Introduction

Landslide has been assessed to be the most common geohazard associated with the urbanization process of rapidly expanding cities in Northwest China despite the use of several ground improvement methods (Shen et al. 2017; Wang et al. 2018a, b; Cheng et al. 2018b, c, d). The Loess Plateau, which covers almost all of Shaanxi province, Northwest China, is susceptible to landslide occurrences (Derbyshire et al. 1994; Derbyshire 2001; Zhou et al. 2002; Xu et al. 2007; Zhang et al. 2009; Zhang and Liu 2010; Wang et al. 2014; Peng et al. 2015; Zhuang et al. 2017; Zhang et al. 2017; Liu et al. 2018; Qiu et al. 2017, 2018). The well-developed joints and/or fractures in the loess provide additional seepage paths for either rainwater or irrigation water to infiltrate into the slope (Du et al. 2014a, b). Stratum erosion occurs when the seepage force is greater than the shear strength of the loess. This stratum erosion, particularly at the toe of the slope, not only leads to a downwards movement of the slope, but also initiates local sliding surfaces that may progressively develop into a global failure phenomenon. These local sliding surfaces could develop further and evolve into a global sliding surface when subjected to a stratum erosion of greater severity, thereby increasing the potential of loess landslide (Zhang et al. 2013, 2014). The raising of the phreatic surface plays an additional and crucial role in triggering loess landslides. Raising of the phreatic surface is subjected to many influencing factors, such as irrigation-water infiltration, domestic groundwater withdrawal, extreme climate, among others.

The south Jingyang platform, located in Shaanxi province, is considered to be one of the world’s most representative landslide-prone areas, with 62 severe loess landslides recorded since agricultural irrigation began in 1976. These loess landslides resulted in significant economic losses and human casualties (Lai et al. 2016, 2017; Cheng et al. 2017a, b, 2018a). Irrigation activities are necessary in Northwest China to enable the conversion of the natural loess into farmland in the very dry climate characteristic of the region. However, irrigation has a significant influence on phreatic surface variation (Shen and Xu 2011; Xu et al. 2012, 2014, 2016; Shen et al. 2013; Wu et al. 2015, 2016, 2017; Zhao et al. 2016; Zhang and Wang 2018). As phreatic surface variation is also subjected to many other influencing factors, the typical pattern of water pressure in this area is rather complex. Many earlier studies reported that the pattern of water pressure can be directly linked to the loess landslide incidents, but these studies did not take other influencing factors into account. Additionally, because of a lack of comparisons between field measurements and laboratory test results with complex real phenomena, a shared interpretation of the triggering mechanism of the loess landslides in the south Jingyang platform is still not available (Fell et al. 2007).

The objectives of this study are: (1) to introduce the types and characteristics of loess landslides in the south Jingyang platform, specifically for landslide LD37, (2) to analyze the field measurements of LD37 (Peng et al. 2017, 2018) in the light of results from constant shear drained (CSD) triaxial test results, and (3) to reveal the triggering mechanism of loess landslide incidents by exploring in depth the data from the comparisons of the CSD triaxial tests results and the field measurements with real phenomena.

Loess landslides in south Jingyang platform

Materials

The geological profile for the southern Loess Plateau and plan view of the study area are shown in Fig. 1. The geology primarily consists of loess deposit, underlain by silty clay deposit with occasional interbedded sand layers. Figure 2 shows the variations in water content, dry density, and void ratio against depth according to the samples retrieved from three geological boreholes (JL1–3) in the vicinity of Jiangliu. This figure also shows that the water contents initially varied from 10 to 22% and then slightly increased with increasing depth due to the influence of desiccation of the upper crust. It can also be noted from Fig. 2 that the dry densities and void ratios varied within ranges of 1.55–1.7 g/cm3 and 0.57–0.75, respectively. As intensive irrigation is a common practice in the study area, due to the demands of rural residents for water, the shearing resistance of the loess may be degraded sharply when subjected to irrigation-water infiltration (Melinda et al. 2004; Lu and Likos 2006; Ng et al. 2016). This degradation of strength promotes the development of sliding surfaces. In addition, as the well-developed joint systems in the loess provide additional seepage channels, the fine-grained soils of loess can even be washed away when subjected to large hydraulic gradients, thereby leading to stratum erosion. Stratum erosion, particularly at the toe of slope, can significantly impact slope stability.

Fig. 1
figure 1

Geological profile and plan view of the study area

Full size image
Fig. 2
figure 2

Variations in water content, dry density, and void ratio against depth. JL1–3 Geological boreholes in the vicinity of the Jiangliu platform

Full size image

Loess landslide types

The loess landslides that have occurred in the study area can be categorized into two types on the basis of the landslide characteristics and topographic investigations, namely, loess flowslide and loess slide (Fig. 3). Loess flowslides behave in a quasi-liquid state that is distinct from similar landslide incidents worldwide; such flowslides can taxi a very long distance at a high velocity (Cascini et al. 2010, 2011; Xu et al. 2013; Oldrich et al. 2014; Leng et al. 2017). The originally deposited strata at shallow depths are likely to be buried below and/or pushed forwards by the landslide deposit, thereby forming a series of thrust structures, as depicted in Fig. 4a, b (Peng 2017, 2018). Liquefaction phenomena due to an inability to dissipate the excess porewater pressure in a very short period of sliding can easily be seen at the interface of the landslide deposit and originally deposited strata, as shown in Fig. 4c, d. Compared with loess flowslides, loess slides behave in a quasi-plastic state and taxi a shorter distance.

Fig. 3
figure 3

Loess landslide types: a loess slide and b loess flowslide

Full size image
Fig. 4
figure 4

Photographs of loess strata. a Large dip-angle silty sand in trench 1 (shown in Fig. 5), b thrust structure in trench 2 (shown in Fig. 5), c sand boiling in trench 2 (shown in Fig. 5), d liquefaction phenomenon

Full size image

Loess landslide LD37

A list of all the available records on loess landslides in the south Jingyang platform is provided in Table 1. Due to the completeness of the data available, we selected only loess landslide no. LD37 (Fig. 5) for an in-depth analysis. The landslide characteristic parameters shown in Table 1 are schematically illustrated in Fig. 6. LD37 is characterized by four landslide phases, with the first and final phases classified as the loess flowslide and the second and third phases classified as the loess slide. The Hmax/Lmax ratio of LD37 varied from 0.20 to 0.44 for the loess flowslides and from 0.73 to 1.31 for the loess slides where Hmax corresponds to the distance from platform surface to the bottom of landslide deposit and Lmax represents the total sliding distance of landslide deposit (equal to Ls +Ld , as denoted in Fig. 6). The Ld value varied from 42.8 to 54.7 m for the loess slides and from 130 to 284 m for the loess flowslides, in which Ld indicates the sliding length of landslide deposit on the surface. The triggering mechanism will be discussed in subsequent sections.

Table 1 Summary of the loess landslides which have occurred in the south Jingyang platform
Full size table
Fig. 5
figure 5

The four landslide phases of LD37 located in the south Jingyang platform

Full size image
Fig. 6
figure 6

Schematic illustration of landslide characteristic parameters. Dd Maximum depth of landslide deposit (m), Wd Average width of landslide deposit (m), Ld Sliding distance of landslide deposit on the surface (m), Ls Projected sliding distance of landslide deposit on the sloping ground (m), Hmax Distance from platform surface to the bottom of landslide deposit (m), Hs Distance from platform surface to the top of landslide deposit (m), Lmax Total sliding distance of landslide deposit (m); ϕα Value returned by the Arctangent relation between Hmax and Lmax (deg.); Vdeposit Volume of landslide deposit (m3)

Full size image

Field measurements and laboratory tests

Methodology

To determine the triggering mechanism of a large number of the loess landslides in the study area, we performed field measurements from displacement markers and crackmeters as well as from observation wells despite the time offset between the field measurements and the occurrence of the four landslide phases of LD37. This time offset due to the difficulty in forecasting loess landslides. Despite this time difference, the field measurements of LD37 were analyzed to characterize the evolution of displacements, tension cracks, and phreatic surface and then compared to the results of the triaxial tests that shears a specimen in a drained condition along the constant deviatoric stress path by increasing the back pressure at a constant rate. The results of the triaxial tests on the specimens are listed in Table 2. Insight gained from the comparisons provide crucial evidence revealing the triggering mechanism of the loess landslides in the study area.

Table 2 Basic physical properties of loess
Full size table

Displacement measurements

As mentioned, the fine-grained soils of loess can easily be washed away when subjected to significant hydraulic gradients, thereby leading to stratum erosion. Stratum erosion, particularly at the toe of slope, has the potential to lead to slope instability. Some pre-landslide signs, such as displacement and tension crack at the top of the slope due to changes in the hydraulic gradient, can thus be captured prior to the occurrence of loess landslide. Figure 7 shows the extent of four landslide phases and the location of the monitoring instruments for LD37, and Fig. 8 shows the evolution of displacements from locations MD01-03 and MD08-09 for LD37 and the change in phreatic surfaces from ZT1-2 for LD31; however, the data at ZT1-2 were not analyzed due to incompleteness. As can be seen, there was little change at MD01-03 and MD08 during the whole period from June 2016 to June 2018. However, at location MD09, which is connected to the tension cracks, there was initially only a small change, but this subsequently increased dramatically, at which time the four stages can be clearly distinguished, namely, the preliminary stage (Stage 1), the gradual development stage (Stage 2), the rapid development stage (Stage 3), and the final stage (Stage 4). The average horizontal displacement rate increased from 0.2 mm/day in Stage 1 to 0.3 mm/day in Stage 2, while the average vertical displacement rate increased from 0.4 mm/day in Stage 1 to 0.6 mm/day in Stage 2. The average horizontal displacement rate then increased by 0.2 mm/day to 0.5 mm/day in Stage 3, while the average vertical displacement rate increased 0.6 mm/day to 1.2 mm/day in Stage 3. These large changes in average horizontal and vertical displacement rates did not develop any further, however, and fell to 0.005 and 0.07 mm/day, respectively, in Stage 4, indicating that the slope had reached a quasi-steady state.

Fig. 7
figure 7

Extent of four landslide phases and location of monitoring instruments for LD37

Full size image
Fig. 8
figure 8

Evolution of displacements from MD01-03 and MD08-09 for LD37 and change in phreatic surfaces from ZT1-2 for LD31. See Fig. 7 for locations

Full size image

The evolution of tension crack expansions or contractions from MDF06-07 and MDF12-13 for LD37 is shown in Fig. 9. At location MDF13, the tension crack first expanded at a rate of 0.25 mm/day and then contracted at a rate of 0.01 mm/day.

Fig. 9
figure 9

Evolution of tension cracks from MDF06-07 and MDF12-13 for LD37

Full size image

CSD triaxial tests

The CSD triaxial test was employed to investigate the behavior of loess in terms of the effect of the rising phreatic surface. Cylindrical specimens of 39.1 mm in diameter and 80 mm in height were retrieved from the geological borehole JL4 depicted in Fig. 1, and CSD tests were performed on loess specimens collected at borehole depths of 6, 12, and 24 m, respectively. CO2 was introduced in the saturation phase to displace air from the specimens, thereby allowing for back pressure saturation to be performed. All specimens were saturated to a value of B ≥ 0.95 (Skempton 1954). The consolidation pressures were obtained by increasing cell pressure and axial load, and K0 = 0.5, earth pressure coefficient at rest corresponding to the ratio of the horizontal effective stress, σ’h, to the vertical effective stress, σ’v, was adopted. The specimens were then sheared in a drained condition along the constant deviatoric stress path by increasing the back pressure u at a rate of 1 kPa/h. The results of the CSD tests are given in Fig. 10 in terms of axial strain (εa) versus mean stress (p’), volumetric strain (εv) versus mean stress (p’), and deviatoric stress (q’) versus mean stress (p’), in which p’ = (σ1′ + 2σ3′)/3 and q’ = σ1′ − σ3′.

Fig. 10
figure 10

Results of constant shear drained (CSD) triaxial tests: aεa (axial strain) vs. p’(mean stress), bεv (volumetric strain) vs. p’(mean stress), cq’ (deviatoric stress) vs. p’(mean stress), dεa (axial strain) vs. u (porewater pressure)/σ3

Full size image

Analysis and discussion

The horizontal and vertical displacement rates shown in Fig. 8 were measured in Stage 2 at location MD09 and averaged up to 0.3 and 0.6 mm/day, respectively. Despite the lack of porewater pressure records, it can be inferred that the seepage induced by irrigation-water infiltration from the top of the slope through joints and/or fractures towards the toe led to washing away of the fine-grained soils (see the reeds growing up at the toe of LD37 in Fig. 11) and the subsequent downwards movement of the slope in a gentle manner, initiating a series of local sliding surfaces. In Stage 3, the average horizontal and vertical displacement rates were measured at values of 0.5 and 1.2 mm/day, respectively; these were nearly twofold higher than those in the previous stage and indicated a further development of the local sliding surfaces associated with a more severe stratum erosion of the fine-grained soils. However, thereafter, in Stage 4, the average horizontal and vertical displacement rates showed a large decline 0.005 and 0.07 mm/day, respectively. From these data, it can be postulated that the effect of particle rearrangement and/or particle inter-locking at the local sliding surfaces effectively impeded subsequent movement of the slope. In some cases, depending upon the severity of stratum erosion, the movement of the slope may develop further, ultimately triggering a landslide incident.

Fig. 11
figure 11

Photos showing reeds growing at the toe of LD37. a Whole view of LD37, b enlargement of the toe area

Full size image

At location MDF13, the tension crack dramatically expanded in Stage 1, at a rate of 0.25 mm/day, then contracted to a rate of −0.01 mm/day in Stage 2, as shown in Fig. 9. Stages 1 and 2 at MDF13 showed a pattern similar to that at MD09 in Stages 3 and 4. This consistency validates the field measurements.

It can be seen from Fig. 10 that the axial strain εa and volumetric strain εv changed only slightly during a steady reduction in the effective mean stress p’, while the deviatoric stress remained constant. This phenomenon is due to the back pressure u not being sufficiently high to make the soil skeleton unstable, yet high enough to be capable of initiating local failure surfaces. The shearing resistance available along the local failure surfaces prevented the axial strain εa or volumetric strain εv from growing. As the back pressure u increased to a level high enough to cause the soil skeleton to become unstable, the axial strain εa and volumetric strain εv also substantially increased, indicating that the loess specimen sheared along the local failure surfaces where the soil particles had been rearranged. Since the soil skeleton remains stable when subjected to small back pressures or the effect of particle rearrangement and inter-locking and became unstable when subjected to high back pressures, u outweighed the effect of particle rearrangement. The developing slow strain–rapid strain cycle can be repeatedly seen in the ε–p′ plot. At CSD3, this developing slow strain–rapid strain cycle was repeated three times, while at CSD1-2, it was seen only once. As the back pressure u increased sufficiently to cause local failure surfaces to be elongated and to become connected with each other to form a global failure surface, the specimen failed as the shearing resistance along the global failure surface ran; at CSD1-2, the ratio of the back pressure u to the cell pressure σ3′ reached 0.7, while at CSD3, it approached 0.8. It is evident that the effect of particle rearrangement and inter-locking enabled the soil skeleton to sustain larger back pressures and/or impeded the development of global failure surface.

The triggering mechanism of the loess landslides in the study area can be summarized as follows. Since the phreatic surface was raised by 13–37 m between 1976 and 1992 and no loess landslides were reported before agricultural irrigation, the loess landslides were most likely triggered by induced seepage of irrigation-water infiltration from the top of the slope through joints towards the toe. Consequently, subjected to a seepage force that was greater than the shear strength of loess, the fine-grained soils were washed away, leading to stratum erosion. This stratum erosion, particularly at the toe of slope, greatly impacted slope stability and initiated a series of local sliding surfaces, with the negligible displacement rates shown in Stages 1 and 2 at MD09 (Fig. 8); this development is different from the CSD triaxial tests undertaken where the local failures were initiated by introducing the back pressure u to the specimen. Although the CSD triaxial test is not capable of taking the effect of stratum erosion into account, its results do reflect the initiation of the local failures by the axial strains εa (or volumetric strains εv) developing at a slow pace, as shown in Stages I, III, and V of CSD3 and Stages I, III, and V of CSD3, and can thus be compared to those at Stages 1 and 2 at MD09.

When subjected to an even greater seepage force, the stratum erosion became more severe and led to instability of the soil skeletons and further development of local sliding surfaces. The slope moved downwards at horizontal and vertical displacement rates that were nearly twofold higher than those in the previous stage, as shown in Stage 3 of MD09. By introducing higher back pressures to the specimen, the CSD triaxial test linked a further development of the local failures to the axial strains εa (or volumetric strains εv) at a rapid pace, as shown in Stages II, IV, and VI of CSD3; therefore, Stages II, IV, and VI of CSD3 can correspond to Stage 3 of MD09.

The developing slow strain–rapid strain cycles in the ε–p′ plot of the CSD triaxial test also indicated that the loess landslides in the study area were a feature of the progressive failure. It is recognized that the severity of stratum erosion determines whether local sliding surfaces can develop into a global sliding surface (Xu et al. 2013; Zhang and Wang 2018). Moreover, although the loess landslides were not triggered by the raising of the phreatic surface, but rather by the seepage induced by irrigation water, the former greatly weakened the available shearing resistance along the local sliding surfaces, thereby increasing the potential of loess landslide. Preventive measures to reduce water infiltration into loess slope should be taken along with the effective management of agricultural irrigation to prevent future similar incidents from reoccurring.

Conclusions

This paper presented the results of an investigation into the loess landslide LD37 for a better understanding of the triggering mechanism of the loess landslides in the study area. Based upon the results and discussions, the following conclusions can be drawn:

  1. (1)

    The field measurements on LD37 and the CSD triaxial test results were analyzed. Insights gained from the comparisons of the LD37 field measurements and the CSD triaxial test results with real phenomena indicated that irrigation-water infiltration induced seepage from the top of slope through joints and/or fractures towards the toe and that this seepage can be considered to be the major contributor to the triggering of the loess landslides in the study area.

  2. (2)

    The seepage washed the fine-grained soils away when the seepage force was greater than the shear strength of loess, thereby leading to stratum erosion and, consequently an unstable soil skeleton. Ultimately, the unstable slope started to move downwards, initiating the local sliding surfaces, with the small displacement rates as shown in Stages 1 and 2 at MD09. Although the effect of stratum erosion cannot be taken into account in the CSD triaxial test, the test did capture the initiation of the local failures by the axial strain εa (or volumetric strain εv) which developed at a slow pace, as shown in Stages I, III, and V at CSD3.

  3. (3)

    When subjected to an even greater seepage force, stratum erosion became more severe, which not only led to an unstable soil skeleton, but also indicated a further development in the local sliding surfaces. The slope thus moved downwards at displacement rates that were nearly twofold higher than those during the previous stage, as shown in Stage 3 of MD09. The results of the CSD triaxial test linked the further development of the local failures with the axial strain εa (or volumetric strain εv) that developed at a rapid pace, as shown in Stages II, IV, and VI of CSD3.

  4. (4)

    The development of slow strain–rapid strain cycles in the ε– p′ plot from the CSD triaxial test results implied that the loess landslides which occurred in the study area were a feature of this progressive failure. Preventive measures to reduce the irrigation-water infiltration into loess slope should be taken along with the effective management of agricultural irrigation to avoid similar incidents in the future.