In general, the error sources of flood forecasting by a runoff model include input data, model structure, and parameter setting. The objective of this study was development of a method to minimize errors due to parameter setting. Hydrological parameters of a distributed runoff model were optimized for each of the fifteen past floods by a huge variety of optimization algorithms and three error assessment functions. Although optimized parameters were widely distributed in the search range, a high correlation was found between the optimized parameters and the runoff characteristics. Furthermore, the predictive accuracy of the optimized parameter sets was validated by applying them to an additional ten floods. The validation showed that the parameter sets optimized by floods similar to the target flood in terms of runoff rate can reproduce the measured discharge with high accuracy.
