EXPERIMENTAL TESTS ON ROCKFILL DAM BREACHING PROCESS M. J. FRANCA, MSc. Student at Technical University of Lisbon (IST) A. B. ALMEIDA, Professor at Technical University of Lisbon (IST) Instituto Superior Técnico Civil Department Av. Rovisco Pais 1049 001 Lisboa Portugal Phone: ++ 351 21 841 81 50 / 8 Fax: ++ 351 21 841 81 50 e-mail: aba@civil.ist.utl.pt Abstract The outflow hydrograph resulting from a dam failure represents the upstream boundary condition of any dam break flood model for valley risk management. Previous studies characterizing a dam breach outflow are mainly concerned with earth dams. Rockfill dams are an alternative solution to create water reservoirs but its breaking characteristics have not been subjected to research and criteria. Each new approach needs experimental work to understand phenomenological aspects to which does not, or can not, exist any analytical approach so far. In order to attenuate the lack of knowledge about the breaching mechanism on rockfill dams, an experimental facility was built at the CEHIDRO Laboratory (I. S. T.). Several tests were carried out to better understand the breaching process and aiming at the characterization of the breach final configuration. Some preliminary but important results were obtained pointing towards new research directions: (i) the discharge that induces the beginning of the breaching process; (ii) the description of the dam body behaviour during the overtopping; (iii) the characterization of the breaching process; (iv) the comparison between the bottom erosion and the lateral erosion rate; (v) the breach final shape; and (vi) the ratio between the breach parameters (final average width and final depth) and the dam height. Key words dam break; rockfill dams; breach modelling; physical experimentation 1. Introduction The probability of a dam failure that may cause a major flood in the downstream valley is very small, however its devastating and catastrophic effects generates a rising concern in the scientific and technology community, and in the organizations responsible for the security and civil protection. These organizations aim essentially to study the phenomenon in order to implement preventive and mitigating actions in the downstream valleys. A hydrodynamic dam break flood model is the main tool to be used on the downstream valleys risk management especially in what concerns risk zoning. Most of the dam safety regulations include the valley inundation map as a mandatory procedure. A dam break model uses the outflow hydrograph resulting from a dam failure as a boundary condition. In the last decades, efforts to predict the outflow hydrograph through the dam breach have been made. One can divide the several attempts in two groups: approaches using historical dam failures data and regression approximations to calculate the hydrograph based on the dam and reservoir general characteristics (usually the volume of the reservoir V R and/or the dam height h D ) Wahl (2001) makes an overall view of these methods; approaches using semi-analytical methods established from the physical laws of breach progress and of reservoir depletion see the
overview by Singh (1996). In CADAM (2000) there are pointed out several conclusions regarding the state of art about dam breaching modelling and it is clear that, the lack of knowledge in this domain is huge. An uncertainty of about 50% in the estimate of the maximum discharge is expected on the results that the existent models can offer. Rockfill dams are a possible alternative solution to create water reservoirs. In the 40 s earth-cored rockfill dams gain popularity and in the 60 s the same happened to the concretefaced ones. Singh (1996) refers the following world records for rockfill dams: Esmeralda (Chivor) dam, 1975, in Columbia (237 m high) and Chicoasen dam, 1980, in Mexico (261 m high). The rockfill Rogun dam in the Kazakhstan, when completed, will be the highest dam in the world with 335 m tall. At the moment, in Portugal, there are seven large rockfill dams, which represent about 7% of the total number of existent large dams. The previous studies characterizing the dam failure hydrographs are, so far, mostly directed to earth dams while the rockfill dam break phenomenon has not been a strong research topic. Nevertheless, some accidents happened with this kind of dams. The following authors presents several cases of rockfill dam accidents: Combelles (1979); Singh and Scarlatos (1988); Serafim and Coutinho-Rodrigues (1989); Walder and O Connor (1997); and Martins (2000). The present paper presents an experimental study that was carried out in the year of 2001 at the CEHIDRO Laboratory (Instituto Superior Técnico Lisbon Technical University), which comprised the construction of an experimental facility to support the setting up of small rockfill dam scale models on it. Several tests were carried out to better understand the breaching process due to overtopping and aiming at the characterization of the breach final configuration. Some preliminary but important results were obtained pointing towards new research directions: (i) the discharge that induces the beginning of the breaching process; (ii) the description of the dam body behaviour during the overtopping; (iii) the characterization of the breaching process; (iv) the comparison between the bottom erosion and the lateral erosion rate; (v) the breach final shape; and (vi) the ratio between the breach parameters (final average width and final depth) and the dam height. The present paper constitutes a first approach based on physical models to the rockfill dam break phenomenon and the experience gathered is certainly useful for future developments. 2. Background Singh (1996), Broich (1999), CADAM (2000) and Almeida (2001) are some of the references where complete state-of-art texts about the numerical dam breaching technology are available. In these ones it is possible to identify about 20 models developed since 1965 until the present days. Beside the numerical modulation and as a complementary activity, several authors developed important experimental work on the earthen dam breaching topic, using small scale models, that is extremely useful contribution on the understanding of the breaching process and the breach final geometry: Coleman et al (1997), Visser (1998), Bechteler and Kulisch (1998), and Loukola et al (1998). There exist several experimental studies about the overflow of rockfill structures that are very useful to establish the link between the existent dam breaching technology and the existent rockfill stability technology. These researchers focused especially on the highly economic flood release solution, which consists in built-in spillways installed on rockfill dams (Olivier, 1967; Stephenson, 1979; Martins, 1981). This solution takes advantage of the capability of the rockfill to remain stable, within a certain range of flow limits, when there is overflow and percolation flow. It is suggested to place a stable channel on the dam crest and on the dam downstream slope and it allows to cutback on the cost of a conventional flood spillway. At the moment exists several prototype of this kind of structures, namely Pit 7,
Laughing Jack, Kemi Lake. In Portugal there exist the Bastelos dam that shows some problems of fine filling in the rockfill interstices. 3. Dam Models In order to respect the similarity conditions the models followed a Froude scale and the choice of the granulometry took into account the similarity conditions related with the turbulent flow imposed inside a rockfill dam when this is under percolation and overflow (Yalin, 1971; Martins, 1984; CUR/RWS, 2000). The dam body was homogeneous and the rock granulometry had an average grain size diameter of D 50 = 18.9 mm, a minimum grain size of 9.52 mm (which corresponds to the normalized siege 3/8 ) and a typical granulometric curve representative of the ones used on real rockfill dams. The unit weight of the grains was 26.2x10 3 N/m 3. The dam models (Figure 1) were 0.5 m high, with slopes of 1.0:1.5 (vertical:horizontal) on both upstream and downstream sides, with a 0.2 m wide and 2.0 m long crest. The dam volume was about 0.9 m 3 and the reservoir had a capacity of about 2.7 m 3. The watertight element was placed over the upstream slope and it was made of a plastic layer on the periphery and of paper on the middle area in order to create a fragile part where the breach could develop. A small pilot channel of about 0.2 m wide was imposed in the middle of the crest in order to guide the breach initiation process and to avoid that the breach felt the wall effect. 4. Experimental Set-Up The experimental facility is composed of a 6.0 m long and 2.0 m wide canal with 1.0 m high walls where the dam models were constructed. The longitudinal bed slope is null. The canal is supplied by an internal pumped scheme at the upstream that allows a maximum discharge of about 100 l/s. One of the canal walls is made of transparent PVC, which allows the lateral observation. A video camera sustained by a tripod was installed to record of the experiments. The inflow discharge is measured by a flowmeter installed at the upstream pipe. In the reservoir of the dam model, the canal bottom is equipped with pressure transducers that allow the continuous measurement of the water levels on the reservoir. The outflow discharge was computed recurring to the mass conservation law and using the measurements of the inflow and water levels in the reservoir. Figure 2 illustrates the schematic of the experimental facility. 5. Results 5.1. Breaking Conditions A constant inflow was supplied to the reservoir to induce the overtopping of the dam and the initiation of the breaching process. Afterwards, the test was carried on until the dam reached a stable configuration. One of the most remarkable results taken from the tests on rockfill dam model is their capability to resist to the overflow and percolation flow when compared with earth dams. In fact, the beginning of the structural collapse of the dam due to the overtopping only happens to discharge values significantly high. With the experiments it was verified the existence of three particular discharges related with the evolution of the model failures: The discharge that induces the beginning of the downstream slope movements, these ones composed typically by minor slides occurred at the slope base - Q u = 19.4 ± 1.7 x 10-3 m 3 /s, which corresponds to a specific discharge q u = 14.9 ± 1.4 x 10-3 m 3 /s/m; The discharge that provokes a general slide of the downstream slope, from the slope base until the crest, and its irreparable ruin Q r = 27.6 ± 1.8 x 10-3 m 3 /s, which corresponds to a specific discharge q r = 19.6 ± 0.0 x 10-3 m 3 /s/m;
The discharge that induces the beginning of the breaching process and consequently of the reservoir depletion Q I 40.0 x 10-3 m 3 /s. The values of these characteristic discharges show that the occurrence of overtopping of a rockfill dam does not mean necessarily the beginning of the collapse of the whole body structure. In fact, only when the overtopped discharge reached the double of the discharge that induces the first movements in the downstream slope, the structure failure begins to have real and dangerous consequences. This is due to the emergence of a breach and the consequent reservoir depletion that induces the flood in the valley. The verification of this difference between the limit discharges (especially Q u and Q I ) allows the dam authority to define a warning time since the beginning of the overtopping, which is anyway an extreme condition to the dam and to the downstream valley, before the beginning of the breaching process. Figure 3.a shows the relation between several phases in the failure process and the crucial moments to the elaboration of the emergence plans. 5.2. Failure Mechanism The rupture mechanism of rockfill dams, according to the tests, is different from what is usually described to earth dams failures. The failure mechanism is synthesised on the Figure 3.b. The dam ruin can be divided in two distinct phases: before the occurrence of the breach, and after the occurrence of the breach. In fact, until the moment when the breach shows up, the dam downstream side experiences several damages due to the percolation and the overflow. The turbulent percolation flow induces slope movements, starting from the basis, in an upward evolution until they reach the crest. The ruin process, before the beginning of the breach, has a two-dimensional character and can be compared with typical landslides occurrences. After the formation of the breach the flow is concentrated on its section; an eroded channel is formed in the breach direction and the control section of the breach moves upwards. The breach cross-section evolution is due, not only to the continuous erosion induced by the flow, but also to the rock slides that occur from time to time when the equilibrium conditions are weak. Other important facts observed in the tests related to the failure process of rockfill dams are the following: the flow over the dam induces several damages on the downstream slope before the beginning of the breach that influences the initial breach configuration these damages are mostly two-dimensional slides along the longitudinal axis of the dam; when the overflow discharge is enough to induce the beginning of the breach (Q I ), a major and sudden slide occurs and the initial breach appears; the overflow induces an initial breach width which is approximately 1.05 times the dam height; the deposition of the rock blocks immediately downstream of the dam has a stabilizing effect, sustaining the failure process this is reflected mainly on the final breach depth which is about 80% of the dam height; the average lateral erosion rate of the breach is about 80% of the average bottom erosion rate. 5.3. Final Configuration The knowledge about the breach final configuration is very important because of the two following main reasons: elaboration of numerical models that considers an evolution (linear or not linear) since the incipient shape of the breach until its final configuration; in the so-called large reservoirs the maximum discharge through the breach is mostly a function of the breach final configuration (Walder and O Connor, 1997). With the current experiments it was possible to achieve the following results: the geometry that best fits the breach final configuration is the parabolic one; the final top width of the breach is approximately 2.25 times the dam height;
the final average width of the breach is approximately 1.70 times the dam height; the final breach depth takes approximately 0.80 times the dam height, what means that 20% of the dam height is not eroded although the continuous discharge imposed; the lateral slope of the breach banks is slightly smaller than the angle of repose; in some cases the breach walls are nearly vertical due to coarser material; the profile along the breach axis presents at the upstream a negative concavity, which corresponds to the flow control section; afterwards a positive concavity and finally, at the downstream a negative concavity, which corresponds to the rock deposition downstream the dam resulting from the its erosion; the length of the rock deposition downstream of the dam is about 1.5 times the dam height. Figure 4 shows a simplified average breach cross-section corresponding to the final stage of the tests. The results obtained agree well with range of variation of the final breach configuration ratios given by Singh (1996): W B,t / W B,b = 1.29; W B,t /h B = 3; and tan θ 1 (W B,t is the breach top width; W B,b is the breach bottom width; h B is the breach depth; θ is the breach lateral slope). 5.4. Outflow Hydrograph In the experiments it was not possible to observe what one could refer as a typical outflow hydrograph due to the dam rupture and this one was never represented by a smooth and regular curve. In fact, the hydrograph obtained from the experiments is characterised by the occurrence of several discharge peaks. These ones correspond to the occurrence of several rockslides, which induce the sudden enlargement of the breach area as a consequence of the drop on the breach bottom and of the widening of the breach width. Figure 5 shows a typical experimental hydrograph. One can remark two main differences between the expected outflow hydrograph from a rockfil dam breaking and the usual one described for earth dams: the occurrence of several peaks of minor amplitude resulting from debris slides on the breach; the unstable character of the outflow hydrograph, formed by several random peaks, which hardly can be modelled as a continuous function. The total failure time observed was typically between 450 and 1200 seconds, since the beginning of the breaching process until the breach reaches its final configuration (equilibrium). Taking as an example a real rockfill dam of about 25 m high (which corresponds to a scaling factor of 50 between the models an the prototype), the correspondent expected failure would be in the range between 54 and 144 minutes. 6. Conclusions With the current experiments it was possible to draw some important conclusions on the rockfill dam breaching process. It was observed the existence of three characteristic discharges related with the evolution of the model failures: one that induces the beginning of the downstream slope movements; other that provokes a general slide of the downstream slope and its irreparable ruin; and finally a discharge that induces the beginning of the breaching process. The knowledge of these discharge values is very useful for the authorities responsible for the risk assessment in the valleys downstream dams, in the implementation of the emergency plans. Some aspects on the failure mechanism of rockfill dams were cleared out and they were pointed out the main differences between both type of abutment dams, earth and rockfill dams: the type of damages induced by the dam crest overtopping before the dam breaching; the sudden character of the breaching initiation and the irregular evolution of the breach and
its outflow hydrograph; and the influence of the rock blocks deposited downstream on the stabilization of the failure process and on the limitation of the breach height. The observation of the tests allowed to predict empirical relations between the breach geometry and the dam height, and to estimate the relation between the lateral erosion rate of the breach and the bottom erosion rate. Two important aspects on the configuration of the outflow hydrograph through the breach, related with its irregular character were described: the occurrence of several peaks due to rockslides on the breach; the impossibility of modelling the hydrograph as a continuous function. From the tests it is possible to predict a range of values between 0.9 and 2.4 hours for the failure time of a rockfill dam with 25 m high. 7. Acknowledgements The authors wish to acknowledge FCT financial support (Praxis 3/3.1/CEG/2688/95). M. J. Franca was supported by FCT (BM/2233/2000). 8. References Bechteler W., Kulisch H. (1998), Description of Test Case Nº. 1 on Dam Erosion, Proceedings of the Munich Meeting CADAM, October. Broich K. (1999), An Overview of Breach Modeling, Proceedings of the Zaragoza Meeting CADAM, November. CADAM - Concerted Action on Dambreak Modelling (2000), Final Report, January. Coleman S. E., Jack R. C., Melville B. W. (1997), Overtopping Breaching of Noncohesive Embankment Dams, Proceedings of the 27 th Congress on the IAHR - Energy and Water: Sustainable Development, San Francisco, August. Combelles P. (1979), Internal Report of the Service de la Production Hydraulique Electricité de France, October. CUR/RWS (2000), Manual on the Use of Rock in Hydraulic Engineering, A A Balkema, Rotterdam. Loukola E., Huokuna M., Xiang W. L., Yahekou (1998), Dam Breach Test Case, Proceedings of the Munich Meeting CADAM, October. Martins R. (1981), Hydraulics of Overflow Rockfill Dams, Laboratório Nacional Engenharia Civil Me 559 (report), Lisbon. Martins R. (1984), Utilização de Enrocamentos em Estruturas Hidráulicas Síntese dos Conhecimento Actuais, Laboratório Nacional Engenharia Civil Me 635 (report), Lisbon. Martins R. (2000), Dam Safety and Protection of Human Lives, International European-Asian Workshop Ecosystem and Flood 2000, Hanoi. Olivier H. (1967), Through and Overflow Rockfill Dams New Design Techniques, Proceedings of the Institution of Civil Engineers, March. Serafim J. L., Coutinho-Rodrigues J. M. (1989), Statistics of Dams Failures: A Preliminary Report, Water Power and Dams Construction, Vol. 41, No. 4, pp. 30-34, April. Singh V. P. (1996), Dam Breaching Modeling Technology, pp. 151-219, Kluwer Academic Publishers, Dordrecht. Singh V. P., Scarlatos P. D. (1988), Analysis of Gradual Earth-Dam Failure, J. Hydraulic Engineering ASCE, Vol. 114 No. 1, pp. 21-42, January. Stephenson D. (1979), Rockfill in Hydraulic Engineering, Elsevier Scientific Publishing Company, Amsterdam. Visser P. J. (1998), Breach Growth in Sand-Dikes, Communications on Hydraulic and Geotechnical Engineering Report Nº. 98.1, Technische Universiteit Delft, Delft.
Wahl T. L. (2001), The Uncertainty of Embankment Dam Breach Parameter Predictions Based on Dam Failure Case Studies, USDA/FEMA Workshop on Issues, Resolutions and Research Needs Related to Dam Failure Analysis, Oklahoma City, June. Walder J. S., O Connor J. E. (1997), Methods for Predicting Peak Discharges of Floods Caused by Failure of Natural and Constructed Dams, Water Resources Research, Vol.33, No. 10, pp. 2337-2348, October. Yalin M. S. (1971), Theory of Hydraulic Models, Macmillan, London. 9. Figures Pilot channel 1.0 1.5 1.5 1.0 Paper 0.50 m Plastic film Figure 1 Views of one of the dam models. 3,25 m 6,0 m Control valve Impervious Reservoir layer Pressure transducer Dam model 0,50 m 1,0 m Video camera Flowmeter Rockfill Grain filter Aquisition system Pumping system Reservoir Figure 2 Experimental facility built at CEHIDRO laboratory.
Reference situation Ocurrence of a extraordinary flood on the dam basin Inflow hydrograph to the dam section (Qi(t)) Downstream side of the dam stable Flow over the dam crest Beginning of the flood propagation to the downstream valley Qi(t) > Qu(t) Reference situation Beginning of the ruin process of the downstream side and of the dam crest First movements on the dowstream side of the dam; ocurrence of minor slides Occurrence of an abno rmal flood Detection of the flood Ocurrence of several slides on the downstream side resulting from the inflow discharge increment Qi(t) > QI(t) Overtopp ing of the da m crest The structure damages reaches the upstream side of the dam Beginning of the breaching process; beginning of the reservoir depletion Beginning of the flood propagation due to the reservoir depletion First slides on the downstream slope Qi(t); QB(t) Beg inning of the breaching p rocess Breach evolution as a function of the inflow discharge and of the reservoir volume The breach progress is a result of lateral slides and superficial erosion Breaching formation Qi(t) = QB(t) Equilibrium: the flow through the breach is not enough to continue the erosion process (a) (b) Figure 3 (a) Critical failure phases for the elaboration of an emergence planning; (b) Failure mechanism on rockfill dams 2,25 hd 1,70 hd h D 0,80 hd Figure 4 Model for the final configuration of the breach based on the CEHIDRO experiments.
50 45 40 Begginning of the breach 45,5 l/s 45,3 l/s End of the test 60,00 50,00 Discharge (x 10-3 m 3 /s) 35 30 25 20 15 10 5 Outflow discharge Inflow discharge Water level 40,00 30,00 20,00 10,00 Water level (x 10-2 m) 0 0 200 400 600 800 1000 1200 Time (s) 0,00 Figure 5 A typical hydrograph obtained during the tests.