Deposit morphology and structure under interactions of sliding mass and erodible layers: experimental insights

Landslides are a kind of highly concerned geological disasters that occurring with complex motion processes and mechanisms. They often significantly affect the human life and properties located in their pathway. In some circumstances, the geological phenomena and structural features generated by the interactions between landslides and their substrates are still unclear, which makes it difficult to be forecasted and mitigated on its effects. In this study, a sandbox experiment was conducted to study the velocity and the displacement of the sliding mass, the geometry of the deposit, and the internal and the external structural characteristics of the deposit under the interactions between the sliding mass and the erodible layer by varying the depth of the erodible layer. Results show that the motion process of sliding mass consists of three stages: falling, shovel push-extrusion, and push-nappe accumulation. In the first stage, the velocity of the sliding mass increases sharply to a peak velocity before colliding with the erodible layer. In the latter two stages, the mobility of landslide is greatly limited by the erodible layers at the foot of the inclined plate, and the secondary acceleration of the sliding mass is observed. The deposits were divided into three zones (Ia, Ib, and II), in terms of the morphological and structural characteristics of their positions. The action forms were mainly pushing and covering in the zones II and I respectively. There were phenomena, such as strata inversion, pushover, and entrainment, which occurred in the deposits—the folds, ridges, and bulge that occurred on the surface of deposits. These structural characteristics reflect the stress states of laboratory landslides in motion from compressing to shearing. The results of this research will provide a valuable theoretical reference for the calculation of the disaster range when erodible layers exist in landslides' motion paths.


Introduction
Landslides, a common geological phenomenon, can rapidly spread on complex terrain and over various substrates, leading to property loss and casualties. These landslides can spread rapidly over complex terrain and substrates with different lithologies after destabilization (Roberts et al. 2021;Roche et al. 2011). Erosion and entrainment of the substrate are often observed during the motion of landslides, increasing their volume significantly (Haas and Woerkom 2016;Lucas et al. 2014;Mergili et al. 2020). Despite the increasing number of experimental, field, and numerical studies on the interactions between landslides and their erodible layers (Crosta et al. 2017;Duan et al. 2023;Iverson et al. 2011;Peng et al. 2018), a systematic study on the stratigraphic structure evolution of deposits at different depths of erodible layers is still lacking. Field investigations are the basis for analyzing the interactions between landslides and their substrates. Geophysical granular flows interact with their substrates in different ways, depending on the mechanical properties of the underlying material (Farin et al. 2014). The mobility and deposit areas of landslides are significantly influenced by erosion (Sovilla 2004), which occurs preferentially along a sloping path, leading to the scraping and erosion of sediment along the path (Conway et al. 2010). However, the erosion phenomenon is also observed along sub-horizontal substrates (Farin et al. 2014;Mangeney et al. 2010). Previous field investigations and laboratory studies on basal erosion have suggested that the erosion action of a landslide depends on its lithology, mechanical conditions, and the geomorphology of the motion path (Iverson 2012;McCoy et al. 2012). The geomorphology and landforms have a great influence on the lateral spreading of landslides. Moreover, the process of bed entrainment can lead to different propagation paths of the slide body in front of the mountain (Cuomo et al. 2016).
Erosion can occur in landslide substrates with different lithologies. When the substrate is rocky, structural phenomena, such as diapiric intrusions, convoluted laminations, faults, recumbent asymmetrical folds, and broken boulders, easily form in the interior regions of the deposits Zeng et al. 2020). These phenomena are considered to be typical evidences of compression (Dufresne et al. 2009). Other structural phenomena can be found in the substrates, such as faulted, folded, or strongly distorted, which suggests strong shear coupling at the flow base (Dufresne 2012;Farin et al. 2014). When the landslide substrate is composed of glacial residue, the sliding mass exhibits higher mobility, primarily due to the concomitant ice melting and entraining of water during motion (De Blasio 2014; Sosio et al. 2012). Typical evidences of erosion can also be found in the deposits, such as boulders with internal shear fracture surfaces, large fragments, striations, and furrows (Dufresne et al. 2019). Similar structural phenomena have also been found in loess landslides (Duan et al. 2018;Xue et al. 2021).The structural phenomena induced by landslides are an important indicator of their motion process and force characteristics (Hungr et al. 2013;Rana et al. 2016). However, due to the complex terrain, infrequent occurrence, high cost, and long measurement time, available field data are scarce (Bai et al. 2021;De Haas et al. 2020;Dietrich and Krautblatter 2019). For this reason, physical model experiments have been conducted by researchers to study the movement process and internal structural characteristics of landslides. Physical experiments (De Haas et al. 2020;Dowling and Santi 2014;Hu et al. 2021;Iverson et al. 2011) demonstrate the important effects of erosion phenomena in landslides. The process of reproducing natural phenomena through laboratory experiments is a common technique to study granular flows and has achieved good performance (Delannay et al. 2015;Duan et al. 2022;Wu et al. 2022). Steady uniform flows of granular material over an inclined bed and landslide geometry with erodible beds on its horizontal plane are often previous facilities. For the former, landslides' erosion process and has been quantified with erodible materials on the inclined bed. (Iverson et al. 2011;Mangeney et al. 2007). Mangeney et al. (2010) revealed that the landslides' runout was increasing linearly with the depth of the erosion bed when the slope angle was larger than the critical one, namely the internal friction angle. The existence of erodible bed would not affect the accelerating stage of landslides but the decelerating stage. For the latter, experiments of granular collapse make it possible to establish scaling law of deposit features with the initial geometries of sliding mass (height, radius, and aspect) (Balmforth and Kerswell 2005;Lajeunesse et al. 2005). Other researchers have focused on the erodible material and laying position as experimental variables, recording the landslide accumulation landforms (Crosta et al. 2017;Dufresne 2012;Lacaze et al. 2008;Shea and van Wyk de Vries 2008). In the study of Crosta et al. (2017), the landslides' evolution, dynamics, erosion, and deposit mode were studied by changing the slope angle and sliding material. The internal structures of the deposit had well been reproduced with a three-colored sand layer of 21 mm. However, how the internal structures vary for thinner and thicker erodible strata is still an unresolved question. We therefore raised questions: whether the internal structures of deposit will be more intensive; whether the increasing in the thickness of erodible beds controls the entrainment and hence the mobility of landslides; what a relation is the landslides' internal and surface structures.
In order to having an attempt to these questions, we plan to change the thickness of erodible beds on the horizontal plane of model experiments. To this end, the objectives of this study are to (1) identify the influence of the depth of the erodible layer on the velocity and displacement of landslides; (2) clarify the influence of the depth of the erodible layer on the deposit morphology; (3) ascertain internal and external structural characteristics of the deposit under the interactions of the sliding mass and erodible layer; and (4) generalize the motion process of the landslide with respect to the erodible layer under its motion path.

Experimental equipment and materials
In this study, a sandbox experimental device was designed to study the process of landslides impacting their erodible substrate. The experimental device consisted of a physical model system and a monitoring system. The physical model system included an inclined plate, a horizontal basin, and a sand container. The inclined plate was 1.5 m × 1.2 m, and the angle of the inclined plate could be adjusted using a bracket. The horizontal basin was also 1.5 m × 1.2 m, with curbs on both sides. The sand container had a side-by-side door, whose height of the center of gravity can be adjusted through the sand container track, which was used to fill the sliding mass. The monitoring system included one 3D scanner and two high-velocity cameras. The 3D scanner had a shooting velocity of 8 frame/s to obtain the detailed motion process of the sliding mass and the digital elevation model data (DEM) of the deposit. Two high-velocity cameras (120 frames/s, 0.4 MPix of resolution) were used to record the profile and bird's eye images during the movement of the sliding mass, respectively. The positions of the aforementioned monitoring equipment are shown in Fig. 1.
A dry medium-fine quartz sand was used as the medium for the sliding mass and erodible layer (Fig. 2). Pre-experiments showed that this kind of sand could characterize fluidization, which could not easily maintain its initial geometry, and would immediately collapse and propagate when constraints are removed during motion (Duan et al. 2020). The coefficient of nonuniformity, coefficient of curvature, average diameter, and specific surface area were 2.39, 1.19, 0.2 × 10 -3 m, and 0.02 m 2 •kg −4 , respectively. The accumulative percentage of particle sizes in the range of 0.075-0.5 mm was 87.71%. The distribution curve is shown in Fig. 3. The internal friction angle φ and cohesion c obtained through direct shear tests were 31.5° and 0 kPa, respectively. The friction coefficient was 0.613. For other direct shear tests, Plexiglas-with dimensions of φ 61.8 mm × 10 mm-was placed in the lower shear box. Utilizing this setup, the interfacial friction coefficient between the Plexiglas and medium-fine sand was measured as 0.403. To better distinguish the stratum structure, the quartz sand used in this study was colored sand. The physical and the mechanical properties of the different colored sands are consistent with one another.

Experimental methods
The angle between the inclined plate and the horizontal basin was 60°, the height of the center of gravity of the sandbox H was 1 m, and the volume of the sliding mass was 3.6 × 10 -3 m 3 . When filling sand into the sand container, the mass and the volume of sand were controlled to ensure the initial density of the sliding mass remained constant. The depths of the erodible layer were paved on the horizontal basin as 0 mm, 4 mm, 8 mm, 12 mm, 16 mm, 20 mm, and 24 mm. Among them, 0 mm indicated that there was no erodible layer on the horizontal basin. The erodible layer was paved using a sand spreader (Fig. 4a). During this process, the colored sand was placed into the funnel of the sand spreader first, and the baffle was then pulled slightly to generate an even gap at the bottom of the sand spreader. The colored sand in the funnel would flow out to form a flat erodible layer when the scraper moved forward along the curbs. The depth of the erodible layer was controlled by adjusting the elevation setter. The thickest erodible layer (24 mm) consisted of six layers of colored sand: blue, purple, gray, red, green, and blue, from top to bottom; each was layer was 4 mm thick. (Fig. 4b, c).
The sliding mass was released from the bottom of the sand container when the switch of the side-by-side door was opened. The motion process and the digital deposit morphology of a laboratory landslide were recorded by the monitoring system. To ascertain the experimental conditions and characteristics of the deposits, we defined the height of the center of gravity of the sand container (H), the angle of the inclined plate (α), the total sliding distance of the sliding mass (L), the depth of the erodible layer (D e ), the length of the deposit (L d ), the width of the deposit (W d ), the depth of the deposit (D d ), the eroded width (W e ), and the eroded length (L e ). The total sliding distance of the sliding mass (L) spans from the bottom of the sand container to the front of the sliding mass along the plates, as shown in Fig. 5.
The motion direction of the sliding mass and its perpendicular direction were defined as the y-axis and x-axis, respectively. After the motion of the sliding mass ceased, a The paving process of erodible layers; c The completed erodible layer of 12 mm is shown as an example, with each layer being 4 mm deep transparent L-shaped piece of Plexiglas was used to visualize the cross sections along the y and x directions of the deposits, which made it possible to observe the internal structure and interactions between the sliding mass and erodible layer. The x-and y-axis cross-sectional images of the deposit were recorded using a camera (Fig. 1).

Limitations of laboratory testing
Due to the scale effect, some important physical and mechanical processes in field landslides could not be reflected in the model experiments, including the dynamics, electrostatic phenomena (Iverson et al. 2004), rock breakage, and seismic effects (Davies and McSaveney 1999) in large landslides. Different structural geological phenomena can be generated under the influence of the slope angle, the material characteristics, and the topographical friction coefficients. Compared to field landslides which are not continuously accelerated and frequently followed by rock fragmentation, the slope angle and the test material selected for the test are fixed and cannot be completely consistent with the practical phenomena. Besides, the complex boundary conditions and the diverse material properties of field landslides are also difficult to reproduce in model experiments (Delannay et al. 2015). This makes it difficult to clarify how a single factor influence landslides' evolution. In this study, we simplified the boundary conditions and material selection, which facilitates testing repeatability (Dufresne 2012;Mangeney et al. 2010). The simplified experimental apparatuses and materials are beneficial to systematically study the effect of the erodible layer depth on the landslide motion and deposit characteristics. In addition, it is easier to describe and quantify the experimental results.
There are no side constraints to the boundary conditions of this study. Our experiments were carried out at 60°, because the kinematic parameters of sliding masses are more discernable Li et al. 2021). The sliding mass will not convert sufficiently into kinetic potential energy and has less impact on the erodible layer without facilitating the extensive development of the internal structure when the angle does not exceed 60 degrees. In our pre-experiments, the erodible layer cannot be penetrated completely at 24 mm thickness, and the deposit boundary is clearer at 3.6 × 10 -3 m 3 . Therefore, such a thickness and a volume are chosen. At 1 m height of the sand container, the sliding mass can possess adequate kinetic energy at the slope break to interact with the erodible bed to produce more observable internal and external deposit structures. This kind of material is demonstrated with similar behaviors of low rheological strength and deposit structures to natural events (Crosta et al. 2017;Duan et al. 2022;Manzella and Labiouse 2009), therefore being as the experimental material.

Characteristics of the movement of the sliding mass
According to the changing characteristics of the sliding mass, the motion process consisted of three stages: falling, shovel push-extrusion, and push-nappe accumulation. At the falling stage, the velocity of the sliding mass could reach 2.7-3 × 10 3 mm/s. At the shovel push-extrusion stage, the velocity of the sliding mass rapidly decreased when the sliding mass collided with the erodible layer (or bare board). The greater the depth of the erodible layer, the faster the sliding mass velocity decreases after the collision. The velocity of the sliding mass decreased about 79.3% within 0.125 s after the collision when D e was 24 mm. However, the velocity decreased from 3024.35 to 1572.64 mm/s within 0.125 s after the collision when the horizontal basin did not contain an erodible layer. Under this condition, the difference in velocity after colliding at the slope break was significantly less than the conditions when the erodible layer exists (Figs. 6 and 7).
At the stage of push-nappe accumulation, the interaction between the sliding mass and erodible layer weakened. Further, the sliding mass propagated forward with a relatively low velocity, which slowly decreased until it stopped. It should be noted that a secondary acceleration of the sliding mass was discerned in this stage under conditions where the D e were 4 mm, 12 mm, 16 mm, and 20 mm. From 0.875 s to 1.0 s, the increased velocity due to the secondary acceleration was 421.6 mm/s and 281.6 mm/s at D e = 4 mm and D e = 12 mm, respectively. However, the timing of the secondary acceleration was relatively late and the increase in value was smaller when the D e = 16 mm and D e = 20 mm. After the transient secondary acceleration, the velocity decreased again. It was noted that there was no secondary acceleration when D e = 0 mm and D e > 20 mm (Figs. 6 and 7).

Characteristics of deposit morphology
The contour map and the photo of the deposit were superimposed, as shown in Fig. 8. When the horizontal basin was paved with an erodible layer, the surface morphology of the deposits was similar to a crescent-shaped sand dune, where distinct upheaval is observed at the front edge of the deposits. By considering the upheaval as a boundary, the deposits were divided into a forward slope and a reverse slope. With increasing D e , the range of the deposit gradually decreased from 3.0 × 10 5 to 1.5 × 10 5 mm 2 , the upheaval gradually approached the slope break, and the distance decreased from 255 to 150 mm. When no erodible layer was paved on the horizontal basin, the deposits performed an ellipse shape with the x-axis being long and the y-axis being short. Its middle part exhibited an upheaval, and the deposit was completely detached from the slope break.
Relevant deposit data was obtained using the digital elevation model (DEM), as shown in Fig. 9. When D e = 0 mm, the length, width, depth, and area of the deposits were 503.35 mm, 775.90 mm, 32.51 mm, and 294.78 × 10 3 mm 2 , respectively. When the erodible layer was on the horizontal plate, the length, width, and area of the deposits were significantly reduced, but the depth increased. With increasing D e , the difference in the length and depth of the deposits was more significant than the width (see Table 1).

Characteristics of the internal structure of the deposit
To observe its internal structures, we cut profiles in the deposit using an L-shaped transparent piece of Plexiglas, as shown in Fig. 9a-c. In the figure, a, b, and c are not only the position of the sectional points, but also the cutting order in each experiment. The measured points in the erosion area were obtained by these profiles of the deposits, such as points D, E, F, and G in Fig. 9b and c. The erosion boundary of the sliding mass could be obtained by fitting these points, where the area and volume of the eroded area could be quantified in terms of the erosion boundary (Fig. 9).
It was found that the erosion distance of the deposit on the y-axis was longer, and the erosion effect was more significant than that on the x-axis when observing the profile of the deposit. The erodible layer was completely penetrated, except for D e = 24 mm. Owing to the interaction between the sliding mass and the erodible layer, phenomena, such as buckles, recumbent folds, thrust shear, horizontal shear, and convolute bedding, were in the interior of the deposit ( Fig. 9c and d).
It is important to note that buckles and recumbent folds developed on the front edges for all deposits with erodible layers (Fig. 9c). As can be seen in the x and y profiles, the positions where the erodible layer formed folds corresponded to the erosion range, namely the core of the horizontal fold. The core of the recumbent folds had a larger distance than the upheaval of the deposits in the horizontal direction at the 4 mm and 8 mm conditions of D e . The positions of both the core of the recumbent folds and the upheaval became closer to the slope break as the D e increased. When the D e was at 8-20 mm, their projected positions in the horizontal plate were almost identical. When D e increased to 12 mm, structures of thrust shear formed. The angle of the shear plate gradually became steeper as D e increased to 20 mm. It is worth noting that secondary shear fracture surfaces, with relatively gentle angles, were formed at D e = 24 mm. When  the D e values were 16 mm and 20 mm, the front sliding mass was convoluted into the recumbent folds during their formation, leading to a sandwich structure. When D e increased to 24 mm, the convoluted bedding structures could hardly be observed (Fig. 9c).
As D e increased, the erosion length, width, and area decreased generally, but the erosion volume increased first and then decreased. The erosion volume reached a maximum when the D e was 20 mm (Table 1).

Relationship between internal and external features of deposits
The deposits were divided into three zones (I a , I b , and II), according to the structural morphology and characteristics of their positions. The position of upheaval was regarded as the boundary between zone II and zone I. Zone II was the region from upheaval of the deposits to the slope break, zone I b was the region from the ridge to the front of the deposits, and zone I a was the region beyond the front of the deposits, where stratum disturbance was present (Fig. 10). From the side view, in zone II, erosion was obvious, and the internal structures formed in the deposits were relatively simple. At D e = 4 mm and 12 mm, the erodible layer was eroded completely and pushed to zone I (Fig. 10a, b and g); when D e = 24 mm, the erodible layer was not penetrated completely, and a slope with a dip opposite to the sliding direction was formed (Fig. 10c, l). There are shear ridges formed on the deposit surface (Fig. 10e,  h). In general, the thrust and covering are the main structures in the zone I. Some complex structures, such as buckles, recumbent folds, thrust shear, horizontal shear, and convoluted bedding, formed in the inner region of the I b zone and the transition zone with the II zone. There were small bulges on the deposit surface in the I b zone (Fig. 10j). Zone I a was not covered by the sliding mass, but developed with folds formed in the erosion layer (Fig. 10 d). Zone I a vanished gradually with the increasing of D e (Fig. 10f, i).

Influence of the erodible layer on the velocity and displacement of landslides
From the velocity and displacement of the sliding masses, it is clear that the presence of the erodible layer on the horizontal basin had a significant restraining effect on the mobility of the sliding mass. Under the restraining effect, not only was the movement distance of the sliding mass reduced, but also the length and the width of the deposits were also decreased. The main reason for this was that the density and depth of the erodible layer increased due to interactions with the sliding mass. It was also for this reason that the topography in the front edge of the deposits elevated and the shear strength of the erodible layer was fully exerted under the force of the sliding mass, which increased its resistance during motion. After the collision between the sliding mass and erodible layer, the kinetic energy of the sliding mass was consumed primarily by shearing and colliding among particles. On the other hand, the friction coefficient in the interior regions of the sand was 0.613, which is larger than the friction coefficient for the sand-plate interface (0.403). Therefore, the mobility of the laboratory landslides without an erodible layer was larger than that with an erodible layer (Crosta et al. 2017;Liu et al. 2020;Yuan et al. 2014;Zhou et al. 2019).
To reduce the influence of material differences on the experimental results, the materials of the sliding mass and erodible layer were both medium-fine sand, with the same physical and mechanical attributes. In some model experiments using dry particles as the erodible layer, experimental results have shown that the erodible layer had an inhibitory effect on the mobility of the sliding mass (Crosta 2013;Crosta et al. 2017;Dufresne 2012). It should be noted that all of these studies set the erodible layer on the nearly horizontal basin. Nevertheless, when the erodible layer was set on a downward sliding path, such as an inclined plate, opposing results were found (Mangeney et al. 2010). Further, the mobility of the sliding mass would increase due to entrainment during collisions with the erodible layer, irrespective of whether the underlying layer was dry or wet, especially when the slope angle of the erodible layer was close to the internal friction angle of the material (Crosta et al. 2015a;Mangeney et al. 2010).
In this study, when the D e was less than 20 mm, the larger the D e was, the more significant the hindering effect was on the mobility of the sliding mass, especially in the planform of the deposits. However, when the D e was 24 mm, the length and the width of the deposits increased, from 287.59 and 630.00 mm to 291.34 mm and 650.60 mm, respectively. At the same time, the erosion volume decreased from 905.14 × 10 2 to 190.40 × 10 2 mm 3 due to the decreasing Fig. 10 Internal geological geometry. a-c profile: The depths of the erodible layer are 4, 12, and 24 mm, respectively; d, f, i are the front fold of the deposits, e, h, k are the ridge on the surface of the zone II; g, l are the cross section of the deposits parallel to the horizontal plane; j is the surface bulge when De = 24 mm. The dividing line of zone I and II is the upheaval, the upheaval to the slope break is zone II, and the rest part is zone I; Zone Ia and zone Ib are bounded by the front of sliding mass. Here the zone Ia is the upheaval to the front, and zone Ib is the front to maximum disturbance range boundary of the erodible layer kinetic energy of the sliding mass and its weakening erosion ability.
The position of the erodible layer may play an important role on the mobility of the sliding mass. Both the mobility and duration of the sliding mass will increase as the increasing thickness of the erodible layer when it is at an inclined plane (Mangeney et al. 2010). However, the increasing mobility will be maximum at D e = 4d to 24d (d: particle diameter) when the sliding mass is compacted (Farin et al. 2014). With the propagating of the sliding mass, its velocity decreases, and consequently decreasing in erosion depth. During its slowly propagating, there is almost no erosion (Farin et al. 2014). The larger the velocity of the sliding mass is, the stronger its erosion ability is, and the deeper the erosion depth is, which is also true for the present study. The difference of this study with those previous studies is that the mobility of the sliding mass is inhibited when the sliding mass is on the horizontal plane (Fig. 7). It can be seen that the position of the erodible layer on the landslides' motion path has significant influences on its mobility and erosion ability.

Internal and external structural characteristics of the deposit
In physical model experiments, it is common to analyze the interaction between the sliding mass and erodible layer through the internal structural characteristics observed from a profile of the deposit (Deng et al. 2015;Paola et al. 2009). The structures of buckles, recumbent folds, thrust shear, horizontal shear, and convoluted bedding revealed in the profile of the deposits in this study indicated that there was an intense interaction between the sliding mass and erodible layer during sliding mass motion Figs. 9 and 10.
With increases in the D e , the effect of the sliding mass on erodible layers changed from pushing to covering (Fig. 10). According to the profile features, the sliding mass in zone II mainly acted by pushing, scraping, and carrying the erodible layer to zone I b . The structures formed in zone II are relatively simple, where some ridges formed on the deposit surface (Fig. 10e, h). However, the ridges disappeared at thicker erodible layers (Fig. 10k). The range of zone II along the y-axis became shorter with increases in D e . Zone I b was mainly characterized by compression deformation, demonstrated by a relatively high compression of the strata and complex internal structures, such as buckles, recumbent folds, thrust shear, horizontal shear, and convoluted bedding. There were small bulges (Fig. 10j) and no fully covered area by the sliding mass (Fig. 10f) in zone Ib. These phenomena were due to that the resistance between the erodible layer and horizontal plane is greater than its internal one when the sliding mass interacted with it. The erodible layer in zone I a is mainly subjected to thrusting. There are obvious ridges formed at the front of the deposits (Fig. 10d). With continuous increases in D e , the extent of compression of the erodible layer in zone I a decreased, and the extension of the y-axis gradually decreased until the area disappeared (Fig. 10c). When the erodible layer is thick, the pushing effect of landslide is weakened, and the front sliding mass will over the erodible layer without strong disturbance. The upheaval formed in the deposits separated zone II from zone I b , which is usually caused by compression and accompanied by rapid decline in velocity of sliding mass. When the inner shear stress of the debris in motion is not sufficient to cause discrete shear and inverse thrust, longitudinal compression would take place, forming the transverse ridges in the debris mass (D'Agostino et al. 2013). This phenomenon is also often found in the field, such as in scenarios where suddenly decreases in the topographic gradient; encounters with soft and/or deformable substrates; or slide into water (Clavero et al. 2004;Johnson 1978;Shea et al. 2007).
Folds are common geological phenomena, which form not only under tectonic stress (Deng et al. 2022), but also under sudden loading caused by turbulence or mountain floods (Rana et al. 2016). The friction resistance or shear stress applied to weak strata is the basis for the formation of folds (Owen 1996). The folds in this study were distributed in the erodible layer close to the upheaval and front edge of the deposit (Fig. 10 I b ). Furthermore, a series of compressive deformation characteristics are also found in zone I a (Fig. 10). Through model experiments with colored sand, Dufresne et al. (2009) reproduced the folds sustaining its initial color sequences and reported that there are bulldozing effects in the substrata when being covered rapidly by the propagating avalanche. A resistance was faced by the lower part of the sliding mass, while the upper part continued to move forward and into the core of the fold when the sliding mass scraped the erodible layer. Under the influence of inertial force, the sliding mass convoluted into the fold moved forward along with the folded erodible layer under the force of the trailing sliding mass, leading to the convolution phenomenon in zone I b . The formation process of this phenomenon can explain the thrust nappe structure in loess landslides .
Thrust shear tends to occur in positions where the stress is most concentrated. These shear stresses may combine with the force of gravity to further deform folds in complex ways (Lowe 1975). In this study, the shear position in the deposits mainly developed in the core and limbs of the recumbent folds.
The secondary shear surfaces in the deposits mentioned in Sect. "Characteristics of the internal structure of the deposit" were caused by the stress concentration and velocity differences in the sliding mass. The secondary sliding surfaces formed by thrust shear primarily existed in the interior of the deformed erodible layer and the contact area between the erodible layer and sliding mass. At the thin erodible layer, there were few secondary sliding surfaces with a shallow burial depth, but a number of complex secondary sliding surfaces in the medium thickness erodible layers (Fig. 10a-c). At D e = 24 mm, the increase in resistance led to an upward movement of the secondary sliding surfaces, which reduced their number. Further, thrust shear is common in landslides, and is observed in the Ancient Longwu Xishan Landslide No.2 (Tian et al. 2022).
Although many structural phenomena were reproduced in this study, there are still some others not observed, such as flash heating and consequently friction weakening (Habib 1975), and fluidization of substrata due to the impact of the sliding mass that results in rearranging of the soil skeletal particles (Sassa 1998). In the study by Zhou et al. (2016), the sliding mass entrains in the manner of impact and scour, in which the water contained in substrata plays an important role during the process. However, our experiments is conducted without the engagement of water, which makes the changing pore water pressure and consequently fluidization of substrata impossible Iverson et al. 2011). In addition, diapiric intrusions, which are considered due to compression  or friction shearing (Phillips et al. 2013) in the substrata, are also not observed. Therefore, further experiments need to be performed to reproduce these phenomena.
In fact, during the interaction between the sliding mass and erodible layer, the sliding mass impacted the erodible layer with a certain velocity. The erodible layer was shoveled, scraped, and pushed by the sliding mass. The erodible layer dissipated the kinetic energy of the sliding mass through the collision and friction among particles, thereby restraining the velocity and displacement of the sliding mass. The base of the moving mass accommodated transport by large amounts of simple shear (Roche et al. 2013). The erosion ability of the sliding mass with a certain volume is finite. When D e was small, the interaction between the sliding mass and the erodible layer was gentle. Therefore, the kinetic energy of the sliding mass dissipated due to fewer collisions and less friction among particles. The sliding mass could push most of the erodible layer on its path a longer distance along the bare board. As a consequence, the position of the fold core was far from the slope break. Owing to the hindrance of the erodible layer on the sliding mass being small, the sliding mass could easily climb over the erodible layer and move horizontally on top of the erodible layer. Consequently, the shear surface was relatively gentle. In contrast, when D e was large, the interaction between the sliding mass and the erodible layer was significant; therefore, there was a large fraction of kinetic energy loss in the sliding mass. The sliding mass had to move upward along the erodible layer under the thrust of the sliding mass and the resistance of the erodible layer. As a consequence, a steeper shear surface was formed in the deposit. The geological structures formed in this study are highly consistent with the structural evidence found in the field. Field evidences suggest a bulldozing action occurred at the front of the Nixu rock avalanche (Zeng et al. 2021). The high-speed avalanche debris in motion plowed the erodible alluvial substrates along a basal decollement like a bulldozer. Complex structural phenomena, such as entrainment and diapiric structure, were formed in the deposits. This work reproduced inner structures seen in natural cases and reflected the kinematics in the final deposits.

Generalization of the sliding mass movement process
The motion process of the sliding mass was divided into three stages, according to differences in their velocity and motion characteristics. In the first stage, the velocity of the sliding mass increased sharply to a peak velocity before colliding with the erodible layer (Fig. 11a).
In the second stage, there was a strong collision between the sliding mass and the erodible layer. The length of the sliding mass became significantly shorter, and the velocity of the sliding mass was significantly reduced. While squeezing, scraping, and pushing the erosion layer forward, the sliding mass was also subjected to increasing resistance due to the erodible layer (Fig. 11b). The disturbed erodible layer moved forward along with the sliding mass, which resulted in an increase in the erosion range and a further compressed erodible layer (Fig. 11c). The horizontal thrust of the sliding mass reached a maximum when all of the sliding mass slid on the horizontal basin. Therefore, secondary shear surfaces were formed at the position of the stress concentration (Fig. 11d). The strong interaction between sliding mass and erodible layer led to the sliding surfaces formed not only on the interface between the sliding mass and horizontal basin, but also at the contact surface between the sliding mass and erodible layers and the inner regions of the erodible layers in the second stage. Therefore, the movement mode of landslide was relatively complex and rich structural features were formed.
In the third stage, the interaction between the sliding mass and the erodible layer became gentle, and the disturbed erodible layer became dense under the pressure of the sliding mass. The erodible layer entrained by the sliding mass eventually stopped moving forward at a lower velocity. During this motion process, there was a small relative displacement between the front sliding mass and the erodible layer. Furthermore, a series of buckles formed at the front of the erodible layer under the squeezing of the sliding mass. In this stage, a secondary acceleration was observed when the D e values were 4 mm, 12 mm, 16 mm, and 20 mm because the trailing sliding mass, under the action of an inertial force, would climb over and surpass the front sliding mass obstructed by the disturbed erodible layer.

Comparisons with previous studies
In sandbox experiments carried out by Dufresne (2012), there was an arc-shaped joint between the inclined and horizontal basins. The angles of the plate in the sliding mass zone and the accumulation zone were set 60° and 0°. Experimental results showed that the effects due to the sliding mass interacting with the erodible layer were mainly horizontal shearing. When the sliding mass pushed and collided with the erodible layer, the deep layer of the erodible layer was disturbed. However, the erodible layer was not completely eroded.  Duan et al. (2021) In a study by Crosta et al. (2017), the experimental device used was similar to that in this study. Structures, such as convolutions, buckles, and thrust shearing, were also found in the deposits (Fig. 12a-c), which further demonstrated that these structural phenomena tend to form under the condition that there was an erodible layer on the horizontal basin. However, with an erodible layer of 21 mm it is hard to answer how the internal structures vary at different D e . In addition, the influence of the D e and the characteristics of the amount of eroded layer have not been quantified. Found in this study, with the increase of D e , the action form of sliding mass on erodible layer changes gradually from pushing to covering. At the same time, there is a certain relation between internal structure and surface structure.
The material of the erodible layer used in this study was dry medium-fine quartz sand. This kind of sand is very loose under conditions without compaction; therefore, it can easily be disturbed. In actual landslides, although the erodible layers differ due to different geo-environments, they all have certain structural properties after long-term consolidation. It is difficult for a landslide to cause a large-scale or deep scraping effect when the landslide impacts the erodible layer (Crosta et al. 2015b;Peng et al. 2018). Therefore, recumbent folds are rarely observed in actual landslides. Conversely, the folds, thrusts, and pushover phenomenon on the erodible layer of landslides are more common and significant in actual landslides.
In addition, the volume of landslide also profoundly affects its erosional ability. For instance, the Yigong landslide primarily comprises weathered granite, and the erodible layer is sandy alluvial deposits . The landslide, with a volume of about 91 × 10 6 m 3 along the gully moving downwards, squeezes the erodible layer, with a volume of about 24 × 10 6 m 3 , to form pushover structures. During this process, the erodible layer is scraped and entrained by the sliding mass, which makes the volume of the landslide continuously increase. The large impact force generated by the landslide takes away the erodible layers in the gully, which results in an outcrop of the bedrock in the eroded area (Delaney and Evans 2015;Zhou et al. 2016). The underlying bedrock prevents deeper erosion and the flow obviously do not lose momentum in these sections (Dietrich and Krautblatter 2019). This field observation is reflected in the experiment conducted herein, when the D e is 4 mm. The impact force of the sliding mass makes the erodible layer at the slope break completely erode, resulting in direct contact between the sliding mass and the plexiglass plate. This experiment reproduces the geological phenomena observed in the field.
Another case, the Nixu landslide is also mainly composed of granite, and its erodible layer is mainly horizontal alluvial deposits, composed of coarse sand and gravel (Zeng et al. 2020). The landslide exhibits an intense erosion ability when moving downwards with a volume of about 47 × 10 6 m 3 , and the erodible layer undergoes strong deformation and distortion. Structures, such as crushed substrate clasts, convoluted lamination, diapiric intrusion, decollement, and sand boils, are manifestations of deformation and distortion, as shown in Fig. 12d-h Zeng et al. 2021). Under this condition, the interaction between the landslide and the erodible layer, as well as the structures in the deposit, are similar to those at D e is 24 mm in this study. These structures reflect the large resistance of the erodible layer to the sliding mass. However, due to the differences in the material properties of the landslide and the erodible layer in the field, the structural phenomenon is more widespread than that in laboratory experiments.
Finally, the Xingyuan landslide consists of silty clay, and the erodible layer is alluvial deposits comprising silty clay, sandy silt, and gravel Peng et al. 2017). The landslide, with a volume of 0.17 × 10 6 m 3 , impacts the erodible layer consisting of silty clay, resulting its liquefaction. The position of the sliding surface in the landslide moves down 5-6 m of the ground surface when the landslide falls onto the erodible layer, then the landslide covers the undisturbed erodible layer (Fig. 12i, j). This is similar to the experimental conditions of D e is 24 mm in this study.

Conclusions
In this study, the motion characteristics of the sliding mass, morphological characteristics of the deposit, and the structural changes inside the stratum were studied by varying the depth of the erodible layer. The following conclusions were drawn: (1) Based on the changes in velocity and displacement fields, the motion process of the sliding mass can be summarized into three stages: falling, shovel pushextrusion, and push-nappe accumulation. In the first stage, the velocity of the sliding mass increases sharply to a peak velocity before colliding with the erodible layer. In the second stage, there was a strong collision between the sliding mass and the erodible layer, with a rapid decrease in speed. In the third stage, the interaction between the sliding mass and the erodible layer became weak. In the latter two stages, the existence of the erodible layer on the horizontal basin had an inhibitory effect on the mobility of the sliding mass. The physical reason is the kinetic energy of the sliding mass is dissipated by the shear strength generated by the deformation of the erodible layer. With an increase in the depth of the erodible layer, there is a corresponding increase in the inhibitory effect, resulting in a reduction of movement distance for the sliding mass.
(2) The deposits were divided into three zones (I a , I b , and II), in terms of the structural morphology and characteristics of their positions. The action forms were mainly pushing in the zone II, and covering in the zone I. With the increase of D e , the action form of sliding mass on erodible layer changes gradually from pushing to covering. (3) There were phenomena, such as strata inversion, pushover, and entrainment, which occurred in the depositsthe folds, ridges, and bulge that occurred in the surface of deposits. These structural characteristics reflect the stress states of the laboratory landslides in motion from compressing to shearing. Furthermore, these phenomena have been confirmed in natural landslides, which showed that physical model experiments could be used to study landslides' motion process and interactions with erodible layers.
It is difficult to observe the whole erosion process of a landslide in field, and few information is inferred from the profile of deposit. From the experiments, we find that there are relations between the landslides' internal and surface structures. This information, combined with the experimental results and the geological phenomena through field surveys, will provide a reference for analyzing the movement and deposit morphology of landslides.