Study on the Identification, Failure Mode, and Spatial Distribution of Bank Collapses after the Initial Impoundment in the Head Section of Baihetan Reservoir in Jinsha River, China

Abstract

After the initial impoundment of the Baihetan Reservoir in April 2021, the water level in front of the dam rose about 200 m. The mechanical properties and effects of the bank slopes in the reservoir area changed significantly, resulting in many bank collapses. This study systematically analyzed the bank slope of the head section of the reservoir, spanning 30 km from the dam to Baihetan Bridge, through a comprehensive investigation conducted after the initial impoundment. The analysis utilized UAV flights and ground surveys to interpret the bank slope’s distribution characteristics and failure patterns. A total of 276 bank collapses were recorded, with a geohazard development density of 4.6/km. The slope gradient of 26% of the collapsed banks experienced an increase ranging from 5 to 20° after impoundment, whereas the remaining sites’ inclines remained unchanged. According to the combination of lithology and movement mode, the bank failure mode is divided into six types, which are the surface erosion type, surface collapse type, surface slide type, bedding slip type of clastic rock, toppling type of clastic rock, and cavity corrosion type of carbonate rock. It was found that the collapsed banks in the reservoir area of 85% developed in the reactivation of old landslide deposits, while 15% in the clastic and carbonate rock. This study offers guidance for the next phase of bank collapse regulations and future geohazards prevention strategies in the Baihetan Reservoir area.

1. Introduction

Reservoir impoundment induces numerous bank collapse geohazards [1,2,3], which lead to bank retreat, siltation of the river channel, and restriction of ship navigation, seriously threatening the safety of people’s lives and properties. The Jinsha River is situated in the upper reaches of the Yangtze River, spanning a length of 2360 km and featuring an impressive water drop of up to 5100 m. It boasts abundant hydraulic resources. Baihetan Hydroelectric Power Station, located downstream on the Jinsha River, is the world’s third largest power station after China’s Three Gorges Hydroelectric Power Station and Brazil-Paraguay’s Itaipu Hydroelectric Power Station. The area has extensive background geohazards due to steep bank slopes, significant vertical height differences, and strong tectonic activities [4,5,6]. It was observed during the impoundment period that a bank collapse occurred in Dawanzi, 22 km from the dam, and severe tunnel deformation occurred [7]. The primary factor contributing to this bank collapse was the fluctuation of reservoir water levels. Therefore, strengthening the systematic investigation, type division, and law analysis of bank collapses during reservoir impoundment is a crucial basis for predicting the extension of bank failures into landslide occurrence and mitigating such geological hazards.

The deformation and damage of bank slopes caused by reservoir impoundment mainly include landslides, avalanches, debris flows, ground cracks, ground subsidence, and bank collapse [8,9]. Among them, bank collapse is a type of geohazard that develops more widely in the initial impoundment, and its severity is reflected in reservoir bank reconstruction [10]. The densely developed collapsed banks resulted in severe socioeconomic problems for the settlements, roads, bridges, harbors, and other buildings. Bank collapse, located at the front of the bank slope, is an obvious sign of the destructive deformation of a reservoir landslide. Previous studies on bank failure mainly focused on the Three Gorges Reservoir and some other small reservoirs, including types of bank collapse, the parameters of bank failure, and the prediction methods [11,12,13]. Thorne et al. (1988) and Osman et al. (1988) state that riverbank retreat is a complicated geological phenomenon that depends on the type of bank materials and the processes and mechanisms of erosion [14,15]. Also, vegetation can increase or decrease the bank stability [14,15]. Tang Minggao et al. (2006, 2007) classified five types of bank collapse according to the field investigation and summarized the evolution modes of their bank collapses [10,16]. Xu Qiang et al. (2009) proposed a bank collapse prediction method suitable for mountain river reservoirs: the Bank Slope Structural Prediction Method (BSSPM), which provides a good prediction model for bank collapse [17].

UAV survey technology is characterized by high measurement accuracy, flexibility, and efficiency. A series of digital products such as orthophoto maps, digital elevation models, and three-dimensional geographic models can be generated and constructed based on drone aerial photography, which has unique advantages with the widespread use of identification, monitoring, and prevention of geological hazards [18,19,20]. UAV aerial survey technology is suitable for large-scale terrain mapping and can reach areas inaccessible by the ground. It provides high-resolution image data that enables professional geologists to accurately identify morphological features and signs of deformation in geohazardous bodies [21]. Therefore, collapsed banks and cracks can be well identified using UAV aerial survey technology [22], thus overcoming the shortcomings of small areas of collapsed banks in reservoirs that are challenging to investigate.

Controlled by factors such as numerous collapsed banks, small scale and rapid change, poor transportation conditions, and the fact that it is not easy to investigate the collapsed bank disasters accurately, previous studies on bank collapse focused on point-to-point investigation and lack of overall analysis of the collapse disaster law of a reservoir bank. Therefore, analyzing the collapsed bank types, failure patterns, and spatial distribution in the reservoir area must be supplemented. The innovation of this study lies in its utilization of the UAV flight method to systematically investigate the bank collapse of Baihetan Reservoir, as opposed to the traditional approach used in conventional bank collapse investigations. This novel methodology contributes to a more comprehensive understanding of regional bank collapse phenomena.

To monitor the slope evolution trend of the bank collapse and to systematically understand the types, the failure mode, and development law of the bank collapse, this paper takes the 30 km-long canyon section of the head of the Baihetan reservoir from the dam to Baihetan Bridge as the study area. For the impact area of the bank collapse caused by impoundment, high-resolution images generated by UAV aerial photography were used for interpretation, and combined with the investigated stratigraphic lithology and bank slope structure, and the bank collapse types and spatial development laws in the reservoir area were summarized. The UAV data can guarantee the precision of the bank collapse interpretation, thereby enabling the conclusion that contributes to ensuring the safety of geological formations near the reservoir water line.

2. Geological Setting

The Baihetan Reservoir, located on the southeastern edge of the Qinghai–Tibetan Plateau, features an alpine canyon geomorphology. The region consists of high mountains with an altitude of 3000~4000 m and “V” shaped canyons with an altitude of 500~1000 m (Figure 1a,b). The bank slope in the reservoir area has a minimum gradient of 5°, a maximum gradient of 43°, and an average gradient of 28°. The reservoir area is dry and has a hot valley climate with significant vertical variation. Annual precipitation reaches 500–800 mm, with the rainy season (May–October) concentrating 86% of the rainfall (Figure 1c). Baihetan Reservoir was impounded in April 2021 with an initial level of 630 m. In October 2021, the water location height reached 816 m and fluctuated, rising to a maximum level of 825 m in late October 2022. The maximum water level of the reservoir in front of the dam was raised by about 200 m, and the water level varied by 60 m during the design operation (Figure 1c). Baihetan Reservoir is tectonically located in the Sichuan Yunnan Diamond Block of China. The major tectonic structures in the area are the SN-trending Xiaojiang faults, the NNW-trending Daliangshan faults, the NW-trending Zemu River faults, and the NNE-trending Lianfeng faults (Figure 2). Controlled by the NE-trending and NW-trending fault structures, the Jinsha River meanders in this section, with the overall flow direction in the SN direction.

The reservoir area has a diverse and extensive stratigraphic lithology, including sedimentary, metamorphic, and magmatic rocks [23]. Reactivation of the old landslide deposit group, carbonate rock group, clastic rock group, metamorphic rock group, and massive basalt rock group are predominantly distributed within the study area. The reactivation of the old landslide deposit group is predominantly Quaternary (Q) residuum, avalanche slope deposits, alluvial deposits, and proluvial deposits. The carbonate rock group is mainly dolomite and limestone assemblages of the simian (Z_bd), the Cambrian (Є ₃e), Ordovician (O₂d), and Permian system (P₁). The clastic rock group is mainly sandstone, shale and mudstone assemblages of the simian (Z_bd, Z_bl), Cambrian (Є₁, Є₂x), Ordovician (O₂q, O₁h), Silurian (S₂s), Devonian (D₂y) Triassic (Td), and Cretaceous systems (K₁x¹). The massive basalt rock group is basalt of the Upper Permian Basalt system (P₂β). The metamorphic group is mainly an assemblage of phyllite, metamorphosed sandstone and slate of the Great Wall System (Pt₁t). In terms of stratigraphic lithological characteristics, the massive basalt rock group is structurally intact, the carbonate and clastic groups are consolidated and contacted in some stratigraphic sections, and the metamorphic rocks, such as phyllite rocks and slate, are knit and crushed loosely by the influence of active faults (Figure 2) [23].

The main factors controlling the stability of bank slopes are the lithologic combination characteristics and the structural types of the bank slope. The combination of both constitutes the basic geological pattern of the slope [24,25,26,27]. Based on the characteristics of the lithological assemblage of the bank slopes, the bank slopes are classified into bedrock bank slopes and soil slopes (

Table 1. Structure division of rock and soil on bank slope.

Rock				Deposit
Angle between the discontinuity and the slope aspect			Basalt structure (IV)	Reactivation of old landslide deposit (V)
0–45°	45–135°	135–180°
Cataclinal slope (I)	Orthclinal slope (II)	Anaclinal slope (III)

). For the bedrock bank slope, the structural type reflects the relationship between the occurrence of the primary stratified structural plane and the bank slope, which controls the deformation and failure mode of the bank slope. According to the angle between slope direction and rock strata inclination, the bedrock bank slope can be divided into the cataclinal slope (0~45°), orthclinal slope (45~135°), and anaclinal slope (135~180°) [24,25,28]. In addition, there are massive basalt structure bank slopes composed of basalt, which are more stable. As for the deposit bank slope, Quaternary (Q) residuum, avalanche slope deposits, alluvial deposits, and proluvial deposits are uniformly divided into the reactivation of old landslide deposit slopes (Figure 3).

3. Data and Methodology

The DOM and DSM before impoundment are 6 m resolution remote sensing image data acquired by Spot, which are used to extract topographic information before impoundment, including slope gradient, profile, slope direction, and other topographic factors, as well as precise topographic data for UAV Terrain Following Photogrammetry.

The UAV survey can obtain accurate DOM and DSM of the bank slopes during the bank collapse. We conducted a photogrammetric survey of the collapsed bank in the reservoir area in May 2022 using a Shenzhen, China-based DJI vertical take-off and landing rotor drone, the Matrice 300 RTK, which is hot-pluggable, has a long endurance and no control points, enabling efficient aerial photography of the reservoir area. A ZENMUSE P1 35 mm focal length lens was used, and the specific UAV aerial flight parameters and camera parameters are shown in

Table 2. Information for the UAV flight parameters.

UAV Type	RGB Camera	Sensor Size	Image Size	Pixel Size	Focal Length
DJI M300RTK	ZENMUSE P1	35.9 × 24 mm	8192 × 5460 px	4.4 μm	35 mm
Front Overlap Rate	Slide Overlap Rate	Relative Flying Height	Flight Mode	Dom Resolution	Dem Resolution
85%	65%	500 m	Terrain following photogrammetry	0.06 m	0.06 m

. The UAV aerial survey area adopts a strip survey area, 30 km for strip flight and 60 km for both sides of the bank. The front overlap rate and side overlap rate of the UAV survey are 85% and 65%. Aiming to obtain the morphology of the collapsed bank in the vicinity of the impoundment line, the relative height of the UAV is 500 m, so that not only can it guarantee a certain amount of GSD, but it can also cover 250 m on both sides of the impoundment line. DJI Smartmap software generated DOM, DSM, and 3D models of the reservoir area with a resolution of 6 cm. The combination of these three data can explain the collapsed bank’s distribution and geometric parameters in the head section of the reservoir during the falling water level stage after the initial impoundment, which was sufficiently applicable for interpreting the collapsed bank.

Based on the field survey and 1:200,000 geological map [29], the study area’s stratigraphic lithology, fault properties, and bank slope structure were determined and updated.

4. Results and Analysis

4.1. Bank Collapse Interpretation Criterion

The collapsed bank has obvious morphological signs and structural features. The trailing edge of the collapsed bank is curved, the overall bottom is enormous, and the top is small. The collapsed bank differs from the surrounding geological body in color tone, texture, vegetation development, and growth condition. According to texture characteristics, the collapsed bank has a more apparent linear texture due to its movement being perpendicular to the contour, and the slower-moving collapsed bank develops tensile cracks after sliding. The profile pattern of the collapsed bank shows a concave shape in micromorphology. According to the Varnes classification standard for landslides, the collapsed bank identified in the reservoir area is classified based on the composition of the bank slope material and the mode of movement [30]. We interpreted the distribution map of the collapsed bank at the head of the reservoir using high-resolution UAV aerial remote sensing data and the characteristics of the collapsed bank (Figure 4a). According to the type of geotechnical body and bank slope structure, we classified the collapsed bank into the following six types, namely, surface erosion type, surface collapse type, surface slide type, bedding slip type, toppling type, cavity corrosion type, as shown in

Table 3. The types and characteristics of bank collapse in the study area.

Material		Characteristics	Type	Code
Deposit: reactivation of old landslide deposit		Erosion of weathered deposit	Surface erosion type	A
		The profile pattern shows two sections, steep at the top and slow at the bottom	Surface collapse type	D
		Surface develops tension cracks	Surface slide type	C
Rock	Clastic rock	Exposure of smooth rock layer faces	Bedding slip type	F
		Rock toppling	Toppling type	B
	Carbonate rock	Cavity development	Cavity corrosion type	E

.

Surface erosion type, surface collapse type, and surface slide type—the lithology of these three types of bank collapse materials is mainly the reactivation of old landslide deposit, whose composition is related to the mother rock. The lithology of the bedding slip-type and toppling-type bank collapse material is mainly clastic rock with an interbedded combination of soft and hard rock. The lithology of the cavity corrosion-type bank collapse material consists mainly of carbonate rocks composed of argillaceous limestone, sandy limestone, and dolomite. The main differences between these types are as follows:

(1) Surface erosion type of bank collapse: Geomorphologically, the reservoir water stripped away the surface soil bodies. The color, texture, and vegetation development of the bank’s geological body on optical remote sensing are missing from the surrounding area. This type of bank failure is named the surface erosion bank collapse (Figure 4b).

(2) Toppling type of bank collapse: Geomorphologically, the bank failure area is on an isolated island surrounded by water on two sides. Bedrock outcrops can be visible on optical remote sensing, and the slope structure investigated in the field is an anaclinal or orthoclinal slope. This type of bank collapse is named the toppling type (Figure 4c).

(3) Surface slide type of bank collapse: Geomorphologically, the boundary between the collapsed and uncollapsed areas is clear, and the topography of the collapsed area becomes steeper than the original. On optical remote sensing, the back edge and cracks of the collapsed bank are clear. This type of bank collapse is called the surface slide type (Figure 4d).

(4) Surface collapse type of bank collapse: Geomorphologically, the upper part of the collapsed bank shows negative topography, the lower part of the slope becomes steeper than the original topography, and the profile morphology shows a zigzag shape. The clear linear texture in the direction of the collapsed bank is essentially parallel to the contours on optical remote sensing. This type of bank is called the surface collapse type (Figure 4e).

(5) Cavity corrosion type of bank collapse: Geomorphologically, the cavity appears in the shaded area. The body of the concave can be visible on optical remote sensing. This type of bank collapse is called the cavity corrosion type (Figure 4f).

(6) Bedding slip type of bank collapse: Geomorphologically, the longitudinal slopes of the bank collapses are jagged and steep. Bedrock outcrops can be visible on optical remote sensing. The plane shows a narrow area of collapsed bank. The structural survey of the slope in the field is a cataclinal slope, and this type of bank collapse is called the bedding slip type (Figure 4g).

4.2. Bank Collapse Modes

(1) Surface erosion type:

The surface erosion type bank collapse refers to the reservoir bank deformation damage that occurs as a reservoir bank slope geohazards. Under the action of reservoir water, surface water, wind, wave scour, and other external agent forces, the bank slope material is gradually washed, abraded, and then carried away (Figure 5a). The main influencing factor is the fluctuation of the water level of the reservoir. The water level of the Jinsha River was around 650 m before the storage operation of the Baihetan Reservoir, after which the water level fluctuated between 780 m and 825 m. Before impoundment, the engineering geological rock group on the surface of the collapsed bank consisted of the reactivation of old landslide deposit, and the slope had been largely stabilized. After impoundment, the surface residual slope deposit softened under alternating wet and dry conditions, exposing the fresh surface, and slope sliding occurs after losing fundamental support. The post-slip slope gradient was similar to the pre-slip slope, and the surface unstable material piled below the water surface (Figure 5b). The size of the surface erosion-type collapsed bank ranges from 84 to 12,495 m² and is primarily curved and pear-shaped. This type of bank collapse is generally not hazardous but is widely distributed on both sides of the reservoir. It can occur in any reservoir slope structure, where some unstable slopes are damaged by surface denudation under fluctuating reservoir water levels. Surface erosion type can be transformed into surface collapse type and surface slip type through reservoir water level fluctuation.

(2) Surface collapse type:

The surface collapse type refers to the reservoir bank slope geohazards that occur near the water level line. Under the long-term action of the reservoir water, the base of the foot of the slope is softened and eroded, the upper material of the bank slope loses its balance, the local downward error collapses, the back wall shows vertical or negative angle, and the river water gradually carries away the damaged material. The surface collapse type is another type of geological hazard that causes damage to banks (Figure 6a). Before impoundment, the engineering geological rock group of the slope consisted of the reactivation of old landslide deposit and the slopes were stable. After impoundment, the groundwater level rises, the reservoir water softens the bottom pile body while entering the fissure, the slope residual deposits are loosened and weakened at the bottom pile body under the action of the fluctuation of the reservoir water level, and the upper pile body experiences a rapid collapse in the absence of supporting force. The surface collapse type is a soil-type collapsed bank with a size ranging from 341 to 23,201 m², showing a circling chair-like geomorphology. The surface collapse type tends to occur at locations where material changes and groundwater occur, such as at the junction of permeable and impermeable layers and coarse-grained and fine-grained soils (Figure 6b). Therefore, this type of collapsed bank is characterized by suddenness and rapid change and is prone to pose a threat to residential buildings.

(3) Surface slide type:

The surface slide type is a bank failure in which the rock and soil mass of a bank slope slips along predominant discontinuity, and a tensile crack appears at the top of the bank and gradually develops into a through-slip surface along the tensile crack (Figure 7a). Before impoundment, the engineering geological rock groups of the bank collapses consisted of mainly sandstone, shale, mudstone, and their reactivation of old landslide deposit. Most of these weathering deposits are farmland, so they are subject to significant factors of artificial disturbance. After impoundment, the reservoir water softened the loose accumulations, tensile and shear cracks appeared at the top and waist of the collapsed bank. Under the influence of reservoir water fluctuation, the collapsed bank is prone to slip on the weak structural surface. When the water enters the cracks, a slip-type collapsed bank geohazards will occur (Figure 7b). Surface slide types mainly target residual slope deposits, with sizes ranging from 516 to 82,026 m² and profile morphology showing arcuate sliding.

(4) Bedding slip type:

The bedding slip type is an unstable bank slope where the rock slope slides along the weak structural surface under the action of reservoir water fluctuation (Figure 8a). This is a type of bank failure hazard controlled by the structural surface. In this type of bank failure, the slope aspect is the same as the inclination of the rock discontinuity, and the angle between the slope aspect and the inclination of the rock discontinuity lies between 0° and 45°. Before impoundment, the lithological composition of the type was mainly layered interbedded sandstone, siltstone, shale, and mudstone, with the surface of the bank slopes covered by thin accumulations, and the bank slopes were stable. After impoundment, the surface residual slope deposits were subjected to slips and slides by the fluctuation of the reservoir water, which entered the bank slopes along the rock discontinuity surfaces, leading to the deterioration and deformation of the rock layers (Figure 8b). The relatively low shear strength of siltstone interbedded with argillaceous siltstone controls the gravitational dislocation along the planar surface of the bedding plane, finally leading to shear-off behavior within the relatively hard rock stratum at the slope foot. The largest bank collapse area in the reservoir’s head section is the bedding slip type of clastic rock. The size ranges from 2368 to 88,383 m², and the area varies greatly. The post-slip profile morphology is jagged, and the length-to-height ratio of the collapsed bank is close to 3:1.

(5) Toppling type:

The toppling type of clastic rock refers to the toppling deformation of rocky bank slopes mainly along the joint surface under the action of reservoir fluctuation, its gravity, and the alternation of dry and wet environments over a long period. It is a type of bank failure hazard controlled by the structural surface (Figure 9a). This type of bank collapse is divided into anaclinal toppling and orthoclinal toppling. The main difference is the range of angles between the slope orientation of the bank and the inclination of the rock discontinuity. The orthoclinal toppling type has an angle between 45° and 135° and the anaclinal toppling type has an angle between 135° and 180°. Before impoundment, the lithological composition was predominantly clastic rocks of layered sandstone composition. The surface weathering is severe, and the joint fissure develops. After impoundment, the reservoir water carries and scours the surface layer of loose material, exposing the reverse and orthogonal rock layers. After the bottom material is hollowed out, the rock layer is tilted along the weak structural surface. The upper rock layer is tilted and deformed along the structural surface under the action of its gravity (Figure 9b). The size of the toppling collapsed bank ranges from 458 to 20,519 m², the clastic group is denuded along the original profile line, and the profile is close to the original profile.

(6) Cavity corrosion type:

The carbonate cavity corrosion type refers to the rock slopes dissolved to form concave cavities of different sizes under the fluctuation of the reservoir water. Ultimately, the bank along the cavity of the weak structural surfaces continued to expand and develop into an unstable bank slope of the bank collapse geohazards (Figure 10a). Before impoundment, the rock group of the bank slope was predominantly carbonate rocks consisting of limestone, dolomite, and muddy limestone, with a small amount of weathered deposits covered on the bank slopes. After impoundment, the reservoir water scours and erodes the surface residual deposits and constantly dissolves carbonate rock materials, forming the uneven size of the concave rock cavity. Under the long-term immersion of reservoir water and the action of waves, the carbonate rocks form dissolution channels and lateral extension fractures, during which the reservoir water and waves will carry detrital materials away from the cavity, making the cavity continue to expand. The size of the collapsed bank ranged from 1389 to 12,000 m², and the top of the profile pattern was similar to the surface collapse type (Figure 10b). The bottom of the profile was concave towards the slope, where the collapsed bank would be subjected to erosion and damage along the concave and turning position.

4.3. Bank Collapse Distribution Laws

From Dazhai Town to Hulukou Town, the total length of the head section of the Baihetan Reservoir is 30 km. Because the topography of the reservoir area cuts to form deep gullies, these areas turn into branch gullies when the water level rises. There are sixteen branch ditches in the study area, including six branch ditches on the left bank and ten branch ditches on the right bank. Influenced by topography and geomorphology, stratigraphic lithology, reservoir water fluctuation, and human activities, bank collapses developed during the initial impoundment, and the distribution of bank collapses has the following characteristics.

According to the bank survey and interpretation results, from the reservoir impoundment in October 2021 to August 2022, 276 collapsed banks were developed in the head section of the Baihetan Reservoir. The left bank developed 141 collapsed banks, the right bank developed 135 collapsed banks, the left bank of the tributary gully developed 56 collapsed banks, the right bank of the tributary gully developed 67 collapsed banks, and the left bank of the mainstem developed 85 collapsed banks. The right bank of the mainstem developed 68 collapsed banks (Figure 11a). The development of weathered deposits in the branch gully is an area where bank collapses developed intensively. The development density of the 30 km long collapsed bank was 4.6 sites/km, and the development density of the left and right bank collapsed banks was 4.7 sites/km and 4.5 sites/km.

In the reservoir’s head section, there are 234 deposit-type and 42 rock-type collapses, accounting for 85 percent and 15 percent of the total collapses, respectively (Figure 11b). The most developed type is the surface erosion type, which accounts for 48 percent of the total number of collapsed banks, and the least developed type is the cavity corrosion type, which accounts for 2 percent of the total number of collapsed banks (Figure 11c).

The area of the collapsed bank is the most obvious manifestation of the collapsed bank after the initial impoundment, ranging from 117 to 88,383 m². The total area of bank collapse hazards in the head section of the reservoir was 1.1 km² (

Table 4. Parameters of bank collapse area in the study area.

Bank Collapse Type	Number of Bank Collapse	Areas of Mapped Bank Collapse (m2)
		Minimum	Maximum	Average	Total
Surface erosion type	133	117	11,060	2115	281,304
Surface collapse type	57	341	23,201	3354	191,184
Surface slide type	44	517	82,026	8550	376,189
Bedding slip type	13	2368	88,383	11,166	145,159
Toppling type	22	458	20,520	3615	79,528
Cavity corrosion type	7	1389	12,000	4679	32,752

), of which the area occupied by the branch bank collapse hazards was 0.43 km² (Figure 11d). The most frequent interval for the distribution of collapsed bank area was 100–2000 m², with 142 collapsed banks developed, accounting for 51% of the total. There are 18 collapsed banks greater than 10,000 m², accounting for 7%. The largest area of the collapsed bank was 88,383 m² for the bedding slip type, and the smallest area was 117 m² for the surface erosion type (Table 4).

Figure 11e analyses the change in the slope gradient of the collapsed bank before and after impoundment, and there is a clear trend of increasing slope after impoundment. Before impoundment, 27.9% of the bank slope is 35~40°, and 25.73% of the bank slope is 40~45°. After impoundment, 43.84% of the bank collapsed at 35~40° and 28.99% at 40~45°. The slope gradient of the collapsed bank increases (Figure 11e).

Bank slope structure is another critical factor influencing the development of bank failures. For the 42 collapsed banks of rocky types, 26% were distributed in the cataclinal slopes, 48% in the orthoclinal slopes, and 26% in the anaclinal slopes (Figure 11f).

BCW (bank collapse width) is the width of the collapsed bank, BCH (bank collapse height) is the height of the collapsed bank, BCL (bank collapse length) is the length of the collapsed bank (Figure 12a), and the magnitude of these three parameters determines the size of the collapsed bank. Seventy-eight percent of the BCW of collapsed banks were between 0 and 60 m, 90% of the BCH of collapsed banks were between 0 and 60 m, and 63% of the BCL of collapsed banks were between 20 and 80 m (Figure 12b). Figure 12c shows the spatial size distribution of the collapsed bank plotted in terms of area as the size of the sphere and the length, width, and height of the collapsed bank as the X, Y, and Z axes. A collapsed bank with a width, height, and length greater than 200 m and an area greater than 6000 m² will be a hazard to roads and residential building facilities.

Five collapsed banks posed an enormous threat to roads and residential buildings (

Table 5. Geometric parameters of bank collapses that pose a serious threat to roads and residential buildings.

Bank Collapse Code	Type	Area (m2)	Length (m)	Width (m)	Hight (m)	Threat Object
1	Surface slide type	7049	116	73	43	Roads
2	Surface slide type	31,254	188	163	92	Storeroom
3	Bedding slip type	88,383	651	199	163	Roads and tunnels
4	Toppling type	6434	200	48	33	Bridge
5	Surface erosion type	9830	172	92	58	Residential buildings

, Figure 13). Figure 13 shows the signs of damage to tunnels (Figure 13a,b), roadways along the river (Figure 13c,d), residential buildings (Figure 13e,f,h), and bridges (Figure 13g) at each of these five bank failures. The collapsed bank caused the leading edge of the tunnel and residential buildings to become accessible from the rock mass, the road along the river to pull apart, and the bridge’s foundation to become unstable (Figure 13), seriously threatening the lives and properties of residents. These buildings and roads are critical to the mobility of the local population, and therefore, these five bank failures require more detailed monitoring and investigation.

5. Discussion

5.1. Geological Conditions Susceptible to Bank Collapse

The bank collapse slope heights were 825 m, and the location is near the waterline. Before impoundment, the slope gradients of the collapsed banks ranged from 15° to 55°, and 83% of the collapsed banks ranged from 25 to 45°. After impoundment, the slope gradients ranged from 25° to 55°, and 86% of the collapsed bank slopes ranged from 30° to 45°. This result indicates that the collapsed bank near the water level line deforms to some extent after impoundment, and the slope gradients become larger than the original gradients.

Regarding the slope types, 234 of the 276 bank collapses were deposit slopes, and 42 were rock slopes, accounting for 85% and 15%, respectively. This result is similar to the distribution pattern of landslides in the Jinsha River Xiluodu Reservoir [28]. The soil bank slopes are composed of the reactivation of old landslide deposit. In the rock bank slopes, the massive basalt is structurally stable and not susceptible to bank failure. Rock slopes where bank failure occurs are clastic slopes composed of sandstone and mudstone, and carbonate slopes composed of limestone and dolomite are prone to inducing bank failure deformation.

In terms of rock slope types, 11 bank collapses were cataclinal slopes, 21 bank collapses were orthoclinal slopes, and 11 bank collapses were anaclinal slopes. These findings suggest that the orthogonal slopes in the head section of the bank collapse are the most deformable of bank collapses, mainly because the orthoclinal slopes have developed joint surfaces, and the reservoir water can soak into the rock surface, inducing the bank collapse.

5.2. Linkages and Differences between Bank Failures and Reservoir Landslides

Among the geological hazards induced by reservoir impoundment, the bank collapse is a widely developed type of geohazard. However, fewer people study bank failures than other geologic hazards such as landslides. Gatto and Doe qualitatively described the collapse of approximately one-third (around 10) reservoirs in North America in 1987 [31]. Much of the bank erosion along most of the reservoirs is caused by water action at the bank toe. Numerous reservoirs have been constructed in China, including the Loess Area Reservoir, the Three Gorges Reservoir, the Xiluodu Reservoir, and the Wudongde Reservoir. Among these, the most extensively researched by scholars regarding bank collapses are the Loess Area Reservoir and the Three Gorges Reservoir. Hence, the bank collapse has emerged as a prominent geological disaster issue. The Loess Area Reservoir belongs to the plain valley type reservoir; the water bank slope remained vertical after the bank collapse, and the accumulation form of the underwater bank slope followed an exponential curve [32]. The Three Gorges Reservoir belongs to the mountain type reservoir; the typical bank collapses are caused by wash and abrasion, toe-erosion collapse, rock break-off and slides, and landslide [17]. The bank collapse at the Baihetan Reservoir shares similarities with the collapses observed in other reservoirs, leading to erosion, collapse, and slope instability following impoundment. These phenomena exhibit certain commonalities. The Baihetan Reservoir belongs to the high mountain canyon type reservoir. Compared to the other two types of reservoirs, this type exhibits a larger water level and higher altitude, leading to a potentially greater scale of bank collapse than the Loess Area Reservoir and the Three Gorges Reservoir.

Bank reconstruction is a proper noun to describe this type of hazard problem. The main difference between bank collapse and reservoir landslides is that the scale and destructive capacity of reservoir landslides are much larger than that of the bank collapse, which is relatively small in scale, integrally destructive into the water, and dense distribution so that the reservoir shoreline is constantly moving back. More than 60 percent of bank failures occur in the reservoir water-level fluctuation zone, prompting the resurrection of ancient landslides and even generating new collapses that harm settlements, roads, bridges, harbors, and other buildings.

5.3. Influence Factors and Research Shortcomings of This Paper

Numerous observations of reservoir bank collapses indicate that bank failures are a long-term and complex geological process that occurs near the water level fluctuation line [33,34]. Several factors affect bank collapse, and they are related to many external factors and intrinsic geological factors. Intrinsic factors include stratigraphic lithology, topography, and bank structure; external factors include water level fluctuations, rainfall, and wind-generated wave effects.

Until now, most research has focused on rivers [35,36,37,38] and mountainous reservoirs [39]. Regarding lithology, river slopes are composed mainly of alluvial and floodplain soils, whereas the slopes of large valley-type reservoirs are composed not only of soils but also of weathered and hard rock layers. Bank slope structure influences the damage pattern of collapsed banks. Simulation experiments show that the increase of the slope gradient deteriorates the slope stability, and the slope angle is the most sensitive factor affecting the width of the bank failure [40,41,42]. Periodic wind waves, rainfall loads, and erosion on the collapsed bank slopes affect slope stability [43]. Field monitoring and investigation can provide insight into the influence of these control factors on bank collapses.

The shortcoming of this paper is that the distribution of bank failures in the head section of the reservoir is only representative of the distribution pattern of bank failures in a small section of the area, but not in the entire reservoir area during the impoundment stage, and more in-depth studies are needed. It can only represent the development model of the bank’s collapse during this period. In order to understand more the change characteristics of bank collapses, multi-temporal high-resolution remote sensing image interpretation is needed to obtain the evolutionary process of catastrophic bank collapse dynamically.

6. Conclusions

The impoundment in the Baihetan Reservoir has resulted in alterations to the geological environment, leading to varying degrees of degradation and instability on both sides of the reservoir. This study analyzed and interpreted remote sensing data from UAV aerial flights in the reservoir area to obtain the deformation pattern and developmental law of the collapsed banks in the head section following initial impoundment. The primary conclusions are as follows:

(1) In the head section of the Baihetan Reservoir, there are 276 collapsed banks on the left and right sides of the 30-km-long reservoir, with a development density of 9 collapsed banks/km. Eighty-five percent of the collapsed banks were developed in the reactivation of old landslide deposit and fifteen percent were in clastic and carbonate rock types. The branch gully is where the collapsed bank develops intensively; 46% of the collapsed banks developed here.

(2) Through the comprehensive field investigation of 276 collapsed banks along the reservoir, six distinct patterns of bank collapse were identified. These include the surface erosion type, surface collapse type, surface slip type, bedding slip type, toppling type and cavity corrosion type.

(3) Different types of bank collapses have different characteristics. Weathered deposits of the surface erosion type undergo denudation under the influence of reservoir water. Surface collapse banks have a vertical profile pattern. Tensile cracks appear at the top and waist of the surface slip type. The bedding slip-type profile shows a jagged shape. Toppling-type bank failures undergo toppling deformation along the discontinuity surface. The cavity corrosion type deformation occurs along the dissolution concave cavity.

(4) The number of collapsed banks in the reservoir area with an area larger than 500 m², 1000 m², and 10,000 m² accounted for 94%, 81%, and 7%, respectively. The slope of 26% of the collapsed banks increased by 5 to 20° after impoundment, while the other slopes remained essentially unchanged. A collapsed bank with a width, height, and length greater than 200 m and an area greater than 6000 m² will be a hazard to roads and residential building facilities.

(5) The head section of the reservoir exhibits significant variability due to differences in lithology and bank slope structure. The development of bank failure hazards is mainly controlled by stratigraphic lithology, bank slope structure, and topography. The steep, dry, and hot canyon landscape provides potential energy conditions for bank collapses. The lithology and slope structure are important factors affecting the development of bank failure. The weathered deposits are the lithological combination with the most development of the bank failure after water storage, and its structure is loose and prone to deformation. The damage of rock types is controlled by the slope structure, with toppling types occurring on anaclinal and orthoclinal slopes, bedding slip types occurring on cataclinal slopes, and cavity corrosion types occurring on carbonatite type slopes.