The main objective of this research is to devise a risk-based methodology for developing emergency operation schedules (EOS). EOS are decision tools that provide guidance to reservoir operators in charge of making real-time release decisions during major flood events. A computer program named REOS was created to perform the computations to develop risk-based EOS. The computational algorithm in REOS is divided in three major components: (1) synthetic streamflow generation, (2) mass balance computations, and (3) frequency analysis. The methodology computes the required releases to limit storage to the capacity available based on the probabilistic properties of future flows, conditional to current streamflow conditions. The final product is a series of alternative risk-based EOS in which releases, specified as a function of reservoir storage level, current and past inflows, and time of year, are associated with a certain risk of failing to attain the emergency operations objectives. The assumption is that once emergency operations are triggered by a flood event, the risk associated with a particular EOS reflects the probability of exceeding a pre-established critical storage level given that the same EOS is followed throughout the event. This provides reservoir operators with a mechanism for evaluating the tradeoffs and potential consequences of release decisions. The methodology was applied and tested using the Addicks and Barker Reservoir system in Houston, TX as a case study. Upstream flooding is also a major concern for these reservoirs. Modifications to the current emergency policies that would allow emergency releases based on the probability of upstream flooding are evaluated. Riskbased EOS were tested through a series of flood control simulations. The simulations were performed using the HEC-ResSim reservoir simulation model. Rainfall data recorded from Tropical Storm Allison was transposed over the Addicks and Barker watersheds to compute hypothetical hydrographs using HEC-HMS. Repeated runs of the HEC-ResSim model were made using different flooding and residual storage scenarios to compare regulation of the floods under alternative operating policies. An alternative application of the risk-based EOS in which their associated risk was used to help quantify the actual probability of upstream flooding in Addicks and Barker was also presented.

O trabalho trata do desempenho dos sistemas de reservatórios segundo a ótica dos usos múltiplos da água. A pesquisa é fundamentada no uso de modelos de otimização de Programação Não-Linear criados com a ferramenta GAMS e resolvidos com o pacote de otimização MINOS, que resultou no modelo SFPLUS. O estudo de caso refere-se ao sistema de reservatórios da bacia do rio São Francisco. Numa primeira fase, o estudo enfoca a otimização do sistema segundo dois métodos para análise de usos múltiplos: o método das restrições e o método das ponderações. Na seqüência, faz-se uma exploração sobre as alternativas para tratamento dos aspectos estocásticos do problema. Os principais resultados indicam que o método das restrições é mais fácil e direto de ser aplicado. Explicita claramente os trade-offs entre os usos competitivos. Porém, permite a análise de no máximo três usos. Por outro lado, o método das ponderações permite avaliar um número maior de usos. Contudo, é necessário estabelecer a priori os coeficientes de ponderação entre os usos competitivos. Quanto à questão estocástica, para o caso estudado, verifica-se que o número de séries sintéticas geradas não influi significativamente na forma das distribuições de probabilidades das variáveis geradas pelo modelo. Contudo, essas distribuições são sensíveis ao fato de se usar séries históricas ou séries sintéticas, principalmente nos seus extremos. Além disso, o uso de séries sintéticas afeta os resultados da análise de usos múltiplos quando comparados com os resultados obtidos a partir de dados históricos. / This thesis deals with multiple purpose reservoir systems operation and its performance. The main aspects are related to trade-off analysis by using non-linear programming optimization models. Therefore, the SFPLUS model were developed with the GAMS software and solved by the MINOS package. The case study refers to the São Francisco river basin reservoir system. Initially, some research is done on two methods of formulating the objective function: the restriction method and the weighting method. Secondly, an alternative analysis on the stochastic aspects involving the problem is presented. According to the main results, the restriction method is an easy and direct approach, by which is possible to express the trade-offs very clearly. However, it is not possible to analyze more than three uses. On the other hand, the main advantage of the weighting method is to allow the analysis of greater number of uses. The problem with the weighting method is concerned with the weighting coefficients, which have to be defined previously. Referring to the stochastic aspects, case study results show that the number of inflow series generated synthetically does not affect significantly the probability distributions of the system performance variables calculated by optimization models. However, the shape of those distributions is sensible to the fact of using historical data or synthetic series, mainly for extreme values. Moreover, the use of synthetic series affects multiple-use results when compared to the results generated with historical data.

O trabalho trata do desempenho dos sistemas de reservatórios segundo a ótica dos usos múltiplos da água. A pesquisa é fundamentada no uso de modelos de otimização de Programação Não-Linear criados com a ferramenta GAMS e resolvidos com o pacote de otimização MINOS, que resultou no modelo SFPLUS. O estudo de caso refere-se ao sistema de reservatórios da bacia do rio São Francisco. Numa primeira fase, o estudo enfoca a otimização do sistema segundo dois métodos para análise de usos múltiplos: o método das restrições e o método das ponderações. Na seqüência, faz-se uma exploração sobre as alternativas para tratamento dos aspectos estocásticos do problema. Os principais resultados indicam que o método das restrições é mais fácil e direto de ser aplicado. Explicita claramente os trade-offs entre os usos competitivos. Porém, permite a análise de no máximo três usos. Por outro lado, o método das ponderações permite avaliar um número maior de usos. Contudo, é necessário estabelecer a priori os coeficientes de ponderação entre os usos competitivos. Quanto à questão estocástica, para o caso estudado, verifica-se que o número de séries sintéticas geradas não influi significativamente na forma das distribuições de probabilidades das variáveis geradas pelo modelo. Contudo, essas distribuições são sensíveis ao fato de se usar séries históricas ou séries sintéticas, principalmente nos seus extremos. Além disso, o uso de séries sintéticas afeta os resultados da análise de usos múltiplos quando comparados com os resultados obtidos a partir de dados históricos. / This thesis deals with multiple purpose reservoir systems operation and its performance. The main aspects are related to trade-off analysis by using non-linear programming optimization models. Therefore, the SFPLUS model were developed with the GAMS software and solved by the MINOS package. The case study refers to the São Francisco river basin reservoir system. Initially, some research is done on two methods of formulating the objective function: the restriction method and the weighting method. Secondly, an alternative analysis on the stochastic aspects involving the problem is presented. According to the main results, the restriction method is an easy and direct approach, by which is possible to express the trade-offs very clearly. However, it is not possible to analyze more than three uses. On the other hand, the main advantage of the weighting method is to allow the analysis of greater number of uses. The problem with the weighting method is concerned with the weighting coefficients, which have to be defined previously. Referring to the stochastic aspects, case study results show that the number of inflow series generated synthetically does not affect significantly the probability distributions of the system performance variables calculated by optimization models. However, the shape of those distributions is sensible to the fact of using historical data or synthetic series, mainly for extreme values. Moreover, the use of synthetic series affects multiple-use results when compared to the results generated with historical data.

Dams and reservoirs are important components of water resource management systems, but their operational sensitivity to streamflow variability may make them vulnerable to climate change. Climate change is likely to affect the magnitude and timing of streamflow, motivating the assessment of potential impacts on dams and reservoirs. Here I examine a case study of Cougar Dam, a multipurpose dam in Oregon, USA, to assess potential impacts of future climate change on operational performance. In the first portion of this study, I examine the historical operation of Cougar Dam, to understand (1), whether operational objectives have been achievable in the past despite operational variability, and (2) how climatic variation is expressed in operational trajectories. By analyzing historical streamflow and operations data using a set of metrics, I characterize variability in past operations and how that variability relates to streamflow. I also employ a reservoir model to distinguish operational differences due to streamflow variability from variability due to other factors that affect operations. I find that operational objectives have been achievable, despite variability in operations and departures from the ideal operational trajectory. Throughout the historical period, flood control operations have almost always kept reservoir outflows below the desired maximum outflow. Although filling occurs 9 days late on average, the reservoir has filled in all but 6 out of 37 years. Although drawdown occurs 47 days early on average, early drawdown does not generally impact recreation and allows minimum outflows to be met every day during all but the driest year. I also find that total seasonal inflow is correlated with measures of operational performance, and that other factors besides climate play an important role in determining operational trajectories. I conclude that operations of Cougar Reservoir are vulnerable to climate change, but that operational flexibility may mitigate some of the potential impacts. In the second portion of this study I assume that current operating rules will be kept in place and I aim to understand what types of operational impacts may be expected, when they may be expected to occur, and whether the operational impacts may necessitate changing operational rules. I employ both a traditional climate impacts assessment approach to assess changes over time as well as a scenario-neutral approach to generalize relationships between streamflow and operations of Cougar Dam. I find that projected increases in winter streamflow could result in up to twice the number of downstream high flows than in the past and that projected decreases in summer streamflow could result in earlier reservoir drawdown by up to 20 days on average. Additionally, filling of the reservoir may occur up to 16% more often or 11% less often than in the past, depending on spring flow magnitude and timing. I also find that there are strong general relationships between total inflow volume and flood control performance, and that there are total inflow thresholds for whether or not the reservoir will fill or will be full enough for recreation in late summer. I conclude that future modification of operating policies may be warranted, but that there will likely be tradeoffs between operating objectives in the future even if operating rules are modified. / Graduation date: 2013