Abstract
This paper presents the method to improve the stability of model parameters and accuracy of the tank model. The tank model includes so many parameters and high flexibility that it is difficult to identify the unique and best parameter for a certain watershed. This instability of model parameters causes instability of accuracy in long-term analysis. To decrease the flexibility and increase the physical fidelity of the model, new restriction conditions are introduced into the parameter identification process. This new condition could improve stability of the model considerably. The result suggests that stable parameters can be identified even from short-term data set by using appropriate restriction conditions.
