Abstract
In a series of previous papers, the well-known FIS/ANFIS systems have been successfully combined with a nonstationary time series modelling for improved predictions of wind and wave parameters. The initial time series is first decomposed into a seasonal mean value and a residual stationary part multiplied by a seasonal standard deviation. Then, the FIS/ANFIS models are applied to the stationary part. Then, they are combined with the already estimated seasonal patterns (mean value and standard deviation) to obtain forecasts of the full time series. In the present paper, different sources of data are combined for the estimation of different parts of the time series (hindcast or buoy for the stationary part and buoy or satellite for the seasonal patterns). In this way, one data from different sources and from different time periods can be combined with very good results. The performance of forecasting procedures is assessed by means of well-known error measures.