Search

Reservoirs considerably affect river streamflow and need to be accurately represented in environmental impact studies. Modeling reservoir outflow represents a challenge to hydrological studies since reservoir operations vary with flood risk, economic and demand aspects. The Brazilian Interconnected Energy System (SIN) is an example of a unique and complex system of coordinated operation composed by more than 160 large reservoirs. We proposed and evaluated an integrated approach to simulate daily outflows from most of the SIN reservoirs (138) using an Artificial Neural Network (ANN) model, distinguishing run-of-the-river and storage reservoirs and testing cases whether outflow and level data were available as input. Also, we investigated the influence of the proposed input features (14) on the simulated outflow, related to reservoir water balance, seasonality, and demand. As a result, we verified that the outputs of the ANN model were mainly influenced by local water balance variables, such as the reservoir inflow of the present day and outflow of the day before. However, other features such as the water level of 4 large reservoirs that represent different regions of the country, which infers about hydropower demand through water availability, seemed to influence to some extent reservoirs outflow estimates. This result indicates advantages in using an integrated approach rather than looking at each reservoir individually. In terms of data availability, it was tested scenarios with (WITH_Qout) and without (NO_Qout and SIM_Qout) observed outflow and water level as input features to the ANN model. The NO_Qout model is trained without outflow and water level while the SIM_Qout model is trained with all input features, but it is fed with simulated outflows and water levels rather than observations. These 3 ANN models were compared with two simple benchmarks: outflow is equal to the outflow of the day before (STEADY) and the outflow is equal to the inflow of the same day (INFLOW). For run-of-the-river reservoirs, an ANN model is not necessary as outflow is virtually equal to inflow. For storage reservoirs, the ANN estimates reached median Nash-Sutcliffe efficiencies (NSE) of 0.91, 0.77 and 0.68 for WITH_, NO_ and SIM_Qout respectively, compared to a median NSE of 0.81 and 0.29 for the STEADY and INFLOW benchmarks respectively. In conclusion, the ANN models presented satisfactory performances: when outflow observations are available, WITH_Qout model outperforms STEADY; otherwise, NO_Qout and SIM_Qout models outperform INFLOW. RESUMO Reservatórios afetam consideravelmente a vazão dos rios e por isso precisam ser adequadamente representados em estudos de impactos ambientais. Simular defluência de reservatórios representa um desafio para estudos hidrológicos já que sua operação depende do risco de inundações e aspectos econômicos e de demanda. O Sistema Interligado Nacional (SIN) é um exemplo de um sistema único e complexo de operação coordenada composto por mais de 160 grandes reservatórios. Então foi proposto e avaliado uma abordagem integrada para simular defluências diárias da maioria dos reservatórios do SIN (138) usando um modelo de Redes Neurais Artificiais (RNA), distinguindo reservatórios de armazenamento e a fio d’água e testando casos em que defluência e nível estavam disponíveis ou não como dados de entrada. Além disso, foi investigada a relação entre as variáveis de entrada propostas (14) e a defluência simulada, sendo as variáveis relativas ao balanço hídrico do reservatório, sazonalidade e demanda indireta. Como resultado, foi verificado que as saídas do modelo de RNA foram principalmente influenciadas pelas variáveis locais de balanço hídrico, como a afluência do mesmo dia e a defluência do dia anterior. No entanto, outras variáveis como o nível de 4 grandes reservatórios representativos de diferentes regiões do país (que infere sobre à demanda pela disponibilidade de água para geração de energia) aparentemente teve uma influência considerável nas saídas do modelo. Esse resultado aponta para vantagens em se usar uma abordagem integrada ao invés de olhar cada reservatório isoladamente. Em termos de disponibilidade de dados, foram testados cenários com (WITH_Qout) e sem (NO_Qout and SIM_Qout) defluência e níveis observados como variáveis de entrada do modelo de RNA. Esses 3 modelos de RNA foram comparados com dois benchmarks simples: a defluência é igual a defluência do dia anterior (STEADY) e a defluência é igual a afluência do mesmo dia (INFLOW). Em reservatórios de armazenamento, as estimativas dos modelos atingiram uma mediana do coeficiente de Nash-Suthcliffe (NSE) de 0.91, 0.77 e 0.68 para os modelos WITH_, NO_ e SIM_Qout respectivamente, comparado com uma mediana do NSE de 0.81 e 0.29 para o STEADY e INFLOW respectivamente. Concluindo, os modelos RNA apresentaram performances satisfatórias: quando dados de defluências dos dias anteriores estão disponíveis, WITH_Qout é superior ao STEADY; caso contrário, NO_Qout e SIM_Qout são superiores ao INFLOW.

River floodplains and reservoirs interact throughout a basin drainage network, defining a coupled human-water system with multiple feedbacks. Recent modeling developments have aimed to improve the representation of such processes at regional to continental scales. However, most large-scale hydrological models adopt simplified lumped reservoir schemes, where an offline routine is run with inflows estimated by the model, with limited consideration of the complementarity between floodplains and reservoirs on altering the hydrological regime at regional scale. This paper presents a novel approach that fully couples river-floodplain-reservoir hydrodynamic and hydrological models, significantly improving the representation of reservoir dynamics and operation in the river-floodplain-reservoir continuum at large scale and across multiple dam cascades. The model is applied to the Parana River Basin with explicit simulation of 31 large dams and river hydraulic variables at basin scale. Three types of reservoir bathymetry representation are compared, from lumped to distributed methods, combined with three reservoir operation schemes and varying degrees of input data requirement within two parameterization scenarios (global and regional setups). The operation schemes were more relevant than the reservoir bathymetry representation to estimate downstream flows and water levels. While the data-driven operation scheme, based on linear regressions between observed water levels and dam outflows, provided the best estimates of both active storage and discharges, the more generic operation reasonably estimated discharges and peak attenuation, albeit not as accurately for active storage. The global parameterization of reservoir operation resulted in poorer performance compared to the regional-based one, but it satisfactorily modeled discharge and peak attenuation. Regarding the reservoir bathymetry representation, a basin scale comparison of the lumped and distributed schemes indicated the inability of the former to represent backwater effects. This was further corroborated by validating the longitudinal water level profile of Itaipu dam with ICESat satellite altimetry data. Finally, the model was used to show the complementarity between floodplains and reservoirs on attenuating floods at regional scale. Large scale models should move beyond offline coupling strategies, and include regional-based, data-driven reservoir operation schemes together with a distributed representation of reservoir bathymetry into river-floodplain hydraulic schemes. This will largely improve the estimation of river discharges, water levels and flood storage, and thus the model ability to represent the regional scale river-floodplain-reservoir continuum.