Journal of Natural Disaster Science

DEVELOPMENT OF TECHNIQUES FOR ON-LINE FORECASTING OF RAINFALL AND FLOOD RUNOFF

Takuma TAKASAO
Professor
and
Michiharu SHIIBA
Instructor, Department of Civil Engineering, Kyoto University

(Received 16 Nov., 1984 and in revised form 18 Feb., 1985)

Abstract

An on-line rainfall forecasting procedure that uses digital radar data is presented. The model can provide not only parallel translation of rainfall distribution, but rotation, shearing strain and expansion of the rainfall distribution pattern.

Our formulation of the model parameter estimation leads to a linear least squares estimation problem that is easily solved with a mini or micro-computer.

An on-line runoff forecasting procedure also is presented. It uses a stochastic state-space runoff model constructed from a given deterministic state-space runoff model. To cope with the nonlinearity of the system model, a statistical second-order approximation technique and the UDUT factorization technique are applied.

Key words

rainfall forecasting, runoff forecasting, radar, statistical second-order approximation