Scenario simulation of the geohazard dynamic process of large-scale landslides: a case study of the Xiaomojiu landslide along the Jinsha River

Large-scale landslides often cause severe damage due to their long run-out distances and having disaster chain effects. Scenario simulation has been adopted in the current work in order to analyze the Xiaomojiu landslide dynamic processes. The landslide characteristics and topography data are obtained via field investigations, whereas high-resolution topographic data (0.17 m) are obtained using an Unmanned Aerial Vehicle. The landslide sliding velocity, deposition characteristics, and flood outburst after a landslide dam failure were obtained using Particle Flow Code (PFC-3D) which introduced the changeable friction coefficient and the HEC-RAS software. The results showed that: 1. The landslide presents a scallop shape with a length of 1110 m, an average width of 950 m, and an area of 1.05 × 106 m2. The average thickness and volume of the sliding body are approximately 50 m and 5.45 × 107 m3. The InSAR (Interferometric Synthetic Aperture Radar) deformation analysis showed that the Xiaomojiu landslide has a maximum annual displacement rate of 60 mm/y and a maximum accumulation deformation of 180 mm since November 25, 2017. 2. The failure process of the Xiaomojiu landslide lasted for 65 s with a maximum velocity of 78.2 m/s. According to the landslide simulation results, the deposited area is approximately 2023 m long, 900 m wide, and has a maximum height of approximately 149 m. 3. A landslide-dammed lake with an elevation of 2940 m and a storage capacity of 4.13 × 109 m3 is formed after the landslide blocks the Jinsha River, and the maximum peak flow rate of the breach is 12051.7 m3/s, 43,451.4 m3/s, 148,635.6 m3/s, and 304,544.7 m3/s for the landslide-dammed failure degrees of 15%, 25%, 50%, and 75%, respectively. These results provide a reference for the risk analysis and mitigation of the landslide.


Introduction
Large-scale landslides often cause severe damage due to their long run-out distances and having a significant disaster chain effect, thus resulting in a large number of casualties, significant property loss, as well as damage to the ecological environment (Yin et al. 2016;Korup et al. 2019;Aslan et al. 2021). The complex geology and high-relief landform along the rivers indicate that the landslide failure and run-out process are complex, resulting in a very high velocity with long run-out distances (Yang et al. 2014;Ge et al. 2019;Guo et al. 2020a;Zhang et al. 2020). Landslides that occur on both riverbanks produce landslide debris that can easily block the river and form a dammed lake. The water upstream of the dam body can then inundate residential areas and infrastructures. Moreover, the severity of dam failures and consequent flood bursts cause significant damage to downstream towns, transportation facilities, communication equipment, and the environment, thereby threatening the safety of human life (Liu et al. 2019;Fan et al. 2020a).
There are many catastrophic landslide occurrences globally, especially in the regions with landforms having higher elevation differences, thus blocking rivers and forming landslide dams, for example, in 1933, a landslide triggered by a magnitude 7.5 earthquake that buried Diexi Town, Maoxian County, Sichuan Province, and blocked the Minjiang River, resulting in the formation of a dammed lake, which broke and caused a massive flood that after 45 days killed more than 2500 people (Zhao et al. 2019). In 2000, a landslide with debris spanning approximately 3 × 10 9 m 3 occurred in Yigong, Tibet, blocking the Yigong River, and forming a 130-m-high landslide dam. The outburst flood from the breakage of this dam was 1.24 × 10 5 m 3 /s causing significant loss downstream (Delaney and Evans 2016). In May 2008, a large-scale landslide occurred in Tangjiashan on the right bank of the Tongkou River triggered by the Wenchuan earthquake. The landslide blocked the Tongkou River and formed a landslide dam with a height that ranged from 82 to 124 m tall and had a storage capacity of nearly 3 × 10 9 m 3 . The landslide posed a significant threat to the lives of nearly a million people downstream (Cui et al. 2009). In October and November 2018, two landslides with volumes of 2.4 × 10 7 m 3 and 8.50 × 10 6 m 3 occurred in Baige Village along the Jiangda River. The landslides traveled a distance of 1,400 m and blocked the Jinsha River. This led to a rapid increase in the water content of the barrier lake (3.85 × 10 8 m 3 ), causing devastating damage to coastal residents and infrastructure up to 1000 km downstream along the Jinsha River (Fan et al. 2020b;Zhang et al. 2020). Meanwhile, the failure of landslides is sudden making landslide dams are extremely dangerous; thus, it is extremely difficult to forecast and alleviate such catastrophes. Therefore, scenario simulation of the geohazards' dynamics processes of large-scale landslides and geohazards' chain of "landslide-blocking river dam failure outburst flood" of the potential landslides is of outmost significance for the risk assessment and mitigation of the geohazard's (Zhang et al. 2019a, b;Liu et al. 2019;Fan et al. 2020a).
With the advancement of computer simulation techniques and geohazard numerical simulations, many scholars have studied large-scale landslide dynamics scenario simulations under different influencing factors (Bandara et al. 2015;Pastor et al. 2015;Yin et al. 2016;Guo et al. 2020b) and flood evolution process simulations (Butt et al. 2013) based on numerical simulation methods in order to understand the mechanisms of large-scale landslide failure and risk assessment. (1) Study the large-scale geohazard chain process based on post-geohazards and compare them with the geohazard characteristics in order to reproduce the geohazard chain process using a numerical simulation (Lucas et al. 2014;He et al. 2015;Liu et al. 2015;Fan et al. 2017;Liu and He 2020). (2) Numerical simulation is also adopted to study the dynamic processes of the geohazards chain process under different triggering factors, such as the dynamic process of landslides caused by earthquakes (Sun et al. 2017;Wang et al. 2018;Sarma et al. 2020), as well as the landslide process caused by rainfall, loading, and unloading (Zhang et al. 2017;Zhang et al. 2019a, b;Guo et al. 2020b;Deng et al. 2020). (3) Numerical simulations were used to carry out regional or single landslide scenario analysis, study the disaster loss and the damage zone caused by the geohazards, and provide a reference for risk analysis (Loreto et al. 2017;Fan et al. 2020b;Mao et al. 2020).
Scenario simulation of the dynamics of large-scale landslides and the detection of the long-runout of geohazard chains for a potential landslide is essential for risk assessment and the mitigation of the landslides, especially for the planning of major projects Fan et al. 2020b). The current work aims to study a potential landslide, namely the XiaoMojiu landslide as the case study. The landslide characteristics and topography data were obtained through field investigations and UAV flights. The landslide model was built using high-resolution topographic data obtained from the UAV. Field investigations and UAV flights obtained landslide characteristics and topography data. The landslide failure process, deposits, the landslide dam characteristics, and the dam breakage flood evolution process are analyzed using PFC3D which introduced the changeable friction coefficient and the HEC-RAS software.

Area setting
The Xiaomojiu landslide is located in the east of the Qinghai-Tibet Plateau on the right bank of the Jinsha River in Jiangda County of the Tibet Autonomous Region (Fig. 1) and is approximately 5 km upstream of the Baige landslide (E 98°41'49'', N 31°07'24''). Since the middle Fig. 1 Xiaomojiu landslide setting area characters (base data from https:// geocl oud. cgs. gov. cn/#/ home) and late Pliocene era, due to strong uplifts of the Qinghai-Tibet Plateau, the height differences between both sides of the Jinsha River have become more prominent, and the valleys have been strongly cut (Chen et al. 2013), creating steep slopes. The landslide's top and toe elevations are 3655 m and 2900 m, respectively, with a height difference of 755 m. The toe of the landslide undergoes strong erosion by the Jinsha River and has a slope of approximately 30-35°.
The geological structure of the Xiaomojiu landslide is closely related to the tectonic evolution of the Qinghai-Tibet Plateau and is affected by the compression and collision of the Indian and the Asia-Europe plates. The eastern Qinghai-Tibet Plateau has strong tectonic deformation and uplift, making the eastern Qinghai-Tibet Plateau is one of the most tectonically active areas. Typically, the faults are highly developed within this area (Cao et al. 2015). The concentrated development area of the geohazards formed by the action is primarily composed of metamorphic basic-ultrabasic rocks, metamorphic clastic rocks, and marble mixtures (Fig. 1). The south-north ductile shear mylonite belt and the strong schistosis belt are highly developed with a series of east-west squeezed imbricated inverses and fold structures (Zhang et al. 2016). The above-mentioned geological background makes the Xiaomojiu landslide a typical high-steep valley structure consisting of a broken loose body structure that provides favorable topography and material for landslide failure.
The study area has a semi-humid climate within the plateau cold temperate zone having an average annual temperature of 7.5 °C and an apparent vertical climate zoning. The average annual precipitation is approximately 650 mm and can reach a maximum of 1067.7 mm. Furthermore, the precipitation has the characteristics of uneven temporal and spatial distribution primarily concentrated in June, July, August, and September. Additionally, the groundwater in the study area is primarily bedrock fissure and pore water and the main sources of the water come from precipitation and melting ice and snow (Chen et al. 2013).

Discrete element method
The discrete element method (DEM) is an effective method for dynamic landslide analysis (Cundall and Strack 1979). PFC-3D (a typical of discrete element method) makes the following assumptions: (1) The particle unit is rigid and all contacts between the particles can be regarded as point contacts; (2) The contact model is a flexible contact, and at the contact point there exists a certain overlap; (3) According to the force-displacement law, the size of an overlap is related to the size of the contact force; therefore, all of the overlaps are smaller than the particle diameter; (4) The contact between the particles can establish the bonding characteristics. The particle motion characteristics are calculated using Newton's second law of motion, whereas the law of force and displacement is used to update the position of the particles and describe the motion of the particles in the PFC3D discrete element method (Eqs. 1, 2, Itasca Consulting Group, Inc 2006): where i = 1, 2, and 3 are the three directions of x, y, and z. F i (t) is the unbalanced force of the particle at time t, g is the acceleration of gravity, m is the particle mass, M is the unbalanced moment, ∆t is the calculation time step, and ω is the angular acceleration. Then, the particle displacement s i and velocity v i are:

Determination of friction coefficient
Many scholars set the friction coefficient as a fixed value while studying the dynamic process of large-scale landslides using numerical simulations, and they rarely consider changes in the friction coefficient (Han et al. 2010;Lucas et al. 2014;Liu et al. 2015). However, numerous study results have shown that the friction coefficient of the rock during the sliding process is indefinite (Han et al. 2007;Yao et al. 2013;Lucas et al. 2014;Dong et al. 2013;Deng et al. 2020). The friction coefficient of a landslide varies due to the dynamic expansion, water hammer effects, sliding separation, or saturated liquefaction.
Other effects, such as the air cushion layer, excessive pore water pressure, and rolling friction reduction, cause a reduction in the friction coefficient resulting in the high-speed and long-run outs of the landslide (Yao et al. 2013;Dong et al. 2013). Han et al. (2010) proposed an empirical equation for the steady-state friction coefficient (Eq. 5) including the velocity based on a rock high-speed friction test and the friction coefficient during the sliding process.
where ss is the steady-state friction coefficient when the speed is v, ss, max is the steadystate friction coefficient at a small slip rate, ss, min is the steady-state friction coefficient when the speed progress to infinity, and v c is the critical speed. When the speed is 0, ss is equal to ss, max ; when the speed tends to be infinite, ss is equals to ss, min .
Based on the friction attenuation empirical formula proposed by Han et al. (2010), the present paper introduced a changeable friction empirical formula based on velocity in the PFC-3D model as well as simulating the landslides movement process. The flow chart is shown in Fig. 2.
The simple model is shown in Fig. 3 and was used to verify the accuracy of the program. The experimental materials of Han et al. (2010) were selected as the material for this verification simulation. The physical parameters are shown in Table 1. The final simulation result is shown in Fig. 4, verifying the reliability and accuracy of the program.

Landslide simulation parameters
In the discrete element method, the macroscopic properties of a particle depend on its mechanical contact properties. Usually, the trail or the empirical method is time-consuming resulting in low efficiency, and it may not be appropriate to apply the properties obtained from the laboratory tests directly into the PFC-3D model (Li et al. 2018).
In order to reasonably obtain the properties parameters of rock mass, the laboratory uniaxial compression tests are carried out on the rock materials of the landslide body (Table 2). Then, a series of parametric simulations tests are performed to match the simulation results with the laboratory test results. The size of the PFC-3D simulation   5).
The sample comprised of 7558 particles with a particle density of 2650 kg·m −3 , the maximum and minimum particle size ratio was 4:3, the contact adopted a linear parallel bonding, the outside of the particles was restrained by the wall command, and the wall stiffness was 1/10 of the particles stiffness. Loading was controlled by assigning speed to the upper and lower walls. The microscopic parameters that need to be calibrated included the particle properties and the bonding relationships. The properties of the spherical particles are determined by three parameters: k n , k s , and μ. The bonding relationship was determined by E c , k n ∕k s , b , b , and , where k n is the normal stiffness of the particle, E c is the parallel bond modulus, k n ∕k s and k n ∕k s are the ratios of the normal stiffness and the tangential stiffness of the particle and the bond, b is the parallel normal bond strength, b is the parallel tangential bond strength, and is the particle bond radius coefficient.
The stress-strain curve was obtained through a uniaxial compression simulation test (Fig. 5). Table 2 shows the differences between the physical and mechanical properties obtained from the laboratory tests and the numerical model parameters. The simulation test results fit well with the laboratory tests. In order to truly reflect the characteristics of force and moment action between the particles in the PFC-3D model, the parallel contact bond model is chosen as the bond model. Based on the properties obtained from the laboratory tests and numerical simulation, the strength parameters of the landslide body calculated according to uniaxial compression of numerical simulation are used in the landslide simulation (Table 3).

Building the simulation model
The UAV model used was a Pegasus E2000, the number of planned routes was 29, the overlap rate of heading and sideways is 80%, the safe flying height is set to 350 m, the flying speed is 10 m/s, and 529 sets of orthophoto and POS data are acquired. The 3D terrain was reconstructed using the Photoscan software, and a DOM (digital orthophoto) with a resolution of 8.15 cm/pixel and a DEM (digital elevation model) with a resolution of 16.86 cm/pixel are generated. Field investigation and remote sensing interpretation methods are also used to analyze the deformation characteristics of the slope; thus, combining three drilling datasets, the thickness of the deformation body is evaluated at 30-80 m with an average of 50 m. Additionally, in order to analyze the speed of different positions of the sliding body, four monitoring points were selected in the upper part (P1-P4), middle part (P5-P8), and the toe (P9-P12) of the sliding body to record the landslide velocity change (Fig. 6).

Out-burst flood simulation
The HEC-RAS software, which uses a two-dimensional unsteady flow based on the Navier-Stokes equation, is the most commonly used method for simulating outburst floods (Hydrologic Engineering Center 2012). Assuming that the fluid is incompressible, the mass conservation (continuity) equation is: where t is time, u and v are the velocity components in the x and y directions, and q is the confluence; therefore, the momentum equation is: where u and v are the velocity components in the x and y directions, g is the acceleration due to gravity, v t is the horizontal viscosity coefficient, c f is the bottom friction coefficient, and f v and f u are the Coriolis parameters in the x and y directions, respectively.

Characteristics of the landslide
The Xiaomojiu landslide is a scallop shape with a length of 1110 m, an average width of 950 m, and an area of 1.05 × 10 6 m 2 . The sliding direction of the landslide is NE46° with the average thickness and volume of the sliding body approximately 50 m, and 5.45 × 10 7 m 3 , respectively (Fig. 7). According to landslide characteristics and the drilling data, the Xiaomojiu landslide can be divided into three zones: the upper sliding deformation zone (I), the middle traction deformation zone (II), and the slope toe stress concentration zone (III) according to deformation characteristics (Fig. 7).
The upper sliding deformation zone (I) is primarily distributed in the area with an elevation of 3326-3462 m and a slope of 22°. This zone's longitudinal length and lateral widths are 345 m and 280-680 m, respectively. Additionally, this zone's area and volume are 2.2 × 10 5 m 2 and 8.5 × 10 6 m 3 , respectively. The zone comprises several-level arc-shaped steps due to the deformation (Fig. 8). The steps are 91-257 m long, and 22.4-64.0 m wide. The vertical dislocation of the steps ranges from 12.6 to 43.5 m.
The middle traction deformation zone (II) is primarily distributed in the area with an elevation of 3127-3326 m and a slope of 38°. This zone's longitudinal length and average lateral widths are 375 m and 1010 m, respectively. Furthermore, this zone has an area of 3.6 × 10 5 m 2 and a volume of approximately 2.5 × 10 7 m 3 . Under the squeezing action of the upper sliding deformation zone, the structure of the rock mass in this zone is significantly broken and weathered with the local convex landform. There are well-developed tension cracks in the southwestern edge of the deformation zone, providing a path for rainwater infiltration. Moreover, a secondary landslide has been found in this area whose length is 246 m, width is approximately 310 m, and a height difference is 131 m (Fig. 9). The secondary landslide is a sign of shallow slippage due to road excavation.
The slope toe stress concentration zone (III) is a distributed area having an elevation of 2900-3127 m and a slope of 35°. This zone has a longitudinal length of 390 m, and average lateral width of 1150 m. The area of this zone is 4.7 × 10 5 m 2 and the volume is approximately 2.1 × 10 7 m 3 . Several local landslides occurred due to erosion caused by the Jinsha River (Fig. 10).

Fig. 7
Landslide deformation zone of the landslide (base imagery and terrain from UAV obtained by authors)

Deformation characteristics
Seventy-six SAR images were collected from the ALOS-2 satellite used for the current study between 2017 and 2020. The SAR images were processed using terrain correction, time-space decoherence factors, and an atmospheric delay phase process (Zhao et al. 2016;Liu et al. 2021), thus obtaining the landslide's annual displacement rate since 2017 (Fig. 11a). Figure 11b shows the displacement of the Xiaomojiu landslide since November 25, 2017. The maximum annual displacement rate of the landslide is 60 mm/y. Since 2017, the deformation rate of the landslide has typically increased and has tended to aggravate the deformation. The maximum deformation reached a value of 180 mm over a 3-year period. During this period, the slope deformation was primarily concentrated in the middle  and upper parts of the slope, indicating that the landslide is induced by gravity or fault activity and has an apparent instability nature. Figure 12 shows the dynamic process of the Xiaomojiu landslide, whereas Fig. 13 shows the friction coefficient distribution of the landslide at different times with the changeable friction coefficients. According to the simulation results, the landslide sliding lasted for approximately 65 s and is comprised of three stages: the initiation phase (Figs. 12a, b and  13a, b), the acceleration phase (Figs. 12c, d and 13c, d), and the deceleration accumulation phase (Figs. 12e, f and 13e, f). The sliding body begins to slide slowly along the bedrock at the initiation phase. The landslide showed an overall instability failure after 4 s with a friction coefficient of 0.25-0.35. At 12-26 s, the sliding body accelerated sliding along the bedrock, and the potential energy of gravity is converted into kinetic energy. At this time, the sliding body started to block the river. The landslide toe particles rushed  Figure 14 shows the plot of the time vs. average velocity of the monitoring points in different parts of the landslide. Comparing the average velocities of different parts of the sliding body, it was found that the particles of the upper-part (P1-P4) of the landslide had three sharp rises and drops in velocity before reaching a peak velocity. The velocities of the toe (P9-P12), middle-(P5-P8), and upper-part particles (P1-P4) of the landslide reached peak values in 26 s, 34 s, and 42 s, respectively. Due to the significant height difference in the movement path of the upper-part particles (P9-P12) of the sliding body, the peak velocity was significantly greater than the peak velocities of the toe (P1-P4) and middlepart (P5-P8) particles. For example, the peak velocity of the P1-P4 monitoring points was 61.4 m/s, the peak velocity of the P5-P8 monitoring points was 70.1 m/s, and the peak velocity of the P9-P12 monitoring points was 78.2 m/s. Figure 15 shows the characteristics of the deposited zone with a changing friction coefficient. From the perspective of the plane distribution, the landslide traveled a distance of approximately 780 m, and the total area of the deposits was approximately 7.9 × 10 5 m 2 . The deposited area presented an uneven oval shape, thin on both sides and thick in the middle. The maximum thickness was 149 m.

Characteristics of the deposited zone
The deposited area blocked the Jinsha River in order to form a large landslide dam according to the simulation results. The landslide dam was spindle-shaped with a

Outburst flood simulation
According to the landslide dynamic simulation results, the height of the landslide dam blocking the Jinsha River was between 2940 and 3000 m. The dam height that was considered the final submerged elevation was 2940 m, and the maximum storage capacity was 4.13 × 10 9 m 3 . The average annual flow in the upper reaches of the Jinsha River was considered to be 957.3 m 3 /s; the dammed lake was estimated to overflow after 50 days. To further analyze the floods from different scenarios of the landslide dam failure, we simulated and will discuss four types of scenarios: 15%, 25%, 50%, and 75% dam failure. Figure 16 shows the flow discharge process at different scenarios of the landslide dam failure. The floods from different scenarios of the landslide dam failure evolved as a single waveform. The flow discharge increased rapidly and reached a peak within 30 min and then decreased slowly, and after 24 h, the decrease was more than 90%. The maximum flow discharge of the dam breach was 12,051.7 m 3 /s, 43,451.4 m 3 /s, 148,635.6 m 3 /s, and 304,544.7 m 3 /s for 15%, 25%, 50%, and 75% of the dam failure, respectively, which exceeded even the ten-thousand-year return flow of the Jinsha River at the landslide site. Four sections that are downstream of the Jinsha River, namely the Sichuan-Tibet Line (Z1, 69 km), the Lawa Hydropower Station (Z2, 140 km), the Batang Hydropower Station (Z3, 163 km), and the Zhubalong Bridge (Z4, 180 km), were selected for flood analysis. The flood discharge process of each section for different scenarios of the landslide dam failure is shown in Fig. 18. It can be seen from the figure that the flood discharge first increased rapidly, then reached a peak flow value, and finally decreased slowly. The flood discharge curves for different scenarios of the landslide dam failure have similar characteristics: the peak discharge gradually decreases with the increasing distance from the landslide dam, the discharge process spreads downward in a single peak form, and the farther the distance from the dam, the wider the peak shape. Furthermore, as the degree of dam failure increases, the flood peak discharge of each section also increases, and the shorter the time it takes the flood to reach each section (Table 4). For the 75% dam failure, the peak flow discharge of each section was the largest and the arrival time was the fastest. Conversely, for the 15% dam failure, the peak flow discharge of each section was the smallest, and the time to reach each section was the slowest.

Discussion
Numerical simulation of geohazard chains is considered the most effective method to reveal the dynamics of landslides and the scenario analysis of potential landslides (Mao et al. 2020;Liu and He 2020). In this area, most of the published articles focused primarily on the simulation analysis of the dynamic process of post-landslides as well as studying the initiation mechanism and characteristics of post-landslides (Yang et al. 2014;Delaney and Evans 2016;Yin et al. 2016;Ge et al. 2019;Guo et al. 2020a, b). In contrast, only a handful of studies concerning the dynamic process and scenario analysis of potential landslides can be found. Moreover, the friction coefficient is often treated as a constant having a fixed value during the process of numerical simulation, which cannot reflect the real dynamic characteristics of the landslide (Han et al. 2007;Yao et al. 2013;Dong et al. 2013). Based on the existing research and the friction attenuation characters obtained from the high-speed rock friction test, the current study has proposed that the friction coefficient is related to the landslide's velocity and is included in the landslide simulation. Figure 17 shows the characteristics of the deposited zone which included both the changeable friction coefficient (model 1) and the fixed friction coefficient (model 2). Concerning plane distribution, the lateral deposited length of model 1 was larger than that of model 2. However, due to the obstruction of the mountain on the opposite bank, the difference between the horizontal and vertical accumulation lengths of the two models was not notable. The edges of the accumulated bodies along the two sides of the river channel appeared to be discontinuously distributed. In contrast, the distribution of model 1 was uniform, indicating that the model combined with the changeable friction coefficient, thus depicting high mobility and continuity of the landslide.
The comparison of the dynamic process of the landslide with changeable and fixed friction coefficients revealed a distinct velocity difference in the landslide during the movement process for the varied and fixed friction coefficients. The maximum average velocity of the landslide with changeable friction coefficient was 55 m/s, 63 m/s, and 65 m/s of the toe, middle, and upper parts of the landslide, respectively, which was higher than those with fixed friction coefficient (31 m/s, 33 m/s, and 42 m/s, Fig. 18). The average velocity of the landslide with changeable friction coefficients was significantly higher than that of the landslide with a fixed friction coefficient. This indicates that the dynamic movement of the landslide with changeable friction coefficients has more dynamic characteristics.

Conclusions
In summary, the current study has established a landslide mode using high-precision threedimensional topographic data (0.17 m) obtained by UAV. Through scenario simulation, the specific dynamic processes of the Xiaomojiu landslide were studied. More specifically, sliding velocity, deposition characteristics, and flood outbursts after landslide dam failure were analyzed using PFC-3D modeling as well as introducing changeable friction coefficient and the HEC-RAS software. The primary conclusions derived from this work are as follows: (1) According to the characteristics of the Xiaomojiu landslide, the area of the landslide is 1.05 × 10 6 m 2 and the volume is 5.1 × 10 7 m 3 with the average depth approximately 50 m. The landslide can be divided into three zones: the upper sliding deformation zone, the middle traction deformation zone, and the slope toe stress concentration zone.
The Xiaomojiu landslide has a maximum annual displacement rate of 60 mm/y and the maximum accumulate deformation has been 180 mm since November 25, 2017, according to the InSAR deformation analysis.
(2) The failure process of the Xiaomojiu landslide lasted for approximately 65 s, and the maximum velocity was 78.2 m/s. According to the simulation, the deposited area was approximately 2023 m long, and 900 m wide, with a maximum depth of approximately 149 m. (3) The landslide-dammed lake formed by the Xiaomojiu landslide failure had a storage capacity of 4.13 × 10 9 m 3 , and the maximum peak flow discharge of the breach was 12,051.7 m 3 /s, 43,451.4 m 3 /s, 148,635.6 m 3 /s, and 304,544.7 m 3 /s for landslide dam failure degrees of 15, 25, 50, and 75%, respectively, which exceeded even the tenthousand-year return flow of the Jinsha River at the landslide site.
Author contributions Jianqi Zhuang, Kecheng Jia, and Jiewei Zhan designed the analysis, developed the model code, and performed analysis. Yi Zu, Chenglong Zhang, Jiaxu Kong, Chendui Du, Yanbo Cao, and Shibao Wang curated data and field investigations. Jianqi Zhuang and Jianbing Peng prepared the manuscript with contributions from all co-authors.
Funding This study was financially supported by the National Natural Science Foundation of China (41941019, 41922054), National Key Research and Development Plan Project (No. 2020YFC1512000), and Fundamental Research Funds for the Central Universities, CHD 300102260302. The authors thank AiMi Academic Services (www. aimie ditor. com) for the English language editing and review services. The funders had no role in the design of the study and collection, analysis, and interpretation of data and in writing or approving the manuscript.

Data availability
The data that support the findings of this study are available from the corresponding author, Jianqi Zhuang, upon reasonable request.

Conflict of interest
The authors have no conflicts of interest to declare that are relevant to the content of this article.
Human and animal rights This research does not involve human or animal participants.