Hindcasting of storm waves using neural networks

Abstract

Cyclone generated waves play a significant role in the design of coastal and offshore structures. Instead of conventional numerical models, neural network approach is used in the present study to estimate the wave parameters from cyclone generated wind fields. Eleven cyclones, which crossed the southern east coast of India between 1962 and 1979, are considered for analysis in this paper. The parametric hurricane wave prediction model by Young (1988) [Young, I.R., 1988. Parametric hurricane wave prediction model. Journal of Waterways Port Coastal and Ocean Engineering 114(5), 637-652] is used for hindcasting the wave heights and periods. Estimation of wave heights and periods is carried out using back propagation neural network with three updated algorithms, namely Rprop, Quickprop and superSAB. In neural network, the estimation is carried out using (1) difference between central and peripheral pressure, radius of maximum wind and speed of forward motion of cyclone as input nodes and the wave heights and periods as output nodes and (2) wind speed and fetch as input nodes and wave heights and periods as output nodes. The estimated values using neural networks match well with those estimated using Young's model and a high correlation is obtained namely (0.99).