Abstract:The variation of tide and people's life are very closely related. The tides are composed of astronomy and violent tides. In this study, we focus on astronomy tide and try to construct a model with artificial neural network to simulate and forecast the astronomy tide at the Tanshui River at Ho-Kou. Tidal analysis has considerable advantage over the analysis of periodic behavior of the system through application of the Fourier-Stieltjes transform. However a large amount of data is needed to establish the forecasting model. In this study, a neural network, Back-propagation (BPNN), is applied for approaching to the estuary water-stage forecasting. And the water-stages calculated by BPNN are compared with that calculated by time series analysis. The results show that BPNN promises a high accuracy to forecast up to six-hour-ahead water-stage.