Autocorrelation and Cross-Correlation Estimators of Polarimetric Variables

Abstract

Herein are proposed novel estimators of differential reflectivity ZDR and correlation coefficient ?hv between horizontally and vertically polarized echoes. The estimators use autocorrelations and cross correlations of the returned signals to avoid bias by omnipresent but varying white noise. These estimators are considered for implementation on the future polarimetric Weather Surveillance Radar-1988 Doppler (WSR-88D) network. On the current network the reflectivity factor is measured at signal-to-noise ratios (SNRs) as low as 2 dB and the same threshold is expected to hold for the polarimetric variables. At such low SNR and all the way up to SNR = 15 dB, the conventional estimators of differential reflectivity and the copolar correlation coefficient are prone to errors due to uncertainties in noise levels caused by instability of radar devices, thermal radiations of precipitation and the ground, and wideband radiation of electrically active clouds. Noise variations at SNR less than 15 dB can bias the estimates beyond apparatus accuracy. For brevity the authors refer to the estimators of ZDR and ?hv free from noise bias as the ?1-lag estimators? because these are derived from 1-lag correlations. The estimators are quite robust and the only weak assumption for validity is that spectral widths of signals from vertically and horizontally polarized returns are equal. This assumption is verified on radar data. Radar observations demonstrate the validity of these estimator and lower sensitivity to interference signals than the conventional algorithms.