Comparison of radar-derived precipitation and rain gage precipitation in northeastern Colorado

Abstract

Accurate precipitation measurement is desired over large areal extents in fine temporal and spatial resolution for a myriad of scientific disciplines and practical applications. Hydrological sciences and federal and local government agencies would benefit from improved precipitation measurements. The question is can radars satisfy this desire for better precipitation measurements. The WSR-88D radar network will provide nearly complete radar coverage of the contiguous United States and has the ability to operationally measure large areal extents in fine temporal and spatial resolutions. Precipitation products derived from the WSR-88D networks are becoming readily more accessible and steadily gaining in popularity and use, often without any reference to accuracy. This study is a comparison of precipitation from the CSU-CHILL multiparameter research radar, National Weather Service's WSR-88D located outside Denver, CO (KFTG), and networks of tipping bucket gages. Comparisons are made to reveal spatial coverage of precipitation, time distribution of precipitation, and quantify amounts of precipitation derived from the two radars and gage networks from three convective precipitation events in northeastern Colorado. This study finds the multiparameter variable, specific differential phase derived precipitation (R(KDP)) compared well with gage precipitation for rainfall accumulations greater than 1 cm. On 20 June 1994 for 12 gages with four-hour accumulated precipitation greater than 1 cm, the R(KDP) to gage precipitation ratio was 0.89. On 21 June 1994 for 3 gages with one-hour accumulated precipitation greater than 1 cm, the R(KDP) to gage ratio was 1.37. For precipitation accumulations less than 1 cm, R(KDP) greatly overestimated gage precipitation which is consistent with previous findings. On 20 June at one gage site (FOR) with a known 30- minute period of mixed phase precipitation, R(KDP) showed an eight percent overestimate of gage precipitation. This result demonstrates R(KDP)' s ability to accurately measure rainfall in mixed phase precipitation. The time distribution of precipitation rates for the radar and gage are in reasonable agreement. In most cases, the radar-derived precipitation captures the temporal pattern of the gage's precipitation event well. However, the amplitude of the precipitation amounts differed appreciably. Of the three reflectivity cut offs used to minimize excessive rain rates, the 53 dBZ reflectivity cut off performed the best in comparison to the gage's peak precipitation. This result agrees with Denver WSR-88D's use of 53 dBZ as their reflectivity cut off for summer convective precipitation in northeastern Colorado. The precipitation derived by the actual WSR-88D precipitation algorithm provided by the WSR-88D Operational Support Facility, R(OSF), consistently overestimated the gage precipitation for two days of convective storms in Colorado. On 21 June 1994 for 64 gages with one-hour accumulated precipitation, the R(OSF) to gage precipitation ratio was 2.13. On 10 August 1994 for 127 gages with two-hour accumulated precipitation, the R(OSF) to gage ratio was 1.80. Over both days, the R(OSF) provided a somewhat smaller overestimate of gage 111 precipitation than this study's elementary use of same level II data and Z-R relationship. This result indicates the details of data processing are extremely important.