Properties of variational data assimilation for synthetic aperture radar wind retrieval
Article first published online: 3 MAY 2008
Copyright 2008 by the American Geophysical Union.
Journal of Geophysical Research: Oceans (1978–2012)
Volume 113, Issue C5, May 2008
How to Cite
2008), Properties of variational data assimilation for synthetic aperture radar wind retrieval, J. Geophys. Res., 113, C05006, doi:10.1029/2007JC004534., and (
- Issue published online: 3 MAY 2008
- Article first published online: 3 MAY 2008
- Manuscript Accepted: 18 JAN 2008
- Manuscript Revised: 21 DEC 2007
- Manuscript Received: 4 SEP 2007
- Synthetic aperture radar;
- wind retrieval;
- variational data assimilation
 The quality of marine wind vector retrieved from variational data assimilation (VAR) of Synthetic Aperture Radar (SAR) backscatter observation is assessed. It is found that the observation is most sensitive to wind speed. The retrieved wind direction from VAR is largely influenced by background wind direction and most of the SAR observation variability is assigned to wind speed. Non-linearity of the Geophysical Model Function (GMF) introduces wind speed bias, modulated by wind direction anisotropy (up-downwind/crosswind difference). The examination of the background wind vector departure from observation reveals two regimes: a quasi-linear response to wind direction for high background wind speed; and a rather monotonic response with two sharp transitions located at crosswind directions for low background wind speed. Information content of SAR observation is estimated using the entropy reduction approach, both analytically and from Monte-Carlo simulations. Crosswind directions have the lowest information content and correspond to those where non-linearity introduces largest discrepancies between analytic and Monte-Carlo estimations. The linear approximation of the GMF needed in the incremental VAR formulation is examined. The retrieved winds using the incremental formulation are in good agreement with those using the non-linear GMF. Monte-Carlo simulations reveal specific situations, around sharp transitions at crosswind directions, where both linear and non-linear VAR formulations may produce more noise than extract information form observations.