Modelling of Significant Wave Height Using Wavelet Transform and Gmdh

Sajjad Shahabi, Mohamad-Javad Khanjani

Monday 29 june 2015

17:30 - 17:33h at Amazon (level 1)

Themes: (T) Water engineering, (ST) River and coastal engineering, Poster pitches

Parallel session: Poster pitch: 3C. Coastal Engineering

Forecasting of significant wave height (SWH) is one of the most important parameter for coastal and ocean engineering operations. This study was proposed a hybrid wavelet-GMDH approach to forecast SWH up to 48h lead time. The original time series of SWH were decomposed into spectral band. After that, these decomposed subseries were given as input to the GMDH model to forecast the SWH. Here, three different indices of efficiency and error indices including the index of agreement (Ia), coefficient of efficiency (CE) and root mean square error (RMSE) were employed to assess of results. The decomposition of original series to multi resolutions time series prepare more accurate results for forecasting of SWH especially in higher lead time such as 24 and 48 h. For instance, Ia increases from 0.35 to 0.80 for level 4 and 7 of decomposition for 48h lead time. Data series obtained from buoys located off east coast of USA (the North Carolina coast) were used to train and test proposed model.