Many investigations have advanced the understanding of important factors contributing in reservoir sedimentation. However, predicting the accumulation of sediment and delta formation in the reservoir is still a challenging problem due to several uncertainties in the modelling process. In estimating reservoir sedimentation, the performance of modelling can significantly be increased by improving the quality of sediment load boundary conditions and model calibration. This study utilizes wavelet neural networks (WA-ANN) and an automatic model calibration method to enhance the performance of the numerical model for estimating reservoir sedimentation and tempo-spatial deposits in the Taberla reservoir located on the Indus River, in northern Pakistan. The model performance was quantified using coefficients of determination (R2) and Nash-Sutcliffe Efficiency (NSE). The results showed that an optimal hydrodynamic calibration (of TELEMAC) using an automatic calibration technique, along with more precise sediment load boundary conditions estimated by WA-ANN, enabled successive morphodynamic model (SISYPHE) to more accurately anticipate the reservoir bed. The calibrated model (for 1983-1984) with R2=0.97, NSE=0.97 also validated the bed levels (for 1985-1990) with R2=0.96 and NSE=0.95. The presented modelling concept could be used to improve/design sediment management strategies for the existing and planned hydraulic structures in other ungauged or poorly-gauged rivers.     «

Many investigations have advanced the understanding of important factors contributing in reservoir sedimentation. However, predicting the accumulation of sediment and delta formation in the reservoir is still a challenging problem due to several uncertainties in the modelling process. In estimating reservoir sedimentation, the performance of modelling can significantly be increased by improving the quality of sediment load boundary conditions and model calibration. This study utilizes wavel...     »