Accuracy aspects in flood propagation studies due to earthfill dam failures

Abstract

Flooding due to dam failing is one of the catastrophic disasters which might cause significant damages in the inundated area downstream of the dam. In particular, there is a need of trustworthy numerical techniques for achieving accurate computations, extended to wide areas, obtained flood mapping and, consequently, at the implementation of defensive measures. In general several key aspects are required for accurate simulations of flood phenomena which are ranging from the choice of the mathematical model and numerical schemes to be used in the flow propagation to the characterization of the topography, the roughness and all the structures which might interact with the flow patterns Regarding general framework discussed before this thesis is devoted to discuss two aspects related to accuracy issues in dam breach studies. In the first part a suitable analytical relation for the description of reservoir have been discussed and the second part the influence exerted by the methods used for computing the dam breach hydrograph on the simulated maximum water levels throughout the valley downstream of a dam, has been investigated. As regards the first aspect, the influence of reservoir morphology on the peak discharge and on the shape of outflow hydrograph have been investigated in the literature. The calculation of the discharge released through the breach requires the knowledge of the water level in the reservoir. It is considerable that the reservoir morphology in computational analyses cannot be expressed exactly by an analytical formula because of natural topography of the reservoir. For this reason, the information about reservoir morphology is usually published as a detail tables or plots which each value of elevation from bottom to top has a corresponding value for lake surface and reservoir volume. However, in the cases for which there is a scarcity of data, analytical expression can be obtained by interpolation of the values of the table. Usually one of the most suitable technique for interpolation data is using polynomial function but unfortunately utilizing this function for solving the problem demand several parameters. Using power function in numerical computations of breach phenomena would be advantageous, because this function is monomial type and only one parameter needs to be estimated. In this thesis, we want to present that this approach is very accurate and suitable to represent the morphology of the reservoirs, at least for dam breach studies. To reach this aim, 97 case studies have been selected from three different geographical regions in the world. The results of this research have been shown that the power function is suitable to obtain an accurate fitting of the reservoir rating curve using a very limited number of surveyed elevations and volumes or areas. Furthermore in this part of the research it has been shown that two points are enough for a good fitting of the curve, or even only one if volume and surface are both available for an elevation close to normal or maximum pool. Results obtained for dam breach calculations using this equation, have the same quality of those achieved using the elevation-volume table. Moreover, this research have been shown that the exponent of power equation can be expressed by a formula which has a precise morphological meaning, as it represents the ratio between the volume which the reservoir would have if it were a cylinder with its base area and height equal to the respective maximum values of the actual reservoir, and the real volume of the reservoir. Regarding the second aspect, over the complexity of the mathematical models which have been used to predict the generation of dam breach hydrograph, it is considerable that the historical observed data of discharge peak values and typical breach features (top width, side slope and so on) have been usually utilize for model validation. Actually, the important problem which should be considered here is traditionally focused on what has been observed in the dam body, because the effects of the flood wave realized in the downstream water levels usually have been neglected. This issue seems considerable because required information for the civil protection and flood risk activities are represented by the consequences induced by the flood propagation on the areas downstream such as maximum water levels and maximum extent of flood-prone areas, flow velocity, front arrival times etc. The water surface data is almost never linked to the reservoir filling/emptying process which can be important information for the estimation of discharge coming from the breach, are available. Moreover, it is quite unusual to have records on the flood marks signs or other effects induced on the river bed, or on the man-made structures, downstream. For this reason finding well documented case study is one of the important part of any simulation study, especially for model validation. One of the few cases in this context is represented by the Big Bay dam, located in Lamar County, Mississippi (USA), which experienced a failure on 12 March 2004. In general analyzing the simplified models for dam breach simulation is the main purpose of this second important activity of the thesis. The simplified model have been utilized in this study, in order to identify a method that, on the basis of the results obtained in terms of simulated maximum water levels downstream, might effectively represent a preferential approach for its implementation not only in the most common propagation software but also for its integration in flood information systems and decision support systems. For the reasons explained above, attention here focuses on the parametric models, widely used for technical studies, and on the Macchione (2008) model, whose predictive ability and ease of use have been already mentioned. To reach this purpose both a 1-D and 2-D flood propagation modelling have been utilizing in this study. The results show that the Macchione (2008) model, without any operations of ad hoc calibration, has provided the best results in predicting computation of that event. Therefore it may be proposed as a valid alternative for parametric models, which need the estimation of some parameters that can add further uncertainties in studies like these.