The Wobaoshi landslide is located in the Ba River basin in the Qinba–Longnan mountainous area. It is located in the village of Baiyanwan, town of Sanhui, Enyang District, in the city of Bazhong, Sichuan Province, China, and the specific location and elevation information are indicated in Fig. 1. The Wobaoshi landslide occurred just on the left bank of the Shilong River, the second-grade tributary of the Ba River, and the boundaries of the landslide are controlled by the local topography of the riverbank. The local geomorphology around the slide is characterized by low cuesta and a structural slope. The stratum consists of interbeds of sandstone and mudstone and belongs to the Upper Jurassic Penglaizhen Formation of the Jurassic series (Chen et al., 2015). The stratum is also called red beds in China (Hu and Zhao, 2006).

This area belongs to the subtropical monsoon region with abundant rainfall, and 75 % to 85 % of total annual rainfall is mostly concentrated between May and October. The monthly average rainfall in 1 year is greater than 100 mm. The maximum monthly rainfall, often occurring in July, is more than 200 mm, and severe rainstorms often occur in the same month. The precipitation in this region gradually decreases after August. The surface water in this area includes fissure water in weathered bedrock and accumulated water in the cracks.

According to the remote-sensing data by the GF-2 satellite and the field surveys, the landslide looks long, flat, and rectangular in shape. The landslide body is nearly 32 m long in the longitudinal (sliding) direction, 160 m wide in the lateral direction, and approximately 30 m thick in the vertical direction, and the total volume is approximately 1.536×105 m3 (Chen et al., 2015). This main body belongs to small- to medium-sized landslides according to the classification proposed the Ministry of Land and Resources of the PRC (2006). Figure 2 shows the schematic map of the Wobaoshi landslide and photographs of five observation points. The landslide lies in the south of the Nanyangchang anticline of the geotectonic outline map of the Daba Mountains (Dong et al., 2006). The landside occurred on a subhorizontal inclining rocky slope. The sliding direction of the landslide is 249∘, and the degree of inclination of the bedrock is 6–8∘. Figure 3 demonstrates the I–I′ cross section of the landslide.

As shown in Fig. 2a, there is 2–3 m thick mixture layer of soil and colluvial deposits, covering the bedrock mass. The major bedrock mass consists of integral and thick sandstone, and the potential bottom sliding surface is in the weak interlayer of silty mudstone. As shown in Fig. 2b and c, the Wobaoshi landslide poses a major threat to residential houses and highways, cracking the houses and uplifting the highways on its front edge; therefore, this landslide considerably threatens the safety of local people's property and transportation. According to Fig. 2d, bent trees grow on the crown of the landslide bodies I and II. The existence of bent trees implies that the geological bodies on the potential sliding surface become unstable, which is also historical evidence of the slow sliding movement of the Wobaoshi landslide.

Totally different from the common geometry of landslides, as shown in Fig. 2, the ratio of the longitudinal length to the lateral width is much smaller than those of common landslides; therefore, this type of geological-hazard mass is often categorized as a bedrock collapse by mistake during the routine field surveys. As indicated in Fig. 3, the two major sliding bodies are almost vertical and look like two parallel walls, a shape which is created by two sets of long and straight structural planes cutting through sub-horizontal sedimentary rock mass perpendicularly into two narrow plates (bodies I and II), and the potential sliding surface is sub-horizontal, parallel with the sedimentary bedding plane. For body I of the landslide, it is 12 m long in the longitudinal direction, 70 m wide in the lateral direction, and 30 m high; for body II of the landslide, it is 16 m long in the longitudinal direction, 65 m wide in the lateral direction, and 28 m high. Cracks I and II, formed by bodies I and II and head rock mass, are filled with clay, gravel, and collapsed debris. When high-intensity precipitation occurs during the monsoon, accumulated water can often be observed in the two cracks, indicating that cracks I and II exhibit favorable water storage conditions.

According to the detailed field surveys and preliminary analysis of Wobaoshi landslide, this landside should be categorized as rainfall-induced translational landslide according to the landslide geometry, lithology conditions, slope structures, and water accumulation situation in cracks (Xu et al., 2010). Based on the previous landslide monitoring cases (Ayalew et al., 2005; Fan et al., 2009), the rainfall, width of cracks I and II, and level of accumulated water in cracks I and II were chosen as key monitoring indicators for the Wobaoshi landslide. The layouts of all the field monitoring instruments are demonstrated in Fig. 4.

As shown in Fig. 4a and c, two non-contact automatic crack meters, LF I and II, are installed on the crown surface of bodies I and II to record the real-time widths of cracks I and II (Liu and Wang, 2015). As shown in Fig. 4b, an automatic rain gauge is installed on the crown of the Wobaoshi landslide to measure monthly and cumulative rainfall values, and two water pressure gauges are installed at the bottom of cracks I and II to measure the water level of accumulated water in cracks I and II. The measurement frequency for the crack width is three times per day, the measurement frequency for the accumulated water level is twice per day, and the monthly accumulative value of precipitation is adopted to indicate the local rainfall amount. All the monitoring data were transmitted to a network monitoring server through the public GPRS network.

The field monitoring work was started from February 2015 and ended as of July 2018. The monitoring work lasted for about 3.5 years; all the monitoring data are consecutive during the field monitoring and qualified for community warning and scientific analysis with reference to the geological data standards issued by the China Association of Geological Hazard Prevention (CAGHP; CAGHP, 2018).

As shown in Fig. 5, for the data processing of water level in cracks, the actual water level, hc, can be calculated using hc=H-hi+hm, where hi is the installation depth of the water pressure gauge, H is the actual depth of the crack, and hm is the measured the water level. For crack I, with the installation depth hi1=24.72 m, the depth of crack I is H1=38 m; thus hc1=13.28 m + hm1. For crack II, with the installation depth hi2=24.85 m, the depth of crack II is H2=35 m; thus hc2=10.15 m + hm2. The initial width of crack I is 5.640 m, and the initial width of crack II is 4.492 m; the first measurement was commenced in January 2015 (Chen et al., 2015).

All the monitoring data and processed results were presented in Tables 1–3 and Fig. 6 and were plotted based on Tables 1 and 2. The plots of Fig. 6a and b denote the comparison plots of the opening widths of cracks I and II, water pressures in crack I and II, and the monthly rainfall with respect to the monitoring time.

According to the above data comparison and analysis, cracks I and II have favorable water storage capability during the monsoon season. According to Fig. 6a, the opening width of crack I increases with the rise of water table positively during the monsoon season, and the crack width waves slightly while the water level is almost static; the same phenomenon happens to the crack II. Therefore, the variation in crack widths is controlled by changes of water levels. As shown by the creep lines indicated both in Fig. 6a and b, the minimum widths of cracks I and II tend to increase year by year, and these values are considerably affected by the amount of rainfall, indicating that the upper part of body I and body II tends to slide outward gradually.

As shown in Fig. 7, plotted with data in Table 3, the absolute width variations in cracks I and II are both approximately 1 m from July to August 2017 (in which the monthly rainfall amount is greater than 250 mm). During the dry season, the crack width narrows with the decrease in the monthly rainfall, and the minimum opening widths of cracks I and II appeared in the months of January during the monitoring period.

Regarding the characteristics of the Wobaoshi landslide, the surface soil layer can be ignored during the establishment of the geomechanical model, and a typical section of plate-shaped bodies I and II of the Wobaoshi landslide was selected, as shown in Fig. 8. A static geomechanical model of the plate-shaped rock bodies is established by using the limit equilibrium method. The basic assumptions of the limit equilibrium method are the plastic behavior for soil mass and validity of the Mohr–Coulomb failure criterion (Vardoulakis, 1983), and a kinematically feasible sliding surface is assumed to define the mechanism of failure. Besides, the ideal elastic–plastic model in the stress–strain state is selected for stability analysis based on associated flow rules (Darve and Vardoulakis, 2004; Labuz and Zang, 2012).

As indicated in Fig. 8, α denotes the dip angle of the sliding surface, hc1and hc2 are the heights of the water levels in cracks I and II, L1 and L2 are the widths of bodies I and II, Lc2 is the distance between bodies I and II, H1 and H2 are the heights of bodies I and II, respectively, and W1 and W2 are the weights of bodies I and II per unit. The stability analysis was commenced from the outer body II; subsequently, that of the inner body I is analyzed.

According to the relation between K, the stability coefficient of the main body, and hc, the height of the water level, the stability coefficient of body II, K2, can be obtained as follows when considering the internal cohesive strength of the sliding surface: 1K2=W2cos⁡α-12γwhc2L2-12γwhc22sin⁡αtan⁡θ+cL212γwhc22cos⁡α+W2sin⁡α. Here, c is the internal cohesion of the sliding surface, γr is the unit weight of the saturated sandstone, γw is the unit weight of water, and W=H⋅L⋅γr. In order to obtain the critical failure height of water level, K2 is set to 1; i.e., body II is set in a critical sliding state. Equation (2) is derived from Eq. (1) and can be used to calculate the critical water level of body II hcr2 when K2 is set to 1: 2hcr2≈12cos⁡αL22tan⁡2θ+8γwW2cos⁡αtan⁡θ-W2sin⁡α+cL2cos⁡α12-L22cos⁡αtan⁡θ. According to the experimental data obtained from the triaxial test of rock cores extracted from the sand–mudstone contact surface of the Wobaoshi landslide, θ, the internal friction angle of the sliding surface, is 11.2∘; c, the internal cohesion of the sliding surface, is 10.2 kPa; and γr, the unit weight of saturated sandstone, is 19.2 kN m-3. According to the cross section of the Wobaoshi landslide (Fig. 2), H=35 m, L=16 m, and α=6∘. All the values are substituted into Eq. (2), so hcr2=13.896 m.

Based on the stability analysis of body II, using Eqs. (1) and (2), the stability coefficient K1 of the inner layer of body I can be obtained using Eq. (3). In addition, hc2′=hc2-Lc2sin⁡α and Lc2=3.8 m; therefore, hc2′=13.499 m: 3K1=W1cos⁡α-12γwhc1+hc2′L1-12γwhc12-hc2′2sin⁡αtan⁡θ+cL112γwhc12-hc22′cos⁡α+W1sin⁡α. Similarly, K1 is set to 1; for sliding body I, H1=38 m, L1=12 m, α=6∘, and hc2′=13.499 m; therefore, the critical water level hcr1 of body I can be calculated using Eq. (3) and hcr1=17.249 m.

The above calculation results indicate that the water pressure in cracks I and II will drive the two plate-shaped bodies to creep slightly when the accumulated water level reaches the critical height, i.e., when hcr1=17.249 m and hcr2=13.896 m.

The water-level-monitoring data from cracks I and II can be used to verify the critical height, calculated by Eq. (2). In order to achieve this goal, two parameters, the opening widths of cracks and actual water level, are adopted to analyze the slipping process of bodies I and II. The two parameters can be calculated with the monitoring data in Table 1, the processing methods are described in Table 3, and all the processed data are listed in Table 3. The relationship between the sudden opening-width increase and the rise of the actual water level in cracks I and II is demonstrated in Fig. 9, plotted with data in Table 3.

The dotted boxes in Fig. 9 denote the fact that when the accumulated water level approaches the critical water level, hcr′, the water pressure in cracks I and II, can make cracks open much wider and cause the main bodies to creep. The measured hcr′ in Fig. 9 can be utilized to verify the relation between the actual water level, hc, and the stability coefficients of the bodies, K1 and K2, obtained using Eqs. (1) and (3), respectively, which are also depicted in Fig. 10.

In Fig. 10, the curves of the hc1-k1 and hc2-k2 represent Eqs. (1) and (3), respectively. The values of hcr′ (measured) in Fig. 10 denote that most measured actual water levels are not higher than the theoretically calculated values. The monitoring data from the Wobaoshi landslide show that when hcr′, the measured value, almost approaches hcr, the theoretical value, water pressure in cracks I and II can cause the main bodies to creep and incline outward and result in wider upper opening of cracks I and II.

Numerical simulation and calculations were performed with respect to the main bodies using the MIDAS GTS NX geotechnical finite-element software. First, the 1:1 main body model presented in Fig. 8 was introduced into the aforementioned software, and the mechanical parameters of the main body model, i.e., the elastic modulus, Poisson's ratio, gravity internal cohesion, and the friction angle, were defined as shown in Table 4. The left and right boundaries were located at a distance of approximately 30 m from bodies I and II, respectively, and the lower boundary was located at sea level to eliminate the boundary effect. A plane strain quadrilateral–triangle mixing element was considered, and the entire model is divided into 13 775 elements and 14 026 nodes. Here, we constrained the vertical and horizontal displacement of its bottom boundary, and the left and right boundary conditions were established to constrain the horizontal displacement. The model used steady-state seepage calculation, and the water levels at the left and right boundaries were 342 and 275 m, respectively. The boundary conditions were set as follows.

In case of the displacement boundary, the left and right boundaries constrained the displacement in the X direction; i.e., TX=0. In case of the bottom boundary, the displacements in the X and Y directions in Fig. 11 were constrained; i.e., TX=TY=0.

The initial displacement state in Fig. 11a is set to zero for performing the following analysis. Figure 11b shows that bodies I and II deform horizontally along the sliding surface under the combined effect of water pressure and seepage. In Fig. 11c, the bodies slide to the maximum distance, where the maximum distance of body II is 0.945 m, which is approximately similar to the value obtained in the monitoring data. In Fig. 11d and e, bodies I and II exhibit the same tendency of tilting and a stop in creeping, owing to the decrease in the water level during the dry season. In Fig. 11e the maximum horizontal displacement is ∼-0.14 m, which implies that the maximum tilting value of body I is consistent with the measured opening widths of crack I in Table 3. Therefore, the calculation results obtained via the numerical simulation can corroborate the above-mentioned geomechanical model and landslide monitoring data.

The monitoring results of the Wobaoshi landslide can be used to validate the rainfall-induced failure mode of the translational landslide (Zhang et al., 1994). The deformation and failure mode for the Wobaoshi landslide were obtained through field monitoring data, geomechanical model analysis, and numerical simulation. Based on the above-mentioned analysis, a schematic drawing of the deformation and failure mode for the Wobaoshi landslide was created, as shown in Fig. 12. In Fig. 12b, the large amount of rainfall during the monsoon season causes cracks I and II to become accumulated with water; when the accumulated water level reaches the critical height, the landslide begins to creep, and cracks I and II open the most. The increased water pressure positively affects the creep initiation of the outermost body (Fan, 2007). Regarding the monitoring data, the accumulated water pressure can drive cracks I and II to open up to by about 1 m, and the consequent gradual creep results in the uplift of residential houses and highways on its leading edge.

For the arrival of rainy season, the plate-shaped body II begins to slide firstly (Fig. 12b) and the water pressure balance in cracks is destabilized; such a situation causes the sliding of body I (Fig. 12c). The failure mode of the Wobaoshi landslide is characterized by the gradual sequential creep from the outer part to the inner part.

As shown in Fig. 12d, the bodies are tilted toward the crown of landslide because of the lower water level and their own weights when there is less rainfall during the dry season, causing the body to fall backward (contrary to the slope inclination). The monitoring data of the Wobaoshi landslide and numerical simulation of the plate-shaped body can be used to verify the deformation and failure mode of the plate-shaped landslide after its occurrence (Xu et al., 2010). As years pass, the cracks at the bottom of the plate-shaped body will increase in size, and the inclination of the body will become severe, which will pose a high risk to the houses and roads located toward the front edge of the landslide.

The stability calculation of the geomechanical model of the body is described in Sect. 3.1, i.e., determination of the critical water height in the crack, hcr, and calculation of the body's stability coefficient, K, which can be determined theoretically by calculating the stratum inclination, shape, weight, and physical properties (unit weight of the saturated volume, γr, internal cohesion of the sliding surface, c, and internal friction angle of the sliding surface, θ) based on the limit equilibrium theory (Lin et al., 2018). Therefore, the stability coefficient of the landslide is observed to exponentially decrease with an increase in the filled water height of the crown crack (Fan et al., 2008; Xu et al., 2010).

The internal friction angle, θ=11.2∘, is considerably low for clay and seems unrealistic. This may be because the clay layer is severely weathered, resulting in a considerably small internal friction angle. Generally, the dilatancy effect obtained via the associated flow law is considerably larger than the actual observation, especially in the case of lateral confinement (Tschuchnigg et al., 2015a). However, in the case of slope stability analysis, lateral infinity is mostly not considered, and the dilatancy effect is not significant (Griffiths and Fenton, 2004). Therefore, it is reasonable to set the dilatancy angle to be equal to the internal friction angle.

With respect to the critical water level, hcr, in Eq. (2), we can observe that the measured critical water level, hcr′, is close to the theoretical critical water level, hcr, validating the calculation equation of hcr in Eq. (2) by comparing with the measured data. Additionally, the measured data in Table 3 are slightly less than the theoretical calculation value. Thus, when compared with the equation to calculate the critical water height proposed by Zhang et al. (1994) and the physical simulation experiment conducted by Fan et al. (2008), the monitoring case of the Wobaoshi landslide shows that for hcr′, the measured data are mostly lower than the theoretical calculated value, hcr, which can destabilize the main body. This instability may be attributed to the fact that the actual cohesion value c′ of the contact surface of sandstone and mudstone is smaller than the cohesive force value c of the sliding surface in Eq. (2) during the creep state of the landslide for a long duration or that the frictional angle of the sliding surface, θ, changes slightly. According to Eq. (2), if c′≤c, hcr′≤hcr means that when hcr′, the measured value, almost approaches hcr, the theoretical value, this condition will cause the main bodies to be unstable and result in wider upper opening of the cracks.

In this study, we propose a long-term monitoring method containing more parameters based on the characteristics of the plate-shaped translational landslides in accordance with the existing field monitoring experience as well as deformation and failure mode exploration.

First, long-term monitoring should be conducted to obtain sufficient monitoring data, mainly including obtaining the accumulated water level in cracks, amount of rainfall, and displacement data on the front edge of the landslide during monsoon as well as focusing on the change of the overall inclination of the body during the dry season. The inclination angle α relative to the sliding surface also changes while the body slides. Thus, an inclination measuring device, a three-axis accelerometer and electronic compass, should be installed on the main body to verify the theoretical model of the deformation mode of the plate-shaped body during the dry season, as indicated in Fig. 12c. Furthermore, a sensitivity analysis of the various parameters, affecting the stability coefficient K of the main body (including the accumulated water level in cracks, internal cohesive force in saturated water, internal friction angle of the sliding surface, and inclination angle of the body), should be conducted based on the monitoring data. Therefore, a detailed analysis and investigation of the deformation and failure mode of the plate-shaped landslide would be beneficial and improve the success rate of landslide warning.