Forecasting near-surface ocean winds with Kalman filter techniques

Anders Malmberg, Ulla Holst and Jan Holst

Centre for Mathematical Sciences
Mathematical Statistics
Lund Institute of Technology,
Lund University,

ISSN 1403-9338
In this paper a statistical forecasting model designed for bounded areas of near-surface ocean wind speed is implemented.
Dimension reduction is achieved by decomposing the covariance structure into one large-scale and one small-scale component using empirical orthogonal functions. The large-scale component is modelled with an AR process and forecasts are calculated by applying a Kalman filter.
The model is suited for stable weather situations as for unsteady situations it requires more frequent wind information. From the prediction variance fields it is possible to identify where unexpected weather usually enter the area.
Key words:
Dimension reduction, principal components, space-time Kalman filtering, forecasting, near-surface ocean winds