Since 22 January 2002, near-surface wind information observed by QuikSCAT has been assimilated in the operational 4D-Var system at ECMWF. A 50 km wind product is obtained by inversion of all available backscatter information in four 25 km cells. JPL-flagged rain contaminated cells are excluded from this procedure. Two out of four wind solutions are presented to the 4-D Var system by the use of a double-well cost function. Quality control (QC) on azimuth diversity is performed which is mainly active in the nadir region. In addition, variational QC dynamically rejects data with large departures from the analysis.

Prior to the assimilation, a global check on instrument performance is performed on 6-hourly data batches. This QC is based on the value of the residual cost function from the 50-km inversion. It is a purely self-consistency check, and therefore, does not rely on any ECMWF fields.

The assimilation of QuikSCAT data has a positive impact on forecast performance, especially on the southern hemisphere. The benefit of QuikSCAT on ECMWF's analyses and forecasts will be discussed.

On average, the analysis of tropical cyclones has been improved. However, this is not true for all cases. Too much good quality data around tropical cyclones is rain flagged, leaving only moderate wind speedsfar from the centre. Experiments using different rain-contamination QC schemes retaining more data in the vicinity of tropical cyclones will be presented. In addition, the impact of using only one wind solution in a purely quadratic cost function required by a new 4D-Var minimization method will be discussed.