Experimental investigation on the characteristics of seepage failure of landslide dams with strongly permeable zones

Landslide dams are formed by river blockages caused by landslides or other slope instability bodies. They exhibit loose structure, poor stability and strong permeability. Large water head caused by water-level increase can trigger seepage deformation of soil and influence the stability of landslide dams, possibly leading to dam breach and catastrophic damage. Various landslide dam structures also result in differences in seepage characteristics. In this study, multiple physical model tests for seepage failure of landslide dams with strongly permeable zones were designed. The influence of the location and gradation of the strongly permeable zones on the seepage of landslide dams was studied. The characteristics and modes of seepage failure of landslide dams with strongly permeable zones were analysed. The experimental results showed that the cyclic evolution failure of piping and downstream slope collapse was an essential failure mode for the seepage-induced failure of landslide dams with strongly permeable zones. Compared with the strongly permeable zone at the bottom of a landslide dam, the piping caused by seepage evidently promoted the slope erosion of the dam with the strongly permeable zone in the middle. As the permeability coefficient of strongly permeable zones increased, piping was faster and easier to form, and piping failure, slope erosion, and slope collapse were more severe. The seepage failure of landslide dams mainly included the emergence of seepage water, piping, slope erosion, and downstream slope collapse. Piping was caused by the erosion and migration of some fine particles of soil in seepage channels in the dam. When the flow drag force could overcome the resistance force among the soil particles, some fine particles and even large particles on the downstream slope surface were continuously eroded. This study provides new insights into the evolution process and breach mechanisms for the seepage-induced failure of landslide dams with strongly permeable zones.


Introduction
Landslide dams are natural dams formed by landslides or other unstable bodies blocking river valleys caused by rainfall, earthquakes, or other forces and loads.Landslide dams are different from artificial earth-rock dams in terms of material composition, internal structure, geometric-size characteristics, and hydraulic and living conditions [1][2][3].Large water head caused by the increase in water level in barrier lakes can trigger seepage deformation of soil and influence the stability of landslide dams, which can lead to dam breach and catastrophic damage.The Allpacoma landslide dam in Bolivia and the Mantaro River landslide dam in Peru broke due to seepage [4,5].The river channel was blocked by a landslide dam, which caused the water level in the barrier lake and seepage pressure to increase continuously and gradually, respectively.Finally, the landslide dam broke completely due to piping.The complex deposition process and inhomogeneous material composition of landslide dams may result in coarse particle separation and relative concentrations in dam materials.The landslide dams potentially exhibit large porosity and permeability in some zones.The Donghekou, Tianchi, Tangjiashan, and Akatani landslide dams were physically explored and analyzed using multi-channel surface wave analysis.The results indicated probable existence of weak and strongly permeable zones in these dams with concentrated fine and coarse particles, respectively [6,7].The permeability of coarse particle concentration zones is significantly different from that of the surrounding dam materials, which is more likely to induce seepage failure.Therefore, investigating the landslide dam with strongly permeable zones is necessary.
A landslide dam is mainly composed of loose rock and soil particles, and the gradation range of dam materials is wide.In a previous study, large-scale model flumes were used in model tests to study seepage failure processes and modes of landslide dams [8].The model test results indicated the seepage failure in the dam to probably cause piping channel formation, leading to the slope surface erosion and eventually leading to the local instability and dam slope collapse [9,10].The formation of dominant seepage path inside the dam depended on the content and distribution of coarse particles [11].When the pores between coarse particles are large and the content of fine particles is low, fine particles cannot fill the pores fully, and the water easily washes them away resulting into piping [12,13].In general, the difference in seepage failure characteristics of landslide dams is caused by the difference in dam material characteristics and internal structure.However, homogeneous dam materials are used in most model tests at present, and the heterogeneity of dam materials, in line with the site conditions, is not considered.The existence of strongly permeable zones can increase the possibility of seepage failure of landslide dams.Therefore, the processes and characteristics of seepage failure of landslide dams with strongly permeable zones are worthy of further investigation.
At present, the seepage failure characteristics of landslide dams with strongly permeable zones have not been experimentally investigated.Therefore, in this research, a series of physical model tests for seepage failure of landslide dams with strongly permeable zones were designed.The influence of the location and soil gradation of the strongly permeable zones on the landslide dam seepage was investigated.The processes and mechanisms of seepage failure of landslide dams with heterogeneous structures were analyzed.This research can provide new insights into the evolution process and breach mechanisms for seepage-induced failure of landslide dams with strongly permeable zones.

Experimental materials
In this study, the gradation of dam materials for flume tests (Figure 1) was an average of some gradations of dam soils obtained from field investigations and historical cases of typical landslide dams.The experimental dam materials, composed of non-cohesive quartz sand with different particle sizes (0.1-100 mm), were mixed and stirred evenly before the test.The permeability coefficient of dam materials obtained by seepage failure tests was 0.037 cm/s.The maximum and minimum dry densities of the dam materials obtained by the relative density test were 2.24 and 1.88 g/cm 3 , respectively.The soil density of the landslide dam is usually low.Zhong et al. obtained the relative density of 0.2 [8].Therefore, the dry density of 1.95 g/cm 3 corresponding to the relative density of 0.2 of the filling soil was used in model tests.This value was close to the actual soil density obtained by drilling tests in the landslide dam site [14].After each test, the dam materials were screened and naturally air-dried before the next test, and the moisture content was about 2.8%.

Experimental apparatus
The experimental apparatus was mainly composed of model flume, reservoir system, data acquisition system, and tailrace collection tank, as shown in Figure 2. The model flume dimensions were 600 cm × 80 cm × 85 cm, and the particles of up to 160 mm, satisfying the wide gradation characteristics of landslide dam materials, could be accommodated by the flume.The toughened glass on both sides of the model flume facilitated convenient observation of the upstream water level and downstream slope.The base slope of the model flume was adjustable, with a range of 0-30°, to simulate the slope of riverbed.The reservoir system consisted of an upstream tank, a water valve, a flowmeter, and a ball float valve.The upstream tank provides the water needed for seepage tests.The water valve and flowmeter were used to control and measure the upstream inflow, respectively.The ball float valve was used to stabilize the upstream water level to the set water level.
The data acquisition system was composed of a camera, pore pressure sensor, turbidity meter, and downstream water-volume measurement device.The variations in the downstream slope and wetting zone were recorded by the camera during the test.The pore water pressure inside the dam was measured by the pore pressure sensor, with an accuracy and range of 0.1 and -100-100 kPa, respectively.The outflow turbidity was measured using the turbidity meter placed in the downstream water tank.The downstream seepage discharge was obtained by measuring the water quantity in a certain time duration by the downstream water-volume measurement device.The tailrace collecting tank was used to collect the downstream discharge and alluvial sediment.

Experimental scheme
The model tests conducted in this study were based on the experience of most model tests and were as similar as possible to the prototype in the length, height, and upstream and downstream slope angles [5,8].The transverse section of the model dam was trapezoidal (Figure 3a); the height, crest width, and bottom width of the dam were 80, 25, and 230 cm, respectively.The longitudinal section of the model dam was rectangular (Figure 3b), and the dam length was 80 cm, which was equal to the flume width.The average base slope of the river in the southwest of China is approximately 2°.Therefore, the bed slope of the model flume was set to 2°.
The experimental scheme for seepage failure of landslide dams with strongly permeable zones is summarized in Table 1.To investigate the influence of the location of the strongly permeable zone on the seepage failure of the landslide dam, a strongly permeable zone with a square cross section was set to pass through the bottom or middle of the model dam.Furthermore, the influence of the gradation of the strongly permeable zone on the seepage failure of the landslide dam was investigated.For this, the particle sizes of 2-60 and 40-60 mm were used for the strongly permeable zone at the bottom and in the middle of the dam, respectively.The permeability coefficients of the strong permeability zones were 0.063 and 0.218 cm/s, respectively, which were approximately 1.70 and 5.89 times the permeability coefficients of the dam materials.The strongly permeable zone at the bottom of the model and layout of pore pressure sensors for the model test are shown in Figure 3.The layout of pore pressure sensors for the strongly permeable zone in the middle of the dam was the same.The sensors p-4-6 were set below the strongly permeable zone in the middle of the dam.

Experimental procedure
Before the model test, a 5-mm thick yellow clay was smeared on both sides and bottom of the model in the flume to avoid dominant seepage between the dam material and flume wall.Then the dam material with given gradation was uniformly filled according to the predetermined dry density.In the filling process, model support plates were used to ensure that the length, height, and upstream and downstream slopes of the model dam were filled in layers according to the preset parameters.Moreover, two separation plates were set vertically to separate the strongly permeable zone (if present in the layer) and other part of the dam.The pore pressure sensors were buried during the filling process according to Figure 3.During the test, the reservoir and data acquisition systems were turned on, and the water level was maintained at 75 cm (5 cm lower than the dam crest) to avoid the overtopping through the upstream ball float valve.The model test was terminated if the obvious seepage failure occurred or long-term stable seepage was observed in the dam.

Results and analysis
3.1.Processes and modes of seepage failure 3.1.1.Seepage failure of the dam with strongly permeable zone at the bottom (model 1).The process of seepage failure of the dam with strongly permeable zone at the bottom is shown in Figure 4, where t represents the test time.At the initial time, the reservoir was impounded upstream of the dam.After 17 min, seepage water emerged from the strongly permeable zone at the bottom of the dam.The wetting zone downstream the dam slope surface continuously widened and extended upward, and new seepage exits appeared at the left and right toe of the dam (Figure 4a).At t=139 min, small cracks appeared at the downstream slope approximately 30 cm below the dam crest, and more seepage exits appeared in the middle area of the downstream slope (Figure 4b).Some fine particles in some seepage channels were washed out, eventually causing piping, and the downstream slope was eroded.For t=139-176 min, the cracks in the middle of the dam continued to extend to both sides, and finally those on the left and right were connected (Figure 4c).Then, the width and depth of the cracks extended continuously, with the final depth and width of the maximal crack reaching 5 and 8.6 cm, respectively.During t=176-445 min, a new crack appeared through left to right approximately 14 cm below the dam crest (Figure 4d).At t=445 min, some areas of the downstream slope began to collapse, and the slide area accounted for approximately 15% of the total area of the slope surface (Figure 4d).The process of seepage failure can be divided into five stages: emergence of seepage water from the strongly permeable zone, development of piping, sudden emergence, extension, and connection of cracks, collapse of partial areas of the slope, and stable seepage.

Seepage failure of the dam with strongly permeable zone in the middle (model 2)
. Figure 5 illustrates the process of seepage failure of the dam with strongly permeable zone in the middle.As shown in Figures 5a and 5b, after 22 min, seepage water emerged from the strong seepage zone in the middle of the dam, and some fine particles were washed out by the seepage water.During t=22-29 min, the soil eroded internally causing piping, and the downstream slope eroded continuously owing to seepage.The maximum erosion depth of the downstream slope was 10 cm, and the wetting zone gradually extended to both the sides.At the piping outlet, fine and some coarse particles were washed out by seepage (Figure 5c).At t=50 min, the soil approximately 22 cm below the damn crest, on the top of the piping channel, collapsed and was washed away (Figure 5d).The piping channel expanded, and the slope erosion deepened.During t=50-144 min, the soil on the left of the piping channel, approximately 5 cm below the dam crest, continued to collapse, and the piping channel continued to  5e).During t=144-193 min, a large number of fine particles of seepage channel were washed out, the soil above the piping channel collapsed on a large scale, and the thickness of the dam crest began to reduce (Figure 5f).During t=226-228 min, the local break occurred approximately 32 cm from the left side of the dam crest, and a breach formed in the middle of the dam crest.Then, the overtopping flow occurred through the breach.Next, the breach scoured and expanded, and finally flood gushed downstream (Figure 5g and 5h).The process of seepage failure can be divided into five stages: emergence of seepage water from the strongly permeable zone, piping and slope erosion, downstream slope collapse, local break of dam crest, and overtopping failure.This failure process was different from the case of the strongly permeable zone located at the bottom of the dam.The piping caused by seepage obviously promoted dam slope erosion with the strongly permeable zone in the middle.When the strongly permeable zone was located in the middle of the dam, a large amount of water flowed out from the middle of the dam with a large scouring force and was more likely to cause slope erosion.Conversely, when the strongly permeable zone was located at the bottom of the dam, a large amount of water flowed out from the bottom of the dam.In addition, lesser piping was observed in the middle and upper parts of the dam compared with the case of the strongly permeable zone in the middle of the dam.Therefore, the dam with a strongly permeable zone at the bottom did not present obvious slope erosion.However, several cracks occurred downstream the slope owing to the piping erosion.In addition, because the particle size of the strongly permeable zone of the model 2 became coarser than that of the model 1, piping channel of the model 2 was easy and quick to form.Consequently, increased number of fine particles were washed out from the piping channel, and the continuous expansion of the piping channel increased the seepage discharge.Therefore, the slope erosion and piping failure of the model 2 were obviously serious.The local break occurred on the dam crest, and the overtopping flow occurred in the breach.The erosion parameters of dams before overtopping are shown in Figure 6.The erosion ratio of slope is ratio of volume of slope erosion and original model dam.The maximum settlement denotes the maximum subsidence depth of the dam crest to the original dam height.The subsidence volume of the dam crest is defined as vertical settlement of the dam crest compared with the original dam.The IOP Publishing doi:10.1088/1755-1315/1334/1/0120227 erosion ratio of slope for dams increases nonlinearly with the testing time.The maximum erosion ratios for Tests 1 and 2 were 1.63% and 5.1 times that of Test 1, respectively.The slope erosion was the most severe in Test 2.Moreover, the slope erosions of Tests 1 and 2 tended to be stable in the later stage of the test and continued to expand in a large area, respectively.The maximum settlement and subsidence volume for Test 1 were the smallest.The settlement and subsidence volume of Test 2 were 4.8 and 6.7 times that for Test 1, respectively.In coclusion, the internal deformation and settlement of dam crest caused by piping are more obvious when the strong permeability zone is in the middle and the particles are coarse.

Processes and modes of seepage failure
The seepage failure of a landslide dam with strongly permeable zones was a progressive failure accompanied by piping and downstream slope collapse cycles in this research.In general, the seepage failure of the dams mainly experienced multiple stages, such as piping, slope erosion, and downstream slope collapse.Thus, the mechanism of seepage failure was analyzed through different stages.
Furthermore, pore water pressure, seepage discharge, and outflow turbidity helped analyze the seepage failure mechanism.The pore water pressure sensors were set at different positions of the dam and obtained the variation in internal pore water pressure during the seepage failure processes, as shown in Figure 7.The variation in seepage discharge and outflow turbidity of landslide dams are shown in Figure 8.   Piping exhibited a progressive failure with time.As upstream water level increased, water flowed downstream through the landslide dams, and the seepage force acted on soil particles in the dam.Interestingly, the strongly permeable zone was a dominant seepage channel.The rapid increase in pore water pressure and seepage discharge in the initial stage indicated the gradual appearance of seepage channels in the dam during t=0-50 min, as shown in Figures 7 and 8.When the seepage force exceeded the resistance of some fine particles, they were washed out through the connected pores of soils by seepage water.The internal erosion in the soil may have caused piping.The migration of fine particles was also reflected by the rapid increase in outflow turbidity during t=0-50 min, as shown in Figure 8. Furthermore, some seepage outlets were eroded into holes by seepage water, and piping channels were continuously expanded.This resulted in increased seepage discharge.Moreover, the expansion of piping channel and collapse of its upper soil occurred continuously and cyclically.The fluctuation in seepage discharge curve (Figure 8) was mainly caused by formation, expansion, and upper collapse of piping channel.At the piping outlet, some fine and coarse particles both were washed out by the seepage water.
Slope erosion was mainly caused by the water flowing on the slope surface.Usually, in the initial stage of seepage failure, the amount of water flowing on the slope surface was less, and the drag force of water flow was limited, overcoming only the resistance force among soil particles and transporting them downstream.With the continuous expansion of piping channels, the seepage discharge on the slope surface gradually increased.The soil on the slope surface continued to be eroded by the drag force of the water flow.Increase number of fine particles and even large particles were continuously washed away.Finally, some gullies appeared on the downstream slope.Besides piping erosion, slope erosion also leads to a significant increase in outflow turbidity, as shown at t=238 and 46 min in Figure 8a and 8b, respectively.
When the sliding force was greater than the resistance force acting on soil particles on the downstream slope: the downstream slope was scoured by seepage, the surface particles were unstable, and the downstream slope gradually collapsed.In addition, the outlet of the piping channel also collapsed.Owing to the loose soil structure, low strength of the piping outlet, and increase in sliding force after soil saturation, the upper soil of the piping outlet collapsed and failed owing to gravity.Then, the piping outlet widened, and the scale of the slope collapse increased.Besides piping erosion, slope collapse was reflected by the increase in outflow turbidity, as shown at t=445 min in Figure 8a and t=144 and 193 min in Figure 8b.The collapse of upper soil of the piping outlet temporarily decreased the seepage discharge, as shown at t=144 min in Figure 8b.When the piping channel formed and expanded continuously, the seepage discharge continued to increase.In particular, the slope collapses gradually extended upward, which may have led to the breach of dam crest and overtopping flood.Even if the dominant seepage and piping occur in the dam, resulting in local failure on the downstream slope, they may not finally lead to overtopping and global break of the dam. Figure 7a illustrates relatively stable pore water pressures that appeared in the later stage in model 1.This indicated the formation of a stable seepage in the dam, which was consistent with the failure process of the model 1 (Figure 4).The sudden decrease in and loss of pore water pressure, at t=140 and 193 min in Figure 7b, indicated that the dam was seriously broken, and the pore water pressure sensors were removed from the dam and exposed to air; thus, the pore water pressure could not be measured.

Conclusions
In this research, a series of physical model tests for seepage failure of landslide dams with strongly permeable zones were conducted.The seepage failure characteristics of landslide dams were analyzed.Following conclusions can be drawn: The cyclic evolution of piping failure and downstream slope collapse is an essential failure mode for the seepage-induced failure of landslide dams with strongly permeable zones.The seepage failure of landslide dams mainly experienced emergence of seepage water, piping, slope erosion, and downstream slope collapse, etc.
Compared with the strongly permeable zone at the bottom of a landslide dam, the piping caused by seepage obviously promoted the slope erosion of the dam with the strongly permeable zone in the middle.With the increase in the permeability coefficient of strongly permeable zones, piping was faster and easier to form, and piping failure, slope erosion and slope collapse were more serious.Finally, overtopping and local and even global dam failure can occur.
Piping was caused by erosion and migration of some fine soil particles in seepage channels in the dam.If the flow drag force could overcome the resistance force among soil particles, some fine particles and even large particles on the downstream slope surface were continuously eroded.When the sliding force exceeded the resistance force acting on soil particles on the downstream slope, the downstream slope would gradually collapse.However, even if the dominant seepage and piping appear in the dam, it may not always result in overtopping and global break of the dam.

Figure 1 .
Figure 1.Grading curves of the dam materials used in the landslide dams.

Figure 3 .
Figure 3. Model design of landslide dam with strongly permeable zones.

Figure 4 .
Figure 4. Process of seepage failure of the dam with strongly permeable zone at the bottom.

Figure 5 .
Figure 5. Process of seepage failure of the dam with strongly permeable zone in the middle.
a. Strongly permeable zone at the bottom of dam b.Strongly permeable zone in the middle of dam

Figure 7 .
Figure 7. Pore water pressure of the dam with strongly permeable zones.

Figure 8 .
Figure 8. Seepage discharge and outflow turbidity of the dam with strongly permeable zones.

Table 1 .
Experimental scheme for seepage failure of landslide dams with strongly permeable zones.