Journal of Natural Disaster Science

PREDICTION OF THE OCCURRENCE OF DEBRIS FLOW AND A RUNOFF ANALYSIS BY THE USE OF NEURAL NETWORKS

Muneo HIRANO, Toshiyuki MORIYAMA and Keiichiro KAWAHARA
Department of Civil Engineering, Faculty of Engineering, Kyushu University, Fukuoka 812, Japan

(Received 4 December, 1995 and in revised form 23 February, 1996)

Abstract

Debris flows are feared for their potential to cause severe disasters, therefore studies on the occurrence and runoff of debris flows are needed. Several practical methods for forecasting a debris flow have been proposed to forecast the debris flow, however, the accuracy of these methods is not enough for practical use. Neural networks are introduced into the prediction of the occurrence of the debris flow and a runoff analysis. The method was verified by applying it to the Mizunashi River in the Unzen volcanic area. The results of this application led to the conclusion that the neural network model can be used to predict of a debris flow. The neural network model also is useful for identifying the critical condition for such a flow. The neural network model, which calculates the hydrograph of debris flow from the time series or rainfall, was verified by the observed data.

Key words

debris flow, neural network, critical rainfall, runoff analysis, Unzen volcano