Classification of hydrometeors and non-hydrometeors using polarimetric C-band radar

Several studies have shown successful retrievals of the microphysical properties of precipitating clouds using polarimetric radar. The algorithms used to automate this procedure are typically constructed within a fuzzy logic framework, where membership functions are produced for each polarimetric variable. Selection of the shape of these functions may be manual or arbitrary. The proposed classification scheme is based entirely on observations. First, expert analysis is used to identify several hydrometeors (e.g., rain, hail, wet snow, dry snow, etc.) and non-hydrometeors (e.g., chaff, ground clutter, clear air echoes, etc.). Radar range, azimuth, and height of measurement are used to isolate the domain of the phenomena of interest. Unconditional probability distribution functions (pdfs) are then computed for each polarimetric variable, their textures, and temperature. Ultimately, these pdfs may serve as membership functions for microphysical identification algorithms. Initial results show good separation between pdfs describing different hydrometeors, and issues concerning radar data quality will be almost entirely resolved.