Incorporation of operational variables in stochastic hydrological dam safety analysis

Resumen

Throughout history, there have been situations in which dams designed and built in accordance with good engineering practice have suffered accidents with catastrophic consequences. Consequently, new regulations and standards tend to increase the hydrological safety requirements of Large Dams, influencing both the design and the adaptation of existing dams to the new criteria required. It is, therefore, necessary to define assessment methodologies that can be applied to all the existing dams in order to identify those that have a smaller safety margin and require action. In Spain, the criterion for hydrological safety is adopted according to a return period associated with the Design Flood. However, there is no rigorous criterion in the definition of this return period, given that the same flood can be characterized by its peak flow, volume or duration; or by a combination of these factors. In the definition of the Design Flood, with the exception of the Design Storm, most of the variables involved are taken as deterministic, when in fact they have a stochastic nature. Variables such as the temporal distribution of rainfall or the humidity conditions in the basin prior to the flood are, among others, determined in a deterministic manner by the designer despite the degree of uncertainty that exists regarding their calculation. Thanks to the advances in computer science, distributed modeling presents an alternative for addressing these issues. Distributed physically-based models coupled with stochastic climate generators provide a framework to simulate hydrological processes in a detailed manner; both temporally and spatially. In addition to the aforementioned variables, there are other extrinsic factors not contemplated in the definition of the return period of the floods associated with the hydrological safety of the dam, such as the initial reservoir level, or the operation and availability of the different spillway outlets of the dam. For this reason, the frequency curves of maximum reservoir levels and of maximum released outflows are considered to be the most representative variables of the hydrological safety; both of the dam and the downstream riverbed. Floods with high return periods may be less compromising than others with a smaller return period if the spillways cannot be properly operated. With regard to the initial level in the reservoir, not only does it affect the design regarding dam safety, but it also maintains a direct relationship with the reservoir's regulation capacity. In many of the existing reservoirs, the flood control volume is increased to alleviate the consequences of extreme floods, which lead to a reduction in the water resources available for the supply of the demands and, consequently, derive in economic losses. However, in the literature and in professional practice, both aspects are generally studied independently. The subject matter presented is of current importance, given the large number of existing dams that must be evaluated to guarantee the safety of the population, as well as the supply of water resources. This doctoral thesis seeks to provide a methodology which, within a stochastic modeling environment, will improve: a) the stochastic generation of floods, b) the operation of the spillway and dam outlets in a flood situation, c) the analysis of the influence of operational variables (initial level in the reservoir and probability of failure of the spillway outlets) and d) the definition of seasonal conservation levels in dams taking into account the regular operation in the reservoir.