Assessment of Censoring Using Coherency-Based Detectors on Dual-Polarized Weather Radar

In Doppler weather radars, signals may exhibit coherency in sample time, whereas noise does not. Additionally, in dual-polarized radars, samples of precipitation echo obtained in the two orthogonally polarized channels are substantially more correlated than samples of noise. Therefore, estimates of auto- and cross correlations can be used individually, collectively, and/or with power measurements to enhance detection of precipitation signals, compared to the approach that uses only power estimates from one channel. A possible advantage of using only estimates of coherency for signal detection is that the detector’s performance is less sensitive to errors in noise power measurements. Hence, censoring is more likely to produce desired false alarm rates even if nonnegligible uncertainties are present in the noise power estimates. In this work these aspects are considered using real data from weather radars. Three novel censoring approaches are evaluated and compared to the censoring approach that uses only estimates of signal and noise powers. The first approach uses only cross-correlation measurements, and the second approach combines these with the lag-1 autocorrelation estimates. The third approach utilizes all estimates as in the previous two approaches in combination with power measurements from the horizontal and the vertical channels. Herein, it is shown that, when more accurate measurements of noise powers are available, the third approach produces the highest detection rates followed by the second and the first approaches. Also, it is corroborated that the first and the second approaches exhibit less sensitivity to inaccurate system noise power measurements than the third one.