Following the devastating Fort Collins flash flood in 1997, an algorithm was developed at Colorado State University (CSU) to estimate rainfall from CSU-CHILL dual-polarization radar data. The algorithm first attempts to remove ground clutter using thresholds on the correlation coefficient r_HV and standard deviation of the total differential phase Y_DP. Specific differential phase K_DP is then calculated from Y_DP. The radar data are interpolated to a Cartesian grid and rainfall estimates at each grid point are determined using an optimization procedure. The procedure picks the "best" estimate of rainfall based on measurement thresholds of K_DP, Z_DR, and Z_H.

During the summer of 2002, a UCAR-COMET grant provided an opportunity to apply the polarimetric algorithm in a real-time environment and quantitatively evaluate the performance in comparison to the standard NWS Z-R technique on selected rainfall events in northeast Colorado. Maps of warm season accumulated rainfall are now generated routinely and are available at the CSU-CHILL web site. Validation of the rainfall algorithms was made using 24-hour accumulated precipitation data from the Community Collaborative Rain and Hail Study (CoCoRaHS), which includes hundreds of volunteers across northeast Colorado. A total of eight events were analyzed from the summer 2002 data set. The events ranged in duration from 3 to nearly 9 hours and the number of rain gauge reports for each event ranged from 32 to 259. Comparisons using standard error and bias statistics showed that, on average, the polarimetric technique performed modestly better than the Z-R algorithm; however, the performance varied significantly from one event to another depending on the presence of hail. The presentation will discuss methods to improve the polarimetric algorithm.