Estimation of Representative Vertical Profiles of Reflectivity in the Regions Between Radar Sites

Precipitation in cold climates, like that in Finland, is quite shallow, in winter frequently only 2-5 km high. In heavy snowfall a vertical reflectivity profile gradient of -10 dB/km is typical immediately above the ground level. It has been found that an optimal Z-R and Ze-S relationship is hardly discernible factor in the accuracy of operational measurements of precipitation at ranges of 0-250 km from a radar. The large bias of 2-15 dB, often observed in gauge- radar comparisons, is fully originated from the effects of the vertical profile of reflectivity (VPR). Thus, we should first perform a vertical reflectivity profile correction and only after that search for possible improvement by applying optimal Z-R and Ze-S relations. At FMI vertical reflectivity profile corrections have been tested for several years but the main problem has been reliable estimation of the VPR at each measurement bin, especially at longer ranges (100-250 km from a radar). We have recently introduced a real time estimation of the 2D field of the VPR and respective correction of radar-estimated surface precipitation in a network of 7 C-band Doppler radars.

Large areas of widespread frontal precipitation can be overhanging i.e. hydrometeors will not reach the ground although precipitation is present in the lowest elevation PPI at longer ranges. Such occasions can be fatal for any correction based on an observed VPR close to a radar as the correction factor tends to be much larger than 1 at longer ranges where the proper correction factor should be 0, i.e. elimination of overhanging precipitation. The first step in the VPR-based correction of surface precipitation at FMI is the diagnosis and rejection of the regions of overhanging precipitation. The diagnosis is based on the measured VPR close to each radar (range 40 km), on the shape of the reflectivity pattern on PPI, and on the comparison of reflectivity patterns in the overlapping areas of neighbouring radars.

The correction factor due to VPR, not equal to 0, is estimated at each composite image pixel as a weighted average of the following components: 1. Time average of the individual VPRs derived from the 3D volume scans close to each radar at 15 minute time intervals. 2. Climatologically shaped VPR adjusted by the actual height of the melting level, obtained from the temperature fields of a high resolution NWP model. 3. The shape of the radar derived, non-corrected accumulated precipitation (period 1-24 h) as a function of range. It is assumed that in a widespread precipitation correct accumulation should be constant as a function of range. The relative weight of each component 1-3 is derived by representativity tests and the validation of the correction method is based on gauge-radar comparisons.