Comprehensive analyses of the initiation and landslide-generated wave processes of the 24 June 2015 Hongyanzi landslide at the Three Gorges Reservoir, China
- 1.7k Downloads
Reservoir landslides pose a great threat to shipping safety, human lives and properties, and the operation of the hydropower station. In this paper, the 24 June 2015 Hongyanzi landslide at the Three Gorges Reservoir is considered as an example to study the initiation mechanism and landslide-generated wave process of a reservoir landslide. The finite difference method and limit equilibrium analysis are used to analyze the deformation and failure characteristics of the Hongyanzi slope. Simulation results show that a large deformation (about 358 mm) happens in the shallow deposits under intermittent rainfall condition, and the slope is in a limit state. At the same time, continuous rapid drawdown of the water level (about −0.55 m/day during 8–24 June 2015) reduced the support and accelerated the drainage of the water for the bank slope. A coupling effect of intermittent rainfall and rapid drawdown of the water level was the triggering factor of the 24 June Hongyanzi landslide. Landslide-generated wave process was simulated using a fluid–solid coupling method by integrating the general moving object collision model. Simulation results show that the landslide-generated wave is dominated by the impulse wave, which is generated by sliding masses entering the river with high speed. The maximum wave height is about 5.90 m, and the wave would decay gradually as it spreads because of friction and energy dissipation. To prevent reservoir landslides, the speed for the rising or drawdown of the water level should be controlled, and most importantly, rapid drawdown should be avoided.
KeywordsHongyanzi landslide Triggering mechanism Fluctuation of water level Landslide-generated wave Fluid–solid coupling model
The Yangtze River is the mother river of China, which is characterized by a large water discharge, and it is very important because it delivers sediment and supports the lives of Chinese people (Saito et al. 2001; He et al. 2008). In recent years, due to increasing demand on the energy supply, irrigation, and flood control from the government departments, several water conservancy projects have been constructed along the Yangtze River, such as a flood protection embankment and a reservoir dam (Wu et al. 2004). The construction of high dam on a river causes the water level to rise and forms a huge reservoir upstream, which may lead changes in local climate and the ecological environment (Yin and Li 2001). Among these changes, sediment accumulation and reservoir landslides are two key issues for the Three Gorges Reservoir (Liu et al. 2004; Huang et al. 2014). Several years of observational data show that the frequency of landslides has significantly increased since the impoundment of the Three Gorges Reservoir, especially for the Three Gorges area extending from Yichang to Chongqing along the Yangtze River (Wu et al. 2004; Tang et al. 2015). These landslides pose a great threat for shipping safety, and they endanger human lives and properties and threaten the operation of the hydropower station.
Numerous catastrophic landslide events have occurred at the Three Gorges Reservoir in recent years. For example, on 12 June 1985, the Xintan landslide, with a volume of approximately 3.0 × 107 m3, occurred 27 km away from the Sandouping dam site of the Three Gorges project. Sliding masses rushed into the river and caused a wave with a height of 80 m (He et al. 2010). On 14 July 2003, the Qianjiangping landslide, with a volume of about 2.4 × 107 m3, occurred on the western side of the Qinggan River (a tributary of the Yangtze) and caused a wave with a height of 20 m. It resulted in 15 death and 9 missing people, and it destroyed 4 corporations and 80 houses (Wang et al. 2004; Yin et al. 2015). On 2 September 2014, the Shanshucao landslide, with a volume of 4 × 107 m3, occurred on the left bank of the Luogudong River (a tributary of the Yangtze) and caused damage to 20 ha of citrus orchard and 27 utility poles, which resulted in direct economic losses of 5.19 million US dollars (Huang et al. 2014). Previous studies show that heavy rainfall and increasing water level are the two triggering factors for these catastrophic landslides (He et al. 2010; Jiang et al. 2015). The threat to life and property comes not only from the landslide itself but also from the landslide-generated wave, which causes huge volumes of sliding mass to rush into the river (Yin et al. 2012). Therefore, studying the triggering conditions and mechanism of landslide-generated waves and other related problems is very important for hazard prevention and mitigation of reservoir landslides.
For reservoir landslides, statistical analyses, field investigations, physical modeling tests, and numerical simulations are always used to study the triggering factors and mechanism (Guzzetti et al. 1999; Bosa and Petti 2011; Zhou et al. 2010). Statistical analyses are generally used to study the relationships between rainfall intensity/duration and landslide density, the fluctuation of water level and landslide frequency, and other factors in a specific region, and these provide the reference triggering conditions for the reservoir landslides, such as critical rainfall intensity, rising drawdown rate of the water level, and other factors (Aleotti and Chowdhury 1999; Chung and Fabbri 1999). The triggering mechanism and mass movement process of a landslide may leave behind landslide traces or deposits, so field investigations are important for studying landslides. Numerical simulation can be used to reproduce the triggering mechanism or dynamic process of a landslide. However, in most cases, using different research methods at the same time is more effective for studying the triggering mechanism or dynamic process of a landslide (Iverson 2000; Huang et al. 2014). Duc (2013) studied the rainfall-triggered large landslides on 15 December 2005 in Van Canh District through field investigations and laboratory tests, and they found that the triggering factor for this landslide was heavy rainfall. The slope stability was related to slope inclination and the thickness of the residual soils in the area. Zhang et al. (2012) used laboratory tests and numerical simulations to study the triggering mechanism of a reservoir landslide and found that water absorption/dissolution may cause a reduction in shear strength of rock masses during impoundment; this results in the initiation of a landslide.
When the sliding masses of a reservoir landslide suddenly rush into the river, a wave will be generated and will pose a great threat to the surrounding residents and their buildings and ships (Chopakatla et al. 2008; Hu et al. 2015). However, the landslide-generated wave process is very short. Observation data is often not collected, and we can only presume the details of the disaster process by using water traces and reports given by local witnesses (Koo and Kim 2008). For landslide-generated wave problems, theoretical analyses are difficult due to strong non-linearities, three-dimensional topography, and flow turbulence (Fritz et al. 2004; Abadie et al. 2010; Yin et al. 2015), and laboratory tests are limited by scale effects (Xu et al. 2015a). Therefore, numerical simulation is more reasonable for studying the dynamic process of a landslide-generated wave (Ataie-Ashtiani and Mansour-Rezaei 2009; Yavari-Ramshe and Ataie-Ashtiani 2015). For example, Yin et al. (2012) conducted a physical modeling test to study the characteristics and evolution of a landslide-generated wave, and they confirmed the presence of a head wave and determined the calculation formulas for prediction of a landslide-generated wave in the Three Gorges Reservoir area. Xu et al. (2015b) conducted a large-scale physical modeling test to study the impact of a landslide-generated wave, and they found that the contact area, landslide volume, and drop height are the main factors that effect the scale of a wave. Choi et al. (2008) conducted a 3D solitary wave run-up analysis with the Reynolds average numerical simulation (RANS) turbulent model to study the dynamics of a landslide-generated wave. Das et al. (2009) studied the impact of slide deformation and geometry on surface waves generated by submarine landslides using a computational model based on Navier–Stokes equations. Montagna et al. (2011) presented a numerical method based on the volume of fluid (VOF) technique to study tsunamis generated by the landslides falling along the flank of a conical island.
In this paper, the 24 June 2015 Hongyanzi landslide at the Three Gorges Reservoir is taken as an example to study the initiation mechanism and landslide-generated wave process of a reservoir landslide. The initiation mechanism of this landslide is studied by using field investigations, laboratory tests, numerical simulations, and theoretical analyses. A fluid–solid coupling method that integrates the general moving object (GMO) collision model is used to simulate the dynamic process of the landslide-generated wave. The initiation mechanism of the reservoir landslide and forecasting of the landslide-generated wave are discussed. The presented conclusions are useful for understanding the initiation mechanism and landslide-generated wave process of a reservoir landslide.
The Hongyanzi landslide
In this section, the hazard of the Hongyanzi landslide at the Three Gorges Reservoir is briefly introduced. Then, the geomorphology and geological conditions for the study area are described. In addition, the several years of observation data for the hydrological and climate conditions are illustrated.
Geomorphology and geological conditions
Mechanical parameters for the main geomaterials of the slope obtained by laboratory tests and engineering experience
Shear strength parameters
Elastic modulus (GPa)
Friction angle (°)
Hydrological and climate conditions
The Yangtze River extends 6397 km with a total catchment area of 1.8 × 106 km2. It is the largest river in China and the third largest river in the world, after the Nile and Amazon. The Yangtze River system is extensive, with more than 7000 tributaries. The main tributaries are the Yalong River, Min River, Jialin River, Wu River, and Xiang River. The average flow of the Yangtze River is 3.1 × 104 m3/s, with significant seasonal fluctuations. The flow has a great impact on the stability of the slopes on both sides.
Landslide initiation mechanism
In this section, the deformation and failure characteristics of the Hongyanzi slope are analyzed by using a finite difference method, and the potential slip surface is determined through numerical simulation and limit equilibrium analysis. Initiation process and mechanism of the Hongyanzi landslide are analyzed from the simulation results; the field investigations for the landslide; and the monitoring data of slope deformation, rainfall, and fluctuation of the reservoir water level.
Slope deformation and failure analyses
Initiation process and mechanism
June is the rainy season for the study area. The daily rainfall values were 57.5 and 44.5 mm on 16 and 17 June 2015, respectively, and the daily rainfall was 4.5 mm on 20 June 2015. Before the catastrophic landslide on 24 June 2015, there was antecedent intermittent rainfall for about a week. Field monitoring results show that on 21 June 2015, a tensile fracture formed at the rear (elevation 276 m) of the Hongyanzi slope and a small slide occurred at the slope toe. On 23 June 2015, a slide occurred with a volume of 1.5 × 104 m3 below the elevation of 175 m. Unfortunately, the deformation of the slope rear increased suddenly on the morning of 24 June 2015, and the arc shape of the tensile fracture was extended to 25–80 m, with a width of 3–10 cm. Monitoring results show that the maximum horizontal displacement of the Hongyanzi slope was about 13 cm at 13:30 on 24 June 2015. At 18:40 on the same day, a catastrophic landslide occurred in the Hongyanzi slope, with a volume of 2.3 × 105 m3. The deformation and failure of the Hongyanzi slope were related to the intermittent rainfall and fluctuation of the water level in the Three Gorges Reservoir.
Simulation of landslide-generated wave
Once the large volume of sliding mass rushed into the river, a wave was generated and it disturbed the balance of the river flow. Here, a numerical method is used to reproduce the dynamic process of the landslide-generated wave.
According to the dynamic process for the landslide-generated wave of the Hongyanzi landslide, a 3D computational fluid dynamics (CFD) code was used to solve the Reynolds-averaged Navier–Stokes equations, and the continuity equation for incompressible flow along with the true VOF method was used to compute the free surface motion (Montagna et al. 2011). A GMO collision model was used to simulate the sliding mass, which was represented as a rigid body under any type of physical motion (Pastor et al. 2009). The GMO collision model is suited for either being dynamically coupled with fluid flow or user-prescribed. It has six degrees of freedom to move or rotate with a fixed point or a fixed axis. The collision model includes collision detection and collision integration. The energy loss of the collision was controlled by the coefficient of restitution. Here, the motion of sliding masses is coupled with the water flow, and the sliding masses were assumed to have GMO rotary motion and collision. The overall restitution coefficient was 0, and the overall friction coefficient was 1.73, based on the characteristics of the Hongyanzi landslide and previous experience (Yin et al. 2015). The renormalization group (RNG) turbulent model was used to simulate the fluid motion once the sliding mass rushed into the water. The RNG turbulent model is derived from rigorous statistical analyses, and it uses equations similar to the k-ɛ model. Empirical values are used for the equation constants (Lindsey et al. 2013).
Considering wave propagation problems, the symmetry boundary was set for the start face and end face of the computational domain in the x direction, and the specified pressure boundary (the air, free surface boundary) was set for the end face in the z direction. The other faces were set at the wall boundary, which means that the normal velocity there must be 0. The vertical gravity was applied to the entire system, and the acceleration of gravity was 9.8 m/s2 in the negative z direction. The sliding masses were set as moving objects and scour material, and the initial state of the moving objects was motionless. After 150 s of calculation time for the fluid–solid coupled interacting model, the simulated results for the landslide-generated wave were obtained.
Formation of the wave
A landslide-generated wave generally contains two parts: (a) the volume-wave generated by the sliding masses along with the lifting of the water level and (b) the impulse-wave generated by the sliding masses entering the river at high speed. Landslide with the volume of 2.3 × 105 m3 sliding into the Daning River at high-speed, huge potential energy is converted into the impulse-wave energy. Due to the relatively shallow watercourse and the wide river near the landslide site, it makes the volume-wave play the secondary role of the landslide-generated wave, and the impulse-wave is the dominant aspect. The wave was limited to the area around the sliding masses only at the beginning, because of the short-time landslide movement, and the peripheral water of the landslide was subject only to the friction force at the boundary of the water.
The dynamic process of the landslide-generated wave can be divided into three stages: (1) the beginning of the interaction between the sliding masses and the river water. Once the sliding masses rushed into the river, the water particles were pushed and expanded by the motion of the masses, which caused the free water surface to rise. The water tongue jumped to the water surface in a parabolic arc, and then, the lower water started to move from the static. The sliding masses splashed at the same time as the wave formation. (2) The formation of the wave along with the continuous falling of the sliding masses into the river caused a cavity to form at the tail of the sliding masses, and there were great difference inside and outside the cavity at the water level. The surrounding water moved into the cavity quickly and caused collisions, which produced a huge splash. At the same time, the energy carried by the sliding masses was transferred to the surrounding water and the wave formed. The largest height of the wave occurred in the landslide direction. The wave spread around in the form of a vibration wave with a small amount of water flow, which had the characteristics of the translation wave. (3) Wave propagation and attenuation. The wave height decayed gradually as it spreads around because of friction and energy dissipation.
Wave propagation process
Figure 13c shows the instantaneous free surface elevation and vectors of flow at 47 s after the Hongyanzi landslide. As shown in Fig. 13c, one peak of the wave was near the landslide area. All free surface elevations were higher than the original surface levels (elevation of 145 m), and the water level did not easily decrease during the propagation. This phenomenon was very significant for the early warning and forecast during the landslide wave. The wave velocity is another important factor associated with early warning; it determines when and where the landslide-generated wave can arrive. The direction of the maximum velocity was consistent with the landslide direction and has the similar attenuation properties as the wave height. Finally, the wave decays gradually as it spreads due to friction and energy dissipation.
Initiation mechanism of reservoir landslides
Landslides are one of the major geomorphological problems in reservoirs (Gutiérrez et al. 2015), especially in mountainous areas. Previous studies indicate that a large portion of the slope movements in reservoirs are related to the reactivation of pre-existing undetected landslides (Wang et al. 2007). New slope movements can occur upstream of a dam reservoir during the impoundment process because of varied groundwater levels along the upstream slope and saturation and softening of the slope soils below the water level (Schwab et al. 2004; Setiawan et al. 2015). For reservoir landslides, the fluctuation of the water level is the key triggering factor (Berilgen 2007). The reservoir bank slopes below the water level are affected by long-term saturation and flow erosion. Small failures happen on the lower slope, which is below the water level. Small failures at the toe of the slope do not have a great impact on the reservoir, but a free surface will form for the upper large volume of slope soils. This plays a key role in the following catastrophic reservoir landslide. The fluctuation of the water level includes two aspects, rising and drawdown, especially for rapid drawdown. The water-level fluctuations may induce changes in the pore water pressure in the bank slopes, which include seepage-induced pore pressures due to transient flow and stress-induced excess pore pressure (Yan et al. 2010; Vandamme and Zou 2013).
During the rapid drawdown process of reservoir water level, the descending speed of the slope internal saturation line lags behind the drop speed of the slope outside water level, and the change in the external water level happens before the slope soils have time to drain (Wang and Qiao 2013). Dissipation of excess pore pressure and reduction of the seepage-induced pore pressure can cause traction force to act on the slope and decreased the slope stability, which may lead to the occurrence of a reservoir landslide (Zhang et al. 2012). Heavy or long-term rainfall is another important triggering factor for reservoir landslides (He et al. 2008). Rainfall infiltration can decrease the shear strength of slope soils and increase pore water pressure in the slope. When continuous rainfall occurs, a transient saturated zone is easily formed above the groundwater level, which can lead the loss of matric suction when the slope soil is saturated. This significantly reduces slope stability and may trigger failure of the bank slope. If rapid drawdown of the water level occurs at the same time, the coupling effect of rainfall and drawdown of the water level on the slope will cause a reservoir landslide to occur more easily. The Hongyanzi landslide is a typical example of this. In most cases, reservoir landslides are usually not triggered by a single factor (such as rainfall or a rising water level). Rather, they are the comprehensive responses to multiple factors affecting the reservoir slope.
For the Hongyanzi slope studied in this paper, the joint effect of intermittent rainfall and rapid drawdown of the water level resulted in the catastrophic reservoir landslide, and the dynamic process for the initiation of this landslide can be summarized in three stages: (1) intermittent rainfall cause the decreasing of shear strength of the soil and the increasing of pore water pressure in the slope, meanwhile tensile fractures are formed at the rear of the slope; (2) along with the intermittent rainfall and drawdown of the water level, a large deformation occurred, and the tensile fracture was further developed; a small local slide occurred at the toe of the slope; (3) the rainfall stopped, but the water level continued to decline. The shear resistance for the slope decreased, and the drainage of water in the slope caused the traction force to act on the shallow deposits. Finally, the result was the catastrophic landslide that happened on 24 June 2015.
Forecasting of landslide-generated wave
Once a large landslide is identified in a reservoir, the following two important questions should be answered: could this landslide lead to a hazardous impulse water wave, and what would be the scale of the landslide-generated wave (Gutiérrez et al. 2015)? The main factors affecting the landslide-generated wave include the volume and geometry of the landslide, its velocity of entry into the water, and the underwater topography of the reservoir (Panizzo et al. 2005; Ataie-Ashtiani and Yavari-Ramshe 2011). Although a landslide-generated wave is a typical secondary disaster of a reservoir landslide, the damage of the impulse water wave may be larger than that of the landslide itself. The effective forecasting of landslide-generated wave and prediction of its scale are very important for hazard prevention and mitigation of reservoir landslides. Usually, some empirical models are used to calculate the characteristic parameters of the landslide-generated wave. These models are usually used in engineering practice, and they can provide rapid evaluation for landslide-generated waves. Empirical models for the prediction of landslide-generated wave are well reviewed by Bornhold and Thomson (2012). However, most of the empirical models are presented from the laboratory test results, and assuming a certain proportion of potential energy is converted into the total energy of the waves (Bornhold et al. 2007). During the forecasting process of landslide-generated wave using empirical model, reasonable empirical model and computational parameters should be carefully selected according to the failure type and characteristics of the landslide; if not, large error may be existed.
Numerical models used for simulation of landslide-generated wave main include rigid body models, viscous flow models, Bingham plastic slide models, impulse wave models, and others. Comparisons of rigid body and viscous slide models show that the former generate much higher waves than the latter for the same input parameters (Bornhold and Thomson 2012). There are also significant differences in the propagation, frequency content and decay of landslide-generated wave by the two models (Bornhold and Thomson 2012). These differences are also existed in other type of numerical models, which mean that each type of numerical models has its own characteristics and scope of application. For the simulation of landslide-generated wave using numerical methods, reasonable numerical model and parameters should be used, same as the empirical methods. According to the forecasting of the landslide-generated wave, the empirical formula can be preliminary used to evaluate the characteristics of the wave. If the scale of the landslide-generated wave is small and it has no influence on the surrounding infrastructures and human beings, no further in-depth analysis is needed. However, if the empirical formula predicts that the scale of the wave is large enough to cause destruction or damage to the surrounding infrastructure and endanger people, then numerical methods can be used to simulate the propagation process of the wave in detail.
The initiation of reservoir landslides is generally related to the fluctuation of the water level and rainfall duration in the reservoir area. Once the sliding masses of the reservoir landslide suddenly rush into the river, a wave is generated, and it can pose a great threat to the surrounding residents and their buildings and ships. Studies of the triggering conditions and mechanism of the landslide-generated wave process are very important for hazard prevention and mitigation of reservoir landslides. In this paper, the 24 June 2015 Hongyanzi landslide at the Three Gorges Reservoir was used as an example to study the initiation mechanism of a landslide-generated wave from a reservoir landslide.
The finite difference method and limit equilibrium analysis were used to analyze the deformation and failure characteristics of the Hongyanzi slope. Simulation results show that a large deformation (about 358 mm) happened in the shallow deposits under intermittent rainfall condition, and a clear potential slip surface formed in the Hongyanzi slope. The safety factor of the Hongyanzi slope under natural condition is 1.230, and it decreased to 1.031 under rainfall condition, which means that the slope reached the limit state after antecedent intermittent rainfall duration. From 8 June 2015 to 24 June 2015, the water level of Three Gorges decline rapidly, at a rate of −0.55 m/day. The rapid drawdown of the water level reduced the support for the Hongyanzi slope, and it accelerated the drainage of water from the slope, which provided the dynamic conditions for the initiation of the landslide. Thus, the initiation of the Hongyanzi landslide was the result of multiple influences, including the intermittent rainfall and the rapid drawdown of the water level for the reservoir deposit slope.
Numerical methods were used to simulate the landslide-generated wave process of the Hongyanzi landslide. The results show that, when the sliding masses rush into the river, the balance of the river is disturbed and the water particles are vibrated. The landslide wave is dominated by the impact wave generated by the sliding masses entering the river at high speed. The maximum wave height of 5.90 m occurs near the landslide (it occurred 12 s after the end of the landslide), and the minimum wave height occurs on the opposite side. The direction of the maximum velocity is consistent with the landslide direction. Finally, the wave decays gradually as it spreads, due to friction and energy dissipation. For the prevention of reservoir landslides, the most important point is that the speed of increase or drawdown of the water level should be controlled. Rapid fluctuations of the water level need to be avoided, especially for drawdown. Also, rainfall infiltration into the slope should be reduced by engineering and non-engineering measures, such as maintaining good vegetation cover, promoting ecological protection, and installing an impermeable membrane cover for the slope. Furthermore, some drainage measures can be adopted for the slope to accelerate the drainage of infiltrated water; drainage holes can be used, for example. For the forecasting of a landslide-generated wave, some empirical formulas can be preliminarily used to evaluate the characteristics of the wave. If the scale of the landslide-generated wave is large enough, then numerical methods can be used to simulate the propagation of the wave in detail. However, for some potential catastrophic reservoir landslides, engineering measures cannot prevent the landslide. In these cases, some mitigation measures should be planned ahead of time, such as relocation of residents and prohibition of navigation.
We gratefully acknowledge the support of the National Natural Science Foundation of China (41472272 and 41102194) and the Science Foundation for Excellent Youth Scholars of Sichuan University (2013SCU04A07). Critical comments by anonymous reviewers greatly improved the initial manuscript.
- Chung CJF, Fabbri AG (1999) Probabilistic prediction models for landslide hazard mapping. Photogramm Eng Remote Sens 65:1389–1399Google Scholar
- Das K, Green S, Basu D, Janetzke R, Stamatakos J (2009) Effect of slide deformation and geometry on waves generated by submarine landslides: a numerical investigation. Offshore Technology Conference, Texas, pp 1–12Google Scholar
- Gutiérrez F, Linares R, Roqué C, Zarroca M, Carbonel D, Rosell J, Gutiérrez M (2015) Large landslides associated with a diapiric fold in Canelles Reservoir (Spanish Pyrenees): detailed geological-geomorphological mapping, trenching and electrical resistivity imaging. Geomorphology 241:224–242CrossRefGoogle Scholar
- Yin KL, Liu YL, Wang Y, Jiang ZB (2012) Physical model experiments of landslide-induced wave in Three Gorges Reservoir. J China Univ Geosci 37:1067–1074 (in Chinese) Google Scholar