Accelerated Movements of Xiaomojiu Landslide Observed with SBAS-InSAR and Three-Dimensional Measurements, Upper Jinsha River, Eastern Tibet

Abstract

Many ancient landslides in the upper reaches of the Jinsha River seriously threaten the safety of residents on both sides of the river. The river erosion and groundwater infiltration have greatly reduced the stability of the ancient landslides along the Jinsha River and revived many large landslides. Studying their deformation characteristics and mechanisms and predicting possible failure processes are significant to the safety of residents and hydropower projects. We used SBAS-InSAR and three-dimensional decomposition techniques in our study. Our results showed that the trailing edge and middle part of the landslide have rapidly deformed. The maximum vertical annual displacement rate was 12 cm/a period from July 2017 to July 2019. Correlation analysis showed that creep deformation is closely related to the river damming of the Baige landslide events and that the rising river level was an important factor in the resurrection and accelerated destruction of the Xiaomojiu landslide. As a result, we predicted the possible failure process of the Xiaomojiu landslide, which might have lasted 80 s and eventually formed a landslide deposit with a height of about 150 m, a length of approximately 1500 m, and an average width of 450 m. Our results provide data references for displacement monitoring and instability risk simulation of large landslides along the Jinsha River.

1. Introduction

Since the Jinsha River is affected by the complexity of geological tectonic movement, especially the deep rivers, steep terrain, and complex geological conditions, it is prone to landslides [1,2,3]. Baiyu-Gongjue is a region in the upper reaches of the Jinsha River that suffers from medium- to large-scale landslides. In late 2018 (11 October and 3 November 2018), Baige landslides occurred successively in Boro Town, Tibet, China, causing upstream inundation and downstream flooding [4,5]. Although timely post-disaster emergency measures alleviated the two blockages, the dam created by the landslide raised the water level in a matter of days, and the high water seeped into the mountains along the Jinsha River, reviving many ancient landslides in the region. Many scholars have studied the evolution and process of the Baige landslides before and after the disaster [5,6,7,8,9,10,11], but they have largely ignored the potential danger of the resurrected ancient landslides along the Jinsha River, which may lead to disaster events similar to the Baige landslides.

Historically, many landslides have been induced by single landslide instability [12]. Due to the complexity of the relationship between internal structures and triggering factors, landslide processes vary even on adjacent slopes with similar geological and topographical conditions. The Xiaomojiu landslide is a large-scale landslide located 5 km upstream of the Baige landslide, with a height difference of about 750 m and an area of about 0.86 km²; it may cause a catastrophic event similar to the Baige landslide once destroyed. However, the attendant ground displacement and possible risk have not been quantitatively monitored or simulated [9]. The impact of the Baige landslide on the resurrection of the Xiaomojiu landslide has not been analyzed in-depth.

Ground displacement is among the most critical parameters for observing changes and movement trends of the landslide [10]. As an effective means to obtain deformation data, Interferometric Synthetic Aperture Radar (InSAR) technology has made remarkable achievements in identifying and monitoring geological hazards such as landslides [13,14,15,16,17,18,19,20,21]. Traditional InSAR and pixel-offset tracking (POT) methods are widely applied in landslide deformation monitoring, which uses phase and intensity information. The former is more suitable for identifying small deformations at the cm level, whereas the POT method is more suitable for monitoring large deformations above the meter level [22]. For example, Fan et al. applied the POT method and calculated the large-scale displacement of the Baige landslide. Their results showed that the cumulative maximum displacement reached 25 m from July 2017 to July 2018 [1]. Their deformation data were along the LOS (line of sight) or azimuth directions. However, the landslide’s evolution process is very complex, and its movement and displacement are often not in the LOS or azimuth directions. Therefore, observational results from a single direction are often not accurate enough to describe the landslide movement [23]. It is necessary to use multi-orbit SAR data to obtain three-dimensional deformation information on landslides through deformation data in three directions.

The water level of the Jinsha River rose in a short time due to the Baige landslide dam. The infiltration of groundwater increased slope weight, reduced soil cohesion, and intensified the slope’s erosion by the river, resulting in the revival of many ancient landslides. The Xiaomojiu landslide is the most active of these large-scale landslides and has potential dangers. Thus, we studied the deformation characteristics and potential hazards of the Xiaomojiu landslide for this paper. The Xiaomojiu landslide deformation data were accurately obtained from three observation angles and based on 96 scenes of SAR data over two years. The landslide’s three-dimensional deformation field was constructed based on the deformation data obtained from three directions and constructed using the 3D decomposition method. Moreover, a time-series displacement of the landslide based on the SBAS method helped analyze the landslide’s movement trends and inducing factors. Finally, we simulated the dynamic process of landslide failure using LA (Landslides Analyst) [24] and predicted possible influence regions.

2. Study Area and Datasets

2.1. Study Area

The Xiaomojiu landslide, located in the upper reaches of the Jinsha River, is characterized by a high mountain canyon located at 31°7’11”, 98°41’48” (Figure 1).

The mean annual temperature in the study area is about 8.0 °C, and the annual precipitation is 627 mm, with a concentrated rainfall from June to August [1]. The elevation ranges from 2200 to 6700 m. Large landslides along the Jinsha River are highly developed due to the vast difference in valley elevation. According to regional geological data, the study area has strong tectonic uplift, and the NW–SE Jinsha River fault dominates the tectonic fault [3]. The Xiaomojiu landslide is several kilometers away from the nearest Bolo-Muxie fault. Strong tectonic movements have caused various geological hazards in this region, including the Baige landslide, which is only 5 km away from the Xiaomojiu landslide [1,7]. The stratum in this landslide area belongs to the Proterozoic (P_txn^a) and primarily comprises metamorphic granite and clastic rocks. With the formation of fractured rock masses under tectonic action and undercutting erosion by rivers, the Xiaomojiu landslide is unstable and highly susceptible to destruction under gravity [10].

2.2. Datasets

To obtain the deformation for a long-term series of the Xiaomojiu landslide, we collected 96 scenes of SAR data (

Table 1. Data and parameters.

Data	Resolution	Span	Number	Source
Sentinel-1	Range 2.3 m, azimuth 14 m	2 years	86	ESA. https://scihub.copernicus.eu accessed on 1 January 2022
PALSAR-2	Range 4.3 m, azimuth 3.8 m	2 years	10	https://alos-pasco.com/ accessed on 1 January 2022
DEM	30 m	-	1	SRTM
Water level	1 month	1 year	13	CGS. https://www.cgs.gov.cn/ accessed on 1 January 2022
Rainfall	10 km	2 years	-	NASA
NDVI	1 km	2 years	-	NASA

and

Table 2. Detail parameters of SAR images.

SAR Data	Time	Number	Path	Frame	Incident	Azimuth
Sentinel-1 (Ascending)	20171008–20190320	44	99	1280	33.8443°	−12.784°
Sentinel-1 (Descending)	20171015–20190327	42	33	487	43.9297°	192.774°
PALSAR-2 (Ascending)	20171127–20190415	10	-	-	36.2779°	−10.341°

) over two years, including 86 scenes collected from ESA Sentinel-1 and 10 scenes from ALOS PALSAR-2 (Figure 1), respectively. The pixel spacing of Sentinel-1 SAR data in the range and azimuth directions are 2.3 and 14 m, respectively. The resolution of PALSAR-2 SAR data is about 4 m. SAR data from three observation directions can determine the three-dimensional deformation of landslides. The SRTM DEM, acquired from USGS, provides 30 m-resolution elevation information. To analyze and identify landslide factors and external geomorphic changes, rainfall and NDVI were obtained from NOAA (National Oceanic and Atmospheric Administration). Water level data of the Jinsha River were acquired from CGS (China Geological Survey).

3. Methodology

3.1. Time Series InSAR Method

In this paper, we used the SBAS-InSAR (Small Baseline Subset InSAR) technique to obtain time-series displacement. Figure 2 shows the processing workflow, and Table 2 shows the detailed data parameters.

Firstly, the general processing of SBAS-InSAR starts from SAR image co-registration [17,18], and the co-registration accuracy reaches 0.1 pixels. Precise orbit files were downloaded from ESA (European Space Agency) and preprocessed. For the selection of differential interference pairs, 36 days and 150 m were used as the time and perpendicular baseline thresholds of Sentinel-1 data. Due to the small amount of PALSAR-2 data, no time baseline thresholds were set. The temporal and perpendicular baselines are shown in Figure 3. After determining the interference pairs, a total of 262 interference pairs were generated. Then, we used the SRTM DEM data to remove the topographic and flattened phases and geocode the InSAR products.

All possible interferometric pairs interfered, and the interferograms’ coherence threshold was set to 0.7. Then, we used an adaptive filtering function of 128 window size based on the power spectral density [25]. We used the minimum cost flow (MCF) algorithm on a triangular mesh for phase unwrapping [26]. After atmospheric errors associated with elevation were removed, high-quality corrected unwrapped interferograms were retained. Finally, time-series deformation through singular value decomposition (SVD) was inverted [27].

3.2. Three-Dimensional Displacement of the Landslide

Landslide movement is a three-dimension deformation process. Determining a movement situation in a single direction is difficult to grasp accurately with InSAR monitoring. The three-dimensional deformation field can derive from different directions of InSAR measurements [23,28]. The estimated 3-D displacements (du, de, and dn) and their variances (${\sigma }_{du}^2$, ${\sigma }_{de}^2$ and ${\sigma }_{dn}^2$) were determined from three direction measurements (d_los,1, d_los,2 and d_los,3) and their errors (${\sigma }_{dlos,1}^2$, ${\sigma }_{dlos,2}^2$ and ${\sigma }_{dlos,3}^2$), as follows:

$$\left[\begin{array}{c}\mathrm{d}u\\ \mathrm{d}e\\ \mathrm{d}n\end{array}\right]=\mathsf{\Gamma }·\left[\begin{array}{c}{\mathrm{d}}_{los,1}\\ {\mathrm{d}}_{los,2}\\ {\mathrm{d}}_{los,3}\end{array}\right]$$

(1)

$$\mathsf{\Gamma }=\left[\begin{array}{c}{\mathsf{\Gamma }}_1\\ {\mathsf{\Gamma }}_2\\ {\mathsf{\Gamma }}_3\end{array}\right]={\left[\begin{array}{c}{a}_1 b_1 C_1\\ a_2 b_2 C_2\\ a_3 b_3 C_3\end{array}\right]}^{-1}$$

(2)

$$a_i=cos{\vartheta }_{inc,i},i=1,2,3;b_i=-sin{\vartheta }_{inc,i}\sin \left({\alpha }_{az,i}-\frac{3\pi }{2}\right);C_i=-sin{\vartheta }_{inc,i}\cos \left({\alpha }_{az,i}-\frac{3\pi }{2}\right)$$

(3)

$$\left[\begin{array}{c}{\sigma }_{du}^2\\ {\sigma }_{de}^2\\ {\sigma }_{dn}^2\end{array}\right]=\left[\begin{array}{c}{\mathsf{\Gamma }}_1 \Sigma {\mathsf{\Gamma }}_1^T\\ {\mathsf{\Gamma }}_2 \Sigma {\mathsf{\Gamma }}_2^T\\ {\mathsf{\Gamma }}_3 \Sigma {\mathsf{\Gamma }}_3^T\end{array}\right]$$

(4)

$$\Sigma =\left[\begin{array}{c}{\sigma }_{dlos,1}^2 {\sigma }_{dlos,1 dlos,2} {\sigma }_{dlos,1 dlos,3}\\ {\sigma }_{dlos,1 dlos,2} {\sigma }_{dlos,2}^2 {\sigma }_{dlos,2 dlos,3}\\ {\sigma }_{dlos,1 dlos,3} {\sigma }_{dlos,2 dlos,3} {\sigma }_{dlos,3}^2\end{array}\right]$$

(5)

where θ_inc,i and α_az,i are the orbit azimuth angle and incidence angle, respectively; du, de, and dn are the displacement in three different directions; Γ is the conversion coefficient; $d_{los,1,2,3}$ are the LOS displacement in three different directions [18].

Although there is a 7-day difference in acquisition dates between adjacent ascending and descending Sentinel-1 SAR data, the deformation distortion error due to the sampling time interval is almost negligible. The time difference between PALSAR-2 and Sentinel-1 data is about one month, so we assumed that the landslide was in uniform motion during this period. We applied the time-domain interpolation method to eliminate the sampling time interval between the three orbits. It is reasonable to derive 3D time-series movement from three independent direction measurements [29].

3.3. Dynamic Processes Simulation of the Landslide

Landslide analysis (LA) is numerical simulation software that combines rigid body motion theory and fluid theory [24]. This software can effectively simulate the landslide’s deposition area and velocity. Its principle is as follows: The Monte Carlo simulation generates blocks, and the displacement is calculated based on these blocks. The stress of the slope is transferred through columns. The fluid model is initially established, and the finite difference method obtains the numerical solution. Finally, the block motion equation based on the Lagrangian description is used to describe the landslide movement.

The input parameter data includes the initial sliding mass thickness and post-sliding terrain data (DEM). The internal friction angle of the sliding mass (degree) and lateral pressure ratio (decimal) are set according to the geotechnical parameters (

Table 3. Calculated parameters on the landslide.

φ (°)	k	hc (m)
35	0.05	0.8

φ is the friction angle, k is the lateral pressure ratio, and h_c is the Cohesion C expressed in the unit of height (C = ρgh_c).

). The original terrain data and the simulation scene set the simulated model resolution and particle count.

The workflow of this study is shown in Figure 4. We used a total of 96 SAR data scenes from three measurement directions in this study. Firstly, we used multi-orbit SAR images to obtain the LOS deformation from three observation directions and construct the three-dimensional deformation field of the landslide. Then, we analyzed the correlation and inducing factors between the landslide time-series deformation and external factors with SBAS-InSAR techniques. Finally, we simulated the dynamic process of the unstable landslide and predicted the possible influence zone based on the LA model.

4. Results

4.1. Three-Dimensional Deformation Field of the Landslide

The three-dimensional geomorphic features of the Xiaomojiu landslide are shown in Figure 5, and the image was obtained from Google Earth. The landslide has a tongue shape in plain view, where most of the material is loosely broken. Based on the topographical features, the landslide belongs to typical large-scale landslides with an area of approximately 0.86 km² and is about 1350 m long and 600 m wide. The main sliding direction is about 46°. The slope angle averages 30° and reaches 40–50° in some places. According to remote sensing image interpretations, the rear edge and local areas on both sides of the landslide are severely damaged, and the surface vegetation is destroyed with the broken rock mass exposed. The middle part of the landslide presents transverse cracks formed by the tensile stress of the sliding mass (Figure 5). A caving zone with an area of about 100 m² can be seen in the middle of the front edge. In addition, a vertical expanding fracture zone was located at the left front of the slope, which was formed by the compressive stress due to the downward slope.

Figure 6a–c shows the LOS displacement of the landslide obtained by the InSAR technique. Due to the NE direction aspect, ascending images of Sentinel-1 and PALSAR-2 were more favorable for deformation acquisition. The results of the two ascending images both showed that the cumulative LOS deformation of the landslide was about 20 cm during the two years, and the large deformation areas were mainly concentrated in the middle and front of the landslide.

In order to compare the differences in displacement obtained from different orbit data, we established three deformation profiles along with the horizon and slope directions (A-A’, B-B’, C-C’ in Figure 7). The observation results from descending images are poor; therefore, the Sentinel-1 deformation results marked by yellow dotted lines showed minimal deformation. The displacement obtained from Sentinel-1 and PALSAR-2 ascending images is similar, and the deformation profiles (A-A’, B-B’) both showed about 10 cm of cumulative deformation. The deformation differences between the two ascending results are within the acceptable range caused by the difference in the satellite orbit angle and the identification ability of the SAR band.

According to the method described in Section 3.2, the three-dimensional velocities were derived to better explain the landslide movement. Figure 8 shows the three-dimensional velocity field of the Xiaomojiu landslide with several areas of rapid deformation. Figure 8a shows that the Xiaomojiu landslide has large-scale vertical subsidence in the middle and front parts with a maximum deformation of 12 cm/a. Moreover, the significant stress failure in the rapid subsidence area will damage the rock mass, which corresponds to the tensile damaged area in Figure 5. Figure 8b,c shows the deformation velocity of the Xiaomojiu landslide in the east-west and north-south directions. The horizontal displacement velocity of the landslide in the middle and front areas was about 8–10 cm/a, whereas the displacement velocity of other areas was relatively minor, about 3–5 cm/a. Through the landslide’s displacement velocity fields in the horizontal and vertical directions, we found that the rear part of the landslide was mainly subsidence. By contrast, the middle and front parts of the landslide have horizontal and vertical movements, which can be judged as push-type landslides.

Error analysis is important for the accuracy of InSAR results. According to Equations (4) and (5), we calculated the 3D deformation decomposition errors. Figure 9 illustrates cumulative errors in the three directions generated by the 3D decomposition. The errors in both the vertical and east-west directions are <5 cm, which are acceptable considering the annual deformation of the landslide of >10 cm. While in the north-south direction, the maximum error in this direction inevitably reaches 12 cm due to the near-northern orbit of the SAR satellite. However, the landslide in this study mainly moves in the east-west and vertical directions, so the error in the north-south direction has little impact on the landslide deformation analysis, proving the reliability of the InSAR results.

4.2. Time-Series Deformation of the Landslide

The time-series deformation of the landslide had adequate data for analyzing landslide evolution and failure process [30]. It was crucial to obtain these data for monitoring the Xiaomojiu landslide. Figure 10 shows the time-series deformation curves of six large deformation points in the landslide area obtained by the SBAS-InSAR method (the six points are marked in Figure 7), and the average displacement velocity is shown in Figure 7a. Based on the time-series deformation and velocity results, the deformation rate within the landslide boundary was generally above 5 cm/a, whereas the areas outside the boundary were almost <2 cm/a. The deformation values of six points from October 2017 to April 2019 (Figure 10) helped analyze the motion state quantitatively. The negative values indicate motion away from the satellite in the LOS direction, and the positive values represent motion toward the satellite. Due to the different observation angles between ascending and descending orbits, it was correct that results from ascending images showed negative values while results from descending images showed positive values. Overall, the deformation in the middle part of the slope is relatively large (Figure 10a–e) and has an average deformation of −10 to −14 cm/a in the ascending image results, whereas that of the toe area (Figure 10f) is relatively minor with a range of −6 to −7 cm/a. It is worth noting that the toe area’s deformation values showed a relatively large deformation of about 2 cm/a, indicating a significant uplift in the region.

Figure 11 shows the time-series deformation along the LOS direction based on PALSAR-2 SAR data and the SBAS-InSAR method. From November 2017 to July 2018, only a small area collapsed at the toe of the slope. There was no apparent displacement in the landslide area. Subsequently, the most apparent deformation signals occurred in the front of the slope. From October 2018 to November 2018, the cumulative deformation gradually increased in the middle part with a velocity of about 5 cm/month, and the cumulative displacement reached 18 cm. After December 2018, the deformation rate of the region in the rear edge also increased rapidly. Beforehand, there was a low deformation rate in this region. In the end, the cumulative deformation reached 20 cm by March 2019.

We divided the motion states into Zone Ⅰ, Ⅱ, and Ⅲ according to the deformation results (Figure 12a). Zone Ⅰ corresponds to the landslide area with a deformation velocity >10 cm/a, located at the rear edge and middle regions and constituting the central crush zone. Figure 12b,c shows two local failure areas: b is the rock mass failure at the rear edge, and c is the tensile cracks by rapid slippage. Zone Ⅱ is the heavy erosion area in the landslide mass, where the rock mass is severely damaged. Mainly on the left edge, the deformation velocity is >8 cm/a. The deformation velocity in Zone Ⅲ is <8 cm/a, which differs from the rapid deformation zones. Therefore, signs of tensile cracks or extrusion uplift can be found in the border area. Figure 12d shows the displacement velocity profile (A-A’) between Zones Ⅰ and Ⅲ with apparent differences in the fracture zones. Figure 12e shows the time-series curves of two points in the rapid deformation area (Ⅰ) and uplift areas of the toe region.

5. Discussion

5.1. Possible Inducing Factors of the Xiaomojiu Landslide

According to the time-series displacement, the Xiaomojiu landslide accelerated in about October 2018. The Baige landslide blocking the Jinsha River during this period was an important factor in inducing the landslide, and external rainfall may also have been one of the trigger factors. Therefore, we collected rainfall and water level data in the landslide area to analyze the relationship between the LOS deformation rate and possible factors (Figure 13). The NDVI data in the landslide area was also collected during a 16-day interval, which helped analyze the damage to landslide surface vegetation (Figure 14). Our results showed an apparent correlation between the deformation rate and water level curves, whereas rainfall seems to have had little effect on the landslide (Figure 13). In late 2018 (11 October and 3 November 2018), the Baige landslide blocked the Jinsha River and formed a high water level upstream. Beforehand, the displacement of the Xiaomojiu landslide was minor, about 2 cm/a. After the water level rose, the landslide entered a stage of accelerated deformation. Although the water level dropped to its original height, the landslide’s acceleration did not slow down. However, during the peak rainfall period of July–August 2018, the landslide rate did not change significantly.

We preliminarily analyzed the impact of rising water levels on landslides. (1) The water infiltrated and softened the rock mass during the high water level, decreasing its mechanical strength significantly. At the same time, water increased the self-weight of the landslide, thus making it prone to slope instability under gravity. (2) After the dam collapse, plunged water levels and fast flow accelerated the erosion of the slope and caused the local small-scale collapse, which greatly reduced the stability of the slope (Figure 5). The NDVI data also show that the vegetation index in the landslide area was rapidly decreasing, reflecting the aggravation of landslide surface damage. Figure 14 shows the NDVI values of the landslide in April 2018 and April 2019; the chosen time eliminates the effect of different seasons on NDVI. Our results showed that the NVDI value of the landslide area decreased from 0.21 to 0.12 after one year, indicating that the destruction process of the landslide’s surface vegetation was aggravated by its displacement.

5.2. Prediction of Potential Slope Failure

A dynamic simulation of possible landslide failures is required to analyze possible disaster risks. Therefore, the numerical simulation method (LA) was used to predict the failure process of the Xiaomojiu landslide. Its principle is described in Section 3.2. By combining the extent of landslide deformation and geomorphological features, we (Figure 5) simulated the kinematic and accumulation characteristics of the landslide. According to the drilling data [31], the thickness of the Xiaomojiu landslide was 40–80 m, with an average of 60 m. The volume was estimated to be 51,600,000 m³. Thus, the average thickness of the sliding body was set to 60 m. The landslide comprises metamorphic rock and gravel soil, so we set the lateral pressure ratio and internal friction angle to 35°, which is somewhat larger than the friction angle of the soil material [32]. The cohesion was calculated as hc = C/ρg ≈ 0.8, and we set a small value of 0.05 as the lateral pressure ratio (k).

Figure 15 shows the dynamic failure process of the Xiaomojiu landslide. The whole motion lasted about 80 s. At 0–20 s, the rock mass begins to slide down the sliding face as gravity’s potential energy is converted into kinetic energy. Approximately 40 s after initiation, the sliding body passes through the river, with the mountain on the opposite bank blocking the leading edge of the rock. At 40–60 s, the sliding body gradually builds up on the riverbed. At 80 s, most of the rock stops moving and is deposited at the foot of the slope, blocking the river and forming a barrier dam. The simulation results indicated that the Xiaomojiu landslide could collapse and form a barrier dam at the Jinsha River with an area of about 450,000 m². The accumulation distribution shows that the influence area exceeds 2 km from the riverbank, and the maximum thickness of the accumulation was about 150 m.

The numerical simulation results showed the landslide’s future trends, and the initial displacement changes reflect the unstable damage in the soft zone of the landslide. We compared the numerical simulation results with the InSAR spatial deformation field. Figure 16 shows the average deformation field of the landslide and the velocity within 5 s of the initial simulated damage. Our results showed that the large deformation region indicated by InSAR is consistent with the initial instability damage region in the numerical simulation results, proving that simulation results are reliable and provide scientific guidance for pre-disaster prediction and early warnings of landslides.

6. Conclusions

We described the deformation characteristics, triggering factors, and possible disaster risks of large landslides in the Jinsha River. Based on 96 scenes of SAR images in three directions, the deformation characteristics of the landslide were obtained using SBAS and three-dimensional decomposition technologies. In addition, we analyzed the relationship between landslide deformation and water level. Furthermore, a numerical simulation of the landslide body was conducted to predict possible disaster risks. The main conclusions are as follows:

(1)

The maximum vertical annual displacement rate of the Xiaomojiu landslide reached 12 cm/a; the rear edge and middle part had rapid deformation areas with an increasing deformation trend.

(2)

The high water level accelerated the deformation of the Xiaomojiu landslide. After the Baige landslide blocked the Jinsha River, the infiltration and erosion of river water accelerated the damage of the landslide.

(3)

The increasing deformation rate implies that the Xiaomojiu landslide is on the verge of failure. Numerical simulation results showed that the whole sliding process of the rock mass might last about 80 s, forming the deposited area with an area of about 450,000 m² and a thickness of 150 m.