October 2001


Following modelling work carried out in the late 1980's both in South Africa and Australia, it became evident that the ability to accurately predict the size of scour holes forming downstream of dam overflow spillways and flip buckets was lacking. In 1991 the CSIR carried out a literature survey on this topic and came to the conclusion that work needed to be done on the physical modelling aspects of plunge pool scour reproduction. The CSIR then approached the Water Research Commission (WRC) for guidance and funding. A contract was signed with the WRC and work on Phase 1 of the project commenced in 1992. The aim of the study was to find the most appropriate model material for plunge pool scour reproduction.

Phase 1 of the project was successfully completed in September 1995. The report on the results CSIR Report EMAS-C 95098, was compiled by Mr K R K Blake. After the resignation of Mr K R K Blake, the project leadership was taken over by Ms E Dunkley.

Phase 2 of the project evaluated the direction set in Phase 1. A model material consisting out of cement based binder with coke nuts (degassed anthracite) as the aggregate was recommended in Phase 1 as best material, although the results were not entirely satisfactory. The temporary unavailability of the degassed anthracite resulted in a series of tests performed with a model material consisting out of cement based binder and sand. Further tests with degassed anthracite were performed once the material was available. Phase 2 of the project was concluded in October 1998. CSIR Report ENV/S-C 98096 describes the tests conducted and the results obtained in this second phase of the study.

Phase 1 and Phase 2 focused on cemented, lightweight materials, such as coke nuts. However the repeatability of the tests when using such material was not very good. The binding strength of the cement and lime, once mixed with water, could not be accurately controlled.

The method developed during the Maguga dam physical model study to reproduce the physical characteristics of a plunge pool scour hole was subsequently proposed by the steering committee. This was applied in the Kariba dam model as Phase 3 of the project, and was concluded in March 2000.

Phase 3 of the project concluded the current investigations into the most appropriate material to be used to reproduce the plunge pool scour below dam outlet structures in scaled physical hydraulic models. The work was carried out under the project leadership of Mr W J van Aswegen. CSIR Report ENV-S-C 2000-038 described the tests conducted and the results obtained for this third phase of the study.

The executive summaries of Phase 1, Phase 2 and Phase 3 are included in this combined executive summary report.



Free trajectory jets are often used as a means to pass excess flood waters from a dam to the river below. This inevitably results in the formation of a scour hole downstream of the dam wall. If this scour hole is extensive, as is often the case, it may endanger the dam's ancillary structures and even the stability of the dam wall itself. It is, therefore, imperative that the ultimate size of the scour hole is known at the design stage of a dam.

The mechanism of scour, however, is an extremely complex process due to the influence of various hydraulic, hydrological and geological factors. Some of these factors are difficult to assess theoretically and limited success has been had when trying to reproduce them experimentally. It is not surprising, therefore, that a reliable method of predicting the size and depth of scour holes is still not available. Although some empirical formulae have been proposed to predict such scour holes, in most cases, scale physical models are constructed to try to reproduce what will ultimately happen in the prototype. This research project set out to investigate the most appropriate materials to be used to model the bed rock below a dam wall, for use in such physical models. At the same time it also looked at the reliability of the available empirical formulae proposed to date.

Nineteen scour prediction formulae were used to predict the scour depths at various known sites and the results compared with the actual scour data. Analysis of these results showed that while certain formulae worked with some degree of success at certain given sites, problems always arose when using the formulae at other sites. It would appear that the proposed formulae do not take into account all the factors influencing the formation of the hole, or that the data used to derive the formulae did not represent the ultimate scour depth. In hindsight, it would seem that a reasonably accurate generic scour prediction formula is not possible. The engineer is thus forced to turn to scale modelling to obtain a more accurate prediction of the depth of the scour hole.

To test the proposed model bed materials, an existing dam had to be modelled in the laboratory, such that its scour hole could be used as a criterion for the tests. However, due to the lack of appreciable depth of scour below suitable dams within South Africa, no sites could be found for this purpose. Therefore, the search had to be extended further north and the Kariba Dam on the Zambezi River was chosen. The scour hole below this dam has extended almost 75 m below the river bed and annual surveys have been kept of the progress of the hole since the dam's construction.

During the extensive model testing phase of this study, a number of aspects were highlighted concerning the accurate reproduction of scour holes. The most important of these was the sensitivity of the formation of the scour hole to the flow patterns in, and around, the eroding hole. The final scour depth was found to vary greatly if the local topography was not exactly reproduced in the model, or if the material being eroded from the hole was not ejected as it would be in reality. Coupled to the latter was the effect of the confinement of the developing hole on the plunging jet. This confinement tended to enhance the erosive power of the jet by concentrating its energy onto a smaller area, resulting in a much deeper scour hole. Thus, the use of cohesive materials to represent the bed rock in the model, which could sustain the steep sides expected in reality, gave deeper, more realistic scour holes. Testing with non-cohesive materials lead to shallower scour holes that, if taken as correct, may result in an under designed system.

This study investigated a wide range of possible materials to represent the bed material in scale models. The criterion against which the materials were tested was their ability to reproduce the scour hole formed at the Kariba Dam. It was found that the loose material collecting in the scour hole adversely affected the ultimate depth of the hole. In reality this material would erode into pieces small enough to be ejected from the hole and transported downstream. To resolve the problem, a very light aggregate (coke nuts) was used, that could easily be ejected from the scour hole once it was freed from the binding matrix. Thus, the scour hole maintained a shape similar to that expected in prototype, thereby ensuring that the flow patterns in and around the hole were correct.

None of the materials tested gave ideal results, however, the one material that did behave remarkably better than all the others consisted of cement based binder with coke nuts (degassed anthracite) as the aggregate, in the ratio (by volume);
1 0.04 7.2 0.26

Due to financial limitations, further investigations into this material could not be carried out. It was, therefore, recommended that this be taken up at a later stage to further develop the ability to accurately predict the depth of scour holes below dams.



The information gathered in this report is a continuation of the work carried out by Mr K R K Blake (CSIR, Environmentek) concerning physical modelling aspects of plunge pool scour reproduction. The aim was to find a mix of material to reproduce the depth and shape of the scour hole below the Kariba dam.

The approach was to pursue the direction set in Phase 1 by Mr Blake and to carry out further investigations into this material by performing additional tests. However, the coke nuts became temporarily unavailable. It was thus decided to test another mix consisting of a cement-based binder with sand. Four tests were carried out, using different proportions of cement: sand: water. Tests with coke nuts, in the ratio recommended in Phase 1, were also performed once the material became available.

Only the mix tested in Test 4, the proportion of which is :
1 0.021 0.124

seemed to give a good enough reproduction of the shape of the scour hole formed below the Kariba dam. The depth of scour was however still insufficient, indicating that weaker sand: cement mix should be tested.

Once coke nuts became available the recommendations in Phase 1 to continue these tests was carried out. The results, however, proved to be different from those obtained by Mr Blake, due to the inconsistency of the strength of the cement used as the binder element.



This report is a continuation of research by the CSIR, which started in 1992. Upon recommendation of the Steering Committee for this project, the WCR extended the budget in May 1999 for additional model tests, based on a method developed during a physical model study of the Maguga Dam, carried out at the University of Stellenbosch.

The method developed during the Maguga Dam physical model study, to reproduce the physical characteristics of a plunge pool scour hole, was applied in the Kariba Dam model. This method involved the use of loose aggregate to model the scour hole. A total of seven tests were conducted in the model. Besides the scour prediction, the tests had as additional objectives:

Test results show an underestimate of between 15% and 23% of the maximum Kariba scour depth. The natural angle of repose for all materials tested was the main characteristic influencing the model test results. If all other material characteristics could stay the same and the angle of repose would change to a value closer to that found in prototype, the correct scour depth and profile might be achieved.

The loose aggregate method of scour prediction has produced the best results to date in the Kariba model. This method, if applied with the necessary expertise and knowledge of how a physical hydraulic model reacts during the scour process, can produce valuable results.


Future research should focus on the development of the binder element used in the model material. The binder element must be able to produce consistent and controllable strengths over a short period of time. A binder element other than normal cement and lime must be considered.

A model material with a gel or resin binder element can be developed in collaboration with the University of Stellenbosch as a joint research project. This is an ideal opportunity as a master's project.

The most current Kariba dam scour information and geotechnical information must be obtained if research is to continue. The Council for Geoscience is currently conducting a comprehensive survey of the scour hole below the Kariba dam wall. The council's study will contain valuable information linking geotechnical aspects and the current scour hole dimensions.