Physical Model Experiments and Numerical Simulation Study on the Formation Mechanisms of Landslides on Gently Inclined Loess–Bedrock Contact Surfaces—A Case Study of the Libi Landslide in Shanxi Province

Abstract

Landslides on gently inclined loess–bedrock contact surfaces are common geological hazards in the northwestern Loess Plateau region of China and pose a serious threat to the lives and property of local residents as well as sustainable regional development. Taking the Libi landslide in Shanxi Province as a case study (with dimensions of 400 m × 340 m, maximum thickness of 35.0 m, and volume of approximately 3.79 × 10⁴ m³, where the slip zone is located within the highly weathered sandy mudstone layer of the Upper Shihezi Formation of the Permian System), this study employed a combination of physical model experiments and numerical simulations to thoroughly investigate the formation mechanism of gently inclined loess landslides. Via the use of physical model experiments, a landslide model was constructed at a 1:120 geometric similarity ratio in addition to three scenarios: rainfall only, rainfall + rapid groundwater level rise, and rainfall + slow groundwater level rise. The dynamic changes in the water content, pore water pressure, and soil pressure within the slope were systematically monitored. Numerical simulations were conducted via GEO-STUDIO 2012 software to further verify and supplement the physical model experimental results. The research findings revealed that (1) under rainfall conditions alone, the landslide primarily exhibited surface saturation and localized instability, with a maximum displacement of only 0.028 m, which did not lead to overall instability; (2) under the combined effects of rainfall and rapid groundwater level rise, a “sudden translational failure mode” developed, characterized by rapid slope saturation, abrupt stress adjustment, and sudden overall instability; and (3) under conditions of rainfall and a gradual groundwater level rise, a “progressive translational failure mode” emerged, experiencing four stages: initiation, development, acceleration, and activation, ultimately resulting in translational sliding of the entire mass. Through a comparative analysis of physical model experiments, numerical simulation results, and field monitoring data, it was verified that the Libi landslide belongs to the “progressive translational failure mode”, providing important theoretical basis for the identification, early warning, and prevention of such types of landslides.

1. Introduction

Rainfall is one of the main factors triggering landslides. Zuo Zibo et al. studied the mechanism of rainfall-induced landslides in accumulation bodies with different gradations through model tests. They found that gradation has a significant impact on landslide initiation [1]. Lin Hongzhou et al. investigated the effect of rainfall characteristics on the instability of soil slopes and reported that rainfall intensity and duration are key factors [2]. Shi Zhenming et al. conducted model tests to analyze accumulated layer landslides, revealing the impact of rainfall infiltration on landslide stability [3]. Li Zhuo et al. studied the influence of antecedent rainfall on slope landslides, emphasizing the importance of the cumulative rainfall amount [4]. Wang Bin et al. simulated the process of intense rainfall-induced landslides in accumulated materials via model tests and explored the critical conditions for landslide initiation [5]. Wu et al. experimentally studied the instability characteristics of loess slopes induced by rainfall [6]. Ching-Chuan et al. examined the responses of the soil moisture content and pore water pressure to rainfall-induced shallow landslides [7]. Chueasamat et al. employed a 1 g physical slope model to study rainfall-induced slope instability [8]. Cui et al. experimentally investigated the migration characteristics of fine particles in widely graded unconsolidated soil under heavy rainfall conditions [9]. Honghua et al. studied the disaster mechanism of completely weathered granite landslides induced by extreme rainfall [10]. These studies have laid the foundation for understanding the relationship between rainfall and landslides, but they lack consideration of the specificity of landslides on gently inclined loess–bedrock contact surfaces. Most physical model experiments for landslide analysis involve the adoption of steeper slopes, whereas research on gently inclined slopes (typically less than 15°) is relatively scarce. The instability mechanism of gently inclined slopes differs significantly from that of steep slopes and involves more complex stress states and deformation processes. Under gently inclined conditions, factors such as the creep characteristics of loess and progressive failure may fulfill more important roles.

Zhang et al. proposed a generalized early warning criterion based on the Deformation Probability Index (DPI) for landslide risk assessment, which provided a new evaluation indicator for landslide monitoring [11]. Additionally, Zhang et al. developed a new early warning criterion based on the Deformation Standardized Anomaly Index for landslide movement assessment, further improving the early warning system [12]. Yang Xiaohui et al. investigated the coupled effect of earthquakes and rainfall on accumulated material landslides via model tests, revealing the synergistic effects of these two factors [13]. F et al. conducted model tests to study rainfall-induced landslides in loose soil bodies in the Wenchuan earthquake area [14]. Liu Peng et al. performed model test research on the initiation mechanism of shallow landslides triggered by rainstorms and proposed new initiation criteria [15]. Chen Lixin et al. studied the hydraulic distribution patterns and initiation mechanisms of typical translational landslides, providing theoretical support for early warning of landslides [16]. Zhao Quanli et al. revised the initiation criteria for translational landslides, thus improving the prediction accuracy [17]. Iverson et al. investigated the acute sensitivity of landslide rates to the initial soil porosity, offering a new perspective for understanding landslide initiation processes [18]. While these studies have provided important reference data for understanding landslide initiation processes, they lack sufficient consideration of the unique characteristics of gentle slope loess landslides. Wang Zhihua et al., via the use of the Fengdian landslide as an example, studied a geomechanical model of gentle slope landslides, providing new insights for landslide stability analysis [19]. Yang Zhongkang et al. conducted stability analysis and threshold research on gentle slope loess landslides, focused on the Liaoji village landslide, and explored the critical conditions for landslide initiation [20]. Li Lin et al. studied risk monitoring and early warning of landslide–debris flow disaster chains via experimental simulations, thereby offering new approaches for comprehensive disaster prevention and mitigation [21]. Liu Linan et al. investigated the rainfall-induced mechanism underlying Jialanpute loess landslide–debris flow in Yili, Xinjiang, revealing the transformation relationship between landslides and debris flows in loess areas [22]. Li et al. performed physical model experiments to assess the risk of rainfall-induced debris landslides [23]. Acharya et al. studied the impact of shallow landslides on the sediment supply [24]. Although these studies have provided important reference data for understanding the landslide initiation process, limitations still exist. The formation of landslides on gently inclined loess–bedrock contact surfaces is typically a long-term, slow process involving the cumulative effect of multiple rainfall events, seasonal changes, and other factors. However, most of the aforementioned studies involving physical model experiments have focused mainly on single heavy rainfall events or short-term effects, making it difficult to simulate long-term evolutionary processes.

Dou Xiaodong et al. investigated the failure mechanism of deep accumulated ancient landslides via laboratory model tests, providing novel insights into understanding landslides under complex geological conditions [25]. Chu et al. studied the deformation behavior and evolution process of multislip band landslides via physical model tests [26]. Zhu Yuanjia et al. conducted numerical simulations of gentle slope landslides under intermittent rainfall and explored the impacts of rainfall characteristics on landslide stability [27]. Yongshuai et al. combined model tests and numerical simulations to study rainfall-induced slope instability processes [28]. Lee et al. simulated rainfall-induced landslides via full-scale flume tests [29]. While these studies have offered new tools for understanding landslide processes under complex rainfall conditions, physical model tests are needed to verify gentle slope loess landslides. Wang Li et al. studied the infiltration characteristics and deformation mechanisms of rainfall-induced landslides in the Three Gorges Reservoir area via one-dimensional and two-dimensional model tests, offering novel insights into the influence of hydrogeological conditions on landslides [30]. Yang et al. elucidated the hydrological mechanisms and thresholds for rainfall-induced landslides via in situ monitoring and unsaturated slope stability analysis [31]. Meng Zhenjiang et al. conducted model tests to study rainfall-induced loess landslides with preset joints, thereby revealing the influences of joints on landslide occurrence and development [32]. Guodong et al. assessed the impact of loess slope excavation under long-term rainfall conditions via model tests [33]. Matziaris et al. used centrifuge models to study the thresholds for rainfall-induced landslides on sandy slopes [34]. In terms of numerical simulation technology, Zhang et al. employed an SVR-based ensemble model to predict landslide displacements in reservoir areas, enhancing prediction accuracy through input parameter optimization, which provided a new approach for landslide numerical simulation [35]. While these studies have offered new perspectives to better understand landslide mechanisms under specific conditions, existing physical model experiments have focused mostly on the impact of surface rainfall infiltration on landslides, with relatively insufficient simulations of dynamic processes such as groundwater level changes and deep groundwater movement. Particularly for landslides on gently inclined loess–bedrock contact surfaces, long-term changes and seasonal fluctuations in groundwater may play crucial roles in their formation and development, but complex hydrological processes can rarely be reproduced accurately in short-term physical model experiments.

Via a review of existing research, current studies have focused primarily on general landslides or steep slope landslides, with insufficient consideration of the specificity of gently inclined slopes (slope < 15°), thus neglecting the unique mechanical properties and deformation processes of gently inclined slopes (<15°). Although gently inclined slopes have relatively low gradients, they still exhibit significant instability due to the unique properties of loesses: loesses are characterized by large pores and well-developed vertical joints, making it susceptible to softening and disintegration upon water exposure, leading to a significant reduction in strength; additionally, weak interlayers often develop at the loess–bedrock contact surface, which can form preferential seepage channels under long-term hydraulic action. Additionally, most studies have focused on the impact of short-term heavy rainfall, with insufficient research on the deformation evolution process of landslides on gently inclined loess–bedrock contact surfaces under long-term continuous rainfall conditions. Furthermore, existing physical model experiments have focused largely on the effect of surface rainfall infiltration on landslides, whereas simulations of dynamic processes such as groundwater level changes and deep groundwater movement are lacking. On the basis of the current research status and existing problems, in this study, physical model experiments were conducted to analyze the formation mechanism of gently inclined loess landslides, using the Libi landslide in Shanxi Province as an example. A physical model suitable for capturing the characteristics of landslides on gently inclined loess–bedrock contact surfaces was constructed to simulate the effects of different rainfall conditions and groundwater level changes on landslide stability. This study aimed to examine the deformation evolution process of gently inclined loess landslides under long-term rainfall and to reveal the critical conditions for landslide initiation.

2. Overview of the Libi Landslide

2.1. Structural Characteristics of the Libi Landslide

The Libi landslide is located on the southern side of the Libi Coal Mine Preparation Plant in Qinshui County, Shanxi Province (112°17′50″ E, 35°41′46″ N), situated at the western foot of the Taihang Mountains and the eastern margin of the Qinshui Basin. The area has convenient transportation access, being approximately 2 km west of Provincial Road S340, 15 km from Qinshui County town, and 45 km east of Jincheng City. The regional transportation location and panoramic photograph of the landslide are shown in Figure 1. The study area is characterized by a temperate continental monsoon climate, with an average annual temperature of 11.3 °C and an average annual precipitation of approximately 600 mm, with rainfall primarily concentrated between July and September. The landslide, which occurred in 2021, is classified as a gentle slope loess landslide. The rear edge of the landslide exhibits an elevation of 843.6 m, whereas the front edge exhibits an elevation of 764.5 m, yielding a relative height difference of 79.1 m. The maximum thickness of the landslide body is approximately 35.0 m, with a volume of approximately 3.79 × 10⁴ m³, categorizing it as a large-scale deep-seated landslide. The landslide demonstrates an elliptical shape in plan view, with a front width of approximately 340 m and a length of approximately 400 m along the main sliding direction, which is 23° west of north. The overall slope of the landslide is gentle, varying between 15° and 20°. The landslide mass primarily comprises loose accumulation layers, including Quaternary Holocene artificial fill (Q₄^ml), loess-like soil (Q₄^dl+pl), upper Pleistocene silty clay (Q₃^dl+pl), silty clay with gravel (Q₃^dl+pl), and boulders (Q₃^dl+pl). According to drilling data and laboratory test results, the characteristics of soil layers from top to bottom are as follows: the uppermost layer is artificial fill, 8–10 m thick, grayish-brown in color, primarily consisting of gravel mixed with silty clay and exhibiting a loose structure, with a natural unit weight of 17.8 kN/m³ and water content of 18.2%; beneath this is a loess-like soil layer, 5–7 m thick, light yellow in color, characterized by large pore structure, with a natural unit weight of 16.5 kN/m³, water content of 16.8%, liquid limit of 28.5%, and plastic limit of 17.2%; below this is a silty clay layer, 4–6 m thick, brownish-yellow in color, with a natural unit weight of 18.5 kN/m³, water content of 20.1%, liquid limit of 32.4%, and plastic limit of 19.8%; the lowest layer is silty clay with gravel, 3–5 m thick, grayish-brown in color, with a natural unit weight of 19.2 kN/m³, water content of 19.5%, and particles < 0.075 mm accounting for 65% of the content (Figure 2). The rear part of the sliding mass mainly comprises silty clay and silty clay with gravel, whereas the front part is dominated by boulders and cobbles. The entire sliding mass exhibits a thickness ranging from 20 to 35 m. The slip zone is located within the highly weathered sandy mudstone layer of the Upper Shihezi Formation of the Permian System. This rock layer has exhibited long-term weathering and softening due to groundwater saturation. The rock mass is extremely fragmented, with the increased clay content causing a significant reduction in the shear strength. This layer forms a weak interlayer component at the top of the sliding bed, providing the primary material basis for landslide development. The sliding bed mainly encompasses moderately weathered sandy mudstone. This rock layer has retained its original structure, with developed joints and fissures, and the rock core exhibits a mostly short columnar shape. Borehole data indicate that the sliding bed is buried at a depth of approximately 35 m, with significant variations in the inclination angle. The rear part is relatively steep (15°–20°), whereas the middle and front parts exhibit gentler slopes (5°–8°), forming a stepped profile. This stepped bedrock topography provided important geological conditions for the occurrence of this landslide.

2.2. Analysis of the Major Factors Influencing the Formation of the Libi Landslide

The formation of the Libi landslide is not only closely related to internal conditions such as the site stratigraphy, lithology, geological structure, and geomorphology but also greatly influenced by external factors such as rainfall infiltration and human engineering activities. The geological conditions of the Libi landslide constitute a crucial internal factor in its formation. From a stratigraphic and lithological perspective, Quaternary loose deposits are widely distributed across the site, comprising from top to bottom, an artificial fill layer (8–10 m), mainly consisting of gravel mixed with silty clay, exhibiting a loose structure and a natural unit weight of 17.8 kN/m³; a loess-like soil layer (5–7 m), characterized by large pore structure and a natural unit weight of 16.5 kN/m³; a silty clay layer (4–6 m), with high compressibility and a natural unit weight of 18.5 kN/m³; and a silty clay layer with gravel (3–5 m), having a natural unit weight of 19.2 kN/m³. These soil layers generally exhibit characteristics of easy softening upon water exposure and low shear strength. The bedrock consists of sandy mudstone from the Upper Shihezi Formation of the Permian System, with its surface being highly weathered, forming a weak interlayer at the top of the sliding bed. This rock layer, having undergone long-term weathering and groundwater saturation, is extremely fragmented with increased clay content, resulting in significantly reduced shear strength. The fresh bedrock primarily comprises moderately weathered sandy mudstone, retaining its original structure but with well-developed joints and fissures. This soft-over-hard structural characteristic provided favorable conditions for the formation of the sliding surface. Regarding the tectonic background, the Libi mining area is located on the eastern wing of the Qinshui syncline, where regional tectonic movements have led to rock layer folding and deformation. The rock layer dip direction basically coincides with the slope direction, forming a typical “dip slope” structure, which makes interlayer sliding more likely to occur.

The study area is located at the western foot of the Taihang Mountains, geomorphologically belonging to the loess hilly region. The landslide mass exhibits an elliptical distribution, with a front width of approximately 340 m and a length of approximately 400 m along the main sliding direction. While the overall slope is gentle, its internal topography shows significant variations. The rear edge has an elevation of 843.6 m, while the front edge is at 764.5 m, yielding a relative height difference of 79.1 m. Multiple gullies have developed within the area, displaying a dendritic distribution pattern with typical cutting depths of 3–5 m. The presence of these gullies not only increases the fragmentation of the terrain but also provides preferential pathways for rainfall infiltration. Floodplains have developed in the middle and lower portions of the slope, facilitating surface water convergence and creating conditions favorable for groundwater recharge. The slope gradient distribution exhibits a distinct zonation: upper section (15°–20°): relatively steep, susceptible to local instability; middle section (8°–15°): more gentle, constituting the main sliding zone; lower section (5°–8°): most gentle, but significantly affected by groundwater. This topographical configuration makes the upper rock and soil mass prone to downward movement under gravitational forces.

Meteorological conditions serve as a crucial triggering factor for the instability of the Libi landslide. The study area is characterized by a temperate continental monsoon climate, with an average annual precipitation of approximately 600 mm, showing extremely uneven distribution throughout the year, primarily concentrated between July and September. The cumulative precipitation from January to September 2021 reached 957.44 mm, exceeding the average for the same period in normal years by more than 30%. Notably, continuous rainfall occurred for four days from 12 to 15 July, with an average daily rainfall of 85 mm and a maximum rainfall intensity of 32 mm/h. Another period of sustained heavy rainfall occurred from 23 to 26 September, with a four-day cumulative rainfall reaching 312 mm, a maximum daily rainfall of 156 mm, and a maximum rainfall intensity of 38 mm/h, all exceeding the local once-in-a-century storm standard (daily rainfall of 125 mm). The groundwater conditions are complex, primarily comprising Quaternary pore water and bedrock fissure water. Monitoring data indicate that the groundwater level exhibits distinct seasonal variation characteristics: early May: average water level depth of 15.6 m; June–July: water level rapidly increased by 4.8 m due to heavy rainfall; late July: reached peak level, followed by gradual decline; late October: returned to 16.8 m. A significant positive correlation exists between groundwater level changes and landslide displacement rates, indicating that groundwater is a key factor controlling landslide movement. Rainfall infiltration affects landslide stability through two pathways: rapid infiltration through surface fissures and gullies, leading to increased water content and reduced strength in shallow soil layers, and groundwater recharge, resulting in water level rise and increased pore water pressure. The coupled effect of these two processes serves as the primary cause triggering accelerated landslide deformation.

Recent human engineering activities in the study area primarily include the following: highway construction initiated in 2019, involving excavation and filling operations in the upper portion of the slope, which altered the original topography and stress state; mining activities at the Libi Coal Mine on the northern side of the landslide, which created mined-out areas and induced surface subsidence, disrupting the original geological structure; and slope excavation and site leveling works during the construction of the coal preparation plant’s industrial site, which increased the slope load. The cumulative effects of these engineering activities have altered the stress distribution and drainage conditions of the slope, constituting significant anthropogenic factors that have exacerbated landslide instability.

The land use types in the study area primarily comprise woodland, bare land, and construction land. The woodland is mainly distributed in the upper portion of the slope, dominated by shrubs and sparse forests, with relatively low vegetation coverage (<30%), shallow root systems, and limited protective capacity. The middle and lower portions are predominantly bare land, lacking vegetation protection, and prone to surface runoff and soil erosion under rainfall conditions. The construction of industrial sites has altered the original land use patterns, increasing impervious surface area and concentrating surface runoff. Site leveling and ground hardening treatments have modified the original infiltration conditions, leading to increased groundwater recharge in localized areas.

To understand the deformation development pattern of the landslide mass, both surface and deep displacement monitoring points were established in the landslide area. Surface displacement monitoring included 14 monitoring points. During the period from November 2020 to October 2021, the maximum cumulative displacement of the slope reached 68.9 mm, with a maximum settlement of 5.5 mm, generally exhibiting greater displacement on the eastern side compared to the western side. Regarding deep displacement monitoring, six and fourteen inclinometer holes were installed in June and August 2021, respectively, with depths ranging from 15.0 to 35.0 m. The first batch of inclinometer monitoring showed a maximum displacement of 30 mm located in the middle surface area; the second batch recorded a maximum displacement of 44 mm, similarly occurring in the middle surface position. The monitoring results indicated that the landslide deformation exhibited significant spatial variation, characterized by “larger in the east, smaller in the west, and maximum in the middle”, with deformation rates significantly increasing during rainfall periods.

The formation of the Libi landslide results from the combined effects of multiple factors. The intrinsic geological and geomorphological conditions provide the material basis for landslide development, anomalous meteorological and hydrological conditions serve as direct triggering factors for landslide instability, and human engineering activities and inappropriate land use have exacerbated landslide development. The coupled effects of these factors ultimately led to the landslide’s transition from a slow creep to a sudden phase of instability.

3. Physical Model Test Design

3.1. Physical Model Test Methods

Physical model testing serves as an essential means for investigating the formation mechanism of landslides on gently inclined loess–bedrock contact surfaces. Through scaled simulation of prototype landslides under laboratory conditions, this approach enables an in-depth study of the influence mechanisms of rainfall infiltration and groundwater level changes on landslide stability. The 1 g indoor model test method was adopted, and a physical model with a geometric similarity ratio of 1:120 was constructed based on similarity theory principles. The test design comprehensively considered the characteristics of the Libi landslide, including large landslide body thickness (20–35 m), rainfall and groundwater level changes as primary triggering factors, and distinct phases in deformation development. Regarding the similarity ratio design, key similarity criteria were established, including geometric similarity ratio (1:120), gravitational acceleration similarity ratio (1:1), and density similarity ratio (1:1), which formed a complete set of similarity relationships. For model materials, a loess from the prototype landslide was selected and processed through sieving and moisture content adjustment to ensure that the physical and mechanical properties of the model materials closely matched those of the prototype soil mass. Boundary conditions were controlled using a rigid model box, with a bottom drainage system to simulate groundwater level changes and an open-top design to facilitate rainfall simulation. The monitoring system configuration established a multi-parameter monitoring system, including pore water pressure, earth pressure, and water content sensors, enabling real-time monitoring of the model’s internal state.

3.2. Design of the Model Box and Rainfall System

A model box was constructed using durable square steel tubes as the main structural support, combined with highly transparent tempered glass. The overall framework was assembled via precise welding techniques, thus ensuring sufficient strength and observational convenience. The dimensions of the model box are 4.0 m in length, 2.2 m in width, and 2.0 m in height, fully considering the experimental requirements and operational space. Inside the box, a partition wall was built with machine-made bricks and concrete, thereby dividing the entire space into two equal-sized sections (4.0 m × 1.0 m × 2.0 m). This design allows two sets of model tests to be conducted simultaneously, which increases the experimental efficiency. The bottom and back of the box were constructed with steel plates nearly 1 cm thick, not only guaranteeing the overall strength of the box but also ensuring favorable waterproofing performance. An open design was adopted for the top to facilitate the implementation of simulated rainfall processes. Both sides were sealed with 12 mm thick tempered glass, with grid lines drawn on the outer surface to facilitate soil deformation observation and recording. To minimize the influence of boundary effects on the formation of the sliding surface, a layer of silicone oil was uniformly applied to the inner side of the tempered glass, thereby reducing sidewall friction. All the gaps in the model box were sealed with silicone sealant to prevent leakage during the experiment. Two drainage pipes were installed in the lower part of the box to drain surface water, thus simulating natural drainage conditions. To simulate groundwater level rise, a water tank was placed in the upper part of the slope on the right side of the model box. The groundwater level can be adjusted by controlling the amount of water in the tank, as shown in Figure 3. To achieve precise control of groundwater levels, the model incorporated a complete groundwater control system. A permeable layer with a thickness of 5 cm was installed 10 cm above the sliding bed surface, constructed using gravel with particle sizes of 2–5 mm to serve as a groundwater seepage channel. The water supply system included a 0.5 m³ elevated water tank, with its bottom connected to the permeable layer via pipelines, allowing water supply pressure to be controlled by adjusting the water level height in the tank. Adjustable overflow pipes were installed at the bottom of the model box, enabling groundwater level control through modifications to the outlet height of these pipes. The initial groundwater level was set at 15 cm above the sliding bed surface (equivalent to 18 m in the prototype according to the 1:120 similarity ratio), which closely matched the natural groundwater level observed in field monitoring. Two methods were employed for controlling groundwater level fluctuations: for scenario two (rapid rise), the rapid water level increase was achieved by increasing the water supply rate while simultaneously raising the overflow pipe outlet; for scenario three (gradual rise), a constant-flow pump was used to control the water supply rate, with the overflow pipe height adjusted gradually according to predetermined rates to ensure a slow and stable water level rise.

The rainfall system primarily encompasses three parts: a control system, a water supply system, and a rainfall system. The control system provided the use of intelligent automatic adjustment technology, enabling precise control of the rainfall intensity within a range of 10 to 200 mm/h, with the set rainfall duration ranging from 0 h to 600 h. The system was equipped with three types of nozzles for producing light, moderate, and heavy rain, which can be used individually or in combination to meet experimental requirements for different rainfall intensities. The rainfall collection subsystem, connected to the control system via rain gauges, achieved real-time rainfall monitoring and self-checking functions, with measurement errors controlled within 2%, thus ensuring the accuracy of the experimental data. The water supply system includes components such as a 220 V AC power supply system, water pumps, a 1 m³ water storage tank, and water supply pipelines, which can source water from near the laboratory and deliver it to the rainfall system. The rainfall system comprises 2 × 2 rainfall units arranged in a rectangular distribution, with each unit containing three nozzles for producing light, moderate, and heavy rain. The nozzles are positioned 2.5 m above the bottom of the model box, ensuring a uniform distribution of raindrops on the simulated landslide surface. The system design facilitated the generation of separate rainfall events on both sides of the model box, enabling the simulation of different rainfall intensity conditions. The rainfall intensity was controlled by calibrating the pressure, achieving a precise simulation of the rainfall process.

3.3. Monitoring and Data Acquisition System

The monitoring and data acquisition system was designed to comprehensively and accurately monitor the dynamic changes in the slope under the effects of rainfall and groundwater. The monitoring system primarily comprises three types of sensors and corresponding data acquisition equipment used to monitor the pore water pressure, earth pressure, and volumetric water content. The parameters of the monitoring and data acquisition system are summarized in

Table 1. Statistical parameters of the monitoring and data acquisition system.

System Component	Equipment Type	Main Parameters	Functional Description
Pore water pressure monitoring	BWMK pore water pressure meter	1. Range: −50~50 kPa	Precise measurement of pore water pressure changes
		2. Dimensions: 21 mm (length), 15.8 mm (diameter)
		3. Accuracy: Error ≤ 0.3 F·S
		4. Sensitivity: 0.2 mV/kPa(2 V bridge voltage) 5. Response time: ≤0.1 ms
Earth pressure monitoring	BWM earth pressure sensor	1. Dimensions: 28 mm (diameter), 6.5 mm (thickness)	Precise measurement of soil pressure changes
		2. Range: −200~200 kPa
		3. Accuracy: Error ≤ 0.2 F·S
		4. Sensitivity: 0.2 mV/kPa (2 V bridge voltage) 5. Overload capacity: 200% F·S
Volumetric water content monitoring	BWM volumetric water content meter	1. Dimensions: 70 mm (length), 40 mm (width), 10 mm (thickness) 2. Probe length: 80 mm 3. Range: 0–100% 4. Measurement accuracy: ±1% (0–50%) 5. Response time: ≤1 s	Comprehensive coverage of the water content range throughout the experiment
Data acquisition (pore water pressure and earth pressure)	YBY-4010 strain analysis system	1. Number of channels: 40	Simultaneous monitoring of pore water pressure and earth pressure data
		2. Sampling frequency: Once every 3 s
		3. Storage capacity: 100,000 data points 4. Resolution: 24-bit 5. Signal amplification: Adjustable from 1 to 10,000
Data acquisition (volumetric water content)	YB-R485 data acquisition instrument	1. Number of channels: 20 2. Sampling frequency: Once every 3 s 3. Storage capacity: 100,000 data points 4. Resolution: 16-bit 5. Signal input range: 0–5 V	Specifically used for collecting volumetric water content data

. Prior to the experiments, systematic calibration was performed on all sensors. For pore water pressure meters, pressure was applied at 5 kPa intervals within the range of 0–50 kPa, and output voltage values were recorded, with measurements repeated three times to obtain average values for fitting the standard curve. For earth pressure sensors, calibration was conducted using standard weights for graduated loading, with measurements taken at 20 kPa intervals within the range of 0–200 kPa and output voltage values recorded, similarly repeated three times for averaging. For volumetric water content meters, calibration was performed using standard soil samples with known water contents (5%, 10%, 15%, 20%, 25%), where sensor probes were fully inserted into the samples and readings recorded, with three tests conducted for each water content to establish the correlation between water content and output signals. The data acquisition system underwent no-load testing before sensor connection, with the sampling frequency set to once every 3 s to verify the normal operation of all acquisition channels.

3.4. Similarity Analysis of the Landslide Physical Model

The 1 g indoor model test method was adopted, with the geometric similarity ratio between the test model and the physical model set to 1:120. The setting of this ratio accounts for both the feasibility of the experiment and the adequate representation of the main characteristics of the prototype landslide. The similarity ratio for both the gravitational acceleration and test density was set to 1:1 to maintain the maximum mechanical similarity between the model and the prototype.

On the basis of the first theorem of similarity, key parameters such as the length, width, and height of the onsite landslide were precisely measured, and the corresponding test model was constructed according to the above 1:120 scale ratio. To meet the kinematic and dynamic similarity requirements, soil samples collected from the study area were employed as model materials. These samples were processed via crushing and sieving to ensure that the physical and mechanical properties of the model materials remained as close as possible to those of the prototype soil. The water content in the model material was controlled at 15%, with a density of 1.64 g/cm³, and the density similarity ratio was controlled at 1 during model construction. The similarity matrix is detailed in

Table 2. Establishment of rainfall scenarios.

Physical Quantity	Dimension	Similarity Ratio
Length	L	1:120
Area	L2	1:1202
Volume	L3	1:1203
Density	ML−3	1:1
Gravitational acceleration	LT−2	1:1
Friction angle	/	1:1
Stress	ML−1T−2	1:120
Rainfall intensity	LT−1	$1:\sqrt{120}$
Time	T	$1:\sqrt{120}$

.

3.5. Rainfall Working Condition Selection

According to meteorological data from Qinshui County, the cumulative precipitation from January to September 2021 reached 957.44 mm, which is more than 30% greater than the average value for the same period in normal years. Notably, two consecutive once-in-a-century extreme rainstorms occurred in July and September. Considering the impacts of rainfall and groundwater level changes on landslides at different time scales, three typical scenarios were designed for the experiment. Scenario one aimed to simulate the effects of short-term intense rainfall, scenario two aimed to model rapid groundwater level fluctuations under extreme weather conditions, and scenario three aimed to capture the cumulative effect of long-term rainfall and gradual groundwater level changes. The rainfall amounts and rainfall schemes are detailed in

Table 3. Designed rainfall duration and rainfall amount.

Date	Rainfall Duration (h)	Rainfall Intensity (mm/h)	Rainfall Amount (mm)
30 January 2024/13:20	0.2	10	2
30 January 2024/15:20	0.25	16	4
30 January 2024/17:10	0.3	16	4.8
31 January 2024/09:00	0.3	10	3
31 January 2024/11:20	0.4	10	4
31 January 2024/13:30	0.4	26	10.4
31 January 2024/15:40	0.25	10	2.5
Total	2.1	/	30.7

.

Scenario one: Rainfall only

On the basis of actual rainfall data from Qinshui County for September 2021, a scaled design was implemented with a similarity ratio of 1:120. The rainfall scheme included seven rainfall events distributed over two days, with a total rainfall amount of 30.7 mm.

Scenario two: Rainfall + rapid groundwater level fluctuations

While maintaining the same rainfall conditions as those under scenario one, the factor of rapid groundwater level fluctuations was introduced in this scenario. The water level rapidly increased via the use of the water tank at the rear of the model box until slope failure occurred. This scenario aimed to simulate the impact of sudden groundwater level changes on landslide stability under extreme rainfall conditions. It focused on the abrupt changes in the internal stress state of the slope during rapid groundwater level fluctuations and the potential for triggering sudden sliding.

Scenario three: Rainfall + gradual groundwater level rise

This scenario involved the same rainfall conditions as those under the previous two scenarios but aimed to simulate a gradual increase in the groundwater level. The water level was controlled via the use of the water tank, thereby first increasing to 10 cm, then decreasing, and finally rising again, This cycle was repeated until the water level eventually reached 20 cm. Scenario three lasted for 6 days, with the rainfall amount and timing identical to those under scenarios one and two. Starting from the third day, a slow water addition method was employed to simulate the progressive increase in the groundwater level.

3.6. Sensor Layout

The sensor design comprehensively accounted for the quantity of available sensors and the limitations of the data acquisition equipment while also considering the requirements of the different scenarios. The layout is shown in Figure 4 and Figure 5. The specific arrangement plan is as follows:

Scenario one (rainfall only): Two layers of sensors were arranged in the middle cross section, with one group placed in the upper, middle, and lower parts of the slope for each layer. The upper layer sensors were buried at a depth of 10 cm, while the lower layer sensors were buried at a depth of 20 cm, resulting in a total of six groups. Each group included one pore water pressure sensor, one earth pressure sensor, and one moisture content sensor, resulting in eighteen sensors in total.

Scenario two (rainfall + rapid groundwater level rise): Two layers of six sensor groups were arranged on both the left and right sides of the slope surface, totaling twelve groups and thirty-six sensors.

Scenario three (rainfall + gradual groundwater level rise): Two layers of six sensor groups were arranged on the left, center, and right sides of the slope, totaling eighteen groups and fifty-four sensors.

3.7. Experimental Procedure

To ensure smooth experiment execution and data accuracy, a detailed experimental procedure was developed, which can be described as follows:

(1) Soil Selection

A loess from the Libi landslide site in Shanxi was selected as the experimental material to better simulate the mechanical behavior of the actual landslide.

(2) Soil Sieving

The collected loess samples were naturally dried in the laboratory and then sieved via standard sieves, as shown in Figure 3. A sieve with a 2 mm aperture was employed to remove larger gravel and plant roots, ensuring soil sample uniformity (Figure 6). The basic physical properties of the sieved soil samples, including water content, density, and liquid and plastic limits, were determined to ensure consistency with the characteristics of the prototype landslide soil. The soil parameters are as follows: natural unit weight, 18.5 kN/m³; saturated unit weight, 19.0 kN/m³; cohesion, 11 kPa; internal friction angle, 9.86°; Poisson’s ratio, 0.3; liquid limit, 30.2%; and plastic limit, 19.1%.

(3) Landslide Model Construction

Based on the scaled landslide profile and prototype stratigraphic characteristics, the model construction began with building the basic shape of the sliding bed using concrete at the bottom, setting the slope gradient according to the prototype scale ratio with 15°–20° in the rear part and 5°–8° in the middle and front parts. Subsequently, a fine sand layer (approximately 2 cm thick) was placed as a permeable layer, followed by the loess-like soil layer (approximately 28 cm thick). The filling process employed a layer-by-layer compaction method, with each layer subdivided into 2–3 thin layers (single layer thickness controlled within 5 cm). Meanwhile, the physical and mechanical parameters of each soil layer were strictly controlled to ensure the model could accurately reflect the stratigraphic structural characteristics and mechanical properties of the prototype landslide. To ensure proper bonding between layers, the surface of the lower soil layer was loosened before subsequent filling.

(4) Sensor Installation

Various sensors were installed according to the previously designed layout plan. A small pit was first established on the soil surface at each predetermined position, and the sensor was placed and carefully covered with surrounding soil. Then, the establishment of the upper soil layer continued. In the filling process, the compaction degree of the soil around the sensors should be consistent with that of the overall soil to prevent local nonuniformities from affecting the measurement results. Pore water pressure sensors and earth pressure sensors were installed with their sensing surfaces facing upwards, whereas moisture content sensors were maintained with their probes in full contact with the soil. All sensor cables were carefully led out of the model box to prevent disturbance to the soil body in the subsequent tests.

(5) System Debugging

The connections of all the sensors were inspected to ensure that each sensor was correctly connected to the data acquisition system. The data acquisition system was started, and each sensor was calibrated and assessed to ensure that it functioned normally and output accurate data. The data acquisition frequency and data storage capacity were adjusted to ensure continuous uninterrupted data recording during long-term experiments. In accordance with the designed rainfall scheme, the position and water pressure of the nozzles were finely adjusted to ensure a uniform rainfall distribution on the model surface. Via calibration tests, the nozzle opening combinations and pressure settings for achieving different rainfall intensities were determined to simulate the designed rainfall scenarios accurately. The sealing of the water tank and pipe connections was inspected, and multiple water addition tests were conducted to ensure precise control of groundwater level changes. Particularly for scenarios aiming to simulate slow groundwater level rise, the water addition rate was carefully calibrated to ensure that the expected water level change process could be achieved.

(6) Rainfall

In accordance with the previously designed rainfall plan, an intelligent control system was employed to precisely control the rainfall intensity and duration. The rainfall process was divided into seven events, with a total rainfall amount of 30.7 mm distributed over two days. A combination of nozzles for producing light, medium, and heavy rain was used to ensure a uniform rainfall distribution on the model surface. In the rainfall process, rain gauge data were obtained in real time to ensure that the rainfall intensity and cumulative rainfall matched the designed values.

(7) Groundwater Level Application

Scenario 2 (rapid fluctuation): Rapid water injection via the water tank at the back of the model box caused the groundwater level to increase quickly. The water level changes were continuously monitored until obvious deformation or signs of instability were observed in the landslide body. Scenario 3 (slow rise): Starting from the third day, a slow water addition method was adopted to simulate the gradual increase in the groundwater level. The water level was first raised to 10 cm and then lowered and finally raised again, and this cycle was repeated until the water level eventually reached 20 cm. The entire process lasted 4 days, with precise control of the water addition rate to ensure the stability of water level change.

(8) Data Collection and Observation

A YBY-4010 strain analysis system and a YB-R485 data acquisition device were used to collect data from the various sensors at a frequency of once every 3 s. Grid marking points were prearranged across the outer surface of the model box to observe rainfall infiltration. Via the use of data from the buried soil pressure sensors, the internal stress changes in the soil were analyzed to infer internal deformation. Close attention was given to the appearance, expansion, and penetration process of cracks on the model surface, and the crack distribution at key time points was recorded. Once obvious deformation or signs of instability were observed, the landslide body movement process was recorded in detail.

(9) Data Analysis

Sensor data, image analysis results, and observation records were synchronized in both time and space. Combining factors such as rainfall intensity, cumulative rainfall, and groundwater level changes, the internal water-force coupling mechanism of the landslide body was analyzed. By comparing data from the different scenarios, the critical rainfall and groundwater level conditions for triggering landslide instability were determined. On the basis of the continuous monitoring data, the complete deformation evolution trend of the landslide from initiation to instability was summarized.

4. Analysis of the Physical Model Test Results

4.1. Analysis of the Slope Seepage and Deformation Characteristics

Under scenario one, with only rainfall action, water infiltration in the slope body occurred from shallow to deep and from top to bottom. After the first day of rainfall, the overall wetting depth of the slope was relatively shallow, reaching only approximately 6 cm in the deepest part of the lower section, as shown in Figure 7a. After the second day of rainfall, the middle and lower parts of the slope body were more deeply wetted, exceeding half the thickness of the slope body, with deeper wetting closer to the lower part of the slope. At the end of the last rainfall event, only a small portion at the toe of the slope was completely saturated, indicating that the slope toe was most susceptible to rainfall infiltration, as shown in Figure 7b. Under purely rainfall conditions, water infiltrated gradually downwards from the surface, with the toe area becoming the most susceptible to rainfall effects due to gravity.

With respect to slope deformation, after the first day of rainfall, only small cracks emerged in the slope surface, with no large-scale deformation observed, as shown in Figure 8a. After the second day of rainfall, the slope surface became moist, small cracks developed in the lower part of the slope, and a gully formed on the left side of the slope toe. However, the landslide as a whole remained relatively stable, as shown in Figure 8b. Despite the occurrence of slight deformation, the slope did not experience large-scale sliding or collapse under rainfall conditions alone, indicating that gently inclined loess-bedrock contact slopes exhibit resistance to rainfall disturbance.

Under the combined effect of rainfall and rapid groundwater level increase under scenario two, the water infiltration rate significantly increased and exhibited rapid and comprehensive characteristics. After the first day of rainfall, the wetting depth due to rainfall in the upper part reached 10 cm, whereas the saturated soil in the lower part significantly increased due to the rapid increase in the groundwater level, almost connecting with the wetted soil in the upper part, as shown in Figure 9a. After the second day of rainfall, the entire slope body was completely saturated, reflecting the significant impact of a rapid groundwater level increase on the slope saturation process, in stark contrast to the gradual saturation under purely rainfall conditions, as shown in Figure 9b.

Under the combined effect of rainfall and rapid groundwater level increase under scenario 2, slope deformation clearly exhibited staged evolution features. After the first day of rainfall, two longitudinal tensile cracks formed in the middle of the slope. Notably, these cracks were 2–3 mm wide, 30 cm long, and approximately 3 cm deep, as shown in Figure 10a. After the second day of rainfall, deformation at the top of the slope intensified dramatically, the longitudinal cracks in the middle part continued to expand, transverse tensile cracks emerged in the upper part, and multiple shear cracks formed at the toe of the slope, indicating that the slope had begun to slide, as shown in Figure 10b. The entire process demonstrated progressive failure characteristics from the surface to the interior and from the local scale to the overall scale, with the middle part and toe of the slope becoming areas of intense and priority deformation.

Scenario three aimed to simulate the combined effect of rainfall and a gradual groundwater level increase, with the wetting depth of the slope exhibiting progressive change characteristics. After the first day of rainfall, the wetting depth of the slope reached approximately 6 cm, reflecting the rapid infiltration process at the initial rainfall stage, as shown in Figure 11a. After the second day of rainfall, as precipitation continued and the groundwater level slowly increased, the wetting depth increased to 21 cm, indicating that water began to penetrate deeper into the slope body (Figure 11b). On the third day, the wetting depth further increased to 28 cm, with the wetted area gradually expanding to cover most of the slope body (Figure 11c). On the fourth day, the wetting depth reached 30 cm, almost approaching the bottom of the slope, with the water infiltration rate decreasing (Figure 11d). After the fifth day of rainfall, the wetting depth only slightly increased to 31 cm, indicating that the slope body was approaching saturation, with the water infiltration rate further decreasing (Figure 11e). After the sixth day of rainfall, the entire slope body was completely saturated, presenting a uniform moist state (Figure 11f). The change in the wetting depth indicated that under the combined action of long-term rainfall and a gradual groundwater level increase, the progression speed of wetting gradually decreased, and the soil water absorption capacity changed over time, revealing the continuous impact of a groundwater level increase on the moisture state of deep soil layers.

Under scenario three, the slope deformation characteristics progressively developed. After the first day of rainfall, an initial deformation occurred in the slope body, with a transverse crack that was 1–2 mm wide, 40 cm long, and approximately 1 cm deep occurring in the middle part and a gully that was 1 cm wide and 5 cm deep developing at the bottom, reflecting the surface structural response to rainfall, as shown in Figure 12a. From the second to third days, the deformation continued to develop, with the middle crack expanding to 3–4 mm wide, 60 cm long, and approximately 2 cm deep and the bottom gully widening to 2 cm and deepening to 10 cm, revealing the continuous impact of the groundwater level increase on the soil structure (Figure 12b,c). After the fourth day of rainfall, the deformation further intensified, with the cracks in the middle and lower parts continuing to expand, and the deep soil layers began to respond (Figure 12d). After the fifth day of rainfall, the slope deformation exhibited accelerated development, with the cracks in the middle and lower parts continuing to expand, while a through-going shear crack up to 70 cm long and approximately 4 cm deep emerged in the lower part, indicating conditions close to possible landslide initiation (Figure 12e). After the sixth day of rainfall, the deformation entered an accelerated stage, with the crack network further developing and increasing in connectivity. Moreover, significant sliding and collapse occurred at the bottom of the slope, with the collapsed area reaching 0.4 m² and the depth ranging from 15 to 20 cm, indicating that under long-term hydrological conditions, the overall strength of the soil body decreased to a critical state, and the slope became globally unstable, as shown in Figure 12f.

4.2. Analysis of Slope Moisture Content Change Patterns

Under scenario one with only rainfall action, in the initial rainfall stage (0–5 h), the surface soil moisture content rapidly increased, reflecting the high porosity and strong water absorption of the loess. The moisture content at 10 cm depth in the upper, middle, and lower parts of the slope quickly increased from 16.7%, 14.9%, and 15.3% to 17.1%, 15.6%, and 15.7%, respectively, as shown in Figure 13a. The rapid response was mainly concentrated in the surface layer, while the moisture content change at 20 cm depth was relatively small, indicating that rainfall initially primarily affects shallow soil layers. In the middle stage of rainfall (5–20 h), the change in moisture content tended to be gradual, entering a relatively stable infiltration phase, reflecting that the surface soil reached a temporary equilibrium state, and water began to slowly penetrate into deeper layers. In the later stage of rainfall (20–27 h), the moisture content showed a slight rebound, such as an increase from 13.3% to 13.9% at 20 cm depth in the middle part, indicating continuous water infiltration into deeper layers and significant changes in the moisture state of deep soil. Throughout the process, the moisture content changes exhibited obvious vertical differentiation characteristics, with the magnitude of change at 10 cm depth generally being greater than that at 20 cm depth, reflecting the vertical infiltration characteristics of the loess.

Under scenario two (rainfall + rapid groundwater level rise), the changes in moisture content were more dramatic and complex. In the initial rainfall stage (0–5 h), the slope moisture content showed a rapid upward trend, with the magnitude of change far exceeding that of scenario one. For example, the moisture content at 10 cm depth in the upper left part of the slope reached 30.9% after 5 h, as shown in Figure 13b; the upper right part increased from 13.9% to 67.1%, and the lower part increased from 17.8% to 21.4%, as shown in Figure 13c. This dramatic change reflects the superimposed effect of rainfall and rapid groundwater level rise, leading to rapid soil saturation. In the middle stage of rainfall (5–20 h), the change in moisture content slowed down but remained at a high level. Fluctuations and decreases in moisture content occurred in some local areas, such as a drop from the peak value to 28.1% in the upper left part. In the later stage of rainfall (20–27 h), under the combined effect of continuous rainfall and a rapid water level rise, significant fluctuations in moisture content reoccurred in some areas, with the upper left part rising from 28.1% to 70.5%. This high saturation state directly led to a significant reduction in soil strength, creating conditions for large-scale sliding.

Under scenario three (rainfall + gradual groundwater level rise), the characteristics of moisture content change reflected the cumulative effect of long-term hydrological actions. In the initial rainfall stage (0–20 h), the surface soil moisture content increased sharply, with the magnitude of change between that of scenarios one and two. For instance, the left slope increased sharply from 16.1% to 70% within 20 h (Figure 13d), the middle slope from 16.1% to 60.3% (Figure 13e), and the right side from 16.1% to 70% (Figure 13f). The moisture content at 20cm depth also showed significant increases, rising from 15.5%, 16.6%, and 15.5% to 73%, 60%, and 69.4%, respectively. In the middle stage of rainfall (20–85 h), the changes in moisture content exhibited obvious spatial differentiation characteristics. For example, the moisture content in the upper left part decreased from a peak of 70% to 58.1%, while the lower part increased from 15.8% to 28%, reflecting the process of vertical wetting and redistribution of water. In the later stage of rainfall (85–171 h), the continuous gradual rise in the groundwater level led to a rapid increase in moisture content in the lower part of the slope, such as an increase from 30% to 63.3% on the left side, and even more dramatically from 26.5% to 69.2% on the right side. This sustained high moisture content state eventually led to a significant reduction in soil strength, creating conditions for landslide initiation.

4.3. Analysis of Pore Water Pressure Change Patterns

Under scenario one, with only rainfall action, at the initial rainfall stage (0–5 h), the pore water pressure rapidly increased, especially in the shallow soil layers. The pore water pressure at the 10 cm depth in the lower part of the slope rapidly increased from 0 to 0.81 kPa, reflecting the direct impact of rainfall infiltration on shallow soil layers, as shown in Figure 14a. The rapid increase phenomenon suggests the possible presence of preferential flow paths, given the vertical joints and large-pore structure of loess. At the middle stage of rainfall (5–20 h), due to water redistribution in the soil and local drainage effects, the pore water pressure slowly decreased, reflecting the dynamic adjustment process of the internal hydraulic balance in the soil. At the later stage of rainfall (20–27 h), the pore water pressure again increased, such as from 0.385 to 0.565 kPa at a depth of 10 cm in the lower part. This change indicates that as continuous rainfall resulted in the gradual saturation of deeper soil layers, the pore water pressure increased again. Throughout the process, the pore water pressure changes exhibited obvious vertical differentiation characteristics, with the magnitude of change in the shallow layer (10 cm) generally exceeding that in the deep layer (20 cm).

Under scenario two (rainfall + rapid groundwater level rise), the changes in the pore water pressure were more dramatic and complex. At the initial rainfall stage (0–5 h), the pore water pressure sharply increased, with the magnitude of change far exceeding that under scenario one. For example, the pore water pressure at the 10 cm depth in the middle left part rapidly increased from 0 to 1.43 kPa, as shown in Figure 14b, while at the 20 cm depth in the lower right part, the value increased from 0 to 0.451 kPa, as shown in Figure 14c. This dramatic change reflects the superimposed effect of rainfall and rapid groundwater level increase, leading to a rapid increase in the soil pore water pressure. At the middle stage of rainfall (5–20 h), the pore water pressure changes exhibited complex fluctuation characteristics. For example, the pore water pressure in the middle left part decreased from 1.315 to −4.63 kPa, with this significant negative pressure phenomenon possibly related to local drainage and soil structure adjustment. This stage reflects the dramatic changes in the internal hydraulic conditions and complex water movement processes in the soil under compound hydrological conditions. At the later stage of rainfall (20–27 h), under the combined effect of continuous rainfall and rapid water level increase, the pore water pressure sharply increased again. For example, in the upper left part of the slope, it increased from 0.04 to 1.775 kPa. This high pore water pressure state directly led to a significant reduction in the effective soil stress, creating conditions for large-scale sliding.

Under scenario three (rainfall + gradual groundwater level rise), the pore water pressure change characteristics reflected the cumulative effect and delayed response of long-term hydrological actions. At the initial rainfall stage (0–20 h), the pore water pressure exhibited a relatively slow but continuous upward trend. For example, the pore water pressure at a depth of 20 cm in the lower right part of the slope increased from 0.26 to 0.34 kPa, as shown in Figure 14f. Although this change rate was lower than that under scenario two, it lasted longer, reflecting the combined impact of long-term rainfall and a gradual groundwater level increase. At the middle stage of rainfall (20–85 h), the pore water pressure changes exhibited obvious spatial differences and fluctuations. The pore water pressure at the 10 cm depth in the upper left part continued to increase from 0.28 to 1.105 kPa, as shown in Figure 14d, whereas the values in the middle and lower parts fluctuated, first increasing and then decreasing. The middle and right parts of the slope also exhibited similar complex change patterns, such as the pore water pressure in the upper middle slope decreasing from 0.105 to −1.125 kPa, as shown in Figure 14e, whereas, at the 20 cm depth in the lower right part, the value continued to increase slowly. At the later stage of rainfall (85–171 h), the continuous gradual increase in the groundwater level led to sustained pore water accumulation and a substantial increase in the pore water pressure, such as an increase from 2.04 to 2.4 kPa in the lower left part and from 0.86 to 1.115 kPa in the middle part. This sustained high pore water pressure state eventually led to a significant reduction in the effective soil stress.

4.4. Analysis of Earth Pressure Change Patterns

Under scenario one, with only rainfall action, at the initial rainfall stage (0–5 h), the earth pressure rapidly increased, especially in the upper part of the slope. For example, the earth pressure at a depth of 20 cm in the upper part rapidly increased from 0 to 4.57 kPa. This rapid response could be attributed mainly to the increase in the soil weight due to rainfall infiltration, as shown in Figure 15a. At the middle stage of rainfall (5–20 h), the change in the earth pressure was gradual, with fluctuations that first increased but then decreased, reflecting the dynamic adjustment process of the internal stress state in the soil, which was related to water redistribution and minor adjustments in the soil structure. At the later stage of rainfall (20–27 h), the earth pressure significantly changed again, reaching a peak value in hour 22 (e.g., 10.02 kPa at the 20 cm depth in the upper part), followed by a sharp decrease. This abrupt change indicates a possible destruction of the soil structure and the beginning of overall sliding.

Under scenario two (rainfall + rapid groundwater level rise), the earth pressure changes were more dramatic and complex. At the initial rainfall stage (0–5 h), the earth pressure sharply increased, with the magnitude of change far exceeding that under scenario one. The earth pressure at the 10 cm depth in the middle left part rapidly increased from 0 to 6.56 kPa, while at the 10 cm depth in the lower left part, the value increased from 0 to 1.92 kPa, as shown in Figure 15b. This dramatic change reflects the superimposed effect of rainfall and rapid groundwater level increase, leading to rapid increases in both the soil weight and pore water pressure. The increase in the earth pressure was mainly concentrated in the middle and lower parts of the slope, which was related to the direct impact of groundwater level increase. At the middle stage of rainfall (5–20 h), the earth pressure continued to increase, but at a lower growth rate. For example, the earth pressure in the middle left part slowly increased from 6.56 to 25.41 kPa, and the earth pressure in the lower right part slowly increased from 6.54 to 13.11 kPa, as shown in Figure 15c. This stage reflects the continuous adjustment process of the soil stress field under sustained hydrological action, with significant differences in earth pressure changes at the different locations. At the later stage of rainfall (20–27 h), the earth pressure changes exhibited complex nonlinear characteristics. The earth pressure in certain areas began to decrease after reaching peak values, such as a decrease from 23.11 to 19.67 kPa in the middle left part. This change suggests that the soil structure may be approaching failure, and large-scale sliding may be imminent.

Under scenario three (rainfall + gradual groundwater level rise), the earth pressure change characteristics reflected the cumulative effect and progressive response to long-term hydrological actions. At the initial rainfall stage (0–20 h), the earth pressure rapidly but relatively gently increased. For example, the earth pressure at various monitoring points on the left side rapidly increased from zero to peak values of 24.43 kPa (upper part), 15.58 kPa (middle part), and 10.49 kPa (lower part). At the middle stage of rainfall (20–85 h), the earth pressure changes exhibited obvious spatial differences and nonlinear characteristics. The earth pressure in the upper left part slowly increased to 23.94 kPa, as shown in Figure 15d. The change trend in the middle part of the slope was more complex, with the value in the upper 10 cm first decreasing to −13.16 kPa and then slowly increasing, whereas the change trend in the lower part continued to decrease slowly, as shown in Figure 15e. The right side of the slope showed a distinct V-shaped fluctuation pattern. For example, the earth pressure in the lower part first decreased to −17.64 kPa and then increased to 18 kPa, as shown in Figure 15f. This complex change pattern reflects the soil strength decrease and local structural adjustments caused by continuous rainfall under long-term hydrological action. At the later stage of rainfall (85–171 h), under the conditions of high saturation and high pore pressure levels, the earth pressure showed an overall accelerating upward trend. For example, the value in the lower left part increased from 24.8 to 33.43 kPa, and that in the lower middle part of the slope increased from 28.8 to 33.44 kPa. This sustained high earth pressure state caused a further reduction in the soil shear strength, eventually resulting in overall slope instability.

5. Stability Analysis of the Libi Landslide via Numerical Simulations

5.1. Establishment of the Landslide Model

To investigate the impact of groundwater level changes on the stability of the Libi landslide and overcome the limitations of physical model tests, numerical simulations were conducted via GEO-STUDIO 2012 software. GEO-STUDIO, incorporating multiple functional modules such as SEEP/W and SLOPE/W, enables a coupled analysis of the seepage field and stress field, effectively simulating the impacts of rainfall infiltration and groundwater level changes on slope stability. The software’s comprehensive constitutive model library and material parameter database can accurately describe the unsaturated characteristics of loess, while its powerful transient analysis capabilities enable the simulation of long-term rainfall and groundwater level variation processes, making it an ideal tool for studying gently inclined loess landslides. A computational model was established on the basis of the 2-2 profile of the Libi landslide, as shown in Figure 16. The left side and bottom of the model were defined as free boundaries, with the slope top serving as the rainfall infiltration boundary. The soil parameters, including Poisson’s ratio, effective cohesion, internal friction angle, saturated unit weight, and saturated permeability coefficient, were selected on the basis of the laboratory test results, as detailed in

Table 4. Soil parameters.

Category	Poisson’s Ratio	Effective Cohesion C/kPa	Internal Friction Angle φ (°)	Saturated Unit Weight (kN/m3)	Saturated Permeability Coefficient (m/s)
Loess	0.3	11	9.86	19.0	4.0868 × 10−6

. A permeability coefficient curve and soil-water characteristic curve were obtained by fitting functions from the SEEP/W module function library. Three simulation scenarios were established: rainfall only, rainfall with a rapid groundwater level increase, and rainfall with a gradual groundwater level increase. Rainfall data from Qinshui County for September 2021 were utilized, and groundwater level changes were set according to the actual monitoring data. The simulation time was set to one month for all the scenarios to comprehensively analyze the slope stability changes under the different conditions. Via this model, the impacts of rainfall infiltration and groundwater level changes on the stability of gentle slope loess landslides could be systematically studied, providing important evidence for revealing their formation mechanisms.

5.2. Geological Material Modeling Process

In establishing the numerical model, the geological material modeling process underwent systematic design and validation. Based on borehole data and engineering geological survey results, the landslide mass was conceptualized into four main soil layers: an artificial fill layer (Q₄^ml, thickness 8–10 m), a loess-like soil layer (Q₄^dl+pl, thickness 5–7 m), a silty clay layer (Q₃^dl+pl, thickness 4–6 m), and a silty clay layer with gravel (Q₃^dl+pl, thickness 3–5 m), while a weak interlayer with a thickness of 0.3–0.5 m was set at the bottom of the sliding mass to simulate the actual slip zone location. Considering the unsaturated characteristics of loess and the nonlinearity of stress–strain relationships, the Mohr–Coulomb elastoplastic model was adopted for the constitutive model to describe the mechanical behavior of the soil mass. The determination of key parameters was primarily accomplished through laboratory tests and field monitoring, where physical parameters were measured through conventional geotechnical tests; strength parameters were obtained through triaxial and direct shear tests; permeability parameters were determined through permeability tests and soil-water characteristic curve tests; and deformation parameters were established through compression tests and back analysis. To ensure parameter rationality, multiple validation methods were employed: back analysis validation using field displacement monitoring data, parameter sensitivity analysis to determine key parameters, and comparative validation with parameters from similar engineering projects.

5.3. Simulation Scenario Settings

On the basis of the actual situation of the Libi landslide, three typical numerical simulation scenarios were established to comprehensively analyze the slope stability changes under different conditions. Scenario one aimed to simulate only rainfall to study the impact of rainfall infiltration on slope stability. Scenario two aimed to simulate rainfall plus rapid groundwater level increase to investigate the acute impact of rapid groundwater level changes on slope stability, and scenario three aimed to simulate rainfall plus gradual groundwater level increase, which is closer to the actual landslide occurrence process. Under scenario three, based on the monitored water level changes, the groundwater level was simulated to increase slowly from 11.19 to 17.3 m, thereby fluctuating with rainfall intensity changes to reflect the actual water level change patterns. The simulation time for all the scenarios was set to one month to capture the dynamic process of slope stability changes.

5.4. Analysis of the Simulation Results

5.4.1. Analysis of the Scenario One Results

Scenario one aimed to capture the stability changes in the Libi landslide under rainfall-only conditions. Via the analysis of the displacement, maximum shear strain, and pore water pressure, a comprehensive understanding of the impact of rainfall on slope stability could be obtained.

On the first simulation day, the maximum displacement occurred in the steeper upper part of the slope, with a displacement of 0.0036 m, mainly manifesting as shallow surface failure due to saturation and wetting of the slope surface soil due to rainfall, as shown in Figure 17. With continuous rainfall, the maximum deformation area gradually expanded downwards. By the fifth day, the maximum displacement area had shifted to the slope toe, reaching 0.0065 m, which was consistent with the phenomena observed in the physical model test; namely, cracks emerged in the upper part at the initial rainfall stage, followed by local failure at the slope toe. The maximum deformation area subsequently slowly migrated towards the middle part of the slope, reaching a maximum displacement of 0.009 m on the thirteenth day. Thereafter, the displacement in the middle part continued to increase, reaching 0.028 m by the end of the simulation, as shown in Figure 18.

The evolution process of the maximum shear strain was basically consistent with the displacement change trend. At the early simulation stage, the maximum shear strain was mainly concentrated in the shallow surface layer of the steep upper slope, which was caused by surface soil saturation and wetting due to rainfall, as shown in Figure 19. As rainfall continued, the maximum shear strain area gradually expanded downwards, reaching the slope toe on the sixth day. The maximum shear strain area subsequently began to shift towards the deep soil layers in the middle part of the slope. By the eighteenth day, the shear strain in the deep soil layers in the middle part reached its maximum, whereas the upper part still exhibited shallow surface failure, and the shear strain at the slope toe decreased. On the thirtieth simulation day, the slope body exhibited obvious layered shear characteristics: the shear strain decreased in the lower part, it remained stable in the middle part, and the process developed from surface single-layer failure to simultaneous shear failure in both the surface and deep layers in the upper part, as shown in Figure 20.

The change in the pore water pressure directly reflects the impact of rainfall infiltration on the soil moisture state. At the early simulation stage, the slope surface layer exhibited a negative pore water pressure due to its relatively dry state, as shown in Figure 21. With rainfall infiltration, the pore water pressure gradually changed from negative to positive values, but the change rate was relatively low. Even by the tenth simulation day, there was no significant change in the slope pore water pressure. This indicates that rainfall infiltration is a gradual process. It was not until the twenty-fourth day that the pore water pressure in the slope body basically turned positive, and this state was maintained until the end of the simulation, as shown in Figure 22.

5.4.2. Analysis of the Scenario Two Results

Rapid groundwater level rise significantly accelerated the landslide deformation process, and this impact can be analyzed from three aspects: displacement, shear strain, and pore water pressure. Regarding the displacement development characteristics, on the first day of the simulation, the maximum displacement occurred in the steeper upper part of the slope, with a magnitude of 0.005 m (see Figure 23), primarily manifesting as shallow surface deformation caused by rainfall. With the rapid rise in groundwater level, by the fourth day, the zone of maximum displacement shifted to the slope toe, reaching 0.01 m, approximately double that of scenario one during the same period. The displacement continued to develop, reaching a maximum value of 0.04 m on the eighteenth day (see Figure 24), about 1.5 times greater than that in scenario one. This displacement evolution pattern reflected a failure mechanism initiating from the slope toe and developing upward, specifically, the slope toe soil first underwent softening due to groundwater saturation, triggering localized instability, after which the unstable zone gradually expanded with the rising groundwater level.

The development of shear strain further confirms this mechanism, exhibiting three distinct phases (Figure 25): an initial peak appearing on day three, rapid growth beginning on day ten, and finally reaching its peak value on day eighteen. Notably, the shear strain in the lower portion of the slope showed the fastest response and highest values, consistent with its characteristics of thinner soil layer thickness and earliest saturation. Compared with scenario one, scenario two demonstrated faster maximum shear strain rate changes and higher peak values, directly reflecting the impact of rapid groundwater level rise (Figure 26).

The evolution of pore water pressure (see Figure 27 and Figure 28) also exhibited rapid change characteristics: in the initial state, values above the water table were negative (unsaturated condition); by the third day, the slope toe area turned positive, indicating that the groundwater level had risen to this location; by the eighteenth day, the entire slope body showed predominantly positive values, suggesting that most areas had reached saturation. In comparison, scenario one did not reach a similar state until the twenty-fourth day. This rapid saturation process significantly accelerated landslide instability. The above analysis indicates that a rapid groundwater level rise is a key factor in accelerating landslide instability, with its influence mechanism operating through rapid changes in soil saturation state, leading to significant strength reduction, thereby accelerating deformation development.

5.4.3. Analysis of the Scenario Three Results

Scenario three aimed to simulate rainfall plus a gradual groundwater level increase, which is closer to the actual landslide occurrence process. The displacement changes under scenario three exhibited evolution characteristics that were intermediate and between those of scenarios one and two, reflecting the impact of gradual groundwater level increase. On the first simulation day, the maximum displacement still occurred in the steeper upper part of the slope, with a displacement of 0.0036 m, comparable to that under scenario one, as shown in Figure 29. This stage mainly demonstrated shallow surface failure, which was still due to slope surface soil saturation and wetting caused by rainfall. With continuous rainfall and a gradual groundwater level increase, the rate of displacement change gradually increased but at a slower pace than that under scenario two. By the sixth day, the maximum displacement area had shifted to the slope toe, reaching 0.009 m. This phenomenon was similar to that under scenario two but occurred later and with a smaller displacement, reflecting the progressive impact of gradual groundwater level increase on slope stability. As the groundwater level subsequently continued to increase slowly, the maximum displacement area gradually shifted towards the middle part of the slope. On the twenty-first simulation day, the displacement in the middle part of the slope reached a maximum value of 0.034 m, as shown in Figure 30. This process indicated that although the gradual increase in the groundwater level did not lead to rapid slope instability, as under scenario two, it still significantly affected slope stability, but this impact involved a relatively slow accumulation process.

The maximum shear strain analysis results of scenario three further confirmed the progressive impact of gradual groundwater level increase on slope stability. The changes in the maximum shear strain in the upper, middle, and lower parts of the slope all exhibited similar patterns: a low peak was reached on the fourth simulation day, as shown in Figure 31; then, it slowly increased until the twelfth day, when it began to increase rapidly; it peaked on the twenty-first day, as shown in Figure 32; and finally, it remained stable until the end of the simulation. This change trend was similar to that under scenario two, but the entire process was slower. For example, the value under scenario two began to increase rapidly on the tenth day, whereas the value under scenario three increased rapidly on the twelfth day; the value under scenario two reached its peak on the eighteenth day, whereas the value under scenario three peaked on the twenty-first day. This difference directly reflects the impact of the groundwater level increase rate on the slope instability process. Although the rate of change was lower, the final maximum shear strain value under scenario three (0.001) was still significantly higher than that under scenario one (0.00055), indicating that even a gradual increase in the groundwater level could significantly affect slope stability.

The pore water pressure changes under scenario three exhibited more pronounced characteristics than those under scenario one but were slower than those under scenario two, directly reflecting the impact of a gradual groundwater level increase. In the initial state, the pore water pressure above the water level line was negative, indicating that the soil exhibited an unsaturated state, as shown in Figure 33. With continuous rainfall and a gradual groundwater level increase, the pore water pressure gradually changed from negative to positive values. By the sixth day, the pore water pressure in most areas of the slope toe had turned positive, indicating that the groundwater level had increased at this location. The associated change rate was higher than that under scenario one but slower than that under scenario two, reflecting the characteristics of gradual groundwater level increase. By the twentieth simulation day, as shown in Figure 34, the pore water pressure throughout essentially the entire slope body had turned positive, indicating that most areas of the slope had reached a saturated state. This process was slower than that under scenario two (which reached a similar state on the eighteenth day) but faster than that under scenario one (which did not reach a similar state until the twenty-fourth day).

6. Formation Mechanism Study of Landslides on Gently Inclined Loess–Bedrock Contact Surfaces

Through a combination of physical model experiments and numerical simulations, this study revealed the formation mechanisms of landslides on gently inclined loess–bedrock contact surfaces under various combinations of rainfall and groundwater conditions, with the primary mechanism manifesting as a progressive failure process driven by hydro-mechanical coupling effects.

Under scenario one (rainfall only), the landslide primarily exhibited a process of surface saturation, crack development, and shallow localized instability. During the initial rainfall period (0–5 h), water mainly affected the surface soil, causing a rapid increase in water content at 10 cm depth (from 16.7% to 17.1%), while changes at 20 cm depth were insignificant, demonstrating typical shallow influence characteristics. As rainfall continued, significant water redistribution occurred within the soil mass, with the pore water pressure evolving from negative to positive values, reaching its peak at the 22nd hour. The soil pressure increased with water content, reaching a maximum of 10.02 kPa at 20 cm depth in the upper section, but did not lead to overall instability. The deformation process exhibited distinct spatial differentiation, mainly manifesting as surface crack development and localized collapse, without forming a through-going sliding surface. This indicates that under rainfall conditions alone, gently inclined slopes maintain a certain self-stability capacity, with failure primarily limited to the shallow surface layer and, resulting in a minimal impact on deep soil masses. This closely aligns with numerical simulation results, which showed a maximum displacement of only 0.028 m, with maximum shear strain mainly concentrated in the surface layer and relatively slow deformation development rates.

Under scenario two (rainfall + rapid groundwater level rise), the landslide exhibited a process of rapid saturation, stress redistribution, and overall instability. The combined effects of rainfall and a rapid groundwater level rise led to quick soil saturation, with the water content in the upper left section dramatically increasing from 13.9% to 67.1% within 5 h. The pore water pressure changed dramatically, rapidly rising from 0 kPa to 1.43 kPa at 10 cm depth in the middle left section, followed by significant fluctuations, reflecting complex hydro-mechanical coupling effects. Changes in soil pressure were even more significant, increasing from 0 kPa to 25.41 kPa in the middle left section, indicating fundamental changes in the internal stress state of the soil mass. The deformation process showed distinct staged and sudden change characteristics, first forming longitudinal tensile cracks (2–3 mm wide, 30 cm long) in the middle section, followed by horizontal tensile cracks in the upper section, and finally, multiple shear cracks at the slope toe, leading to overall instability. The numerical simulation results showed that the maximum displacement reached 0.04 m, 1.5 times that of scenario one, with a more rapid failure process developing from the slope crest downward, ultimately forming a through-going sliding surface. Based on both the model test and the numerical simulation results, the failure mode under this scenario can be categorized as a “sudden translational failure mode”, characterized by the following features: under the combined effects of rainfall and a rapid groundwater level rise, softening first occurs in the rear portion of the slope, and as the groundwater level continues to rise rapidly, the saturated zone quickly expands downward, accompanied by sudden adjustments in internal slope stresses, increased unit weight of soil in the middle and rear portions, and increased sliding force. Under the weight of the overlying soil mass, sudden overall sliding occurs. This failure mode is characterized by rapid development, deep failure depth, and wide affected area, often leading to disastrous consequences within a very short time.

Under scenario three (rainfall + gradual groundwater level rise), the landslide formation mechanism exhibited typical progressive failure characteristics. The water infiltration process proceeded from fast to slow, with the initial saturation depth reaching only 6 cm on the first day and complete saturation achieved by the sixth day. The water content changes demonstrated clear cumulative effects, increasing from 16.1% to 70% within 20 h in the left slope section, with spatial distribution showing distinct non-uniformity. The evolution of the pore water pressure was gradual but continuously accumulative, increasing from 2.04 kPa to 2.4 kPa in the lower left section during the 85–171 h period. The soil pressure exhibited significant nonlinear characteristics with different variation patterns at different locations, such as a “V-shaped” fluctuation in the lower right section (decreasing from 0 kPa to −17.64 kPa before rising to 18 kPa). The deformation development underwent a complete evolutionary process: from initial fine cracks (1–2 mm wide, 40 cm long), to continued expansion in the middle phase (3–4 mm wide, 60 cm long), and finally forming through-going shear cracks (70 cm long, 4 cm deep), leading to overall instability. The numerical simulation results indicated that this progressive failure process resulted in a maximum displacement of 0.034 m, between that of scenarios one and two, but with a more complete and typical failure process. Based on both the model test and the numerical simulation results, the failure mode under this scenario can be categorized as a “progressive translational failure mode”, characterized by the following features: under continuous rainfall and a gradual groundwater level rise, the soil water content gradually increases and strength begins to deteriorate; due to the gentle slope characteristics, the soil mass generates continuous horizontal thrust under gravity but without an obvious deformation initially; when the accumulated pore water pressure reaches a certain level, a localized translational deformation first appears at weak points, characterized by the horizontal displacement exceeding the vertical displacement, forming independent small translational bodies; as the groundwater level continues to rise slowly, various local translational deformation zones gradually expand and interconnect, forming translational sliding surfaces that are often approximately parallel to the bedrock surface; once the translational sliding surface is fully connected, under continuous hydraulic action and thrust from the upper soil mass, the sliding body exhibits primarily translational movement characteristics, resulting in a large-scale horizontal displacement. This failure mode is characterized by slow development, high predictability, and sufficient warning time.

According to monitoring data, the accumulated precipitation in the Libi landslide area reached 957.44 mm from January to September 2021, more than 30% higher than the average for the same period in normal years, with relatively continuous rainfall patterns. Groundwater level monitoring data showed distinct seasonal variation characteristics: the average water depth was 15.6 m in early May, and slowly rose by 4.8 m during June–July due to heavy rainfall influence. This slow but continuous water level change pattern highly corresponds with scenario three. Regarding the deformation process, monitoring data indicated that the landslide underwent a slow deformation phase lasting several months: during the period from November 2020 to October 2021, the slope’s maximum cumulative displacement reached 68.9 mm, with a maximum settlement of 5.5 mm. These progressive deformation characteristics closely align with the experimental results of scenario three. Therefore, it can be concluded that the formation mechanism of the Libi landslide highly corresponds with the formation mechanism under scenario three conditions; namely, under the long-term coupled effects of continuous rainfall and a gradual groundwater level rise, the landslide underwent a progressive translational failure process.

7. Conclusions

To reveal the formation mechanism of landslides on gently inclined loess–bedrock contact surfaces, taking the Libi landslide in Shanxi as the research object, a systematic study was conducted using a combination of physical model experiments and numerical simulations, yielding the following conclusions:

(1)

A specialized model test system for loess–mudstone contact surface landslides was designed, comprising three parts: a model box, rainfall system, and monitoring system, ensuring that the simulation results align with actual conditions.

(2)

Under rainfall conditions alone, this type of landslide exhibited a process of surface saturation, crack development, and localized instability, with water primarily affecting surface soil while deep changes remained insignificant. The soil pressure increased with the water content but did not lead to overall instability. The numerical simulation showed a maximum displacement of only 0.028 m, with the maximum shear strain mainly concentrated in the surface layer, indicating that under pure rainfall conditions, gently inclined slopes maintain a certain self-stability capacity.

(3)

Under conditions of rainfall plus a rapid groundwater level rise, this type of landslide exhibited a typical “sudden translational failure mode”. The failure process manifested in three stages: the softening of the upper slope forming localized weak zones, the rapid downward expansion of weak zones, and finally, the sudden overall translational sliding under the thrust of overlying soil mass.

(4)

Under conditions of rainfall plus a gradual groundwater level rise, this type of landslide exhibited a “progressive translational failure mode”, experiencing four stages: the initiation stage (increased soil water content and strength deterioration), the development stage (localized deformation in the rear portion), the acceleration stage (progressive forward expansion of rear deformation zone under continuous groundwater rise), and the activation stage (overall landslide movement). Through a comparative analysis of monitoring data, it can be concluded that the Libi landslide formed under the coupled effects of rainfall and a gradual groundwater level rise, conforming to the characteristics of the progressive translational failure mode.