Author(s): Sungwon Kim

Linked Author(s):

Keywords: Reliability analysis; Uncertainty; Input data information; Sensitivity analysis

Abstract: Sensitivity analysis is used to eliminate input data information uncertainty of EDRNNM, which was developed during previous study for flood stage forecasting at Musung station of Wi-stream, one of IHP representative basins in South Korea. Uncertainty analysis is performed to determine sensitivity level, which is calculated by the change of output node according as each input node changes. ST5 Dongkok station is proved to have a little sensitivity as compared with the other stations of Wi-stream. The improved EDRNNM is consisted of four input nodes, which are organized by the reduction of ST5 Dongkok station from five initially chosen input nodes. Flood stage forecasting accuracy of the improved EDRNNM does not reduce as compared with that of EDRNNM with five input nodes. The reliability of flood stage forecasting is proved to have a high accuracy. Therefore, the optimal EDRNNM is suggested by elimination of structural and input data information uncertainty from previous and this study. Basically, the new methodology is introduced that the uncertainty embedded structure and input data information of ANNs-based model can be easily eliminated. Therefore, it enables the engineers or hydrologists to prevent the unnecessary data collection and operate flood stage forecasting system with economic benefits such as lower costs in this study.

DOI:

Year: 2005