Introduction

Seafloor instabilities (such as sediment failures and subsequent mass movements) are common in river delta areas worldwide (Walsh and Nittrouer 2009; Denommee et al. 2016) due to multiple factors such as rapid sediment accumulation and erosion (Zhang and Askarinejad 2019), relatively low permeability of fine-grained sediments, and external loads from earthquakes and storm waves (Henkel 1970; Hampton et al. 1996). These failures often lead to significant environmental and economical consequences as they can potentially cause catastrophic damage to seabed infrastructure such as offshore oil platforms, pipelines, and submarine cables (Coleman et al. 1980; Hooper and Suhayda 2005; Dong et al. 2017), and they are also important drivers of deltaic sediment flux (Obelcz et al. 2017). As a result, seafloor instabilities have been the focus of numerous marine geological studies.

The occurrence process of seafloor instability is a key element for failure mechanism explanation, landslide classification and definition, submarine geohazard assessment, and early-warning system development (Vanneste et al. 2014; Hornbach et al. 2015; Urlaub and Villinger 2018). However, studies on seafloor instabilities are conventionally carried out through geotechnical methods and geophysical surveys using, e.g., geotechnical testing, backscatter, multibeam, side scan sonar, and various sub-bottom profiling techniques (Vanneste et al. 2014; Miramontes et al. 2018; Roesner et al. 2019), which can acquire information of an existing submarine landslide for evaluating its deposits, locations, size, runout distance, volume, and potential preconditioning factors (Masson et al. 2006; Leynaud et al. 2017; Dong et al. 2020). While these geomorphological characteristics of seafloor instabilities can be well documented from geophysical surveys, the initiation and evolutionary process of failures are insufficiently understood based on this dataset alone. To learn more about the mechanisms and processes of seafloor instabilities, more field data on the seabed response (e.g., seabed deformation and pore pressure) under storm waves is essential. However, in situ monitoring of seafloor instabilities, especially the in situ measurement of seafloor deformation, is extremely difficult due to the harsh marine environment during storms. Consequently, the lack of adequate in situ long-term observation technology has become one of the biggest obstacles in this field.

In the past approximately 30 years, with the development of submarine monitoring technologies, in situ long-term observation methods related to submarine landslides have gradually been realized worldwide. These methods include seafloor deformation measurements, pore pressure observation, etc., utilizing technologies such as absolute pressure gauges (Phillips et al. 2008), tiltmeters, tri-axial accelerometers (Prior et al. 1989; Fabian and Villinger 2008; Stegmann et al. 2012), acoustic ranging techniques (Blum et al. 2010), fiber-optic strainmeters (Blum et al. 2008), pore water pressure probes (Sultan et al. 2007), and in situ hydrophones (Chadwick et al. 2012). Although there are many ways to observe the seafloor deformation associated with submarine landslides, the random occurrence of seafloor instabilities often leads to high investments in field observations and a low data return rate. The in situ observation of the seafloor instability process is thus still in a relatively preliminary stage and needs further development.

In the subaqueous Yellow River Delta, submarine landslides are common even over gradients of 0.3–0.4° due to high deposition rates (Prior et al. 1989) and with sediments exhibiting extremely inhomogeneous strength under wave action (Liu et al. 2013). Research on seafloor instabilities in the subaqueous Yellow River Delta started with the project “Sediment dynamics at central south Bohai and the Yellow River entrance,” which was a collaboration between teams from China, the USA, and Canada from 1985 to 1987 (Prior et al. 1989). Topographical evidence of submarine landslides and their distribution in the modern subaqueous Yellow River Delta was investigated using side-scan sonar and sub-bottom profiles. Classification and characterization methods of the submarine landslide system in the subaqueous Yellow River Delta were subsequently proposed based on this data (Yang et al. 1994). It is worth noting that Prior et al. (1989) conducted early in situ observations using two bottom-supported observatories containing tri-axial accelerometers, tilt sensors, and pressure sensors, finding that existing submarine landslides would reactivate during storms. This was preliminarily attributed to wave-induced seabed liquefaction. Subsequent to this project, a series of studies on submarine slope instability in the subaqueous Yellow River Delta were carried out through geophysical surveys (Feng et al. 2004), flume testing (Xu et al. 2016), geotechnical testing, and numerical modeling (Liu et al. 2006; Wang and Liu 2016).

Wave-induced submarine landslides have generally been ascribed to sediment liquefaction and shear failure under storm waves (Rahman 1991, 1997; Zhang et al. 2016). However, the actual occurrence process is insufficiently understood. In order to further explore this question, an innovative instrument system, the Submarine Landslide Monitoring (SLM) system, was developed (Wang et al. 2018) to detect the sediment deformation process, especially during the initial stage of wave-induced seafloor instability. More than 4 months of in situ observations were carried out in the subaqueous Yellow River Delta, and satisfactory field data of seabed deformation was recorded. This serves as a recent advance to the work of Prior et al. (1989), whose packages showed only oscillatory and vertical motion, with cumulative horizontal deformation.

The present study first presents the results obtained from the in situ observation carried out in the subaqueous Yellow River Delta in 2014–2015, with the initiation and evolution process of wave-induced seafloor instabilities then being interpreted from this data set. Wave flume experiments are also presented and used to verify the patterns and physical mechanisms observed in situ. These research findings advance the understanding of the sliding processes and physical mechanisms of wave-induced seafloor instabilities in the Yellow River Delta area. New insight is also into the submarine landslide occurrence process, which is expected to benefit investigations of river delta seafloor instabilities not only in this location but also in other delta areas worldwide involving silty substrates.

General settings

The subaqueous Yellow River Delta is located to the north of the Shandong Peninsula (Fig. 1). The Yellow River is the world’s largest contributor of fluvial sediment load to the ocean (Ren and Shi 1986; Wang and Aubrey 1987) and has diverted 12 times since 1855, gradually forming the modern delta.

Fig. 1
figure 1

Map of the Yellow River Delta, China, with the Chengdao study area marked by a black rectangle and the geophysical survey area indicated by a red rectangle. S1 and S2 are in situ observation sites, while P is a wave observatory platform of the Sinopec Shengli Oilfield

Full size image

The sedimentation process in the subaqueous Yellow River Delta is characterized by rapid deposition, inhomogeneous consolidation, and liquefaction, with the main sediment types being clayey silt, sandy silt, and silty sand. In the 1980s, a large-scale submarine landslide system was identified in the subaqueous Yellow River Delta. This system contains bottleneck slides, translational plate slides, and rotational slides based on their topographical characteristics (Yang et al. 1994). These slides can be classified as silt flows and bottom sediment collapses (Prior et al. 1986b), which are consistent with liquefaction but without visualized in situ evidence.

The Chengdao area is central to the Sinopec Shengli Oilfield, where many pipelines, cables, and platforms are located. This area was the mouth of the Yellow River from 1964 to 1976 (Fan et al. 2006). After river channel migration, low sediment supply has resulted in the erosion and retreat of the original pro-delta apron. The in situ sediments progressively became consolidated and densified under wave action (Jia et al. 2011).

Cyclic wave loading on the seabed surface can lead to slope failures, which are hazardous to the oil and gas facilities in the region (Yang et al. 1990; Sun et al. 2008). For example, a storm wave-induced slope failure caused the breakage of two submarine power cables in 2003 (Fig. 2). The slope failure area is about 370 thousand m2, and the material volume involved is about 1.11 million m3. The sub-bottom profiles show that the post-failure mass areas are characterized by disturbed sediment strata, which are acoustically transparent (Fig. 2). The spatial feature of the disturbed mass appears to be a bowl-shaped shallow arc in two dimensions, with a depth of about 3 m. However, the mechanisms and processes behind these failures remain to be explained.

Fig. 2
figure 2

Multi-beam bathymetry map and sub-bottom profiles of a slope failure which occurred in 2003 at the Chengdao area, which caused the fracture of two power cables. The survey was conducted in 2017; sub-bottom profile surveying lines cover the whole failure area at 80-m intervals, with three typical profiles being shown. The slope failure was initially surveyed in 2004 (Xu et al. 2009)

Full size image

Set-up of in situ observation and flume verification testing

Direct evidences of the wave-induced seafloor sediment deformation were obtained through in situ observation, and the initiation process and evolution pattern of the seafloor instability were derived. The study further validated and detailed the sediment dynamic process through wave flume simulation experiments. The detailed study method and materials of the in situ observation and the flume experiments are outlined in the sections “In situ observation” and “Wave flume verification experiments.”

In situ observation

Observation sites

The in situ observations were carried out at 2 sites in the Chengdao area, as shown in Fig. 3; one was in a disturbed area (S1, 38° 10′ 15.3″ N, 118° 54′ 53″ E) while the other was in an undisturbed area (S2, 38° 11′ 44″ N, 118° 51′ 31.6″ E). The two sites are approximately 7 km apart. S1 was located at the edge of an erosion monadnock with a mean water depth of 6.8 m and a disturbed sediment thickness of 4.8 m. The S2 site was located in a region with relatively flat terrain with a mean water depth of 9.5 m, next to a 4.5-m-thick disturbed area. In spite of a scarp with steep gradient being located northeast of the S1 site, both S1 and S2 were on a flat area.

Fig. 3
figure 3

Multi-beam water depth maps, sub-bottom profiles, and cone penetration test (CPT) results from the observation sites. a The site in the disturbed area (S1). b The site in the undisturbed area (S2). c Cone resistance (MPa) and side friction (kPa) as a function of depth (m)

Full size image

Cone penetration test (CPT) results suggest higher strength for the sediments at S1 than those at S2, especially within the top 1 m. The sediment engineering geological properties (e.g., shear strength) of the mass disturbed in the past are in general better than that of the undisturbed mass in the subaqueous Yellow River Delta, (Xu et al. 2009). The disturbed soil mass was once thought to be related to the liquefaction process, in the compaction/consolidation phase during which the sediments progressively become denser than before failure (Prior et al. 1989; Sumer et al. 2006). The sediments at both S1 and S2 can be classified as clayey silt, with an average silt content of 77.3% and clay content of 18.5% measured in the cores. The mean values of soil properties obtained from geotechnical testing are summarized in Table 1.

Table 1 Physical and mechanical properties of seabed soils to a depth of 5 m
Full size table

Instruments and emplacement

The Submarine Landslide Monitoring (SLM) system (Wang et al. 2018) used in the observation consists of two parts: a monitoring tripod for recording hydrodynamic data (e.g., waves, tides, and currents) and a Shape Accel Array (SAA) for recording seabed deformation at different depths. The tripod, shown in Fig. 4a, carries a variety of sensors including a wave-tide gauge, a current meter, a turbidity meter, and an echosounder. The main part of the SLM system is an SAA that was manufactured in Canada by Measurand Inc. and consists of a rope-like array of tri-axial microelectro mechanical system (MEMS)-chip accelerometers (Fig. 4b).

Fig. 4
figure 4

The instruments used in the observation. a The monitoring tripod for recording hydrodynamic data. b The Shape Accel Array (SAA) for recording seabed deformation at different depths. Note: L is length, Φ is diameter

Full size image

The SAA can be placed vertically into the sediments for measuring the horizontal deformation at different depths, being 4 m long and having a resolution of 2 arcsec and an accuracy of 0.18 mm. The SAA was installed into the seabed using offshore drilling techniques due to the high sediment strength in the Chengdao area. The SAA was embedded between 1.23 and 5.23 m below the mudline, while the hydrodynamic monitoring tripod was deployed less than 10 m from the SAA.

Wave flume verification experiments

Instruments and preparations

The wave flume experiments were carried out in a laboratory wave flume that was 0.5 m wide, 0.7 m deep, and 14 m long (Fig. 5). Regular waves were generated by a piston-type wave maker at one end of the flume (the frequency of the wave maker was regulated to generate regular waves with different wave heights), and a wave dissipating slope was placed at the opposite end. The water depth was maintained at 0.4 m. The sediments were placed in a sediment tank with transparent walls that was 0.5 m wide, 0.6 m deep, and 3.8 m long.

Fig. 5
figure 5

Schematic of the laboratory wave flume

Full size image

The wave heights were measured with a wave gauge. The pore pressure dynamic response under wave loading was measured by piezoresistive pore pressure sensors, 20 mm in diameter and 60 mm in length. The pore pressure sensors were set in the center of the sediment tank at depths of 0.1 m, 0.2 m, 0.3 m, and 0.4 m. An electronic penetrometer was used to measure sediment strength prior to and after each round of wave actions.

The sediments used in the flume experiment were collected from the subaqueous Yellow River Delta. Soil particle size distribution analysis showed that the sediments contain 30.8% sand, 54.2% silt, and 15.0% clay. The collected sediments were air-dried and crushed, and the large particles (gravels and shells) were removed before being used in the flume experiment. The sediments were thoroughly mixed with laboratory-made standard seawater (3.5% salinity) proportionally using a blender in order to achieve the desired water content and to form saturated uniform slurry to fill the sediment tank. The initial bed prepared with this slurry was left to consolidate under gravity for 10 days before the laboratory-made standard seawater was gradually added into the wave flume up to a level of 0.4 m above the soil surface to conduct the wave action experiment. Sediment properties after consolidation but before wave action are listed in Table 2.

Table 2 Physical properties of the sediments sampled before and after wave loading. All values are taken from the average of data obtained in the vertical direction
Full size table

Experimental process

After adding seawater into the flume and settling for 24 h, wave loading was imposed by the wave maker. Three designated wave conditions (with wave heights of 0.05 m, 0.1 m, and 0.15 m) were carried out. After each wave action, there was a resting period of 12 h before the next wave action series. The associated wave parameters are listed in Table 3.

Table 3 Wave conditions used in the flume experiments
Full size table

Sediment strength testing and sampling were carried out during the experiments. Depth profiles of penetration resistance (every 0.05 m into the seabed) were measured every hour during each test run. Sampling was performed at the end of each wave height series using pushed-in-place PVC tubes. The physical properties of the samples taken and the results obtained are summarized in Table 2. During the test runs with wave heights of 0.1 m and 0.15 m, part of the sediment bed failed under the cyclic wave loading, with the interface between fluidized sediments and unfluidized sediments being recorded by marking with dashed lines on the sediment tank wall and photographing. Detailed descriptions of the sediment property test methods and experiment procedures can be found in Jia et al. (2014) and Liu et al. (2013, 2017).

Results

Wave-induced seabed failure-field evidence

The in situ observation was conducted from December 2014 to April 2015 with winter/spring storms occurring frequently. The seabed deformation data was only acquired successfully at the S1 site for the full period. The data recorded, in terms of cumulative deformation, significant wave height, and seabed elevation change, as shown in Fig. 6, suggest that seabed deformation occurred many times throughout the observation, and correlated strongly with storm waves. These results indicate that storm waves are the dominant cause of seabed deformation.

Fig. 6
figure 6

The in situ observation results at S1. a Time series of cumulative horizontal deformation (mm) of the Shape Accel Array (SAA), which is relative to the initial shape of the SAA at 0:00 on December 29, with differing colors denoting various depths. b Significant wave height data (m) from an observatory platform, located about 13 km away from the observation site S1, having slightly larger water depth but the same tendency in wave height. c Time series of seabed surface change (cm); the data was converted from the relative height of the echosounder, fixed on the hydrodynamic monitoring tripod, to the seabed surface. d Flow velocity (m/s), which responded significantly to waves and tides. e Suspended sediment concentration (SSC, g/L), which was calibrated with the turbid liquid samples prepared using the experimental silts from the Yellow River Delta. Note: the yellow shadow bars indicate periods of tripod sinking, while for the two major deformation events A and B, the orange shadow bars indicate the erosion phases and the blue shadow bars indicate the deposition phases

Full size image

Two significant deformation events occurred on 21 January 2015 (event A) and 7 February 2015 (event B) respectively; during these 2 days, the significant wave heights were the maximum observed (> 3 m) over the entire observation period. During these two storms, the hydrodynamic monitoring tripod sunk at least 14 mm into the sediments, indicating that the sediment strength was greatly reduced during the storm.

Data analysis of the tripod sinking is as follows. The elevation change of the seabed surface is derived from the relative height from the echosounder (Autonomous Echologger AA400, fixed to the tripod) to the seabed surface. Generally, if the tripod sinking process is not considered, the decrease and elevation of the seabed surface represents the erosion and siltation process, respectively, with erosion occurring during storms and siltation occurring during calm sea states. During events A and B, the height of the seabed surface increased; however, the significant wave height, flow velocity, and suspended sediment concentration were all large; the aforementioned factors were indicative of sediment deposition not occurring. The footings of the tripod were three circular steel plates; each was 0.6 m in diameter. Steel needles fixed on the footings of the tripod (three on each footing, 0.5 m in length) ensured the stability of the tripod sitting on the seabed, such that the possibility of tripod sideslip caused by storm waves and currents, which might cause changes in the relative height of the sensor to the seabed, was also minimized. The tripod weighted approximately 200 kg; hence, the seabed only needed 2.4 kPa bearing capacity to ensure the stability of the tripod, if any additional bearing capacity provided by the steel needles is neglected. The foundation bearing capacity at S1 was estimated to be greater than 130 kPa (Wang et al. 2018) if the in situ undrained strength was adopted. It is therefore inferred that a significant decrease in sediment strength occurred during events A and B, resulting in the sinking of the steel plate footings. On this basis, it is concluded that the relative elevation “increase” of the seabed was induced by the sinking of the tripod.

At S2, we also obtained evidences of tripod sinking (shown in Fig. 7), in two events which occurred on December 10 and 15, 2014, respectively. The tripod sequentially sank 8.7 cm, then a further 16.5 cm into the seabed, and when the tripod was recovered, the total amount of tripod sinking was 60 cm measured by the diver.

Fig. 7
figure 7

The tripod sinking events during two storms at S2, marked by yellow shadow bars. a Significant wave height data (m) from S2. b Time series of seabed surface change at S2 (cm), with data being converted from the relative height of the echosounder, fixed on the hydrodynamic monitoring tripod, to the seabed surface

Full size image

Under the same wave conditions, the seabed surface change at S1 was low (within 0.5 cm), which was determined by the different sediment strengths between the two observation sites. The shallow sediments at S2 have a lower strength than that at S1, and thus have a greater probability of sediment failure during a storm. Sediments at S1 were disturbed and had high strengths. Nevertheless, reactivation of seabed failure during large storms was observed at S1, which is consistent with the observation of Prior et al. (1989).

Based on the geophysical survey data after the observation, the existing disturbance area at S1 experienced repeated storms and seabed deformations with no enlargement and significant change. There was also no disturbance for the seabed stratum at S2. This indicates that the degree and scale of seabed failure caused by the storm during the observation period were not large enough to trigger large-scale mass movement and sediment stratum disturbance. Nevertheless, this was the first time the deformation process of sediments was recorded at various depths during storm events by means of in situ observation. The observations show that the storm waves directly triggered the seafloor sediment failure in the subaqueous Yellow River Delta, causing strength reduction in sediments and consequently tripod sinking and seabed deformation.

Seabed failure evolutionary process

Figure 8 presents the SAA data recorded during event B at the S1 site. The results show that seabed failure occurred three times during this storm process, and the evolution of the maximum deformation depth interface moves from shallow to deep, and then migrates upward from deep to shallow layers (the shallow-deep-shallow pattern). This indicates that the process of wave-induced seabed failure followed by restored stability may occur multiple times in a single storm.

Fig. 8
figure 8

The SAA deformation, tripod sinking, and wave height data recorded during 7 February 2015 (event B) at S1 site. The seabed sliding deformation depth varied from shallow to deep, and back to shallow again. The deformation cases basically aligned with tripod sinking, and reactivated three times in a single storm. a The absolute deformation versus depth in 30-min intervals at S1 on 7 February 2015 (event B). The maximum deformation depth was approximately determined by the displacement at which depth the deformation was greater than 0.35 mm. b Sinking displacement of the tripod (cm), which was converted from the relative height of the echosounder to the seabed surface. c Significant wave height (cm) during the storm, with the data from S2. Note: the yellow shadow bars are periods of seabed deformation (cases a, b, and c), the red bars mark the maximum deformation depth, the scale of the absolute deformation plots at 14:00 and 14:30 are 0–14 mm and 0–4 mm, respectively, and 0–2 mm for the others

Full size image

The significant wave height during event B reached 3.3 m, with the maximum cumulative seabed deformation reaching 13 mm and the tripod sinking at least 14 mm into the seafloor. The seabed failure was a continuous and repeated process (as shown in Fig. 8) according to the deformation and tripod sinking data which were analyzed over 30-min intervals. The SAA recorded three sediment deformation cases during the day (cases a, b, and c; the maximum absolute deformation in case c reached 10.7 mm and the maximum depth of deformation was 4.73 m), and tripod sinking occurred twice corresponding to the occurrence times of cases a and c. The occurrence of seabed deformation and tripod sinking did not exactly match, possibly due to the tripod being separated from the SAA sensor by about 10 m in situ and the different critical conditions of the two cases. However, these features were generally a manifestation of the seabed failure and reflected the process of such failures.

The horizontal cyclic shear stress in seabed sediments was produced through the alternating actions of the wave crest and trough. While sediment failure occurs, the state of sediments changed from elastoplastic to non-Newtonian fluid behavior, and cyclic oscillation occurs under the action of the cyclic shear stress (Liu et al. 2013; Zhang et al. 2018a). According to the in situ observation results, the storm-induced seabed deformation process resembles an oscillating motion. Taking the seabed deformation event B as an example, the sediment deformation directions during the period were inconsistent at each depth, and the deformation at the same depth did not develop in a certain direction. Figure 9a–d shows the deformation direction evolution process at four depths. The y-axis and the x-axis represent the two coordinate systems of the SAA on the horizontal plane, that is, the top view of the four depth planes. Figure 9e shows the distribution of the deformation in the y-axis and x-axis versus depth, showing that the deformation directions of each depth were inconsistent. This was because the interface of the maximum deformation depth was not unique, but evolved in a shallow-deep-shallow pattern, thus causing inconsistencies in the deformation direction at various depths (detailed verification tests resulted in a modification of the explanation of Wang et al. (2018); the influence caused by the SAA deformation in the shallow-layer sediments to those in the lower layer is almost negligible). The partial enlarged views in Fig. 9a, b show the development of the SAA deformation direction at the same depth, exhibiting reciprocal and irregular features. This is because of the irregular waves in situ, preventing the wave direction from being consistent like in a regular wave and generating regular cyclic shear stress such that the deformation direction exhibits a disorder and irregularity.

Fig. 9
figure 9

The deformation direction evolution process of the sediments, showing that the failed sediments acted in cyclic oscillation, and the deformation direction changed with time and depth. a–d The deformation direction at four depths at S1 on 7 February 2015 (event B). The red arrows represent the deformation direction evolution process. The y-axis and the x-axis represent the two coordinate systems of the SAA on the horizontal plane, i.e., the top view of the four depth planes. Partial enlarged views are shown in a and b. e The distribution of displacement (mm) of the y-axis and x-axis along depth at S1 during event B with the color changing with time

Full size image

Recurrence of wave-induced sediment failure in flume testing

The flume experiment results show that the failure zone gradually expanded after sediment failure was initiated under wave actions. Once the maximum range was reached, the region of the failed sediments shrunk back, corresponding to the shallow-deep-shallow development pattern observed in the field. The failed sediments underwent oscillating motion governed by wave action. After the experiment, the sediments were reconstructed, the stratum was re-structured, and the strength was improved.

In the tests with larger wave heights (H = 0.10 m, 0.15 m), sediment failure was observed, with the sediment states changing from elastoplastic solid to non-Newtonian fluid behavior. The failed sediments experienced cyclic oscillation, in a movement state consistent with the water particle trajectory (ellipse). Figure 10a shows the oscillating sediment interface in a deep-colored arc shape photographed through the transparent sidewall of the soil tank, with sediments below the interface being stable. Macroscopic vertical micro-drainage channels developed near the interface, as discussed by Liu et al. (2017). The pore water discharged upwards through the micro-channels and carried fine-grained sediments to form light-colored channels. The sediments near the boundary gradually become coarser, forming a macroscopically deep-colored interface, with the interface expanded vertically and laterally as wave action continued. When the maximum depth was reached, as shown in Fig. 10b, the deeper sediments no longer displayed failure and the interface stopped expanding. The failed sediments began to re-consolidate and the interface gradually recovered from deep to shallow layers (Fig. 10c, d). New stratums of the failed sediments were re-structured in the process of interface recovery, resulting in graded bedding, convolute bedding, and parallel bedding (Xu et al. 2016). In general, the evolution process of sediment failure under wave action can be summarized into two stages: the initial failure zone expansion and the later failure recovery. The evolution of the failure interface on the glass wall of the soil tank was observed to move from shallow to deep, and then migrated upward from deep to shallow layers.

Fig. 10
figure 10

Photographs of the flume experiment, showing the development of the sediment failure process. a Sediment failure occurred under wave action, and a clear deep-colored arc interface appeared between oscillating sediments and stable sediments. b The interface migrated downward and reached the maximum depth. c and d The failed sediments began to re-consolidate and the interface gradually migrated upward. The post-failed sediments displayed new stratum features different from the original bed. e The final sediment stratum formed by the failure process of expansion and recovery. Note: For clarity, images a–d are all selected at the same place of the tank and at the same scale (cm), while the image in e shows a more clear stratum from another flume experiment, which was conducted on December 21, 2015, via the same method as that in this article

Full size image

Figure 11 shows the pore pressure development around the startup of sediment failure during the flume experiment. In the early stage (before 3250 s), the state of sediments near the pore pressure sensor was stable, and there was no sign of instability. The pore pressure showed a continuous accumulation trend as a whole, but the accumulation rate was slow, and the pore pressure still remained stable during the period of 2700–3000 s. At around 3250 s, sediment failure occurred near the sensor, the oscillation of the soil grains started, and the pore pressure increased suddenly with a faster accumulation rate.

Fig. 11
figure 11

Time series of pore pressure development around the startup of sediment failure during a flume experiment, which was conducted in a similar experiment in the same flume on December 21, 2015 (with a wave height of 0.16 m). The sensor was buried at 0.12 m depth below the mudline in the sediments. The red line is the smoothed curve

Full size image

The strength of the failed sediments will increase relatively during the recovery process (Xu et al. 2016). At the end of the experiment, the penetration resistance of the seabed was measured along the parallel section of the flume, and a contour map of the penetration resistance is shown as Fig. 12. It can be seen that the entire soil bed exhibited an inhomogeneity in penetration resistance. The area of high strength corresponded to post-failed sediments, while the areas of the undisturbed sediments and the continuous oscillating sediments were with lower strength.

Fig. 12
figure 12

Spatial isopach map of sediment penetration resistance (N) tested after the flume experiment (the lower left corner of the sediment tank is the coordinate origin; units: cm)

Full size image

Discussion

Trigger mechanism of sediment failure initiation

The startup mechanisms of wave-induced sediment failures and further mass movements mainly include liquefaction failure and shear failure (Rahman 1991, 1997; Zhang et al. 2016). Liquefaction is an extreme form of seabed instability (Jeng 2001). When this occurs, sediments lose all of their structural strength unless a dilative response is exhibited, and behave as a heavy liquid with no rigidity. In this state, the sediments can be readily transported as fluids. Storm waves inducing shear stress at a point within the sediments may be significant enough to overcome its shearing resistance, causing it to fail without liquefaction. This type of seabed instability, referred to as shear failure, may produce a horizontal displacement of sediment mass. Shear failure is more likely to occur when the seabed consists of cohesive sediments, while liquefaction failure is expected mainly for seabeds with non-cohesive sediments (Rahman and Jaber 1986; Rahman 1991; Jeng 2001). The sediments in the subaqueous Yellow River Delta are silty soil, displaying properties between those of sand and clay. The real nature of wave-induced seabed failure in situ the Yellow River Delta is therefore very complex.

Regarding the initiation mechanism of wave-induced silty sediment failure, the following paragraphs present an analysis from the view of pore pressure in the wave flume experiment as shown in Fig. 11. The startup process of the silty sediments reflected by the pore pressure does not exactly match the liquefaction or shear failure mechanism. It is not rigorous to describe the process by the terms “liquefaction” or “shear failure” alone. Sumer et al. (2006) used loose silty sediments prepared by natural sedimentation to simulate the process of liquefaction failure in a wave flume test. The pore pressure in the loose sediments accumulated rapidly under the action of waves and took only a few seconds to liquefy. The situation, however, is merely applicable to new deposits, while the in situ sediments in the Chengdao study area are much denser under long-term wave action, and may be over-consolidated (Jia et al. 2011). It is difficult for a large accumulation of excess pore pressure to occur under wave action because of the high density state, and it is unreasonable to justify the sediment failure via liquefaction theory alone.

The initiation mechanism of the silty sediment failure induced by waves in the subaqueous Yellow River Delta (especially in the Chengdao area) can be explained by the coupling effects of cyclic wave loading (Fig. 13). The waves on one side generate cyclic shear strains in the sediment and will gradually recombine the sediment grains at the expense of the pore volume of the sediment. The latter effect will “pressurize” the water into the pores, leading to a gentle accumulation of excess pore pressure, and the buildup only develops due to the dense sediment state and the partial dissipation of the accumulated excess pore pressure. The internal pore pressure thus accumulates, and the sediment strength decreases accordingly, and when the wave-generated shear stress at a point within the sediment exceeds the degraded shear strength, sediment failure occurs. At this time, the large shear strain causes the sediment grains to rearrange and the pore volume to increase, resulting in water being “inhaled” into the pores and sharing much of the effective stress between the particles, resulting in the pore water pressure increasing rapidly under load. In summary, the increase of pore water pressure under wave action reduces the shear strength of the sediment, and the wave-induced shear stress drives the initiation of sediment failure.

Fig. 13
figure 13

Time series of pore pressure development around the startup of sediment failure and the conceptual model for sediment responses during the process. The difference in seabed volume over time represents the strain of sediments

Full size image

Physical mechanisms of the sediment failure evolutionary process

The results of in situ observation and flume experiments show that the depth of wave-induced seabed failure follows a shallow-deep-shallow development pattern. This suggests that the sliding surface of the submarine landslide is not constant in the subaqueous Yellow River Delta, especially as the seafloor instability process manifests with a gentle slope angle and low sliding speed. The physical mechanisms of the sediment failure process under wave action are of great significance to revealing the evolution of the sliding interface during submarine landslides.

Once sediment failure is initiated, the shallow soil undergoes periodic and reciprocating oscillations. The reciprocation and periodicity depend on the alternating action of the wave peak and trough. Therefore, the deformation displacement obtained in situ may be smaller than the actual amplitude of oscillation. The structure of failed sediments becomes loose, exhibiting a plastic flow state with large deformation (resembling a non-Newtonian fluid), and the soil strength is significantly reduced, resulting in a decrease in foundation bearing capacity and permitting the in situ observation instruments to sink (the tripods in this study and the observation package of Prior et al. (1989)). The failure of the shallow-layer sediments makes the seepage action intense at the failure interface, and the effective stress is reduced additionally; thus, the failure of the underlying layer is more likely to occur. Therefore, the failure depth and the horizontal extent of the failure zone continuously increase under the continuous action of wave loading.

The wave-induced cyclic shear stress in seabed sediments decays with depth. According to the principle of limit equilibrium, as the failure interface reaches a certain depth, where wave-induced shear stress is insufficient to promote the continued expansion of the failure zone, the expansion stops. From this point onwards, the pore pressure dissipation rate in the sediments (enclosed by the interface) exceeds the cumulative rate (Sumer et al. 2006) and re-consolidation governs the transition of the sediment to stability. During the re-consolidation process, the excess pore water dissipates upward along the pore pressure (seepage) gradient (Liu et al. 2017), and as the sediment grains settle down through the water and gradually join together to form a stronger structure, the effective stress recovers. This process is also called dynamic “compaction” (Sumer et al. 2006) or dynamic “solidification” (Miyamoto et al. 2004). The process gradually develops from the interface and migrates upwards until it reaches a new equilibrium again. The waves in the flume experiment are regular forward waves, with constant wave runs during each test wave train; therefore, the development process of sediment failure (the shallow-deep-shallow pattern) occurs only once. Waves in situ are irregular waves, so failure may repeat in the post-failed sediments triggered by waves with larger wave height, and multiple “reactivations” of sediment failure during one storm may be observed. The physical and mechanical properties of the sediments after dynamic consolidation are expected to change. Compared with the unfailed sediments, the water content and void ratio are expected to be lower, while the density/compactness (Table 2) and the soil strength would increase (Fig. 12). The engineering properties of the sediments would be significantly improved (as shown by the contrast of penetration strength in in situ observation points S1, S2, Fig. 3), and the sediments subjected to stronger wave actions are more significantly improved, as compared in Table 2.

With the cumulative excess pore pressure in the failed sediments, a pressure gradient in the vertical direction is generated, and seepage occurs with pore water dissipation. The sediment fine particles are transported upward along the seepage channel under the drag of the seepage force (Foda and Tzang 1994), separating from the soil skeleton and forming a so-called “internal sediment suspension” (Tzang 1998). A floating mud layer composed of fine-grained sediments will be formed on the surface of the seabed after the wave-induced sediment failure process (Zhang et al. 2018b; Jia et al. 2020), and the fines content in the sediments will be significantly reduced (Xu et al. 2016). All of the above indicate that the sediment failure process has a transformative effect on the seabed. The original sediment structure is broken and recombined during the process, and migration of sediment grains may occur.

Implications for seafloor instabilities in the subaqueous Yellow River Delta

The seafloor instabilities in the subaqueous Yellow River Delta mainly include in situ collapse and translational slides such as silt flow, etc. (Prior et al. 1986b; Feng et al. 2004), and are controlled by the coupling effect of wave-induced sediment liquefaction and shear failure. According to the results of in situ observations, wave flume tests, and geophysical surveys presented in this study and others, the seafloor instability development pattern is summarized in Fig. 14.

Fig. 14
figure 14

Schematic model for seafloor instability evolution including collapse depressions (a–c) and translational slides (d–f) in the subaqueous Yellow River Delta. The sliding surface dynamically changes during the process. a, b Initiation and expansion stages of sediment failure on the horizontal seafloor, where the collapse depression feature gradually appears; c retraction stage of the sediment failure, where the depth of the depression increases. d, e Initiation and expansion stage of sediment failure on a slope, where the failed sediments gradually transport downslope and spill over the initial arc failure area. f Erosive silt flow tongue slides downslope and the “cut and fill” structures are formed

Full size image

Once wave-induced sediment failure starts up, the failure zone will gradually expand. However, the failed sediments cannot slide downslope where the seabed is flat and horizontal. In this scenario, it can only undergo a circular arc shearing slip in situ, which is also a form of submarine landslide (Terzaghi 1956; Coleman and Garrison 1977; Xu et al. 2009). During the retraction stage of the failure zone, the sediment reconsolidates dynamically under the action of waves and becomes denser. The seabed surface is thus lowered, eventually forming a geological depression feature. Collapse will occur more frequently at the estuary of the Yellow River (Prior et al. 1986b), where the newly deposited sediments are looser and have more pores.

When the slope angle reaches a certain value, the failed sediment mass will transport downslope under the action of gravity. A main scarp will be formed at the source area, and the transported sediments will be deposited in the downstream of the landslide, before spilling over and continuing to slide to form a silt flow tongue (Yang et al. 1994). The overflow of high-concentration sediments will cause erosion on the seabed (Gee et al. 2005) and form a silt flow gully, which will then be buried by later sedimentation, forming complex “cut and fill” structures (Prior et al. 1986a). The depth of sliding surface changes dynamically in this failure cycle, characterized by the two stages of failure zone expansion and retraction during failure and caused by the increase of the failure depth induced by the erosive silt flow tongue in an event of overflow runout.

After mass movement, the initial seabed slope is re-constructed to form a more stable gentle slope. Although the disturbed seabed sediments may fail again during the next severe storm, it is difficult to start up a further mass movement on the gentler and more stable slope. This explains the observations in this study and those exhibited in Prior et al. (1989), specifically how the existing landslide features experienced reactivation without enlargement or upslope retrogression.

Conclusions

In situ observations of seafloor instabilities were carried out with displacement sensor arrays and hydrodynamic tripods in the subaqueous Yellow River Delta. Wave-seabed interaction experiments were also conducted in a wave flume with silty sediment. The in situ observation methods are advanced, with which wave-induced seabed lateral deformation was recorded at different depths for the first time. The initiation and evolution process of wave-induced sediment failure during seafloor instabilities were thus identified and the physical mechanisms were discussed. Furthermore, seafloor instability process patterns including collapse depression and translational slides were identified. The key conclusions are:

  1. 1.

    The initiation of wave-induced sediment failure in the subaqueous Yellow River Delta is controlled by the coupling effect of liquefaction and shear failure, i.e., wave-induced accumulation of pore water pressure reduces the shear strength of the sediments and the wave-induced shear stress drives the initiation of sediment failure.

  2. 2.

    Once sediment failure is initiated, the shallow soil undergoes periodic and reciprocating oscillations on the flat seabed. The reciprocation and periodicity depend on the alternating action of wave peaks and troughs.

  3. 3.

    The wave-induced sediment failure process includes an expansion stage and a retraction stage of the failure zone. The sliding interface develops following the shallow-deep-shallow pattern.

  4. 4.

    There could be multiple reactivations of sediment failure followed by restored stability during a single storm in the subaqueous Yellow River Delta.

  5. 5.

    The seafloor instabilities in the subaqueous Yellow River Delta, characterized by disturbed sediment strata and with shallow arc spatial features, mainly include in situ collapse depressions and translational slides. The sliding surface depth dynamically changes, controlled by the sediment failure process.