抄録
We are proposing a simple yet very efficient data-driven model based on an artificial neural network to predict the water level at Hirakata Station without using rainfall-forecast information. The proposed model is based only on the observed upstream water levels at Katsura Station on the Katsura River, Mukaijima Station on the Uji River, and Inooka Station on the Kizu River. The proposed model uses a simple, single hidden layer on a feed-forward neural network, and it does not require much input data, but the prediction is sufficiently stable and reliable up to 9 hours of lead time, which can be very useful information for practical flood warning.
