Introduction

Rainfall is one of the most important factors in slope failure. Of more than 90 slope failures in China listed by Sun (1988), over 95% were caused by or closely related to rainfall. According to Zhong (1999), 15 geohazards in 27 case studies were induced by rainstorms. Recently, increasing attention has been paid to rain-induced slope failure and its mechanisms (Au 1998; Ng and Shi 1998a; Cho and Lee 2001; Ying et al. 2002; Sun et al. 2009; Lin et al. 2009; Hossain 2010; Schnellmann et al. 2010), methods for evaluating slope stability (Ping et al. 2004; Wei et al. 2006; Zhu et al. 2006; Liu et al. 2007; Lou 2007; Wu et al. 2008), and forecasting of rainfall-induced slope failure (Ng and Shi 1998b; Miller et al. 2009). These studies have demonstrated that rainfall can cause a change in the slope seepage field, decrease the matric suction of soil, and increase the pore water pressure and soil unit weight. In addition, an increase in water content softens soil and decreases the shear strength of the slip zone. Slope stability is affected by various parameters including total rainfall, rainfall intensity and duration, rainfall patterns, soil properties, and the topography of the slope.

It is well recognized that intense rainfall can trigger shallow landslides on homogeneous soil slopes. Both theoretical and experimental research has suggested that the depth of rainfall infiltration in a homogeneous soil mass is extremely limited. Indeed, in situ monitoring showed rainfall infiltration depth of less than 2.0 m in a homogeneous expansive clay soil (Zhang et al. 2003). Based on finite-element method (FEM) calculations, it was proposed that the infiltration depth with 12 h of rainfall was only 2.0 m in a filling subgrade consisting of silty clay soil (Zhang and Wang 2004). Using the theory of unsaturated seepage, Chen et al. (2009) deduced a formula for calculating the maximum depth of rainfall infiltration from conservation of energy, and illustrated that the maximum depth of infiltration in silty clay soil was less than 4 m.

Many deep soil landslides have been reported during or shortly after periods of intense rainfall (Liu 2014; Sun 1988; Guthrie et al. 2009; Wu et al. 2013; Azzoni et al. 1992; Chen and Wu 2018). However, the mechanisms of rainwater infiltration into deep soil (over 20 m) and landslide triggering remain debated. For many deposits consisting of coarse soil, such as rocky landslide deposits and alluvial/diluvial deposits, the saturated permeability is approximately 10−4 to 10−6 m/s, and the unsaturated permeability is much lower. If those deposits were considered homogeneous, it would be difficult for rainwater to infiltrate the deep soil (over 20 m depth) and trigger failure of the deep deposit slopes during or shortly after periods of intense rainfall.

However, it is possible that deep deposit slopes are not always homogeneous; for example, cracks or crushed stone soil with high hydraulic conductivity could offer natural channels for rapid rainfall infiltration of deep soil. Tang et al. (2015) showed that fractures and holes in loess could quickly form infiltration channels that significantly influenced loess slope stability. Sun et al. (2009) showed that macropores and fissures in soil may supply preferential paths for water infiltration into deep soil. Li et al. (2016) suggested that gravel and rock fragments in the top layers of landslide deposits triggered by the 2008 Mw8.0 Wenchuan earthquake play an important role in rainwater infiltration.

The studied landslide deposit lies on the right bank of Lancang River, at a distance of 5 km from the dam of the Gushui Hydropower Station downstream. It has a volume of 4.7 × 107 m3. After a heavy rainfall that occurred on 7–9 October 2012, many cracks emerged on the surface of the landslide deposit, indicating that reactivation of the old landslide deposit was triggered. Such a phenomenon could block the Lancang River and threaten the safety of residents on both banks.

As a case study, the objective of this work was to analyze the rainfall infiltration process in an old landslide deposit with a particular geology, namely the presence of volumes of highly permeable material deposited locally in the landslide body, surrounded by less permeable soils. We investigated the geological characteristics of the old landslide deposit and performed mechanical and hydraulic characterization. Using a FEM model and a physical experimental model, we studied the infiltration process in the landslide deposit and the variation in stability during rainfall.

Landslide deposit characteristics

The study area is situated in Foshan Town, Deqin County, Yunnan Province, China. The geographic coordinates are E 98°47′–98°36′ and N 28°33′–29°10′. It is in the eastern margin of the Qinghai-Tibet Plateau, and belongs to the geomorphic unit of the Longitudinal Valley in Western Yunnan. The terrain is high in the north and low in the south. The study area of the landslide deposit is characterized by a monoclinal structure. The layer of bedrock inclines to the inside of slope with an occurrence of N20–30°W/SW∠70–80°. Toppling deformation of rock mass is common in the study area.

The studied landslide is located on the right bank of the Lancang River. It is an old landslide, formed many years before people lived there. The elevations of the rear and front edges are 2860 and 2060 m, respectively, producing a maximum elevation difference of 800 m. The minimum and maximum landslide accumulation widths are 150 and 700 m, respectively. The landslide is approximately 850 m long, and the accumulation area is 7.8 × 105 m2. The maximum thickness of the accumulation is 110 m, with a mean thickness of 60 m and volume of 4.7 × 107 m3 (Figs. 1 and 2).

Fig. 1
figure 1

Aerial map of the study area

Full size image
Fig. 2
figure 2

Engineering geological section of the landslide

Full size image

The landslide terrain is steep, with a mean slope of approximately 30°. The rear and front edges of the landslide are even steeper, with slopes of approximately 35°–40°; a platform in the middle of the landslide has a gentler slope of approximately 7°–12°.

The area experiences a semi-arid climate, with mean annual precipitation of 494.7–522.4 mm. Heavy rainfall occurred on 7–9 October 2012, with 3-day cumulative precipitation of up to 262.5 mm—the largest rainfall in the area since the meteorological record was established in 1952. After the rainfall, many cracks appeared on the landslide surface, indicating the possibility of reactivation. The local government relocated residents for their safety and instructed the builder of a planned hydropower station located 5 km upstream to survey and assess the site. Those adits and drill hole surveys (Fig. 1) were performed by the builder. In addition, the authors were invited to conduct a cooperative investigation and study in December 2012.

Field investigation on landslide deposit

The builder of the hydropower station made 25 drill holes and five horizontal adits (Fig. 1) to survey the formation and structural characteristics of the landslide deposit from October to December 2012.

According to the horizontal adit and drill hole surveys, the landslide deposit consists mainly of densely compacted clayey gravel soil (Fig. 3). In addition, there are two zones consisting mainly of crushed stone soil (Figs. 1 and 3): (1) a zone close to the river with a width of 380 m, length of 130 m, and maximum thickness of 45 m; and (2) the remainder, with a width of 270 m, length of 150 m, and maximum thickness of 30 m.

Fig. 3
figure 3

Characteristics of clayey gravel soil and crushed stone soil: a, b, c crushed stone soil; d, e, f clayey gravel soil

Full size image

The clayey gravel soil is composed mainly of cobbles (60–200 mm diameter, weight percentage > 10%), gravels (2–60 mm, 50%–60%), sands (0.075–2 mm, 10%–20%), and fines (< 0.075 mm, 15%–20%). The crushed stone soil is relatively loosely compacted, and consists mainly of boulders (200–630 mm, 20%–35%), cobbles (25%–40%), gravels (25%–30%), sands (approximately 10%), and fines (< 2%). The grain size distribution is shown in Fig. 4.

Fig. 4
figure 4

Soil grain size distribution: a clayey gravel soil; b Crushed stone soil

Full size image

The old slip zone of the landslide is moist and consists mainly of pebbly clay with a thickness of approximately 0.5–2 m. Gravel in the slip zone is subangular, with a grain size of approximately 2–20 mm; the weight percentage of the gravel is less than 10% (Fig. 5). According to the engineered section of the landslide (Fig. 2), the rear of the slip zone is relatively steep, with a dip angle of 30°–33°. The middle of the slip zone has a dip angle of 18°–25°, while the front of the slip zone is nearly horizontal, with a dip angle of approximately 1°–3°.

Fig. 5
figure 5

Characteristics of the original slip zone

Full size image

The bedrock is composed of metabasalt and sandy slate of the lower Permian Jidonglong formation, with a normal occurrence of approximately N20–30°W/SW∠70–80° (Fig. 2).

At present, 26 cracks have been found on the surface of the landslide (Figs. 1 and 6). The cracks can be divided into two groups according to strike. One group is transverse, extending 15–145 m along a strike of 280°–330°. These cracks have widths of 0.15–1 m and extend to a measurable depth of 0.2–1.5 m. The remaining cracks are with lengths of 20–125 m along a strike of 30°–75°, widths of 0.2–2 m, and extend to a measurable depth of 0.2–2 m.

Fig. 6
figure 6

Representative cracks in the landslide: a crack L4; b crack L5; c crack L6; d crack L9; e crack L12; f crack L19

Full size image

Tests for soil characterization

The saturated permeability of the clayey gravel and crushed stone soils was tested in situ using a single ring method shown in Fig. 7. The saturated permeability is shown in Table 1.

Fig. 7
figure 7

In situ testing for saturated permeability of soil: a crushed stone soil; b clayey gravel soil

Full size image
Table 1 Derived parameters of the Van Genuchten model
Full size table

For laboratory testing of the physical, mechanical, and hydraulic properties of the soil, samples of crushed stone soil and clayey gravel soil were taken at 26 m and 58 m from the entrance of horizontal adit PD1003, respectively, and the samples of the slip zone were taken at 113 m from the entrance of adit PD1003 and 145 m from the entrance of adit PD1031, respectively (Figs. 3 and 5).

Matric suction data for the clayey gravel, crushed stone soil, and pebbly clay (slip zone) were collected under different water content conditions (Fig. 8) using a full-suction double-cell extractor for SDSWCC (stress dependent soil–water characteristic curves; Earth Products China Ltd). The soil sample was 90 mm in diameter and 25 mm in height. For clayey gravel soil and crushed stone soil, particles with diameters larger than 5 mm were removed. After sampling, suction saturation was performed in a vacuum bucket. Dehydration test data were obtained by a pressure plate extractor. Vertical pressure tests were applied stepwise at 10, 20, 35, 50, 100, 200, 350, 500, and 700 kPa.

Fig. 8
figure 8

Soil–water characteristic curves for the landslide

Full size image

Using the Van Genuchten model (Van Genuchten 1980), which is listed as Eq. (1), to fit those measured data, the soil–water characteristic curves were derived. The derived parameters in Eq. (1) are listed in Table 1, and the soil–water characteristic curves are shown in Fig. 8.

$$ \theta ={\theta}_r+\left({\theta}_s-{\theta}_r\right)/{\left[1+{\left(\alpha {p}_c\right)}^n\right]}^m $$
(1)

where θ is the volume water content of the soil; θr is the residual water content of the soil; θs is the saturated water content of the soil; pc is the matric suction; and α, n, m are curve fitting parameters (note: “α” has units of 1/pressure, m = 1-1/n).

The permeability coefficients of unsaturated soil (Fig. 9) were calculated from Fig. 8 using the approach described by Fredlund and Rahardjo (1993):

$$ {k}_w{\left(\theta \right)}_i=\frac{k_s}{k_{se}}{A}_d\frac{\theta_s^2}{N^2}\sum \limits_{j=1}^m\left\{\left(2j+1-2i\right){\left({u}_{\alpha }-{u}_w\right)}_j^{-2}\right\} $$
(2)

where kw(θ)i is the predicted water coefficient of permeability for a volumetric water content, θi, corresponding to the ith interval (m/s); i is the interval number, which increases as the volumetric water content decreases (e.g., i = 1 identifies the first interval, which is close to the saturated volumetric water content, θs); i = m identifies the last interval corresponding to the lowest volumetric water content on the experimental soil–water characteristic curve, θr; j is a counter from “i” to “m”; ks is the measured saturated coefficient of permeability (m/s); kseis the calculated saturated coefficient of permeability (m/s); m is the total number of intervals between the saturated volumetric water content, θs, and the lowest volumetric water content on the experimental soil–water characteristic curve, θr (i.e., m = 20); (ua− uw)j is the matric suction corresponding to the jth interval (kPa); and Ad is the adjusting constant, which was assumed to equal 1 in this study.

Fig. 9
figure 9

Unsaturated permeability coefficient as a function of matric suction

Full size image

The shear strengths (c, φ), elastic modulus, and Poisson ratio of the crushed stone soil and clayey gravel soil were tested using a coarse soil consolidated-drained triaxial test (Fig. 10). The samples of crushed stone soil and clayey gravel soil were 30 cm in diameter and 60 cm in height; particles with diameters larger than 60 mm were removed. The confining pressure was set as 80, 100, 150, 200, and 300 kPa.

Fig. 10
figure 10

Coarse soil consolidated-drained triaxial test samples (after test): a clayey gravel soil; b crushed stone soil

Full size image

A consolidated drained triaxial test was used for testing the shear strength, elastic modulus, and Poisson’s ratio of the slip zone. The slip zone samples had a diameter of 61.8 mm and height of 125 mm. The confining pressure was set as 30, 50, 100, 150, and 200 kPa. The natural water content and porosity of the soil were tested using an oven drying method. The testing results are shown in Table 2.

Table 2 Physical and mechanical parameters of the landslide
Full size table

Physical model experiment and FEM validation

To investigate the mechanism of infiltration and the effect of crushed stone soil, a physical experimental model was built (Fig. 11). The clayey gravel soil and crushed stone soil in the model were sourced from the landslide; particle compositions were in accordance with the parameters in Fig. 4. The crushed stone soil was covered by a 15-cm layer of clayey gravel to enable comparison of infiltration between soil types.

Fig. 11
figure 11

Structure of the physical experimental model

Full size image

Four pore water pressure devices (HC-25 Micro Pore Water Pressure Sensors, Earth Products China Ltd.; measuring range: −10 to 15 kPa, accuracy: ±0.1 kPa) were installed, two (P1, p2) at the top (depth = 10 cm) and two (P3, p4) at a depth of 60 cm. Six 503DR moisture sensors (InstroTek/CPN; measuring range: 0–100%, accuracy: ±2%) were installed; M1 and M2 were installed at the top (depth = 10 cm), M3 and M4 were installed at a depth of 40 cm, and the final two were installed at the bottom (depth = 60 cm; Fig. 11b).

The rain intensity was set to 10 mm/h for 48 h. The artificial rainfall system consisted of a supercharging unit, a rainfall unit, and a control unit. The rainfall unit consisted of four sprinklers, which were set above the physical model. By controlling water pressure, the rainfall intensity was held at 10 mm/h during the test period. The test procedures were as follows. (1) Soil was compacted into layers according to the natural density, with PVC tubes buried at depths corresponding to monitor placement. (2) After compaction, the PVC tubes were removed, and sensors were placed in the correct positions. (3) Sensors were connected to the data acquisition system and adjusted until a suitable data stream was achieved. (4) Rain was generated over the model according to the designed rain intensity and duration. (5) Values of water content and pore water pressure were collected every 15 min. A digital camera was used to video the entire test process. (5) The test ended when the data from all monitors was stable. During the process of filling the physical model, particles with a diameter larger than 50 mm were removed.

Figure 12 shows the progression of the wetting front as it moved from the surface to the bottom of the model during rainfall. The wetting front moved down 3–7 cm in the first 2 h 47 min of rainfall, increasing in velocity when it moved into the crushed stone soil. It moved down approximately 30 cm and 15 cm in the crushed stone soil and clayey gravel soil, respectively, over the next 1 h 3 min. The wetting front moved much faster in the crushed stone soil than it did in the clayey gravel soil, and developed a distinctive kink at the border of the two soils. Comparing its shape at 3 h 50 min, 5 h 16 min, and 7 h 45 min, the wetting front not only moved down vertically, but also expanded into the surrounding soil in a horizontal direction; this increased the velocity of the wetting front in the surrounding clayey gravel soil.

Fig. 12
figure 12

Movement of wetting front during physical experimental testing. a Initial moment; b 2 h 47 min; c 3 h 50 min; d 5 h 16 min; e 7 h 45 min; f 9 h 30 min

Full size image

Figure 13 presents variations in water content and pore water pressure in the model during the test. In terms of volumetric water content (Fig. 13a), sensors M1 and M2, at a depth of 10 cm, increased to 17.3% and 16.5%, respectively, after rainfall of 13.5 to 14 h. The water content of sensor M4 at a depth of 40 cm increased to 27.6% after 15 h of rainfall. The water content at M3 increased to 27.5% after 27.5 h. Sensors M5 and M6 were both installed at 60 cm depth; M6 was placed 10 cm under the crushed stone soil. The water content of M5 and M6 increased to 29.2% and 27.3%, respectively, after 20 h of rainfall. The variation in water content indicated that it increased faster and earlier in the clayey gravel soil at the same depth where was closer to the crushed stone soil than elsewhere after the wetting front moved into the crushed stone soil.

Fig. 13
figure 13

Variations in water content and pore water pressure in the physical experimental model: a variation in water content; b variation in pore water pressure

Full size image

Figure 13b shows the variations in pore water pressure monitored by sensors. The pore water pressure at sensors P1 and P2, at a depth of 10 cm, increased from initial values of −0.2 and −0.3 kPa to 0.6 and 0.15 kPa, respectively, after 1 h of rainfall. The pore water pressure at sensors P3 and P4, which had an installation depth of 60 cm, began to increase after 10–12 h of rainfall, and increased to 5.1 kPa and 5.02 kPa, respectively, at the end. The high pore water pressure formed at the location of P3 and P4 can be attributed to the increase in the water table.

A finite-element method-based (FEM) model was built as a validation model. Then, if the calculation results of the FEM were similar to the test results of the physical model, another model could be built for the real landslide deposit in the same way to investigate the real rainfall infiltration process.

The physical and mechanical parameters of soil in the FEM validation were assigned according to Table 2. The original water table was assumed at the bottom of the model, and the initial water content of soil was assigned as natural water content (Table 2).

The governing equation (Richards equation) for rainwater seepage in unsaturated-saturated soil was:

$$ \frac{\partial }{\partial {x}_i}\left[\frac{1}{\gamma_w}{k}_{ij}{k}_r(p)\frac{\partial p}{\partial {x}_j}+{k}_{i3}{k}_r(p)\right]=\left(C(p)+\alpha {S}_s\right)\frac{\partial p}{\partial t} $$
(3)

where kij is the permeability tensor coefficients of saturated soil; kr(p) is the coefficient of relative permeability; P is the pore water pressure; γw is the unit weight of water; Ss is the specific storativity; C(p) is the specific water capacity of the soil; and α is a parameter for expressing the saturated or unsaturated condition of the soil (α = 0 means that the soil is unsaturated, and α = 1 means that the soil is saturated). The initial conditions for Eq. 3 were those described by Mao et al. (1999):

$$ P{\left(x,y\right)}_{t= 0}={P}_0\left(x,y\right) $$
(4)

The boundary conditions for Eq. 3 were:

$$ \Big\{{\displaystyle \begin{array}{l}{\left.P\left(x,y,t\right)\right|}_{\varGamma 1}={P}_0\left(x,y,t\right)\\ {}{\left.-\left[{k}_r(p){k}_{ij}\frac{\partial p}{\partial {x}_j}+{k}_{i3}{k}_r(p)\right]{n}_i\right|}_{\varGamma 2}={q}_n\\ {}-\left[{k}_r(p){k}_{ij}\frac{\partial p}{\partial {x}_j}+{k}_{i3}{k}_r(p)\right]{n}_i{\left|{}_{\varGamma 3}=0,P\right|}_{\varGamma 3}=0\\ {}-\left[{k}_r(p){k}_{ij}\frac{\partial p}{\partial {x}_j}+{k}_{i3}{k}_r(p)\right]{n}_i{\left|{}_{\varGamma 4}<0,P\right|}_{\varGamma 4}<0\end{array}} $$
(5)

where ni (i = 1, 2, 3) is the direction cosine; Γ1 is the pressure condition; Γ2 is the flux boundary condition; Γ3 is the saturated escape boundary condition; Γ4 is the unsaturated escape boundary condition; and qn is the normal flux of the boundary, with the direction to the outside being positive.

The maximum infiltration capacity of the soil during rainfall was calculated by the equation:

$$ R(t)={k}_r(P)\left({k}_{ij}\frac{\partial p}{\partial {x}_j}+{k}_{i3}\right){n}_i $$
(6)

where rain intensity q(t) of less than R(t) indicates that rainfall completely infiltrated the soil, namely R(t) = q(t). Rain intensity q(t) larger than R(t) indicates rainwater accumulation on the landslide surface. Rainwater does not pond easily on the surface of this landslide; therefore, ponding water was removed from the model. Runoff may erode soil on the surface of a landslide; however, this study was focused on surface water infiltration and slope stability, so the effect of runoff was not considered. The R(t) was calculated by Eq. 6, where the rainfall infiltration boundary condition was set to P = 0 (Wei et al. 2006).

Figure 14 presents the variations in water content and movement of the wetting front in the FEM validation model. The movement of the wet front in the FEM model was similar to that in the physical model, only slightly slower. However, the wet front close to both sides of the physical model moved down more slowly than it did in the FEM model, which is attributed to the boundary effect for the physical model. In the physical model, the boundary between the soil and model box was filled with glass glue to prevent rainfall infiltration along the boundary. The variation in water content in the FEM model was similar to that in the physical model, while the water content was slightly greater in the surficial soil than in the physical model. This indicates that FEM was a reliable method for investigating the effect of crushed stone soil on the infiltration process. Hence, on the basis of the numerical model that was previously validated, we were able to build a FEM model for the real landslide deposit aimed at reproducing the effect of crushed stone soil on rainfall infiltration.

Fig. 14
figure 14

Variations in water content and movement of wetting front in FEM: a 2 h 47 min; b 3 h 50 min; c 5 h 16 min; d 7 h 45

Full size image

FEM results for the real landslide deposit

In the FEM model for the real landslide deposit, the Γ2, Γ3, Γ4 (formula 5) were selected as the rainfall boundary condition, the physical and mechanical parameters of soil were assigned according to Table 2, and the rainfall intensity was set as 87.5 mm per day for 3 days. According to field investigation, the ground water table is below the old slip zone. We assumed the initial water table in the FEM model to be as shown in Fig. 2, and assigned the initial water content of soil as natural water content (Table 2). The FEM model is shown in Fig. 15.

Fig. 15
figure 15

FEM for rainfall infiltration in real landslide deposit

Full size image

Figure 16 presents the variations in water content and the movement of the wetting front within the landslide during the rainfall of October 2012. On the first day, the water content in the surficial clayey gravel soil varied only slightly from the initial value of 0.02, whereas it increased to 0.5–0.6 (i.e., near saturation) in the crushed stone soil. It increased to 0.05–0.25 in the clayey gravel soil 15–35 m below the crushed stone soil. One day later, the water content of the surficial clayey gravel soil (5–8 m depth) increased to 0.05, while it remained at 0.55–0.6 in the crushed stone soil. The water content in soil below the crushed stone soil, with a depth of 20–80 m (i.e., 20–45 m below the crushed stone soil), increased to 0.05. On the third day, when the rainfall ended, the water content in the surficial clayey gravel soil (depth = 7–10 m) increased to 0.05. It remained in the range of 0.55–0.6 in the crushed stone soil, and increased to 0.2–0.3 (i.e., nearly saturated) in the underlying clayey gravel soil. Over the next 3 days, the water content of the surficial clayey gravel soil showed no significant variation; it decreased to 0.05–0.15 in the crushed stone soil (depth = 5–25 m) and had high values of 0.5–0.6 only at the bottom. It increased to 0.2–0.3 in the underlying clayey gravel soil (i.e., below the crushed stone soil and above the slip zone), and the range of the area with high water content expanded along the slip zone. The water content in the slip zone increased to 0.35–0.45 on day 6. From day 6 to day 12, the water content in the surficial clayey gravel soil did not vary, and it decreased gradually in the crushed stone soil. It should be noted that in reality, surficial soil should decrease gradually after rainfall owing to evaporation; however, we did not consider evaporation during the FEM calculation. The two zones, with a high value of 0.25–0.3 of water content, continued to expand along the slip zone, and showed a trend toward integration.

Fig. 16
figure 16

Water content variations and movement of the wetting front in the landslide during the rainfall of October 2012

Full size image

As shown in Fig. 16, the wetting front in soil was calculated according to the variation in water content from the natural value (Table 2). During the rainfall event, the wetting front in the surficial clayey gravel soil moved down to a depth of 7–10 m, and moved no further until the rainfall ended. In the crushed stone soil, the wetting front arrived at the base after 6 h of rainfall. It moved through the crushed stone soil and into the underlying clayey gravel soil after 1 day of rainfall. On the first day, the wetting front arrived at a depth of 30–80 m under the two areas covered by the crushed stone soil. By the time rainfall ended, it had arrived at the slip zone underlying these two areas. Over the next 9 days, the wetting fronts underlying these two areas did not move across the slip zone, but expanded along the slip zone. On day 12, the two wetting fronts met and were integrated.

In terms of velocity, the wetting front moved down 7–10 m through the surficial clayey gravel soil during the 3 days of rain; it moved down 50–70 m at the locations that were covered by the crushed stone soil. It took only 6 h for the wetting front to move through the crushed stone soil. These results clearly show that the wetting front moved faster in the crushed stone soil, followed by the clayey gravel soil covered by the crushed stone, and more slowly in the surficial clayey gravel soil, The difference in permeability between the clayey gravel soil and crushed stone soil can be attributed to the difference between the two soils. However, the reasons for the much faster movement of the wetting front in clayey gravel soil that was covered by crushed stone soil than in the surficial clayey gravel soil need to be studied and discussed further.

Figure 17 presents the variations in pore water pressure in the landslide deposit during the rainfall. On the first day, high pore water pressure formed in the crushed stone soil, with maximum values of 100–200 kPa at the bottom. One day later, the maximum value of the pore water pressure increased to 200–300 kPa. By the third day, the pore water pressure in the crushed stone soil close to the river decreased as the water moved down into the underlying clayey gravel soil; the maximum value of pore water pressure at the bottom decreased to 200 kPa. When rainfall ended, the pore water pressure in the crushed stone soil decreased rapidly, reaching less than zero at most locations by day 6, except for the bottom of the crushed stone soil.

Fig. 17
figure 17

Pore water pressure variations and movement of the wetting front within the landslide during the rainfall of October 2012

Full size image

The variations in pore water pressure in the clayey gravel soil should be considered separately for two different conditions: clayey gravel soil not covered by crushed stone soil, and clayey gravel soil covered by crushed stone soil. For the former, the pore water pressure was less than zero and did not vary significantly during the rainfall. For the latter, two zones with pore water pressure of 0–200 kPa formed in the underlying clayey gravel soil after 1 day of rainfall. The two zones both had a thickness of approximately 15–25 m. Over the next 2 days, the ranges of the two zones widened rapidly, and the infiltrated water arrived at the slip zone by the time the rainfall ended. On day 6, the pore water pressure at the bottom of the two zones increased to over 300 kPa. Over the next 6 days, the pore water pressure in the two zones continued to increase, reaching a maximum value of over 400 kPa. The two zones continued to expand along the slip zone until they were integrated on day 12.

FEM results, including pore water pressure, volumetric water content, and matric suction, were saved and loaded into another stress–strain FEM model built for the analysis of landslide deposit stability. Therefore, the variations in landslide deposit stability over time were obtained using the strength reduction method, which was originally used by Matsui and San (1992) and Griffiths and Lane (1999). In the FEM model, the following equation proposed by Vanapalli et al. (1996) was implemented for calculating the shear strength of the unsaturated soil:

$$ \tau =\left(\sigma -{u}_a\right)\tan {\varphi}^{\prime }+\left({u}_a-{u}_w\right)\left[\left(\frac{\theta_w-{\theta}_r}{\theta_s-{\theta}_r}\right)\tan {\varphi}^{\prime}\right]+{c}^{\prime } $$
(7)

where c’ and φ’ are the effective cohesion and effective angle of the internal friction, respectively; σ is the total stress; θw is the volumetric water content; θs is the saturated volumetric water content; θr is the residual volumetric water content; and ua and uw are the pore air pressure and pore water pressure, respectively.

It was assumed that the soil in the FEM model was included in the elastic-perfect plastic behavior, and the Mohr-Coulomb shear strength criterion was adopted as the constitutive law. The physical and mechanical parameters of the soil were assigned according to Table 2. The FEM model was built using ANSYS software.

Figure 18 presents the characteristics of the plastic zone in the landslide deposit on the 12th day under the condition that the reduction coefficient of shear strength is 0.95. This indicates that the most critical failure surface for the landslide deposit was along the old slip zone during the intense rainfall.

Fig. 18
figure 18

Characteristics of plastic zone in landslide deposit (shear strength reduction coefficient of 0.95) on the 12th day

Full size image

Figure 19 shows that the safety factor decreased from an initial value 1.10 to 1.08 when the rainfall ended on day 3. It continued to decrease rapidly, reaching 0.98 over the next 5 days, and 0.95 on day 12. The decrease in the safety factor was correlated with the high pore water pressure formed in the slip zone.

Fig. 19
figure 19

Variation in the safety factor of the landslide during and after rainfall

Full size image

Discussion

As an old landslide, the landslide may have formed many years earlier. According to the plane shape, the front part of the landslide is convex to the left bank of the river, causing the river bed to swing to the left side. This indicates that the landslide would likely have blocked the Lancang River when it failed. The formation of the landslide may be related to the downcutting and erosion of the Lancang River. A similar case has been studied in Catalonia, Spain (Ferrari et al. 2011).

Prior October 2012, residents had lived and farmed on the deposit for many years, and eight houses were built here. According to their representation, there were not as many cracks or any other signs of failure on the surface of the landslide deposit before the rainstorm, except in the front part close to the river. Hence, the secondary failure in front of the landslide could be related to erosion of the river (Fig. 1). However, most of the new cracks that appeared over the landslide deposit may be attributed to the intense rainfall.

The landslide safety factor decreased to 0.98, close to a limit equilibrium state, 5 days after rainfall ended; this is consistent with in situ investigation results suggesting that the landslide was under a reactivation condition following the rainfall event. These results confirm that the major triggering factor for the reactivation of the landslide was the rainfall of 7–9 October 2012.

The safety factor decreased rapidly from day 3 to day 9; high pore water pressure (300–400 kPa) was formed in the old slip zone during that period. However, this high pore water pressure formed only in the two slip zone segments underlying the crushed stone soil; in other parts of the slip zone the pore water pressure was less than zero. These locations were also consistent with the arrival locations of the wetting front, which confirmed that the formation of high pore water pressure at the slip zone was caused by the infiltration of rainwater through the crushed stone. Interestingly, lower levels of infiltration occurred elsewhere, despite smaller deposit thicknesses.

Two different vertical sections of the landslide (Figs. 16 and 17) were compared, one where the clayey gravel soil was covered by crushed stone soil (section A), and the other where clayey gravel soil made up the total thickness (section B). During the rainfall, water infiltration was faster and deeper in section A than in section B. While high pore water pressure (maximum value of over 200 kPa) formed at the border between the crushed stone soil and the clayey gravel soil in section A, in section B the pore water pressure was less than zero. This indicates that more water was able to infiltrate the crushed stone soil due to the high hydraulic conductivity, and the lower permeability of the surrounding soil was conducive to the formation of high pore water pressure at the bottom of the crushed stone soil. Both factors caused faster and deeper water infiltration to occur in section A.

In terms of Figs. 12 and 14, when rainwater moved into the crushed stone soil, the velocity of the wetting front increased and it formed a kink at the border. This allowed more water to move into and accumulate in the crushed stone soil, which subsequently spread into the surrounding soil. This can be attributed to the accumulating water forming a higher hydraulic gradient.

Based on this study, the effects of locally accumulated crushed stone soil on rainfall infiltration can be summarized as follows:

  1. (1)

    Rainfall and runoff water from the landslide surface accumulated and formed a water-filled basin with high pore water pressure due to the high hydraulic conductivity of the crushed stone soil. The accumulated water offered a stable resource for infiltration of the surrounding soil.

  2. (2)

    A water-filled basin with high pore water pressure induced higher infiltration of the surrounding soil due to the higher gradient.

When the rainwater arrived at the old slip zone, it did not move down continuously, but accumulated at the slip zone (Figs. 16 and 17). This indicates that the old slip zone had an important effect on the seepage of infiltrated water, causing the infiltrated water to accumulate at the old slip zone and form high pore water pressure due to the lower permeability of the slip zone. The high pore water pressure was the main factor triggering reactivation of the old landslide deposit.

Our results suggest that precipitation and the distribution area and thickness of crushed stone soil were the key factors affecting water infiltration and ponding in the crushed stone soil; however, it is also possible that slope angle affects rainfall infiltration.

Conclusions

This study used an old landslide deposit in Yunnan Province, China, to investigate the impact of locally accumulated crushed stones on rainfall infiltration. In particular, we analyzed the infiltration process of rainwater and variations in the landslide deposit safety factor. The crushed stone soil played a key role in the rapid infiltration of rainfall deep into the landslide deposit. Rainfall and runoff water accumulated and formed high pore water pressure in the crushed stone soil for high hydraulic conductivity. High pore water pressure caused higher infiltration of the surrounding soil due to the higher gradient. This was the main reason for the much greater infiltration depth at the position covered by crushed stone soil than at that covered by clayey gravel soil. The original slip zone with low permeability was another key factor, causing the infiltrated water to accumulate at the slip zone and form high pore water pressure. The high pore water pressure that formed at the slip zone was the main factor triggering reactivation of the landslide along the old slip zone during the rainfall.