ERAD 2010 - THE SIXTH EUROPEAN CONFERENCE ON RADAR IN METEOROLOGY AND HYDROLOGY The use of radar data in the verification of high resolution prognostic rainfall fields Daniela Rezacova 1 , Petr Zacharov 1 , Zbynek Sokol 1 1 Institute of Atmospheric Physics AS CR, Prague, Czech Republic Abstract Merging radar data with surface rainfall measurements can produce QPE data sets suitable for verification of precipitation nowcasts and forecasts as the spatial resolution of present high-resolution NWP models is of the same order as radar-based QPE fields. The difference between the diagnostic radar- based QPE and the prognostic rainfall fields are especially pronounced if heavy local convective rainfalls are forecast. Assimilation of radar data can reduce the forecast error with the time limits of assimilation effect; the last depends on precipitation event and is difficult to estimate. In this paper we will summarize the results obtained by using radar-based QPE for a spatial verification of heavy convective precipitation nowcasts and forecasts and for the estimation of forecast uncertainty. Prognostic data from heavy convective events were obtained by using the COSMO model with 2.8 km horizontal resolution. Following verification related topics will be included in the paper: (1) radar-based traditional vs. spatial verification techniques, (2) the technique producing radar-based QPE and its effect on verification results, (3) the influence of radar data assimilation on time dependent verification results, (4) the evaluation of forecast uncertainty by using verification results and the ensemble approach. The results were obtained during the COST 731 collaboration. 1. Introduction Present NWP models with horizontal resolution 2-3 km are able to describe the physics of convective motions in clouds and the well parameterized cloud microphysics can simulate the convective precipitation origin. Nevertheless, the model resolution still doesn't correspond to the spatial resolution of convective processes of the order of 100-1000 m and the model generates convective circulations different from the reality. It obviously causes the dissimilarities in the precipitation fields. The precipitation fields generated by the model with high resolution correspond to spatial resolution of radar measurements and the model fields also optically resemble the radar fields. As the model doesn't reproduce the convective circulation well, the precipitation can't be forecasted correctly and the precipitation fields are incorrect in time and space. In the past the determination of several verification scores were the most common verification technique and the interpolated gauge measurements served as the verification data. These techniques are often marked as "traditional verification methods". They include the use of contingency table and the skill scores derived from the table or scalar error measures, like e.g. RMSE. The scores and error measures based on grid-to-grid comparison are useful for synoptic scale up to mezoscale. The application of the scores is questionable at the precipitation forecasts with horizontal resolution below 10 km. The development of new verification methods began about ten years ago and was motivated by development of NWP models with high temporal and spatial resolution, ensemble forecasting and data from radar and satellite measurements (e.g. Ebert and McBride, 2000; Casati et al, 2004; Davis et al, 2006; Casati et al, 2008; Ebert, 2008; Gileland et al, 2009). 2. Set-up The nonhydrostatic fully compressible NWP model COSMO has been developed at the German Weather Service (DWD) to meet high-resolution regional forecast requirements and to provide a flexible tool for various scientific applications on a broad range of FIG. 1. 3h precipitation totals for 15 July 2002 (top row) and 23 May 2005 (bottom row). The left column represents the forecast of the LM COSMO, the right column represents the gauge-adjusted radar observation. The totals correspond to the lead time 0014 (17-20 UTC). The rainfall values in mm/3 h are indicated in the legend.