Author(s): Go Ohno; Takahiro Sayama; Yukinobu Oda
Linked Author(s):
Keywords: Water level prediction rainfall distribution watershed Neural Network safety management
Abstract: In river constructions, it is necessary to evacuate not only workers but also construction materials before a runoff of a river. To evacuate safely, the workers need to be warned about the upcoming water level rise, 24 hours in advance. The authors have been investigating a method to predict the water level by Neural Network using the spatial distribution of observed rainfall. As a result, the prediction accuracy of this method tended to be good when the cumulative rainfall for 6 hours before the runoff of the river was high and the rainfall was uniformly occurring in the watershed. In this paper, water level predictions were conducted among 56 river watersheds in Japan in order to examine whether the tendency was similar to the above. Also, the relationship between the characteristics of the watershed and the prediction accuracy was examined.
Year: 2025