Estimation of reservoir inflow with significant lateral inflow by using the adjoint equation method

Abstract Reservoir inflow estimation is important in flood control operation. Reservoir inflow includes three components, namely, upstream, tributary, and lateral inflow. However, the lateral inflow, which has a significant contribution to reservoir inflow during flood seasons, is difficult to estimate in regions where precipitation or meteorological data are lacking. A reverse model based on the adjoint equation method (AEM) was proposed to improve the estimated accuracy of reservoir inflow without using rainfall data. The model inverted the lateral inflow on the basis of observed hydraulic characteristic data, such as water levels and discharges. The AEM, which is considered an optimization method, determines the optimal lateral inflow by minimizing misfits between the observed and simulated data. The capability of the AEM was verified by two synthetic examples with varied boundary conditions. After running the forward model to solve one-dimensional Saint-Venant equations, the resulting water level hydrographs were used as known observations for the inversion. The practical applicability was tested by reproducing a historical flood event of the Three Gorges Reservoir by using water level data and outflow hydrograph. Results demonstrated that the AEM could successfully invert lateral and reservoir inflows by using observed data and has practical applicability in complicated reservoirs.