Abstract
The present study investigates the relationship between the accuracy of predicted rainfall intensity by using the advection model and the patterns of parameters as well as the number of preceding data. The advection model uses nine parameters, and they are assigned by the regression method using several preceding data. Each set of nine parameters has kinematic meaning and it is defined as a pattern of prediction. By using the patterns of parameters, uncertainty of prediction decreases, and the accuracy of the rainfall prediction improves. In order to improve the short-term rainfall prediction, this study investigated the method to select appropriate pattern using the data from high resolution radar and the advection model. The present study found out that movement and time development of rainfall predicted by the advection model are characterized by Eigen values of the matrix whose elements are a subset of the parameters. Moreover, it shows that the Eigen values provide information on which pattern should be used.
