Sliding-Surface Liquefaction and Undrained Steady-State Shear-Strength

Abstract

The sliding-surface liquefaction (SSL) is one of the most important concepts to understand and reduce rapid landslide disaster risk. SSL occurs even in dense sandy layers. SSL is caused by a series of phenomena, 1) grain-crushing due to shearing under overburden pressure in the shear zone, 2) volume reduction in the shear zone, 3) generation of high pore-water pressure in the shear zone, and 4) liquefaction of the shear zone material.

After SSL, a mass of soil layer above the liquified sliding surface moves at high speed. Motion of sandy layer above the liquified sliding-surface was physically simulated under the corresponding normal stress by the undrained stress-control ring-shear apparatus. The shear stress acting on the sliding-surface and the generated pore pressure near the sliding surface are monitored. After failure, shear strength mobilized on the shear zone decreases by the effects of increased pore-water pressure, then, the shear strength working on the shear surface together with the generated pore pressure reaches a certain constant value which is the undrained steady-state shear-strength (USS). After reaching to USS, only shear displacement is increased without any change of shear stress and pore pressure. It is a common feature of rapid and long-travelling landslides which poses a high risk to people living in/near slopes.

1 Introduction

A new Open Access Book Series “Progress in Landslide Science and Technology (P-LRT)” was launched for the platform to implement the Kyoto Landslide Commitment 2020 in 2022. P-LRT has a new category of ICL landslide lessons. This article is the first ICL landslide lesson in P-LRT. This article consists of the following four parts.

Background information of SSL in the context of landslide risk reduction

• Concept of the sliding-surface liquefaction

• Initiation mechanism of landslides and Types of landslides based on the stress path

Tools to support SSL and USS

• Physical landslide simulator: Formation of sliding surface and post failure motion

• Development of undrained dynamic loading ring shear apparatus

• Development of LS-RAPID to simulate long travel distance using the post failure strength reduction and the USS

Original test results for this article

• Experiment of standard sand “Silica sand No.6”

• Experiment of natural soils (example of Volcanic debris taken from Unzen)

Different case studies of sliding-surface liquefaction from previous research

• 2006 Mega-slide in Leyte, Philippines, triggered by a small earthquake after long rainfall.

• Large-scale landslides triggered by the 2004 Mid-Niigata earthquake, Japan

• Hypothetical submarine landslides in Suruga Bay, Japan

• Large-scale landslides triggered by the 2011 typhoon “Talas” in Kii Peninsula, Japan

• 1792 Historical mega-slide in Unzen, Japan.

2 Background Information of Significance of Sliding-Surface Liquefaction

2.1 Background

Landslide disasters are classified in the following two types,

• 1. Disasters give damages to roads, railways, brides, dams mining pits, and other engineering facilities. This type of landslide disasters have studied with funds to maintain the social infrastructure.

• 2. Human disasters by rapid and often long-travelling fatal landslides. Funding of research on fatal landslides is not always easy.

The International Consortium on Landslides (ICL) together with the landslide community has made efforts to obtain support from the United Nations, International and national stake holders. Supports from the United Nations and International organizations to the ICL and the landslide communities are realized the establishment of the International Programme on Landslides (IPL) in 2006, the Landslide Sendai Partnerships 2015–2025 in 2015, and the recent framework is the Kyoto Landslide Commitment 2020 for global promotion of understanding and reducing landslide disaster risk.

The founding of this new open access book series is owed to the Kyoto Landslide Commitment 2020 and the KLC2020 official promoters. The sliding-surface liquefaction is one of key to understand the fatal rapid landsides. Before entering in to technical contents, we wish to outline the framework which has enabled to launch the IPL, the Sendai landslide Partnerships, and the Kyoto Landslide Commitment 2020 with thanks for all supporting organizations as below.

2.2 Kyoto Landslide Commitment 2020

Kyoto 2020 Commitment for Global Promotion of Understanding and Reducing Landslide Disaster Risk (Kyoto Landslide Commitment 2020: KLC2020): A Commitment to the ISDR-ICL Sendai Partnerships 2015–2025, the Sendai Framework for Disaster Risk Reduction 2015–2030, the 2030 United Nations Agenda Sustainable Development Goals, the New Urban Agenda and the Paris Climate Agreement was launched in Kyoto, Japan on 5 November 2020 (Sassa 2021).

Figure 1 is the logo of the ICL (left) and the Kyoto Landslide Commitment 2020 (right). The Kyoto Landslide Commitment 2020 was established based on the ICL and its scientific programme IPL. The ICL was founded in 2002, the IPL was launched in 2006 with the MoU between the ICL and each of the following five United Nations Organizations and two global stakeholders.

• United Nations Educational, Scientific and Cultural Organization (UNESCO)

• United Nations Office for Disaster Risk Reduction (UNDRR)

• World Meteorological Organization (WMO)

• Food and Agriculture Organization (FAO)

• United Nations University (UNU)

• International Science Council (ISC)

• World Federation of Engineering Organizations (WFEO)

Based on the IPL and the 2015 Sendai Landslide partnerships (Sassa 2016), the Kyoto Landslide Commitment 2020 has been launched in 2020. This framework is expected to continue to 2030 and beyond. Blue arrow symbolizes this process (Sassa et al. 2022).

Fig. 1

Logos of the ICL and the Kyoto Landslide Commitment 2020

The Kyoto Landslide Commitment 2020 was signed by 90 United Nations, Global, International, and national organizations by 5 November 2020. Representatives from all 90 organizations are invited to join these launching sessions. Under the participation of all KLC2020 partners and greetings from major organizations, the Kyoto Landslide Commitment 2020 was established by the 2020 Kyoto Declaration “Launching of Kyoto Landslide Commitment 2020.” All 90 partners and participants of this launching session are reported in Landslides, Vol.18, No.1 issue (Sassa 2021).

The ICL is a host organization of KLC2020 and also the Open Access Book Series “Progress in Landslide Research and Technology” for KLC2020. The IPL and also Kyoto Landslide Commitment 2020 is a global and long-standing cooperation framework as long as landslide disaster risk threatens people’s lives worldwide and the society.

2.3 The Sliding-Surface Liquefaction in Landslides

A landslide is a downslope movement of rock or soil, or both. The movement velocity can be very different. Cruden and Varnes (1996) classified the velocity of landslides in the following seven classes (Table 1).

Table 1 Seven velocity class of landslides

Class 1 to Class 5 are slower than 3 m/min. Though these landslides may damage infrastructure, but the risk to human lives are not serious. However, landslides at the velocity class No.6 (50 mm/s to 5 m/s) are dangerous. Especially, Velocity class No.7 (faster than 5 m/s) landslides are very rapid and fatal.

Landslides as a result of the sliding-surface liquefaction move very fast, similar to skiing or sliding on snow or skating on ice. Snow under rapidly moving skiing and sled or ice under skating blade changes from snow/ice to water by rapid loading. When water entering speed under skiing, sliding or skating blades is greater than the velocity of water drain under skiing or sled or skating blades, undrained loading state is formed. Namely water (liquid) exists under skiing or skating.

Let us examine what is the mechanism of rapid motion of the skater (non-liquefied mass) on the ice using the Fig. 2. When a skater stands on a skating blade and moves on the ice by kicking another skating blade, the surface of the ice is liquefied to water by loading and shearing by the moving skating metal edge. The water generated under the metal edge laterally drains to both sides of the metal edge as well as forward and backward. At a slower speed than a critical value, water drainage is faster than the water generation speed. In this case, water cannot exist between the skating edge and the ice. Frictional resistance is that of contact between the metal edge and the ice. Whereas, in the case of faster speed than a critical value, a water layer remains between the ice and the metal edge. As such, frictional resistance is close to the value of water. The skater can slide very fast on the liquefied ice layer. The sliding-surface liquefaction beneath the skating blade is the reason for the rapid motion of the skater. The mechanism of the sliding-surface liquefaction in landslides is analogous to that of the ice skating. Only difference is in the skater and the moving landslide mass on the sliding-surface liquefaction layer. The ice corresponds to the saturated soil layer. The ice melts by loading and shearing. The saturated soil layer is liquefied by grain crushing due to shearing.

Fig. 2

Sliding-surface liquefaction in ice-skating

Sassa et al. investigated the landslide in Leyte Island in Philippines in 2006 which was triggered by a very small earthquake after a long rainfall. The landslide was rapid and the travel distance was long and wide. It destroyed one village and killed around 1000 people. We investigated this landslide and took samples and conducted the ring shear test seismic loading using the monitored earthquake record (Sassa et al. 2010). The mechanism was investigated using the undrained dynamic-loading ring-shear apparatus which was developed by K. Sassa and colleagues of Kyoto University. It is explained in Tools to support the concepts of the sliding-surface liquefaction (SSL) and the undrained steady-state shear-strength (USS) in more detail. This ICL Landslide Lesson introduces sliding-surface liquefaction and undrained steady-state shear-strength. We will explain, with the illustration of grain-crushing, volume-reduction and pore-pressure increase, the instance of the sliding-surface liquefaction and steady-state shear strength triggered by a very small earthquake—using Fig. 3.

Fig. 3

Illustration and experimental data for the Sliding-surface liquefaction

Figure 3b shows a time series data of undrained dynamic loading ring shear test for the sample taken from the Leyte Landslide. The lines indicate monitored normal stress (black line), monitored shear stress acting on the shear surface (red line), the pore pressure monitored near the sliding zone (blue line), the shear displacement at the center of the inner ring and the outer ring of the apparatus (green line), respectively. T1 shows the start of seismic loading, T2 shows the start of shear displacement when pore pressure started to develop by grain-crushing due to shearing. T3 shows the start of the steady-state shearing, in which normal stress, shear stress, and pore pressure were constant, but only shear displacement proceeded.

Figure 3a is an illustration of grain-crushing of volcanic debris due to shearing. The left figure illustrates before applying the seismic loading at, or before T1. The right figure illustrates the situation at the steady-state shearing at i.e. after T3. Angular volcanic grains are crushed due to shearing under a high-normal stress. Then, volume reduction proceeds. The shear box is fully saturated and shearing is maintained under undrained condition.

The situation explained above demonstrate that the pore pressure should rise due to the tendency of volume reduction. In the case of drained condition, a greater volume reduction will occur, but pore pressure does not build up. Due to the tendency of volume reduction and the resulting pore pressure increase, liquefaction occurs in a narrow shear zone. It is different from liquefaction of saturated and loose sand mass during an earthquake. Like in the case of skating, often, the whole landslide mass above the sliding surface is not always liquefied.

2.4 Landslide Initiation Mechanism and Landslide Types by Shearing Mechanism

The landslide initiation mechanism from the stress path at the sliding surface is explained using Fig. 4. The upper figure of Fig. 4 shows a section of slope and one vertical column element and its working force elements. A mass of the column (m) x gravity (g) is working on the soil column. The stresses acting on the bottom of the column, namely sliding surface, are 1) normal stress: mg·cosθ. 2) shear stress: mg·sinθ.

Fig. 4

Landslide initiation mechanism by pore pressure increase

The lower figure of Fig. 4 is a graph between normal stress and shear stress. Initial stress without pore water pressure is Point I (σ0, τ0). The length of the line from the origin (0) to Point I is mg in the case of no porewater pressure acting on the sliding plane. When ground water level increases by rainfalls, the stress point I will move to the left. When the stress point will reach the failure line, landslide will be initiated. Deep landslide needs a large increase of pore-water pressure until failure. Shallow landslides are generated by a smaller pore-water pressure increase. But the failure line is the same. Mechanism of rain-induced landslide are the same. The stress path is also changed under earthquake wave. In this case, stress path is Not the straight line like the Fig. 4. If pore water pressure is high, and the initial stress point is near the failure line before an earthquake, a small earthquake triggers a landslide. The Leyte landslide case (Fig. 3) is the case. Earthquake shaking is the last push to the failure.

2.4.1 Shearing Types of Landslides in Stress Path

Sassa published the Geotechnical Classification of Landslides (Sassa 1985, 1989), the sliding surface liquefaction was explained in his further publication (Sassa 1999). The extremely slow moving “Creep” is a geomorphic process, but not the target of landslide risk reduction.

It is important to understand the mechanism of the landslide initiation and the post-failure motion for understanding and reducing landslide disaster risk, focused by the KLC2020. Four types of shearing of landslides are classified in the stress path which shows the shift of the stress on the sliding surface from a stable state to failure, and from the failure to the steady state motion in the domain of shear stress (vertical axis) and normal stress (horizontal axis).

Type I: Peak strength slide

This type is called as “Slope Failure” in Japan and others. This type of landslides occurs in steep and stiff slopes at the inclination steeper than the failure line during motion. Most cases are the fist-time slide (virginal slide) on the slope. After failure, the stress pass drops to the failure line during motion. The gravity force by the stress reduction from the peak to the stress during motion is spent for the acceleration of the landslide mass. Therefore, the speed is fast. If the sliding surface is saturated, and shearing causes volume reduction generating high-water pressure, it is classified as Type IV-sliding surface liquefaction.

Type II: Residual state slide

This type of landslide repeatedly moves. The soil layer on the sliding surface is in a residual state, no stress reduction takes place at failure. So, no acceleration will occur. Slow and very small distance of movement occurs, but soon stops due to the increase of stability by the movement onto a flatter slope. The movement is very limited and the ground surface is gentle, and often the soils are fertile and wet (water rich). Those residual state landslide areas are used for paddy field to product rice and pasture for grazing cattle. The movement is gentle and poses no/less risk to cattle and working farmers, although those areas are not always suitable to construct houses and buildings. Due to rainfall and ground water rise during rainy season, the stress will move leftward due to pore-water pressure increase, and may reach the failure line at residual state and start movement. However, it will stop after a limited movement.

Type III: Mass liquefaction

In saturated coastal areas consisting of a loose sand deposit, sometimes landslides occur during earthquakes.

Earthquake shaking causes liquefaction of loose sand deposit. A liquefied sand layer starts to move even in a gentle slope. In the mountain slopes, sometimes those mass liquefactions will occur. Along a concave slope such as extension of torrent, surface water caused by rainfall will penetrates into the ground and flow into the underground of the concave slopes. A long-time ground water flow gradually erodes fine particles within the sandy layer. Then, a loose structure of sand layer is produced. During earthquakes or during heavy rainfalls, such a loose structure may be failed, and volume reduction occurs. The mass liquefaction occurs in the loose-structure part of the sand layer. The upper soil layer will start to move on the liquefied sand mass. After earthquakes or heavy rainfall, a new valley is formed, suddenly.

It is a phenomenon similar to if a snake went out from the slope or a valley deposit went out. This phenomenon is called as Snake-off or Valley-off in Japan (Fig. 5).

Fig. 5

Shearing types of Landslides. F.L.P failure line at peak strength, F.L.M failure line during motion, F.L.R failure line at residual state

Type IV: Sliding-Surface-Liquefaction

This type of landslide occurs without having loose structure of sands. The soils in the shear zone must be saturated, and the soil grains can be crushed during shearing under the given overburden pressure. Strong grains can be crushed during shearing at a greater overburden pressure. This type of landslides occurs in deep slopes as well as shallow slopes. The effective stress will move during rainfalls or/and earthquakes from the initial stress point to the failure line at peak. Failure occurs when the stress point reaches the failure line at peak. When sand/gravel grains are subjected to grain-crushing by shearing, volume reduction of the sand/gravel grains should occur. If the soil layer is saturated, a high-pore-water pressure is generated. Namely the sliding-surface liquefaction occurs in the shear zone. Thereafter the soil layer above the liquified sliding surface will move rapidly. During this rapid motion of landslides, sometimes the soil layer is kept with standing stress; sometimes the whole mass is liquefied by shaking during the motion or collision to Sabo dams (check dams), natural steps, sharp curves, or sharp change of the width of the valley route and others. When the whole moving mass is liquefied, it is called debris flow.

3 Tools to Support SSL and USS

The study of the sliding-surface liquefaction (SSL) required a development of the testing apparatus to reproduce the sliding-surface liquefaction within the apparatus. The study of the undrained steady-state shear strength (USS), the key parameter controlling the post-failure motion required an apparatus to reproduce the long-range shear displacement reaching to a steady state.

To assess the landslide disaster risk due to SSL and USS, an integrated computer simulation model for the initiation and motion using the SSL and USS parameter is needed. A difficulty of landslide simulation is how to model the transient phase from the stability analysis of the stable slope before motion to the dynamics of a moving landslide mass including solid (sand) and liquid (water).

This section devotes to the concept of a landside ring-shear simulator, development of a series of undrained dynamic-loading ring-shear apparatus and development of an integrated landslide simulation model using the landslide dynamics parameters measured or estimated by the developed undrained dynamic-loading ring-shear apparatus (UDRA).

3.1 Physical Landslide Simulator—Formation of Sliding-Surface and Post-Failure Motion

The concept for the development of the undrained dynamic-loading ring-shear apparatus is to physically re-produce the formation of sliding surface and post failure motion within an apparatus.

Image is shown in Fig. 6. Sample is taken from the shear zone in the slope and/or the shear zone which will be formed within a soil deposit by a rapid (undrained) loading due to the moving landslide mass.

Fig. 6

Concept of Landslide Ring-shear Simulator

When the landslide mass initiated from the steep slope reaches onto an alluvial deposit, saturated at a certain depth, a high-water pressure will be generated within the ground water level, and a soil layer will be sheared. To estimate the landslide dynamics, parameters of landslide dynamics are measured by reproducing this process.

Samples taken from the landslide sites (potential sliding surface) are set up in the ring-shear box. They are saturated and consolidated under the in-situ overburden pressure. Then, normal stress, shear stress and pore-water pressure in the site are loaded on the sample.

• To investigate rain-induced landslides, porewater pressure is increased to reproduce the initiation of sliding surface and post failure motion of the rain-induced landslides.

• To investigate earthquake-induced landslides, seismic stresses are loaded on the initial stresses before landslides.

• To investigate landslides triggered by an earthquake after rainfalls (typical example is the 2006 Leyte landslide in Philippines), initially the pore pressure within the shear box is increased, seismic stress due to earthquake is loaded on the sample within the shear box, and the resulting behavior (mobilized shear stress on the sliding surface, generated pore water pressure near the shear zone, post-failure motion-shear displacement) are monitored.

Landslide velocity after failure is important for the study. The landslide velocity is controlled by pore-water pressure generated in the shear zone. The key element to physically reproduce the post-failure motion is how to reproduce the pore pressure in the shear zone, and how to monitor the generated pore pressure under moving rings. The objectives of developing the undrained dynamic-loading ring shear test were how to keep undrained state of the ring shear box in a better way and how to monitor the pore-pressure near the shear zone.

The undrained dynamic-loading ring-shear apparatus was successfully developed. The undrained ring-shear testing reproduces the sliding-surface liquefaction and the steady-state shear strength which are the results of undrained shearing, pore-water pressure generation by grain crushing, which are not always understood. It is very important to coin the term “sliding surface liquefaction”, the key of rapid and fatal landslides; then, developing the apparatus in DPRI, Kyoto University. Sassa et al. (2004) developed the undrained transparent shear box for DPRI-7 model, focusing on demonstration to the community.

DPRI-7 has a normal stainless-steal shear box for research and a transparent shear box for demonstrating to the community. Let us show the test result of a transparent shear box using Fig. 7 and the video on the initiation, failure, acceleration and steady state motion. It is a visual example to present the capability and performance of the undrained dynamic loading ring-shear apparatus.

Fig. 7

Test result of undrained speed control shear test by DPRI-7 (Sassa et al. 2004). Sample: Silica sand No.1, Normal stress: 200 kPa, Maximum shear displacement: 30 m, Speed control: 0–200 m/s within 0–5 s. Video link: https://us06web.zoom.us/rec/share/DQdfHQhL-F6QhvAJvAe8GUERe%2D%2DIdXdGpfpYA8djeMjTql-WtsuwvFzGmWo3i5At.8E-DdvJWe1KVIt_X

Photo (a) of Fig. 7 shows the transparent shear box. A stainless-steel metal edge is attached at the bottom of the upper shear box. A rubber edge is attached on the top of the lower shear box. To keep undrained condition, two metal edges and rubber edges must be parallel, both the inside and the outside shear boxes in the lower shear boxes as well as the upper shear boxes. To ensure this condition, the rubber edge attached to the inner and the outer shear boxes of the lower shear box was processed by the lathe. The distance of the upper shear box and the lower shear box is always adjusted by the servo-control piston with precise monitoring of the distance by a gap sensor (GS) with a precision of 1/1000 mm. The plastic shear box is easily deformed by temperature change. We could not successfully perform the undrained test of transparent shear box for many times. It is not the practical test.

Used sample was silica sand No.1. It is coarse sand, and easy to observe grains and the grain crushing. Test was shear speed control test. Shear speed was increased from 0 cm/s to 200 cm/s for 5 s. It was programmed to stop shearing when the shear displacement reaches 30 m. The normal stress was kept constant at 200 kPa.

Photo (b) presents the motion of lower shear box. The grains along the shear zone are crushed due to shearing under 200 kPa normal stress. Photo (c) presents muddy water going upward and downward due to high pore-water pressure. Photo (d) presents steady state shearing.

Monitored data during the test are shown in the right column.

• (a) Present time series data of normal stress in the black color, and shear stress mobilized at the shear surface in the red color, and the pore water pressure in the blue color.

• (b) The shear displacement is shown in the black color and the shearing speed is shown in the red color.

Comparing two figures (a) and (b), the peak shear strength appeared at the 0.154 cm shear displacement, whereas 0.154 shear displacement is not visible in Fig. (b). The servo-control normal stress is not kept this instant at failure (a). At around 10 cm shear displacement, pore-water pressure suddenly stated to increase which should be the result of grain crushing and volume reduction in the shear zone. Excess pore-water pressure is generated and increased at 10 cm to 1000 cm shear displacement. Then, the value of pore-water pressure became constant, namely steady-state of shearing was created and continued until the end of the test (3000 cm). (c) in the right column presents the stress path of this test.

When the shear displacement is started, the shear stress is jumped to the failure line at peak at 0.154 cm shear displacement. The stress path drops to the failure line during motion at 1 cm shear displacement. Thereafter, the stress point goes down along the failure line during motion; then stops at 3000 cm shear displacement. The apparent friction angle at the final value was 3.3 degrees. It is very small compared to the peak friction angle and the friction angle during motion. A rapid landslide occurs and the movement continues until 3.3-degree slope.

Figure 8 presents the grain size distribution. The black line shows that of the original sample, namely sample before shearing. The red line presents the grain size distribution in the lower shear box. The finer grains increased because the finer grains produced in shear zone are falling in the lower shear box as seen in (c) and (d).

Fig. 8

Grain size distribution of the original sample (Silica sand No.1) before shearing, and the sample taken from the lower shear box (fine crushed particles fell down from the shear zone to the lower shear box)

Please click the link to Video. You can observe this process from (a) to (d), and 200 cm/s speed undrained shearing.

3.2 Development of Undrained Dynamic-Loading Ring-Shear Apparatus

The development of untrained dynamic-loading ring shear test was aimed on how to better keep undrained state of the ring shear box and how to monitor the pore pressure near the shear zone (Table 2).

Table 2 A series of dynamic-loading ring-shear apparatuses developed by Sassa and colleagues

3.2.1 An Overview of a Series of the Undrained Dynamic-Loading Ring-Shear Apparatus in the Chronological Order

• 1. Preparation stage: DPRI-1 before the undrained condition

K. Sassa developed the first ring shear apparatus (termed as DPRI-1). It was reported in Sassa (1984) in 4th ISL in Toronto, Canada and Sassa (1988) in 5th ISL in Lausanne. The outline features are presented below.

• Aim: to reproduce debris-flow motion under a certain normal stress within a rotational channel.

• Target: Debris flows frequently occurred in Volcano Usu, Volcano Sakurajima and others in Japan

• Shear box: 300–480 mm in diameter and transparent acrylic shear box for eye-monitoring flow of debris

• Maximum normal stress: 40 kPa

• Loading system: Rubber tube by air compressor and regulator which was installed between the loading plate and the top cap of the upper shear box

• Loading system: Speed-control motor

• Maximum shear speed in the center of shear box: 100 cm/s

• Gap control system: Manual gap control by measuring the position change of the upper loading plate

• Sealing of sample leakage: Silicon rubber

• Undrained condition: Not possible. No pore pressure monitoring.

• 2. Transient stage: DPRI-3 from drained ring-shear apparatus to the undrained ring-shear apparatus. It was reported in Sassa et al. (1992) in 6th ISL in Christchurch, and in Sassa (1996) in 7th ISL in Trondheim.

• Aim: to reproduce earthquake-induced landslides.

• Target: Ontake landslide triggered by the 1984 Naganoken-Seibu earthquake in Japan.

• Maximum normal stress: 500 kPa

• Loading system: Loading piston with air servo-valve, compressor and loading frame to support piston

• Monitoring: two load cells, one for loading pressure, another for side friction within the shear box

• Shear box: 210–310 mm in size with a transparent acrylic outer ring

• Loading system: Stress-control and speed-control motor

• Maximum shear speed in the center of shear box: 37 cm/s

• Gap control, undrained condition, and pore-pressure monitoring

• Sealing of water leakage: Polychloroprene (®Neoprene) rubber edge attached on the lower ring (Rubber hardness Index, 45_JIS)

• Gap control: Servo-gap control system by measuring the position change of the upper loading plate and adjusting by servo-motor

• Undrained condition and pore-water pressure monitoring: It was improved from 6th ISL in Christchurch, 1992 to 7th ISL in Trondheim,1994 Successful undrained condition: 400 kPa

• Pore pressure was monitored by a needle inserted close to the shear zone in 1992. It was monitored from the gutter (4 _ 4 mm) along the whole circumference of the upper-outer shear box 2 mm above the gap in 1994

• Major reports: Sassa et al. (1992) in 6th ISL and 7th ICL in 1996

• 3. Undrained ring-shear apparatus: DPRI-6. It was reported in Sassa et al., Vol.1, No.1 of Landslides in 2004.

1995 Hyogo-ken Nanbu earthquake caused a big disaster in Kobe, Japan. It occurred in very dry season. In the condition of no rainfall, a rapid landslide triggered in the densely populated urban area killed 34 people. We received a budget to study earthquake-induced landslides. Then, DPRI-5 and DPRI-6 was developed. The size of shear box is different. But other features are similar.

• Aim: to develop landslide ring-shear simulator for earthquake-induced landslides.

• Maximum normal stress: designed for 3000 kPa. However, normal stress servo-control system does not function well. Tests were conducted upto 750 kPa by changing the normal stress load cell.

• Normal stress loading system: Loading piston by oil servo-valve and oil-pressure pump and loading frame to support piston

• Monitoring: two load cells: one for loading pressure, the other for side friction within the shear box

• Shear box: 250–350 mm in size with non-transparent outer and inner rings of stainless steel. Maximum shear speed in the center of shear box: 224 cm/s

• Rubber edge in the gap: Polychloroprene rubber with Rubber hardness Index, 45_ JIS

• Gap control: Piston with oil servo-valve and oil pressure pump by measuring the position change of the upper loading plate. To observe the effect of extension of central axis, a displacement guide was installed within the central axis

• Successful undrained condition: 550 kPa

• Pore pressure is monitored along the entire circumference of the upper-outer shear box 2 mm above the gap (same with DPRI-3)

• 4. Undrained dynamic-loading apparatus for megaslide (3 mega pascal): ICL-2. It was reported by Sassa et al. in Vol.11, No.5 of Landslides in 2014, and Sassa & Dang in Landslide Dynamics, 2018.

• Aim: to develop a landslide ring-shear simulator for mega-slides (3 MPa normal stress) and able to be maintained in a developing country

• Targets: 1792 Unzen Mayuyama landslide killing 14,528 people

• Maximum normal stress: 3 MPa

• Normal-stress loading system: Loading piston by oil servo-valve and oil pressure pump which is retained by the central axis (no loading frame)

• Monitoring: One load cell with automatic side-friction canceling

• Shear box: 100–142 mm in diameter with non-transparent outer and inner rings of stainless steel

• Shear speed in the center of shear box: up to 50 cm/s

• Rubber edge in the gap: Polychloroprene rubber with Rubber hardness Index, 90_JIS

• Gap control: Mechanical jack driven by a servo-control motor with feed-back signal of gap sensor for the position change of the upper loading plate. (To avoid the effect of extension of central axis, a displacement guide is installed within the central axis). Successful undrained condition: 3 MPa

• Pore pressure is monitored through three-layer metal and felt cloth filter along the whole circumference of the upper-outer shear box 2 mm above the gap.

• 5. ICL-2 with a reduced loading stress (1000 kPa).

ICL-2 apparatus is used in countries other than Japan. Most users do not need normal stress more than 1000 kPa.

ICL-2 CS is the version of reduced normal stress (1000 kPa). To protect possible damage between the bottom of the upper shear box and the lower shear box due to reducing the height of the rubber edge for long use, a Teflon holder pressed the rubber edge in ICL-2. Instead of Teflon holder, Teflon O-ring was installed in the inner ring, and it protects the contact between the bottom of the upper shear box and the top of the lower shear box. The rubber edge holder is only stainless-steel holder which supports the rubber edge from the inside to resist the horizontal force from the soil sample to the rubber edge.

ICL-2CS is a slightly modified version from ICL-2. The vertical load cell was replaced by 1/3 capacity version. And the structure of sealing rubber edge was modified.

3.2.2 Difference of Loading Structures Between DPRI-6 and ICL-2 Which Enabled the Undrained Test Under a High Normal Stress (3000 kPa)

We targeted to shear soil under 3000 kPa normal stress for DPRI-6. The mechanical structure of DPRI-6 is shown in Fig. 9.

Fig. 9

Mechanical structure of DPRI-6

The oil pump and the oil piston (OP1) and oil-servo valve (SV), Normal stress monitoring load cell (N1), Side-friction monitoring load cell (N2), and the oil piston (OP2) and oil-servo valve were ready. To shear samples under the high-normal stress, two torque and speed-control servo-motors (Servo Mother No.1 and No.2) are installed. However, both DPRI-5 and DPRI-6 failed to test the undrained condition at more than 400–600 kPa. Servo-control of the normal stress loading was not possible. The cause of servo-control of normal stress was the elastic deformation of the beam to support the loading normal stress, and also the elastic extension and compression of two long shafts to support the horizontal beam. The elastic response is faster than the servo-control. Vibration occurs and the gap between the upper shear box and the lower shear box is also affected. The gap is controlled to be a constant—to keep contact pressure at the rubber edge and the bottom of the upper ring shear boxes more than pore water pressure inside the sample box. When the pore-pressure generated in the sample exceeds the contact pressure, pore-water dissipates out of the sample box. Namely undrained condition is not kept.

In order to enable to do undrained ring shear test up to 3000 kPa, we replaced the loading frame which consisted of one horizontal beam and two vertical shafts with the central axis that is pulled by one oil piston, three vertical loading rods (two rods are seen in Fig. 10, really three rods), which gives normal stress on the sample. This system has two merits.

• 1. It does not include a long deformable horizontal beam and two long shafts.

• 2. DPRI-6 needs two load pistons and two loading cells (one for loading normal stress another for side friction in the shear box. ICL-2 needs only one piston and one load cell. Namely side friction is automatically cancelled out.

• 3. Instead of oil pump and oil-servo valve, the same sized another oil piston is used for normal stress control. The stress was controlled by the servo-control motor. This system is very quiet, whereas oil pumps in DPRI-3-6 were noisy.

Fig. 10

Mechanical structure of ICL-2

The loading system in ICL-2 is explained by Fig. 11 (left). The normal stress working on the sliding surface gives upward force. This upward force is balanced by the downward force acting at the central axis. DPRI-6 needed two load cells (one for normal stress, another for side friction). In this system, the side friction is automatically canceled out. Then, one load cell is enough.

Fig. 11

Simple structure to see the forces in ICL-2 (left) and how to measure the stress working on the shear surface (right)

Rubber edge keeps contacts to the upper shear box. This contact force is necessary to keep the undrained state of the shear box. As we can see in the figure, the normal stress measured by the load cell N is the sum of the normal stresses acting on the sample and the rubber edge contact force. The real normal stress must be modified from the measured normal stress.

It was investigated in the following way. The sample was consolidated by a normal stress. Then, pore water pressure gradually increased. The water pressure when the instant of lifting of the loading plate (namely a water film or a thin water layer is formed) should be the same with the normal stress working on the shear zone. Table 3 shows the relationship between the measured normal stress, the water pressure at the instant of lifting the loading plate, and its ratio (α).

Table 3 Measured normal stress, the pore water pressure equal to the real normal stress, and ratio (α)

α = measured water pressure u (equal to the actual normal stress on the sliding surface)/measured stress by N. The ratio changes by the value of normal stress from 0.85 to 0.96. The undrained test needs the sealing rubber edge. The effect must be corrected by rubber edge correction factor (α) (Fig. 12).

Fig. 12

Relationship between the measure normal stress and the correction factor (α). Real normal stress = α × (normal stress measured by the load cell (N))

3.2.3 Electronic Servo-control System of ICL-2 for Undrained State of Shearing, and Seismic-Loading Test, and Pore-Pressure Control Test

Figure 13 presents the servo-control system of ICL-2.

• 1. Gap servo-control: the position of the upper shear box must be controlled very precisely to keep the rubber edge contact force greater than the generated pore-pressure. The position is adjusted by the control signal (CS) from the computer through serve amp (SA). The precise location is measured by the gap sensor (1/1000 mm precision). To avoid the extension and compression of central axis, a hole is made in the center of the central axis. A narrow bar (orange color) keeps touching the gap sensor through the hole in the center of the central axis. The current precise location is informed by the servo-amp as feedback signal (FS). If CS and FS is not the same, servo-amp sends the further CS to servo-motor to adjust the location. Then, the gap, namely the contact force of the rubber edge is kept constant even during seismic loading.

• 2. Shear stress servo-control: in the case of cyclic loading test or seismic loading test, the computer sends CS to shear stress loading Servo-motor. Shear stress acting on the shear surface is monitored by shear load cell (S1 and S2), and the monitored shear stress is informed to the shear stress servo-amp as the feedback signal (FS). If FS is not the same value with the CS, further modified CS will be sent until reaching CS=FS. S1 and S2 shear load cells retain the rotation of the upper shear box against the rotational force transferred from the sliding surface. Thus, the value corresponds to the shear stress acting on the shear surface.

• 3. Normal stress control: Control signal is sent from the computer to the servo-amp and the normal stress loading servomotor. The oil pressure is changed within the loading piston (LP-1). LP-1 pressure is transferred to LP-2. LP-2 loading is given to the sample on the shear box. Normal stress loading on the shear surface is retained by the vertical load cell (N). The value of N is sent to the servo-amp as the feedback signal (FS). If FS is not the same with CS, further CS is sent to Servo-motor and continues until the FS=CS.

• 4. Pore-pressure control test: to reproduce the rain-induced landslide, pore pressure is increased by the control signal from the computer to the Servo-amp and to the servo-motor to create a pore-water pressure. The increased pore pressure is monitored by a pore pressure gauge installed in the shear box near the sliding surface. The monitored value is sent to SA as a feedback signal. When CS-FS is different, further CS will be sent and continues until FS=CS.

Fig. 13

Electronic servo-control System of ICL-2

3.2.4 Development of a Series of Rubber Edge Structure for Undrained Water Sealing

Water sealing and pore pressure measurement near the sliding surface have progressed from DPRI-3. Many trial types of water sealing and pore pressure monitoring were examined in DPRI-3. Figure 14 presents the most successful version DPRI-3 final. The stair shape of rubber edge was attached on the lower ring by Acryl Glue. The thickness of Acryls glue cannot be constant. Then, top surface of the rubber edges of inner ring and outer ring was processed by lathe post attachment.

Fig. 14

DPRI-3 Final: Stair shape of the polychloroprene rubber edge (Rubber Hardness Index is 45_JIS) attached onto the lower ring.

• Successful undrained condition up to 400 kPa under 0.1 Hz cyclic test.

• Successful pore-pressure measuring through the gutter along the entire circumference in the upper ring

The three layers filter (two metal filters and one felt cloth filter) were set in the inlet of pore water in the sample. Pore-pressure was monitored by a pore pressure transducer with a small-scale sensitive diaphragm. Water is collected through the metal filters filled in the gutter along the whole circumference in the upper ring. The lower ring of DPRI-3 was an acryl transparent shear box.

Figure 15 presents the structure of water sealing used in DPRI-6. Basic structure was the same with DPRI-3. The shape of filters is greater to collect more water. The undrained condition is better kept for seismic loading.

Fig. 15

DPRI-6: Stair shape of the polychloroprene rubber edge (Rubber Hardness Index is 45_JIS) attached onto the lower ring

After the rubber edge is attached, the upper surfaces of inner and outer rubber edges need to be processed by a lathe or file to ensure that the rubber surface is at the same height everywhere. Successful undrained condition up to 550 kPa under realistic seismic wave loading. Successful pore-pressure measuring through the gutter along the entire circumference in the upper ring.

Figure 16 presents the structure of ICL-2. Processing by lathe or file after attaching the rubber edge is troublesome. And it cannot be maintained in developing countries. In ICL-2, lath or file is not needed. Rubber edge is simply placed in the lower ring. Then, the rubber edge is pressed by Teflon holder, and the Teflon holder is pressed by stainless steel holder.

Fig. 16

ICL-2 The polychloroprene rubber edge (grey) (Rubber Hardness Index is 90_JIS) was pressed by a polytetrafluoroethylene (Teflon) ring holder (pink) which was pressed by a stainless-steel ring holder (dark blue). No glue was used. The rubber edge was simply placed and pressed. Glue to attach the rubber edge is not used. No need for the process of rubber edge surface by a lathe or file after rubber edge setting. A great progress in the maintenance of the apparatus! Successful undrained condition—up to 3000 kPa. Successful pore-pressure measurement—up to 3000 kPa

ICL2 tests soil at 3000 kPa normal stress. In this case, lateral pressure acts from the soils to the rubber edge. The force deforms/bends the rubber edge to the outside. Then, water leakage occurs. As such, support to prevent the outer deformation of the rubber edge was necessary. Then, Teflon and steel holder support the rubber edge. However, the distance between the holder and the bottom of the upper ring was very close. Therefore, rubber edge height is gradually decreased by wear during rotation. The direct contact between the stainless holder and stainless upper ring may damage the bottom of the upper ring. Teflon holder is installed between the rubber edge and stainless-steel holder to prevent this.

Figure 17 shows the latest rubber edge water sealing structure. The Teflon holder in the ICL-2 was replaced by the Teflon O-ring in the inner ring. It protects the direct contact between the stainless-steel holder and the upper ring. The rubber edge is supported directly by the stainless holder. It is more effective to protect the rubber edge deformation by strong lateral pressure caused by 3000 kPa normal stress.

Fig. 17

ICL-2 CM. This is the latest version of sealing. To protect the damage contacting of the bottom of upper shear box to the top of the lower shear box, a Teflon O-ring (red circle) was placed in the inside part around the central axis. Then, the metal holder (without the Teflon holder) directly supports the back side of rubber edge which is pushed from sample

3.3 Development of an Integrated Landslide-Simulation Model for Rapid Landslides (LS-RAPID)

LS-RAPID is an integrated simulation model capable of capturing the entire landslide process starting from a state of stability to landslide initiation and movement to the mass deposition. This section provides an overview of the use of LS-RAPID to simulate landslide case histories around the world, provides a manual for readers to begin using the program for simulations, and describes the use of the program for several models. Detailed explanation and how to use LS-RAPID with video tutorials is published as LS-RAPID Manual with Video Tutorials in the category of Teaching Tools of Vol.1, No.1 of “Progress in Landslide Research and Technology (P-LRT)”.

Shear resistance mobilized during motion is not as simple as static slope stability analysis, because it will be mostly determined by pore-water pressure generation in the shear zone during shearing. The undrained dynamic-loading ring-shear test can simulate various cases of landslide initiation and motion, and measure the pore-water pressure generation and resulting shear resistance mobilized at the sliding surface. However, a hazard assessment needs the areal distribution of a landslide mass. The areal distribution of the landslide mass is estimated though numerical simulation by inputting the key measured or estimated parameters obtained from the undrained dynamic-loading ring-shear apparatus.

Sassa (1988) initially proposed a numerical simulation model for the motion of landslides. However, the landslide dynamic parameters were not directly measured by the ring-shear apparatus. Sassa et al. (2010) proposed a model simulating both landslide initiation and motion within the same model using the landslide dynamics parameters measured by the well-developed undrained dynamic-loading ring-shear apparatus. A computer programming company has modified it to include user-friendly input and output systems. This version of the model is called LS-RAPID and is commercially available to any user.

The basic concept of this simulation is explained using Fig. 18. A vertical imaginary column is considered within a moving landslide mass. The forces acting on the column are (1) self-weight of column (W), (2) Seismic forces (vertical seismic force Fv, horizontal x–y direction seismic forces Fx and Fy), (3) lateral pressure acting on the side walls (P), (4) shear resistance acting on the bottom (R), (5) the normal stress acting on the bottom (N) given from the stable ground as a reaction of normal component of the self-weight, (6) pore pressure acting on the bottom (U).

Fig. 18

Concept of LS-RAPID

The landslide mass (m) will be accelerated by an acceleration (a) given by the sum of these forces: driving force (Self-weight + Seismic forces) + lateral pressure + shear resistance. The relation is expressed by Eq. (1). Here, R includes the effects of forces of N and U, and works in the upward direction of the maximum slope line before the motion and in the opposite direction of landslide movement during the motion.

$$ am=\left(W+ Fv+ Fx+ Fy\right)+\left(\frac{\partial {P}_x}{\partial x}\Delta x+\frac{\partial {P}_y}{\partial y}\Delta y\right)+R $$

(1)

Equation (1) can be expressed into x and y directions (Sassa et al. 2010) as shown in Eqs. (2) and (3). The assumption that the sum of landslide mass that flows into a column (M, N) is the same with the change or increase of height of the soil column, will give the relationship presented in Eq. (4). The used symbols and parameters in Equations are explained in Table 4. Further related explanation is referred to Sassa and Dang (2018).

Table 4 Symbols and parameters of the LS RAPID model equations

$$ \frac{\partial M}{\partial t}+\frac{\partial }{\partial x}\left({u}_oM\right)+\frac{\partial }{\partial y}\left({v}_oM\right)={\displaystyle \begin{array}{l}g\kern0.5em h\left\{\frac{\tan \alpha }{q+1}\left(1+ Kv\right)+ Kx\kern0.5em {\cos}^2\alpha \right\}\\ {}-\kern2px \left(1+ Kv\right)k\kern0.5em g\kern0.5em h\frac{\partial h}{\partial x}\\ {}-\kern2px \frac{g}{{\left(q+1\right)}^{1/2}}\\ {}\cdot \frac{u_o}{{\left({u}_o^2+{v}_o^2+{w}_o^2\right)}^{1/2}}\left\{{h}_c\left(q+1\right)+\left(1-{r}_u\right)h\kern0.5em \tan {\varphi}_a\right\}\end{array}} $$

(2)

$$ \frac{\partial N}{\partial t}+\frac{\partial }{\partial x}\left({u}_oN\right)+\frac{\partial }{\partial y}\left({v}_oN\right)={\displaystyle \begin{array}{l}g\kern0.5em h\left\{\frac{\tan \beta }{q+1}\left(1+ Kv\right)+ Ky\kern0.5em {\cos}^2\beta \right\}\\ {}-\kern2px \left(1+ Kv\right)k\kern0.5em g\kern0.5em h\frac{\partial h}{\partial y}\\ {}-\kern2px \frac{g}{{\left(q+1\right)}^{1/2}}\\ {}\cdot \frac{v_o}{{\left({u}_o^2+{v}_o^2+{w}_o^2\right)}^{1/2}}\left\{{h}_c\left(q+1\right)+\left(1-{r}_u\right)h\kern0.5em \tan {\varphi}_a\right\}\end{array}} $$

(3)

$$ \frac{\partial h}{\partial t}+\frac{\partial M}{\partial x}+\frac{\partial N}{\partial y}=0 $$

(4)

Landslide initiation is a matter of the stability analysis and the landslide motion is a matter of dynamics at a certain friction during motion. The transient state between the stability and the dynamics is most important. The initiation, the transient state, and the steady state of motion of landslides are illustrated in Fig. 19 in the stress path.

Fig. 19

Stress path from the landslide initiation to the steady state motion

I: Initial state on the sliding surface. Rainfalls move the stress point by the increasing pore-water pressure; earthquakes move the stress point by additional seismic stress. When the stress reaches the failure line (FLP) at the peak strength state (point F), shear failure will occur. When the shearing causes the volume reduction of the soil layer, excess pore-water pressure will be generated. Then, the stress point moves to the failure line during motion. Thereafter, the stress path goes down and reach the steady state (point SS).

Figure 20 shows a series of undrained loading ring shear test results on the Osaka formation (deposits of weathered granite in Osaka area) where a rapid landslide occurred in Nikawa area of the Nishinomiya city and 34 people were killed by the 1995 Hyogo-ken Nanbu Earthquake. Seven undrained speed control tests were conducted at different normal stresses. Granit is composed of three different minerals. Granitic sands are easy to be crushed due to shearing. Stress paths of seven different tests converge to a steady state shear strength at a certain normal stress. It means that grain crushing was continued until a certain normal stress, namely normal stress at the steady state (σss). The mobilized shear stress is the steady-state shear-stress (strength) (τss). We tested many samples other than Osaka formation. Some soils at greater normal stresses can be crushed out during shearing and may not reach the same steady-state stress mobilized at a lower normal stress. In order to investigate this, a least crushable sample (a single mineral hard-sand: Silica sand) and a crushable volcanic sand were tested and described in Sect. 4.

Fig. 20

Test results of undrained speed-control tests of the Osaka Formation sample (Sassa et al. 2014a, b; Okada et al. 2000)

In the case of real landslides, the initial stress is not zero shear stress. Figure 21 shows the cyclic loading test result of the Higashi Takezawa landslide triggered by the 2004 Mid-Niigata Prefecture earthquake.

Fig. 21

Relationship of Shear stress and shear displacement in the saturated undrained shearing. Samples taken from the Higashi-Takezawa landslide triggered by the 2004 Mid-Niigata Prefecture earthquake

Figure 21a presents the stress path, Fig. 21b presents the time series data, and Fig. 21c presents the shear stress and shear displacement relationship. Shear stress starts to decrease at DL (lower shear displacement and shear stress reaches the steady state at (DU) modelling the transient state by a linier function. Figure 22 demonstrates another example of shear stress and shear displacement relationship. The transient state can be modeled by a liner relationship between DL (5 mm) and DU (110 mm).

Fig. 22

Test data for the Unzen Landslide soil—modeling of the transient state from peak to the steady state

Materials in Figs. 21 and 22 are different, one for tertiary sand from the Higashi-Takezawa landslide, Niigata (DL = 5 mm and DU = 110 mm), another is volcanic debris in Unzen volcano. Nagasaki (DL = 6 mm, DU = 100 mm). However, the values and shapes are similar.

Modeling of the transient state from peak to Steady state

Based on the result of Figs. 21 and 22, the transient state of sear stress is approximately expressed by the following Eq. (5).

$$ {\uptau}_D={\uptau}_P-\frac{\mathit{\log}D-\mathit{\log} DL}{\mathit{\log} DU-\mathit{\log} DL}\cdotp \left({\uptau}_p-{\uptau}_{SS}\right) $$

(5)

The concept of Eq. (5) is that the shear resistance (τD) at shear displacement between DL and DU is estimated by shear resistance at peak (τp) and shear resistance at steady state (τss). This is the practical and simple method.

In the past, Sassa et al. (2010) and Sassa and Dang (2018) expressed the shear resistance in the transient stage using tan φa, cohesion c and pore pressure ratio ru in the Eq. (6).

$$ DL\le D\le DU: $$

$$ \mathit{\tan}{\varphi}_a=\mathit{\tan}{\varphi}_p-\frac{{\mathit{\log}}_D-{\mathit{\log}}_{DL}}{{\mathit{\log}}_{DU}-{\mathit{\log}}_{DL}}\times \left(\mathit{\tan}{\varphi}_p-\mathit{\tan}{\varphi}_{a(ss)}\right) $$

$$ c={c}_p\left(1-\frac{{\mathit{\log}}_D-{\mathit{\log}}_{DL}}{{\mathit{\log}}_{DU}-{\mathit{\log}}_{DL}}\right) $$

$$ {\gamma}_u={\gamma}_u\frac{{\mathit{\log}}_{DU}-{\mathit{\log}}_D}{{\mathit{\log}}_{DU}-{\mathit{\log}}_{DL}} $$

(6)

However, τss is the steady state shear resistance obtained from the undrained ring shear test. It is not calculated by friction angle, cohesion and pore pressure ratio. Grain shape and grain size distribution can be dramatically changed by grain crushing and pore water pressure can also change. The shear stress in the transient state is not possible to be calculated from c, φ. It is reasonable to decide it based on the experiment using Eq. (5). Then, the programme was changed from (6) to (5). Concept is different, but no practical change occurred. Then, we have decided to use the relationship of Eq. (5).

Landslide depth and the apparent friction angle at the steady state

In the case of undrained steady state, shear resistance (τss) is almost constant, the excess pore pressure (Us) in deep landslides (A) is high, whereas in the medium-depth landslides (B), the excess pore pressure is moderate. Very shallow landslides (C) do not generate any excess pore pressure. The landside mobility corresponds to the apparent friction angle (φa). Here, tan (φa) = steady state shear strength/initial normal stress. Thus, in the same soil, deeper landslides show great mobility. The medium landslides show moderate mobility. Figure 23 illustrates the landslide depth and the apparent friction angle at the steady state.

Fig. 23

Landslide depth and the apparent friction angle at the steady state

Pore pressure ratio Bss

One of key parameters of the mobility used in LS-RAPID is pore pressure ratio Bss.

Figure 24 presents the relationship between the pore pressure rate (Bss) and the steady state shear resistance:

Fig. 24

Pore pressure rate (Bss) and the steady state shear resistance

Fig. 25

Effect of saturation on the pore pressure parameter B-Value (Ratio of the generated pore pressure for confining pressure increment by the undrained triaxial test (Sassa 1988)

Fully saturated state Bss = 1.0 Dry state Bss = 0

Partially saturated state Bss = between 0 and 1.0

Pore pressure rate Bss is B value at the steady state. It is similar to the pore pressure parameter B value.

B-value was measured for different degrees of saturation for the 1984 Ontake debris flow study. Figure 25 presents the relationship between the B-Value and the Degree of Saturation using the Triaxial test results. The pore-pressure parameter B is 0.2 for 95% degree of saturation, 0.5 for 96–97% degree of saturation. Bss = 1.0 for the test result of fully saturated ring shear test (B = 0.95 or more). As seen in Fig. 25, B value higher than 0.95 could not be produced. Bss = 1.0 is around B = 0.95. During the grain crushing, water volume will not change, but void in the sand structure in the shear zone will decrease. So, the degree of saturation in the shear zone at the steady state may be increased from the initial stage before the motion.

3.3.1 Video of the Initiation and Motion of Rain-Induced Landslides and Earthquake-Induced Landslides in LS-RAPID

Figure 26 shows LS-RAPID rain-induced landslide on a simple slope. The used rainfall record is 2016 Aranayake landslide in Sri Lanka (front cover of Vol.1, No.1 of P-LRT).

Fig. 26

LS-RAPID rain-induced landslide on a simple slope. Video Link: https://us06web.zoom.us/rec/play/Vl90Nvu9TPzmwIerczFBkGBWsewvKxJX_f_tnSFfpnZPyH5LAE9NmIkVVKsAs5RIjKDhr3xAMnwyGnWy.9ONvy5uiiohjS7qq?continueMode=true

Figure 27 shows LS-RAPID for an earthquake-induced landslides on a simple slope. The used earthquake record is 2008 Iwate-Miyagi Earthquake Record.

Fig. 27

LS-RAPID for an earthquake induced landslides on a Simple slope. Video Link: https://us06web.zoom.us/rec/share/Vjdll3lzRLHWykSmUKNJI5qsh0o90uEdRtJTeohG5X7sL1R7I0nM2RjdSAQHSOjA._40ADQ-lCyhgOimq

The upper figure of Fig. 26 LS-RAPID rain-induced landslide presents the instant of initial local failure at the top of a slope which is circled by a red color line (Maximum velocity at this local failure = 0.2 m/s. The whole landslide area is shown in the blue color line. The lower figure shows the landslide mass after deposition (velocity = 0 m/s).

Clicking the video link, you can see the whole movement from the initiation to motion and the deposition.

The upper figure of Fig. 27 LS-RAPID for an earthquake induced landslides presents the state when the initial landslide mass has been created by the progress of local failures such as the upper figure of Fig. 26.

Both cases can be seen by video. LS-RAPID can reproduce a local failure at a weak point in the slope. When a part moves more than DL (usually a few mm), a failure occurs. After DL, the shear resistance is reduced to the steady state shear strength. When shear resistance is reduced in a part, the adjacent elements may move more than DL. Another failure will occur. Such a progressive failure can be seen. Then, the whole landslide body is formed and starts to move as a landslide body onto the downslope. In case where the downslope layer or the layer in the flat area is saturated, the undrained loading causes the pore pressure generation within the saturated soil layer. The landslide mass moves together with a scraped soils above the shear surface. However, when downslope layer is dry or less saturated, the shear surface will be formed within the moving landslide mass which moves leaving a soil mass below the shear surface and stop shortly. Figure 28 illustrates three cases for the initiated landslide mass moving over the downslope.

Fig. 28

Three cases for the initiated landslide movement over the downslope (Sassa 1988)

3.3.2 Three Cases for the Initiated Landslide Movement over the Downslope

Sassa (1988) proposed three cases for downslope movement of a landslide mass as seen in Fig. 28.

Case A: It is the case where a landslide mass moves on a concrete channel or a bed rock. Pore-water within the landslide mass does not dissipate. The movement will continue without leaving soil or scraping soils.

Case B: It is the case where a landslide moves on a saturated soil layer (such as torrent deposit or alluvial deposit seen in Fig. 6—Concept of Landslide Ring-shear Simulator). Pore pressure is generated due to the undrained loading within the saturated deposit. In this case, a sliding surface is created within the soil layer (around ground water surface). Then, the soil layer above the sliding surface is included in the landslide mass. So, the landslide mass is increased during motion. Sometimes the landslide mass around the top of torrent is increased to 5–10 times of initial landslide mass.

Case C: It is the case which a landslide mass moves on less saturated or dry soil layer. In this case, pore water of the moving landslide mass dissipates to the dry or less saturated ground. The shear surface is formed within the moving landslide mass. Then, the landslide mass moves downslope leaving a soil mass below the sliding surface, and gradually terminates its motion and stabilize.

Bss is a very important parameter to control the motion of landslides. In the relation of Fig. 24 Pore pressure rate (Bss) and the steady state shear resistance, the value of the steady state shear resistance τ_ss are much affected. Bss = 0, τ_ss is close to the shear strength at the failure line during motion.

4 Original Test Results for This Article

During the course of writing the ICL landslide lesson on sliding-surface liquefaction (SSL) and undrained steady-state shear strength (USS), we conducted original tests focusing on SSL and USS for one standard sand (silica sand No.6) and a natural sand causing the sliding-surface liquefaction.

As an example of natural sands, we went to the Unzen volcano and took a sample from the 1792 Unzen landslide site (around 15,000 fatalities). The 2006 Leyte landslide (around 1000 fatalities) occurred in a volcanic deposit. The grain size distributions of both samples are shown in Fig. 29. Silica sand is a hard quartz sand. The Unzen sand is volcanic andesitic fragile sand.

Fig. 29

Grain size Distribution of Silica sands and the Unzen sands

4.1 Experiment of Standard Sand “Silica Sand No.6”

Sliding-surface liquefaction occurs only in the undrained condition. All measured parameters are affected by pore pressure values measured in the undrained state. Under the condition of no pore-water pressure, namely in the drained state, we estimated the failure line during motion and measured the friction angle during motion of both samples.

Figure 30 presents the results of a test measuring the failure line in the drained state and the friction angle during motion. Firstly, normal stress of 1000 kPa was applied. Then, the sample was sheared at a constant speed (1 mm/s). The shear stress reached the failure line during motion. Then, normal stress was decreased to zero at the rate of 0.2 kPa/s for 5000 s.

Fig. 30

Results of the saturated constant-speed drained shear test. Sample: Silica sand, Shear speed: 1 mm/s, Normal stress decreasing rate: 0.2 kPa/s. (a) Time series data. (b) Stress path data

Shear stress reached the failure line during motion due to the initial 1000 s of shearing, namely 1000 mm shearing.

Then, normal stress decreased at a rate of 0.2 kPa/s. The time series data of normal stress, shear stress, shear displacement, and pore water pressure (0) are shown in the left figure, and the stress path from the initial stress point (normal stress = 1000 kPa, and shear stress = 0) to the failure line during motion, then, gradually decreasing to zero stress. The friction angle during motion was 35.2 degrees.

Three undrained speed-control tests and two undrained stress-control tests were conducted on the Silica sand as presented in Table 5.

Table 5 A series of tests on Silica sand

Figure 31 shows the test result of Test number 1 of silica sand. After saturation and consolidation at 250 kPa, a constant speed of shearing (1 mm/s) started in the undrained state. Figure 31a presents the stress path. The black line represents the stress path of the total stress. The stress point moved up until failure, then, went down to the shear stress at the steady state shear strength.

Fig. 31

Saturated and undrained speed control test for Silica sand No.6. Normal stress = 250 kPa, BD = 0.95, Shear speed = 1 mm/s. (a) Stress path, (b) Time series data, (c) Shear stress–shear displacement relationship

The red line represents effective stress path. When shearing started, pore water pressure was generated with the progress of shear displacement due to volume reduction by gradual deformation of sand structure. Then, the stress point reached the failure line at peak (36.9 degrees). Once dilatancy occurred, namely the sand volume tended to increase, then the stress point moved to the right direction to the failure line at peak. Grain crushing started in the shearing zone. Excess pore pressure was generated and the stress point went down along the failure line during the motion (35.2 degrees). No further grain-crushing state, namely steady state was reached at the shear stress of 40 kPa.

Figure 31b presents the time series data of normal stress (black), shear stress (red), pore pressure measured near the shear zone (blue), and the shear displacement (purple). One can find that pore pressure generation started at the progress of shear displacement and reached its maximum value around 2000–4000 s, namely at 2–4 m shear displacement. Grains of Silica sands are very hard comparing to the Unzen volcanic sands and granitic sands. However, quartz sands were gradually crushed and caused volume reduction, and the effective normal stress working on sand grains were reduced at the progress of grain crushing. Then, grain crushing was terminated and the steady state was reached. Namely, shearing was continued without any changes in stress.

Figure 31c presents the relationship between shear stress and the shear displacement. Shear speed line is added.

This graph presents the peak strength (τp) at which shear stress reduction started, and the steady state shear strength (τss), and the shear displacement at the peak (DL), the shear displacement at the start of the steady state (DU) can be seen.

In the LS-RAPID, the experimental curve from DL to DU is replaced by a line connecting two points at DL and DU. This transient state between the peak failure point and the steady state shear state is very difficult to analyze from theoretical analysis. During the state, many factors of grain size, grain shape, the grain size distribution, and pore water pressure change. The experimental estimation based on the peak strength (τp) and the steady state shear strength (τss), shear displacement (DL and DU) and the current shear displacement (D) in the Eq. (5) are described in Sect. 3.

$$ {\uptau}_D={\uptau}_P-\frac{\mathit{\log}D-\mathit{\log} DL}{\mathit{\log} DU-\mathit{\log} DL}\cdotp \left({\uptau}_p-{\uptau}_{SS}\right) $$

(5)

Figure 32 (Normal stress = 500 kPa), Fig. 33 (Normal stress = 1000 kPa) show the test results of Test number 2 and the results of Test number 3 are presented in a set of three graphs, (a) Stress path, (b) Time series data, (c) Shear stress–shear displacement relationship.

Fig. 32

Saturated and undrained speed control test for Silica sand No.6. Normal stress = 500 kPa, BD = 0.96, Shear speed = 1 mm/s. (a) Stress path, (b) Time series data, (c) Shear stress–shear displacement relationship

Fig. 33

Saturated and undrained speed control test for Silica sand No.6. Normal stress = 1000 kPa, BD = 0.95, Shear speed = 1 mm/s. (a) Stress path, (b) Time series data, (c) Shear stress–shear displacement relationship

Number 1 (250 kPa) and Number 2 (500 kPa) show very similar steady state shear strengths 40 kPa and 46 kPa. DL are 10 mm and 8 mm, respectively. DU are 1300 mm and 1100 mm.

However, Test number 3 (Normal stress = 1000 kPa) showed a higher steady state shear strength (145 kPa) in Fig. 33. Probably all crushable grains crushed in this high normal stress. Figure 20 for Osaka formation presents the same steady state between 100 and 630 kPa. Stress range was less than 1000 kPa.

Figure 34 shows the test result of shear-stress control test under the normal stress (500 kPa). It uses the torque control mode of servo-motor. After saturation and consolidation, shear stress was increased at the speed of 0.5 kPa/s.

Fig. 34

Saturated and undrained stress control test for Silica sand No.6. Normal stress = 500 kPa, BD = 0.95, Shear stress increment = 0.5 kPa/s. (a) Stress path, (b) Time series data, (c) Shear stress–shear displacement relationship

Figure 34a presents the stress path. The black line shows the stress path of the total stress. The stress point moved up until failure, then, went down to the shear stress at the steady state shear strength.

The red line shows the effective stress path. When shear stress was increased at 0.5 kPa/s, the effective stress path almost vertically moved up to the failure line at peak. Pore pressure generation is minimum until failure. After the failure at the failure line at peak (37.8 degrees), pore pressure started to increase due to the progress of grain crushing. Then, the stress path shifted to the failure line during motion, then reached the steady state (τss = 70 kPa). The friction angle of the failure line during motion was 35.2 degrees.

Figure 34b graph showed the time series data of the normal stress (black), the shear stress (red), the pore pressure (blue) and the shear displacement (purple). The shear displacement acceleratedly increased after failure. A rapid increase of the pore pressure was found at the same time of shear displacement increase.

Figure 34c graph presents the relationship between shear stress and the shear displacement. Shear speed curve (it was increased from 0 to around 120 mm/s during the test) is added. This graph presents the peak strength (τp = 375 kPa) at which shear stress reduction started, and the steady state shear strength (τss = 70 kPa), and the shear displacement at the peak (DL = 5 mm), the shear displacement at the start of the steady state (DU = 2300 mm) are well seen.

Figure 35 shows the test result of shear-stress controlled test under the normal stress (1000 kPa). After saturation and consolidation, shear stress was increased at the speed of 0.5 kPa/s.

Fig. 35

Saturated and undrained stress control test for Silica sand No.6. Normal stress = 1000 kPa, BD = 0.94, Shear stress increment = 0.5 kPa/s. (a) Stress path, (b) Time series data, (c) Shear stress–shear displacement relationship

Figure 35a presents the stress path. The black line is the stress path of the total stress. The stress point moved up until failure, then, went down to the shear stress at the steady state shear strength.

The red line shows the effective stress path. When shear stress was increased at 0.5 kPa/s, the effective stress path almost vertically moved up to the failure line at peak. Pore pressure generation is minimum until failure. After the failure at the failure line at peak and during motion (35.8 degrees). Pore pressure started to increase due to the progress of grain crushing, the stress points went down along the failure line and reached the steady state (τss = 65 kPa).

Figure 35b graph showed the time series data of the normal stress (black), the shear stress (red), the pore pressure (blue) and the shear displacement (purple). The shear displacement rapidly increased and reached 10 m. A rapid increase of the pore pressure was found at the same time of shear displacement increase.

Figure 35c graph presents the relationship between shear stress and the shear displacement. Shear speed curve (it was increased from 0 to around 150 mm/s during the test) is added. This graph presents the peak strength (τp = 655 kPa) at which shear stress reduction started, and the steady state shear strength (τss = 65 kPa), and the shear displacement at the peak was DL = 3 mm, the shear displacement at the start of the steady state was DU = 1000 mm.

The results of three shear-speed-control tests and two shear-stress-control tests are plotted in a combined stress path graph in Fig. 36. All stress paths of five tests reached the same failure line during motion at 35.0 degrees.

Fig. 36

Stress path of three speed controlled tests (250 kPa, 500 kPa, 1000 kPa) and a stress controlled test (500 kPa, 1000 kPa) for Silica sand No.6

As explained in Fig. 30, we conducted a saturated constant-speed Drained test to investigate the friction angle during motion. No pore pressure existed through the whole drained test. The failure line during motion was manifested by decreasing normal stress (0.2 kPa/s) at the constant speed of shearing (1 mm/s). The friction angle during motion obtained by this test was 35.2 degrees. The value of undrained shear tests of Fig. 36 was almost the same at 35.0 degrees.

The results of three shear-speed-control tests and two shear-stress-control tests are plotted in a combined shear stress–shear displacement relationship in Fig. 37. All five DL (shear displacement which shear strength reduction started) are 3–8 mm. All four DU are 1000–2000 mm.

Fig. 37

Shear stress–shear displacement of three speed controlled tests (250 kPa, 500 kPa, 1000 kPa) and a stress controlled test (500 kPa, 1000 kPa) for Silica sand No.6

The steady state τss in the speed controlled tests are 40–70 kPa for normal stress of 250–500 kPa in Figs. 31 and 32. However, the steady state τss in the speed controlled tests was 145 kPa for the normal stress of 1000 kPa in Fig. 33. Probably all crushable grains crushed out before reaching the steady state. In contrast, the steady state τss in the stress-controlled test for the normal stress of 1000 kPa was 65 kPa in Fig. 35 which was the same with 250–500 kPa tests. This large difference of both tests is in shear speed. Shear speed in the speed controlled test was 1 mm/s, and the shear speed of the stress controlled test was at maximum of 150 mm/s. The steady-state shear strength can be affected by shear speed. Natural phenomena are under the stress-controlled condition. So, for the reproduction of natural phenomenon, the stress controlled test is suitable. By contrast, constant speed test is suitable to precisely monitor pore-water pressure during motion. Both tests have different suitability.

4.2 Experiment of Natural Soils (Example of Volcanic Debris Taken from Unzen)

We went to the Unzen landslide which caused a very rapid and long-travelling landslides. The moving mass entered into the Ariake sea and caused a big Tsunamis which killed around 5000 people on the opposite bank (Kumamoto Prefecture) and islands around the Ariake sea. Figure 38 presents the total view of the Unzen Landslide. Red arrows show the head scarp of the landslide. Weathered debris were falling from the head scarp and entered into torrents. The torrent deposits are the weathered materials which remained on the slope after the landslide. They are same materials which caused the landslide.

Fig. 38

Sampling of Unzen sands in Shimabara, Japan and the photo of Sample No.1 (Unzen-1). Upper photo: General view of Unzen landslide. Lower photo (Left): Sampling site within the central gulley. Lower photo (right): Close up photo of samples in the field

We went to the central torrent and took samples at the point of No.1 with yellow mark in the Fig. 38. The lower photo shows a part of the torrent. Large boulders are also included in the torrent deposit, but those are not suitable to the ring shear testing. We took volcanic sand deposits near the torrent wall. The lower right photo is the close up of the sample. As a test sample, less than 2 mm grains were used.

Figure 39 presents a test measuring the failure line in the drained state and the friction angle during motion. Firstly, normal stress of 1000 kPa was loaded. Then, the sample was sheared at a constant speed (1 mm/s). The shear stress reached the failure line during motion. Then, normal stress was decreased to zero at the rate of 0.2 kPa/s for 5000 s. The whole test was conducted in the drained condition. No pore water pressure existed in the shear zone. Namely, the failure line during motion was manifested in the drained state. The friction angle during motion of the Unzen sand was 38.6 degrees. It is much higher than the silica sand which is 35.2 degrees.

Fig. 39

Results of the saturated constant-speed drained shear test. Sample: Unzen volcanic sand, Shear speed: 1 mm/s, Normal stress decreasing rate: 0.2 kPa/s. (a) Time series data. (b) Stress path data

Three stress controlled tests and three speed controlled tests as shown in Table 6 were conducted on the Unzen sand to investigate the sliding-surface liquefaction and the undrained steady-state shear resistance.

Table 6 A series of tests on Unzen sand

Figure 40 shows the test result of Test number 1 of the Unzen sand. After saturation and consolidation at 250 kPa, a constant rate of shear stress (0.5 kPa/s) was loaded in the undrained state. Figure 40a presents the stress path. The black line shows the stress path of the total stress. The stress point moved up until failure; then, it vertically went down to the shear stress at the steady state shear strength. The red line shows the effective stress path. When shear stress was increased at the rate of 0.5 kPa/s, pore water pressure was generated and the stress path moved to the left direction. However, the stress point reached the failure line during motion (38.3 degrees), pore pressure decreased due to dilatancy and the stress path moved to the right direction. The peak shear strength was mobilized on the failure line at peak (42.2 degrees). After peak, pore pressure was generated and the stress path moved to the failure line during moion (38.3 degrees). In the progress of shearing, further pore pressure increase continued until the steady state shear strength of τ_ss = 14 kPa.

Fig. 40

Saturated and undrained stress-control test for Unzen sample. Normal stress 250 kPa, B_D = 0.94, alpha = 0.9, Shear stress increment = 0.5 kPa/s. (a) Stress path, (b) Time series data, (c) Shear stress–shear displacement

Figure 40b presents the time series data of normal stress (black), shear stress (red), pore pressure measured near the shear zone (blue), and the shear displacement (purple). One can find that pore pressure generation started at the progress of shear stress and reached a small peak at around 53 kPa, then, once pore pressure was decreased. However, at the same time of the start of the shear displacement, a rapid pore pressure due to grain crushing was generated. Pore pressure value was close to the normal stress. Hence, the typical sliding-surface liquefaction occurred. The shear displacement acceleratedly increased.

Figure 40c presents the relationship between shear stress and the shear displacement. Shear speed line is added.

This graph presents the peak strength (τp = 211 kPa) at DL = 7 mm from which shear stress started to decrease until the steady state shear stress (τss = 14 kPa).

Figure 41 presents the test result of saturated and undrained stress contolled test for Unzen sample. Normal stress was 750 kPa, Shear stress increment was the same at 0.5 kPa/s. The friction angle at peak was 41.8 degrees, the friction angle during motion was 38.0 degrees. The peak shear strength was 473 ka, and the steady state shear strength was 15 kPa. Approximating the stress reduction curve in Log-scale by two lines, the start of steady state motion (DU) was around 80 mm.

Fig. 41

Saturated and undrained stress control test for Unzen sample. Normal stress 750 kPa, BD = 1.0, Shear stress increment = 0.5 kPa/s. (a) Stress path, (b) Time series data, (c) Shear stress–shear displacement

Figure 42 presents the test result of saturated and undrained speed-controlled test for Unzen sample. Normal stress was 250 kPa, Shear speed was 1 mm/s. The friction angle at peak and the friction angle during motion were the same at 36.1 degrees. The peak shear strength was 137 kPa. The steady state shear strength was 18 kPa which was almost the same for the previous three stress controlled tests.

Fig. 42

Saturated and undrained speed control test for Unzen sample. Normal stress 250 kPa, BD = 0.9, Shear speed = 1 mm/s. (a) Stress path, (b) Time series data, (c) Shear stress–shear displacement relationship

The shear stress reduction curve is not smooth. For the range from shear displacement of 100–700 mm, shear stress reduction was minimum. After that, the shear strength reduction rate was the same with that during 8–100 mm shear displacement. Because of this irregular state, the DU (the shear displacement between two lines approximating the shear displacement curve was around 1000 mm).

Figure 43 presents the test result of saturated and undrained speed-contolled test for Unzen sample. Normal stress was 500 kPa, Shear speed was 1 mm/s. The friction angle at peak and the friction angle during motion was the same at 36.3 degrees. The peak shear strength was 214 kPa. The steady state shear strength was 18 kPa which was the same with Fig. 43.

Fig. 43

Saturated and undrained speed control test for Unzen sample. Normal stress 500 kPa, B_D = 0.95, Shear speed = 1 mm/s. (a) Stress path, (b) Time series data, (c) Shear stress–shear displacement relationship

The shear stress reduction curve is also not smooth. For the range from shear displacement of 100–700 mm, shear stress reduction was minimum. After that, the shear strength reduction rate was same with that during 7–100 mm shear displacement. Such irregular shape did not appear in Silica sand. The Unzen volcanic sand seems to be not homogeneous.

Figure 44 presents the test result of saturated and undrained speed-controlled test for Unzen sample. Normal stress was 1000 kPa, Shear speed was 1 mm/s. This sample also did not show the friction angle at peak. The friction angle during motion was 42.0 degrees. The peak shear strength was 502 kPa. The steady state shear strength was 66 kPa which was the largest. However, DU = 100 mm, and the shear stress reduction curve was smooth.

Fig. 44

Saturated and undrained speed control test for Unzen sample. Normal stress 1000 kPa, B_D = 0.99, Shear speed = 1 mm/s. (a) Stress path, (b) Time series data, (c) Shear stress–shear displacement relationship

The test results from Fig. 40 to Fig. 45 are somehow similar and also different. To view the general shear behavior, all six undrained test results are combined in the stress path (Fig. 45) and also in the shear stress–shear displacement relationship (Fig. 46).

Fig. 45

Stress path of three speed control tests (250, 500, 1000 kPa) and two stress control tests (250, 750 kPa) for Unzen sample

Fig. 46

Shear stress–shear displacement of three speed control tests (250, 500, 1000 kPa) and two stress-control tests (250, 750 kPa) for Unzen sample

Figure 45 presents the combined stress paths of all six undrained shear tests; three shear-stress controlled tests and three constant shear-speed controlled tests. The friction angle during motion was 38.7 degrees. This value was almost same with 38.6 degrees in the Fig. 39 which was measured by the drained normal-stress decreasing test independent of any pore water pressure.

Figure 46 presents the combined shear-stress and shear-displacement curves of all five undrained shear tests; three shear-stress control tests and three constant shear-speed controlled tests. The shear displacements DL, which started shear stress reduction, namely at the peak shear strength, were 5–8 mm. The steady state shear strengths were 14–18 kPa (except 1 kPa data). The start of the steady state to be used in LS-RAPID is 80–200 mm. Some test results suggested DU = 400 mm (Figs. 40 and 43). However, DU = 100–200 mm seems to be practical from Fig. 46. It is 10% of Silica sand (DU = 1000–2000 mm). When DU is short, the materials are brittle. A local failure within the slope expands to progressive failure, and create the landslide block. It often becomes rapid landslides.

5 Case Studies of Sliding-Surface Liquefaction from Previous Research

Section 5 of this landslide lesson are devoted to introduce various types of landslides which are rapid landslides affected by sliding-surface liquefaction.

The examples include the five cases including the 2006 Mega-slide in Leyte Philippines which was first application of a new integrated landslide simulation model “LS-RAPID”, the 1792 historical landslide causing the largest landslide disaster and the largest volcanic disaster and the largest landslide-induced tsunami in Japan, large-scale landslides triggered by 2004 Mid-Niigata earthquake, in which the undrained ring shear testing proved that the sliding surface of rapid landslide was formed by the sliding surface liquefaction at the bottom of sand layer, a hypothetical submarine landslide with sampling from 300 m below the sea floor near the Suruga trough (a part of Nankai trough) by an IODP (Integrated Ocean Drilling Progamme) drilling, and large-scale landslides induced by 2011 typhoon “Talas”. These are typical landslides caused by pore-water-increase. Those five cases are introduced from our previous articles published in journal “Landslides” and some others.

• 4.1 2006 Mega-slide in Leyte, Philippines, triggered by a small earthquake after long rainfall.

• 4.2 Large-scale landslides triggered by the 2004 Mid-Niigata earthquake, Japan

• 4.3 Hypothetical submarine landslides in Suruga bay, Japan

• 4.4 Large-scale landslides triggered by the 2011 Typhoon “Talas” in Kii Peninsula, Japan

• 4.5 1792 Historical mega-slide in Unzen, Japan.

5.1 2006 Mega-Slide in Leyte, Philippines, Triggered by a Small Earthquake After Long Rainfall

This landslide and its analysis of the initiation and the motion by the undrained ring-shear testing using the sample from the landslide and the integrated landslide simulation model (LS-RAPID) is introduced from the following paper.

“Sassa K, Nagai O. Solidum R, Yamazaki Y, Ohta H (2010) An integrated model simulating the initiation and motion of earthquake and rain induced rapid landslides and its application to the 2006 Leyte landslide. Landslides 7:219-236”.

5.1.1 Landslide Field Investigation

A rapid and long-traveling landslide occurred in the Leyte Island, Philippines on 17 February 2006 in the southern part of Leyte Island, Philippines. The landslide caused 154 confirmed fatalities and 990 people missing in the debris. The International Consortium on Landslides (ICL) and the Philippine Institute of Volcanology and Seismology (PHIVOLCS) organized investigation with a joint Japanese and Philippine team of 22 scientists and engineers. The team investigated the landslide from the ground and from a chartered helicopter. The landslide volume was 20 million m³ (Catane et al. 2007). Figure 47 presents the location of landslide in Philippines and a front view of the landslide source area (upper photo) and the total landslide moving area (lower photo) taken from a helicopter chattered by the joint investigation team.

Fig. 47

Photo of the 2006 Leyte landslide. Left: The location of the Leyte landslide in Philippines. Photo of the landslide source area. S: Sampling point in the landslide flow mounds which traveled from the source area and deposited in the flat alluvial deposit. Bottom: A widely spread landslide mass which destroyed the whole village including a school, a church and many houses, and residents of around 1000 people

A planer hard rock is outcropped at the left side of the head scarp. Other parts of the slope seem to be weathered volcanoclastic rocks or debris. The landslide mass moved from the slope and deposited on the flat area. Many flow-mounds or hummocky structure were found.

Figure 48 presents the central section of this landslide and the sampling point (landslide flow mound S in the upper photo of Fig. 47). The section of the central line of the landslide was surveyed by a non-mirror total station and a ground-based laser scanner in the field and compared with a SRTM (Shuttle Radar Topography Mission) map before the landslide. The red-color part shows the initial landslide mass while the blue-color part presents the displaced landslide debris after deposition. The length of landslide from the head scar to the toe of the deposition is around 4 km. The inclination connecting the top of the initial landslide and the toe of the displaced landslide deposit is approximately 10°, which indicates the average apparent friction angle mobilized during the whole travel distance. The value is much smaller than the usual friction angle of debris (sandy gravel) of 30–40°. Therefore, it suggests that high excess pore-water pressure was generated during motion, namely the sliding-surface liquefaction.

Fig. 48

Central section of the landslide from the top to the toe (a). The landslide flow mound and the sampling from the bottom of the flow mound (b)

The material of the flow mound is volcanoclastic debris, including sand and gravel. We observed the material in the source area by eye observation from the surface and by hand scoop excavation in the valley-side slope after the landslide. It consisted of volcanoclastic debris or strongly weathered volcanoclastic rocks. It is regarded to be the same material (either disturbed or intact) observed in the flow mound shown in Fig. 48. Therefore, we took a sample of about 100 kg from the base of the flow mound shown in the point “S” in the section of Fig. 47) and the photo of Fig. 48b. The location is in the center of travel course and just below the source area. Then, we transported the material to Japan and subjected it to the undrained dynamic-loading ring-shear test.

5.1.2 Triggering Factors of the Landslide

We examined the triggering factors of the Leyte landslide. A small earthquake occurred near the site at the time of occurrence of the landslide, and the location of the hypocenter of this earthquake was estimated by the US Geological Survey (USGS) and the PHIVOLCS. According to PHIVOLCS, the earthquake occurred at a location (10.30 N, 124.90 E) 22 km west of the source area of the landslide, 6 km deep, with magnitude Ms 2.6, at 10:36 hours on 17 February 2006. The seismic record at Maasin is shown in Fig. 49b. E-W component (blue color) was the major direction of seismic shaking, other two components are around a half magnitude. Using the standard attenuation function between peak ground acceleration and hypocentral distance (Fukushima and Tanaka 1990), the peak ground acceleration at the landslide site was estimated at 10 gal for this magnitude. We then estimated the expected peak acceleration at the bottom of the landslide mass as about 60–200 gal from the following consideration. (1) Three to five times amplification of ground accelerations at the sliding surface due to the difference in the compression (P) wave speeds between soft volcanoclastic debris (Vp = 0.5–1.5 km/s) and hard volcanic bed rock (Vp = 2.5–5 km/s) that outcropped in the head scarp, because the amplification level is proportional to the velocity contrast between two layers. Though the shear (S) wave speeds of the volcanoclastic debris and the bed rock are unclear, similar level of velocity contrast to the P wave is expected at the sediment/bed rock interface. (2) An additional magnification of 2–4 times is expected in the landslide site due to the focusing of seismic waves onto the mountain ridge. Namely the total magnification of this site will be 6–20 times.

Fig. 49

Triggering factors of the Leyte landslide. (a) Daily rainfalls and accumulated rainfalls from 8 to 18 February 2006. (b) An earthquake (Ms = 2.6) monitored at Maasin, Leyte, located 22 km from the landslide

Heavy rainfall (459.2 mm for 3 days on 10–12 February and 571.2 mm for 5 days on 8–12 February 2006) occurred in this area before the day of the landslide as shown in Fig. 49a. This rainfall should have increased the ground-water level and pore-water pressure inside the slope. However, the peak ground-water level had likely depleted before the occurrence of the landslide on 17 February because the rainfall on 13–17 February was small (total 99.0 mm for 5 days). We simulated the ground-water level using a tank model that had been developed to simulate the ground-water level in the Zentoku landslide, Japan (Hong et al. 2005), which had a depth and inclination similar to that of the Leyte landslide. When inputting 10 days’ precipitation records at the nearest monitoring station in Otikon (about 7 km west of the landslide) on 8–17 February, the peak ground-water level occurred on 13 February 2006. Because the peak ground-water level had already passed when the landslide occurred on 17 February, we deduced that a small earthquake was the final trigger of the landslide.

5.1.3 Undrained Dynamic-Loading Ring-Shear Test

Based on this consideration, a dynamic-loading ring-shear test on the sample taken from the landslide was conducted as follows:

The sample was set in the shear box (250 mm inside diameter, 350 mm outside diameter) of DPRI-6, and fully saturated (BD = 0.98). The stress acting on the sliding surface of the deepest part (around 120–200 m) is very high. However, because of the capacity of this apparatus: the sliding surface was assumed for the test to be 35 m deep and at an inclination of 25°. The unit weight of the soil was assumed to be 20 kN/m³. In the preliminary test to increase pore-water pressure until failure, the failure line of this material was obtained. It was 39.4° in the friction angle and almost zero cohesion. In the simulation test of a rain- and earthquake-induced landslide, the normal stress corresponding to that of 5 m lower than the critical ground-water level (i.e., further 5 m rise of ground-water level shall trigger the landslide) was first loaded on the sample. Then, the shear stress due to the self-weight of the soil layer was loaded. It is the stress point shown by the white circle in Fig. 50a. Using three components of seismic record observed at Massin (PHIVOLCS, Code number: MSLP, Latitude:10.1340, Longitude: 124.8590, Elevation: 50.0), normal stress and shear stress acting on the shear surface of 35 m deep with 25° inclination on the direction of the Leyte landslide were calculated so that the peak seismic stress may correspond to the range of seismic acceleration of 60–200 gal, but as low as possible to avoid failure before seismic loading.

Fig. 50

Undrained seismic loading ring shear test. (a) Effective stress path. (b) Time series data of normal stress, shear stress, pore pressure and shear displacement. (c) Mobilized shear-stress in the ring shear testing, The maximum = 314 kPa, the minimum = 235 kPa

The test results are presented in Fig. 50a–c. Figure 50a presents the stress path of the test. The effective stress path showed a complicate stress path like a cloud. The stress path reached the failure line repeatedly. Therefore, this small seismic stress failed the soil and gave a repeated shear displacement during the period of stress reaching the failure line. It generated a pore water pressure (blue color line) due to grain crushing and volume reduction, and it was accelerated in progress of shear displacement (green color line). Namely the sliding-surface liquefaction phenomenon occurred. The value reached a very small steady state stress (red color line). This process is presented in the time series data around the failure in Fig. 50b. The mobilized apparent friction coefficient defined by steady-state shear resistance divided by the total normal stress was 0.016 (0.9°). The loaded seismic stress is so small, the control was not easy.

The monitored shear resistance is shown in Fig. 50c. When the stress reaches failure line, the balance of shear stress and shear resistance is manifested as acceleration. As seen in the figure, the magnitude of increment of shear resistance seems to be slightly smaller than the decrement. The level of seismic stress is around 40 kPa, the shear stress due to gravity is around 275 kPa. The ratio is 0.145. Namely the seismic coefficient K = 0.145. The estimated magnitude of stress is to load 60–200 gal, namely K = 0.06–0.20. This level will be around the filed condition. It was the minimum possible value to avoid failure before seismic loading in some trial test because of stability of initial shear stress loading. This geotechnical simulation test could physically reproduce the rapid landslide motion triggered by a small seismic shaking in a high pore pressure state due to rain falls. Thus, the combined effect of rainfall and earthquake was confirmed and it was examined in the new integrated computer simulation LS-RAPID.

5.1.4 LS-RAPID Simulation for the Leyte Landslide

The most important parameter is the steady-state shear resistance (τss). The steady-state shear strength is very low as less than 10 kPa in Fig. 50. The testing condition is 100% saturation and the loading stress corresponding to 35 m deep much shallower than the real landslide: 120–200 m). DPRI-6 planned to test normal stress and pore pressure of 3000 kPa. However, the maximum testing normal stress in undrained state was 400–600 kPa. Therefore, testing was conducted below 400 kPa. 3000 kPa test can be conducted by ICL-2 by changing the normal stress loading system from the loading frame type such as triaxial apparatus to a new loading structure without the loading frame which is explained in 3.2.2 and Figs. 9 and 10.

Because the testing condition (35 m) is much shallower than the real landslide and the used sample may be more weathered than that in this deep landslide body, we selected τss = 40 kPa as a practical value for this landslide.

Various combinations of values of factors can be considered. It is not easy, but we assumed the followings: The landslide is deep and the material seems to be intact in the source area as seen in Fig. 47. Then, we estimated that the peak friction and peak cohesion before motion in the source area should be high (tan ϕp = 0.9, cp = 100–300 kPa); the part of head scarp shown in Fig. 47 would be not saturated because it is close to the ridge, probably there is less ground water to generate excess pore water pressure. Then, Bss = 0.1–0.2 was given in this area, and the middle part was probably more saturated (Bss = 0.4–0.6) and the lower part in the paddy field on the flat area was probably well saturated (Bss = 0.9–0.97); The landslide body was stiff in the top, and moderate in the middle and much disturbed in the lower part and on the flat area (lateral pressure ratio k = 0.2–0.7).

Various magnitude of seismic shaking using the wave forms of EW, NS, and UD recorded at Maasin, Leyte were given in addition to pore pressure ratio of 0.15. The border to create a rapid landslide existed between seismic coefficient K_EW = 0.11 and 0.12. Then, we gave K_EW = 0.12. We used the ratio of magnitudes of seismic records of EW, NS and UD, K_NS = K_UD = 0.061. The seismic shaking of three directions of EW, NS and UD were given in this simulation. A 3 m unstable deposit was assumed in the alluvial deposit area. Blue balls shown in Fig. 51a–e are the unstable soil deposits (initial landslide body) in the source area and also unstable deposits in the alluvial flat area. The critical height (Δhcr = 0.5 m) to reduce shear strength from peak to the steady state was given.

Fig. 51

Simulation result of the initiation and the motion of the Leyte landslide by LS-RAPID. 1 Mesh, 40 m; Area, 1960×3760 m; Contour line, 20 m; 3 m unstable deposits in the alluvial deposit area (blue balls)

A series of motion (a–e) was presented in the case of K_EW = 0.12, K_NS = K_UD = 0.061, r_u = 0.15 in Fig. 51.

• a: ru rises to 0.15 and earthquake will start but no motion.

• b: Continued earthquake loading triggers a local failure within the source area as presented in red color mesh

• c: An entire landslide block (red color block) is formed and moving

• d: The top of landslide mass goes on to alluvial deposits

• e: Termination of the landslide motion

The air photo taken from the chartered helicopter and the simulation results presented in 3D view from a similar angle are presented in Fig. 52. The travel distance and the major part of landslide distribution were well reproduced by LS-RAPID using landslide dynamics parameters which were measured by the undrained dynamic-loading ring-shear apparatus.

Fig. 52

Landslide moving area (photo-left and simulation result-right)

5.2 Large-Scale Landslides Triggered by the 2004 Mid-Niigata Earthquake, Japan

This landslide and its analysis of the initiation and the motion by the undrained ring-shear testing using the sample from the landslide and the integrated landslide simulation model (LS-RAPID) is introduced from the following paper.

This case study is a typical earthquake induced landslides. The sliding surface of this landslide was located between the sandy layer and a silt layer below it. Dynamic-loading ring-shear tests were conducted on soils taken from both sand layer and silt layer. Initially people thought that the sliding surface was formed in the outcropped silt layer after the landslide. However, the undrained cyclic and seismic loading tests revealed that silt layer is very strong against earthquake loading. The sliding surface was formed within the sand layer where the sliding-surface liquefaction occurred. This case study is introduced from the following paper.

Sassa K, Fukuoka H, Wang f, Wang G (2005) Dynamic properties of earthquake-induced large-scale rapid landslides within past landslide masses. Landslides 2:125-134

5.2.1 Outline of the Higashi-Takezawa Landslide

The 2004 Mid-Niigata Prefecture earthquake (M6.8) triggered 362 landslides more than 50 m wide, and 12 large-scale landslides of more than 1 million cubic meters, and the total volume of all landslides was about 100 million cubic meters. The area is the Tertiary landslide area in Niigata Prefecture. So many landslides exist in the Tertiary weathered mudstone (silt) area (Sato et al. (2004)). The behavior of the Tertiary landslides through the snowy period in Niigata Prefecture, Japan was reported by Niigata prefecture. https://www.jstage.jst.go.jp/article/jls2003/41/1/41_1_37/_pdf/-char/en

Those landslides were triggered by the snow melting water in the snowy area. The permeability of silt was very low. The slow and long-term water supply caused the landslide movement. By contrast, short-term strong rainfalls caused shallow landslide and debris flow.

Landslides triggered by the Mid-Niigata Prefecture were reactivation of past large-scale landslide masses. Two major landslides of the Higashi Takezawa landslide and its neighboring Terano landslide occurred within the previous landslide mass. As explained in this section, the sliding surface of this earthquake-induced-landslide was formed in the sand layer.

Figure 53 presents the view of the Higashi-Takezawa landslide. It created the largest landslide dam that has posed the risk of debris flow disaster by dam failure. The head scarp of the current landslide is shown on the figure by red arrows, and the head scarp of the previous landslide is shown by a curved line with red arrows. The building in the foreground was an elementary school. This landslide mass rapidly moved around 100 m, and hit the opposite bank of Imokawa River. A part of landslide mass spread cross the road and hit the school. We investigated the head scarp of this landslide. A very straight, gently-dipping (around 20 degrees) stiff silt (stone) layer outcropped as shown in Fig. 53. This layer seemed to be a part of the sliding surface of this landslide. It is relatively impermeable and groundwater flowed over this stiff silt layer. The sand layer over this stiff silt layer was soft. It was probably a part of previously moved landslide mass.

Fig. 53

Photos of the Higashi-Takezawa landslide triggered by 2004 Niigata-ken Chuetsu Earthquake. (a) A Photo of the general view of the landslide taken from the chartered helicopter (K. Sassa). The earthquake-induced landslide occurred within the previous landslide. (b) Photo of the exposed silt stone layer under the sand layer in the source area. The slope is gentle. And the silt-stone layer was hard to take sample. The undrained ring shear test was conducted on the silt taken from the nearby Terano landslide

The central section of the Higashi-Takezawa landslide is presented in Fig. 54a. The black line presents the ground surface before the landslide, and the green line shows the ground surface after the landslide. The mass A (red dots) and the mass B (black dots) are the unstable mass before the current landslide movement.

Fig. 54

Central sections of the Higashi Takezawa landslide (a) and the Terano landslide (b)

The initial average slope angle between P1 and P2 was around 14.6 degrees, and the average slope angle between P1 and P3 was around 13.5 degrees. The mobilized energy line between P1 and P4 (highest point) was around 7.5 degrees. Hence, this rapid landslide occurred in a gentle slope less than 15 degrees, and the mobilized average apparent friction angle during motion was 7.5 degrees. The large difference between the energy line and the center of gravity of the moving mass suggests a rapid motion. Block A initially started to move due to seismic loading and resulted in excess pore pressure generation. Thus, the movement of block A imposed an undrained loading to block B. This undrained loading effect can be greater than earthquake loading for the Block B. Then, the block B started to move with the block A and crossed the Imokawa River and hit the opposite bank. Mud found on the wall of elementary school was jumped from the landslide mass due to the strong impact between the rapid landslide mass and the river embankment. Figure 54b shows the section of the Terano landslide.

The main body of the landslide block at the toe was composed of sands (T1) that seem to be similar to the samples of H1 and H2 of the Higashi-Takezawa landslide. The head scarp of this landslide was also investigated, and in contrast to the Higashi Takezawa landslide, the silt sample (T2) taken from the head scarp was well weathered and soft. The head scarp and sample T2 was inside the previously moved landslide mass as suggested by the section of Fig. 54b.

The grain size distributions of samples T1, T2 and H2 are shown in Fig. 55. Samples H2 and T1 are rather similar though sample T1 includes a slightly greater portion of finer grains. While silt sample T2 is quite different from samples T1 and H2, it is much finer than those sands.

Fig. 55

Grain size distribution of the tested soils

Using the samples from the Higashi-Takezawa landslides (H2) and from the Terano landslide (Terano sand T1 and Terano silt T2), 1) a seismic-loading test was conducted on the H2 sample using the monitored earthquake record at the monitoring site NIG019 in K-NET (NIED), and 2) cyclic loading tests were conducted on the Terano sand and the Terano silt to investigate their undrained dynamic properties.

5.2.2 Seismic Loading Test on Sands from the Higashi-Takezawa Landslide

Seismic loading test was performed to simulate the initiation of the Higashi Takezawa landslide using the real monitoring record of the Mid-Niigata earthquake in the closest site. The monitored record was obtained from one observation station of K-NET (Kyoshi Net), which is the strong earthquake motion monitoring network by the National Research Institute for Earth Science and Disaster Prevention (NIED). K-NET is a system which sends strong-motion data on the Internet. The data are obtained from 1000 observatories (25 km mesh) deployed all over Japan. The nearest monitoring site to the Higashi Takezawa landslide is the observation station of NIG019 at Ojiya, around 10 km west of the Higashi Takezawa landslide, and WNW 7 km from the epicenter of the main shock. The Higashi Takezawa landslide is ENE 3.6 km from the Epicenter. The location of the epicenter is between the NIG019 and the Higashi Takezawa landslides, but closer to the landslide.

The real acceleration and its wave form that affected the Higashi Takezawa landslide cannot be known in detail as it is complex due to inter alia topography, geology, underground structure, and distance from the fault and the epicenter. Therefore, the monitored earthquake record in NIG019 was applied as the first approximation of input acceleration to the sliding surface. Using three components of acceleration records, normal stress component and shear stress component on the bottom of landslide were calculated based on the section of the Higashi Takezawa landslide.

Figure 56 presents the result of the undrained real-earthquake-wave loading test on the Higashi-Takezawa sand (H2).

Fig. 56

Undrained real-earthquake-wave loading test on the Higashi-Takezawa sand (H2) (BD = 0.98). (a) Time series data, and (b) Stress path

The result of seismic loading of the sample was presented as the time series monitoring result of loaded normal stress, mobilized shear resistance (not the same with the applied shear stress because failure occurred in the sliding surface at the failure line), generated pore pressure and the resulting shear displacement as seen in Fig. 56a. Due to seismic shaking, pore water pressure was generated, then shear failure occurred and shear displacement started. As shear displacement progressed, a typical sliding surface liquefaction phenomenon occurred, namely grain crushing along the shear zone proceeded and a higher pore pressure was generated with progress of shear displacement. The sand was originally a marine deposit and much stronger than volcanic deposits such as pyroclastic flow deposits and pumice, and also weathered granitic sands. However, the shearing stress under 40 m overburden pressure was high enough to cause grain crushing and resulting volume reduction. It caused sliding surface liquefaction. The mobilized apparent friction angle was only 2.5 degrees at the steady state.

The stress path of the test is shown in Fig. 56b (in which a red color stress path shows the effective stress path and a green color stress path presents the total stress path). The apparent friction angle was obtained from the ratio of mobilized steady state shear resistance divided by the initial normal stress. The effective friction angle mobilized in the post failure process is not so clear in Fig. 56b because seismic stress was reproduced by the undrained ring-shear test, but pore water pressure monitoring was delayed due to the permeability of soils. Therefore, instead of real seismic loading, cyclic loading test (1 Hz: 1 cycle/s) was conducted on the same sample. In this test, normal stress was kept constant. The effect of the loaded normal stress was cancelled by the generated pore water pressure at the undrained state. The cyclic shear stress was loaded. The test result is presented in Fig. 57. In this 1 Hz test, pore-pressure monitoring can follow the shear stress change. 1 Hz cyclic shear stress can be seen in Fig. 57a time series data and also in Fig. 57b stress path data. From the stress path data, we can know that the friction angle during motion of this sample was 36.9 degrees.

Fig. 57

Undrained cyclic loading test of the Higashi Takezawa sand (H2) (B_D = 0.98). (a) Time series data, and (b) Stress path

5.2.3 Cyclic Loading Tests for the Sand (T1) and Silt (T2) of the Terano Landslide

In order to investigate the landside dynamics behavior of the sand layer and the silt layer, we took sample from the sand layer (T1) and the silt layer (T2) in the same Terano landslides. Sample position is shown in Fig. 54 Central sections of the Terano landslide (b). The grain size distributions of both samples are shown in Fig. 55.

Test results are presented in Fig. 58 for the sand sample (T1) and in Fig. 59 for the silt sample.

Fig. 58

Undrained cyclic loading test on the Terano sand (T1) (BD = 0.98). (a) Time series data, and (b) Stress path

Fig. 59

Undrained cyclic loading test on the Terano silt (T2) (B_D = 0.98) (a) Time series data, and (b) Stress path

During the test, normal stress was kept constant, and shear stress of sine curve of 1 Hz loading frequency was applied. The shear stress was increased step by step until 15 cycles. All test conditions were the same for T1 sand and T2 silt.

The peak friction angle of the sand sample was 38.4 degrees, and the friction angle during motion was 35.7 degrees.

In the test of T2, no pore pressure was generated through the test. The stress path only vertically moved. The peak friction angle was 33.3 degrees.

The shear displacement in the sand acceleratedly increased after failure and even after the termination of cyclic loading. The apparent friction angle in the steady state was 3.8 degrees. By contrast, the shear displacement increased step by step during the cyclic loading test, but the movement stopped after cyclic loading. Shear resistances before loading and after loading were the same (around 100 kPa).

From these two tests, the following is estimated.

In the snowy period, pore pressure gradually increased. When the stress path reached the peak failure line of 33.3 degrees, landslide motion was initiated. The peak failure line of the sand layer was 38.4 degrees. This layer was stable during those pore pressure increases. In the tertiary mudstone area, sand layer and silt layers alternated.

During the earthquake loading, sand layer generated excess pore pressure due to grain crushing by seismic shaking. However, no excess pore pressure was generated in the silty layer. Then, the layer was stable during the earthquake.

Figure 54 presents the central section of the Higashi-Takezawa landslide and the Terano landslides. Landslide debris (mass) produced by snow-melting or rain triggered landslides moved by earthquake loading. The sliding surface was formed possibly in the border of sand and silt as seen in Fig. 53 (lower photo). But from a mechanical point of view, the sliding surface of this earthquake-induced landslides was formed in the sand layer (such as the bottom of the sand layer), and the sliding surface of the water-induced landslide was formed in the silt layer (such as the top of the silt layer).

5.3 Findings by the Investigation of Two Landslides

• 1. Two rapid major landslides triggered by the 2004 Mid-Niigata earthquake were investigated. Both created landslide dams on the Imokawa River.

• 2. Cyclic loading ring shear test on sands from the Higashi Takezawa landslide and the Terano landslide proved that both sands can be liquefied by the sliding surface liquefaction under 20–40 m over burden pressure. Since silt taken from the Terano landslide proved that the silt was not subjected to excess-pore-water-pressure generation by seismic/cyclic loading, and the shear displacement was only limited when the shear stress reached or exceeded the failure line, the movement stopped at the end of cyclic loading.

• 3. A dynamic loading ring shear test to simulate the HigashiTakezawa landslide using the monitored seismic record in Ojiya (NIG019) on the sand collected from the head scarp of the landslide reproduced the initiation of rapid landslide resulting from the sliding-surface liquefaction. The mobilized friction angle at the steady state was only 2.5 degrees.

5.4 Hypothetical Submarine Landslides in Suruga Bay, Japan

The 2011 off the Pacific coast of Tohoku Earthquake so called the 2011 Tohoku earthquake (Mw 9.0) occurred. The magnitude of this earthquake was Mw 9.0; it was the largest in the Japanese earthquake monitoring history. This earthquake induced an enormous tsunami which caused major disasters along the north-eastern coast of Japan. Another big earthquake was concerned along the plate boundaries such as the Nankai Trough. This section is devoted to possible contribution from landslide scientists to this type of mega-disaster. The paper was published in the following.

• 1) Sassa K, He B, Miyagi T, Strasser M, Konagai K, Ostric M, Setiawan H, Takara K, Nagai O, Yamashiki Y. Tutumi S (2012) A hypothesis of the Senoumi submarine megaslide in Suruga Bay in Japan—based on the undrained dynamic-loading ring shear tests and computer simulation. Landslides 9:439–455

Coastal and submarine landslides can cause tsunami. The submarine landslides in Suruga Bay should have caused big tsunami and also may cause big tsunami in the future. It was presented on the World Tsunami Awareness Day Special Event on 5th November 2021 at the Fifth World Landslide Forum in Kyoto, Japan. It was published in the following paper by Springer Nature in 2021

• 2) Doan Huy Loi, Kyoji Sassa, Khang Dang, and Toyohiko Miyagi (2021) Simulation of Tsunami Waves Induced by Coastal and Submarine Landslides in Japan. Understanding and Reducing Landslide Disaster Risk. Vol.1 (edts: Sassa K, Mikos M, Sassa S, Bobrowsky P, Takara K, Dang K): 295–327.

5.4.1 Area of Investigation

The Nankai Trough (Fig. 60) is characterized by destructive earthquakes that recur along the plate boundary megathrust where the Philippine Sea Plate subducts under SW Japan. The authors examined 250 m mesh bathymetry of the sea floor along the Nankai Trough from Kii Peninsula to Izu Peninsula. Many submarine slope deformation features are seen there. Some are similar scale with the Senoumi feature (shown in Fig. 61). However, the social significance is the highest in the Senoumi case because the head scarp is located at the coast of densely populated Yaizu city which Japanese transportation arteries (the Shinkansen and the Tokyo–Nagoya highway) pass through.

Fig. 60

Location of investigation area and drill sites. Empty triangle: Nankai Trough, yellow square the target area of Suruga Bay, red circles two sampling points (C0018 drilling site of the IODP-Expedition 333 and the Omaezaki hill)

Fig. 61

A hypothetic mega-slide (the area called as Senoumi) in Suruga Bay. A characteristic bathymetric feature called as Senoumi-stone flower sea which is delineated by red arrows. This feature was possibly created by a submarine megaslide. Possible border of the head scarp of landslide, Epicenter of 2009 Suruga Bay earthquake, small submarine landslide induced by this earthquake, and the coast line are found in the left figure. The section of the A-A’ crossing the hypothetic landslide the Suruga trough which is an extension of the Nankai trough

White circles in Fig. 60 mark gullies identified due to their distinctive shape. The Senoumi forms a step on the western side of the Suruga Bay with a gully (Fig. 61). The section A-A′ passing through this gulley is shown in the right-side figure of Fig. 61. The slopes in the steep parts of the head scarp and toe in the section A-A′ are 8–12°. The Senoumi Megaslide hypothesis was made from the following consideration of this feature.

The shape of the Senoumi feature differs from those of most landslides: the exit to the Suruga Trough is very narrow, relative to the width of Senoumi depression. If it would be formed by a blockslide, its mass could not move out through this narrow exit. However, when a large strength reduction occurs within the soil mass after failure, this type of feature can be created (such as quick clay landslide in Norway: Photo is shown in Sassa et al. 2012). This shape is possible to occur when landslide mass is almost liquefied after failure and move/flow downward with a very small shear resistance while the landslide expands upward retrogressively. This area must have been disturbed by tectonic movements and shear bands must develop near the border (Nankai trough-Suruga trough) between the Eurasian Plate and the Philippine plate. Many mega-landslides as well as mega-earthquakes often occur near the borders of plates, because the fractures of rocks and soil layers reduce strength of the layer and the ground water flow through the fractures may provide excess pore water pressure and promote weathering of the layer.

Both sides of Suruga Trough (Omaezaki hill and Izu peninsulas) are composed of the Neogene layers. The base of the Senoumi area is likely formed by Neogene layers. The soil layer of the Senoumi area before landslide can be deposits transported by the Oi River. We sampled and tested Neogene soils taken from Omaezaki hill. The deposits may include Neogene deposits and also the volcanic ash. Then, we tested samples from Omaezaki shown in Fig. 62 and also volcanic ash taken from IODP drilling C0018 in Fig. 60.

Fig. 62

Assumption of fracture zones in Senoumi. The deep-sea floor in the blue color is called as Suruga Trough which is a part of Nankai Trough

Figure 62 presents the location of the Suruga Trough, the extension of Nankai Trough shown as the white line with a line of white triangles. This is the plate border. Because of high-stress, shear bands and fold structures parallel to the Suruga trough such as Izu Peninsula. The shear bands are often created in a set of two (parallel and perpendicular directions) directions. Oi River and the Gully along A-A’ will be one of shear bands perpendicular directions of the Suruga Trough. The shear bands are crushed, therefore, those part are week and also pore-water pressure through shear band system from the higher elevation (Fig. 6 schematic figure of excess pore-pressure in continental shelf, in Sassa et al. 2012). Many similar gullies are found within the continental shelf in Fig. 60.

5.4.2 Sampling from Volcanic Ash by IODP Drilling

The depth of Senomi landslide is between 200 and 600 m. We plan to use a sample taken from a few hundred meters before the sea floor. Such a deep drilling in the sea floor is conducted by the Integrated Ocean Drilling (IODP). And drilled cores of IODP around Japan are stored in the Kochi Institute for Core Sample Research of JAMSTEC (Japan Agency for Marine Earth-Science and Technology). Sassa visited the Kochi Institute and searched the cores. One of the nearest IODP drilling site was C0018. The sampling way from a drilling ship and the drilling core and its geology. The pink color part at the bottom of MTD is volcanic ash. The photo of this volcanic ash layer, its grain size distribution and the microscopic photo are shown in Fig. 63.

Fig. 63

IODP drilling and the core of sample of the volcanic ash in around 189 m in which a sliding surface was formed

This volcanic ash layer correlates to the “Pink” volcanic ash sourced from Kyushu Island, Japan and is dated to 0.99–1.05 Ma (Hayashida et al. 1996). Comparable volcanic ash layer as cored at IODP C0018 drill site are likely to have been deposited also in Suruga Bay, where no deep-drill hole is available yet. We therefore used a sample taken from this fine-grained volcanic ash layer at the base of landslides (MTD) drilled at IODP Site C0018 as analog material for potential Suruga Bay sliding surfaces and tested it by the undrained ring-shear apparatus. ICL-2 which can test 3 MPa was not yet developed. So, we tested this sample by ICL-1 which was developed in 2011 and is able to test until 1 Mega normal stress in the undrained cyclic-loading and seismic-loading condition.

We tested the Omaezaki silt and the Omaezaki sandy silt which were sampled from the Omaezaki hill in Fig. 62. The grain size distribution of three samples is shown in Fig. 64. The Omaezaki silty sand and the IODP volcanic ash are very similar in its grain size.

Fig. 64

Grain size distribution of IODP volcanic ash, Omaezaki silt, and Omaezaki Silty-sand

5.4.3 Undrained Basic Cyclic—Loading Tests

Cyclic-loading tests were conducted on IODP volcanic ash and Omaezaki sand and Omaezaki silt. Frequency is 0.1 Hz. Figure 65 presents the test result of cyclic-loading ring-shear tests of three samples: (a)—IODP volcanic ash, (b)—Omaezaki sand, (c)—Omaezaki silt. As the initial stress, we used the normal stress = 1000 kPa, the initial shear stress = 160 kPa. This is 160/1000 = 0.16. It is the slope angle = 9.09°. This angle corresponds to the slope of the central section of A-A’ of the Senoumi hypothetic landslide (8–12°).

Fig. 65

Result of undrained cyclic loading tests. (a) IODP volcanic ash, (b) Omaezaki Neogene sand, (c) Omaezki Neogene silt

The shear stress was increased step by step in 10 cycles until 400 kPa increment (large enough to cause failure). Shear stress was gradually increased in 10 cycles. Then, a certain cycle of loading was kept constant to confirm pore pressure generation and decreased step by step. This procedure was decided by the preliminary tests to confirm to study shear failure, and post-failure strength reduction due to pore pressure generation until the steady state. When soils are fully saturated, effective normal stress should be constant in the undrained condition. Soils were regarded as in the undrained condition during earthquake loading because loading is very quick to dissipate generated pore pressure. Therefore, only shear stress was loaded, while normal stress was kept constant by means of automatic servo-control system.

The left column figures of Fig. 65 show the time series data. In this figure, the black color line shows normal stress, the red color line presents mobilized shear resistance, the blue line shows the generated pore water pressure, the purple line presents the resulted shear-displacement, and the green line shows the control signal to stress servo-control motor (stress control motor) by pre-decided computer program. The shear stress motor is to load the ordered stress. The control signal and mobilized shear resistance is the same until failure. However, after failure, power provided by servo-control motor is spent for frictional energy consumption and acceleration. The mobilized shear resistance after failure gets smaller than the control signal.

The right column figures show the stress path of total stress (blue line) and the effective stress (red line). The total normal stress is kept constant, while the effective normal stress is shifted by the generated pore water pressure. When the stress path reached the peak failure line, failure occurred. The following findings are obtained from these test results.

• 1. IODP volcanic sand generated a high pore-water pressure during cyclic loading, and strength reduction was the largest. The shear speed was accelerated to the maximum speed of ICL-1 (5.4 cm/s) and the planned 3–5 m shear displacement was reached. Then, shearing was stopped to avoid ablation of rubber edge. The mobilized steady-state shear strength was 72 kPa (shear displacement curve is seen in Fig. 66). Pore pressure ratio reached 0.8 (normal stress = 1000 kPa, pore-water pressure = 800 kPa). The peak friction angel was 36.2°.

• 2. Neogene silty–sand also generated a high pore-water pressure. The shear speed was accelerated after failure, but it decelerated after the stop of cyclic loading. The control-signal corresponding to 160 kPa (shear stress due to the self-weight of soil layer) was close to the steady-state shear resistance. The difference between stress and shear resistance will create acceleration. Therefore, no acceleration occurred after the stop of cyclic loading. The peak friction angle was 39.5°. It is the greatest value in these three soil samples.

• 3. Neogene silt failed at a lower peak shear resistance comparing to two other samples. Pore pressure generation during cyclic loading was small. Shear displacement occurred step by step during cyclic loading after failure, but it stopped in the decreasing cycles at around 600 mm. The peak friction angle was 36.8°.

Fig. 66

Shear stress–shear displacement curves of the cyclic-loading ring shear tests on IODP silty sand (ash), Omaezaki sand, and Omaezaki silt

5.4.4 Seismic-Loading Tests Using the 2011 Tohoku Earthquake Wave Record

We planned to investigate the landslide dynamic characteristics of the IODP sample subjected to the strong earthquake of the 2011 Tohoku Earthquake record by the undrained dynamic loading ring shear apparatus.

Figure 67 shows three components of seismic record of 2011 Tohoku Earthquake. The largest measured earthquake acceleration 2933 gal as the resultant acceleration of EW, NS, and UD components and 2699 gal as a single component were recorded at station MYG004 at Tsukidate in Miyagi Prefecture which is 176 km west from the epicenter. Since mega earthquakes similar to the 2011 Tohoku earthquake occurred in the past and, likely, will occur in the future. The seismic record of NS component of MYG004 was used for the ring shear simulation test as an example of mega earthquake loading.

Fig. 67

2011 Tohoku earthquake record (MYG004)

Examination of loading shear stress

We examined the testing plan of seismic loading ring shear test based on the result of cyclic loading test of Fig. 65a, b. Initial normal stress and shear stress were decided to be the same with those of the cyclic loading test, namely 1000 and 160 kPa, which represented the normal stress and shear stress due to gravity on a slope of 9.09° which corresponds to the slope angle of A-A’ section in Fig. 61. According to the test of Fig. 65, additional 400 kPa of shear stress may fail the sample. So, we used 0.3 times of MYG004 (2699 gal in NS component wave), namely the maximum acceleration 810 gal in this ring shear simulation test. The same value was recorded in FKS009 in Ono of Fukushima Prefecture (217 km from the epicenter).

Examination of loading duration

The test of Fig. 65 was conducted at 0.1 Hz. The servo-shear stress control motor (400 W) of ICL-1 (we did not yet produce the powerful ICL-2) cannot reproduce the high-frequency loading using the recorded data (Fig. 67). Preliminary tests were conducted to investigate the time required to reproduce the seismic wave form of MYG004 by increasing the shaking time by factors of 10, 20, and 30. We found that a 30-fold increase in time scale could reproduce shear stress changes similar to the recorded wave form. The ring shear test was conducted under the undrained condition. Pore pressure value was not affected by time because any of pore pressure dissipation did not occur. The same stress path would be obtained in 30 times longer test with that in the real time test. The comparison of the monotonic (corresponding to 0.0 Hz) undrained shear stress loading test and 0.2, 0.4, and 1.0 Hz cyclic undrained shear stress loading test presented almost the same relationship between stress and shear displacement and also almost the same stress path between the curve of monotonic loading test and the curve connecting peak values of cyclic loading test (Trandafir and Sassa 2005). Then, the test was conducted in the 30 times longer time period.

Results of undrained dynamic-loading ring-shear tests using the 2011 Tohoku earthquake wave form

Figure 68 presents both of seismic loading test results, the left figure for IODP volcanic ash and the right figure for Omaezaki Neogene sand. The IODP ash generated a high pore-water pressure in the first main shock. The large shear displacement started in the second main shock and an accelerating motion was created. Namely, this ring-shear simulation test visualized that a rapid landslide would be triggered on 9.09° slope, while the Neogene sand generated a smaller pore water pressure during the first main shock. A large pore water pressure was generated at the second main shock and a rapid motion was started. But when the level of seismic loading got smaller, the movement was decelerated. It is similar to cyclic loading test of Neogene sand (Fig. 65). The Neogene sand has a greater peak friction angle (39.5°) than that of IODP volcanic ash (36.2°). The Neogene sand did not fail during the first main shock; therefore, the soil mobilized the same value of shear resistance with the loaded shear stress. As seen in Fig. 68b, the green line (control signal) and the red line (mobilized shear resistance) are overlapped until the second main shock. From this ring shear test, slopes with the same or greater angle than 9.09° will fail by 810 gal of earthquake having the same wave form as that of MYG004. Neogene sand will move by this seismic loading, but the movement will stop after seismic loading. The relationship is very similar to the cyclic loading test in Fig. 65.

Fig. 68

Undrained seismic-loading ring-shear test using the 2011 Tohoku earthquake record wave form. Left (a): IODP Volcanic ash; Right (b): Omaezaki Neogene sand

5.4.5 Reproduction of the Movement of the Hypothetical Senoumi Landslide by LS-RAPID

We have prepared Fig. 69 to implement LS-RAPID.

• (a) Firstly, we estimated the ground surface before the landslide

• (b) Current elevation of the ground and the sea floor.

  Red zone was the estimated initial landslide mass caused by pore-pressure supplied through T-shape shear bands (Fig. 62, red colored cross shape part)

• (c) The difference of both elevations, namely landslide mass.

Figure 70 presents LS-RAPID simulation result. Blue colored cells are no motion. Red colored cells are moved parts.

Fig. 69

Reproduction of the topography before the Senoumi landslide for LS-RAPID. (a) Elevation of the ground before the landslide. (b) Current elevation after the landslide. Red zone was the initial landslide mass caused by pore-pressure supplied through T-shape shear bands (Fig. 62). (c) Depth distribution of unstable mass (different between a and b)

Fig. 70

Landslide simulation result for IODP volcanic ash by the 1.0× Tohoku earthquake record (MYG004)

(a) A red color block MJ (major landslide) moved at the fist shock. (b) The initial landslide block (MJ) progressively expanded to the south (downward) and the landslide started to enter into Suruga trough. (c) Flowed out landslide mass moved downward along the Suruga trough. (d) Landslide debris moved out to Suruga trough.

The simulation result for IODP volcanic ash using 1.0×MYG004 seismic record is shown in Fig. 70. Blue balls represent soil columns stable or less than 0.5 m/s moving velocity. Red balls show columns with values greater than 0.5 m/s velocity. Figure 70a is the situation soon after the first shock. The first shock moved the excess pore pressure acting area (major landslide block, MJ) and failed a part of bottom-left red-colored area (minor landslide block, MN). The left figure of Fig. 70a presents the pore pressure ratio (upper graph) and loaded acceleration (lower graph). Figure 70b shows the landslide moving area soon after the second main shock. All columns in the whole simulation area were subjected to movement with the velocity more than 0.5 m/s. Figure 70c is the instant of 400 s. Landslide mass is flowing down along the Suruga trough from north to south (see the vertical white arrow), while the part of bottom of the simulation area is moving to the Suruga trough (see the horizontal white arrow). Figure 70d shows the termination of movement at 4463 s (1.2 h) from the initiation.

Further application of this undrained ring-shear testing on the Senoumi hypothetic landslide

Based on the landslide simulation (LS-RAPID), we implemented the LS-Tsunami analysis (Sassa et al. 2016). The concept of LS-Tsunami is introduced in Fig. 100 of this article. The result was published in Loi et al. (2021). Figure 71 is a result of LS-Tsunami analysis by this Senoumi hypothetic landslide.

Fig. 71

LS-Tsunami simulation results when considering the hypo-thetical Senoumi landslide in the IODP volcanic ash triggered during the 1.0 Tohoku earthquake record (MYG004). In this figure, the red, yellow, light blue, white and dark blue represent the wave heights corresponding to more than 5 m, 3–5 m, 3 to −3 m, −3 to −5 m and less than −5 m, respectively

Figure 71a–h presents the propagation of the tsunami caused by the Senoumi landslide based on the ring shear test result (Fig. 65a) of the IODP Volcanic ash sample and the 1.0× Tohoku earthquake record (Fig. 67).

Figure 71a at 30 s shows the tsunami soon after the first shock. It appears to be only a local tsunami occurring in Suruga Bay. Figure 71b at 80 s shows the tsunami soon after the second shock. The negative tsunami heights are shown in blue and positive tsunami heights are in red. Tsunami waves expanded onto the opposite bank (Figure 71c). Figure 71d illustrates that the tsunami wave will hit the Fuji area, which is the third largest city in terms of population in Shizuoka Prefecture. Omeazaki area was attacked by tsunami wave at 860 s (Figure 71e). Figure 71f shows that tsunami reached the Ota River. Along the Ota River, four tsunami deposit layers were found in the excavation site (Fujiwara et al. 2019). The right figure on the bottom (Fig. 71h) at 5550 s shows the tsunami wave when the landslide stopped moving.

5.5 Landslides Triggered by the 2011 Typhoon “Talas” in the Kii Peninsula, Japan

5.5.1 Case Studies of the Kuridaira and Akatani Landslides in Kii Peninsula

The Kii Peninsula (Fig. 72a), located in the southeast of Japan, is known as one of the regions frequently prone to deep-seated catastrophic landslides due to its complex climatic condition and geological and morphological settings in fault and fracture zones of Cretaceous to Neogene Shimanto accretional complex (Hayashi et al. 2013; Tien et al. 2018). The area has suffered the most destructive landslide disasters in the years of 1889, 1953 and 2011. In August 1889, the worst storm associated heavy rainfalls triggered more than 1000 large landslides and 33 landslide dams with a total volume up to 200 million m³ were formed in the peninsula. The fatal disasters of landslides, flooding and dam breach killed 1492 lives and forced 2500 people to move out of the Totsukawa village to a new place in Hokkaido (Tabata et al. 2001; Inoue and Doshida 2012; Chigira et al. 2013). In 1953, the Arida-Kawa disasters of flood and landslide dam breach that were induced by heavy rainfall caused 1064 deaths (Tabata et al. 2001). In contrast, the second most severe sediment-related disasters that were triggered by the 2011 record historical rainfall claimed 98 people, completely destroyed 379 houses and partially damaged to 3159 houses (FDMA 2012). In the latest event, about 100 million m³ of landslide sediments was produced from more than 3000 landslides. Among them, large deep-seated landslides exceeding an area of 10,000 m² were formed in 72 locations with a total of 17 huge landslide dams. Five largest landslide dams (namely Kuridaira, Akatani, Nagatono, Kitamata, Iya as shown in Fig. 72a) have been put under urgent countermeasures by Ministry of Land, Infrastructure, Transport and Tourism (MLIT) due to high risks from overflow, debris flow, erosion and dam breach (Hayashi et al. 2013; SABO 2013).

Fig. 72

(a) Location of the Kii Peninsula, (b) the Kuridaira landslide (modified from an aerial photograph taken by the Kii Mountain District Sabo Office), and (c) Google Earth image showing the Akatani landslide area

Case studies of this research are the two largest deep-seated landslide dams in Kuridaira and Akatani valleys upstream of the Kumano River in the Totsukawa village, Yoshino District, Nara Prefecture, Kii Peninsula (Fig. 72b, c). The two landslide cases were selected because of not only the biggest size and the highest risk to potential hazards resulting from river blocking, but also the typical features in term of geology, topography, trigger and failure characteristics. The Kuridaira landslide was about 100 m deep, 750 m long and 600 m wide, with the largest volume of 23.0 million m³. The Akatani landslide had the second biggest volume of 10.2 million m³ with 67 m in depth, 900 m in length and a width ranging from 300 m at the head and to 500 m at the toe (Tien 2018). The Kuridaira landslide created the highest dam on the left bank of the Kuridaira River and on the right tributary of the Taki River joining to the Kumano River. The Akatani landslide-dammed lake is formed on the right-bank tributary of the Kawarabi River joining to the Kumano River. The mass movement of Kuridaira and Akatani landslides blocked the valley courses and formed natural dams up to 100 m high a catchment area of 8.7 km² and 85 m high in a catchment of 13.2 km², respectively (SABO 2013). The storage capacity is of 7.5 and 5.5 million m³ for the Kuridaira and Akatani natural lake, respectively. Although two natural lakes were created by the river blockage in the Akatani valley, the Akatani East landslide dam is about 1.5 km downstream of the Akatani dam was breached just after its formation. The Fig. 72c also presents the Nagatono landslide dam that is behind and and on a nearby slope of the Akatani landslide.

The landslide areas are characterized by broken formations and mixed rocks of Hidakagawa Group in the Shimanto Belt (Tien 2018). The geological map and the cross-section of these two landslides created by Kii Mountain District Sabo Office are denoted in Fig. 73. The Kuridaira and Akatani landslides are characterized by the structure of mixed rocks, broken formations and fractured wedge-shaped discontinuities that are favorable conditions for the build-up of groundwater table (Chigira et al. 2013; Tien et al. 2018). The Kuridaira slope composes debris materials produced from fractured shale and sandstone-dominated rock masses (Fig. 73a). The Akatani slopes is mainly characterized by interbedded sandstone- and mudstone-rich material layers (Fig. 73b). The sliding surfaces of the landslides were mainly formed along the North-West dip direction of the strata on the planes of weakness, such as faults or bedding planes. The sliding plane of the Kuridaira slope was formed in interbedded sandstone and shale layers while it was existed along the interbedded layers of sandstone and mudstone rocks for the Akatani slope. As can be seen in Fig. 73, there existed an abundant condition of groundwater in the study areas with a high level of groundwater table in the Kuridaira slope.

Fig. 73

Geological plan (left) and a cross section (right) of the landslides: (a) Kuridaira and (b) Akatani (redrawn and modified from the figures created by Kii Mountain District Sabo Office)

The Kii Peninsula landslides were induced by a record-breaking accumulative precipitation exceeding 1800 mm during the 2011 Typhoon Talas. Rainfall data recorded between August 31 and September 4 at Kamikitayama AMeDAS Station 17 km northeast from the Kuridaira slope is presented in Fig. 72a. The maximum 72-h rainfall was 1650.5 mm and the maximum daily rainfall reached 661 mm on September 3. The accumulative precipitation triggering the Kuridaira and Akatani mass movements were 1516.5 mm and 1746 mm, respectively (Fig. 74). These high cumulative values of rainfall not only caused a saturation of slope materials, but also brought about a high pore water pressure to trigger the slides (Tien et al. 2018). Seismic recordings of the landslides revealed that the Kuridaira landslide occurred at 23:06:13 on September 3 in duration of 100 s while the Akatani landslide took place at 7:22:00 on September 4 during 70 s (Yamada et al. 2012). The average speed of slope movements was about tens of meters per second, in which, the maximum velocity of the movement of the Akatani landslide was calculated to be 80–100 km/h (Chigira et al. 2013).

Fig. 74

Rainfall Data at Kitayama Station during the Typhoon Talas in Sep. 2011 (Tien et al. 2018)

Various researches have investigated the geological, topographical and morphological features of the 2011 Kii landslides, there has been still a lack of the understanding of the physical mechanism of the deep-seated landslides in the Kii Peninsula (Tien 2018; Tien et al. 2018). This study aims to clarify physical failure mechanisms of the two rainfall-induced landslides and particularly to investigate the liquefaction behaviour of sliding-surface samples that govern the rapid travel of the landslides. For these objectives, a series of laboratory experiments on collected samples was carried out by using the undrained high-stress dynamic-loading ring shear apparatus. The study will not only be helpful for understanding the sliding mechanisms of the Kuridaira and Akatani landslides, but the findings may be employed for the landslide hazard assessment of slopes with the similarity of geology and geomorphology such as in the Kii Peninsula (Tien et al. 2018). Understanding failure mechanism and computer modeling of landslides are indispensable for taking countermeasures in mitigation and preparedness phases against the future landslide disasters.

5.5.2 Soil Sampling and Laboratory Experiment Setting

Because the Akatani landslide has been continuously posing high risks from debris flows and rock failures, the access to the Akatani landslide area was not allowed in the site surveys. Site investigation and soil sampling were carried out in the Akatani East landslide area that were very similar to geological, topographical and morphological features of the Akatani landslide. Based on geological features of the sliding surface, four soil samples, including samples K1 (sandstone-dominated material) and K2 (fractured shale) of the Kuridaira landslide and samples A1 (mudstone-rich material) and A2 (sandstone) in Akatani slope area were collected for ring shear tests (Tien 2018; Tien et al. 2018). Sample K1 contains both coarse-grained clastic rocks and clay-size materials in brown color. Sample K2 comprises small grain-size angular fragments of broken shale in black greenish color and is very brittle to break with hands. Sample A1 is light-gray sandstone-rich materials while sample A2 contains landslide debris of dark-gray mudstone-rich materials. For the above-mentioned objectives, this paper only presents test results and its detailed analysis for shear behaviours of sample K2 and A1. The locations of soil sampling for studying the Kuridaira and Akatani landslides are respectively denoted in Fig. 72b, c, while the photographs of the sampling are shown in Fig. 75. Soil samples were passed through a 2 mm size sieve before ring shear testing. The grain-size distribution of these samples (K2 and A1) are presented in Fig. 76. The unit weight of sample K2 and A1 are 19.5 and 20.2 kN/m³.

Fig. 75

Photos of soil sampling on the sliding surface of Kuridaira (a) and Akatani-East (b) landslides

Fig. 76

Grain-size distribution of tested soil samples

The ring shear simulator with a high undrained capability of 3.0 MPa was employed to investigate the failure mechanisms, changing process of pore water pressure, mobilized shear strength and possible liquefaction for large displacement at steady state (Sassa et al. 2014a, b). In this study, ring shear tests were then performed up to 1500 kPa and 1000 kPa of normal stresses that correspond to over 100 m and near 70 m depth of the sliding surface, respectively. The initial stress conditions were calculated as the depth by slope inclination and soil weight. All of tests were conducted on fully saturated samples with B_D ≥ 0.95.

First, the samples were consolidated at the initial stress state of normal stress and shear stress simulating forces acting on the soil mass at the sliding surface under drained condition. The sample was then sheared in different modes of shearing including undrained shear stress controlled tests and drained pore water pressure controlled tests. The pore water pressure simulating the buildup of groundwater table was increased by using a backpressure control device with a servo-motor. Mobilized shear resistance, pore water pressure change, deformation and displacement of the samples were monitored during the shearing. The analysis of test results is presented in the following section.

5.5.3 Ring Shear Test Results

• (a) Undrained shear-stress control tests for sample K2

Undrained shear stress controlled tests (SSC) were conducted on sample K2 under three different normal stresses (σ) of 500, 1000 and 1500 kPa. For each test the samples were first consolidated at the designed normal stresses in the drained condition. Shear stress was then loaded gradually at a given rate (Δτ/s) of 1–2 kPa/s until the failure in SSC tests. The stress path (a), time series (b), and shear stress–shear displacement curves are presented in Figs. 77, 78 and 79. The tests indicated the failure occurred as the effective stress path reached the failure line. After the failures, the shear strength reduced quickly in progress with an acceleration of shear displacement due to excess pore water pressure generation. The shear stress continued moving along the failure line to certain values of the steady-state shear resistance. All stress paths almost had the same friction angle at peak, namely 44.1° (Figs. 77a and 78a) and 43.8° (Fig. 79a). The friction angle during motion for sample K2 were 42.3° (Fig. 77a). The test at higher normal stress shows the higher mobilized shear resistance at peak and higher steady-state shear resistance. Values of the mobilized shear resistance at peak (τp) were 276, 459, 756 kPa whereas the steady state shear resistance (τss) were 25; 37 and 95 kPa for the 500, 1000 and 1500 kPa tests. The rapid motions were observed when the samples stayed at steady state with low values of residual strength, which indicate the rapid landslide motion. The maximum shear speed reached to about 175 mm/s (Fig. 79c). Test results indicate the occurrence of liquefaction phenomena within the sliding surface due to undrained high-speed shearing condition.

Fig. 77

Saturated and undrained stress-controlled test for Kuridaira sample (stress path (a) and time series (b), Shear stress–shear displacement (c)), normal stress 500 kPa, BD = 0.95, Shear stress increment = 2 kPa/s

Fig. 78

Saturated and undrained stress controlled test for Kuridaira sample (stress path (a) and time series (b), Shear stress–shear displacement (c)), normal stress 1000 kPa, BD = 0.96, Shear stress increment = 2 kPa/s

Fig. 79

Saturated and undrained stress controlled test for Kuridaira sample (stress path (a) and time series (b), Shear stress–shear displacement (c)), normal stress 1500 kPa, BD = 0.96, Shear stress increment = 2 kPa/s

The undrained test results for three different normal stresses, ranging from 500 kPa to 1500 kPa, are plotted in Fig. 80. Although there is a small difference in the values of steady-state shear resistance, all the effective stress paths reached the failure line. The sliding surface liquefaction took place as the failure line approached the steady state stress point. The tests show slight negative excess pore pressure generation at the steady state. The reason might be because of a low permeability and dense clay-sized grain materials created by grain crushing within the shear zone. The relation of the shear strength reduction and the progress of shear displacement is presented in Fig. 81. The failure of all soil samples exhibited three main periods including pre-failure, transient state and steady state that were separated by the values of DU and DL (Sassa et al. 2010).

Fig. 80

Combined stress path of three shear stress controlled tests (500, 1000, 1500 kPa) for Kuridaira sample K2

Fig. 81

Shear stress–shear displacement of three stress controlled tests (500, 1000, 1500 kPa) for Kuridaira sample K2

The DU value defines the shear displacement at the starting point of shear strength reduction (at peak strength) and the DL value indicates the shear displacement at the ending point of shear strength reduction (the shear strength for the initiation of steady-state shear resistance). The values were DU = 2 mm and DL = 150 250 mm for Kuridaira sample 2.

• (b) Pore-water-pressure controlled tests

The rainfall-induced landslides in Kuridaira and Akatani slopes were reproduced in the drained pore water pressure controlled tests. The designed initial stress conditions were estimated from a representative sliding plane depth of about 70 m for the two landslides. Thus, the saturated samples were first consolidated to 1000 kPa normal stress. Under the drained condition, the initial shear stresses were then reproduced at 600 kPa and 620 kPa for Kuridaira and Akatani landslides, respectively. The ratios of shear stress and normal stress were 0.60 and 0.62 that correspond to the natural slope angles of 31 degrees for Kuridaira slope and 34 degrees for Akatani slope, respectively. The simulation of the increment of groundwater level due to rainfall was implemented by using a back pressure control device. Pore water pressure increment simulating rainfall-induced landslide was maintained at a constant rate of 1.0 kPa/s for sample K2 and 0.5 kPa/s for sample A1 (Figs. 82 and 83). The failure of sample K2 and sample A1 initiated as the controlled pore pressure values were about 360 and 340 kPa, respectively. It indicates that the critical pore water pressure ratios (ru) triggering the failures were 0.36 for sample K2 and 0.34 for sample A1, whereas the friction angles at peak were 42.7° for sample K2 and 42.4° for sample A1, respectively. After the failure, both samples remained at small values of apparent friction angle of 3.4° and 5.5°. The steady state shear resistances were 60 and 94.5 kPa for sample K2 and A1, respectively. Both tested samples indicate a high level of landslide mobility because of a great loss of shear resistance under shearing. In the tests, the exact values of excess pore pressure generation were not completely monitored due to the drained condition during shearing. The excess pore pressures were monitored because the generation rate due to grain crushing was larger than the dissipation rate. In addition, it was observed that pore pressure slightly fluctuated due to both generation of positive and negative pore pressures at a large shear displacement in the post-failure moving stage.

Fig. 82

Saturated and pore water pressure controlled test for K2 sample (stress path (a) and time series (b), Shear stress–shear displacement (c)), normal stress 1000 kPa, shear stress 600 kPa, BD = 0.96, Pore water increment = 1.0 kPa/s

Fig. 83

Saturated and pore water pressure controlled test for A1 (stress path (a) and time series (b), Shear stress–shear displacement (c)), normal stress 1000 kPa, shear stress 620 kPa, BD = 0.95, Pore water increment = 0.5 kPa/s

The shear stress–shear displacement curves of pore water pressure controlled tests are presented in Figs. 82c and 83c. The values of DU and DL were 4 and 1500 mm for sample K2 while they are 5 and 1500 mm for sample A1. Before the end of shearing at the prescribed shear displacements, the maximum values of shear velocity reached 115 and 140 mm/s, respectively. The simulation demonstrates the groundwater contribution resulting in high pore water pressure that triggered the rapid landslides.

5.5.4 Sliding-Surface Liquefaction Behavior

The results from undrained shear-stress controlled tests on sample K2 indicate that liquefaction phenomena within the sliding surface resulted from much grain-crushing and excess pore water pressure generation, causing the significant drop of its shear strength at a large displacement (Figs. 77, 78 and 79). In the drained pore water pressure controlled test, the SSL took place at steady state as the shear strength reduced from peak of 42.7° to residual friction angle of 3.4° for sample K2 and from 42.1° to 5.5° for sample A1. The evidences of SSL in the ring shear tests for sample K2 are presented in Figs. 13a and 13b, in which the soil samples were mostly crushed and liquefied to be finer grained particles. The liquefied materials within the shear zone indicate much grain crushing during a high-speed shearing. The sliding surface liquefaction highly agrees with the onsite observations of liquefied materials within the sliding plane of the Kuridaira landslide (Figs. 84c, d). The SSL behavior of sample K2 and A1 was the main reason for the rapid motion of the landslides. This behavior of sliding-surface liquefaction is helpful for the understanding of the rapid moving and long run-out landslides in the Kii Peninsula.

Fig. 84

Sliding surface liquefaction behavior of shale samples under different normal stresses in ring shear tests

5.6 1792 Historical Mega-Slide in Unzen, Japan

The 1792 Mayuyama landslide in the Unzen volcano is a very rapid mega-slide in Japan that caused the largest landslide disaster, the largest volcanic disaster, and the landslide induced tsunami disaster in Japan. The landslide and the landslide induced tsunami reportedly killed a total of 15,153 people. 10,139 people were killed in the Shimabara area. Many other people were killed on the opposite banks by the landslide-induced tsunami wave; 4653 people in the Kumamoto Prefecture, 343 people in Amakusa Island and 18 people in other areas (Usami 1996).

Though this event occurred 230 years ago, the events were well documented. The objective of the development of the high-stress undrained ring shear apparatus (ICL-2) which can test 3 MPa normal stress test is to study the Mayuyama Mega-slide killing 15,153 people. As stated in Table 2, we tried to develop the undrained high-normal stress apparatus (2 mega (DPRI-5) and 3 mega (DPRI-6)). However, we could not succeed in the undrained test more than 400–600 kPa. To succeed in 3 MPa undrained ring shear test, we were able to develop ICL-2 which can test 3 MPa undrained ring-shear test by changing the loading structure from Fig. 9 (DPRI-6) to Fig. 10 (ICL-2) which was published in Sassa et al. (2014a, b).

The Unzen-Mayuyama landslide caused a big tsunami and killed around 5000 people in the opposite bank. To make an assessment of the landslide induced tsunami hazard, we have developed a new landslide-induced tsunami simulation model (LS-Tsunami) which was reported in Sassa et al. (2016).

The organizing committee of the Fifth World Landslide Forum planned a field trip to the Unzen-Mayuyama landslides, Hiroshima urban landslides, earthquake-induced landslides in Aso volcano area. The field trip was cancelled due to COVID-19, but the content of the field trip was contributed to Vol.1, No.2 of the Progress in Landslide Research and Technology which is published. Reviewing those three papers, Sect. 5.6 was written as the final chapter of the ICL Landslide lesson: Sliding-surface liquefaction and undrained steady-state shear strength.

• 1. Kyoji Sassa, Khang Dang, Bin He, Kaoru Takara, Kimio Inoue, Osamu Nagai (2014) A new high-stress undrained ring-shear apparatus and its application to the 1792 Unzen–Mayuyama megaslide in Japan, Landslides 11:827–842.

• 2. Kyoji Sassa, Khang Dang, Hideaki Yanagisawa, Bin He (2016) A and its application to the 1792 Unzen-Mayuyama landslide-and-tsunami disaster. Landslides 13:1405–1419.

• 3. Daisuke Higaki, Kiyoharu Hirota, Khang Dang, Shinji Nakai, Masahiro Kaibori, Satoshi Matsumoto, Masataka Yamada, Satoshi Tsuchiya (2022) Landslides and Countermeasures in Western Japan: Historical Largest Landslide in Unzen and Earthquake-induced Landslides in Aso, and Rain-induced Landslides in Hiroshima. Progress in Landslide Research and Technology, Vol.1, No.2 (in print).

5.6.1 Outline of the Unzen-Mayuyama Landslide

Figure 85 presents the image of the Unzen-Mayuyama landslide (from Google Earth) in the Shimabara city of Nagasaki Prefecture. S1 is the sampling point for the landslide source area. S2 location was selected outside of the landslide moving area to eliminate the effect of the displaced landslide mass.

Fig. 85

1792 Mayuyama landslide in the Unzen Volcano and sampling points

Within Fig. 85, one may find many islands in the Ariake Sea which were parts of the Unzen-Mayuyama landslide. Initially much greater number of islands were formed. However, most of them disappeared by sea erosion after the event. But some landslide mounds still exist as islands as seen in the photo.

Figure 86 presents the section of this landslide which was revised from Inoue (1999) and the MLIT Unzen Restoration Office (2002). The Unzen-Mayuyama mountain consists of volcanic lava rock and unconsolidated eruption products (sand and debris). The section of the initial main landslide block is shown by a red dotted mass, and the secondary sliding block pushed forward by the motion of the initial landslide mass is shown by a black dotted soil layer. The main sliding block and the secondary sliding block were interpreted by the authors. The lines of slope angle of 6.5° and the 28.1° were drawn for the undrained dynamic-loading ring shear test to physically simulate the initiation of the landslide (red dotted area) and the movement of the secondary sliding block (black dotted area) continued to the coast due to the undrained loading from the moving landslide mass.

Fig. 86

The central section of the Mayuyama landslide before/after failure

This megaslide moved very rapidly and entered into the Ariake sea, then caused a very big tsunami wave. Landslide mass and the landslide-induced tsunami cased an enormous landslide and tsunami disaster in Shimabara city and Kumamoto prefecture. Figure 87 is the map showing this big disaster around Ariake sea which was modified from the Unzen Restauration Office (2003).

Fig. 87

Disasters caused by the Mayuyama landslide

The total number of deaths was 15,153 people. The size of circles is proportional to the number of human fatalities in the area. The legend for the number of deaths is shown in the right-top corner (B). (C) The greatest deaths are Shimabara town around the castle (5251 people). (D) The second largest deaths are the southern part of Shimabara Peninsula (around 3500 people). (E), (F), (G) Tsunami-Dome-Ishi (a stone showing the tsunami reaching point, was set to record the tsunami by the community in Kyodomari (E), Umedo (F), and Otao (G) of the Higo (Kumamoto) Han area. The tsunami-Dome-Ishi in Kyodomari was moved for the construction of a road, but its former location is marked on the road retaining wall (by the regional education committee). The Tsunami-Dome-Ishi is limited in Higo (Kumamoto) Han area. These tsunami records are reliable. (H), (I) Stone pillars for memorial service for deaths by Tsunami in Futsu (H) and Mie (I) in Shimabara Han area. Tsunami reaching points recorded by Tsunami-dome stones were: Kyodomari: 9.6 m, Umedo: 14.9 m, Ohtao: 22.5 m, Futsu: 57 m (probably this is too high), Mie: 7.7 m.

This value was compared with the Tsunami simulation model (LS-Tsunami) based on the landslide movement estimated by the landslide simulation model (LS-Rapid) in Fig. 99.

5.6.2 Sampling from the Unzen-Mayuyama Landslide

We took two samples from the landslide site. Sample S1 was taken from the landslide source area and sample S2 was taken from the coastal area outside the landslide area to represent a soil overridden by the landslide.

Sample S1 was taken from a sand layer exposed along a torrent gully in the source area of the landslide. The close-up photo of the sampling site with a hammer. The large debris covered the fine volcanic debris. We took the fine and homogeneous sands for testing. Grainsize is different. Both are the same material. The grain-size distribution of Sample S1 and Sample S2 is shown in Fig. 88b. Sample 1 is finer than Sample 2.

Fig. 88

Photo of Sample 1 in the source area of the Mayuyama Landslide (a) and grain size distribution of Sample 1 and Sample 2 (b)

5.6.3 Undrained Monotonic Stress-Control Ring-Shear Tests on Sample S1

A new ICL-2 undrained ring-shear apparatus was used to test samples taken from the Unzen. This apparatus was developed to investigate landslides with 100–300 m depth. The shear motor of ICL-2 is 11 kW, which is almost 30 times larger than the 400 W motor of the ICL-1. Loading method is the shear-stress increasing test with a constant stress increment (Δτ = 1–2 kPa/s). Loaded normal stresses are from 0.3 to 3.0 Mpa. Four tests were conducted with different normal stresses. The test results are presented in Figs. 89, 90, 91 and 92.

Fig. 89

Saturated and undrained stress controlled test for Unzen sample S1. Normal stress 375 kPa, BD = 0.93, Shear stress increment = 1 kPa/s. (a) Stress path, (b) Time series data, (c) Shear stress–shear displacement relationship

Fig. 90

Saturated and undrained stress controlled test for Unzen sample S1. Normal stress 1030 = kPa, BD = 0.95, Shear stress increment = 2 kPa/s. (a) Stress path, (b) Time series data, (c) Shear stress–shear displacement relationship

Fig. 91

Saturated and undrained stress controlled test for Unzen sample S1. Normal stress 1970 kPa, BD = 0.96, Shear stress increment = 2 kPa/s. (a) Stress path, (b) Time series data, (c) Shear stress–shear displacement relationship

Fig. 92

Saturated and undrained stress controlled test for Unzen sample S1. Normal stress 2900 kPa, BD = 0.96, Shear stress increment = 2 kPa/s. (a) Stress path, (b) Time series data, (c) Shear stress–shear displacement relationship

Figure 89 presents the stress paths and time-series data for 375 kPa normal stress test. The loaded shear stress is increased at the rate of 1 kPa/s. Stress path went up and leftward initially. The stress path moved vertical, then rightward due to negative pore pressure, and it reached the peak shear resistance on the failure line at peak (39.8°). After peak, the stress path shifted to the failure line during motion (38.5°) and went down along the line and reached the steady state shear resistance (τss = 20 kPa). Time series data (b) visualized the negative pore-water pressure due to dilatancy near the peak, a rapid pore pressure increase at the same time of the start of shear displacement. Shear displacement crushed grains in the shear zone, then, a minimal volume reduction generated a high pore-water pressure. Shear speed in the shear stress–shear displacement relationship (c) showed the accelerated motion after failure at 273 kPa. After failure at DL = 7 mm, shear stress was reduced rapidly and reached to the steady state shearing at 20 kPa. These three set of graphs demonstrated a typical sliding surface liquefaction.

Figure 90 presents the test result for 1030 kPa normal stress. The stress path is similar to Fig. 89, but there was no negative pore pressure, although a constant (no increase, and no decrease) pore pressure was measured just before failure. The peak shear strength was 697 kPa, and the steady state shear strength was 45 kPa. The post failure stress drop was very large. Thus, the shear speed acceleratedly increased after failure. This is very similar to Fig. 89.

Figure 91 presents the test result for 1970 kPa normal stress test. The failure line at peak and the failure line during motion were almost the same (40.7° and 39.8°). The peak shear strength was 1472 kPa, and the steady-state shear resistance was 80 kPa.

Shear resistance decreased to only 5%. Shear speed is the fastest of 25 cm/s due to rapid stress drop.

Figure 92 presents the test result of 2900 kPa, close to 3 MPa. It is very large normal stress, but still a complete undrained state was kept during the rotation of the lower ring at the speed of 18 cm/s. The peak failure line and the failure line during motion is almost same (40.3°). The peak shear strength was 1571 kPa and the steady-state shear resistance was 120 kPa. The steady state shear resistance was 7.6% of the peak strength. This is also the typical sliding surface liquefaction.

All stress paths are plotted in the same figure in Fig. 93. As shown, the four tests overlapped along the failure line during motion at 39.8°.

Fig. 93

Combined stress paths of saturated and undrained stress controlled tests for Unzen sample S1

All shear stress–Shear displacement curves are plotted in the same figure in Fig. 94.

Fig. 94

Combined shear stress and shear displacement relationship of the undrained stress controlled tests for Unzen sample S1

The peak shear strengths are much different in the four normal stress tests from 375 kPa to 2900 kPa. However, the steady-state shear resistances were much closer. In the case of Unzen volcanic sands, the undrained shear behaviors are classified as the sliding surface liquefaction. These test results verified the mechanism of rapid landslide motion of the Unzen-Mayuyama Megaslide causing a big tsunami wave. The total number of deaths was more than 15,000, which were the largest landslide disaster and the largest volcanic disaster in Japan.

5.6.4 Initiation Mechanism of Landslides

Three types of initiation process were examined for the Unzen case on Sample 1 and Sample 2

• 1. Landslide initiation by pore-pressure increase.

  Pore-pressure control test was conducted as the basic information.

• 2. Landslide initiation by volcanic seismic shaking.

  From the historical review, the volcanic earthquake (M = 6.4 ± 0.2) occurred in Shimabara city and it was the trigger of the landslide.

• 3. Initiation of the motion of deposits downslope and alluvial deposits by the moving landslide mass (such as Fig. 6).

5.7 Case for the Landslide Initiation Mechanism by Pore-Pressure Increase (Fig. 95)

The first basic test for this landslide (Fig. 95) was to trigger landslide failure by increasing only the pore-water pressure.

Fig. 95

Pore-pressure-controlled test on the Unzen sample S1 to simulate a rain-induced landslide

Firstly, the sample was saturated (BD value, 0.98), then consolidated to 3.0 MPa normal stress and 1.5 MPa shear stress in a drained condition. This preparatory stage was to reproduce the initial stress in the slope, and is shown as a black line in Fig. 95. This initial stress corresponded to a slope of arctan (1.5/3.0) = 26.5°. This is a similar slope to the landslide block in Fig. 86. Then, in order to simulate the pore-pressure-induced landslide process, the pore-water pressure was gradually increased at a rate of Δσ = 5 kPa/s. Failure occurred at a pore-water pressure of 1.2 MPa (a porewater pressure ratio ru = 1.2/3.0 = 0.4). The friction angle at failure was 39.4°.

5.8 Case for the Landslide Initiation by Volcanic Seismic Shaking (Fig. 96)

The third test (Fig. 96) was a seismic-loading ring-shear test to simulate the landslide initiation of the Mayuyama Landslide by the combined effect of pore-water pressure and earthquake shaking. Initially, the sample (S1) was saturated (BD = 0.94) and consolidated to 3 MPa in normal stress and 1.5 MPa in shear stress (the corresponding slope angle was arctan (1.5/3.0 = 26.6°). Then porewater pressure was increased up to 800 kPa, (a pore-water pressure ratio ru = 800/3000 = 0.27) as the initial slope condition. An exact value remains unknown, but it must have been smaller than 0.4. A preparatory test (Fig. 95) showed that ru = 1.2/3.0 = 0.4 was a critical pore-water pressure which could cause a landslide without an earthquake. The earthquake which triggered the 1792 Unzen–Mayuyama Landslide was estimated to be magnitude M = 6.4 ± 0.2, with a seismic intensity of VII during the earthquake; in the Japanese standard, this corresponds to a seismic acceleration of more than 400 cm/s², as explained above.

Fig. 96

Seismic-loading ring-shear test on the Unzen sample S1. (a) Stress path, (b) Time series data

The maximum recorded seismic acceleration in the 2008 Iwate–Miyagi earthquake was 739.9 cm/s², which caused the Aratozawa Landslide. We loaded the N–S component of the 2008 Iwate–Miyagi earthquake record (maximum acceleration is 739.9 cm/s²) at MYG004 as the additional shear stress. For precise pore-pressure monitoring, as well as to maintain servo-stress control, a five times slower rate was used in applying the recorded seismic acceleration. The test result is shown in Fig. 96.

The green line indicates the control signal. The maximum value is 2469 kPa (1500 + 969 kPa) and the minimum value is 369 kPa (1500−1131). The loaded acceleration (a) is calculated from the ratio of seismic acceleration and gravitational acceleration: a/g = 969/1500 or a/g = 1131/1500, because ma = 969 kPa and mg = 1500 kPa, expressing the landslide mass at unit area as m. The acceleration corresponds to +633 and −739 cm/s². Therefore, the control signal for shear stress sent to the ring-shear apparatus exactly corresponded to the monitored acceleration record. As Fig. 96 shows, failure occurred around 1825 kPa, namely, a/g = (1825−1500)/1500 = 0.22; the necessary acceleration to failure was 216 cm/s². This test result suggested that a lesser earthquake shaking (of around 216/633 = 0.34) than occurred in the Iwate–Miyagi earthquake could have caused failure under a slope condition with a pore-pressure ratio of 0.27. The steady-state shear strength was 157 kPa.

5.9 Case for the Initiation of the Motion of Deposits Downslope/Alluvial Deposits by the Moving Landslide Mass (Fig. 97)

ICL-2 was used to simulate undrained loading on the black-dot layer in the lower slope continued to the coast (Fig. 86) using the displaced mass from the upper slope (red-dot block). Figure 97 presents the test result of an undrained dynamic loading test simulating this scenario. Firstly, the initial normal stress and the initial shear stress (σ₀ = 1000 kPa, τ₀ = 150 kPa, corresponding to an 8.5° slope) were loaded in the drained condition to reproduce the initial stress state at the bottom of the black-dot layer. The normal stress was increased to 2790 kPa (which is close to the 3 MPa capacity of the apparatus) as the undrained load, although a 400-m-deep initial landslide will result in a greater normal stress. If the lower slope mass can resist the undrained load from the upper slope without raised porewater pressure, it may resist around 2374 kPa (2790×tan (40.4°)) in this dynamic loading. However, as seen in Fig. 97, a high porewater pressure was generated in the undrained loading. The sample failed at 720 kPa and its steady-state stress was 80 kPa. The landslide mass from the upper slope should scrape off a layer of the lower slope and move together as a combined greater mass toward the sea. This scenario was proven to be possible by this landslide simulation test using ICL-2.

Fig. 97

Undrained dynamic loading test on Sample 2 B_D = 0.97. Initial stress (Normal stress = 1000 kPa, Shear stress = 150 kPa). (a) Stress path, (b) Time series data

5.9.1 Observation of the Shear Zone State Between Undrained Test and the Drained Test (Fig. 98)

Sample S2 was taken from Volcanic debris. The grain shape is angular. We compared the shear zone after shearing under the high normal stress of 3 MPa.

Fig. 98

Samples after shearing. Left: undrained test. Right: Drained test (pore-pressure control test)

The left photo shows the shear zone of S2 after the undrained stress control test (Fig. 92), the right photo shows the shear zone of pore-pressure control test (Fig. 95). Pore-pressure control test was conducted under the drained condition. Pressure of pore-water was controlled, while pore-water moved from/to the shear box through the pore water control system (Fig. 10). Shearing proceeded in almost liquefied state in the undrained test (Figs. 92 and 96), whereas shearing in the pore pressure control (drained test) proceeded not in the liquefied state. So, grain crushing under the pore-pressure control test should be greater than that of the undrained shear test. The right photo visualized a silty shear zone which was formed by grain crushing of the sample. We could not find the shear surface in the left photo. The steady state was reached at 10–20 cm shear displacement. During the steady state, effective normal stress, pore pressure and shear stress were constant. Only shear displacement progressed without further grain crushing.

5.9.2 LS-RAPID Simulation (Fig. 99)

We applied LS-RAPID to estimate the motion of the 1792 Unzen-Mayuyama Landslide.

Fig. 99

Movement of the Mayuyama landslide estimated by LS-RAPID simulation

Triggering factors used were pore-water pressure + earthquake shaking which represent the case for the landslide initiation by volcanic seismic shaking (Fig. 96). The parameters used in the computer simulation are listed in Table 7.

Table 7 Parameters used in computer simulation (LS-RAPID and LS-Tsunami) of the Unzen-Mayuyama landslide

The simulation result is presented in Fig. 99.

Red color balls represent the moving mass, while blue balls represent the stable mass.

• At 11 s, the pore-water pressure reached 0.21 and the earthquake started, but no motion occurred.

• At 17 s, the main shock of earthquake struck the area and failure occurred within the slope. Failure began from the middle of the slope.

• At 26 s, the earthquake has almost terminated. The whole landslide mass was formed during the earthquake shaking.

• At 64 s, the landslide mass continued to move after the earthquake ended and entered into the sea

• At 226 s, the landslide mass stopped moving and was deposited.

In Fig. 99, the simulated deposition area was compared with the map made by the Unzen Restoration Office of the Ministry of Land, Infrastructure and Transport of Japan (2002) based on topographic survey (right-bottom of the figure). To compare both landslide motions, the section of line A in the right-bottom figure and the E–W section (almost the same as line A) of the computer simulation were compared. Both movements were very similar. The travel distances from the head scarp of the initial landslide to the toe of the displaced landslide mass were around 6 km. The travel distance in the computer simulation was 6.6 km, while the distance based on the field investigation of the deposits was 5.9 km.

5.9.3 LS-Tsunami Simulation (Fig. 100)

Figure 100a presents the basic principle of the landslide-induced tsunami simulation model and Fig. 100b presents the tsunami wave calculated by LS-Tsunami (Sassa et al. 2016).

Fig. 100

Landslide-induced tsunami simulation model (LS-Tsunami) (a) and its application for the Tsunami wave induced by the Mayuyama Landslide (b)

The basic principle of the landslide-induced tsunami model (Fig. 100a)

The left part of Fig. 100a is the same as the principle of the integrated landslide simulation model (LS-RAPID) of Sassa et al. (2010). The only difference here is that the landslide mass is located beneath a water surface. The right part of Fig. 100a illustrates the elevated water mass (which causes a tsunami wave) by the upward displacement of the interface between the landslide and the water due to the movement of the landslide mass over the sea floor.

When the landslide mass enters or travels across the sea floor, the elevation of the submarine ground surface (which is the interface of landslide mass and submarine water) will be increased. The submarine ground surface elevated by the landslide mass will raise the water mass above the landslide-water interface by an amount, Δh, in a unit time, Δt. The shear resistance (τ_sw) between soil and water at this interface should be much smaller than the shear resistance within the soil. Hence, this resistance, τ_sw, was ignored in the simulation. The effective stress acting on the surface between the bottom of the soil column and the top of the sea floor was regarded as being unaffected by the increase in water height for the following two reasons: (1) The surficial deposit on the sea floor is unconsolidated and hence is expected to be permeable enough to transmit the water pressure change due to water height change to the bottom of the soil column above the sea floor; and (2) the soil column of the moving mass is both unconsolidated and water saturated, and hence the transmission of water-pressure change from the top of soil column to the bottom of the soil column is fast (elastic wave velocity of water is around 1.5 km/s) and can be regarded as instantaneous.

When a landslide mass enters into water, the buoyant force by water will act on the soil mass. Namely, the unit weight of soil mass (γ_t) on land will change to the buoyant unit weight of soil mass, the difference between γ_t and γ_w (unit weight of seawater) under water. In the case of partial submersion, the buoyant force by water will act only on the part of soil mass below the water surface. No tangential force between soil and water is considered because the tangential force by water will be negligible compared to the shear resistance mobilized in the shear surface of large-scale landslides.

There are many parameters to regulate landslide motion on land and submarine landslide motion. The values of such landslide dynamics parameters were mostly measured by the undrained ring shear testing on the samples taken from the landslide source area and its moving area. The parameters used in tsunami simulation (LS-Tsunami) are the unit weight of seawater and Manning’s roughness coefficient for basal resistance between water and ground.

The Result of LS-Tsunami-Propagation of Tsunami Wave from the Landslide to the Opposite Bank (Fig. 100b)

Figure 100b presents the contour of the maximum tsunami height at each mesh with the tsunami height records in the Kumamoto side coast and in the Shimabara side coast. The coastline is shown in red where the tsunami wave moved up to the land. Table 8 presented the reported and simulated tsunami heights at five sites along the Kumamoto and the Shimabara coasts. Differences between the computer simulation and the historical records for the five specific points were less than 5.7 m for Manning’s coefficient 0.025 m^–1/3/s, 4.6 m for Manning’s coefficient, 0.020 m^–1/3/s, and 5.2 m for Manning’s coefficient 0.015 m^–1/3/s except the case at Futsu. Futsu is a special case where the first direct tsunami wave from the source of tsunami hits the cliff of the Futsu terrace and probably amplified due to the gulf-shaped coast line.

Table 8 Comparison of tsunami heights between the simulation result and the historical record

The simulation of LS-Tsunami used 100 m mesh data. It cannot include the effect of local amplification of tsunami height due to the sharp topography in this part of the coast. LS-Tsunami uses the average elevation for each 100-m mesh. The elevation of Tsunami-Dome-Ishi on the slope is not always close to the average height of calculated mesh. In the case of 15–20° slope, 25-m (one fourth of one mesh) difference in horizontal distance cause 6.7–9.1-m elevation difference. The maximum 5.7-m difference in Table 8 is within the elevation difference on the slope. Considering these conditions, the estimated area of tsunami and the tsunami heights along the coast by LS-Tsunami matches with the historical record. It can be understood that the result of this trial test on the Unzen-Mayuyama Landslide-and-tsunami disaster presented a reasonable extent of reliability and a meaningful extent of precision within the variation of Manning’s coefficient and the resolution of ground elevation. Carefully examining the value difference in different Manning’s coefficient, the effect of the coefficient is greater in the long travel distance of tsunami and gives closer tsunami heights in Kyodamari and Umedo.