Correction for clipping of Doppler spectra from clouds and other atmospheric targets

Computing a Doppler spectrum usually increases the radar signal to noise ratio (SNR) considerably. The noise is distributed over all Doppler cells and the signal from atmospheric targets is concentrated in a limited interval. Storing the entire Doppler spectrum is often not done to reduce data storage, instead the moments of the Doppler spectrum are computed: total power, average velocity and Doppler width. One way to prevent the signal to be drowned in noise again, when calculating the moments, is to apply clipping or thresholding. Clipping simply means ignoring Doppler cells that have a power close to the noise level per Doppler cell. A picture showing the idea of clipping is presented. Clipping is very effective in removing noise, however a drawback is that also some signal is removed. If the clip level is close to the peak power in some Doppler cell, most of the signal power will even be removed. The width of the Doppler spectra is even stronger affected. If the distribution is not symmetrical, also the average velocity will change due to clipping. The effect of clipping is studied on radar cloud measurements made with the Delft Atmospheric Research Radar, Darr. This is a sensitive 9-cm radar with a high Doppler resolution of 3.5 cm/s. Normally 256 Doppler cells make up one Doppler spectrum. This spectrum is clipped at 7 dB above the noise per Doppler cell. The Doppler width (standard deviation) for the clouds in this study is between 18 and 30 cm/s. When an accurate quantitative value for the signal power is needed the calculated average power must be corrected for the effect of clipping. To derive a function for the correction of the measured total power, the data processing of a Doppler radar has been simulated using a known signal.