by Claudia Acquistapace, Robert Banks, Gordon Diefenthal, Marvin Forde, and Francesco De Angelis

The aim of the group project was to evaluate the accuracy of precipitation estimation at the ground during a severe storm event, which happened on 09.06.2014 (Pentecost storm). A local relation between reflectivity (Z) and rain rate (R) was derived for the JOYCE site using independent observations of rain rates and Z values from a Parsivel disdrometer located at the site. It is the aim of this derivation to provide independent observations of rain rates from the reflectivities of a 3D radar composite by the University of Bonn. Various Z-R relations were developed: Marshall-Palmer, local, convective, and two relations using the polarimetric moments, Z_DR and K_DP.

Figure 1: Estimation of rain rates from the radar composite: Marshall-Palmer, local, convective, and two relations using the polarimetric moments, Z_DR and K_DP have been used. Results are averaged over the complete storm window of approximately 3 hr and evaluated against independent rain gauges.

Fig. 1 shows the performance of the different average Z-R relations over the complete storm window of approximately 3 hr as evaluated by independent rain gauge observations. The parameterizations behave differently and among them Marshall-Palmer seems to be the one that performs the best for these atmospheric conditions. It has to be concluded that a reliable local Z-R relationship is difficult to derive and depends on a variety of situation-specific factors; most of them are not covered by the disdrometer measurement.

Furthermore, different accumulation times ranging from 30 minutes up to 3 hours have been compared. The results of the local Z-R relationship varied greatly throughout different 30-minute time windows during the storm. This is due to a wider spread of measurement points and a high sensitivity to local outliers. As expected, the amount of outliers decreases with increasing accumulation time. Further, the uncertainty of the fit can be reduced by longer accumulation times. As a result, an accumulation time of at least 1 hour is suggested.