統計的二次近似理論を適用した流出予測システムの構成

Abstract

This research discusses how to build a stochastic state-space runoff model from a given deter-ministic one by introducing some noise terms into the latter in order to compensate observationand model errors, and develops the technique for the construction of a real-time stochastic runoffprediction system by the application of the filtering and prediction theory to the stochastic model.In order to cope with the non-linearity of the system, the statistical second-order approximationtheory is applid. This theory found in Mahalanabis-Farooq's work is extended to multidimensionalcases in this research. The numerical technique for the realization of the second-order approxima-tion is developed by the use of the U-D factorization of covariancce matrices and the Hermite-Gauss integration formula.An example in which the application of our method improves prediction performance is pre-sented.