It is a customary practice at NCEP to evaluate the timeliness of availability and the quality of every new data set before using them in operational models. Such an effort is particularly needed with Quikscat because the design of its antenna system is different from the ones used on the previous scatterometers. It uses two antennas with satellite look angles of 40 and 46 degrees to scan the ocean surface. The radar design suggests that there are three regions of the swath where the accuracy may deteriorate; two regions are located at the outer 200km edges of the swath where only one scanning antenna can retrieve measurements, and the third region is near nadir where the viewing angles are small, suggesting that wind direction may be less accurate.
It was expected that the backscatter in the Ku-band will be affected by atmospheric attenuation due to rain and cloud liquid water as well as being subjected to distortions at the ocean surface due to rain. In addition, the wind retrievals suffer from the well-known directional ambiguity problem because the inversion process from backscatter measurement to wind vector is not unique and may provide up to 4 vector solutions which are ranked in order of their Most Likelihood Estimate (MLE). But the MLE may not be adequate for the best wind selection. Hence, one of the most common practices to reduce errors in the selection of the best wind vector has been to use an independent ocean surface wind field as background information to re-rank the solutions. The closest solution in this case is called the nudged solution. NESDIS, which provides the Quikscat data to NCEP, uses the 6-hour forecast from the NCEP's operational global Aviation model to provide this background wind field. A median filter is also applied at each Quikscat cell to provide for further consistency in the wind vector retrievals over the swath.
This paper discusses the results of the evaluations carried out to determine the quality of QuikSCAT data. Two types of evaluations have been carried out; the first one was performed using collocated fixed buoy wind data with QuikSCAT wind retrievals and the second one was performed using ocean surface model generated wind field analyses matched with QuikSCAT swath data. The latter evaluation allows for the sensitivity of relative error due to cell location.
Based on these evaluation, the following conclusions can be made: The QuikSCAT ocean surface wind vector data meet the accuracy specifications (speed - rms < 2 m/s, direction - rms < 20 degrees) only if retrievals are eliminated for cells along the outer 200km edges of the swath and for cells with a rain probability of greater than 10%.
But, nudging and filtering force the retrieved winds to be too similar to the model background winds, thus removing independent information content from the satellite retrievals. Because rain contamination often occurs in wind retrievals in weather active regions (storms and fronts), reliable retrievals are often not available where they are needed most.
Supplementary URL: