QuikSCAT measurements and ECMWF wind forecasts - an explorative study

Anders Malmberg

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

ISSN 1403-9338
The aim of this paper is to study the covariance structure for the difference between satellite measurements and meteorological forecasts of near-surface ocean winds. This type of covariance enters in assimilation techniques such as Kalman filtering.

I all, six residual fields are studied. Us of a Box-Cox transformation leads to approximately symmetrically distributed residual wind speed fields. For these, two isotropic covariance models are explored, both belonging to the Matérn family. Substantial variations in the parameter estimates for the different fields are found.

Analytical expressions for the covariance functions of the original fields are hard to obtain and hence a simulation based approach is suggested. However, these simulated fields do not resemble the original residual fields very well. Hence, the isotropic covariance model should be reconsidered in further research and the present weather conditions might be considered in an extended dynamical model.
Key words:
ECMWF forecasts, QuikSCAT observations, residual wind speed, Box-Cox transformation, variogram parameters