Multiple peak processing algorithm for identification of atmospheric signals in Doppler radar wind profiler spectra

For most wind profilers, wind data is computed from the so-called moments of the strongest peak in the Doppler spectrum. This single peak moment estimation technique works well as long as contamination and noise signals are small with respect to the atmospheric signal. However when the atmospheric wind signal is obscured or dominated by other sources, the single peak moment-finding routine produces often erroneous wind data which may either be removed by subsequent quality checking algorithms or may remain as false data. A new peak processing algorithm that allows the simultaneous identification and characterization of several peaks in a Doppler spectrum is presented. The peak which most likely represents the wind signal is selected using a spatial, modal and temporal continuity scheme. The characteristics of this multiple peak processing method include automatic real-time processing, independent of profiler type and site location. It is based solely on the original spectral data and requires no interpolation or extrapolation, it can be applied operationally just as any of the existing standard algorithms. The multiple peak processing method was tested using a long-term data set. The resulting winds were compared with those obtained by using the single peak processing method. The newly developed algorithm leads to a significant increase in the availability of wind data. The improvement is about 9 % for the high mode and about 16 % for the low mode. (In the high mode, the wind profiler is operated with a long pulse width. Because a long pulse contains more power than a short one, echoes are obtained from a greater range [i. e. height, hence, high mode]. However, the price for this is a lower vertical resolution because the longer pulse occupies a larger spatial volume. Consequently, short pulses provide high vertical resolution within a reduced range [low mode] while long pulses cover a greater height at reduced vertical resolution). The results were also compared with radiosonde and with sodar data. For the low mode, the quality is about the same for single peak processing method and the multiple peak processing method. For the high mode, however, there is a significant increase in data quality when applying the multiple peak analysis.