Another method for ground target identification and filtering using spectral processing of dual-polarization returns

To eliminate ground targets, one often filters near-zero velocity echoes using time-domain or frequency-domain filters. At McGill, because our S-band radar scans rapidly, we would have to filter everything between +/- 5 m/s to eliminate ground targets, and that would eat too much of the weather echoes. Furthermore, new uses of radar data such as data assimilation are extremely sensitive to contamination of any kind and to improper corrections.

Here, we extend the work of Moisseev and Chandrasekar (2009) by identifying clutter, noise, and precipitation returns at the spectral component level: We decompose the radar signal into its Fourier components, and determine which of these are due to noise, non-meteorological returns, or meteorological returns. This is achieved by combining using fuzzy logic whether these Fourier components have characteristics matching each of these three types of echoes, parameters being evaluated including signal power, differential phase and reflectivity, and the standard deviation of the ratio of the complex H and V return with their spectral neighbors. Once this is done, spectral components identified as clutter are replaced by values expected for weather assuming that the Fourier spectrum follows a Gaussian function, unless contamination is too great, in which case the data are being flagged as unrecoverable. Though occasionally, weak precipitation spectrum components are wrongly identified as clutter, the approach seems to be working remarkably well given the limits imposed by our fast antenna scanning speed.