Slope protection effect of typical vegetation in the Three Gorges reservoir area under extreme rainfall

Summary

In recent years, vegetation plays a key role in landslide stability under extreme rainfall in the Three Gorges Reservoir area, so it is very important to identify the mechanism of vegetation slope protection. This study takes wildcat landslide in Three Gorges Reservoir area as the research object, using indoor landslide model test and building monitoring systems such as stress field, displacement field, and soil erosion, to illustrate the protective effect of typical vegetation. Furthermore, Bermuda cover effectively reduces pore water pressure, pore soil pressure, displacement, and turbidity. In particular, the three stages of interception and buffering of rainfall by stems and leaves, infiltration and absorption of rainfall by the root system, and the reinforcement of the slope against sliding forces by the root system have been divided. Moreover, these findings offer valuable preliminary insights for guiding landslide mitigation strategies in the Three Gorges Reservoir area.

Graphical abstract

Highlights

• •Bermuda roots can increase the strength of root-soil complex
• •Bermuda covering landslide can reduce the internal stress, displacement, and soil erosion of slope
• •Vegetation slope protection mechanism can be divided into three stages
• •These findings offer valuable preliminary insights for guiding landslide mitigation strategies in the Three Gorges Reservoir area

Physical geography; Hillslope processes; Erosion; Nature conservation

Introduction

Characterized by a humid, middle subtropical monsoon climate, the Three Gorges reservoir area presents a fragile ecological environment susceptible to soil erosion from persistent rainfall and extreme storm events. This erosion leads to a cascade of ecological engineering challenges, including slope instability and landslides. In addition, increasing population growth in the reservoir area has driven intensified land development and utilization, resulting in a sharp decline in vegetation coverage. Therefore, soil erosion has become increasingly severe, leading to mounting economic losses.^1,2,3,4,5 Therefore, assessing landslide stability under extreme rainfall conditions and the protective role of vegetation is crucial for effective environmental management and mitigation strategies.

The infiltration of precipitation into soil can trigger a cascade of events resulting in slope instability and landslide occurrences. This process is primarily driven by an increase in soil water content, which therefore reduces shear strength and elevates pore water pressure.^6,7 The resulting softening of potential sliding surfaces in the soil structure creates conditions conducive to landslide initiation. Specifically, extreme rainfall events significantly exacerbate these effects, leading to accelerated deformation and a higher risk of landslides.^8,9,10,11 Empirical analyses utilizing physical models have offered valuable insights into the effect of rainfall on slope stability. Sun et al.¹² demonstrated that varied rainfall patterns induce deformation and failure processes in loess slopes. Their findings indicate that compromised drainage resulting from slope deformation can lead to excessive pore water pressure, effectively reducing effective stress and loess strength, finally triggering landslides. Similarly, Ering et al.¹³ analyzed the Malin Hills mudslide event, highlighting the critical role of rainfall intensity and duration in influencing slope stability. Further research by Dong et al.¹⁴ explored the effect of seepage boundaries on fractured soil interfaces, indicating that boundary conditions result in varied seepage field configurations and hydraulic characteristics. Large-scale physical model testing conducted by Wang et al.¹⁵ and Zhan et al.¹⁶ highlights the significance of monitoring pore water pressure during rainfall events. The development of excessive pore water pressure in the slope serves as a crucial early warning indicator of potential landslide initiation. This observation is corroborated by Zhang et al.,¹⁷ who observed a lag effect between rainfall infiltration and the occurrence of loess landslides through the monitoring of parameters such as pore suction, volumetric moisture content, pore water pressure, and slope deformation. Advanced analytical methods have also been employed to assess rainfall-induced landslides. Kumar et al.¹⁸ utilized finite element analysis, limit analysis, and limit equilibrium methods to assess the three-dimensional stability of slopes under rainfall conditions. In addition, Liu et al.¹⁹ employed geomechanical model tests with micro-soil pressure sensors and pore water pressure sensors to analyze the stress-deformation response and evolution of landslides, offering a comprehensive understanding of the mechanisms governing slope instability.

The utilization of vegetation offers a solution to address the limitations of conventional engineering methods in slope stabilization. Beyond its effectiveness in soil retention and slope reinforcement, vegetation also presents significant aesthetic and environmental benefits. Extensive research by scholars worldwide has substantiated the significant effect of vegetation on soil conservation and slope integrity^{20,21,22,23,24,25,26}(Shen et al. 2022; Sun et al. 2022; Chen et al. 2005; Liao et al. 2006; Zhao et al. 2017; Zhang et al. 2019; Masieb et al. 2021). Waldron et al.²⁷ hypothesize that the root systems of plants act as a strengthening agent, converting shear stress in the soil into tensile stress in the roots through interfacial friction, thereby augmenting the shear strength of the slope. Ghestem et al.,²⁸ through in situ shear tests on vetiver grass, demonstrated a linear correlation between the shear strength of the soil-root matrix and root density, highlighting the effectiveness of plant roots in enhancing soil shear strength. Jiao et al.,²⁹ based on vegetation surveys and soil erosion monitoring on the Loess Plateau, concluded that substantial vegetation cover effectively reduces erosion caused by intense rainfall events. Xu et al.³⁰ revealed, through comprehensive field monitoring and comparative analysis, that prolonged, low-intensity rainfall exerts a more negative effect on expansive soil slopes compared to brief, high-intensity rainfall. Li et al.³¹ employed numerical analysis and field measurements to compare the behavior of unsaturated soil slopes incorporating capillary barrier systems and deep-rooted grass with that of original slopes. Their findings suggest that capillary barrier systems play a more crucial role than deep-rooted grass in maintaining slope stability. Rahardjo et al.³² and Lim et al.³³ evaluated the unsaturated properties of capillary barrier soil and its influence on slope stability under rainfall conditions. Their research demonstrated that soil with root systems effectively reduces water infiltration into the slope surface during rainfall events.

While extensive research has explored rainfall-induced landslide mechanisms and the strengthening impact of vegetation’s root systems on soil strength, the internal influence of vegetation on landslide bodies under extreme rainfall remains poorly understood. To address this gap, this study evaluates the Wildcat landslide in the Three Gorges Reservoir area. Triaxial compression tests were conducted on both rootless soil and root-soil composites, followed by indoor landslide model rainfall tests on vegetated slopes. The study employed multiple stress sensors to monitor internal stress changes in the slope, displacement meters and 3D laser scanners to observe internal and surface displacement, respectively, and analyzed water content variations at different slope positions during the landslide process, along with turbidity changes in runoff sediment samples post-slope runoff. Through comparative analysis, the research aims to explain the protective effect of vegetation on the Wildcat landslide under extreme rainfall conditions.

Materials and methods

General situation

Characterized by its mountainous terrain and susceptibility to erosion, the Three Gorges Reservoir area is a focal point for soil and water loss in the upper Yangtze River basin. Its steep slopes, coupled with loose soil layers, make the region particularly vulnerable to the erosive forces of concentrated, high-intensity rainfall events, leading to substantial hydraulic erosion. Situated in this area, the Wildcat landslide area experiences a humid subtropical climate marked by seasonal changes. Influenced by monsoon patterns, the region exhibits a dry winter season and a summer season characterized by abundant rainfall.

The Wild Cat Face Landslide, situated on the left bank of the Niugan Mafei Gorge in the Xiling Gorge segment of the Yangtze River’s Three Gorges, poses a significant geohazard. This massive landslide body, located on the northern slope of the Yangtze River valley in Zigui County, Hubei Province, China, holds the distinction of being the largest landslide in proximity to the Three Gorges Dam in the Three Gorges Water Conservancy Hub area. Its geographical coordinates place it 12 km upstream from the old town of Zigui (Guizhou Town) and 17 km from the Three Gorges Dam construction site. The landslide comprises the residences of over 200 individuals, all residing above the 280-meter elevation mark in the administrative division of Miaohe Village, Maoping Town, Zigui County. The local economy, primarily reliant on citrus and green tea cultivation, is classified as underdeveloped due to limited transportation infrastructure, resulting in the classification of the area as a poverty-stricken mountainous region. The landslide’s topography exhibits a unique pattern: a steep frontal edge transitions into a gentler middle section before rising again to a rear section of slightly greater inclination than the middle. Accurately, the landslide’s central point is located at 110° 51′35.5″ E longitude and 30° 53′35.4" N latitude. The slopes in the valley generally maintain a 40° gradient, with certain sections reaching steepness levels of 70°–90°. A detailed representation of the landslide’s landform is depicted in Figure 1.

Figure 1.

Overall View of the Front Face of the Wild Cat Face Landslide

Meteorological data from the Zigui County Meteorological station indicates an average annual rainfall of 1025mm, with a historical maximum of 1430mm (1963) and a peak daily rainfall of 358mm (August 8, 1975). Rainfall is primarily concentrated between May and October, primarily in the form of rainstorms, resulting in 120–140 days of precipitation per year. Conversely, the average annual evaporation rate ranges from 800mm to 1000mm. Temperature extremes have been recorded at 41.7°C for the highest and ∼8.9°C for the lowest.

Test materials

The experimental soil samples were extracted from both the undisturbed and landslide zones at a 155m elevation in the Three Gorges Reservoir area. Following the “Chinese Standard for Geotechnical Test Methods” (GB/T 50123 2019),³⁴ a series of fundamental geotechnical analyses were performed on the undisturbed soil to determine its intrinsic physical properties and structural composition, alongside those of the landslide zone soil. Adhering to rigorous standard specifications, direct shear tests, soil moisture content tests, soil natural density tests, and soil specific gravity tests were executed. The resultant data, including the fundamental physical and mechanical parameters of both the undisturbed and landslide zone soils, are presented in Table 1.

Table 1.

Physical and mechanical parameters of landslide soil

Soil sample	Dry density/(g.cm−3)	Natural moisture content/(%)	Permeability coefficient	Specific gravity	Cohesion/kPa	Internal friction angle/(°)
Sliding soil	1.62	38.10	8.2 × 10−5	2.585	20.36	31.97
Plip soil	1.76	5.09	1.37 × 10−4	2.691	9.17	33.54

For this slope protection study, we selected Bermuda grass (Figure 2) due to its prevalence and representativeness as a herbaceous cover in the Three Gorges Reservoir region. Bermuda grassis a perennial grass species belonging to the Poaceae family and is also referred to as dog’s tooth grass, Bahama grass, or couch grass. Its distribution spans temperate zones globally, with particular abundance in southern parts of North China and some presence in Xinjiang, Jilin, and Qinghai provinces. Its cultivation extends to the southern United States, Europe, Africa, and parts of Asia. Characterized by a robust root system and resilience to pruning, trampling, and drought conditions, Bermuda grass features prostrate stems of varying lengths and lanceolate leaves with keeled sheaths and lacerate ligules bearing short hairs. The leaf blades typically range from 1 to 10 cm in length and 0.2 cm in width. These attributes make it an ideal candidate for vegetation restoration efforts in the Three Gorges Reservoir subsidence area.^35,36,37 To assess its tensile properties, we conducted single root tensile tests utilizing an HP-100 Edelberg digital push-pull force meter (Figure 3) with a 100 N range and 0.50% accuracy. The results, presented in Table 2, indicate a root diameter range of 0.13–0.50 mm for Bermuda grass, with maximum tensile strength and root tensile strength for individual roots ranging from 0.9 to 5.1 N and 25.99–67.84 MPa, respectively.

Figure 2.

Bermuda Grass

Figure 3.

HP-100 Edelberg digital push-pull force meter

Table 2.

Tensile test parameters of Bermuda grass

Type	length/cm	Root diameter/mm	Maximum tensile resistance/N	Tensile strength/MPa
Bermuda grass	0.5–4.5	0.13∼0.50	0.9∼5.1	25.99∼67.84

Test equipment

Utilizing a saturated/unsaturated soil stress path triaxial testing system, triaxial compression tests were conducted on both root-free soil and soil-root composites. The cylindrical samples possessed dimensions of 61.8mm in diameter and 125mm in height. To accommodate varying soil strengths, two axial force ranges were employed: 0-3kN and 0-10kN, both offering an accuracy of 0.1% of the respective range. Axial displacement measurements, with a range of 0∼50mm, were obtained with an accuracy of 0.25% of the range. Back pressure and pore water pressure, crucial for simulating in situ conditions, were controlled in a range of 0∼3MPa, each achieving an accuracy of 0.1% of the range. Finally, confining pressure was regulated between 0 and 2MPa, also with an accuracy of 0.1% of the range.

This study employed a designed indoor physical model to assess landslide behavior. The model, constructed with dimensions of length × wide × height = 3 m × 2 m × 2m, utilized transparent organic glass sidewalls to facilitate direct observation of deformation and failure mechanisms during experimentation. A flowmeter, positioned at the inlet of the rainfall sprinkler system, offered accurate data for determining rainfall intensity. The sprinkler system itself employed atomizing nozzles, and rigorous pre-experiment testing ensured rainfall uniformity exceeding 80% throughout the model. The physical model was conceptualized to represent the fundamental structural components of a landslide: the sliding bed, sliding zone, and sliding body (Figure 4). The sliding bed, constructed with ordinary bricks and surfaced with a layer of cement mortar for impermeability, simulated the bedrock upon which the landslide occurs. Maintaining consistency with the slope of the studied landslide, the model was set at a 30° incline. Both the sliding zone and sliding body were filled with soil in layers and sections, with the filling quantity determined based on accurate calculations of soil density and volume.

Figure 4.

Landslide model and layout plan

To comprehensively analyze the three-dimensional spatial variations of the internal seepage, stress, and displacement fields in the landslide across different elevations, as well as their two-dimensional variations at various points on the same plane, a suite of monitoring instruments was embedded in the landslide soil. These instruments included BW16–0.5 Soil Pressure Sensors (①), BWK18-1 Pore Water Pressure Sensors (②), YWC-20 Displacement Sensors (③), and MS-10 Soil Moisture (Temperature) Sensors (④). The monitoring scheme comprised four profiles (I, II, III, and IV), each instrumented with soil pressure and pore water pressure measures at 0.1m intervals along the same vertical plane from the base of the slope upwards. Besides, at a height of 0.3m above the base, each profile was equipped with a soil pressure measure, a pore water pressure measure, and a displacement measure. Moisture sensors, positioned at the midpoint of each profile, offered moisture content measurements at 5-min intervals. Data acquisition from the soil pressure and pore water pressure sensors was facilitated by the YBY-4010 strain testing and analysis system. Complementary to the embedded instrumentation, electronic monitoring devices, 3D laser scanners, and high-speed cameras were deployed to reflect the overall displacement, deformation, and failure characteristics of the landslide.

Test method

In the triaxial testing procedure, a soil sample is carefully extracted and promptly stripped of its plastic wrap. A membrane-lined cylinder is then employed to encase the sample with a rubber membrane, ensuring its integrity. This intact specimen is subsequently subjected to triaxial compression testing in a conventional strain-controlled triaxial apparatus, with the entire test conducted under ambient temperature conditions. The specific triaxial compression test type selected is the unconsolidated undrained (UU) shear test, employing confining pressures of 50 kPa, 100 kPa, and 200 kPa. Adhering to established geotechnical testing standards, the axial strain rate for the triaxial compression test is maintained in a range of 1%–3%, which represents 1.25–3.75 mm min⁻¹. To acquire the stress-strain curves of both root-free soil and root-soil composites under triaxial compression, and compare their strengths, a loading rate of 1.25 mm min⁻¹ is implemented during the testing process.

To accurately replicate real-world precipitation patterns in the Three Gorges Reservoir region, researchers collected and analyzed relevant meteorological data. This analysis, coupled with a study of 1-h maximum rainfall events (ranging from 55mm to 110mm) in the Hubei section of the reservoir, indicated a significant concentration (89%) of intense rainfall days below 50mm. Specifically, 51.4% of these events fell in the 20-29.9mm range, while 37.6% registered between 30 and 49.9mm. Rainfall exceeding 50mm constituted only 10% of the observed events, with sporadic occurrences of extreme precipitation surpassing 90mm during the summer months. Calculations determined the rainfall intensity for 50-year and 100-year events in the Three Gorges Reservoir area to be 50 mm/h and 75 mm/h, respectively. Based on these findings, this study employed a controlled indoor environment to simulate rainfall conditions, focusing on the extreme scenario of a 100-year event (75 mm/h intensity for 1 h duration). Two slope conditions were established for comparison: a bare slope control and a slope strengthened with herbaceous vegetation, as depicted in Figure 5. The initial moisture content and bulk density of the slope fill were set at 11.11% and 1805 kg/m³, respectively. Construction of the test slopes involved a manual layering and compaction approach, with each slope reaching a height of 34cm through four layers. Firstly, a 1cm sliding band was implemented, followed by the application of sliding soil in three 11cm layers. The fill quantity for each layer was determined based on the soil’s predetermined density and volume. A wooden hammer ensured uniform compaction across the soil layers. A ring knife facilitated sample extraction from the filled area, enabling verification of the compacted soil layer’s density and moisture content. This rigorous process guaranteed that the model materials in each layer mirrored the weight and moisture content of the prototype slope. Following the construction of the unprotected bare slope, the surface was prepared for vegetation. Bermuda grass, cultivated for a period of two months, was transplanted onto the upper surface of the slope utilizing a block planting method. This method involved extracting mature Bermuda grass sod, which was divided into smaller sections. For each section, a small pit, slightly exceeding the dimensions of the grass block, was excavated on the designated slope. The Bermuda grass sod was then carefully positioned in the pit, covered with soil, and firmly compacted. To promote complete integration of the root system with the surrounding soil, a transparent film was utilized to cover the slope after planting. Following a two-week stabilization period, the experimental phase was carried out.

Figure 5.

Slope stacking surface diagram

(A) Unprotected bare slope, (B) vegetation covered slope.

Prior to initiating the experimental protocol, the prepared soil bed was saturated utilizing an atomized rainfall nozzle and allowed to settle for a 24-h period. Following this equilibration phase, the rainfall experiment began, with displacement sensors and a 3D laser scanner employed to simultaneously monitor internal displacement deformation in the landslide body and displacement deformation of the slope surface, respectively. Simultaneously, pore water pressure sensors and soil pressure sensors facilitated the monitoring of stress in the landslide body, while moisture sensors tracked changes in the internal moisture content of the slope throughout the landslide process. Besides, the sensors measured changes in water turbidity at the outlet following runoff generation. Upon completion of the experiment, a comparative analysis was conducted to assess the stability of unprotected slopes and Bermuda grass-protected slopes under rainfall conditions, explaining the deformation and failure mechanisms of landslides and the positive effects of vegetation in soil stabilization and slope protection.

Result and analysis

Analysis of triaxial compression test results

Triaxial compression tests, the results of which are displayed in Figure 6, were conducted to generate stress-strain curves for both rootless soil and root-soil composites. The data demonstrates that, irrespective of the presence of root systems, the maximum principal stress difference exhibited by the samples increases proportionally with confining pressure. This observation indicates that elevated confining pressure significantly enhances the overall strength of rootless soil. However, this increase in principle stress difference eventually plateaus in rootless soil. Conversely, the stress-strain curve for root-soil composites progressively steepens, suggesting advancements in both overall strength and deformation resistance in the composite. In addition, under identical confining pressures, root-soil composites exhibit significantly greater principal stress differences compared to rootless soil. Specifically, peak growth rates at confining pressures of 50kPa, 100kPa, and 200kPa were observed to be 40.5%, 14.5%, and 10.8%, respectively. These findings illustrate that soil compression deformation increases in response to rising axial strain. Therefore, the root system establishes a resistance relationship with external loads, thereby facilitating the enhancement of soil strength and enabling the root system’s deformation-constraining effect to take place. The root-soil composite plays a crucial role in resisting external loads, and without the contribution of the root system, overall soil strength improvement would be unattainable. Besides, plant roots can be conceptualized as strengthening fibers possessing a certain degree of tensile strength. This strengthening material, arranged longitudinally and interwoven in the soil, forms a robust root-soil composite reinforcement layer. This layer effectively restricts soil deformation caused by external loads, thereby augmenting the overall stability and strength of the soil.

Figure 6.

Stress-strain relationship curve under triaxial compression

Utilizing the Mohr-Coulomb theory, a comprehensive analysis of the triaxial compression test’s stress-strain curve yielded two key shear strength index parameters for the soil, as presented in Table 3. The data indicate 63.3% increase in cohesion in the root-soil composite when compared to soil lacking root presence, emphasizing the significant role of roots in enhancing soil cohesion. These findings suggest that the primary mechanism by which root-soil composites enhance shear strength is through the reinforcement effect of roots on the surrounding soil matrix.

Table 3.

Statistical results of shear strength indicators

Sample	Cohesion/kPa	Internal friction angle/(°)
Rootless soil	37.14	19.03
Root-soil complex	61.52	16.56

Analysis of monitoring data for unprotected slope landslides

Monitoring of changes in soil pressure

The evolution of soil pressure during the landslide process can be categorized into three phases: gradual increase, rapid increase, and stabilization. Figure 7 illustrates the monitoring curves of soil pressure changes in of the landslide mass. In the horizontal direction, as rainfall progresses, water gradually infiltrates the soil, leading to an increase in soil water content and hence, an increase in the internal bulk density of the soil. In addition, the pressure exerted by the overlying soil on the soil pressure sensor increases due to the intensification of rainfall, resulting in increase in soil pressure. As the downward trend continues, the internal soil is subjected to gradual compaction, further contributing to the progressive increase in soil pressure. In addition to vertical infiltration, rainfall also percolates from the top to the bottom of the front edge. Therefore, the soil at the front edge reaches saturation first, as evidenced by the most significant response of the soil pressure sensor EP₇ at section IV. During the formation and evolution of landslides, rainfall-induced cracks initially appear at the toe of the front edge slope, leading to collapse and sliding. The soil pressure change curve of EP₈ exhibits a more prominent response due to the development of cracks at the slope toe during the rainfall event. Moreover, the soil pressure change curve of EP₂ increases to a level similar to that of EP₈. This phenomenon can be attributed to the formation of cracks on the surface of the trailing edge when the rainfall duration reaches approximately 13 min, accompanied by a sliding phenomenon on the slope surface from the trailing edge to the leading edge. Accordingly of soil movement, the EP₂ sensor on the surface of profile I becomes exposed to the soil.

Figure 7.

Monitoring curve of pore soil pressure changes in profiles I-IV

Analysis of the vertical soil pressure fluctuations across landslide profiles I-IV indicates a consistent trend: surface soil pressure invariably surpasses internal soil pressure, irrespective of profile location. This phenomenon is attributable to the differential impact of rainfall infiltration. While precipitation gradually permeates the slope’s interior, the surface experiences continuous erosive forces exerted by rainwater. The velocity of rainwater impacting the slope’s surface exceeds its infiltration rate, leading to an elevated pressure response recorded by surface soil pressure sensors due to the combined effects of impact and erosion. Specifically, a significant difference is present between internal soil pressure EP₁ and surface soil pressure EP₂ at profile position I. This discrepancy arises from the significant changes of the soil at the rear edge during the landslide’s deformation and failure processes, characterized by extensive cracking and collapse. The sliding process initiates with surface soil collapse originating from the rear edge and progressing toward the middle front edge, while the internal soil structure remains relatively unaltered until the landslide’s mass movement. Therefore, EP₂ directly reflects the squeezing effect exerted on the exposed soil surface.

Monitoring of changes in pore water pressure

The pore water pressure variations in the landslide body can be categorized into four stages: slow increase, rapid increase, sharp decrease, and stabilization. An analysis of the pore water pressure variation curve along the horizontal direction of the landslide body (Figure 8) indicates that during the initial phase of rainfall infiltration, the presence of pore gas in the soil maintains the landslide in a stable state. As rainfall progressively infiltrates the slope, forming internal seepage channels, the soil gradually transitions to a saturated state. Therefore, the permeability of the landslide soil increases, initiating slight changes in pore water pressure. With persistent rainfall, the soil reaches full saturation, resulting in a rapid increase in water content in the landslide body, which in turn leads to a sharp rise in pore water pressure. Following the soaking process, the soil softens, and its shear strength experiences a sudden reduction. This instability causes landslide damage, and the sliding motion of the soil expels a portion of the water from the slope, resulting in a decrease in pore water pressure. Upon the cessation of the landslide, the pore water pressure stabilizes, coinciding with the stabilization of the slope. The graph illustrates that P₁<P₃<P₅<P₇, and P₂<P₄<P₆<P₈, indicating a gradual increase in pore water pressure from the top to the foot of the slope, both on the surface and in the interior of the slope. The pore water pressure reaches its maximum at the front edge of the slope foot. This phenomenon can be attributed to the formation of seepage channels in the slope due to rainfall infiltration. Moreover, the combined effect of rainfall and the downward slope trend causes the soil at the front edge of the landslide to reach a saturated state first. Accordingly, the pore water pressure at position IV of the profile exhibits the most rapid and significant increase. The relatively gentle middle edge of the landslide facilitates the infiltration and flow of rainfall toward the front edge, while the steep slope surface at the top and rear edge hinders this process. Accordingly, the increase in pore water pressure at section I is slightly slower compared to sections II and III. These findings suggest that the variation of pore water pressure in response to rainfall is dependent on the soil’s position on the slope, with the growth rate of pore water pressure following the order of leading edge>middle edge>trailing edge.

Figure 8.

Monitoring curve of pore water pressure changes in profiles I-IV

Analysis of pore water pressure variations along the vertical axis of landslide profiles I-IV indicates a direct correlation between proximity to the slope’s surface and the rate of pore water pressure increase. This phenomenon can be attributed to intense rainfall events where the rate of rainwater infiltration into the soil is outpaced by the rate of precipitation reaching the slope surface. Therefore, the surface soil of the landslide experiences rapid saturation, leading to a sharp rise in pore water pressure. Conversely, deeper soil layers exhibit a slower rate of pore water pressure increase under rainfall conditions, which explains the prevalence of rainfall-induced shallow landslides. Specifically, at the rear edge, where slope failure initiates, the surface-level pore water pressure sensor P₂ becomes dislodged due to soil movement. However, the subsurface pore water pressure sensor P₁ remains embedded in the soil, subject to ongoing rainfall infiltration and erosion. This results in a higher pressure reading at P₁ compared to P₂ in profile I, aligning with observed changes in soil pressure at the rear edge of the same profile.

Analysis of landslide monitoring data under vegetation action

Monitoring of changes in soil pressure

To assess the effect of vegetation on slope stability, controlled rainfall experiments were conducted on an indoor landslide model equipped with varying vegetation coverage. The objective was to analyze the resulting soil pressure fluctuations along profiles I-IV (Figure 9). Observations indicated that the presence of vegetation induced a pattern in soil pressure changes, characterized by a steady, incremental rise, finally reaching a state of equilibrium. During the initial rainfall phase, the vegetation canopy effectively intercepted precipitation, causing a majority of the water to remain on the surfaces of leaves and stems or infiltrate into the root zone. Therefore, minimal water penetration into the deeper soil layers resulted in negligible changes in soil pressure. As the rainfall persisted, gradual water seepage through the root system and into the slope interior ensued, leading to a progressive increase in soil pressure. This increase continued until the occurrence of a landslide event, after which soil pressure stabilized. Comparative analysis with soil pressure curves from barren slopes demonstrated that vegetated slopes exhibited a slower rate of soil pressure change, while the overall trend remained similar. In addition, a significant reduction in peak soil pressure was observed in the presence of vegetation, with maximum values on vegetated slopes measuring approximately half those on unprotected slopes. This phenomenon can be attributed to the interception and channeling of rainfall by vegetation, leading to reduced soil infiltration, reduced moisture content, and a decrease in bulk density. Accordingly, the soil pressure values under vegetated conditions were significantly lower.

Figure 9.

Monitoring curve of pore soil pressure changes in profiles I-IV

Study of vertical soil pressure fluctuations across profiles I-IV under vegetated conditions indicates minimal difference between internal soil pressure and surface pressure near the midslope edge. Conversely, at the rear edge of profile I, a significant elevation in surface soil pressure is observed compared to internal pressure. This phenomenon can be attributed to the vegetation cover’s influence on water flow. Specifically, at the rear edge, the majority of precipitation is diverted away by the vegetation’s stems and leaves, with the steep incline further impeding water infiltration into the slope’s interior. Therefore, the internal deformation response at the trailing edge is delayed. During the landslide deformation phase, the trailing edge surface experiences significant collapse and sliding initially. Further analysis of soil pressure at profile position IV demonstrates an initial stage where surface soil pressure EP₈ exceeds internal soil pressure EP₇. However, this trend reverses in the later stage, with EP₇ surpassing EP₈. This can be explained by the initial limited rainwater infiltration due to vegetation cover, resulting in rainwater impacting the surface soil and increasing its pressure while internal soil changes remain negligible. As rainfall duration reaches approximately 45 min, a significant portion of infiltrated water flows toward the front edge of the slope’s base, generating landslide deposits. This overall slope movement induces soil displacement and compression toward the front edge of the base, leading to a gradual rise in internal soil pressure.

Monitoring of changes in pore water pressure

Vegetation significantly affects pore water pressure changes in slopes, exhibiting four phases: initial stability, gradual rise, rapid decline, and eventual stabilization. In the early stages of rainfall, the presence of vegetation intercepts precipitation through its canopy, resulting in near-zero horizontal pore water pressure in the slope. As rainfall progresses, the pore water pressure trends reflect those observed in unprotected slopes, as depicted in Figure 10. Rainfall infiltration occurs both vertically and laterally, with the highest pore water pressure values observed at the slope’s base. The rate of pore water pressure increase follows a consistent pattern, with the front edge exhibiting the most significant rise, followed by the middle and rear edges, respectively. This suggests that vegetation exerts a minimal effect on the overall horizontal pore water pressure distribution. Specifically, vegetation significantly contributes to the sharp decline in pore water pressure observed during rainfall events. The peak pore water pressure in unprotected slopes is approximately double that of vegetated slopes. This phenomenon is attributed to the vegetation’s ability to attenuate raindrop impact, thereby reducing soil permeability and reducing rainfall infiltration.

Figure 10.

Monitoring curve of pore water pressure changes in profiles I-IV

While vegetation influences slope stability, the fundamental behavior of pore water pressure remains consistent with that observed in unprotected slopes. Rainfall induces initial saturation at the slope surface, leading to consistently elevated pore water pressure at the front edge compared to the interior. Vegetation cover does not prevent the primary failure mode, which involves the sliding of the rear edge surface. Therefore, the surface-level pore water pressure sensor P₂ is displaced due to soil movement, whereas the interior sensor P₁ remains embedded despite erosion and infiltration caused by rainfall. This results in a gradual increase in pore water pressure at the P₁ position in Profile I, attributed to ongoing rainfall infiltration following the rear edge surface slide.

The preceding analysis indicates that under extreme rainfall events, vegetated slopes exhibit stress field changes comparable to those observed on bare slopes. However, the presence of vegetation significantly reduces pore water pressure and soil pressure in the slope, effectively delaying stress field fluctuations. This phenomenon arises because, on bare slopes, rainfall directly infiltrates the soil, leading to significant changes in the internal stress field due to the absence of significant surface runoff. Conversely, on vegetated slopes, rainfall is intercepted by the vegetation canopy, resulting in a partitioning of water between infiltration and surface runoff. This partitioning mechanism, facilitated by vegetation, attenuates the effect of rainfall on the internal stress field of the slope.

Analysis of landslide moisture content and turbidity

Analysis of the characteristics of moisture content changes

During the landslide experiment, soil moisture content at four profile locations on the slope was measured at 5-min intervals, generating a monitoring curve illustrating the change in water content across profiles I-IV, as depicted in Figure 11. The data indicate that the landslide sensor experiences significant fluctuations under the erosive force of a 75 mm/h rainfall intensity. The overall trend of these fluctuations can be categorized into three phases: stable, rapid, and gentle. This pattern arises due to the initial stage of rainfall, where water has not yet infiltrated the slope. The second stage corresponds to the infiltration phase, characterized by a rapid rise in internal soil moisture content. Finally, the change in moisture content gradually stabilizes as the slope reaches saturation. Specifically, the period of rapid moisture content increase under vegetated slope protection significantly lags behind that of the bare slope. This delay can be attributed to the interception effect of vegetation stems and leaves during the initial rainfall stage. However, in the later stages, the effect of vegetation cover is evident, leading to a more significant change in slope moisture content compared to the bare slope. This phenomenon is due to the combined effect of water absorption by vegetation roots and the enhanced infiltration capacity of vegetated soil, resulting in a more rapid increase in internal moisture content in vegetated slopes.³⁸ The data indicates a peak moisture content of 34.5% for the bare slope and 46.7% for the vegetated slope, with both peaks observed at position IV of the profile. This observation highlights that regardless of vegetation presence, the most significant change in moisture content occurs at the slope’s foot. This is because the moisture increase at this location results not only from the vertical infiltration of rainfall but also from the seepage of water flowing downslope.

Figure 11.

Monitoring curve for changes in water content of slopes in sections I-IV

Analysis of the characteristics of moisture content changes

In the conducted experiment, the initial runoff time exhibited a significant difference between the bare slope and the vegetated slope, with the former initiating runoff at 13 min and the latter at 18 min. Subsequent to the onset of slope runoff, sediment samples were systematically collected from the outlet at 3-min intervals. Following a weighing procedure, each sample is subjected to a drying process in a 105°C oven for a duration of 24 h. This established the dry mass of the solid sediment component in the turbid water. Turbidity, defined as the ratio of sediment concentration to the runoff sample, was then computed utilizing the formula exhibited in Equation 1:

$$T=\frac{M_s}{M_t}$$ (Equation 1)

In the formula, $T$ represents turbidity, %; $M_s$ is the dry mass of solid in the runoff sample, g; $M_t$ is the mass of runoff sediment samples taken every 3 min, g.

Figure 12 illustrates a consistent decline in turbidity over time, exhibiting three phases: a phase of rapid decrease, followed by a period of gradual change, and resulting in a phase of stabilization. This observed pattern can be attributed to the relations between rainfall and soil moisture content. During the initial phase, as rainfall persists, soil moisture content progressively increases, leading to a conversion of rainfall into runoff. This increased runoff, possessing augmented erosive capacity, results in higher turbidity levels. As the rainfall event progresses into its later stages, the soil moisture content approaches saturation, and surface runoff stabilizes, causing turbidity to gradually decline and reach a state of relative equilibrium. In addition, a comparative analysis indicates significantly lower turbidity values for vegetated slopes compared to bare slopes, with a reduction ranging from 1.81% to 3.28%. This disparity can be attributed to the erosion mitigation effects of vegetation cover. The presence of vegetation shields the slope surface from the disruptive impact of raindrops, with stems and leaves effectively intercepting and buffering rainfall. Besides, the root systems of plants contribute to soil stabilization and slope reinforcement, thereby attenuating the erosive forces of rainfall and runoff. Therefore, soil loss on vegetated slopes is minimized, leading to reduced turbidity levels in comparison to bare slopes.

Figure 12.

Landslide turbidity variation curve

Analysis of deformation and failure characteristics of landslides

Analysis of maximum displacement variation characteristics of landslides

Analysis of displacement variations in sections of the landslide soil during the rainfall event indicates insightful trends, as depicted in Figure 13. The initial phase of rainfall induces surficial erosion of the slope by impacting water droplets, leading to the gradual formation of minor cracks at various points on the slope. This phenomenon results in subtle displacement fluctuations. As the rainfall persists and intensifies, the slope experiences the formation of large cracks, resulting in eventual collapse and landslide events. Therefore, displacement exhibits a rapid rise followed by stabilization. Comparative analysis highlights the accentuated displacement observed in the bare slope scenario relative to the vegetated protection slope condition. Specifically, the maximum displacement reduction rates at positions I, II, III, and IV in the vegetated protection slope demonstrate reductions of 38.3%, 22%, 11.8%, and 63.6%, respectively, when contrasted with the bare slope. In addition, the bare slope surfaces S1 and S4 exhibit comparatively larger displacement magnitudes. This observation is attributed to the initiation of cracks and significant deformation at the leading edge of the slope, followed by the occurrence of large cracks and landslides at the trailing edge. The deformation rate at the middle edge lags behind that of the front and rear edges. This pattern suggests that the extreme rainfall event exerts a significant effect on the deformation stability of both the leading and trailing edges of the slope. Conversely, under the vegetated slope protection condition, the maximum landslide displacement at various positions demonstrates relative stability. This phenomenon can be attributed to the role of vegetation stems and leaves in intercepting and attenuating raindrop impact, coupled with the soil stabilization and slope protection function of the root system. These combined effects reduce the erosive intensity of rainfall and runoff, thereby reducing soil loss on the slope surface and preserving the integrity of the surface soil layer. Accordingly, the deformation of landslides under vegetated conditions is less severe compared to bare slopes, and displacement changes at each position remain insignificant.

Figure 13.

Maximum displacement change curve of landslide

3D laser scanning analysis

A 3D laser scan of the post-failure topography generated a displacement cloud map, indicating the landslide’s surface movement (Figure 14). Analysis of the map demonstrates that rear edge displacement exceeded that of the leading edge in both vegetated and bare slope scenarios. Specifically, slopes with vegetation cover exhibited less overall surface displacement compared to bare slopes. Despite the complete destabilization of the slope due to rainfall, vegetated areas displayed reduced soil slippage at the rear edge, with more significant cracking and significantly less displacement at the leading edge. Conversely, bare slopes lacked cracking and failed as a cohesive mass, resulting in greater momentum and significantly larger displacement distances at both the front and rear edges.

Figure 14.

3D laser scanning landslide displacement cloud map

(A) Bare slope, (B) vegetation covered slope.

A comparative analysis of surface displacement across varying slopes under extreme rainfall conditions indicates the significant effect of vegetation on landslide mitigation. While both vegetated and non-vegetated slopes exhibit damage during intense rainfall, the presence of vegetation significantly reduces both the volume of displaced soil mass and the extent of slope movement.

The landslide was partitioned into three sections: left, middle, and right. Utilizing three-dimensional scanning data points extracted from the landslide’s middle section, a position-distance relationship curve was generated, as depicted in Figure 15. The positions 0–3000 on the x axis correspond to the location from the landslide’s rear edge to its front edge. Analysis of the curve indicates a significantly smaller sliding distance for the vegetated slope protection compared to the bare slope. The graph’s peaks, from left to right, primarily coincide with the rear and front edges of the landslide. Specifically, the sliding distance of the vegetated slope protection is 0.038 m less than that of the bare slope, with a 0.023-meter reduction observed at both the rear and front edges. In addition, the peak sliding distance at the rear edge surpasses that of the front edge by 0.095 m on the bare slope and 0.08 m on the vegetated slope protection. This finding aligns with the results obtained from the three-dimensional laser scanning cloud map and is consistent with the overall deformation and failure patterns of the landslide.

Figure 15.

3D scanning of the position distance relationship curve in the middle of a landslide

Analysis of landslide deformation characteristics

Utilizing Geo-slope software, an analysis of slope stability under rainfall conditions was conducted for both bare and vegetated slopes. Results indicated a decline in the factor of safety for bare slopes, decreasing from an initial value of 1.84 to a final stabilized value of 1.21. Comparatively, vegetated slopes exhibited a reduction in the factor of safety from 2.14 to 1.44. The presence of vegetation significantly enhanced slope stability, with an average increase of 17.9% in the stability coefficient observed during rainfall events.

The destabilization and failure progression of unprotected slope failures can be categorized into three phases: crack initiation and propagation, crack coalescence and breakthrough, and trailing edge collapse leading to mass wasting, as illustrated in Figure 16. Upon the beginning of precipitation, in approximately 3 min, surficial soil layers are subject to erosion by runoff, exposing underlying lithological features. Thereafter, at the 5-min mark, localized depressions appear on the slope surface due to the effect of raindrops. As the rainfall duration reaches 6 min, the soil structure at the leading edge of the potential slide mass is subjected to rapid saturation and loses stability, resulting in the formation and upward extension of incipient cracks. Around the 9-min mark, similar crack patterns appear near the slope’s midsection. Upon reaching the 13-min mark, the slope’s toe experiences the initiation of soil flow, followed by the downslope movement of the upper portion of the trailing edge. The trailing edge, characterized by instability, serves as the primary driving force for the impending landslide. With sustained precipitation over a 16-min interval, continued infiltration leads to an expansion of the saturated zone in the slope, thereby increasing the driving forces and reducing the shear resistance of the soil mass. This progressive decline in stability culminates in the initiation of mass wasting. During this phase, the displaced material migrates along the failure surface toward the leading edge, forming an accumulation zone. Over the following 26 min, the slope continues to fail along the leading edge, resulting in complete destabilization. Simultaneously, water accumulates in the upper section of the debris mass, leading to the saturation of the entire landslide body.

Figure 16.

Landslide deformation and failure diagram of vegetation covered slope

(A) Bare slope, (B) vegetation covered slope.

Discussion

Under the protection of vegetation cover, the deformation and damage processes of landslides are significantly decelerated, albeit the deformation and damage modes remain consistent with those observed on bare slopes. The vegetation slope protection mechanism can be categorized into three stages: “interception and buffering of stem and leaf rainfall, infiltration of rainfall into root system water absorption, and reinforcement of slope sliding root system,” as illustrated in Figure 15. During the initial phase of rainfall, vertical infiltration predominates as the primary form of water movement. The rainfall interception effect of plant stems and leaves, coupled with the weakening of the erosion reduction function and the inhibition of surface runoff, safeguards the slope surface, resulting in no significant changes to the slope body. As the water absorption effect of vegetation roots gradually allows rainwater to permeate the soil, water in the soil progressively seeps from the top to the bottom of the slope. Upon reaching 15 min of rainfall, the upper portion of the slope foot temporarily attains saturation, leading to an increase in soil weight and sliding force, accompanied by the formation of cracks at the front edge. With the persistence of rainfall for 27 min, cracks manifest at the top of the rear slope, concurrent with collapse, and these cracks continue to widen and deepen. At the 30-min mark, the upper layer of the slope experiences sliding, emphasizing the crucial role of the enhanced infiltration effect of vegetation in triggering shallow landslides, particularly under extreme rainfall conditions. However, due to the reinforcement effect of the root system, soil strength is augmented. Therefore, when the landslide is delayed for 45 min, the slope undergoes comprehensive sliding along the sliding surface. It can be inferred that irrespective of the presence or absence of vegetation protection on the slope, under the effect of heavy rainfall, the precipitation induces cracks and collapse of the slope, forming the sliding surface, which serves as the critical factor in inducing landslides.

Conclusion

• (1)Triaxial compression testing indicated a significant difference in principle stress between root-soil composites and soils devoid of root structures. Specifically, peak growth rates under confining pressures of 50kPa, 100kPa, and 200kPa were measured at 40.5%, 14.5%, and 10.8%, respectively. Specifically, the cohesion of the root-soil composite exhibited a 63.3% increase compared to its non-root counterpart. These findings emphasize the role of root structures in soil composites in enhancing shear strength through reinforcement mechanisms.
• (2)Under extreme rainfall events, irrespective of vegetation coverage, pore water pressure consistently registers higher at the slope surface compared to its interior. Implementing Bermuda grass as a protective covering on landslide-prone slopes significantly reduced both pore water pressure and soil pressure in the slope compared to bare slopes. Besides, the presence of vegetation delayed the stress field change process in the slope.
• (3)Through the embedding of displacement sensors at various sections of landslide bodies, coupled with 3D laser scanning tests, a significant reduction in displacement was observed in slopes adorned with vegetation. Compared to bare slopes, the maximum displacement reduction rates at of vegetation-protected slopes were 38.3%, 22%, 11.8%, and 63.6%, respectively. This pattern indicates the most significant displacement changes occur at the leading and trailing edges of the landslide body.
• (4)The fluctuation of water content throughout the landslide process can be categorized into three stages: stable, rapid, and gentle. In vegetation-protected slopes, the rapid phase of water content change in the downhill segment exhibits a significant lag compared to bare slopes. Upon reaching a state of destruction and stabilization, the water content of vegetation-protected slopes surpasses that of bare slopes. Regardless of vegetation presence, the most significant change in water content is consistently observed at the slope’s base. Turbidity demonstrates a consistent decline over time, with vegetation-protected slopes exhibiting a 1.81%–3.28% lower turbidity level compared to their bare counterparts. This observation can be attributed to the interception and buffering effect of surface vegetation’s stems and leaves, which effectively reduce soil loss on the slope surface.
• (5)Analysis of diverse slope failure modes under extreme rainfall conditions indicated an average increase of 17.9% in the stability coefficient of vegetation-protected slopes compared to bare slopes. In addition, landslide deformation and failure can be categorized into three stages: crack generation and deepening, crack expansion and penetration, and rear edge collapse leading to overall sliding. The effect of vegetation protection on slopes can be similarly segmented into three stages: interception and buffering of rainfall by stems and leaves, infiltration of rainfall into the root system for water absorption, and reinforcement against sliding offered by the root system.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Other
Mean annual precipitation data	China Meteorological Data Network	http://data.cma.cn/

Resource availability

Lead contact

Requests for further information and resources should be directed to the lead contact, Can Wei (wc1995@ctgu.edu.cn).

Materials availability

This study did not generate new materials.

Data and code availability

All data can be obtained from the lead contact, provided the request is reasonable.

The code related to the attribution model can be accessed by reaching out to the lead contact.

Method details

Indoor experiment details

This study employed a designed indoor physical model to assess landslide behavior. The model, constructed with dimensions of length × wide × height=3 m × 2 m × 2m, utilized transparent organic glass sidewalls to facilitate direct observation of deformation and failure mechanisms during experimentation. A flowmeter, positioned at the inlet of the rainfall sprinkler system, offered accurate data for determining rainfall intensity. The sprinkler system itself employed atomizing nozzles, and rigorous pre-experiment testing ensured rainfall uniformity exceeding 80% throughout the model. The physical model was conceptualized to represent the fundamental structural components of a landslide: the sliding bed, sliding zone, and sliding body. The sliding bed, constructed with ordinary bricks and surfaced with a layer of cement mortar for impermeability, simulated the bedrock upon which the landslide occurs. Maintaining consistency with the slope of the studied landslide, the model was set at a 30° incline. Both the sliding zone and sliding body were filled with soil in layers and sections, with the filling quantity determined based on accurate calculations of soil density and volume. The experimental soil samples were extracted from both the undisturbed and landslide zones at a 155m elevation in the Three Gorges Reservoir Area. Following the "Soil Test Methods Standard" (GB/T 50123 2019),³⁴ a series of fundamental geotechnical analyses were performed on the undisturbed soil to determine its intrinsic physical properties and structural composition, alongside those of the landslide zone soil. Adhering to rigorous standard specifications, direct shear tests, soil moisture content tests, soil natural density tests, and soil specific gravity tests were executed.

The leaf blades typically range from 1-10 cm in length and 0.2 cm in width. These attributes make it an ideal candidate for vegetation restoration efforts in the Three Gorges Reservoir subsidence area.^35,36,37 To assess its tensile properties, we conducted single root tensile tests utilizing an HP-100 Edelberg digital push-pull force meter with a 100 N range and 0.50% accuracy. The results, presented in Table 2, indicate a root diameter range of 0.13-0.50 mm for Bermuda grass, with maximum tensile strength and root tensile strength for individual roots ranging from 0.9-5.1 N and 25.99-67.84 MPa, respectively.

Background climate

To accurately replicate real-world precipitation patterns in the Three Gorges Reservoir region, researchers collected and analyzed relevant meteorological data. This analysis, coupled with an study of 1-hour maximum rainfall events (ranging from 55mm to 110mm) in the Hubei section of the reservoir, indicated a significant concentration (89%) of intense rainfall days below 50mm. Specifically, 51.4% of these events fell in the 20-29.9mm range, while 37.6% registered between 30-49.9mm. Rainfall exceeding 50mm constituted only 10% of the observed events, with sporadic occurrences of extreme precipitation surpassing 90mm during the summer months. Calculations determined the rainfall intensity for 50-year and 100-year events in the Three Gorges Reservoir area to be 50mm/h and 75mm/h, respectively. Based on these findings, this study employed a controlled indoor environment to simulate rainfall conditions, focusing on the extreme scenario of a 100-year event (75mm/h intensity for 1 hour duration). The mean annual precipitation data of Three Gorges Reservoir area can be found from the China Meteorological Data Network (http://data.cma.cn/).

Published: May 22, 2024