Includes bibliographic references. ; The finite element model of an arch dam was calibrated for dynamic behaviour using the measured natural frequencies and mode shapes as benchmarks. The properties were extracted from the structure using ambient vibration testing techniques. Besides the geometry and general material properties of the dam wall concrete and foundation rock, the measured frequencies and mode shapes depend on the conditions at the dam site namely water level, temperature and the interactions between the several components of the dam. This study, however investigated the effects of the water level and to some extent, the effect of dissimilar foundation abutment material properties on the natural frequencies of the dam. A dam is continuously in harmonic motion due to some environmental factors such as wind. Either due to this movement of the dam itself or the internal movement of the reservoir water, a dynamic interaction occurs between the water and the dam wall where the movement of the one medium affects the other. A study conducted in the early twentieth century deduced that it is only part of the reservoir water that can be assumed to interact with the dam. It is from the same study that the Westergaard added mass concept was born which says that the interacting water mass can simply be added to the dam wall mass, a procedure from which the extraction of the dynamic properties can ensue as normal. This added mass formulation was derived on a basis of some assumptions which include a rigid and vertical dam wall and the incompressible water body. The added mass concept was extended to account for flexibility and curvature of the upstream dam wall in more recent studies. The extended version of the Westergaard method is normally referred to as the generalized Westergaard method. The original Westergaard added mass formulation was used to account for the dam wall- water interaction in the double curved Roode Elsberg dam model. This proved to be problematic as this dam is highly asymmetrical and has diverging reservoir walls, the characteristics of which are not catered for in the original Westergaard added mass method. The combined effect of using the original Westergaard method and these deficiencies in the formulation resulted in the model's natural frequencies being lower than the field ones, for the same ambient conditions. On the basis of literature, a factor of 0.8 on the added masses was applied on all the original Westergaard added masses to account for the effect of the diverging reservoir walls. The remaining masses were then reduced until a good correlation of the field frequencies and model frequencies was achieved. This was done to account for the effects of the flexibility of the dam and the curved upstream dam wall. All in all, a factor of 0.25 on the masses calculated using the original Westergaard added masses was applied to account for all the above-mentioned effects. This factor compares favourably with literature even though in literature it is rarely mentioned what effects are being accounted for when this factor is introduced. This work hence raises awareness about the shortcomings of the Westergaard method when used for model calibration and how those shortcomings can be accounted for. In summary, these shortcomings are brought about by assuming a prismatic and infinite reservoir, while in reality this is not always the case. It appears that these shortcomings affect the results of the added mass approach when used as a tool to represent the dam-water dynamic interaction in arch dams.