REAL-TIME FLOOD FORECASTING USING A PARTICLE FILTER COMBINED WITH THE RRI MODEL AND A RIVERBED EVOLUTION MODEL

Abstract

 In the past studies on a data assimilation, few studies have evaluated rainfall-runoff processes and hydraulic processes at the same time, and there are no studies using riverbed evolution models for real-time prediction. The objectivites of this study is to improve water level of accuracy using a paticle filter combined with the RRI model and riverbed evolutions. A particle filter is simultaneously and sequentially estimated riverbed evolution at the water level gauging and slope water depth, which is the initial conditions for the RRI model. In addition, a riverbed evolution model is predicted the riverbed according to discharge by the RRI model for the next six hours. The target river is Seri River in Shiga Prefecture, Japan. As a result, the calculated water level is assimilated to the observed one at the present time. Furthermore, the forcasted water level considering the riverbed evolution can be accurately computed for the next six hours.In conclusion, the practicality and validiry of the method we proposed in this study were verified.