Optimization of Nonlinear VAD Method in the Low-level Wind Retrieval

The performance of nonlinear velocity azimuth display method in the vertical wind profile retrieval at low levels (below 2 km) is quantitatively examined by combing the theoretical analysis and cases observed by SoWMEX S-Pol radar and Yangjiang radar in Guangdong Province. Results show that the general structure and evolution of the low-level wind profile can be reasonably deduced by traditional nonlinear VAD method. The root mean square error can be used to evaluate orders of velocity azimuth display (VAD) fitting, but small error does not always mean the better performance especially with big continuous data absence, and a specific example is given. When setting the VAD fitting order to 3 instead of 2, coefficients which represent the horizontal wind u and v are closer to the wind derived from radial velocity image. However, when the fitting order comes to 4, coefficients lost their physical meaning. The wind direction differs a lot and the speed is much smaller than the value before. At the same time, the root mean square error decreases compared with the order of 3. Besides, data used in nonlinear VAD fitting come from the whole volume, which decreases quite a lot and leads to nonlinear VAD fitting error when the volume coverage pattern (VCP) only has some lower elevations (e.g., two elevations). Therefore, the retrieved wind could contain large error in certain situations, such as for a region with large continuous data absence or a volume scan with fewer elevations.After carefully evaluating the impact of the corresponding parameters on the nonlinear VAD retrievals by analyzing radar measurements, a modified nonlinear VAD method is proposed which takes account of the maximum fitting order in horizontal (VAD) and vertical adaptively according to the size of continuous data absence and the number of sweeps in a volume scan. VAD fitting is abandoned when the data absence is larger than 90°; the order is set to 3 when the data absence is between 60° and 90°; and the order is set to 4 when the data absence is smaller than 60°. The order of nonlinear VAD fitting is reduced when the VCP only has low elevations. Apply the method in two cases: One is a front case passing through Taiwan, China, the other is a typhoon case landfall in Guangdong Province, with both of them having nonlinearity in the low level wind profile. The wind profile after adjusted can significantly improve the wind retrieval, as compared with the traditional nonlinear VAD. Both wind speed and direction from modified nonlinear VAD agree with those from sounding observations, with the root mean square of the wind less than 2 m·s^-1, which is obviously better than nonlinear VAD before adjusted.