Keywords

Introduction

Globally, rainfall and earthquakes are two of the most prominent causes of landslides, which result in thousands of injuries and deaths, and billions of dollars of property damage and revenue losses each year. In recent history, the 2014 Oso Landslide in Washington, which was said to be triggered by heavy rainfall, is among the worst landslide disasters in the United States (Keaton et al. 2014; Iverson et al. 2015). Researchers, including Tiwari et al. (2016), Tran (2016), Collins and Znidarcic (2004), Kim et al. (2012), Orense et al. (2004), Ling and Ling (2012), Lu et al. (2012), and Tiwari and Caballero (2015), have studied the effect of various parameters such as slope geometry, soil density, and intensity of rainfall on the stability of rainfall-induced slope failures. Furthermore, databases containing over 3400 landslides after the 2011 Tohoku Earthquake in Japan (Wartman et al. 2013), over 14,000 landslides after the 2015 Gorkha Earthquake in Nepal (Tiwari et al. 2016), over 56,000 landslides after the 2008 Wenchuan Earthquake in China (Dai et al. 2011) and over 10,000 landslides after the 1999 Chi-Chi Earthquake in Taiwan (Khazai and Sitar 2004) clearly demonstrated the disastrous consequences of earthquake-induced landslides. Following strong ground shaking, a major concern for residents and the local officials in the region is the potential damage and consequences that could result if additional landslides are triggered during the future rainstorms and monsoon seasons. Despite this, there has been little research geared at understanding the influence of post-earthquake rainfall on the stability of slopes. This study aims to fill that gap by examining the effect of rainfall and seismic shaking on model soil slopes and numerical simulations using the GeoStudio software suite.

The soil used in the model soil slopes was collected from a housing development project in Mission Viejo, California. At this project site, slope movements were reported to have occurred on January 20, 2005 near the residences on Encorvardo Lane on the eastern side of Ferrocarril. Heavy rainfall from December 28, 2004 through January 11, 2005 resulted in almost 254 mm of precipitation. It was following this storm that initial movements were recorded on a 70 ft tall graded slope constructed in 1967. Model soil slopes tested in this study were prepared from compacted fill slope material obtained from this housing development project. Several standard laboratory tests were also conducted to determine the material properties. This paper will detail the influence of post-earthquake rainfall on the stability of slopes by presenting the results of the model soil slopes and the numerical simulations conducted.

Materials and Methods

The soil used in this study was collected from the housing development project in Mission Viejo, California. It was used to measure the grain size distribution (ASTM D422), the index properties (ASTM D4318), maximum dry density and the optimum moisture content (ASTM D1557), the saturated coefficients of permeability (ASTM D5084), the soil water characteristic curve and the shear strength parameters (ASTM D3080). Table 1 outlines the material properties for the soil properties.

Table 1 Soil properties
Full size table

The slope was constructed in a 1.2 m × 1.2 m × 1.2 m Plexiglas container mounted on a shake table. At the bottom of the container, a drainage layer consisting of a 5-cm-thick gravel layer and perforated plastic pipe drainage network were installed. To prevent the soil from clogging the drainage layer and the perforated pipe, a geo-textile was placed on top of the gravel drainage layer. The soil used to construct the model slopes was first sieved through a #40 sieve using a mechanical shaker and mixed with sufficient water to have an initial moisture content of 12%, corresponding to the field moisture content. The soil was placed in 5 cm thick lifts compacted to a void ratio of 1.2 to create a slope inclined at 40°. The completed model slope is pictured in Fig. 1.

Fig. 1
figure 1

Slope model used in this study

Full size image

Tensiometers and accelerometers were installed at various locations and depths (Figs. 2 and 3) within the slope to instrument the slope in order to measure the pore water pressure during the seismic loading and the post-earthquake rainfall event. Decagon T5 tensiometers with an active surface area of 0.5 cm2 and a 5 mm diameter ceramic tip were used in this study. They are capable of recording pore water pressures ranging from +100 to −85 kPa. The holes made to install the tensiometers were backfilled with the slope material and the surface was covered with a bentonite slurry to prevent preferential movement of water through those locations.

Fig. 2
figure 2

Locations of tensiometers

Full size image
Fig. 3
figure 3

Plan view of slope showing locations of the accelerometers

Full size image

The slope was then subjected to different ground motions using the shake table at California State University, Fullerton. The 1.5 m × 2 m shake table uses a 25 kN actuator and has a ±6.5 cm horizontal capacity. The model slope was subjected to sinusoidal motions with accelerations ranging in amplitude from 0.1 g to 0.3 g with frequencies between 1 and 3 Hz. Each motion was applied for 10 to 50 cycles of loading. The ground motion recorded at Station 90095 during the 1994 Northridge Earthquake was also applied to the slope model. The ground motions applied are shown in Fig. 4. Following the seismic loading, a rain simulator system was placed on top of the slope. The system was used to apply rainfall at an intensity of 3.6 cm/h until the slope became completely saturated, as determined from the location of the wetting front. A photograph of the rain simulator system applying rainfall to the model slope is presented in Fig. 5.

Fig. 4
figure 4

Applied cyclic loading

Full size image
Fig. 5
figure 5

Rain simulator system on slope model

Full size image

A second model was prepared following the procedure described above and instrumented as illustrated in Fig. 2. In this model, the slope was subjected to rainfall without any seismic excitation. The results obtained served as baseline in order to compare the effect of the seismic loading on the observed responses.

The slope geometry and material properties were used to perform numerical simulations using the SEEP/W, SIGMA/W, and QUAKE/W modules in the GeoStudio software suite. SEEP/W was used obtain the pore-water pressure distribution. The seismic loading was simulated using QUAKE/W, in which the input motion was the same as that used in the experimental models, as shown in Fig. 4. The seepage patterns as a result of the rainfall were simulated using SIGMA/W.

Results and Discussion

The movement of the wetting fronts with time for the two models prepared in the laboratory that were subjected to post-earthquake rainfall and to rainfall without earthquake shaking are compared and presented in Fig. 6. The locations of the tensiometers, denoted by T, are also shown in this figure. Unfortunately, all except one of the tensiometers malfunctioned in the slope model that was subjected to rainfall without any shaking (Fig. 6b). Figure 6 also contains the values of the seepage velocity computed, based on the time required for the wetting front to reach Point A. It is clear from Fig. 6 that more time was required for the complete saturation of the model that was subjected to earthquake loading than the time required for complete saturation of the model that was not subjected to any shaking. Specifically, the seepage velocity in the slope subjected to post-earthquake rainfall was 14.6 cm/h in comparison to the seepage velocity of 25.7 cm/h in the slope subjected to rainfall without earthquake loading. A comparison of the results from the tensiometers (Fig. 7) also agrees with the observations made of the wetting fronts. Specifically, the results show that approximately 2.7 h were required for the suction to reduce to a value of zero at the location of T1 in the slope subjected to rainfall without earthquake loading, while approximately 3.2 h were required in the slope subjected to post-earthquake rainfall.

Fig. 6
figure 6

Comparison of the wetting fronts and seepage velocities for slopes subjected to a post-earthquake rainfall and b rainfall without earthquake loading

Full size image
Fig. 7
figure 7

Comparison of suction recorded at tensiometer 1 (T1) for slopes subjected to a post-earthquake rainfall and b rainfall without earthquake loading

Full size image

Results from Numerical Simulations

Figure 8 shows the movement of the wetting front with time obtained from the numerical simulations conducted as part of this study. The results in Fig. 8 were used to calculate the seepage velocity to point A and compared with the experiment results. As it can be observed from Fig. 6b and Fig. 8, the seepage velocity from the numerical simulations are similar to those obtained from the experimental models.

Fig. 8
figure 8

Movement of the wetting front with time from numerical simulations

Full size image

SLOPE/W was used to calculate the factor of safety of the slope subjected to post-earthquake rainfall. The variation in the factor of safety during the earthquake loading is presented in Fig. 9, while the results for the factor of safety during the post-earthquake rainfall are presented in Fig. 10. Figure 10 also contains the results for the factor of safety for the slope subjected to rainfall without any earthquake loading. The figure shows that the factor of safety for the slope subjected to rainfall following earthquake shaking was higher than the slope subjected rainfall without any earthquake shaking. This can be attributed to the fact that the seepage velocity in the slope subjected to post-earthquake rainfall was lower than in the slope subjected to rainfall without earthquake loading.

Fig. 9
figure 9

Factor of safety computed from SLOPE/W during earthquake loading

Full size image
Fig. 10
figure 10

Factor of safety computed from SLOPE/W in a slope subjected to rainfall following earthquake shaking and a slope subjected to rainfall without any earthquake loads

Full size image

Conclusions

In order to determine potential causes for the computed factors of safety, soil was collected from a landslide site and used to prepare the laboratory slope models in a Plexiglas container. The model slopes, instrumented with tensiometers and copper wires, were subjected to rainfall using a rain simulator system. A second model prepared to the same slope geometry and density was constructed on top of a shake table and subjected to a series of sinusoidal earthquake loading functions, as well as to the motion recorded from the 1995 Northridge Earthquake, before being subjected to rainfall at the same intensity as the previous model. The results showed that the slope subjected to post-earthquake rainfall required more time to become completely saturated and had a lower seepage velocity than the slope subjected to rainfall without earthquake shaking. The results suggest that the earthquake shaking was responsible for increasing the density of the slope material and thereby effectively reducing the permeability of the material, resulting in a more stable slope.