ABSTRACT
Surface meteorological variables are essential in the characterization of ocean-atmospheric interactions. Air Temperature (AT), Relative Humidity (RH), Sea Level Pressure (SLP), Sea Surface Temperature (SST) and Wind Speed (WS) are surface meteorological variables monitored by buoys of the the Prediction and Research Moored Array in the Tropical Atlantic (PIRATA) Project. In this work, a year-ahead prediction procedure based on knowledge of previous periods is coupled with regression via Support Vector Machines (SVM). The procedure is focused on seasonal and intraseasonal aspects of AT, RH, SLP, SST and WS. Data from a PIRATA buoy is used to feed the SVM models using information about curvatures of each variable and the prediction models are assessed by means of the Mean Absolute Error (MAE). The obtained results indicate that the used methodology is a promising technique for the prediction of meteorological variables.
