抄録
In recent years, damage caused by torrential rain is occurring frequently in global scale. In order to prevent the damage caused by this, more accurate forecast system is needed. However, it is difficult to predict guerrilla rainstorm that occurs locally within widespread torrential rainfall. Therefore, it must be predict accurately the amount of torrential rain in anticipation of few minutes as much as possible. We develop short term forecasting method of growth and disappearance phenomenon of rainfall that based on the fact that torrential rainfall tend to develop and disappear within the period of 30 minutes to 1 hour. To measure the rainfall level and velocity of nucleus of rain drop, we use the high resolution radar. For an immediate and rapid and precise prediction of guerrilla rainstorm, at first, we develop the state transition algorithm to predict the rain-level. But for the short term prediction, it is not suitable for general methods. To overcome the weakness of the general state transition methods, we introduce the rapid state-transition ones based on short-period sampling data. Furthermore, to achieve the more precise prediction scheme, we also introduce the estimation method of the coordinates of center of gravity movement of rain cloud. The effectiveness of the newly proposed hybrid prediction algorithm that consist of short term state transition method and estimation of center of gravity of rain cloud was verified by numerical simulation for actual rain-level data. As a result, the prediction accuracy was remarkably improved.
