Computational & Technology Resources
an online resource for computational,
engineering & technology publications
Civil-Comp Proceedings
ISSN 1759-3433
CCP: 82
Edited by: B.H.V. Topping
Paper 45

Back-Propagation Neural Networks for Prediction of Storm Surges

C.P. Tsai+, T.L. Lee*, T.J. Yang* and Y.J. Hsu$

+Department of Civil Engineering, National Chung Hsing University, Taichung, Taiwan
*Department of Construction Technology, Leader University, Tainan, Taiwan
$Marine Meteorology Center, Central Weather Bureau, Taipei, Taiwan

Full Bibliographic Reference for this paper
C.P. Tsai, T.L. Lee, T.J. Yang, Y.J. Hsu, "Back-Propagation Neural Networks for Prediction of Storm Surges", in B.H.V. Topping, (Editor), "Proceedings of the Eighth International Conference on the Application of Artificial Intelligence to Civil, Structural and Environmental Engineering", Civil-Comp Press, Stirlingshire, UK, Paper 45, 2005. doi:10.4203/ccp.82.45
Keywords: tide, storm surge, typhoon, artificial neural network, back-propagation procedure.

The regular rise and fall of the astronomical tide is caused by the continuous gravitational attraction of the moon and sun, which is predicted accurately by a conventional method of the harmonic analysis. Contrary to the astronomical tide, the storm surge is caused by meteorological factors, which are defined by the difference between the actual observed water level from the tidal gauge and the predicted astronomically-induced tide during the storm. This meteorological tide is mainly induced by tropical low pressures and strong winds, which may pile-up the mean water level to several metres. This increase of the water level due to storms may cause severe flooding in the coastal area, particularly when the meteorological tide coincides with the high waters of spring tides. A comprehensive review of the global nature of storm surges was given by Murty [1].

The prediction of the storm surge is more complicated than the astronomical tide owing to the ambiguous uncertainties of the prediction of the storm. The numerical hydrodynamic models or the empirical methods were conventionally used to estimate the storm surges. But the accuracy of the predicted models is still to be improved. Alternatively, this paper proposed an application of the artificial neural network (ANN) to the storm-surge prediction. A supervised neural network with the back-propagation procedure is adopted in the study.

In the present ANN model, the prediction of the water level during the storm is done by a time series that includes the three major parameters affecting the storm surge including the atmospheric pressure, wind velocity and wind directions, input hour by hour. The total water level during the storm is to be predicted in this study. The corresponding time series of the astronomical tide obtained from the harmonic analysis is input to the input layer. Therefore, there are four variables in the input layer. In the output layer, there is only one variable, which is the total water level.

To illustrate the applicability of the proposed method, the hourly water-level data due to three typhoons collected from a tidal station were used in the study. The three typhoons are named as the Vamco (8/19-8/20, 2003), the Krovanh (8/22-8/23, 2003) and the Dujuan (8/31-9/1, 2003). The advancing path of the Vamco is quite different from the other two. The time series of the water level during the storm in the neural network is forecast based on the training of the water level due to the previous storm. Based on the results, it is found that the neural network is efficiently applied to forecast the storm surge though their typhoon paths and the scales of the wind velocity are different. The root-mean-square errors of the predictions are less than 0.1m, and the correlation coefficients are higher than 97%; which shows that a good performance is achieved by using the artificial neural network.

T.S. Murty, "Storm surges-meteorological ocean tides", Canadian Bulletin of Fisheries and Aquatic Sciences 212, 876-897, 1984.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £80 +P&P)