Failure Mechanism and Movement Process Inversion of Rainfall-Induced Landslide in Yuexi Country

Abstract

Shallow landslides are one of the main geological hazards that occur during heavy rainfall in Yuexi County every year, posing potential risks to the personal and property safety of local residents. A rainfall-induced shallow landslide named Baishizu No. 15 landslide in Yuexi Country was taken as a case study. Based on the field geological investigation, combined with physical and mechanical experiments in laboratory as well as numerical simulation, the failure mechanism induced by rainfall infiltration was studied, and the movement process after landslide failure was inverted. The results show that the pore-water pressure within 2 m of the landslide body increases significantly and the factory of safety (F_s) has a good corresponding relationship with rainfall, which decreased to 0.978 after the heavy rainstorm on July 5 and July 6 in 2020. The maximum shear strain and displacement are concentrated at the foot and front edge of the landslide, which indicates a “traction type” failure mode of the Baishizu No. 15 landslide. In addition, the maximum displacement during landslide instability is about 0.5 m. The residual strength of soils collected from the soil–rock interface shows significant rate-strengthening, which ensures that the Baishizu No. 15 landslide will not exhibit high-speed and long runout movement. The rate-dependent friction coefficient of sliding surface was considered to simulate the movement process of the Baishizu No. 15 landslide by using PFC^2D. The simulation results show that the movement velocity exhibited obvious oscillatory characteristics. After the movement stopped, the landslide formed a slip cliff at the rear edge and deposited as far as the platform at the front of the slope foot but did not block the road ahead. The final deposition state is basically consistent with the on-site investigation. The research results of this paper can provide valuable references for the disaster prevention, mitigation, and risk assessment of shallow landslides on residual soil slopes in the Dabie mountainous region.

1. Introduction

Landslides rank among the most destructive geological hazards, causing considerable casualties and economic losses globally each year [1]. Among the various triggering factors of landslides, rainfall infiltration is the primary cause [2]. The characteristics, mechanisms, and destruction processes of rainfall-induced landslides have become hot and difficult topics in relevant research fields, attracting the attention of scholars worldwide [3,4,5,6]. Studies have shown that rainfall-induced landslides are mostly shallow soil landslides, which are generally small in scale and often occur in clusters, with losses comparable to those of large landslides [7]. Currently, the academic community has reached a broad consensus on the mechanical mechanisms of rainfall-induced shallow landslides. On one hand, rapid infiltration reduces matric suction and raises pore-water pressure, decreasing effective stress and soil shear strength [8,9,10,11,12]. On the other hand, rainfall infiltration increases the unit weight of the landslide body, thereby increasing the sliding force, and is also related to the migration and accumulation of fine particles under seepage [13,14,15,16]. Despite extensive research, understanding the initiation and progression of rainfall-induced landslides remains difficult due to diverse geological conditions and unpredictable triggers. As a region prone to geological disasters in Anhui Province and a key area for the prevention and control of collapses, landslides, and debris flows, Yuexi County, with its topography mainly consisting of medium and low mountains, is significantly affected by these geological disasters, posing potential risks to the personal and property safety of the local people [17]. Therefore, revealing the mechanism of rainfall-induced shallow landslide geological disasters in this area is of great theoretical significance for disaster prevention and mitigation.

Key parameters after landslide failure, such as velocity, displacement, and deposition patterns, are critical for understanding landslide dynamics, predicting impact zones, and conducting risk assessments under similar geological conditions [18]. Currently, commonly used methods include empirical models, mechanical models, physical model tests, and numerical simulation. Empirical models assess the movement characteristic parameters of landslides based on the statistical patterns between the movement distances and geometric factors of a large number of occurred landslide samples, but they do not involve dynamic parameters (e.g., velocity, runout distance, impact area, deposition, etc.) [19]. In terms of mechanical models, the main approach is to establish dynamic or kinematic equations based on the sliding block model to obtain analytical or numerical solutions for sliding velocity. The models consider the mechanical parameters or rheological parameters (e.g., viscosity) of the sliding surface soil, making them more physically meaningful compared to empirical models [20,21]. Physical model tests can intuitively observe characteristic parameters such as the slope movement velocity, movement distance, and accumulation process [22]. However, they are time-consuming, costly, and susceptible to the effects of model size; hence, numerical simulation methods have gained favor among scholars. For example, Zhang et al. (2023) [23] used the FLAC/PFC coupled method to simulate the movement process of the Guizhou Guanling landslide; Zhang Weijie et al. (2024) [24] utilized the Smoothed Particle Hydrodynamics (SPH) analysis method to simulate the sliding distance of flow-like landslides; and Zhang Wei et al. (2024) [25] adopted the landslide dynamic depth-integrated coupled material point method to predict the sliding distance, velocity, depth, and other sliding characteristic parameters of the model landslide. In this coupled method, the landslide body is generally treated as a depth-integrated, single-phase continuum, and the density, viscosity, and volume fractions of the mixture are generally assumed to be constant. It is worth noting that in similar studies, the variability of the shear strength of the sliding surface soil is rarely considered, such as rate and stress dependence [26].

Given the above analysis, this paper takes a rainfall-induced shallow landslide in Yuexi County, located in the Dabie Mountain area—specifically the Baishizu No. 15 landslide—as an example. Based on field geological surveys and laboratory tests, the GeoStudio V2024.1 software was used to simulate and calculate the seepage field, stability, shear strain field, and displacement field of the landslide under actual rainfall infiltration conditions. This approach aims to reveal the triggering mechanism of rainfall-induced instability and the destruction of the landslide. The rate-step ring shear tests were conducted on the soil–rock interface samples of the landslide to reveal the shear rate effect on the residual strength. Finally, the PFC^2D discrete element numerical method was used to simulate the landslide movement process, considering the rate dependence of the residual strength of the sliding zone soil, and to obtain the characteristic parameters of its movement evolution.

2. Landslide Case Study

2.1. Engineering Geological Overview of the Landslide

The Baishizu No. 15 landslide is located in Baishizu, Dianqian Town, Yuexi County, Anhui Province (116°3′18″ E, 30°43′9″ N). According to the geological disaster rapid report from the Anhui Provincial Institute of Geo-Environment Monitoring, the Baishizu No. 15 landslide occurred on July 6, 2020. The slope faces 210°, with an inclination ranging from 40° to 60°, and the sliding surface dips approximately 45° in the same direction. The landslide is tongue-shaped in plan, with a rear wall height of 1 to 4 m after the landslide. The landslide body is approximately 38 m long, with an average width of 30 m, a thickness of about 2 to 6 m, an area of about 1100 m², and a volume of about 4000 m³, classified as a small landslide. The Google Earth image and the overall view of the landslide are shown in Figure 1 and Figure 2, respectively.

According to the field investigation, the landslide body material consists of silty clay with broken stones from the residual slope accumulation of the Quaternary, with broken stone content less than 40%. Except for a few particles with a diameter greater than 1.0 m, most broken stones have a diameter of 0.5 to 35 cm. The underlying bedrock is strongly to moderately weathered gneiss of the Chengjiahe Rock Group (Pt₁c) from the Lower Proterozoic, with a rock layer attitude of 190°∠46°. The sliding surface is located within the residual slope accumulation layer of silty clay with broken stones, and the sliding zone characteristics are not obvious. However, obvious sliding scratches can be seen on the left wall of the landslide. No groundwater outcrops were observed in the landslide area and the adjacent area during the investigation. The engineering geological section along the main sliding direction is shown in Figure 3.

According to the monitoring data from the meteorological station in Dianqian Town, Yuexi County, the daily and cumulative rainfall from 5 June to 6 July 2020 (a total of 10 days) is shown in Figure 4. As can be seen from Figure 4, the rainfall on 5 July and 6 July 2020, was 132 mm and 148.9 mm, respectively, both reaching the level of heavy rainstorms. The cumulative rainfall 5 days and 10 days before the landslide occurred was 359.5 mm and 898.7 mm, respectively. After experiencing two heavy rainstorms, the landslide was triggered, making it a typical rainfall-induced landslide.

2.2. Sampling and Basic Physical Properties of the Samples

To ensure the representativeness of the samples, as shown in Figure 2, samples were taken from three locations: the rear wall of the landslide (S1), the soil–rock interface within the rear accumulation of the landslide (S2), and the left wall of the landslide (S3). Due to the severe disintegration of the landslide body, only a small number of undisturbed ring knife samples were obtained at S2, while several disturbed samples were taken at these three locations. The undisturbed ring knife samples obtained at S2 were sealed and brought back to the laboratory for the determination of natural water content, dry density, soil particle specific gravity, liquid plastic limits, saturated permeability coefficient, etc. The disturbed samples were used for indoor shear mechanical tests. The basic physical properties of the three samples are shown in

Table 1. Physical properties of the samples.

Samples	ρd/ (g·cm−3)	Gs	w/ (%)	wL/ (%)	wP/ (%)	IP	Ks/ (cm·s−1)
S1	/	2.72	/	45.32	26.81	18.51	2.11 × 10−4
S2	1.30	2.71	19.16	35.51	19.95	15.56	1.12 × 10−4
S3	/	2.71	/	29.00	15.94	13.06	2.07 × 10−3

Note: ρ_d is dry density; G_s is soil particle specific gravity; w is natural water content; w_L is liquid limit water content; w_P is plastic limit water content; I_P is plasticity index; K_s is saturated permeability coefficient (samples S1 and S3 were measured at a dry density of 1.30 g/cm³).

. According to the liquid plastic limits in Table 1, the plasticity chart of the three samples is shown in Figure 5. It can be seen from Figure 5 that samples S1, S2, and S3 are all low liquid limit clays (CL).

The mineral composition of the fine particles (<0.075 mm) of the three samples was determined using X-ray powder diffraction, and the results are shown in

Table 2. The mineral composition of the three samples.

Samples	Quartz/%	Potassium Feldspar/%	Plagioclase/%	Clay Minerals/%	Components of Clay Minerals/%
					K	I	C	I/S	C/S
S1	6.0	5.8	1.3	86.2	50.9	2.8	4.6	41.7	0
S2	17.6	5.5	6.5	70.4	52.8	6.4	5.1	30.5	5.2
S3	15.2	6.0	9.1	69.7	53.6	7.7	5.8	25.4	7.5

. As can be seen from Table 2, the main mineral components of the three samples are clay minerals, quartz, potassium feldspar, and plagioclase. The clay mineral purification analysis shows that the clay minerals of the three samples mainly consist of kaolinite, illite, chlorite, and illite/montmorillonite, with samples S2 and S3 containing a small amount of chlorite/montmorillonite. Taking sample S2 as an example, its XRD diffraction pattern is shown in Figure 6.

The microscopic morphology of the fine particles of the three samples was observed using a scanning electron microscope, and the results are shown in Figure 7. As can be seen from Figure 7, the fine particles of the three samples are mainly flaky particles and their aggregates (sample S1 contains some needle-like particles), reflecting the layered structure characteristics of clay minerals.

3. Shear Mechanical Properties of the Samples

The shear mechanical properties of the soil are key factors controlling the occurrence mechanism of landslides and the movement process after destruction. The shear strength of the soil involves peak, residual, and long-term shear strength. For new landslides, when the shear stress on the sliding surface exceeds its peak shear strength, the landslide fails and becomes unstable. During the movement process after the landslide becomes unstable, the residual strength controls the kinematic mechanism of the landslide. Therefore, direct shear tests were first conducted on the three disturbed samples to obtain peak strength indicators, providing a comprehensive reference for the selection of strength parameters in the subsequent numerical simulation study of landslide stability evolution mechanism. Then, ring shear tests were conducted on the soil–rock interface samples (S2) of the landslide to obtain residual strength indicators and the rate dependence of residual friction coefficients, providing a parameter basis for the subsequent simulation of the landslide movement process after the failure.

3.1. Direct Shear Test

The direct shear test was conducted using the STSJ-5A type intelligent electric four-link direct shear apparatus. Considering the size of the shear box, the three samples were all passed through a 2 mm sieve, and then, four saturated ring knife samples were prepared for each sample at a dry density of 1.30 g/cm³. Under four levels of normal stress (50 kPa, 100 kPa, 150 kPa, and 200 kPa), shear was performed at the same shear rate (0.8 mm/min). Kuenza et al. (2004) [27] believe that when the gravel content in the soil does not exceed 40%, its shear characteristics are mainly controlled by the matrix material. Therefore, the shear mechanical characteristics of the samples after removing particles larger than 2 mm are basically unaffected, and the test results can represent the shear mechanical characteristics of the landslide body.

The peak shear strength envelopes of the three samples are shown in Figure 8. As can be seen from Figure 8, the effective peak cohesion (c′_p) of samples S1, S2, and S3 are 28.43 kPa, 4.4 kPa, and 2.2 kPa, respectively, and the effective peak internal friction angle (φ′_p) are 19.83°, 25.52°, and 28.40°, respectively. Obviously, the effective peak cohesion of sample S1 is significantly higher than that of samples S2 and S3, and the effective peak internal friction angle is the smallest. This is because sample S1 has a higher content of fine particles and clay minerals.

3.2. Ring Shear Test

The ring shear test was conducted on sample S2 using the THJ-200 type automatic ring shear apparatus from Hunan Yaxing Testing Technology Co., Ltd. (Changsha, China). The inner and outer diameters of this type of ring shear apparatus are 10 cm and 15 cm, respectively, with a maximum loading height of 6 cm, a shear area of 98.125 cm², and a maximum normal stress of 2 MPa that can be applied. There are two shear modes: stress control and rate control. Drainage conditions can be controlled by the drainage valves on the upper and lower shear boxes. The test content includes (1) consolidated drained shear tests under different normal stresses (at the same shear rate) to obtain the residual strength indicators of the samples and (2) rate step drained shear tests under the same normal stress to obtain the rate dependence of the residual strength of the samples.

The sample S2 was passed through a 2 mm sieve and dried in an oven. The following test procedures were carried out: (1) Based on the sample volume, the mass of the dried sample was weighed at a dry density of 1.30 g/cm³, and the sample was filled into the shear box using the dry falling method. After the sample was filled, the upper and lower drainage pipe valves were opened, and water was passed to saturate the sample (saturation time 48 h). (2) After the sample was saturated, it was consolidated under a normal stress of 200 kPa. After the consolidation was stable, the sample was sheared to the residual state at a shear rate of 10°/min (0.01818 mm/s) under 200 kPa, 300 kPa, 400 kPa, and 500 kPa using the rate control mode to obtain the residual strength indicators. (3) After the above tests were completed, the sample was re-consolidated under 200 kPa, and then, the rate step drained ring shear test was carried out. The shear rates were set to 10°/min, 20°/min, 100°/min, 200°/min, 500°/min, 1000°/min, and 2000°/min. The sample was sheared to the residual state at each shear rate level, and then, the next shear rate was applied until the test was completed.

Figure 9a shows the consolidated drained shear process curve of saturated sample S2 under four normal stresses (200 kPa, 300 kPa, 400 kPa, and 500 kPa). As can be seen from Figure 9a, the shear stress of sample S2 first increased and then reached the residual state under normal stresses of 200 kPa to 400 kPa. This may be due to the coarse particles interlocking during the formation of the shear band, followed by subsequent fragmentation.

The shear stress at the stable stage in Figure 9a was taken as the residual strength. According to the Mohr–Coulomb strength theory, the residual strength indicators of the three samples were obtained, and the residual strength envelope is shown in Figure 9b. As can be seen from Figure 9b, the effective residual cohesion (c′_r) of sample S2 is 16.79 kPa, and the effective residual internal friction angle (φ′_r) is 11.57°. Compared with the peak shear strength indicators of sample S2 (c′_p = 4.4 kPa, φ′_p = 25.52°), the residual cohesion increased significantly after long-distance shearing, but the internal friction angle decreased significantly. This is because after long-distance shearing, soil particles are oriented along the shear surface, reducing the friction strength. However, due to the compression of the volume, the connection between soil particles is tighter, and the intergranular bonding force is stronger, resulting in greater cohesive strength.

Under the same normal stress (200 kPa) and drainage conditions, a rate step ring shear test was carried out on the sample S2 that had reached the residual state. The shear stress–time curve during the entire rate step ring shear test process is shown in Figure 10. According to the rate-and-state-dependent friction law (RSFL) [28], the relationship between the stable friction coefficient ${\mu }_{ss}$ (i.e., the residual friction coefficient) and the sliding (or shear) rate v can be expressed as a logarithmic function:

$${\mu }_{ss}={\mu }_0+\left(a-b\right)lg\left({\displaystyle \frac{v}{v_r}}\right)$$

(1)

where $v_r$ is the reference rate; ${\mu }_0$ is the friction coefficient corresponding to the reference rate $v_r$; and $\left(a-b\right)$ is the friction parameter in the RSFL, defined in this paper as the shear rate effect coefficient. The magnitude of $\left(a-b\right)$ reflects the type and strength of the residual strength shear rate effect of the soil, with $\left(a-b\right)$ > 0 indicating rate strengthening and vice versa indicating rate weakening.

Based on Equation (1), with a shear rate of 10°/min (0.01818 mm/s) as the reference rate $v_r$, the shear rates were normalized, and the fitted curve is shown in Figure 11, with the following expression:

$${\mu }_{ss}=0.1718+0.1137lg\left({\displaystyle \frac{v}{v_r}}\right)$$

(2)

where $v_r$ = 1.818 × 10⁻⁵ m/s.

The shear rate effect coefficient $\left(a-b\right)$ in Equation (2) was 0.1137, indicating that the residual friction coefficient of sample S2 (taken from the soil–rock interface of the landslide body) had a significant motion rate dependence and exhibited a clear monotonic rate strengthening effect. This pattern controlled the movement characteristics of the Baishizu No. 15 landslide after destruction. The control mechanism is as follows: The landslide initially slides under acceleration after failure. As the sliding zone soil decays from peak strength to residual strength, the landslide accelerates to a certain speed. Subsequently, due to the rate strengthening effect of the residual friction coefficient, the landslide gradually decelerates, and at the same time, the residual friction coefficient decreases. Then, the landslide enters a new round of acceleration and deceleration, repeating this process until the sliding stops. Therefore, the movement speed of the Baishizu No. 15 landslide exhibits an oscillatory characteristic with the sliding time, and the landslide will not exhibit high-speed and long runout movement characteristics.

4. Impact of Rainfall Infiltration on Landslide Stability

In this paper, the SEEP/W, SIGMA/W, and SLOPE/W modules in the GeoStudio software were used to simulate the stability of the Baishizu No. 15 landslide and the response characteristics of stress, strain/displacement to rainfall infiltration, in order to reveal the triggering mechanism of this rainfall infiltration on the landslide.

4.1. Numerical Model and Parameter Selection

Based on the field investigation results of the Baishizu No. 15 landslide, a two-dimensional numerical model was established for the main sliding section, as shown in Figure 12. The model materials consist of three soil and rock layers: silty clay with broken stones (M1), strongly to moderately weathered gneiss (M2), and weakly to unweathered gneiss (M3). The initial groundwater level of the landslide is shown as the blue dashed line in Figure 12. The model has a total length of 140 m and a height of 75 m. Three monitoring sections (Z1, Z2, and Z3) were set at different positions in the numerical model to study the changes in pore-water pressure, shear strain, and displacement of the landslide body at different locations and depths with the duration of rainfall. Note that the monitoring sections Z1, Z2, and Z3 corresponded to the rear, middle, and the front part of the landslide, respectively.

Comprehensive consideration was given to the indoor physical and mechanical test results of the landslide body samples (see Table 1 and Figure 8), and the physical and mechanical parameters of the soil and rock bodies as well as the hydraulic parameters required for seepage and stability calculations were determined with reference to empirical values, as shown in

Table 3. Material parameters of the numerical model.

Materials	γ/(kN·m−3)	c/kPa	φ/(°)	E/kPa	Ks/(cm·s−1)	θs
M1	17.5	7.0	19.83	2800	2.11 × 10−4	0.50
M2	21.0	21.0	35.00	2800	1.12 × 10−4	0.15
M3	23.0	30.0	70.00	5000	1.00 × 10−7	0.0022

Note: γ is the natural unit weight; c is the cohesion; φ is the internal friction angle; E is the elastic modulus; K_s is the saturated permeability coefficient; and θ_s is the saturated volumetric water content.

. The material constitutive model used the Mohr–Coulomb model, and the stability calculation method used the Morgenstern–Price method.

4.2. Simulation Conditions

The actual rainfall before the occurrence of the Baishizu No. 15 landslide was used as the calculation condition. The simulation period was from 0:00 on 21 June 2020 to 23:00 on 6 July 2020, with a total calculation duration of 16 days. The simulation was carried out with a time step of 1 h, totaling 384 time steps. By setting the above rainfall conditions on the upper surface boundary of the model, as shown in Figure 13, the transient simulation of the seepage field was carried out, and the stability, stress field, strain field, and displacement field of the landslide were calculated at each time step. Based on the results of the numerical simulation, the seepage field, stress field, strain field, and the changes in displacement in the XY direction of the landslide body were analyzed at different time steps, including the initial state (time step 1), the 4th day (time step 96), the 8th day (time step 192), the 12th day (time step 288), and the 16th day (time step 384). Additionally, the trend of the landslide stability coefficient Fs with the duration of rainfall was determined for each condition.

4.3. Results and Analysis

(1) Variation of stability coefficient with rainfall infiltration.

Figure 14 shows the variation curve of the stability coefficient (F_s) of the Baishizu No. 15 landslide with the duration of rainfall during the numerical simulation period. It should be noted that the hourly rainfall was accumulated to the daily rainfall in Figure 14.

As can be seen from Figure 14, the stability variation trend of the Baishizu No. 15 landslide has a good correspondence with the rainfall. During the continuous rainfall of three days on 21, 22, and 23 June, the stability coefficient of the landslide gradually decreased. In the following three days with basically no rainfall, the stability coefficient of the landslide slightly increased. During the continuous rainfall of three days on June 27 to 29, the stability coefficient of the landslide decreased accordingly. Under the heavy rain on July 5 and 6, the stability coefficient of the landslide dropped sharply to F_s < 1, and the landslide became unstable on July 6. This result is consistent with the actual occurrence time of the landslide, indicating that the established numerical model is reliable.

The decrease in the stability coefficient of the landslide under rainfall infiltration is due to the reduction in the matric suction of the slope rock and soil (i.e., the increase in pore-water pressure), leading to the reduction in the effective stress and shear strength of the rock and soil. Figure 15 shows the variation curves of pore-water pressure in the rock and soil at different depths with the duration of rainfall.

As can be seen from Figure 15, under the actual rainfall conditions, the pore-water pressure within 2 m below the ground surface at the three sections gradually increased, with the response at section Z3 (near the front of the landslide) being the most rapid. Except for section Z3, the pore-water pressure of the rock and soil below 2 m at sections Z1 and Z2 (i.e., the middle and rear part of the landslide) basically did not change. The above results indicate that the influence range of rainfall infiltration is mainly concentrated within 2 m below the ground surface of the landslide. However, the significant increase in pore-water pressure within 4 m depth at the slope foot indicates that this area may be a convergence zone for rainfall infiltration, leading to a reduction in the shear strength of the soil in this area and making it a weak zone for the stability of the entire landslide body.

(2) Distribution of maximum shear strain.

When the landslide body deforms, the degree of deformation at different points in the slope is generally not the same. Shear strain is used to describe the degree of deformation at a point in the rock and soil. The occurrence and development process of maximum shear strain during the incubation of a landslide reveals the formation process of the slip surface and the deformation and destruction mechanism of the landslide. Figure 16 shows the cloud map of the maximum shear strain of the Baishizu No. 15 landslide during instability. As can be seen from Figure 16, the landslide has a concentration of shear strain at the slope foot, indicating that the landslide will first occur at the slope foot and form a shear outlet and then traction in the entire landslide body will cause it to become unstable and destroyed, showing a “traction type” failure mode.

(3) Landslide displacement field.

Figure 17 is the cloud map of the XY direction displacement of the Baishizu No. 15 landslide during instability. As can be seen from Figure 17, the landslide forms a displacement concentration area in the shallow part (<2 m) of the front and slope foot, which is consistent with the concentration area of the maximum shear strain in Figure 16.

To more clearly show the evolution characteristics of the landslide displacement, the variation curves of the cumulative displacement in the XY direction at different depths at the three sections with the duration of rainfall were drawn, as shown in Figure 18. As can be seen from Figure 18, with the infiltration of rainfall, the displacement in the depth range of 2 to 4 m of the landslide body increased significantly, and the displacement increase at section Z3 was the most obvious, reaching about 0.5 m.

5. Simulation of the Movement Process After Landslide Destruction

As a representative method of discrete element numerical simulation, PFC (Particle Flow Code) is widely used in the simulation of large deformation destruction of rock and soil bodies, such as the disaster process of collapses and landslides. In this paper, the PFC^2D 9.00 (2024 Itasca Consulting Group, Inc., Minneapolis, MN, USA) software was used to invert and simulate the movement process after the destruction of the Baishizu No. 15 landslide, obtaining the movement speed, movement distance, and accumulation status of the landslide and comparing them with the actual investigation results.

According to the main sliding section engineering geological map shown in Figure 3, a PFC^2D numerical model of the Baishizu No. 15 landslide was established, and three monitoring points were sequentially set on the sliding surface, as shown in Figure 19. The landslide model adopted a ball–clump–wall model, in which the landslide body was a coupled model of ball and clump, consisting of 4959 particles and 141 clumps (each clump composed of 29 pebbles). The bonding between particles is parallel, the bonding between particles and clumps is parallel, and the bonding between clumps is linear; the slide bed and the rear stable area were modeled as a wall, which was bonded linearly with the particles. The friction coefficient between particles was assigned according to Equation (2) to realistically reflect the rate effect of the residual friction coefficient of the sliding zone soil.

5.1. Numerical Model and Micro-Parameter Calibration

The rationality of the micro-parameters of the model has a direct impact on the accurate simulation of the accumulation range and movement distance of the landslide. Domestic and foreign scholars usually use the consistency of the strength parameters obtained from an indoor test and numerical simulation of rock and soil bodies to calibrate the meso-parameters of the PFC numerical model [28,29]. In this paper, the meso-parameters of the numerical model were determined by comparing the results of the PFC^2D biaxial compression test and the direct shear test (sample S2). Figure 20 is the PFC^2D biaxial compression test numerical model. Under the micro-parameters listed in

Table 4. Physical properties of the sample.

Parameter Types	Values
Particle density ρ/(kg·m−3)	2000
Particle contact modulus Ec/MPa	50
Particle stiffness ratio kn/ks	1.0
Particle friction coefficient μ	0.46
Parallel bond modulus Ep/MPa	1.5
Parallel bond stiffness ratio knb/ksb	1.0
Normal tensile strength of parallel bond/kPa	34
Tangential tensile strength of parallel bond/kPa	34
Parallel bond cohesion/kPa	34
Parallel bond friction angle/(°)	45
Parallel bond radius coefficient	1.0

, the stress–strain relationship of the sample failure process under confining pressures of 50 kPa, 100 kPa, and 150 kPa was obtained, as shown in Figure 21.

Based on Figure 21, the peak deviator stress values (σ₁ – σ₃) at the three confining pressures were determined. Under the Mohr–Coulomb strength criterion, the residual cohesion and internal friction angle were 19.71 kPa and 12.99°, respectively, which basically matched the residual shear strength indicators obtained from the indoor ring shear test of the soil–rock interface soil (sample S2). This indicates that the micro-parameters listed in Table 4 can be used for the simulation of the movement process after the destruction of the Baishizu No. 15 landslide. However, it should be pointed out that the particle friction coefficient μ in Table 4 was defined according to Equation (2) during the simulation of the landslide movement process.

5.2. Simulation Results and Analysis

(1) Landslide accumulation characteristics.

The total movement time of the simulated Baishizu No. 15 landslide was about 36 s. The landslide displacement cloud map and accumulation state at different times are shown in Figure 22. As can be seen from Figure 22, the Baishizu No. 15 landslide showed a clear movement zoning characteristic, with the sliding distance in the middle-front part of the landslide being significantly greater than that in the middle-rear part. After the landslide started, it first occurred at the slope foot, and then, traction in the overall landslide body caused it to slide down. A shear crack was formed at the rear after 3.5 s. Subsequently, with the progress of the simulation, the landslide showed overall downward movement, but the movement displacement at the front was greater than that at the rear. After 25 s, the displacement of the shallow particles of the landslide body was significantly greater than that of the deep particles. The landslide finally stopped moving at about 36 s, with the landslide body forming a landslide wall at the rear and accumulating as far as the platform at the front of the slope foot, but it did not block the road ahead. The final accumulation state was basically consistent with the on-site investigation results. The maximum sliding distance of the landslide body was about 13 m, mainly distributed in the shallow part of the middle of the landslide body (see the red particles in the figure).

(2) Landslide movement rate characteristics.

Figure 23 shows the variation curves of the movement speed of the three monitoring points (P1, P2, and P3) on the sliding surface with time. Monitoring point P1 is located at the rear of the landslide body, monitoring point P2 is located in the middle of the landslide body, and monitoring point P3 is located at the front of the landslide body.

As can be seen from Figure 23a, the slope body at the rear of the landslide showed accelerated movement after the landslide started and then experienced accelerated and decelerated movement to reach a peak of about 1.8 m/s. Subsequently, the landslide movement gradually decelerated until it stopped. Although the movement speed at monitoring point P1 generally first accelerated to the peak and then decelerated to stop throughout the movement process, it showed a certain oscillatory change characteristic.

As shown in Figure 23b,c, the landslide bodies in the middle and front parts (monitoring points P2 and P3) both showed obvious oscillatory change characteristics, indicating that the landslide body was in a cycle of accelerated and decelerated sliding. From the entire process of the movement speed evolution of the two monitoring points, it still showed acceleration to the maximum speed and then gradual deceleration until it stopped. Compared with monitoring point P1, the maximum movement speeds at these two monitoring points were smaller.

From the movement speed size of the three monitoring points, the entire movement process after the destruction of the landslide was relatively slow and did not show high-speed movement characteristics. The movement characteristics of the Baishizu No. 15 landslide were related to the “positive rate effect” of the residual friction coefficient of the soil. When the landslide body accelerated, the friction coefficient increased with the increase in speed, thereby increasing the anti-sliding force, and then, the landslide decelerated, repeating this process until the sliding stopped. The above simulation results show that considering the rate dependence of the residual strength of the soil can truly reflect the movement characteristics after the destruction of the Baishizu No. 15 landslide.

6. Discussion

6.1. Implications for the Landslide Prediction and Early Warning in Yuexi Country

As shown in Figure 14, the factor of safety (F_s) of the Baishizu No. 15 landslide decreases remarkably from 2 July 2020 and drops to the value less than one at 6 July 2020. According to Figure 4, from 2 July to 6 July in 2020, the accumulated rainfall is 359.5 mm. Moreover, the rainfall in two days before the landslide occurred had all reached the level of heavy rainstorms, i.e., 132 mm in 5 July and 148.9 mm in July 6, respectively. This suggests that for the landslides in metamorphic gneiss rock area in Yuexi Country, the rainfall threshold for landslide early warning can be set as a 5-day cumulative rainfall no less than 360 mm, with the previous two days both reaching the level of heavy rainstorms (e.g., >100 mm/d). Ma et al. [30] reported that a one-time rainfall of 171 mm and a maximum rainfall of 40.5 mm per hour induced 116 shallow landslides in Quannan Country of Jiangxi Province in China in 2019. Zhou et al. [31] demonstrated that accumulated rainfall of 312.9 mm in 4 days was the rainfall threshold for landslide early warning in Chongqing, China. In addition, our results show that the increase in pore-water pressure was most pronounced within 2 m in the middle-to-rear part of the landslide in the case of actual rainfall. Especially, in the front part (i.e., near the slope toe) the increase of pore-water pressure can occur within a depth range of 4 m below the landslide surface. This indicates that in the metamorphic gneiss rock area in Yuexi Country, the above rainfall level can saturate the soil within a range of two meters below the surface. However, at the front part of the slope, the accumulation of seepage may lead to saturation of the soil, resulting in a decrease in shear strength and thus causing the instability of the slope. For instance, there was the maximum displacement of approximately 0.5 m occurred at the front part of the Baishizu No. 15 landslide.

With the change in global climate and the frequent occurrence of extreme weather, clustered and sudden shallow landslides are increasingly becoming one of the significant dynamic erosion factors contributing to the destruction of geological environments in mountainous areas, which make enormous challenges for landslide prediction and early warning [32,33,34]. As a typical rainfall-induced landslide in Yuexi County, the research framework reported in this paper on the Baishizu No. 15 landslide will provide valuable references for understanding the mechanism, establishing an early warning system, and carrying out prediction under similar geological environmental conditions in the region.

6.2. Implications for the Assessment of Post-Failure Movement of Landslides in Yuexi Country

In fact, during the post-failure movement of landslides, the residual strength controls the kinematic characteristics of landslides as the sliding zone soil undergoes large displacement shear deformation [35]. Extensive laboratory test results indicate that the residual strength of soil often exhibits complex shear rate effects, manifesting as rate strengthening, rate weakening, or a combination of both [36,37,38,39,40]. Therefore, numerical simulations of post-failure landslide movement must incorporate the rate dependence of shear strength in sliding zone soils to obtain more realistic results. The rate-stepping ring shear test results of sample S2 from the soil–rock interface of Baishizu No. 15 landslide demonstrate that the residual strength exhibits significant monotonic rate strengthening characteristics, which dominate the post-failure movement behavior. Numerical simulations also reveal that back-analysis of post-failure landslide movement considering the shear rate dependence of residual strength in sliding zone soils yields results closer to real-world observations. Therefore, when conducting landslide risk assessments by evaluating post-failure movement processes in similar geological environments in Yuexi County, it is essential to test the rate-dependency of residual strength in landslide materials, particularly at the soil–rock interface between weathered layers and bedrock.

7. Conclusions

This paper took the typical rainfall-induced landslide in Yuexi County, Anhui Province—the Baishizu No. 15 landslide in Dianqian Town—as the research object and used a comprehensive method of field investigation, indoor testing, and numerical simulation to study the mechanism of rainfall infiltration-induced landslide instability and the movement process after destruction. The main conclusions are as follows:

(1) The Baishizu No. 15 landslide is a typical rainfall-induced new landslide that occurred in the residual slope accumulation soil layer in the metamorphic rock area of the Dabie Mountains. The soil samples taken from the rear wall of the landslide, the soil–rock interface of the landslide body, and the left wall of the landslide are all low liquid limit clays, with basically the same mineral composition, mainly consisting of clay minerals, quartz, potassium feldspar, and plagioclase. The residual friction coefficient of the soil–rock interface of the landslide body shows a monotonic “positive rate effect”.

(2) Under the actual rainfall conditions, the pore-water pressure within 2 m of the landslide body increased significantly. The stability coefficient of the landslide had a good correspondence with the rainfall; it dropped sharply to 0.978 under the heavy rain on July 5 and 6, 2020, and the landslide became unstable on July 6. This result is consistent with the actual occurrence time of the landslide.

(3) The landslide showed a concentration of shear strain at the slope foot and gradually expanded backward, first occurring at the slope foot to form a shear outlet, and then, traction in the entire landslide body caused it to become unstable and destroyed, showing a “traction type” failure mode. The displacement in the depth range of 4 m of the landslide body increased significantly, and the displacement increase at the slope foot was the most obvious, reaching about 0.5 m.

(4) The PFC^2D simulation showed that the movement time after the destruction of the Baishizu No. 15 landslide was about 36 s, the maximum sliding distance was about 13 m, and the movement speed along the sliding surface showed obvious oscillatory characteristics, which was related to the rate strengthening characteristic of the residual friction coefficient of the sliding zone soil. After the movement stopped, the landslide body formed a landslide wall at the rear and accumulated as far as the platform at the front of the slope foot but did not block the road ahead, and the final accumulation state was basically consistent with the on-site investigation results.