Geohazards in the three Gorges Reservoir Area, China – Lessons learned from decades of research

1. Introduction

The Three Gorges Dam on the Yangtze River has the largest installed hydropower capacity in the world, and the impoundment of the massive Three Gorges Reservoir (hereinafter referred to as TGR) that began in 2003 required the relocation of 1.2 million people (Yin et al., 2016). Water levels subsequently rose as much as 107 m over the impounded ground, creating a 660 km-long backwater area (Wang et al., 2014) exposed to geohazards. Even prior to the impoundment, an average of approximately five major geohazard events per year had occurred in this area. These geohazards were mostly in the form of landslides and rock avalanches, but debris flows, ground fissures, ground subsidence, karst collapses and earthquakes had also been reported (Wang and Li, 2009).

The reservoir impoundment has clearly increased the frequency of geohazard events in mountainous, tectonically active TGR area. Most significant was the reactivation of large landslides, whose newly submerged toes became subject to fluctuating hydrologic conditions (Song et al., 2018). An extreme case was the fatal Qianjiangping landslide and its associated, 30-m high impulse wave, which occurred shortly after the initial TGR impoundment, causing 24 deaths, destroying 346 houses, and capsizing many ships (Jian et al., 2014; Tang et al., 2017). The resettlement program for the relocated people required many new development projects and constructions in TGRA, and the associated excavations, road building and side casting produced many oversteepened, potentially unstable slopes. Moreover, many faults, karst caves and mining chambers exist in TGRA, and they can be affected by the pressure of the impounded water. As a side effect, the frequency and magnitude of earthquakes has clearly increased in the TGRA, similarly to what had been observed in other cases of induced seismicity following large reservoir filling (Simpson et al., 1988).

A great deal of scientific research has been undertaken to identify and mitigate TGRA geohazards, starting from the planning phase of this huge construction (Three Gorges Dam) project and continuing until the present time. More than 4,000 articles have been published on various aspects of TGRA geohazards. These articles focused mainly on 1) Characterizing TGRA geohazards; 2) Evaluating how they evolved and behaved; 3) Designing, testing and installing preventive stabilization measures, and 4) Monitoring geohazards and implementing early warning protocols to prevent potential catastrophic events. This paper attempts to summarize the massive corpora of research work, with the goal of gaining insights toward effective prevention and mitigation of TGRA geohazards. The paper also discusses challenges facing the engineering geologists in the prevention and mitigation of TGRA geohazards and offers suggestions for further research toward the goal of disaster reduction.

2. Geological setting and geohazards in TGRA

2.1. Geological and environmental background of TGRA

TGRA includes the regions of Yichang, Zigui, Xingshan, Badong, Wushan, Wuxi, Fengjie, Yunyang, Wanzhou, Kaixian, Zhongxian, Shizhu, Fengdu, Changshou, Fuling, Wulong, Yubei, Banan, Chongqing and Jiangjin (Fig. 1). The total lengths of the banks of the mainstream (the Yangtze River) and its tributaries are about 660 km and 1840 km, respectively.

The landscape of TGRA was shaped by major tectonic events. Yanshan orogenyin the Late Jurassic formed the terrain skeleton of mountains. Following the Himalayan orogeny in the Neogene, long-term erosional processes gradually transformed the area into the present-day landscape with moderate- to low-altitude mountains and river valleys. The section from Fengjie to the west of Zigui is topographically the highest, creating the famous Three Gorges (Qutang Gorge, Wu Gorge, and Xiling Gorge). The elevation declines westwards and eastwards from the highest part, forming hilly landscape and moderate-altitude mountains, respectively. The trend of the mountains is controlled by major geological structures.

The age of the geological units varies from the pre-Sinian to the Quaternary. There are no units of the Upper Silurian, Lower Devonian, and Upper Carboniferous age. The so-called red strata are widespread in TGRA, accounting for approximately 72% of the total length of the TGR banks. Here, the red strata refer to sandstone, mudstone, and sandstone interbedded with mudstone layers. The red strata were deposited in the Jurassic and Triassic periods. The red strata of Jurassic age are predominant in TGRA, mainly exposed in the west of Fengjie and eastern Zigui (Fig. 1). The Triassic age red strata appear only in some parts of Badong and Zigui regions (Badong Formation). In addition to the red strata, other sedimentary rocks (limestone, marlstone, and dolostone) are also present in the area between Fengjie and Zigui. These hard rocks form the steep gorges and valleys in the Fengjie-Zigui area. Metamorphic complexes and magmatic rocks crop out in a relatively small area near the dam site.

The main geological structure of TGRA is a prominent fold belt, which reflects the influence of multiple tectonic events. Starting from the west part of TGRA, the fold belt changes its trend from north-south to east-west direction and joins the Zigui syncline in the east. Ample evidence of intensive tectonics is observed in the area east of Fengjie, which includes large-scale structures such as the Huangling anticline, Zigui syncline, Guandukou syncline, Xiannvshan fault, Jiuwanxi fault and Tianyangping fault.

The TGRA is in the middle part of China and has a wet subtropical climate; it is warm and humid with abundant rainfall. Monsoon leads to a notable variation of heat from city to city during the year, while the rainfall is mainly concentrated in summer. The largest precipitation is in Wanzhou, with about 1930 mm annually (He et al., 2008). Moving away from Wanzhou to either part (east or west), the precipitation shows a decreasing trend. The smallest precipitation (about 996 mm annually) is in Zigui (Fig. 1).

Since the impoundment, the water level of the TGR progressed through three stages (Fig. 2). The first stage was the trial reservoir impoundment from April 2003 to September 2006 when the water level was raised from 69 m to 139 m ASL (above sea level) within the first two months, and then varied slightly. The second stage was from September 2006 to September 2008, when the water level was raised from 139 m to 156 m ASL in one month, and subsequently varied annually between 145 m and 156 m ASL. The third stage began when the maximal water level of the reservoir was raised to 172 m ASL in 2008, keeping the fluctuation between 145 m and 172 m ASL from 2008 to 2010. Thereafter the water level is managed to fluctuate annually between 145 m and 175 m ASL. During the fluctuation process, the water level goes through a sequence of stages, including slow drawdown, rapid drawdown, low level, water rising and high level.

2.2. Characteristics of TGRA geohazards

Landslides, rock avalanches, debris flows, ground fissures, and ground subsidence are the main instability phenomena in TGRA. There are 4429 geological hazards in TGRA from Yichang to Jiangjin along the Yangtze River. Among them, 4256 are landslides and rock avalanches with a total volume of about 4.24 billion m3, and the rest includes 58 debris flows, 42 ground fissures and 73 cases of ground subsidence.

2.3. Characteristics of landslide and rock avalanche hazards

The characteristics of landslide and rock avalanche hazards in TGRA can be summarized as follows:

(1)
Spatial distribution

Geohazards tend to concentrate along the Yangtze River and some of its tributaries but are most common to the east of Wanzhou (Fig. 1). Landslides and rock avalanches are particularly abundant in Zigui and Badong counties (Fig. 3). Considering the tributary rivers, the geohazards are most frequent in the catchments of Xiangxi, Guizhou, Qinggan, Caotang, Meixi, and Wujiang rivers, accounting for 44.3% of the total number of landslides and rock avalanches and 63.4% of the total volume of the geohazards present along the Yangtze River tributaries.

(2)
Slide-prone strata

Lithology is a major factor controlling the distribution of TGRA geohazards (Li et al., 2018). The sandstone, mudstone, and sandstone interbedded with mudstone layers of the Triassic Badong Formation and the Jurassic strata are known as the most slide-prone strata, bearing numerous rock avalanches and landslides. The volume of the rock avalanches and landslides identified in these strata accounts for 87.3% and 91.1%, respectively, of the total volume of TGRA rock avalanches and landslides.

No large rock avalanches or landslides are associated with the Precambrianmagmatic and metamorphic rocks, and only small bank slump occurred in the weathered materials near the dam. Moreover, slope instabilities are relatively infrequent in the carbonate rock area that hosts the famous Three Gorges. Less than 8% of the total volume of identified landslides and rock avalanches are associated with these rocks.

(3)
Influence of geologic structure

The landslides in the TGRA are likely to occur where strata dip at moderate angles toward the rivers, particularly on the flanks of anticlines and cores of synclines. These settings are ideal for the development of large consequent bedding landslides. Two prominent examples of such TGRA landslides are the Qianjiangping and the Huangtupo landslides developed, respectively, in the Zigui and the Guandukou synclines.

2.4. Characteristics of reservoir-induced earthquakes in TGRA

Seismic records from Hubei Institute of Seismology, China Earthquake Administration, show that 192 earthquakes with magnitude (M) ≥3.0 occurred in TGRA between 1995 and 2018 (Fig. 4). Among these, 170 earthquakes had magnitudes of 3.0 ≤ M < 3.9, while 18 had 4.0 ≤ M < 4.9 and four had M ≥ 5.0.

As illustrated in Fig. 4, the frequency and intensity of earthquakes were low before the first TGRA impoundment (2003). Based on the monitoring data from 1995 to 2001, the average number of M ≥ 3.0 earthquakes per year was 2.4, except in 2001 (11 events). In those seven years, the magnitude of the largest earthquake was 4.4 (1997 event), and only three events had magnitude >4.0.

Seismic activity did not change during the trial impoundment period (June 2003 to September 2006), when the water level first rose to 135 m ASL. Moreover, seismic activity increased only slightly during the second impoundment stage (September 2006 to September 2008), when the water level rose to 156 m ASL. However, seismic activity increased significantly following the final impoundment stage (after September 2008); approximately 14 events per year with M > 3.0 now appears to be a norm, compared to 2.4 events per year prior to the impoundment. In addition, four events with M ≥ 5.0 occurred in this stage, compared to none between 1995 and 2008. Clearly, both frequency and intensity of the earthquakes have risen since the full impoundment in September 2008.

3. Trends of research on TGRA Geohazards

3.1. Research and engineering work relevant to TGRA geohazards

TGRA is extremely important for the study of geological disasters. Several large-scale research projects were funded to examine the key scientific and technological issues for the prevention and control of geohazards in this area. Important research results were obtained, including those related to the triggering mechanism of geohazards, geotechnical properties of special lithologies, prevention and control methods, monitoring and large-scale field experimental stations. The major research and development activities and their outcomes are summarized as follows:

(1)
In-situ investigation, multi-parameter monitoring, and physical model tests were conducted to study the behavior and evolution process of TGRA geohazards. The established tempo-spatial evolution mechanisms and deformation-failure modes show how rainfall and reservoir water level fluctuations affect the geohazards (Tang et al., 2015a, Tang et al., 2015b, Tang et al., 2015c; Wang et al., 2016; Wu et al., 2019; Song et al., 2018). Prediction and evaluation models for landslide-driven impulse waves in the reservoir were also developed (Yin et al., 2012; Huang et al., 2012).
(2)
The engineering geological properties of the slide-prone strata and the key areas of large landslides were examined. The degradation and disintegration processes under the effect of the reservoir water fluctuation were studied, focusing on the rock masses from the red strata of the Jurassic age and the Triassic Badong Formation (Jiao et al., 2014; Shen et al., 2019). The shear and rheology properties of the slip zones in large landslides were investigated through in-situ triaxial creep tests, which provided data for further analysis of TGRA landslides (Tan et al., 2018a, Tan et al., 2018b, Tan et al., 2018c).
(3)
Different control methods, including deep drainage, anti-sliding piles and anchor cables were successfully applied to mitigate large TGRA geohazards. The successful mitigation projects, including those to stabilize the Lianziya unstable rock mass and the Hongshibao landslide, provided valuable experience for the prevention and control of geohazards in the area. Through these on-site demonstration projects, standards were established for the prevention and control of TGRA geohazards.
(4)
A modern monitoring network of TGRA geohazards was constructed by combining systems of automatic monitoring, real-time information release and remote transmission. An early warning system was established based on the identification of the geohazard evolution stage (Xu et al., 2008). Many TGRA geohazards were successfully forecasted, such as the Xintan landslide, based on comprehensive monitoring and warning systems.
(5)
Large-scale field experimental stations for the study and monitoring of TRGA geohazards were constructed. Among them was the Badong field experimental station (Tang et al., 2015a, Tang et al., 2015b, Tang et al., 2015c), which is the largest experimental facility ever built inside a landslide body. The Badong station was constructed to provide support to research, teaching, and risk communication of TGRA geohazards. The Badong experimental station has marked one of the most significant advances for the discipline of engineering geology in China.

3.2. Trends of TGRA geohazards research based on scientific literature survey

Numerous research results have been published in journals or other media to elucidate TGRA geohazards. A bibliometric analysis was applied to a large sample of relevant articles from Science Citation Index Expanded (SCIE) database and Chinese National Knowledge Infrastructure (CNKI) database. These two databases cover the vast majority of articles that are related to TGRA geohazards.

The time span of this survey was limited to the period of 1981–2019, since few articles were found before 1981 in the above databases. Before the search process, a series of terms related to the geohazards were selected. Publications were found by searching the term “three gorges” and one of following terms: “geohazard”, “hazard”, “landslide”, “slope”, “rockfall”, “rock avalanche”, “ground fissure”, “earthquake”, “debris flow”, “impulse wave”, “tsunami”, “rock mechanics”, “rockmass mechanics”, “soil mechanics”, “factor of safety” and “susceptibility” in the title, abstract or keywords of a paper. Eventually, 1002 publications and 2857 publications were identified in SCIE and CNKI respectively.

The statistics regarding the five most investigated types of geohazards, i.e., landslide, earthquake, impulse wave, rock avalanche and debris flow, are shown in Fig. 5. Landslide is the most frequently studied geohazard topic in the TGRA literature, followed by earthquake, impulse wave, rock avalanche and debris flow. Articles in English and in Chinese reveal the same trend.

The most frequently used keywords were obtained from an open-source search results clustering engine (CARROT2; https://project.carrot2.org/). Fig. 6 shows the results of this search, in which “Reservoir landslide” is the most frequently used keyword. Focusing on TGRA geohazards, the most mentioned terms in these articles include landslide monitoring, landslide stability, landslide deformation, landslide displacement and mechanism.

4. Assessment of TGRA Geohazards

4.1. Assessment of reservoir landslides and their evolution mechanisms

Evolution mechanism is an important aspect in the study of reservoir landslides. The dynamic characteristics of landslide evolution and the influencing factors have been widely investigated in TGRA-related studies.

4.1.1. Dynamic characteristics of TGRA landslide evolution

(1)
Deformation process of landslides

Deformation and failure processes of reservoir landslides were studied via site survey, monitoring, physical model test and numerical simulation. Many TGRA landslides were affected by the periodic variations of water level and rainfall. Some landslides deformed intermittently and showed step-like cumulative displacement curves, such as the Baishuihe landslide (Li et al., 2010), Shuping landslide (Wu et al., 2019) and Outang landslide (Yin et al., 2016). Other TGRA landslides deformed continuously and had linear-like curves of cumulative displacement, such as the Majiagou landslide (Ma et al., 2017a, Ma et al., 2017b).

Some investigators used physical model tests and numerical simulations to study and reconstruct the evolution process and deformation mechanism of landslides, while others used finite element methods to study the coupling effect of rainfall and reservoir water level variation (Jiang et al., 2011; Zhao et al., 2017). Rheology model was adopted to study the mechanism of the Anlesi landslide (Jian et al., 2009). Universal discrete code was applied to study the effects of rainfall and water level fluctuations on the behavior of the Quchi landslides (Huang et al., 2018).

(2)
Mechanical characteristics of reservoir landslides

Based on the role of water in the deformation of reservoir landslides, TGRA landslides can be classified as seepage-induced or buoyancy-induced landslides. In the seepage-induced landslides deformation mainly occurs when the reservoir water declines (Fig. 7A), while in the buoyancy-induced landslides deformation mainly occurs when the reservoir water level is high (Fig. 7B). The seepage-induced landslides are more prone to form in low permeability sliding material. Porewater pressure is difficult to dissipate when the reservoir water level drops suddenly, and this produces outward seepage force that tends to destabilize the landslide. The two types of deformation behavior are illustrated by the Baishuihe landslide (Fig. 7C) and the Muyubao landslide (Fig. 7D).

In addition to the hydro-mechanical effects caused by the fluctuation of water level, the mechanical properties of slip zone were found to be a controlling factor in the deformation and failure process of landslides. Ring shear tests were conducted to investigate the residual strength of the slip zone of the Qianjiangping landslide (Wang et al., 2008). Rheology tests, including ring shear creep test (Wang et al., 2018) and triaxial creep test (Miao et al., 2014; Li et al., 2019a, Li et al., 2019b), were conducted to characterize the long-term strength and creep behavior. A large-scale in-situ triaxial creep test was carried out in the slip zone of the Huangtupo landslide at the Badong in-situ experimental station (Tan et al., 2018a, Tan et al., 2018b, Tan et al., 2018c). The weakening of the slip zone material was studied by wetting-dry cycling tests, simulating the condition of periodic water level fluctuation (Jiao et al., 2014).

4.1.2. Effect of rainfall and fluctuating water level on the evolution of TGRA landslides

(1)
Effect of rainfall

Rainfall was found to be one of the most important causative factors of TGRA landslides. Different workers used field monitoring to investigate the water infiltration mechanism and effect of rainfall on slope stability. Observation wells were constructed in typical colluvial landslides (Tang et al., 2015a, Tang et al., 2015b, Tang et al., 2015c) and decomposed granite slope (Zhang et al., 2000). The rainfall infiltration process was studied using the monitoring data on suction and temporary pore pressure variations during the rainfall. Additionally, the influence of rainfall characteristics (amount, intensity, distribution pattern) on slope stability were studied (Yang et al., 2017; Liu et al., 2018).

(2)
Effect of reservoir water level flunctuation

The periodic TGR water level fluctuation can alter the seepage field, the stress field and the material properties, resulting in the change of slope stability. Such findings were reported by Hu et al. (2012), Zhang et al., 2014a, Zhang et al., 2014b and Zhao et al. (2017). The effect of rising and falling water level was also studied through model test or numerical simulation considering different rates of water level variation (e.g., Sun et al., 2017). Further, Liao et al. (2005)demonstrated that slope stability was closely related to the variation of the permeability coefficient of rocks.

(3)
Combined effect of rainfall and water level fluctuation

In many cases, the evolution of reservoir landslides depends on the combined (coupled) effect of rainfall and water level fluctuation. Some conceptual models were established to calculate the phreatic line (Wu et al., 2009; Sun et al., 2016a, Sun et al., 2016b). Numerical simulations were used to study the characteristics of seepage field based on saturated-unsaturated theory (Miao et al., 2018). Many TGRA landslides have been continuously monitored. Based on the monitoring data, the influence of rainfall and water level fluctuation on the landslide deformation was analyzed (Li et al., 2010; Tang et al., 2015a, Tang et al., 2015b, Tang et al., 2015c; Wang et al., 2017). In addition, data miningmethods, such as grey relational grade analysis, two-step cluster analysis and Apriori algorithm analysis, were applied to examine and quantify the influence of rainfall and water level variation on landslide deformation (Ma et al., 2017a, Ma et al., 2017b; Tan et al., 2018a, Tan et al., 2018b, Tan et al., 2018c).

4.1.3. Stability evaluation methods

The methods used in the evaluation of TGRA landslides transitioned from qualitative to quantitative approaches and from the deterministic to the probabilistic approaches.

(1)
Qualitative methods

The qualitative methods, including engineering geological analogy, graphic method and expert system, are typically used in the early stage of landslide evaluation (Zhou, 1996; Zhu et al., 2014). Moreover, classification studies based on database management system (DBMS) and geological information system (GIS) were employed to evaluate the stability of slopes (Fourniadis et al., 2007).

(2)
Limit equilibrium methods

The limit equilibrium methods applied in the stability evaluation of TGRA landslides include Swedish method, Janbu method, Bishop method, M-P method and Sarma method (Ma and Liang, 2002; Liu et al., 2007a, Liu et al., 2007b). Some of these methods were further modified based on simplified boundary conditions and/or additional assumptions (Yong et al., 2016; Wang et al., 2017; Tang et al., 2017). Furthermore, the two-dimensional limited equilibrium method was extended to three-dimensional limited equilibrium method (Wang and Xu, 2013; Sun et al., 2016a, Sun et al., 2016b).

(3)
Numerical simulation methods