For doppler velocity data quality control

1. INTRODUCTION * It has been increasingly well recognized in recent years that radar data quality control is critical for radar data assimilation. To support the operational radar data assimilation at the National Centers of Environmental Prediction (NCEP), continued efforts have been made in developing efficient quality control techniques for Doppler radial-velocity observations from operational WSR-88D radars (Gong et al. 2003; Liu et al. 2005; Zhang et al. 2005; Xu et al. 2009a,c). The earlier developed techniques were combined into an automated quality control package for pre-operational tests with real-time KTLX radar observations at NSSL (Liu et al. 2003; Xu et al. 2005) and then operational tests at NCEP (Liu et al. 2009). During the tests, some particularly difficult cases were encountered in which the three-step dealiasing method (Gong et al. 2003) failed to detect or correct severely aliased radial-velocities. Having diagnosed each failure in details, the following two reasons have been identified. (i) The modified velocity azimuth display (VAD) technique (Tabary et al. 2001) adopted in the first step is not sufficiently robust to deal with severely aliased radial velocities, so the refined reference radial velocities produced by the traditional VAD method in the second step cannot always ensure the seed data qualified at the end of the second step to be absolutely free of alias errors, and the errors can propagate via the point-to-point continuity check in the third step. (ii) The point-to-point continuity check in the third step relies on qualified seed data, and it cannot pass a non-qualified date area (including data holes) that has 5 or more gates (2.5 km) along the previous and current beams from the flagged data point that is being checked. Besides, the point-to-point continuity check may also occasionally produce false dealiasing and propagate the error into neighborhood areas. To eliminate the above sources of dealiasing failure, an alias-robust VAD algorithm was developed (Xu et al. 2009b) and is used to replace the modified VAD method for the preliminary reference check in