systems evolved, the higher erosion rate on the east side drove the drainage divide to migrate westward and continues to the present (Figure 8B). In the future, the divide will keep migrating westward for ~2.2–3.5 km to reach a steady state, assuming all conditions remain unchanged.

5.2. Implications for the Initiation of the Chenghai Fault

Our newly revealed drainage divide stability and numerical modeling results offer an independent constraint on the tectonic evolution of the Chenghai Fault. The current Chenghai–Jinsha drainage divide has migrated westward by ~2–3 km since the onset of the Chenghai normal faulting (Figure 2A and Figure 6A). Such a short migration distance indicates the young age of the drainage divide. Meanwhile, our numerical modeling results further demonstrate that it took only a few million years for the divide to migrate to the current location (Figure 6). The timescale decreases as the uplift rate on the eastern edge and the uplift radio increase. Even in the case with a low uplift rate of 0.5 mm/yr and uplift rate ratio of 0, the divide migration to ~2.5 km would have taken only 9 Myr. Considering the complex active fault systems in the southeastern Tibetan Plateau and a relatively short distance from the eastern edge, the western edge may also be affected by the vertical uplift, thus resulting in a larger uplift rate ratio (i.e., $U_{\beta }/U_{\alpha }$). This would further shorten the time required for divide migration (Figure 6). In another example, the Xizhou Shan of the Shanxi Rift in North China, the newly emerged main drainage divide has migrated ~3–4 km away from the fault zone over the past ~7 Myr based on the numerical modeling [41]. Moreover, the Chenghai Basin has preserved fluvial–lacustrine facies sedimentary deposits from the Pliocene and Early Pleistocene (i.e., the Xigeda Formation or Sanying Formation in this region), indicating the initiation of the Chenghai Fault during this period [51,52]. Therefore, all these results support that the Chenghai Fault initiated in the Pliocene.

The younger activity of the Chenghai Fault provides new insights into the regional deformation. Affected by the southeastward motion and clockwise rotation of the southeastern margin of the Tibetan Plateau, the fault systems in the southeastern plateau have undergone extensive kinematic reversal [5,9,10,11,12,13] (Figure 9). After the Red River Fault turned into dextral shear, the northern part of the Chuan–Dian block entered the shear-extensional deformation stage (Figure 9). The Chenghai Fault has also initiated extensional deformation, which constitutes the basin–mountain boundary in this region. Right-lateral faulting on the Red River Fault and normal faulting in the northern Chuan–Dian block would be the consequences of the present extrusion phase (Figure 9) [5]. The younger activity of the Chenghai Fault supports a later onset of kinematic reversal along the Red River Fault. As the structural deformation propagates towards the central part of the Chuan–Dian block, the Chenghai Fault is controlled by two boundary faults and exhibits sinistral strike slip movement [23]. Therefore, the Chenghai Fault should be strongly active, and has strong seismic activity in the late Quaternary.

6. Conclusions

We use χ-plot and Gilbert metrics methods to analyze the stability of the Chenghai–Jinsha drainage divide at the northwest Chuan–Dian block. The results show that this divide is migrating westward away from the Chenghai Fault and will further migrate ~2.2–3.5 km westward to reach a steady state, given all geological conditions remain unchanged. The migration of the Chenghai–Jinsha drainage divide is driven by the initiation of the normal faulting along the Chenghai Fault. Normal faulting controls the subsidence of the Chenghai Basin, causing the west-flowing rivers to reverse their flow towards the Chenghai Basin. The Chenghai–Jinsha drainage divide first occurred at the eastern edge and migrated towards the geometric center. Our numerical modeling results suggest that it took a few million years for the Chenghai–Jinsha drainage divide to migrate to its current location. It implies that the Chenghai Fault initiated in the Pliocene.