SINGULAR VECTOR METHOD ON SHORT-TERM RAINFALL PREDICTION USING RADAR FOR HYDROLOGIC ENSEMBLE PREDICTION

Abstract

Ensemble rainfall short-term prediction with flood simulation in urban river basin is presented. Considering the small scale of storm event, the radar echo extrapolation by translation model was used for predicting rainfall from X-band polarimetric radar. A new attempt of the ensemble prediction system by using initial condition perturbed by singular vector was introduced to find perturbations of advection vector, which linearly grow most rapidly. The methodology was demonstrated throughout case studies in Kofu urban river basin, Japan, with convective and stratiform precipitation system cases. SV found two leading perturbations over three hour prediction, giving five ensemble members. The members were subsequently considered as uncertain input of distributed hydrological model for developing ensemble flood prediction. Verified with observed data, this approach could serve as a reliable and effective system for operational flood disaster prevention in urban area.