How Fill Material Affects the Overtopping Process for Earthfill Dams


This article has been evaluated and edited in accordance with reviews conducted by two or more professionals who have relevant expertise. These peer reviewers judge manuscripts for technical accuracy, usefulness, and overall importance within the hydroelectric industry.

By Johnson Malisa, F.W. Mtalo and Leif Lia

For many small earthfill dams, significant overtopping potential exists if inflow exceeds the reservoir storage and spillway outflow discharge capacities. In the Arusha region of Tanzania, which contains about 150 small earthfill dams, several cases of failure have been reported.

This overtopping risk can never be completely eliminated. Thus, the challenge is to determine how embankments will perform before an overtopping event, primarily by determining the influence of soil materials on the process and rate of erosion during overtopping and breaching.

Erosion begins once the forces exerted by the overflowing water exceed the resistive forces of the exposed material. The extent of damage caused is a function of the embankment materials and overtopping depth and duration.

Discussions of dam overtopping must distinguish between the erosion characteristics of cohesive and noncohesive soil.1 During overtopping of a dam made of cohesive material, the erosion phenomena result in the formation of an overfall.1,2 The overfall migrates upstream and advances through the embankment crest, breaching the dam. For noncohesive soil, the tractive stress analysis method may be appropriate, in which erosion occurs on a uniform but gradually reducing gradient or as a gradual upstream migration on a constant gradient.

Physical model tests indicate that the headcut advances progressively headward as its base deepens and widens. Breaching of the dam occurs when the headcut migrates through the upstream crest, which is called “time of breach initiation.” The point at which erosion reaches the toe of the upstream slope is called “time of breach formation.”

The upstream headcut advance is influenced by:

– Insufficient soil strength to stand vertically due to height of the headcut face;

– Stress relief cracking and induced hydrostatic pressure in the stress crack; and,

– Loss of foundation support for the vertical face due to the waterfall flow plunging effect and its associated lateral and vertical scour.

This type of erosion is a three-dimensional process, in which upstream migration is associated with lateral widening.1 The rate of widening is a function of the headcut migration rate, and both are important in determining the timing and amount of water discharge through the breach.3

Background of dam breach modeling

A comparison of dam breach models shows the wide variability of results.4 Other summaries and analyses of dam breach models have been published.5,6

Breach formation through embankment dams can be modeled using process-based models.7 Most models are based on steady state sediment equations related to homogeneous bank and adopted breach growth mechanisms. The modeler needs a significant number of assumptions and simplifications to simulate the breach, all of which can greatly affect the results.

In general, the breaching process is dominated by the interaction of flow hydraulics over the embankment and through the breach, the erosion process, soil properties, and geotechnical mechanisms. Simplification in any of these processes affects the results, especially geotechnical simplification concerning the breach widening process, which depends on removal of the lateral support of the breach side slope as a result of flow.

Further research into the breach process, including full-scale tests and scaled-down flume tests, is one of the top priorities for developing a better understanding of the dam break problem.5,8,9

Experimental set-up, procedures

Physical model testing was conducted in an outdoor laboratory flume at the University of Dar Es Salaam in Tanzania. The flume was 10 meters long and 1.2 meters wide, with side walls 1 meter high. Inflow of 5 to 15 L/sec was supplied using pumps. The flume walls were mainly made of bricks, with a bottom of precast concrete slabs. Part of the wall was made of plastic glass, to allow lateral observation of the model during testing.

The homogeneous embankment models were made up of clay, silt, and fine sand. Material was selected based on the type of soil found in embankment dams in Arusha. Materials used have a plasticity index (PI) of 23 to 42 and liquid limits (LL) of 44 to 88. The soils were compacted in layers using hand-held timber and steel plate compactors. The soils were compacted to a different percentage of the Proctor value, except where the effect of compaction and moisture content were investigated.

A rolling carriage mounted on the flume side walls with an attached point gauge was used to obtain longitudinal bed profile, cross-sections, and water surface elevation. Inflow was supplied from the laboratory reservoir using a 4 inch-diameter (9.7 cm) pipe. A digital flow meter was attached to the pipe to record inflow. Outflow was measured using a V-notch installed downstream of the flume and the reservoir volume-elevation relation. The V-notch had a capacity of 17 L/sec, which was not adequate for all flows. Therefore, the reservoir elevation and storage records were an essential part of evaluating the outflow. The flume bottom was equipped with manually read piezometers for continuous water level measurement.

Before testing, the reservoir was filled to a known depth, depending on the dam height. During testing, the reservoir was completely filled and the embankment was overtopped. Inflow discharge stabilized slowly and then was maintained at a relatively defined constant flow. Maximum overtopping head attained was recorded for each test. Flow conveyance notches were cut on top of each model to direct flow during initiation of overtopping.

Pre-test soil material determination

Samples were taken for testing from the soil batch before model construction (particle grain size distribution and Atterberg limits) and from the compacted layers during construction (maximum dry unit weight, and optimum moisture content). The soil types used in these tests were:

– M1, reddish brown silty clay classified as CL in the Unified Soil Classification. The average maximum dry density determined at optimum moisture content of 17 percent during the laboratory compaction test was 1.88 mg/m3.

– M2, darkish silty clay classified as CL. The average maximum dry density determined at optimum moisture content of 19.5 percent during the compaction test was 1.85 mg/m3.

– M3, brownish silty clay classified as SM. The standard compaction test on this soil shows that the soil has a maximum dry density of 1.95 mg/m3 at an optimum moisture content of 14 percent.

Results and discussion

During overtopping, flow on the downstream slope initiated rill and microrill erosion. This developed a network of rills that gradually formed a main rill(s). This observation seemed to agree with previous research.10,11 The study showed the occurrence of significant undercutting, which was not reported by other researchers, produced by turbulence and hydraulic stresses surrounding the jet impingement area immediately downstream of the developed headcut during breach initiation.

Breach morphology during the breach formation phase was of a bellmouth shape. The hourglass shape and streamlines of the flow show good agreement with literature results.12 The breach development and propagation also showed good agreement with other experiments.13

Figure 1 gives centerline profiles for an embankment model test conducted using soil M1 and an overtopping head of 5 to 10 cm, which shows typical headcut development. The erosion profile showed the development of several small headcutting steps on the downstream slope during the early stages of the test, but these coalesced into one headcut by mid-test.

The large overfall dominated the breaching process and finally migrated to the upstream crest, resulting in a total breach, lowering of the embankment elevation, and reservoir emptying. Figure 1 also shows the cross sections taken at the embankment crest station for the same test. The lateral erosion rate varied greatly over time during a breaching event, which is different from the assumption used in most breach models.

Most numerical models assume much more simplified breach opening morphologies than those observed during actual breaching events. On the other hand, it is easier to model static erosion using symmetrical breach opening geometries rather than continuous erosion dynamics, which normally is associated with the formation of complex geometries. The observation here shows the breach opening shape is complex and changing over time depending on the rate of headcut migration and widening. The final breach shape depends on stability of the side slope of the initiated breach channel, which is a function of the geotechnical properties of the dam model.

Different phases occurred in a complex manner during tests of the models made of cohesive material. These results show that the breaching process can be presented in three phases: breach initiation, breach formation/development, and reservoir depletion.

Detailed observations from the breach formation are:

– Headcutting was seen on the downstream face, contrary to the smoothing process observed in noncohesive material tests;12

– Crest overtopping initiates rill and microrill erosion on the downstream slopes of the embankments;

– This erosion changes into a network of rills that gradually develop into a main gully, which consists of multiple cascading overfalls. This main gully was observed to be similar to erosion channel for the noncohesive material;12

– The main gully migrates upstream through the downstream slope until one large headcutting channel remains;

– The headcut migrates toward the upstream crest of the embankment, and any further upstream migration of the headcut results in significant lowering of the crest elevation and increased discharge from the reservoir, leading to full breach development;

– As headcut migration continues, the erosion channel widens. The initial rapid widening and continued migration upstream are due to the turbulence and hydraulic stresses surrounding the area immediately downstream of the headcut;

– The hydraulic stresses erode the material at the base, resulting in undercutting, which eventually causes instability of the erosion channel banks, leading to discrete mass block failures. The failed masses are instantaneously transported by the breach channel flow, which leads to breach widening and further headcut migration;

– The above processes continue until a significant portion of the dam model is eroded and inflow to the model has been stopped;

– Upstream slope erosion at the entrance to the eroding channel was observed, producing a bellmouth shape. Slumping was observed on the upstream face, but the slumping rate was much lower than that observed during breach initiation; and,

– In the case of significant compaction of cohesive material, the breach widening erosion rates and final width were smaller than those observed in noncohesive material.

The effect of varying parameters on breaching

The effect of various parameters was investigated using the physical model tests. The base model had a crest width of 20 cm and upstream and downstream slopes of 1:2. Results are:

– Material grading. The M3 material, with relatively large average grain size, was more erodible than the M1 material. This accelerated the erosion process and led to a higher peak outflow and significantly wider final breach width;

– Compaction. Two compaction efforts were used during model construction, where one effort was half the number of blows of the normal effort used in other tests at the same lifting head and thickness of the layer to be compacted. The decrease in compaction effort accelerated the erosion process and led to a higher peak outflow and wider final breach width;

– Moisture content. One sample had moisture near the optimum (17.4 percent), and a second had a lower moisture content (12 percent), achieved by sun drying the soil before construction. Because the compaction efforts for these tests were basically the same, these tests show that the moisture content variation has a significant effect on the outflow hydrograph and breach top width;

– Downstream embankment slope. The upstream slope was kept constant at 1V:2H and the downstream slope was changed from 1V:2H to 1V:3H. The test with 1V:2H led to an earlier time to peak and a slightly higher peak outflow than the 1V:3H slope. Increasing the slope had some effect in delaying the breach initiation time, but the peak outflow and final breach width were similar for both slopes;

– Crest width. Crest width was increased from 0.2 meter in the base model to 0.3 meter. Increasing the crest width had almost no effect on the peak outflow and erosion rates. The increase in crest width was found to have a significant effect on time to peak at this scale; and,

– Breach location. The initial breach notch was cut once at the center and then at a distance of 10 cm from the flume wall on one side for the second test. The effect of breach location is significant at this scale. The peak outflow was lower and the breach width smaller for the breach location on the side.

Headcut formation and migration analysis

Understanding how a headcut forms and moves is important to embankment dam erosion modeling. The migration rate of the headcut is a function of the soil material properties, hydrodynamic forces, and embankment geometry. For the headcut migration analysis, test material was placed in layers to the full depth and width of the flume. Factors that were varied included overfall height of the fill and moisture content of the material used. The tests described below were used to evaluate the effect of inflow discharge and soil properties on the headcut migration rates.

The headcut migration tests were conducted in the same flume using materials M1-H, M2-H, and M3. The M1-H and M2-H materials had slightly different properties from those used in the previous embankment model tests. The inflow discharge, overfall height, and water content were all varied under similar compaction efforts used in the embankment model scaled tests. The upstream face of the fill was protected from surface erosion using a thin layer of soil-cement mixture. Headcut position was monitored with time and water surface profiles, and flow measurements were taken at irregular intervals based on significant changes in headcut position.

The tests showed that the erosion process is influenced by soil properties and hydraulic parameters. Several erosion-resisting forces of the compacted cohesive soil – such as soil strength, dry density, and overfall height – controlled the headcut erosion more than the discharge variation. In some tests, the headcut migrated with a sloping face, which is a sign of stress detachment. In some tests, the headcut migrated with a vertical face where the tension cracking and larger mass failure were observed.

Effect of density and soil moisture variation

The headcut test was conducted with varying moisture content, which directly affected the dry density and strength of the compacted cohesive soil, but with constant flow rate and overfall height. The soil materials tested were observed to have a strong relationship between the dry density and unconfined compressive strength, which increased as density increased. At the overfall, the flow over the headcut created a reverse roller that removed material from the overfall base. When significant material had been removed, mass block failure occurred and the headcut migrated upstream.

Although stream power (discharge) was not observed to be highly correlated to headcut advance, the unconfined compressive strength and dry unit weight were strongly correlated with the headcut migration rate (see Figure 2).

The results showed that the compaction water content has a significant effect on headcut advancement.

Effect of overfall height and flow rate

The effect of overfall height and flow rate were studied using a constant moisture content and controlled compaction for the two soil types. The tests were performed using overfalls with average heights of 0.3 meter, 0.45 meter, and 0.6 meter. The discharges were also varied in each overfall. It was observed that the headcut advance rate was slightly influenced by the discharge and overfall height variations (see Table 1). Both sloping and vertical erosion mechanisms took place on the headcuts. The erosion with sloping surface was mostly observed with the lower moisture content tests, while the high moisture content experienced erosion along a nearly vertical face. Undercutting, tension cracking, and mass failure of relatively large blocks of soil were also observed.

Effect of soil type

Headcut migration rates were significantly different for each of the soil types (see Figure 3). The rate for soil M3 was rapid, about 0.12 meter/min, as compared with the other soil types. The reason behind this is the low clay content of M3 when compared with that of M2 and M1. The rates of headcut migration in all tests were observed to increase significantly during the breach formation phase.

Effect of compaction

The effect of compaction on headcut migration was studied using three compaction efforts for material M1-H:

– Low: Wood plate compactor lifted 0.2 meter high (achieving 50 — 60% compaction);

– Medium: Steel plate compactor lifted 0.2 meter high (achieving 61 — 70% compaction); and,

– High: Steel plate compactor lifted more than 0.3 meter high (achieving 71 — 90% compaction).

Figure 4 shows that an increase in compaction at a natural moisture content significantly reduces headcut migration.

Headcut migration prediction

Determining the rate of headcut migration is one of the keys to predicting a cohesive embankment failure during overtopping. Flume tests allowed the headcut advance to be observed in a setting where discharge, overall height, compaction energy, and compaction water content could be controlled. The simple relationship for headcut migration prediction uses the energy concept at the overfall as the deriving mechanism.14,15 There is a simple model describing headcut migration based on material-dependent coefficient C and hydraulic attack parameter A:14

During breach initiation (at left), the undercutting mechanism resulted in mass slumping, causing rapid breach widening. During embankment overtopping (at right), an hourglass shape was observed.

Equation 1



– dX/dt is headcut migration; and,

– A equals (qH)1/3, where q is specific discharge of overflow in m3/sec/meter and H is headcut overfall height in meters.

There is no defined approach for determining coefficient C other than that based on observed migration rates, discharge, and headcut overfall height.14

The flume headcut migration test results were used to develop a relationship for coefficient C. The relationship was then compared with the results obtained during the embankment model tests. The flume measurement data are shown in Table 1 for the headcut advance tests and Table 2 for the embankment overtopping tests.

The rate of headcut migration was reduced significantly for the model compacted at moisture content close to optimum. This shows that compaction moisture content has a significant effect on the erosion properties of the material used.

The results from the headcut advance tests where compaction was similar to the effort used during embankment model tests indicated that moisture content has a significant effect on headcut migration (see Table 1). By changing the moisture content from 15.4 percent to 19.5 percent, the associated change in the headcut migration was more than 50 times.

There is a reasonable correlation between coefficient C and compaction moisture content. The relationship in Equation 2 was developed for coefficient C based on headcut advance test results.

Equation 2



– mc is the moisture content.

Results from embankment overtopping tests also show that the compaction moisture content plays a significant role in influencing the rate of headcut migration.

Values of coefficient C predicted from Equation 2 compare well with values of C computed from the embankment overtopping test results. These results indicate that a defined relationship between water content and coefficient C for a given compaction effort can be employed in the predictions and is independent of the embankment material texture for the range of soil types tested.14

The results in this study can be considered a step to developing a universal relationship for coefficient C for the most important parameters observed to significantly influence headcut migration rate (compaction water content and compaction effort). Increased water content (up to an optimal value) and increased compaction effort increase resistance of a soil to headcut migration.

Conclusions and recommendations

The equation developed in this study relating moisture content to coefficient C appears to be independent of soil texture for a specific compaction level. The headcut migration studies show promising effort in developing the universal relation for C for the soil parameters observed to have significant influence on headcut migration rate.14 Coefficient C will play a significant role in assessing the breaching potential for existing and proposed embankment dams. The results of the headcut migration test can be used in developing an emergency action plan in case of dam failure upstream of well-developed areas in the Arusha region.

Further large-scale tests on headcut migration need to be done to develop a universal relationship for C to the compaction water content and effort. This prediction will also be used to evaluate breaching potential of existing and proposed earthfill dams. The improvement of breach prediction models should take into consideration the integration of updated breach, hydraulic, and material parameter effects observed during embankment overtopping tests. This kind of integration will give more realistic breach hydrographs, which are needed for more accurate predictions of flood routing downstream of the dam.


1 Ralston, D.C., “Mechanics of Embankment Erosion during Overflow,” Proceedings of the National Conference on Hydraulic Engineering, American Society of Civil Engineers, Reston, Virginia, USA, 1987.

2 Al-Qaser, G., and J.F. Ruff, “Progressive Failure of An Overtopped Embankment,” Proceedings of the National Conference of Hydraulic Engineering, American Society of Civil Engineers, Reston, Virginia, USA, 1993, pages 1957-1962.

3 Hanson, G.J., K.R. Cook, W. Hahn, and S.L. Britton, “Observed Erosion Processes during Embankment Overtopping Tests,” American Society of Agricultural Engineers, St. Joseph, Michigan, USA, 2003.

4 Morris, M.W., Concerted Action on Dambreak Modeling Final Report, 2000,

5 Wahl, T.L., “Prediction of Embankment Dam Breach Parameters: A Literature Review and Needs Assessment,” Water Resources Research Laboratory, U.S. Bureau of Reclamation, Denver, Colorado, USA, 1998.

6 Mohamed, M.A.A., P.G. Samuels, G.S. Ghataora, and M.W. Morris, “A New Methodology to Model the Breaching of Non-cohesive Homogeneous Embankments, Proceedings of the 4th CADAM Workshop, Zaragoza, Spain, 1999.

7 Graham, W., “Dam Failure Inundation Maps – Are they Accurate?” Proceedings of the 2nd CADAM Workshop, Munich, Germany, 1998.

8 Fread, D.L., “Dam Breach Modeling and Flood Routing: A Perspective on Present Capabilities and Future Directions,” Proceedings of International Dam Breaching Processes Workshop, USDA-ARS, Hydraulic Engineering Research Unit, Stillwater, Oklahoma, USA, 1998.

9 Fread, D.L., “BREACH: An Erosion Model for Earthen Dam Failures,” National Weather Service Report, National Oceanic and Atmospheric Administration, Hydrology Office, Silver Spring, Maryland, USA, 1991.

10 Leopold, L.B, M.G. Wolman, and J.P. Miller, Fluvial Processes in Geomorphology, W.H. Freeman and Co., San Francisco, California, USA, 1964.

11 Dietrich, W.E. and T. Dunne, “The Channel Head,” in Channel Network Hydrology, Wiley, Chichester, UK, 1993.

12 Coleman, S.E., and D.P. Andrews, Overtopping Breaching of Noncohesive Homogeneous Embankments, Department of Civil and Resources Engineering, University of Auckland, Auckland, New Zealand, 2000.

13 Hahn, W., G.J. Hanson, and K.R. Cook, “Breach Morphology Observations of Embankment Overtopping Tests,” Proceedings of the 2000 Joint Conference on Water Resources Engineering and Water Resources Planning & Management, American Society of Civil Engineers, Reston, Virginia, USA.

14 Temple, D.M., “Estimating Flood Damage to Vegetated Deep Soil Spillways,” Journal of Applied Engineering in Agriculture, Volume 8, No.2, 1992, pages 237-242.

15 Temple, D.M., and J.S. Moore, “Headcut Advance Prediction for Earth Spillways,” Transactions of the American Society of Agricultural Engineers, Volume 40, No. 2, 1997, pages 557-562.


Malisa, Johnson, and F. Mtalo, “Dam Failure Incidents-Tanzania Experience,” International Seminar on Stability and Breaching of Embankment Dams, October 2004.

Johnson Malisa is lecturer and F.W. Mtalo is a professor with the University of Dar Es Salaam in Tanzania. Malisa was the main researcher for this study and Mtalo was coordinator of the research project. Leif Lia, a professor with the Norwegian University of Science and Technology, was supervisor of the research project.

More HRW Current Issue Articles
More HRW Archives Issue Articles


Previous articleU.S. Judge Redden to step down from long-running Columbia salmon case
Next articleDraft EIS reviews Yakima water plan including new dams, civil work

No posts to display