1. Introduction
Rivers and streams are dynamic systems and are continuously changing their structure due to different flow conditions. Riverbank erosion [
1] commonly refers to the removal of bank material by flowing water or carried sediment. This is a phenomenon that derives from two categories of factors [
2]: natural (e.g., climate parameters such as precipitation type, intensity, and variability, or soil properties like water content and shear strength and type of vegetation) and human actions [
3] (e.g., construction of dams, logging, and intensive grazing).
The phenomenon of riverbank collapse incorporates a variety of bank and riverbed deformations in the affected sections (e.g., the longitudinal erosion and deposition of material on the riverbed or the transverse deformation of riverbed channels) [
4]. These deformations assume different shapes and typically alter the cross-sectional morphology of the river affected. This phenomenon plays a very important role in the evolution of rivers, and despite the fact that some stable rivers have a healthy amount of erosion from which they benefit, unstable rivers and the erosion taking place on their banks are a cause for economic, environmental, and social concern: for example, (i) people are forced to migrate due to land erosion; (ii) riverbank collapse causes the loss of large areas of farmland; (iii) riverbank collapse can change the original boundary conditions of the river as well as the water and sediment conditions when large amounts of sediments enter the river channel and siltation occurs; (iv) the sediments eroded due to riverbank collapse are one of the main sources of river sediments and make the river muddy, originating environmental and ecological problems. Therefore, this phenomenon is a great concern for the society and researchers should accurately characterize and assess the several causes that typically accelerate this phenomenon that leads to major impacts, and they should identify feasible solutions for mitigation and adaptation strategies to be implemented.
For non-cohesive riverbanks, the sediment particles on the bank slope are mainly affected by the thrust on the bank, the uplift force, and the gravity effect generated by the water flowing in the river [
5]. When the slope of the riverbank is larger than the underwater angle formed by the deposition of eroded sediment, the soil within the upper layer typically collapses along a sliding surface, usually in the form of “shallow collapse” [
6]. Particle entrainment can be quantified using the magnitude of the shear stress and the particle size [
7,
8] for each soil type. On the other hand, cohesive riverbanks, more commonly found worldwide within river streams, are not only subject to these forces but also to the inter-particle cohesion magnitude. Lohnes et al. [
9] completed a study to analyze the stability of cohesive riverbanks by calculating the ratio of the driving forces to provide resistance on the collapse surface, and proposed a hypothetical collapse model to be associated with cohesive riverbanks. However, the model that Lohnes et al. [
9] developed does not take into consideration the effects of tensile cracks [
10], pore water pressure [
11], and hydrostatic pressure, and only assumes that the collapse surface can pass through the slope foot and therefore this method can only be applied to relatively steep cohesive riverbanks [
12,
13]. On this basis, Osman et al. [
10] established an additional collapse model for cohesive riverbanks, which takes into account the effects of tensile cracks and assumes that the collapse surface passes through the bank slope foot. The study conducted by Darby et al. [
14] also considered the actual topography of the riverbank, plus the pore water pressure and the hydrostatic pressure. Few years later, Rinaldi and Casagli [
15] introduced the suction component of saturated and unsaturated parts of the riverbank into the identified collapse mode. Over the last decade, the bank stability and toe erosion model (BSTEM) proposed by the US National Sediment Laboratory has been widely used to numerically quantify the erosion within riverbanks. In this model, the process of riverbank collapse is divided into two parts [
16]: (a) the slope foot erosion and (b) the riverbank stability analysis. However, other studies have been completed to provide alternative numerical and mathematical solutions. Huang et al. [
17] also established a mathematical model for the collapse, quantifying the factors affecting the riverbank stability. Wang and Kuang [
18] derived an equation for calculating the critical height of the initial and secondary riverbank collapse while Xia et al. [
19] established a secondary collapse model to be used for cohesive riverbanks that analyzes the forces applied on the bank using the soil typical of the lower Yellow River.
Furthermore, the composite riverbank of an alluvial river generally exhibits a dual structure of two single layers (with different thickness and same distribution) or a tiered structure (with different thickness and distributions). For this last configuration, cantilever collapse [
20] and piping collapse [
21] are more likely to occur. Xia et al. [
22] studied the dual structure experienced within the lower Jingjiang River and specified the three stages of dual structure cantilever collapse, quantifying the influencing factors and analyzing the collapse process. Once the riverbank collapse has taken place, the eroded material is transported within the river and this process is composed of three steps: (i) riverbank collapse and movement of eroded particles; (ii) sediment deposition and its interaction within the riverbed; and (iii) sediment transport along the river. Nagata et al. [
23] combined the riverbank collapse model with a two-dimensional mathematical model to simulate the deformation process of the river channel. Darby and Delbono [
24] combined a two-dimensional model with the collapse mode of cohesive riverbanks to calculate the deformation process of curved channels. Simon et al. [
25] comprehensively considered the impact of sediment accumulation from eroded banks and collapsed banks on the riverbed, using the BSTEM model to estimate the riverbank collapse and material sedimentation under different flood conditions. Darby et al. [
26] predicted the evolution of river channels composed of fine sand by coupling an infiltration model and the collapse model. Jia et al. [
27] combined the dual structure collapse mode with the three-dimensional model to simulate the evolution of Shishou Bay in the lower Jingjiang River. Nardi et al. [
28] simulated the evolution of rivers where bed is composed by medium grained sand. Xiao et al. [
29] simulated the river evolution process under the influence of vegetation by combining non-cohesive riverbank collapse models with the shallow water equation. Xia et al. [
30] established a mixed model of two-dimensional riverbed deformation in the orthogonal curvilinear coordinate system, and simulated the evolution process of the wandering sections of the lower Yellow River. Jia et al. [
31], based on the Osman model, built a three-dimensional water sediment model with bank deformation considered and effectively simulated the horizontal oscillations of river channels caused by cohesive riverbank collapse. These studies demonstrate that the riverbed erosion is a phenomenon clearly observed but they do not consider the characterization of the motion of the sediment eroded and the load in the riverbed. Without considering these aspects, it is very difficult to replicate the riverbed erosion phenomenon with numerical models due to the paucity of existing datasets useful for calibration and validation of numerical tools. Mathematical models developed to date can be used to obtain riverbed and riverbank deformations but are applicable only under limited boundary conditions. Furthermore, it is still very challenging to numerically replicate multiple conditions associated with various slopes and flow conditions because of the paucity of high resolution localized data for this type of phenomenon. Hence, more field or experimental studies are needed to calibrate and validate the dynamic features associated with riverbank erosion under numerous conditions.
Focusing on previous studies based on physical models, Yu et al. [
32] used a flume test to qualitatively analyze the interactions between material eroded due to hydraulic effect and its deposition in the river. Yu and Guo [
33] studied the coupling relationship between material eroded and its transportation and deposition in the riverbed using a curved channel flume. Wu and Yu [
34] revealed new insights on this natural phenomenon for cohesive riverbank collapse, non-cohesive riverbank collapse, and riverbed deposition via experimental tests. Deng et al. [
35] simulated the collapse process of the upper Jingjiang Riverbank by combining the longitudinal deposition of the riverbed surface with the secondary collapse mode of cohesive riverbanks. Yu et al. [
36] studied the interactions between bank collapse and riverbed deposition for different near-shore riverbed compositions by using a curved channel flume to complete the experimental tests.
Physical experiments can qualitatively measure the amount of material eroded and the amount of material deposited within the close riverbed sections to fulfill the gaps previously described. This paper presents the results obtained with an experimental flume constructed at the Key Laboratory of Water and Sediment located within the School of Environment of Beijing Normal University, investigated the phenomenon of riverbank collapse utilizing a variety of slopes (45°, 60°, 75°, and 90°) for cohesive banks, and characterized the transport of riverbank eroded material within rivers under dissimilar flow conditions (45 and 60 L/s). The datasets provided led to new insights for riverbank erosion under novel specific physical and hydraulic conditions and can be used by numerical modelers to validate relationships between variables associated with this natural phenomenon.
4. Conclusions and Discussion
The conclusions can be summarized as follows:
- (1)
Experimental tests conducted have enabled the characterization of the riverbank collapse process within four different slopes simulated. This phenomenon can be subdivided into multiple steps: (i) the foot of the bank is frequently affected by the washing effect of the water flowing within the river; (ii) small fragments start to fall down along the bank and a groove is usually gradually formed on the bank toe. The groove becomes deeper and deeper as the water flow continues to wash the bank. Due to the strong bonding effect, the cohesive soil in the upper part of the groove is suspended. When the gravity of the suspended soil is greater than its anti-sliding force and exceeds its stable slope, cracks form along the bank slope. The cracks deepen until the collapse occurs and part of the bank falls in blocks inside the river. When the bank slope is steep, its collapse surface is almost flat, which is consistent with the collapse observed at Dengkou River section [
37].
- (2)
Topographic surveys were completed to characterize the riverbank collapse under different slope scenarios and the bank collapse magnitude is directly correlated with the water flow rate; in fact, the greater is the flow rate, the faster the collapse occurs and faster is the time for the collapse to initiate. Moreover, the pore water pressure monitored during the occurrence of the riverbank collapse, is an indicator of how quickly this phenomenon arises (e.g., for gentle slopes, the internal pore water pressure of soil fluctuates for a longer duration, particles interact more frequently, and the shear resistance seems to be stronger than the shear resistance of steep slopes).
- (3)
The process of sediment transformation from bank collapsed materials to river sediment was analyzed through monitoring the sediment concentration at typical sections and the tailgate. It was found that the change of sediment concentration with time is basically consistent regardless of the initial bank slope. At the beginning, the sediment concentration increases slowly, and then rises dramatically. The rapid rise of the curve corresponds with the occurrence of the collapse along the riverbank. When the high amount of debris enters the river, it mixes with the water, increasing the sediment concentration along the stream and reaching the maximum value recorded. Consequently, it gradually decreases and becomes stable. The rapid increase of sediment concentration due to collapse is between 35 and 45 min after the start of the experiment, but for the tailgate section, the maximum sediment concentration is between 40 and 50 min. The maximum sediment concentration of the tailgate section arrives later than that of section #3. This is due to the fact that when collapsed soil enters into the water, the water requires time to carry it downstream. Additionally, it was observed that the maximum sediment concentration of the tailgate is smaller than that of section #3. That is due to the typical sediments transport process within river.
- (4)
When the sediment enters the river, part of it is carried downstream by the current, and the remaining part deposits at the foot of the bank. The deposited sediment, under the action of water flowing, is further activated and then transformed into bed load and suspended load, which is transported downstream as well within the water. In terms of quantity, about 7–10% of the sediment typically deposits locally, and the rest continues to move downstream with the water flowing, part of which (70–77%) forms suspended sediment and another part (15–20%) forms bed load.
Despite the insights provided with this study, it is fundamental to state its limitations considering the complexity of the riverbank collapse phenomenon. To study the mechanisms of riverbank collapse, it is necessary to consider not only the longitudinal erosion and the siltation effect of the riverbed due to water and sand, but also the transverse deformation of the river channel (or the collapse of riverbank). In natural rivers, the longitudinal erosion and siltation of the riverbed, riverbank collapse, and sediment transportation are simultaneous, hence it has been very complex to simulate all these phases. Additionally, since the mechanism of bank collapse and the interaction between sediments and silt bed are not completely explored yet, it is difficult to accurately simulate the process of sediment transportation in the bank collapse section by using either a mathematical or a physical model. However, we were able to provide new insights regarding the longitudinal deformation and the transverse deformation of the river, involving the degree of deformation in a certain period of time under various flow conditions, furnishing details about the amount of riverbed erosion, the amount of siltation, and the amount of riverbank collapse. Hence, this study can be a considered as a preliminary phase to investigate bank erosion and sediment transportation mechanisms due bank collapse. Future research should also address other limitations. Firstly, bank shapes in natural rivers are irregular, so more bank angles should be considered. Secondly, the bank is typically eroded by sediment gravity flow in natural rivers, so the effect of sediment concentration on bank collapse should be added when simulating the water flowing, which is never completely clear. However, results obtained with this study should be used to calibrate existing and new numerical models and such recommendations can support current best practices for river management and environmentally sustainable restoration, preventing riverbank destabilization.