Stability Analysis of the Huasushu Slope Under the Coupling of Reservoir Level Decline and Rainfall

Abstract

The coupling of water level fluctuations and heavy rainfall in the Three Gorges reservoir area poses a significant threat to the stability of bank slopes, especially in landslide areas with complex geological conditions. In this study, the Huasushu slope in Fengjie County, Chongqing, was taken as the research object and, based on a field investigation and monitoring data, two- and three-dimensional numerical models were constructed to analyze the response mechanism of the slope under the combined effects of different reservoir water level decreases and rainfall. In addition, the safety coefficients under each working condition were calculated using the Morgenstern–Price method. The results show that it is difficult to trigger significant deformation with a single water level drop or rainfall. However, when the reservoir water level drops more than 10 m within a short period of time and is superimposed with strong rainfall, the landslide body is prone to plastic zone extension and significant displacement, showing typical strain localization characteristics. The three-dimensional model further reveals the spatial distribution characteristics of the landslide deformation area, which helps to accurately identify potential destabilization locations. The research results provide theoretical support for the construction of early warning systems for reservoir bank slopes and have reference value for the development of disaster mitigation engineering measures based on the coupling mechanism of rainwater and reservoir water.

1. Introduction

China has been prone to geological disasters since ancient times due to its complex and diverse terrain, geological conditions, and climate types. The Three Gorges reservoir area, China’s largest hydropower project, has played an essential role in flood control and power generation since its construction. However, the unique geological and climatic conditions of the reservoir area pose many risks of geological disasters. To study geological disasters in this area and their impact on slope stability, substantial in-depth research has been conducted on landslides and collapse formation mechanisms. The main factors affecting the stability of reservoir slopes include rainfall, reservoir water level fluctuations, the physical and mechanical properties of the rock–soil mass, and human activities.

Studies have reached consistent conclusions that fluctuations in reservoir water levels significantly impact slope stability. Japanese studies have shown that about 60% of reservoir landslides occur after sudden drops in reservoir water levels. In comparison, 40% occur during rises in water levels, especially during the initial filling of the reservoir [1]. This indicates that water level fluctuations—especially rapid drops—can have a significant impact on slope stability.

Kun Song et al. [1,2,3] verified that the rate of water level drop is negatively correlated with the stability coefficient: the faster the drop, the more obvious the decrease in the stability coefficient. This is due to a drastic change in pore water pressure, reducing the shear strength. Q. Deng et al. [4] determined the instability mechanism for the slope of a bank on a dichotomous structure through a physical model of water level rising and falling using a monitoring technique.

In terms of theoretical and numerical analyses, Emmanouil Steiakakis [5] used the Van Genuchten model to calculate the slope stability of a pit lake and showed that a decreased water level reduces slope stability. Chardphoom et al. [6,7] used unsaturated seepage theory to confirm the significant effect of the water level’s rate of change on the coefficient of safety. Based on this theory, Jiao et al. [8] proposed a dynamic analysis framework of seepage force to analyze the effect of water level decline on slope stability, quantitatively assessing stability pairs under precipitation conditions. Mao et al. [9] found that a larger unsaturated zone improves slope stability. Sun Guanhua et al. [10,11,12] verified the stability of three slopes based on the Morgenstern–Price method and the Bell algorithm, verifying the reliability of this slope analysis method for the Three Gorges reservoir area. Mehmet M. Berilgen et al. [13] analyzed the effect of reservoir water level drops on slope stability using a finite element model, emphasizing the effect of the transient seepage field on the slope.

In addition, as another predisposing factor, rainfall is closely related to slope deformation. Sultan Kocaman et al. [14] analyzed rainfall data from severely affected areas through optical and radar data analysis, and showed that rainfall action is also a major predisposing factor for slope deformation. This further confirmed the mechanism of landslide instability under the action of multiple factors. Zong et al. [15] used Automatic Dynamic Incremental Nonlinear Analysis (ADINA)-Structure software (V1.0) to show that high-intensity rainfall increases the bottom movement of slopes. Poulos H G et al. [16,17,18] showed that slopes are susceptible to instability under intense rainfall through numerical modeling.

Regarding coupling effects, A. Mori et al. [19] coupled snow melting with rainfall, finding that increased water content under the influence of rainfall decreases the safety factor in the snow melting stage. Zhou et al. [20] combined seepage force with the limit equilibrium method to construct a multifactor dynamic prediction model. Zhao N. et al. [21] analyzed monitoring data for the displacement field of Baijiaobao slope over the last 10 years, revealing the deformation evolution mechanism of this slope under three scenarios: rainfall, reservoir water level decline, and coupling of the two. Y. Zhu et al. [22,23,24] studied the Baihetan reservoir area and explored the combined effects of water level fluctuation and rainfall, deepening our understanding of the landslide mechanism. Other scholars have further analyzed landslide causes through numerical simulations [3,25,26,27,28,29,30].

Although many studies have confirmed the impact of reservoir water level fluctuations on slope stability, these studies have focused on specific engineering cases and lacked a general analysis of slope behavior under different geological conditions and climatic influences. In addition, most of these studies have relied on two-dimensional or simplified models that only partially consider the three-dimensional complexity of slopes and the joint effects of multiple factors, such as water level fluctuations and rainfall. These simplifications limited the accuracy of predicting slope behavior in actual environments. To overcome these deficiencies, this study considers the Huasushu slope in Fengjie County; uses numerical analysis methods; establishes a model for the joint effect of reservoir water level fluctuations and rainfall; constructs two-dimensional and three-dimensional numerical models; conducts a detailed analysis of slope deformations under different working conditions; and proposes a landslide prediction model. Our results not only provide a theoretical basis for analyses of the Huasushu slope, but may also serve as a reference for slope stability control projects in similar geological conditions.

2. Study Area

Huasushu slope is located in Fengjie County, Chongqing City, Hefeng Township, Wenfeng Village, on the left bank of the Daxi River slope section. It is in the Dabashan table edge fold belt and the Sichuan Taikou intersection of the composite tectonic zone, and is characterized by complex geological conditions. The right boundary of the slope is a ridge, the left boundary is a gully, the back edge is a steep cliff, and the front edge extends to the bank of Daxi River, located at geographic coordinates 109°31′28.75″ E (longitude) and 30°55′22.09″ N (latitude). The elevation is about 214 m above sea level; the elevation of the front edge is about 155 m; the back edge reaches 310 m; the direction of the main slip is 151°; and the topography as a whole presents a “slow on top, steep underneath” characteristic. The leading edge of the slope is in direct contact with the Yangtze River and is subject to scouring, forming a high and steep critical surface. A detailed map of the Huasushu landslide is shown in Figure 1.

Huasushu is a secondary large-scale slope, with a total area of about 17.40 × 10⁴ m² and a volume of about 522 × 10⁴ m³, consisting of upper loose accumulations and the muddy siltstone and sandstone of the Badong Group from the Middle Triassic in the lower part of the country. It has a loose structure and has many fissures. Slip surfaces have mostly developed along the contact surface between the slip body and the underlying relatively spaced cemented rocks, and seepage zones form easily, prompting slip behavior. Tension cracks have developed on the surface of the slope, with a width of 6 cm locally, showing certain creep–slip and traction deformation characteristics. This shows that the slope body is in a slowly developing sub-stable state.

The formation of the slope can be traced back to 2006. At the initial stage, it showed deformation characteristics indicating the collapse of a steep wall at the rear edge, and the slope body was dominated by the deformation of a local small-scale slope. With the construction of Fengyen Highway, significant digging and filling construction changed the original equilibrium state of the slope, reduced its slip resistance, and formed a high and steep airside, creating conditions for the further development of landslides. This resulted in tension cracks in the road pavement approximately 28 m long and 3~6 cm wide. These cracks spread in a direction basically perpendicular to the overall slip direction of the slope.

After the initial period, the reservoir’s annual cycle of rise and fall and regional heavy rainfall jointly influenced the water level of the Three Gorges project water storage area, causing the slope’s evolution process to enter the next stage [31]. Given that the water storage increased to 175 m, the slope deformation clearly intensified. Long-term high-water-level immersion weakened the shear strength; the seasonal water level rapidly declined, inducing unloading; and rainwater infiltration triggered pore water pressure rise. All three of these factors jointly deformed the slope [32,33]. According to years of on-site monitoring data, deformation in the slope body is due to annual cumulative displacement, often caused by June–September rainstorms and water level drops during the same period of superposition. This reaches a peak, and the deformation rate increases by up to several millimeters per day; thus, the non-flooding season rate significantly slows down, showing a “low speed–accelerate” pattern. The deformation rate can then increase by several millimeters per day, due to the superposition of heavy rainfall and water level drops in September.

Before the Three Gorges Dam was constructed, the main stream of the Yangtze River in the study area flowed faster, water level fluctuations in the riverbank were small, and the erosion at the foot of the slope was stable [1]; after the dam was completed, the water level of the Fengjie section significantly rose, and the river channel changed from straight to wide and slow. Furthermore, the bank area entered a cyclic “inundation–exposure” alternating state, which resulted in the adjacent section of the slope being frequently affected by the water level change. The ground surface was also affected by the water level change, and the water level was changed from straight to slow. Frequently affected by water level changes, the groundwater recharge and drainage conditions have also changed. Indeed, hydrogeological structures tend to be complex. This transition has significantly increased the frequency and sensitivity of landslide activities, such that the slope body has transitioned from its original, more stable state into a sustained deformation stage. Thus, it has become a typical water-level-sensitive slope in the reservoir area.

3. Numerical Analysis Model

3.1. Model Parameters

The Mohr–Coulomb model was utilized to research the behavior properties of the Huasushu slope. According to the results of our geological exploration, the direction of the II–II section plane represents the overall direction of the slope’s movement. Therefore, this vertical section plane was selected for simulations (Figure 2).

Field samples and local engineering experience were used to determine the model’s parameters, including the deformation modulus and the cohesion and friction angle of the sliding surface. These parameters are listed in

Table 1. Physical and mechanical parameters adopted for the finite element model of the Huasushu slope.

Part	Volumetric Weight (kN/m3)	Deformation Modulus (MPa)	Porosity (-)	Poisson Ratio (-)	Cohesion (kPa)	Internal Friction Angle (°)	Saturated Permeability Coefficient (cm/s)
Slope mass	20.0	18	0.36	0.25	70	20	1.1 × 10−2
Sliding belt	19.2	12	0.21	0.28	30	12	9.3 × 10−6
Landslide bed	24.0	17,000	0.16	0.15	1000	30	1.4 × 10−6

.

3.2. Boundary Conditions

The boundary conditions adopted for seepage in the slope under rainfall and changing reservoir water levels are shown in Figure 3. The water head boundary of the model used the submerged anterior border of the slope, and the seepage flow was caused by rainfall at the surface of the slope mass. When the rainfall intensity exceeded the seepage speed of the rock–soil mass on the slope surface, the seepage speed of the rock–soil mass was taken as the flow boundary; when the rainfall intensity was less than the seepage speed of the rock–soil mass on the slope surface, the rainfall intensity was taken as the value of the boundary flow. The bottom and sides of the model were assumed to be free seepage boundaries, while the bedrock was considered impervious due to its extremely low permeability. The boundary conditions are shown in Figure 3.

3.3. Numerical Model

3.3.1. Two-Dimensional Numerical Model

For the main section plane of the slope at Huasushu, GEO-SLOPE finite element software (v2024.2.1.28) was used to divide the slope into a finite element grid using quadrilateral elements. There were 5297 nodes and 5255 elements (see Figure 4 for the grid map). This two-dimensional simulation model had the advantage of being computationally efficient and suitable for preliminary slope stability analysis.

3.3.2. Three-Dimensional Numerical Model

The simulation results of a two-dimensional model can provide the basis for a preliminary analysis. However, the actual damage mechanisms of a slope are more complex. Therefore, a three-dimensional model was established to more accurately reflect exact characteristics and damage mechanisms. Based on the geological conditions and topographic and geomorphological characteristics of the slope, the scope of the three-dimensional numerical simulation model was selected as follows: approximately 832 m along the direction of the Yangtze River flow and approximately 768 m perpendicular to the moving direction. The calculation domain included the landslide mass, sliding belt, and bedrock and was divided into 54,510 tetrahedral elements with 61,600 nodes (see Figure 5).

Under the effects of water level changes and rainfall, the rock and soil body of a slope will undergo significant elastic–plastic deformation due to different degrees of unloading and loading processes, during which the rock body may be damaged by tension or compression shear. The yield criteria usually adopted in rock mechanics include the D-P criterion and the Mohr–Coulomb criterion [34,35,36], in which the yield surface of the D-P criterion is circular in the π-plane and conical in the principal stress space. There is no cusp problem, so it is better to deal with it; the yield surface of the Mohr–Coulomb criterion is an irregular, six-edge vertebral surface in the principal stress space, and it is an unequal angle hexagonal surface in the π-plane, so it is better to deal with it when considering the cusp. The hexagonal shape leads to some difficulties in dealing with the sharp corners. However, due to the continuous progress in numerical computation technology, the problem of sharp corners can be easily solved; thus, the composite criterion combining the Mohr–Coulomb criterion and the tensile damage criterion is used in the present calculation to determine the yield damage of the rock body under the loading effect of reservoir storage and rainfall conditions. The shape of the Mohr–Coulomb criterion in the principal stress space is shown in Figure 6, and the shape of the composite criterion in the (${\sigma }_1$, ${\sigma }_3$) plane is shown in Figure 7. Point A to point B is the Mohr–Coulomb yield criterion, $f_s$ = 0, where $f_s$ can be expressed as the equation below.

$$f_s={\sigma }_1-{\sigma }_3{N}_{\phi }+2c\sqrt{N_{\phi }}$$

(1)

• $\phi$—angle of internal friction.
• $c$—cohesive force.
• $N_{\phi }={\displaystyle \frac{1+\sin (\phi )}{1-\sin (\phi )}}$.

In Figure 4, point B to point C is the pull-down criterion, $f_t$ = 0, and $f_t$ can be expressed by the following equation:

$$f_t={\sigma }_3-{\sigma }_t$$

(2)

• ${\sigma }_t$—tensile strength.

The plastic potential function is represented by the shear plastic flow function, $g_s$, and the tension plastic flow function, $g_t$, where

$$g_s={\sigma }_1-{\sigma }_3{N}_{\psi }$$

(3)

• $\psi$—shear angle.
• $N_{\psi }={\displaystyle \frac{1+\sin (\psi )}{1-\sin (\psi )}}$

$$g_t={\sigma }_3$$

(4)

To more accurately simulate rainfall infiltration, the SEEP/W module in GeoStudio was utilized to simulate and analyze infiltration lines representing changes in groundwater levels. Using numerical simulation techniques, changes in pore water pressure within the slope were determined by analyzing infiltration lines [37]. Subsequently, the SLOPE/W module was used to integrate geotechnical parameters for limit equilibrium analysis.

3.4. Working Condition Specifications

A water level chart was constructed based on the water level dispatch of the Three Gorges Project. During the operation of the Three Gorges Reservoir, the water level periodically rises and falls. Changes in the reservoir water level are shown in Figure 8. There are two ways for the reservoir water level to fall: one is a large, slow fall before the flood season, where the reservoir water level gradually drops from 175 m to 145 m; the other is a rapid but smaller fall during the flood season, rapidly dropping from 162 m to 145 m.

The rainfall intensity in the area where the landslide occurred is shown in

Table 2. Rainfall intensity in Huasushu slope area.

Rainfall Intensity	50-Year Rainfall
Flood season	74.850 mm/d
Non-flood season	47.571 mm/d

, extracted from rainfall data for Fengjie County. Based on the water level regulation and rainfall data, computational analyses of landslides with five working conditions are carried out using the finite element method under two- and three-dimensional conditions. The specific working conditions are listed in

Table 3. Working conditions and load combinations.

Working Condition	Reservoir Water Level	Rate of Water Level Rise and Fall	Rainfall Conditions	Water Level Status
1	175 m	\	\	Static
2	175 m dropping to 145 m	1.2 m/d	\	Slow drop
3	175 m dropping to 145 m	1.2 m/d	Non-flood season, 50-year rainfall
4	162 m dropping to 145 m	2.0 m/d	\	Rapid drop
5	162 m dropping to 145 m	2.0 m/d	Flood season, 50-year rainfall

.

4. Three-Dimensional Slope Stability Analysis

4.1. Analysis of Results of Pore Pressure Simulations

Under the working conditions of Case 1 (reservoir static water level is 175 m), the results for the pore pressure shown in Figure 9a demonstrate that the pore pressure in the zone below the 175 m water level was directly affected by the reservoir’s water level.

Combining the pore pressure simulation results for Case 2 (a slow decrease in water level from 175 m to 145 m) and Case 4 (a sudden decrease in water level from 162 m to 145 m) (Figure 9b,d), the calculation results show that in the absence of rainfall, the pore pressure within the slope was solely affected by the decreased water level. The wading area at the anterior border of the landslide was the main affected zone. As the water level decreased, the seepage field at the anterior border changed significantly, and the groundwater level also decreased, resulting in the slow dissipation of pore pressure, producing stress hysteresis. As the posterior border of the slope was far away from the zone where the reservoir water level changed, the pore pressure in the central and posterior regions changed only slightly and was negligible.

For the 50-year rainfall data coinciding with a drop in water levels, the results for Case 3 (slow water level drop from 175 m to 145 m) and Case 5 (a sudden water level drop from 162 m to 145 m) (Figure 9c,e) show that the groundwater distribution at the front of the slope was affected by both the change in the reservoir water level and rainfall. By contrast, the pore pressure in the middle and rear parts of the slope was mainly affected by rainfall. However, for the entire slope mass, the changes in pore pressure were relatively small.

The pore pressure cloud diagrams for the five working conditions shown in Figure 9 indicate that a drop in the water level had a significant impact on the wading area at the anterior border of the slope, while having a lesser effect on the non-wading area in the middle and rear of the slope. Rainfall had a more significant impact on the surface layer of the slope.

4.2. Analysis of Stress Simulation Results

According to the stress cloud diagram of Case 1 (the static water level of the reservoir is 175 m) (Figure 10), when the reservoir increases to 175 m, the maximum value of the first principal stress is 2.10 × 10³ kPa, and the minimum value is −2.51 × 10³ kPa; the maximum value of the third principal stress is 2.12 × 10³ kPa, and the minimum value is −8.24 × 10³ kPa. The overall stress level is lower, the slope is relatively stable, and there is no obvious stress concentration phenomenon.

In Case 2 (the water level slowly decreases from 175 m to 145 m) and Case 4 (the water level suddenly decreases from 162 m to 145 m) (Figure 11, Figure 12 and Figure 13), the distribution of the first and third stresses of the slope body changed. The maximum value of the first principal stress decreased from 6.55 × 10³ kPa to 5.58 × 10³ kPa, and the minimum value increased from −3.83 × 10³ kPa to −4.13 × 10³ kPa for both conditions. Compared with the high water level (175 m), the stress level inside the slope decreased overall during the lower water level (162 m). This suggests that the decline in the high water level has a greater detrimental effect on slope stability, mainly because the high water level provides a greater reservoir level load, the stress redistribution is more intense when the water level declines, and the slope is more susceptible to destabilization damage. The decline in reservoir water causes the seepage field to change; the mutual coupling of the seepage field and stress field is the main reason for the change in the stress field in this case. With the continuous decline in the reservoir water level, the three-dimensional stress distribution shows the overall change trend in the internal concentration of the slope body, and the potential destabilization area of the landslide gradually becomes clear.

When the water level drops and encounters rainfall in 1 of 50 years, the stress cloud maps (Figure 12 and Figure 14) of Case 3 (the water level slowly drops from 175 m to 145 m, superimposed on rainfall in 1 of 50 years) and Case 5 (the water level suddenly drops from 162 m to 145 m, superimposed on rainfall in 1 of 50 years) are combined. The impact of Case 3 on the stress field is much larger than that of Case 5, with the maximum value of the first principal stress decreasing from 9.07 × 10³ kPa to 6.44 × 10³ kPa, and the minimum value of the third principal stress increasing from −2.21 × 10³ kPa to −1.2 × 10³ kPa, indicating that Case 3 is the most unfavorable case. The superposition effect of the rainfall leads to extremely significant tensile action within the slope body, while the high water level exacerbates the stress redistribution under the heavy rainfall. This leads to stress redistribution.

Combined with the pore water pressure results, it can be seen that there is a coupling effect between the seepage field and the stress field. The change in the reservoir water level and rainfall together change the seepage field, which triggers the corresponding change in the stress field under different working conditions.

4.3. Analysis of Results of Displacement Simulations

Figure 15a,b show the horizontal and vertical displacement under Case 1, respectively. In the water-filled state, significant horizontal displacement occurred in the middle and lower parts of the slope mass with values of 0.0002–0.0018 m. The horizontal displacement in other places is relatively small. This was mainly due to the significant deformations of the lower part of the slope caused by the inundation of the slope’s anterior border. At the same time, large-scale downward vertical displacement occurred in the middle and upper part of the slope mass with values of 0.004–0.013 m. The displacement direction of the leading edge of the slide is vertically downward, and the displacement value is 0~0.0005 m, mainly due to the steeper upper and middle part of the slide, which makes the landslide produce a large downward vertical displacement under the action of self-weight.

An analysis of the horizontal displacement cloud maps (Figure 16a, Figure 17a, Figure 18a and Figure 19a) for Cases 1 to 5 shows that the middle and lower parts of the slope mass experienced a certain degree of horizontal displacement in all of the working conditions considered, while the horizontal displacement in other areas was relatively small. When there was no rainfall, the horizontal displacement caused by the drop in the reservoir water level was more significant than that in the underwater storage conditions; i.e., the horizontal displacement under hydrostatic conditions was the smallest. This was because a drop in the reservoir water level changed the surrounding seepage field, increasing the seepage force and sliding force, which increased the horizontal displacement in the slope. After rainfall, the horizontal displacement increased significantly compared with the case without rainfall, especially in the middle and lower parts of the slope mass, where the horizontal displacement was more significant than at the upper part. This was because rainfall further increased the seepage forces acting within the slope, making the horizontal displacement in Cases 3 and 5 more significant than in in Cases 2 and 4. Case 3 had the largest horizontal displacement with a maximum value of 0.041 m.

Comparing the vertical displacement maps from Case 2 to Case 5 (Figure 16b, Figure 17b, Figure 18b and Figure 19b) shows that the slip body generates vertical displacement in all cases, and there is a significant change compared with Case 1. The direction of vertical displacement in the middle and upper parts of the slope is downward, and the direction of vertical displacement in the lower part is upward. Under rainfall conditions, the vertical displacement in Cases 3 and 5 is significantly larger than that in Cases 2 and 4 when the reservoir water level drops; especially in Case 3, the vertical downward displacement of the upper and middle parts of the slope body is the largest, reaching 0.05 m, while the vertical upward displacement of the lower part of the slide body is 0.001 m. This is mainly due to the downward slippage of the slope caused by the drop in the reservoir water level, coupled with the combined effect of self-weight and rainfall. This results in the leading edge of the slope being squeezed, thus increasing the slope’s size and vertical displacement.

4.4. Analysis of Simulation Results of Equivalent Plastic Strains

The equivalent plastic strain distribution cloud maps of the main sliding section plane for Cases 1 to 5 are shown in Figure 20. Under hydrostatic conditions, the equivalent plastic strains in the slope were mainly distributed in the middle and upper parts of the sliding belt and the leading edge of the slide zone, with a maximum value of 0.001. Under heavy rainfall conditions, the range of equivalent plastic strain distribution in Case 6 increased compared with Case 5, with the maximum value increasing from 0.003 to 0.013. The range of equivalent plastic strains in Case 3 increased compared with Case 2, and the maximum value increased from 0.005 to 0.025.

According to the numerical results for Cases 1 to 5, a landslide occurred only under Case 3 (a slow drop in reservoir water levels from 175 m to 145 m under 50 years of rainfall). Due to a drop in the reservoir water level and heavy rainfall, the seepage field within the slope mass changed significantly. The seepage force within the slope mass (caused by rainfall infiltration and the water level difference between the inside and outside of the slope mass) also increased significantly, causing significant displacement and changes in the equivalent plastic strains. Compared with the remaining four working conditions, the deformations in the slope in Case 3 increased the most. Therefore, Case 3 was the most detrimental to the stability of Huasushu slope.

5. Two-Dimensional Slope Stability Analysis

Results of Slope Stability

A two-dimensional stability analysis of Huasushu slope was carried out using the Morgenstern–Price method based on seepage field data. The stability factor for Case 1 was 1.077. A slope was considered stable when its stability factor exceeded 1.0. Case 1, therefore, represented a stable state under hydrostatic conditions; that is, the slope remained stable when the reservoir water level was not changing.

The stability coefficients under other working conditions were calculated using the finite element model, which coupled the saturated–unsaturated seepage and stress fields to analyze the impact of reservoir water level fluctuations and rainfall. The results are shown in Figure 21.

Figure 21a shows that the slope stability coefficient demonstrates a continuous decay trend with time when the reservoir level slowly decreases from 175 m to 145 m. The slope stability coefficient for the 50-year rainfall conditions decreased from 1.077 to 1.022, indicating that rainfall infiltration significantly increased the risk of slope instability. The mechanism can be attributed to the dynamic response of the seepage field triggered by the coupling of the reservoir water level decline and rainfall infiltration. Notably, the effect of rainfall on slope stability is not obvious at its early stage, showing that the effect of rainfall intensity on slope stability at this stage has a time lag characteristic.

As shown in Figure 21b, the slope stability coefficients still show a decreasing characteristic when the reservoir water level decreases from 162 m to 145 m. Comparative analysis shows that the slope stability coefficient under the 50-year rainfall conditions is significantly differentiated after 2 d. Compared with the single reservoir water level condition, the stability coefficient under the coupled rainfall condition intensifies up to 1.03, revealing the lag effect mechanism of the unsaturated matrix suction dissipation under the decreasing water level conditions.

The coupling analysis showed that the most dangerous condition for the Huasushu slope’s stability was Case 3 (a large decrease in reservoir water level from 175 m to 145 m combined with 50 years of rainfall). However, under these conditions, the minimum value of the slope stability factor was still 1.022, indicating that the slope was essentially stable.

6. Numerical Landslide Predictions

Numerical simulations demonstrated that Case 3 (the reservoir water level slowly reduced from 175 m to 145 m, accompanied by the 50-year rainfall) was the most detrimental to the Huasushu slope’s stability.

We extracted data from surface displacement monitoring points GPS2-1 and GPS2-2 (Figure 22) on the main sliding section plane of the slope and the equivalent plastic shear strains at the representative points on the sliding belt under the most unfavorable working condition (Figure 23 and Figure 24). A numerical prediction model for the deformations and failure of the slope was established by fitting these data.

To obtain a prediction model for the equivalent plastic strain, representative points A, B, and C on the sliding belt were selected for analysis, as shown in Figure 25. The trend of the equivalent plastic strain at each point is shown in Figure 26.

Comprehensive monitoring of displacement and equivalent plastic strain trends shows that under Case 3, the displacement at monitoring point GPS2-1 at the middle and upper parts of the landslide was significantly larger than at the front monitoring point, GPS2-2, indicating that the displacement was mainly concentrated in the middle and upper parts of the slope.

In Case 3, the equivalent plastic strains at the rear point, C, and central point, B, of the sliding belt both exceeded zero, while the equivalent plastic strain at the front point, A, was zero, indicating that the slope damage mainly occurred in the middle and upper parts, but an overall landslide did not occur. Further analysis shows that the rear of the slope is steeper and exhibits a larger range of tensile shear damage, while the middle of the slope is slower, and the scope of the damage is less extensive.

7. Discussion

7.1. Discussion of Landslide Mechanisms

The results of this study are consistent with the literature, verifying that water level decline and strong rainfall are the main factors triggering landslides [38,39]. The simulation shows that under a single condition of rapid water level decline, the displacement of the landslide body is small; meanwhile, when this decline is superimposed with strong rainfall, the landslide body shows critical plastic strain, and the deformation significantly increases, which is consistent with the mechanism characteristics of landslides triggered by the typical rainfall–reservoir level coupling effect. The predicted damage pattern is also consistent with the observed landslide characteristics in similar geological environments.

7.2. Model Strengths and Applicability

A three-dimensional numerical model can more accurately identify the location of slope instability than a two-dimensional analysis, especially in a region where deformation is locally concentrated and shows higher spatial resolution ability. This study shows that there is a certain threshold for triggering the landslide response: when the water level drops and the rainfall intensity exceeds the critical value at the same time, the slope will be significantly displaced. This finding is of great significance for the development of scientific warning thresholds and identifying high-risk periods.

From the perspective of engineering, the results of this study emphasize that (1) landslide risk assessment should consider both the reservoir water level and rainfall changes; (2) the strain concentration trend in the local weak zones in a slope should considered; and (3) the numerical prediction model can provide technical support for a real-time early warning system for landslides, with the prospect of engineering integration and application.

7.3. Limitations and Future Prospects

This study uses numerical simulation as the main analytical tool, which can reveal landslide evolution laws. However, it still has the following limitations: (1) the model was not calibrated and verified through physical tests or long-term field monitoring data [40]; (2) the assumption that the soil body is homogeneous does not consider the non-homogeneity and structural complexity of the soil body in reality [41]; and (3) this study focuses on a short-term strong rainfall scenario and does not explore the impact of long-term rainfall on the stability of the slope with a lag or the hysteresis effect of long-term weak rainfall on slope stability [42].

Future research can be strengthened in the following ways: combining field monitoring data and physical tests to improve simulation accuracy and extrapolation abilities; introducing soil heterogeneity parameters and exploring their influence on slope response mechanisms; and constructing dynamic early warning systems based on multi-source data (e.g., reservoir level, rainfall, and surface displacement) and combining it with weather prediction to achieve real-time hierarchical early warnings of landslide risk in reservoir areas, as well as proposals for emergency responses.

8. Conclusions

The following conclusions were drawn from our analysis of three-dimensional cloud maps of stress, displacement, pore pressure, and equivalent plastic strain for Huasushu slope and two-dimensional slope stability calculations for Cases 1 to 5:

(1) The three-dimensional deformation and failure analysis showed that the combined effect of a drop in the reservoir water level and heavy rainfall can significantly change the seepage field within this slope, leading to displacement in the slope mass. The deformations in the Huasushu slope and the increased equivalent plastic shear strains reached peak values when the reservoir water level gradually dropped from 175 m to 145 m and the 50-year rainfall occurred concurrently. Therefore, Case 3 was the most adverse for the Huasushu slope’s stability. In this case, slope displacement was mainly concentrated in the middle.

(2) In Case 3, the slope experienced local tensile–shear failure in the middle sections and the anterior border, but not an overall failure. The sliding belt in the middle and rear sections was steeper than in the anterior border, and the scope of the damage was more extensive. The horizontal displacement at surface monitoring point GPS2-1 was 5 mm with a vertical displacement of 30 mm.

(3) Our analyses demonstrated that an increase in the rate of decline in the reservoir water level will reduce the slope stability factor, which will decrease further with an increase in the intensity and duration of rainfall.

(4) According to the two-dimensional stability analysis, the most dangerous condition for the occurrence of a landslide at Huasushu is Case 3 (a slight decrease in the reservoir water level from 175 m to 145 m accompanied by 50-year rainfall). Under these conditions, the minimum value of the slope stability factor is 1.022, indicating that the landslide is basically in a stable state. The two-dimensional calculation results of the landslide further verify the correctness of the calculation results for the three-dimensional model.

(5) Comprehensive two- and three-dimensional simulations and analyses of their results showed that changes in the reservoir water level had little effect on the overall stability of the Huasushu slope, but may lead to local damage in the wading area. Meanwhile, heavy rainfall has a more significant impact on overall slope stability. Therefore, in subsequent monitoring, increased attention should be paid to heavy rainstorms.