Improvement of wave height forecast in deep and intermediate waters with the use of stochastic methods

Theocharis, Z; Memos, C; Koutsoyiannis, D

dc.contributor.author	Theocharis, Z	en
dc.contributor.author	Memos, C	en
dc.contributor.author	Koutsoyiannis, D	en
dc.date.accessioned	2014-03-01T02:54:12Z
dc.date.available	2014-03-01T02:54:12Z
dc.date.issued	2006	en
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/36711
dc.title	Improvement of wave height forecast in deep and intermediate waters with the use of stochastic methods	en
heal.type	conferenceItem	en
heal.publicationDate	2006	en
heal.abstract	In recent years, data assimilation and artificial neural network techniques have been used in a number of wave height forecast improvement efforts. In this work we present the application of linear and non-linear stochastic techniques to show that WAM background errors can be reasonably predicted by using a limited number of buoy observations. Re-run of the wave model is not required. The first assessment, conducted in the Aegean Sea, refers to the improvement of the significant wave height prediction in deep water. The results were checked against four pilot-study monitoring stations. The assessment had a two-fold scope. First, a study conducted in a time domain fashion using four stochastic models whose explanatory variables are the WAM prediction and the measured wave height at previous steps. Two bivariate linear models, a trivariate linear model and two versions of a non-linear bivariate model were used and resulted in a significant forecast improvement, irrespectively of the application time period and of the location of the prediction. The coefficients of determination increased from approximately 0.7 (WAM) to over 0.9, suggesting that this method may be suitable for operational use. The second part of the application consists of a space-wise study including spatial stochastic modeling and wave information transfer aiming at expanding the improvement described above in space and especially in coastal regions. We found that wind information can help to improve the said prediction in time and space without using measurements or satellite observations, except for a calibration period. The applied stochastic methods show a somehow limited but steady improvement. To avoid the Aegean Sea complexity and peculiarity, further examination was conducted in two locations of the Indian Ocean. A non linear transformation in the stochastic models which is related to the swell content optimizes the improvement of the wave height prediction in intermediate waters using the offshore measurement. The improvement of the wave height prediction is reaching high coefficients of determination (~0.9).	en