Some numerical prediction methods for the wind speed in the sea based on ERS-1 scatterometer wind data

Three functional models, polynomial, autoregression and autoregressive movi ng average models, are fitted and compared in fitting and predicting the se a wind speeds based on the ERS-1 radar altimeters. the robust maximum likel ihood theory is introduced in fitting the functional models. To purify the satellite scan data for the sea winds and construct the data sets with equi valent intervals, we deleted the data which responds to the ice and land, t hen using the retained pure sea data sets to fit a simple polynomial and in terpolate the sea data that was deleted. By many trial computations, we fin d that the autoregressive moving average model is not only more complex but also less accurate than the autoregression model in our data sample. The r obust autoregression model has not only the best inner precision, but also the most accurate prediction. The accuracy is very little changed with the increase of the distance between the measurement points and the predicted p oints.