One problem of the JPL QC approach is that some of the good quality QuikSCAT wind data over weather producing storm areas may be discarded. However, over the weather active areas where rain occurs, ocean surface winds would be important. To address this deficiency, a new quality control procedure based on analyses of maximum likelihood estimates has recently been developed by Portebella and Stoffelen (2002), which is capable of differentiating good from bad quality data, especially over the rainy storm areas. In an attempt to test this new quality control procedure for assimilating the QuikSCAT wind data at NCEP, a global data assimilation experiment was recently conducted, and preliminary results show that applying this new quality control procedure to the data in the assimilation leads to a better improvement in the short range tropical wind forecasts as compared to those from the control experiment which uses JPL QC'ed QuikSCAT wind data. Results of these assimilation and forecast experiments will be presented.
In October, 2002, a higher resolution global forecast model (T254, L64) with improved model physics in convection was implemented. After this new implementation, a series of assimilation experiments have been conducted to test effective scales of superobed resolutions associated with QuikSCAT wind data. In particular, a data assimilation experiment was conducted to test higher resolution (half degree longitude-latitude superobed) QuikSCAT winds, and results showed a further improvement in the forecasts of wind and mass fields in the tropics. Hence, the half degree longitude-latitude superobed QuikSCAT wind data were implemented at NCEP GDAS on March 11, 2003. Most recently, in anticipation to the arrival of ADEO-II SeaWind data (12 km spatial resolution), another data assimilation experiment was conducted using the full resolution (25 km) QuikSCAT wind data at NCEP. Results of these assimilation and forecast experiments designed to test effective scales of spatial resolutions associated with QuikSCAT wind data will also be discussed.