According to the International Commission on Large Dams (ICOLD) [1
], embankment dams, constructed with locally excavated rock-fill or earth-fill materials consist of 78% of the total existing dams worldwide. The dams made for over 50% from coarse-grained material are described as rockfill dams and represent 13% of the entire world’s dam population. Specifically, in Norway, more than 360 large dams (over 15 m high) are present and over half of them are rockfill dams (Figure 1
Because of the possible disastrous consequences of a dam break, dam safety is crucial. ICOLD [1
] states the overtopping phenomenon as the principal cause of embankment dam breakage (as the primary factor for 31% of the total number of ruptures, and 18% as a secondary factor). Thus, having rockfill dams equipped with defence mechanisms to protect the dam structure against unexpected overtopping or leakage events is essential from a dam safety aspect.
During the overtopping of an embankment dam, the downstream slope of the dam is subjected to highly destabilizing dynamic forces. These are produced by turbulent overflow (overtopping of the crest) and throughflow mechanisms. Here, throughflow entails flowing through the supporting fill of the dam due to overtopping of the impervious core (Figure 2
a), with the top of the core at a lower level than the dam crest. Under throughflow conditions, the high-velocity turbulent flow within the dam can induce internal erosion processes, particularly if the internal stability criteria are not fulfilled with well-graded material. Furthermore, throughflow can induce external erosion in the exit zone and destabilization of the downstream embankment slope because of the pore pressure increase. If the crest is overtopped (Figure 2
b,c), the downstream slope is then subjected to high-velocity, turbulent surface flow, provoking a progressive external erosion process that can lead to a dam breach.
Ripraps are one of the most employed defence mechanisms for several in-stream hydraulic structures such as embankment dams, spillways, streambeds, riverbanks, bridge piers, and abutments [3
]. For rockfill dam engineering, ripraps are encountered both on the upstream embankment and on the downstream slope. This enables protection against erosion from wave impacts and ice-induced forces on the upstream slope and protection against external erosion from accidental leakage or overtopping events on the downstream slope.
Two types of riprap structures can be distinguished on rockfill dams: placed riprap and dumped riprap. While dumped riprap consists of randomly placed stones, placed riprap corresponds to an interlocking arrangement of stones on the dam shoulder. Thanks to this specific building pattern, placed riprap has displayed greater resistance against overtopping phenomena [8
]. Nonetheless, placed riprap construction remains more expensive than dumped riprap from an economic perspective. Also, these protective structures can be combined with the presence of toe support. A better understanding of the riprap structure and resistance against overtopping would help improve the reinforcement and building techniques.
In Abt and Thornton [11
], the progress in research on riprap design against overtopping is detailed, and important authors and works [6
] are mentioned, proposing riprap design relationships for overtopping flow conditions for multiple stone sizes. Monteiro-Alves et al. [15
] study the failure of the downstream shoulder submitted to overflow from a large number of experimental models with modification of specific parameters (dimensions, material size. Also, it is important to reference scientific publications on placed riprap designs that include toe support [16
]. They demonstrate that the transport of individual stones in placed riprap does not automatically induce the failure of the whole structure and that placed ripraps display greater stability than dumped ones.
For some years, in the hydraulic laboratory of the Norwegian University of Science and Technology (NTNU), Trondheim, several experimental models have been set up. The main objective of those experiments was to investigate the failure mechanisms of riprap on a steep downstream slope exposed to overtopping events as well as to compare the resistance of different designs according to the overtopping discharge level. In this research article, a review of nine different designs is introduced and their different associated failure mechanisms as well as resistance against overtopping are detailed and discussed. The aim is to bring forth the impact of each of the following characteristics on the dam stability: presence or absence of toe support and throughflow as well as dumped riprap versus placed riprap structure.
3. Data Analysis
Before analysing the data for each setup, a sum-up of the critical discharge values (qc
), their standard deviation (
), the quantity of executed tests, and the failure mechanisms associated with each setup is displayed in Table 2
. The 3 failure mechanisms are displayed in Figure 9
. The standard deviations are computed according to the number of tests, critical discharge value for each test, and average critical discharge. It must be highlighted that the standard deviation values must be considered with care as the sample sizes are quite small for statistical analysis. For unsupported riprap models, the standard deviation computation cannot apply since only one model from these setups was exposed to increasing discharge values, the other ones were directly exposed to the critical discharge value obtained from the pilot test (respectively 40 and 60 L·s−1
for dumped and placed riprap model). More information on that specific point can be found in [9
3.1. Riprap Model Supported at the Toe
For riprap models supported at the toe, the comparison of critical discharges for dumped and placed riprap (Table 2
) demonstrates that the placed riprap (qc
= 245 L·s−1
) has almost seven times higher stability than the dumped one (qc
= 40 L·s−1
), even though this value must be considered carefully since only one test was carried out with dumped riprap model.
For the placed riprap model, Hiller et al. [8
] observe that the removal of a single riprap stone does not necessarily affect the structural integrity of the whole riprap layer. The loose stones were easily identified because of their trembling motion during the overflow, and those rocks could more easily be removed by the flow. Nevertheless, it must be highlighted that some stones stabilized after some time, thanks to the compaction of the riprap, and did not get dragged downstream. Considering 1D displacement in the x-direction, the failure of the model occurred at the transition between the crest and the downstream slope, where a gap was forming. This gap formation is the direct consequence of the compaction of the riprap layer on the chute in the flow direction, increasing along with the discharge values. The displacements of the riprap stones were provoked by flow-induced vibrations. The deformation of the riprap layer could be observed in both x and z directions (Figure 9
a) as explained in [24
]. It is then interesting to point out that the mechanism of progressive 2D deformation of placed riprap stones supported at the toe can be compared to the mechanism of buckling observed in a slender-long column, pinned at one end and free at the other [24
]. The explanation lies in the interlocking forces generated between the riprap stones that create a bearing structure able to resist certain levels of deformations.
For the dumped riprap model, what can be described as a surface erosion process was observed (Figure 9
c). As for the placed riprap model, some individual stones were eroded by the action of destabilizing turbulent flow forces. However, what was described as a bearing structure for placed riprap model does not exist in such a setup. Thus, dumped riprap failure occurs much sooner, being the result of progressive unravelling external erosion.
3.2. Riprap Model Unsupported at the Toe
For riprap models unsupported at the toe, the comparison of critical discharges for dumped and placed riprap (Table 2
) demonstrates that the placed riprap (qc
= 60 L·s−1
) has 1.5 times higher stability than the dumped one (qc
= 40 L·s−1
). However, as explained in [9
], only one experimental test was carried out for placed and for dumped models with increasing discharges (qi
= 20 L·s−1
= 20 L·s−1
) so no standard deviations could be estimated. The other experimental models were directly exposed to the critical discharge value since the smartstone batteries’ life used in these experiments was limited [9
For the placed riprap model, minor displacements along the x and z axes were recorded prior to failure initiation. After the first overtopping event, the hydraulic drag and lift forces rearrange the individual stones, leading to a compaction of the riprap layer. This more compacted protection layer gained stability and formed a unified structure. So as the discharge increases, the destabilizing forces increase and some of these are partially transferred to the filter layer below as frictional forces and towards the unsupported riprap toe lying on the geotextile membrane. The sliding of the whole riprap layer occurs when the hydrodynamic forces exceed the limiting values of the static frictional forces between the toe stones and the horizontal platform [9
For dumped riprap model, the failure mechanism is comparable to the one for the supported dumped riprap. The individual riprap stones undergo progressive erosion by the flow forces. They first resist through the self-weight of the stones and from the frictional forces between individual rocks and at the riprap-filter interface. Failure initiates when the hydrodynamic forces exceed the resultant of these self-weight and frictional forces.
3.3. Half Dam Model with Unsupported Riprap
For the half dam model with riprap unsupported at the toe, the placed riprap (qc = 17.5 L·s−1) setup has been demonstrated to be almost two times as resistant as the dumped riprap setup (qc = 30 L·s−1).
Concerning the failure mechanisms, the observations that can be made for these tests are comparable to the descriptions of the ones for the riprap model unsupported at the toe. The placed riprap setup undergoes a sliding failure with initial compaction of the riprap layer during the first overtopping stage. It can be pointed out that no important movements of the riprap stones nor of the toe stones were observed because of the throughflow. Also, the sliding mechanism started at the toe section and the rockfill shoulder of the model was stable, suggesting that the critical component is the riprap protection layer for failure initiation. The failure of dumped riprap model, as for the previous setups, could be attributed to the progressive erosion of individual riprap stones.
3.4. Half Dam Model with Riprap Supported at the Toe
The half dam models with placed riprap supported at the toe were found to fail at an average discharge value of qc = 100 L·s−1. At first sight, the mechanism is comparable to the one described for the riprap model supported at the toe, even though the failure occurs sooner. Indeed, a gap was forming on the top of the downstream slope, at the transition with the crest and this is where the initiation of the failure could be observed. The displacement of the riprap layer in both x and z directions also suggests a buckling deformation similar to that observed on the supported toe riprap model, this will still have to be confirmed in a future research article.
3.5. Full Dam Model with Unsupported Riprap
The full dam models with placed riprap unsupported at the toe were found to fail at an average discharge value of qc
= 48 L·s−1
, one additional test was performed without geotextile at the base, resulting in a failure discharge of 25 l·s−1
(not included in Table 1
and Table 2
). The full dam model with dumped riprap unsupported at the toe was found to fail at an average discharge value of qc
= 20 L·s−1
. The failure mechanism observed is quite similar to the half dam tests. There are some deviations in failure discharge both for the dumped and placed riprap models, these deviations are most likely primarily attributable to differences in the construction of the riprap layer.
4.1. Role of the Toe Support
Looking at the failure mechanisms, the toe support has no impact on the surface erosion mechanism of individual stones for dumped riprap models. However, for the placed riprap models, the sliding phenomenon described for unsupported riprap is accompanied by a 2D buckling process for the supported ones. This was observed for both models built on a ramp and on a half dam. This buckling process was first described by Ravindra et al. [24
This process is not observed within unsupported models. In such cases, some hydraulic forces are directed towards the riprap toe. Then, the static frictional forces between the toe stones and the geotextile (on the horizontal platform, Figure 4
) increase to counter the increasing hydrodynamic forces transferred towards the toe. When the hydrodynamic forces exceed these static frictional forces, a displacement of the toe stones occurs and the complete riprap layer undergoes a progressive slide on the underlying filter layer. The importance of toe stone friction is also demonstrated well by the single test performed without the geotextile in place, resulting in a much lower failure discharge due to the lower sliding friction between rocks and the smooth metal platform.
It is interesting to point out that the experimental models equipped with supported placed riprap are much more resistant than the unsupported placed riprap ones (Table 2
). Indeed, the average critical discharge value for supported models on a ramp (qc
= 275 L·s−1
) are almost 5 times more resistant than the unsupported models on a ramp (qc
= 60 L·s−1
). Also, the half dam models with supported placed riprap are 3 to 4 times more resistant (qc
= 100 L·s−1
) than the half dam models with unsupported riprap layer (qc
= 30 L·s−1
). It is noteworthy that the gain in resistance is huge with the simple addition of support at the toe of the structure. Nonetheless, toe support brings no benefits to dumped riprap models where the protective layer does not act as a bearing structure. The supported and unsupported dumped riprap models both failed at qc
= 40 L·s−1
4.2. Difference between Placed and Dumped Riprap for Dam Stability
The way of building a riprap layer has a significant impact on the whole structure’s resistance. Both the failure mechanisms and the critical discharge when the failure occurs are quite different according to the type of riprap in place. In fact, when the protective layer consists of interlocking placed riprap stones, the observed associated failure mechanism is always a sliding process (associated with a buckling deformation for the supported riprap layer). On the other hand, dumped riprap models were always associated with surface erosion processes of individual stones and smaller slides of multiple stones that lead to failure.
Previous studies on placed riprap had already concluded that the dislodgement of individual riprap stones does not necessarily imply the failure of the whole layer [17
]. This compilation of experimental test results also moves in that direction, highlighting that the placed riprap layer acts as a unified structure thanks to the interlocking forces and the remaining stones still offer an important resistance against turbulent flow forces.
However, these structural differences not only have an impact on the failure mechanism but also on the resistance to overtopping discharges. While the supported placed riprap models (qc
= 275 L·s−1
) are almost 7 times more resistant than the supported dumped riprap ones (qc
= 40 L·s−1
), in particular, because of the buckling process described earlier, all the other unsupported placed riprap models (built on a ramp, on the half dam, and on the full dam) demonstrated a resistance 1.5 to almost 2.5 times superior to dumped riprap ones (Table 2
4.3. Impact of Throughflow
Even though the failure mechanisms for all models appear to initiate with surface erosion (Figure 9
c) or sliding processes (Figure 9
b) of the protective riprap layer, the throughflow within the dam shoulder has an important impact on the structures’ stability, as previously demonstrated by [28
]. When comparing the average critical discharges for models without throughflow (built on a ramp) and models with throughflow (built on half dam shoulder), structures with throughflow are 2 to 2.75 times less resistant than the ones with only overflow (Table 2
). Full dam models also demonstrated a smaller resistance compared to riprap models built on a ramp, confirming the role of throughflow. The flow inside the shoulder increases the pore pressures, cumulating with the drag and lift forces from the overflow, and the destabilization of the riprap layers is enhanced.
Dumped riprap models showed a non-significant difference between half and full dam models. However, from our experiments, full dam models with unsupported placed riprap (qc = 48 L·s−1) demonstrated a resistance 1.5 times greater than what could be observed from half dam models (qc = 30 L·s−1). Such a difference could lie in a difference in the construction process from the builders.
4.4. Recommendations and Limitations
First, the repeatability of the experimental results and critical discharge values can be discussed. From Table 2
, it can be observed that some results could not be associated with standard deviations on critical discharge values, either because of a lack of tests or modification of the overtopping procedure. Even the standard deviations that could be computed must be considered carefully since the number of tests remains quite limited for statistical analysis. The variability of qc
for similar models shows that even though the building procedure is the same on paper, the perfect repeatability of each test cannot be granted. According to the experience of the builder, the exact arrangement of the individual stones cannot be repeated the same way, either for placed or for dumped riprap. It is more likely that if more stones are loose in a model, they can collapse at lower discharge levels and leave part of the structure unprotected. Also, some of the full dam profile tests were done with a pilot channel in place to help provoke a breach along one wall.
The reader should also be aware that the validity of these results is confined to models with these specific material physical parameters, dimensions, and great slope value (S = 0.67) and that different outcomes could be obtained from different materials, different riprap stone shapes, and milder slopes. Especially, results that one would obtain with rounded shape stones would certainly be quite different since the interlocking pattern would not be effective with such shapes. Note also that these tests were all carried out under conditions of a fixed foundation, a dumped riprap placed on soil rather than bedrock would still need a toe structure to prevent undercutting of the toe stones, either by way of a buried toe or a horizontally extended toe structure as is common in river ripraps and rock weirs. The reader eager to learn more could also be interested in reading the recent research article [15
], presenting the results from an important quantity of failure tests due to overtopping or throughflow. In further studies, it would be pertinent to focus on some specific scale effects such as the viscous scale effect, friction scale effect, and aeration scale effect that were not taken into consideration in these research works.
Globally, even if placed riprap suggests a better resistance against overtopping events, the sliding failure mechanism described seems to occur much more abruptly than the more progressive erosion surface observed with dumped riprap models. This specific point should be considered when coming to real dams and to the concerned issues located downstream of the hydraulic work.
Finally, it is worth noting that many scientific research articles were issued from all these experimental tests and that more are expected to come, considering the use of pore pressure and load data as well as structure from motion and particle image velocimetry techniques. Especially, the buckling process in the half dam supported placed riprap model will have to be studied carefully and compared to the one described in the supported placed riprap model without throughflow.
This article displayed nine different experimental models carried out within the last few years at the Department of Civil and Environmental Engineering at NTNU. These nine models implied variation of specific characteristics: the presence of dumped or placed riprap, presence or absence of toe support, and presence or absence of downstream and upstream rockfill shoulder. These models were all submitted to overtopping events with increasing water discharges until the complete failure of the structure.
Even though overtopping should always be avoided at all costs, this research has described the respective resistances against overtopping of each model as well as their associated failure mechanisms. The results of the tests highlight the importance of placed riprap protective layer in the dam resistance against overtopping processes as well as the use of toe support for placed riprap models. Also, the results underline the importance of studying the riprap resistance when built above a dam shoulder, to take into consideration the throughflow mechanisms that induce an increase in the pore pressure and destabilizing flow forces. In fact, structures with throughflow were 2 to 2.75 times less resistant than the ones with only overflow (without rockfill shoulder). This research also shows that placed riprap undergoes an abrupt sliding failure mechanism, with a buckling phenomenon when supported at the toe, while dumped riprap goes through a process of smaller slides and surface erosion.
The results from each test as well as the associated scientific discussion, when corroborated by additional data to come on complementary tests or from the bibliography, could be precious to enhance the comprehension of riprap stability on rockfill dams but also to provide recommendations for dam design and reinforcement methods for existing dams.