A Variational Correction Method as an Alternative to Forced Rejection of Sidelobe-Contaminated Bistatic Doppler Measurements

Abstract

Abstract The problem of sidelobe contamination of bistatic apparent Doppler velocity measurements involved in a bistatic Doppler radar network is examined. So far in the context of 3D wind field analysis, by combining a traditional Doppler radar with one or more bistatic receivers, identification and hence removal of regions of high degrees of contamination were necessarily crucial steps to obtaining reliable wind fields. This study proposes an alternative solution to the forced rejection of bistatic Doppler data suspected to be contaminated by sidelobe echoes, on the basis of restoring the nonmeasured “actual” (i.e., noncontaminated) bistatic Doppler velocity from both monostatic radar and bistatic receiver measurements. The correction method is based on a modeled expression of the observed bistatic apparent Doppler velocity defined as the reflectivity-weighted average of actual Doppler velocity of particles within individual volume samples, including the antenna gain pattern of both transmitting and receiving radars. The searched actual Doppler velocity is a solution of an underdetermined inverse problem that can be handled as a constrained linear inversion problem, through a variational least squares analysis method. The performances of the proposed method are analyzed, using simulated radar observations involving one remote receiver. An example of application to experimental data collected by the Deutsches Zentrum für Luft und Raumfahrt (DLR) bistatic Doppler radar network within a moderate precipitation system observed on 8 May 2000 in Germany is also presented. Pseudo-Doppler observations of a tropical squall-line system are used to quantify the effective improvement of the correction method on the bistatic Doppler velocity and hence the retrieved 3D wind field. Statistics of the differences are presented between observed and idealized (sidelobe free) velocity structures on the one hand, and corrected and idealized velocity structures on the other hand. Clearly shown is the very low level of the corrected minus idealized differences (mean and standard deviation) against the significantly high level of the observed minus idealized differences. As previously observed, maximum correction occurs in regions of potentially high gradients of reflectivity. It is also found that regions of low observed minus idealized differences remain unchanged after correction, which means that the sidelobe-correction method only acts on needed regions and does not introduce any artificial modification.