貯留モデルによる実時間流出予測に関する基礎的研究

Abstract

The authors investigate models and methods for real-time prediction of flood runoff. Then, they describe runoff system in a stochastic continuous-discrete state-space model. Multiplicativecolored noise is incorporated into the model as model error, and multiplicative white noise as observa-tion error.Kalman filter theory, statistical linearization and a simulation of rainfall prediction are employedin the prediction algorithm. The present method is rather practical. The accuracy of the runoffprediction is quantitatively investigated in relation to that of the rainfall prediction and model error.