Characteristics and mechanism of landslides on highway land�ll along Xiaolangdi Reservoir of the Yellow River:A case study

: On February 17, 2023, a landslide occurred on a highway slope located on the bank of 11 Xiaolangdi Reservoir of the Yellow River, with a volume of about 1,200 m 3 , which directly led to 12 the interruption of the road and had a serious social impact. In this paper, the area where the 13 landslide was located was studied and analyzed by means of field investigation, remote sensing 14 images, laboratory tests and numerical simulation. The evolution pattern of landslide groups under 15 the influence of human engineering activities is traced, and the disaster - causing mechanism of 16 landslides is investigated in terms of climatic factors, water level changes, stress changes, and the 17 nature of the fill. The results indicate that: (1) Water level changes can lead to a decrease in the 18 stability coefficient of ancient landslides, but have no direct effect on the occurrence of landslides. 19 The change of soil stress caused by artificial excavation will change the distribution of soil plastic 20 region, which makes the potential sliding surface changing; (2) Rainfall and temperature are the 21 key elements affecting the fill - type landslides, and the alternating period between winter and 22


Introduction
With the ongoing development of society and the economy, the alteration of the natural environment through human engineering activities has become increasingly significant.Moreover, climate change has rendered the geological environment more fragile, giving rise to an annual escalation in various geologic disasters induced by human engineering activities.According to the United Nations Office for Disaster Risk Reduction's (UNISDR) forecast, it is anticipated that the occurrence of sizeable and medium-sized disasters around the world will rise to 560 annually by 2030.This translates to an average of 1.5 disasters every day, posing severe threats to both engineers' safety and the wellbeing of individuals and their assets.
Reservoirs, road, and railway construction encompass a vast range of areas and involve lengthy construction periods and long-lasting effects, making them the most prone to trigger geological disasters.Among these, landslides within the reservoir fallout zone garnered significant attention and display a certain periodicity.In severe cases, they can obstruct waterways and create weirs, further instigating secondary disasters (Xu et al. 2009, Liu andHe 2018).The deformation of bank slopes around reservoirs is strongly linked to alterations in water levels.Numerous studies have revealed that the primary factor leading to landslides in these slope areas can be attributed to variations in internal and external infiltration pressure of the soil, caused by fluctuations in water levels (Han et al 2018, Dai et al 2022).Furthermore, such water level changes initiate buoyancy forces in the soil mass (Yi et al 2022).
In addition, it is important to consider alterations in soil properties and the water table resulting from rainfall (Wang et al., 2022;Hou et al., 2022).Recently, the majority of research on landslide tragedies affecting reservoir escarpments has centred on the Yangtze River Basin in China, particularly the Three Gorges Reservoir (Wang et al., 2004;He et al., 2010;Zhang et al., 2020;Deng et al., 2023).The landslides are predominantly characterized by Quaternary debris accumulations, with a few occurrences in the red beds (Yin et al., 2020).
There is a dearth of research on the risk of landslides on the banks of reservoirs in the Yellow River Basin of China, particularly in the midstream where the river traverses the Loess Plateau.
The geological environment in this area is extremely delicate and vulnerable to geological calamities (Hu et al., 2022;Lan et al., 2022).There are two significant hydropower stations in the central stretches of the Yellow River: Sanmenxia Hydropower Station and Xiaolangdi Hydropower Station.Both are positioned in the southern terminus of the Loess Plateau and have brought forth considerable catastrophes during development and storage activities (Kong et al., 2017;Liu et al., 2021).And there is also a large number of mining areas in the reservoir area (Data from Shanxi Coal Geological Bureau), which makes the evolution of geological hazards in the region more complex.
Additionally, linear projects such as highways and railways inevitably cross complicated terrain, resulting in a significant amount of excavation and backfill operations, which alter the terrain features and create a potential disaster hazard.During the initial construction phase, the road ditches slopes are shaped and reinforced to maintain their stability, and therefore, most road ditch landslides occur during the construction stage (Wei et al, 2012).Large-scale landslides also occur during the operational period on some roads with lower protection levels (Tang et al., 2019), but their incidence remains low.However, there are fewer studies concerning the development mechanisms and evolutionary patterns of road landslides, with a significant focus on regional risk assessment (Dhakal et al, 2020;Hussain et al, 2021).
Compared to road ditch slopes, embankment slopes consist mostly of man-made fill, making them vulnerable to geological disasters like subsidence and landslides due to differences in the nature of the rock and soil bodies, as well as constraints in construction conditions.This phenomenon is especially significant in the Loess Plateau (Jie et al, 2021;Yao et al, 2022).
Compared to slopes under natural conditions, the original structure of artificial fill has been lost, with changes to particle grading and altered physico-mechanical properties (Meng et al 2020).The more homogenous characteristics of artificial fill facilitate sliding surface prediction, but the influence of various factors makes accurate landslide timing prediction difficult.
Currently, numerous studies have examined reservoir landslides and highway excavation and fill landslides.Additionally, various scholars have used remote sensing to track the landslide development process.However, few studies have investigated the chain of landslide hazards induced by engineering activities under different factors.In particular, scarce reports are available on the deformation of bank slopes of reservoirs situated in the Yellow River Basin.Building upon the shortcomings of prior research, this study conducts a thorough investigation and analysis of landslides along the highway's slopes adjacent to the Xiaolangdi Reservoir of the Yellow River.
The research considers the historical landslides and the reservoir's receding zone.The study explores the climatic features and stress adjustments resulting from filling and excavation.A retrospective analysis of landslide clusters in the area is conducted, and the disaster-causing mechanisms and evolution mode are examined by integrating indoor experiments and numerical simulations.This study has the potential to enhance the findings in geohazard research whilst offering a disaster tracing methodology.This holds significant theoretical and practical benefits for future engineering construction and disaster prevention efforts.

Location
On 17th February 2023, a sizeable landslide took place on Highway G241 in Yuanqu County, Yuncheng City, Shanxi Province (Figure 1c).The landslide is situated in the southeast corner of the Loess Plateau, in the reservoir region of Yellow River Xiaolangdi Reservoir (N35°05′ 27.027″, E111°49′28.831″), with a mean elevation of 400 m.The landslide is situated 1 km from the Yellow River and close to the highway.The loess accumulation has a height of 50 m, a slope degree of 35 ° and measures 400 m in width and 200 m in length (Figure 1a).The main slip direction is NE150 °.Geological conditions in the area are complex with a peak acceleration of ground shaking of 0.10 g.The location of the landslide is in the proximity of the Yellow River.
The geological conditions wherein the landslide is situated are intricate, and the maximum acceleration of ground vibration is zero.The location of the landslide is on the second terrace of the Yellow River with a basic seismic intensity of VII degree.The surrounding topographic conditions are relatively simple with a height difference of approximately 100m and the ground's integrated slope ranges mainly between 10-25°.The lithology stratum is varied and the structure of the rock and soil bodies complex.The first slope near the road has neocene sandstone and sandstone outcrops.Near the road, there are sandstone outcrops of the Neoproterozoic system on the first-grade slope, which comprise easily weathered rock layers.A large thickness of fresh, brownish-red soil is piled on top of it.The soil has been artificially loaded, with a flat top surface and traces of ploughing.There is a small amount of snowmelt pooling at the front of the road.
Below the road lies the highway landslide unloading treatment area, equipped with comprehensive interception and drainage facilities.

Landslide groups
The area affected by the landslide is an ancient landslide, which has undergone multiple phases of sliding under subsequent geological activity.Based on the topographic results, it is evident that eight distinct signs of landslides have appeared on the ancient landslide body and this landslide is labeled as L8 (Figure 1d).The volume and characteristics of nine landslides were determined and are presented in Table 1.Among the successive landslides, L1 was the first developed landslide.With the construction of the reservoir in 2002, the water level of the Yellow River rose sharply, resulting in the landslides of L2, L5 and L6.Subsequently, numerous secondary slides were induced by engineering activities and water level changes during the following two decades.By examining InSAR surface deformation data from the previous three years, it was discovered that the slopes of the reservoir banks have experienced ongoing deformation (Figure 1b).Additionally, the leading edge of the ancient landslide body supporting the site of the current landslide has also undergone deformation, indicating that the entire area is currently in an unstable state.

Meteorological characteristics
Meteorological data for the study area from 2020 to 2023 indicate a temperature range of -20 to 40℃.The maximum daily rainfall in 2020 will remain between 40-50mm, while 2021 will experience up to 90mm in daily rainfall.In 2022, the maximum daily rainfall will decrease somewhat, but the overall temperature will remain above 80mm.Heaviest rainfall is primarily expected from June to September.According to an analysis of meteorological data from the month in which the landslide occurred, there was a small amount of precipitation between February 9 and February 14.This caused a significant decrease in temperature, with the average temperature dropping below 32℉.Following the end of the rainfall, the temperature began to rise, and the landslide occurred on February 17.

Laboratory test
Considering the impact of freeze-thaw cycles on bedrock, a freeze-thaw deformation test was conducted on sandstone.A fresh rock outcrop was chosen in the field, where the weathered rock surface was removed to expose the underlying fresh block.Following sample retrieval, several core samples measuring φ50×100mm were drilled along the bedding direction, dried at 105℃, and placed in a dryer.Using an X-ray diffractometer to test the mineral composition of the sample, it was determined that the rock consists primarily of quartz, feldspar, and calcite as secondary minerals, with small amounts of copper, pyrite, and chlorite present.According to the Engineering Rock Mass Test Standard (GBT50266-2013), the sample was initially soaked in an aqueous solution for 48 hours, vacuumed, and saturated.Thereafter, the surface water was wiped dry, and strain gauges were affixed in various positions.The top, side, and front of the sample were pasted with strain gauges, and numbered accordingly, as illustrated in Figure 5.After attaching the strain gauge, all samples were placed in an ambient chamber set to -20 °C for 4 hours.It is important to note that when the ambient temperature falls below 0 °C, the humidity defaults to 0%.Later, the temperature was raised to 0 °C and maintained for another 4 hours.Subsequently, the humidity was adjusted to 85% and the temperature to 20 °C for 4 hours.The humidity was kept unchanged, whereas the temperature was increased to 40 °C for 4 hours.This cycle of freezing and thawing, and alternating hot and cold cycles marks the end.The Yellow River The Land and the strain abruptly escalates.Subsequently, with each increase in the count of cycles, the rock sample's deformation rapidly escalates.In this experiment, we tested the deformation of three regions using a total of six strain gauges to measure the tensile and compressive deformation of the bedding in those regions.The results showed that T-1 and L-1 experienced substantial deformation, with T-1 having the largest shape variable.While other parts did not experience significant deformation, the overall results still followed a certain pattern.The variables are sorted in descending order from largest to smallest.It is evident that the bedding experiences significant expansion deformation during temperature fluctuations.As the number of cycles increases, the bedding's deformation transitions from initial elastic deformation to plastic deformation and culminates in a large, open fracture.After the 20th cycle, the specimen experienced deformation and temperature changes that resulted in the detachment of the strain gauge.The maximum deformation at the conclusion of the test was 2500.

Fig.8 Freeze-thaw cycle deformation curve
Taking T-1 as an example, it exhibited noticeable differences in deformation under varying temperature conditions within a single cycle (Figure 9).The process can be divided into five this stage, the deformation remains unchanged and does not increase significantly, indicating that the water has already frozen completely.As the temperature shifts from -20℃ to 0℃, the deformation decreases to some extent due to the partial melting of water within the rock layer.
This allows for recovery of the bedding deformation which marks the beginning of the weak freezing stage (b-c).The temperature then increases from 0℃ to 20℃ in the c-d stage, resulting in complete melting of the water in the bedding.However, the freezing and melting of water cause a change in the arrangement mode of mineral particles and pore characteristics in the bedding.
During this process, the humidity in the environmental chamber increases to 85%, and a significant amount of water enters the porosity of the bedding, leading to a further increase in deformation.During the d-e stage, the rock undergoes the warming process which leads to increased activity capacity of water molecules.As a result, water in the air keeps entering the rock bedding.It is essential to note that this stage involves the continuous rise in temperature.However, due to the decreasing porosity of the bedding, the ability of water to enter the pores reduces.
Although deformation increases at this stage, the rate of growth decreases notably.

Fig.9 Single cycle curve
In brief, bedding deformation is largely attributed to freezing conditions.As temperature rises, the deformation slowly recovers, though it proves challenging to reach the original state.Freezing With an increased number of cycles, the freezing deformation augments along with larger pores enabling water ingress.Under the cumulative effect, a sudden change ultimately occurred, and the rock strata began to expand considerably.
As mentioned earlier, 16 hours represent a cycle, and the maximum micro-strain is recorded with 16h as the unit to draw the maximum strain-time curve (figure 10).The figure displays a parabolic change characteristic indicating that the maximum deformation amount and time change irreversibly due to multiple cycles, leading to the gradual accumulation of the deformation of the bedding.Through this curve, it is evident that as time and circulation increase, the bedding will experience significant deformation and eventually crack, resulting in visible cracks.bedding.An increase in the deformation rate correlates with a higher degree of response of the bedding to temperature.The overall deformation curve slope characterizes the deformation rate.
The deformation rate curves at various temperatures are obtained by comparing the data from the 1st, 5th, 10th, 15th, and 20th cycles (figure 11).It is observed that the deformation rate notably decreases with temperature.Particularly, the deformation rate drastically drops from -20℃ to 0℃.
Moreover, higher temperatures lead to a narrower deformation range.With an increase in the number of cycles, the impact of temperature variation on deformation weakened gradually, leading to a transition of the curve from exponential to linear change.By referencing survey data and conducting a site survey during the highway construction phase, we were able to create a two-dimensional section and construct a numerical geological model.Technical abbreviations will be explained upon first use.The model spans 1000m and consists of four different strata, namely limestone, sandstone, mudstone, and loess in ascending order.After the reservoir was constructed, the area underwent four significant engineering actions: (1) a substantial slump caused by the reservoir water level increase; (2) unloading brought on by the building of the G241 highway; (3) unloading caused by expressway construction; and (4) the back end being filled with earth.In the numerical simulation, we simulate the evolution of landslides by controlling the changes in groundwater level, excavation unloading, and pile load.
To clarify the characteristics of stress and displacement during the evolution process, we set up 9 monitoring points on the slope surface to track these changes (figure 12).The numerical calculation is separated into five stages: Stage 1 represents natural conditions To calculate the slope's stability using the strength reduction method, only gravity is considered as an external force.
The shear plastic slip area of the slope under varying conditions is depicted in Figure 13.
Based on the available historical survey data, it can be inferred that an ancient landslide took place in this region.This conclusion can be corroborated through numerical simulation.As shown in

Discussion
It is noted in this paper that the landslide cluster is situated near Xiaolangdi reservoir on the Yellow River, and the primary landslide is a loess-bedrock type.Due to various geological processes and engineering activities in the subsequent timeframe, several small loess landslides occurred with the primary slip face being in the loess layer.This section concerns the fluctuation area of reservoir water levels, which has experienced increased human engineering activities over the last 20 years, making it a significant factor in inducing landslides.Based on the preceding research and analysis, we have classified three distinct landslide mechanisms: frost heave deformation of bedrock bedding (figure 15b), stress load changes on slopes (figure 15c, d), and solid-liquid-gas three-phase changes to water found in soil (figure 15e).
As a weak interlayer, the bedding of sandstone easily accumulates water and experiences freezing and melting at high and low temperatures, respectively, leading to changes in its structure.
Laboratory tests focusing on freeze-thaw cycles demonstrate that the bedding undergoes significant expansion deformation, which cannot be ignored over time as it will cause the overlying soil to lift and trigger landslides.
In addition, engineering activities are a significant contributing factor to landslides.The construction of a reservoir leads to a rise in water levels and a consequent sharp increase in pore water pressure in the soil, resulting in a decrease in effective stress and soil strength.This decrease makes landslides more likely.In addition, water infiltration can reduce the friction of the soil-rock interface, leading to landslides caused by gravity.In the construction of highways, excavation is an inevitable process.While excavation unloading is an effective method for controlling landslides, excavation carried out in different areas and time periods can alter the distribution of soil stress, potentially leading to landslides.In this particular landslide, the upper soil was excavated first, which had a beneficial effect on overall stability.With the excavation and removal of the lower soil in the later stages, the upper soil no longer experiences the back pressure from the slope foot and becomes susceptible to landslides due to external influences.
A significant amount of overloading at the top behind the ancient landslide is the determining factor that triggers the landslide.In comparison to the Q3 loess, the density of the artificial fill is greater and the permeability is subpar, making it challenging for water to infiltrate the fill.This characteristic causes the creation of a water-rich layer between the fill and the loess, resulting in significant water accumulation.Water exists in the soil in three phases: solid, liquid, and gas.This causes the soil structure to shift with changes in temperature.Following winter rainfall, the surface soil becomes wet, and as temperatures increase, the surface water evaporates.The difference in water concentration, caused by evaporation and gas release, creates a "vacuum pump" effect, moving water rapidly upward and resulting in sudden landslides.However, the evaporation and movement of water take time, and this is influenced by the particle composition and pore characteristics of the fill material.As a result, the landslide happened on the third day after the rain stopped.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 6
Fig.6 Remote sensing image of water level change in Xiaolangdi Reservoir from 2002 to 2022

Figure 8
Figure 8 displays the general deformation outcomes of the test.The figure clearly portrays a stages: frozen expansion stage (o-a), stable freezing stage (a-b), weakly frozen stage (b-c), fast melting stage (c-d), and warming stage (d-e).It is evident from the figure that the o-a stage comprises a swift deformation phase, where the freezing speed of water in the bedding is rapid, resulting in expansion of the rock layer due to the volume increase.The stable A-B stage ensues since the low moisture content in the bedding limits the degree of expansion through freezing.At expansion with larger pores, thereby facilitating environmental water intrusion.

Fig. 12
Fig.12 Numerical calculation model prior to reservoir construction, Stage 2 describes the condition of rapid water level rise after reservoir construction, Stage 3 pertains to local excavation at the top of the slope, Stage 4 involves large area excavation in the middle of the slope, and in Stage 5, the top of the slope trailing edge is stored.The problem modeled is a two-dimensional strain problem.Lateral and rotational constraints are applied on the left and right sides, while fixed constraints are applied to the bottom.

Figure 13a ,
Figure 13a, prior to the reservoir's establishment, the ancient landslide utilized the soy-rock

Fig. 15
Fig.15 Landslide evolution and mechanical mechanism sandstone were analyzed individually.Technical abbreviations, such as Q3 and N2, will be explained upon first mention.The test results are listed in table 2.