Sediment yield is a complex phenomenon of weathering, land sliding, and glacial and fluvial erosion. It is highly dependent on the catchment area, topography, slope of the catchment terrain, rainfall, temperature, and soil characteristics. This study was designed to evaluate the key hydraulic parameters of sediment transport for Kali Gandaki River at Setibeni, Syangja, located about 5 km upstream from a hydropower dam. Key parameters, including the bed shear stress (τb), specific stream power (ω), and flow velocity (v) associated with the maximum boulder size transport, were determined throughout the years, 2003 to 2011, by using a derived lower boundary equation. Clockwise hysteresis loops of the average hysteresis index of +1.59 were developed and an average of 40.904 ± 12.453 Megatons (Mt) suspended sediment have been transported annually from the higher Himalayas to the hydropower reservoir. Artificial neural networks (ANNs) were used to predict the daily suspended sediment rate and annual sediment load as 35.190 ± 7.018 Mt, which was satisfactory compared to the multiple linear regression, nonlinear multiple regression, general power model, and log transform models, including the sediment rating curve. Performance indicators were used to compare these models and satisfactory fittings were observed in ANNs. The root mean square error (RMSE) of 1982 kg s−1, percent bias (PBIAS) of +14.26, RMSE-observations standard deviation ratio (RSR) of 0.55, coefficient of determination (R2) of 0.71, and Nash–Sutcliffe efficiency (NSE) of +0.70 revealed that the ANNs’ model performed satisfactorily among all the proposed models.