Previous studies of hydrometeor classification using dual-polarization radar have usually used research data sets with high vertical resolution and have utilized many different variables in determining the classification. Automated hydrometeor classification is a valuable tool for both research and operational meteorologists. This study examines the effects on the resulting classification of several practical constraints usually associated with operational radars versus research radars: fewer radar variables, coarser vertical resolution, and greater uncertainty in absolute calibration.

A simplified hydrometeor classification program is run using the reflectivity (Z) and differential reflectivity (ZDR) fields from volume scans as input. The hydrometeor types identified are placed in five categories: rain, snow, super-cooled rain, graupel, and hail. Six storm cases observed by the Kwajalein K-POL radar from June 2001 to November 2001 were selected from the TRMM Ground Validation data set. These cases were selected based on the percentage area of radar echo above 20 dBZ in the 1 degree elevation scan. Approximately 99 hours of data in 484 scan volumes were processed, with individual storm cases ranging from 5 hours to 22 hours in duration. Additional volumes obtained by the CSU CHILL and NCAR S-POL research radars were also examined. The polar-coordinate volume scans were interpolated to Cartesian grid with 2 km x 2 km horizontal resolution and 1 km vertical resolution out to a horizontal range of 80 km.

Results indicate that operational classification may be practical, especially in distinguishing regions of graupel and hail from rain and snow. Since supercooled rain typically occurs in shallow layers, it is often difficult to discriminate in coarse resolution data. The algorithm results are very sensitive to the estimated height of the freezing level. Other sensitivities include the absolute calibrations of Z and in particular ZDR.