STUDY ON A PARTICLE-FILTER APPLICATION METHOD FOR FLOOD FORECASTING CONSIDERING RAINFALL-RUNOFF CHARACTERISTICS

Abstract

 To develop a real-time flood forecasting system for small and medium-sized rivers, we investigated the characteristics in the case of applying a particle filter to a rainfall runoff model that expresses runoff processes with different time constants such as surface flow, intermediate flow, and river channel flow. The results found that the forecasted water level decreased due to the time delay of the runoff processes when the particle filter acted to make negative corrections to the slope water depth in the increment period. We also indicated that it is possible to prevent the accuracy of water-level forecasting from decreasing by adding an algorithm that sequentially corrects the correction coefficient of the particle filter and changes the initial distribution of the state value.