Validation of QuikSCAT and ASCAT wind vectors by dropwindsonde data from DOTSTAR and T-PARC field experiments

The accuracy of 10-meter wind vectors derived from the QuikSCAT and ASCAT satellite near tropical cyclones is validated against soundings from the GPS dropwindsondes deployed by DOTSTAR (Dropwindsonde Observations for Typhoon Surveillance near the Taiwan Region) during 2003-2007 and T-PARC (THORPEX Pacific Asian Regional Campaign) in 2008. To maximize the database, the surface to 40-m wind speed in the dropwindsondes is averaged and interpolated to the 10-m wind speed by linear regression. For the result of the accuracy of QuikSCAT wind as comparing with the dropwindsondes data observed during 2003-2007, the root-mean-squared differences between QuikSCAT and dropwindsonde data were 2.6 m s-1 or 18% (wind speed) and 17¢ª (wind direction) based on 896 matching samples. Further analyses also indicate that the QuikSCAT data slightly underestimates the wind speed of medium wind regime (between 10 and 17.2 m s-1) and possesses some clockwise directional bias in the high wind regime (above 17.2 m s-1). In summary, this study suggests that the QuikSCAT wind vectors below tropical storm wind strength (17.2 m s-1) are accurate enough for forecasters to determine the critical wind radius of 34-kt wind, while a new error bound of the QuikSCAT wind estimate in high-wind regimes near tropical cyclones is suggested to be set at about 4 m s-1. Since the QuikSCAT has failed in operation in November 2009. There is still ocean surface vector wind data from the European ASCAT system which provides a partial mitigation for QuikSCAT. The ASCAT was launched in October 2006. The follow-up research will focus on the examination of the accuracy the ASCAT wind vector by the dropwindsondes deployed by both the DOTSTAR (2007-2009) and T-PARC (2008) field experiments. We anticipate that this work will provide a basic knowledge on the accuracy of satellite-derived wind products, especially in TCs, for use in forecaster analysis and for initializing numerical forecast models.