Sea surface temperature forecasts using on-line local learning algorithm in upwelling regions

When forecasting geophysical time series, long records are not always avail able. Therefore, the possible physical events may not be evenly represented in the available dataset and a unique neural network modelling the phenome non of interest may not be accurate. As an alternative, a local learning al gorithm is presented in this paper. It is applied to the prediction of sea surface temperature in an upwelling area. It provides encouraging results w hen applied to forecasts for eight days ahead. (C) 2000 Published by Elsevi er Science B.V. All rights reserved.