As part of continuous modernization of the nationwide network of the NEXRAD weather radars, US National Weather Service has decided to add polarimetric capability to existing operational radars. Herein, we report on tests of the proof-of-concept which was developed on the NSSL's research WSR-88D radar; operational demonstration started in March 2002. An algorithm for classification of meteorological and non-meteorological radar echoes provides one of the key products that are delivered in real-time to the Norman NWS office for evaluation. In this paper, basic principles of the algorithm are outlined and results of classification for several cases are presented with special emphasis on rain/snow discrimination. Current real-time version of the algorithm enables discrimination between radar echoes caused by (1) ground clutter and anomalous propagation, (2) biological scatterers (including insects and birds), (3) dry snow, (4) wet snow, (5) stratiform rain, (6) convective rain, and (7) rain/hail mixture. The classification algorithm utilizes polarimetric radar data collected at two lowest elevation angles, 0.5 deg and 1.5 deg, to produce a field of classified scatterers at the elevation scan of 0.5 deg. These fields are regularly supplied to the NWS operational staff for evaluation and feedback. Special attention is given to the discrimination between rain and snow for cold season events. This is a real challenge because polarimetric contrasts between light rain and dry aggregated snow are quite small, and direct application of the fuzzy logic methodology does not provide satisfactory discrimination. Combining principles of fuzzy logic and pattern recognition enables much better rain / snow delineation than a fuzzy logic scheme alone. Such improvement will be illustrated for several cases of mixed precipitation observed with the R & D polarimetric WSR-88D radar during winter season in central Oklahoma.