We have attempted to identify parameters of a tank model for the Saba River basin, where the great sediment disaster occurred in July 2009, with observation data in order to present a highly accurate method providing a critical line for sediment-related disasters (CL). The identification of the model parameters can be done with high accuracy using the shuffled complex evolution method which is one of the global optimization methods and the proposed model is more reproducible than the Japan Meteorological Agency (JMA) model. We set up a CL by using soil water index calculated by this model and found that the accuracy of CL evaluated by our model is higher than that evaluated by JMA's model. It is found that the construction of an appropriate tank model based on the observation data is very important for improving the accuracy of sediment-related disaster predictions.