RIVER LEVEL ESTIMATION USING ARTIFICIAL NEURAL NETWORK FOR URBAN SMALL RIVER IN TIDAL REACH

Abstract

Prediction of water level in small rivers is great interest for flood control in an urban area located in the river mouth. The tidal river water level is affected by not only flood discharge but also tide, atmospheric pressure, wind direction and speed. We propose a method of estimating river water level considering these factors using an artificial neural network model for the Kanda River located in the center of Tokyo. The effects by those factors are quantitatively investigated. As for the effects by the atmospheric pressure, river water level rises about 7cm per 5hPa increase of the pressure regardless of river discharge under the conditions of 1m/s wind speed and north wind direction. The accurate rating curve for the tidal river is finally obtained.