The Zhongbao landslide is located on the north side of the Yancang river, about 59 km to the Wulong District, Chongqing, China (Figure 1A). In plan view, the shape of the Zhongbao landslide can be approximately likened a tongue, being 850 m in length and 200 m in width (Figure 1B). The thickness of the landslide is estimated to be 25–47 m, and its’ volume approximately 4.6 × 106 m³ (Figure 1C). The slope of the landslide surface is 25°–45°. The main sliding direction is 210°, approximately perpendicular to the Yancang River. The Zhongbao landslide began to slide at 16:00 on July 23, 2020, accelerated and slid into the river on July 27, and destroyed three buildings and main roads. Over 1.0 × 104 m³ of soil and rock mass slipped into the river and formed a barrier lake.

The field investigation indicates that the main geologic units of the area are deposits and shales. The uppermost part of the surficial deposits are eluvial deposits (up to 0.5∼3.5 m thick), whereas stratigraphically lower layers comprise mostly alluvial deposits and colluvial deposits. The eluvial deposits consist of clay and 10∼45% of gravel clasts (Figures 2A,B). The diameter of the gravel clasts ranges from 50 to 300 mm. In-situ exposed outcrops over 10 m high (undisturbed by the landslide) crop out near the western limit of the depictions of the lithological structure (Figure 2C). Dominant pale-yellow shales, 0.2∼0.5 m thick, are found in the pelitic structure with developed joints. The main component of shales is smectite and other clay minerals, which are easily weathered and softened by water. The rocks dip 25° toward 130°. Two dominant joints are found in the study area (Figure 2D). The first joint dips 75° toward 5°–15° with a width of 3∼15 mm and a length of 1.5–4.9 m, and the joint spacing is 1.30–4.20 m. The second one dips 80° toward 210°–230° with a width of 2–12 mm and a length of over 10 m, and the joint spacing is 1.10–2.70 m. The joints are partially filled with clay.

The Zhongbao landslide was first found to have deformed on July 22. Several cracks were recorded on the uphill side of the landslide after a continuous intense rainfall. As the landslide deformed, the eastern boundary of the landslide was generated, then a great many tension cracks formed at Part II while Part I gave way (Figure 4B). The uphill slope slid over 20 m, and the downhill slope suffered local deformation. On July 24, the sliding mass was transported to the middle and downhill slopes, causing the downhill slope to gradually collapse (Figure 4C). On July 25, there was a strong rainfall, the downslope had obvious deformation, and the sliding mass gradually slid into the Yancang River. The barrier lake and dam were formed (Figures 4D,E).

The steep Part I firstly deformed and collapsed, which destroyed the main road. The upstream portion of Part II gradually deformed and pushed the middle portion to slip. Due to the cliff at the east side of the landslide, the sliding direction of the landslide gradually changed from 165° in the source area to 220° in the transportation area, and finally to 180° in the accumulation area. According to the deformation characteristics of the landslide, the landslide area could be divided into five areas (Figure 9):

Part I: This portion is located at the northeast side of the landslide (Figure 5). Reactivated deposits of Part I covered Part II. This part is 140–245 m wide and 230 m long, about 10 m thick with its largest elevation difference being 102 m. The deformation characteristics of the road showed that the accumulation moved with Part II, inferring that Part I failed before Part II. Due to the steep slope, engineering projection, and continual rainfall, Part I slipped with velocity and compressed Part II, which accelerated the movement of Part II. Therefore, the uplift area and puddle were found in part II.

Source area (Figure 5): The length and elevation difference of this area is approximately 207 and 46 m, respectively. Under the effect of rainfall and part I, the rear section of Part II gradually deformed and developed into scratch (Figure 6A) and tension cracks. Those cracks became an infiltration path for rainfall which damaged the yale structure and decreased the shear strength of the soil and rock. The rock became severely fragmented (Figure 6B). The deformation of the landslide thus accelerated. Many scarplets and cracks were found in this portion. Moreover, the boundary between the source area and the transportation area was an uplift zone with an average height of 3.0 m (Figure 5).

Transportation area (Figure 7): Owing to the rear section sliding and pushing the rock soil mass in the transportation area, some buildings and houses in the middle section only formed tension cracks but were not completely destroyed (Figure 7). Some parts were uplifted. In the plane view, this area had an approximately rectangular shape with a length of 340 m and an average width of 175 m. Some pools were found on the landslide. The road debris showed that it slid about 99 m but kept good continuity.

Accumulation area (Figure 8): The rear part became steep again, and the dip direction of the exposed rock was 15°. Thus, the sliding direction of the landslide changed. The sliding mass slipped and mobilized surficial deposits in the accumulation area by scraping, entraining, and pushing them downward. Finally, the reactivated deposits accumulated in the toe of the landslide and formed a barrier dam and barrier lake. The thermal infrared monitoring found that the temperature field appeared abnormal in the place where the groundwater was exposed. Moreover, the temperatures of the accumulation area were higher than in other areas.

Secondary-landslide area: Three secondary-landslides (1–3# in Figure 5) were found at the west side of the landslide because of the V-shaped valley. No. 1 secondary-landslide was caused by the source area sliding along the direction of 230°. No. 2 is located on the cliff, so could be from the collapse of the sliding mass. No. 3, caused by a change in the sliding direction of Part II, occurred and buried a small road.

1) Visual observation of surface deformation

The radar technology was used to evaluate the volume and deformation characteristics of the landslide because it can remove the trees on the landslide surface and obtain the rear geography of the landslide. Figure 9 shows that Part I had an obvious boundary with 4.1 × 102 m². Its sliding mass accumulated in Part II and slipped to the transportation area of Part II. The downhill slope continued onto progressive retrogressive collapse and slid into the river. Thus, this landslide was still unstable. Moreover, a cliff was observed on the northeast side of the landslide with over 1 × 106 m² colluvial deposits where there was also the occurrence of Part I.

2) Surface displacement

An emergency monitoring was started from July 27 to Aug 2. Ground-based radar was 2.8 km away from the landslide (Figure 3). The surface deformation of the landslide was scanned from July 27 to October 2. Unfortunately, the monitoring was non-continuous due to the power issues and a change in the device location. Therefore, two periods are selected to analyze the deformation characteristics of the landslide debris.

Figure 10 is the monitoring results of the landslide surface from 16:00 July 27 to 06: 00 July 28. The uphill and middle of the landslide showed no obvious deformation, but the downhill section had large deformation with the largest displacement being 200 mm (Figure 10A). The displacements of four monitoring points in the downslope are shown in Figure 10B. It is observed that the displacement of P2 and P4 slowly increased and reached about 50 mm. However, the displacement at P1 and P3 had the largest values. The velocity of P1 gradually decreased after July 28. The displacement at P3 sharply increased at 22:00 July 27 and then tended to be stable. Moreover, local deformation was also found at the cliff on the northeast side of the downslope (approximately −50 mm).

Figure 11 shows the displacement of the landslide surface from 11:00 to 17:00 July 28. It was found that the largest displacement of the landslide is about 70 mm, which is clearly smaller than that in Figure 10 (Figure 11A). The displacement of the point at P was monitored (Figure 11B). The displacement increases with a velocity of 12 mm/h, which indicated that local collapse was still occurring Figure 11C shows that the upslope also had small deformation (>40 mm) and the whole landslide was stable.

In consequence, after the landslide collapse, this landslide tended to be stable except for local collapses. The emergency works were taken in the source and transportation area before July 28, and then, the barrier dam was excavated on July 28. The traditional monitoring equipment was installed on the landslide surface.

In Figure 12, the rainfall and displacement from Aug 12 to Oct 1 was recorded. The second wet month (from September to October) was characterized by a 333 mm cumulative rainfall. The downhill slope (G05, G06) had large deformation with a velocity as high as 6 cm/day (G06). The increment displacement of the G06 suddenly decreased tending to zero, but the G05 went on with a constant rate of 3.8 mm/day. Clearly, the rainfall reactivated the landslide. Moreover, the increase of the displacement at G06 and G05 generally lagged 1 day behind the rainfall. Another portion of the landslide had a smaller displacement (<30 mm/month), but also responded to the rainfall and gradually increased.

Water has a significant effect on the strength of shale rock. The increase of the water content could lead to mineral particle dissolution and the increase of clay minerals. As a result, the shale structure becomes loose, the mic-cracks gradually increase, and finally, the rock degrades and softens (Feng and Ding, 2005; Cornelius et al., 2011; Zhu et al., 2015; Ling et al., 2016; Liu et al., 2016; Xue et al., 2018; Li et al., 2019). Moreover, a filed investigation found that the primary porosity of shale in the landslide area was well developed, and three sets of the superior fractures caused the loosening of the rock. Therefore, after a constant rainfall (July 15–18), the rainfall infiltrated into the landslide and decreased the shear strength of the rock. Further, underground water had a lubrication effect on the rock layers, so that the landslide began to deform on July 22. The displacement of the monitoring results also showed that the failure landslide had a large deformation after a strong rainfall (Sep 6). Moreover, the lag time of the landslide deformation after failure (about 1 day) was shorter than that of the landslide before failure response to rainfall (about 6 days), which could be because of the loose deposits after landslide collapse accelerated the rainfall infiltration. According to the deformation characteristics of the landslide, the deformation process could be divided into three stages:

Creep deformation stage: Rainfall caused a shallow landslide (Part I). A continuous rainfall lasted for almost 20 days and increased the groundwater level. As a result, the weight of the sliding mass was increased and the shear strength of the soil decreased. Moreover, the main road provided beneficial conditions for Part I failure. The reactivated deposit impacted and compressed the upslope of part II which accelerated the deformation of the landslide (Figure 4B). Besides, the gentle slope in the middle of the landslide is conducive to the accumulation of groundwater. Therefore, under the action of groundwater immersion, the landslide gradually moved. Thus, the scratch and shear cracks were generated at the uphill side of the landslide (Figure 6). The uphill section gradually generated obvious deformation. The deformation direction of part I is 230°.

Accelerated deformation stage: After the rear part of the landslide was initialized, it gradually slid along the sliding zone and compressed the middle section. Although there was no rainfall, the surface water continually infiltrated along these shear cracks, and the shear strength of the soil continually decreased under groundwater immersion. The sliding zone rapidly developed and the landslide slid over 90 m (Figure 4C). The sliding direction was 210°.

Accumulation stage: the landslide slides out from the steep ridge of the front edge and pushes and scrapes the shale (Figure 4E), which accumulates at the foot of the landslide and a barrier lake is formed. Moreover, the scarplet in front of the landslide led to the deposit mainly accumulating in the landslide toe and forming a secondary landslide (Figure 9) which still deformed after the landslide failure, especially when heavy rainfall occured.

The pale-yellow shale samples, with different water content and angles of bedding planes, were tested to obtain the compressive strength by uniaxial compression test (shown in Figure 13). All curves had the same shapes that as the water content increased, the compressive strength gradually decreased (Yang et al., 2007; Wei et al., 2015. The angle of bedding planes also had significant importance on the strength, because the clay minerals filled in the bedding plane, the content of which could be over 35% clay minerals (Fang et al., 2019), such as kaolinite, montmorillonite, illite, absorbed water and swelled, causing expansive pressure. As a result, the shale rock gradually split and disintegrated along the bedding plane (Rybacki et al., 2015; Rybacki et al., 2016; Teng et al., 2017), which also provided further seepage channels. Moreover, two clay layers (0.5–1.5 m thickness) were observed in the shale rock layer, one of which was the sliding surface. According to the other reports, the shear strength properties of the clay have a significant decrease when the moisture content increases (Dafalla, 2013; Li et al., 2018b; He et al., 2018), and are more sensitive than the rock. Therefore, after intense rainfall, the landslide slipped along the clay layer, and the shale rock was fragmented in the sliding mass.

After the landslide failed, the barrier lake formed in front of the landslide increased the groundwater level of the accumulation area (Figure 14). The reservoir impoundment could accelerate the landslide deformation and failure (Macfarlane, 2009; Huang et al., 2016; Li et al., 2018b; Wang et al., 2019). When the lake flooded the downhill portion of the deposit, the shear resistance was reduced (He et al., 2018), especially for the destroyed deposits, so that the deposit still had a strong response to the rainfall. In the collapse at the shear outlet of the landslide, several springs with a flow velocity of 0.1–0.6 m³/s indicated that the seepage field rapidly declined which could decrease the stability of the landslide (Berilgen, 2007; Della Seta et al., 2013; Hu et al., 2017; Sui and Zheng, 2017; Zhang et al., 2018b; Wu et al., 2020; Yang et al., 2020). The stability of the accumulation area was evaluated using Geo-slope software. Firstly, we evaluated the safety factor of the accumulation area under rainfalls (Figure 15). The safety factor of the accumulation area had a strong response to the rainfall and presented a certain linear change as the displacement increased. During an intensity continuous rainfall, the safety factor significantly decreases, because the groundwater level rapidly increases. If the dammed lake is excavated and drained during this stage, the landslide could slide again and form a larger barrier dam.

After landslide failure, the re-activated deposit and the barrier lake could cause secondary economic losses (Fan et al., 2015; Gischig et al., 2016; Liu et al., 2020), so emergency engineering is necessary (Xiao et al., 2021). This paper constructed an integrated monitoring system for landslide failure. The monitoring equipment is non-contact measurements such as Ground-based radar, Unmanned Aerial Vehicles, Airborne Lidar, and thermal infrared temperature monitoring. These devices can scan the whole landslide surface and obtain point cloud data to evaluate the stability of the landslide debris in detail. The monitoring results found that deformation mainly occurred in the scarp at the uphill side of the landslide, and the deformation gradually tend to become stable after July 28. Those results provided guidance for rescue work and the excavation of barrier lakes. However, these devices are too expensive and difficult to operate if used for long-term monitoring of landslides, and also in addition only obtain deformation information of the landslide surface. Therefore, After landslide failure gradually stableizes, conventional sensors should be installed into the landslide to measure the displacement of the typical landslide profiles, such as GNSS, GPS, deep monitoring, and so on. Those devices can measure the multi-filed information of the landslide for a long time without interruption, which could provide information for reconstruction work. In this paper, the monitoring for the failing landslide shows that the failing landslide debris is still responsive to rainfall, especially at the front and middle parts of the landslide. Thus, engineering activities and human activity should be avoided as far as possible in the middle and front parts as the landslide could slip again and form a dammed lake. Moreover, over 1 × 106 m² of colluvial deposits on the northeast side of the landslide could be initialized and fail under the long-term effects of rainfall, which could cause extensive damage to the landslide. Therefore, the most effective monitoring method is non-contact devices, such as ground-based radar, unmanned air vehicles, three-dimensional laser scanners, and so on.