ERAD 2010 - THE SIXTH EUROPEAN CONFERENCE ON RADAR IN METEOROLOGY AND HYDROLOGY Assimilation of extrapolated radar reflectivity into a NWP model and its impact on a very-short range precipitation forecast Zbynĕk Sokol, Petr Pešice, Petr Zacharov Institute of Atmospheric Physics, AS CR, Prague, Czech Republic, sokol@ufa.cas.cz 1. Introduction Severe convective storms occur frequently in central Europe during summer. Most of these storms produce high rainfall rates, but their impact on the environment is negligible. Some storms, however, are accompanied by torrential rains that can accidentally cause local flash floods, which significantly impact inhabitants. It is very difficult to forecast the development of such events and their locations. Recent high-resolution NWP models are supposed to be able to directly simulate at least the larger-scale elements of organized convection without using convective parameterization. The NWP models must include assimilation of the latest data, especially those data that contain information on convective processes in the atmosphere (e.g., radar, satellite, and lightning observations), to have a chance to forecast rapidly developing convective storms. The data assimilation not only adds the current data into the NWP model, but should also initialize convective-scale events. Number of papers from the last decade show that the assimilation of radar data (reflectivity or derived Doppler velocities) into a NWP model can significantly improve precipitation forecasts for the next several hours (e.g., Macpherson, 2001; Tong and Xue, 2005; Leuenberger and Rosa, 2007; Milan et al., 2008; Stephan et al., 2008; Dixon et al., 2009; Sokol, 2009; Sugimoto et al., 2009; Zhang, 2009). Although the positive effect of the assimilation usually disappears within a few hours, the improvement is useful for nowcasting of precipitation. Forecasts of such models, however, are time-consuming; therefore, they are not commonly used operationally for forecasts with lead times of one to three hours. However, with increasing computer power, one can expect that such models will substitute currently employed nowcasting models in the foreseeable future. This paper deals with the nowcasting of convective precipitation, and it proposes a method assimilating observed and extrapolated radar reflectivity. 2. NWP model The COSMO NWP model, version 4.8, was applied. This is a non-hydrostatic, compressible model formulated in advection form. The numerical solution uses a two-level integration scheme based on the Runge–Kutta method (Doms and Schaettler, 2002; Steppeler et al., 2003). The model was integrated in the area shown in Fig. 1 with a horizontal resolution of 2.8 km and with 50 vertical levels. The model time step was 30 s. The model was run without the parameterization of deep convection. The parameterization of shallow convection was included. The precipitation processes made use of five classes of hydrometeors (rain water, cloud water, snow, ice, and graupel). The initial and lateral boundary conditions were derived from the prognostic fields of the COSMO-EU model, which started integration at 0000 UTC. This model is operated by the German Weather Service. The horizontal resolution of the COSMO-EU model is about 7 km. The boundary data were obtained by the linear interpolation of prognostic data, which were available at each hour of the integration time. 3. Assimilated data and assimilation method Radar reflectivity fields at 2 km above sea level (CAPPI 2 km), which are the standard operational products of the Czech Hydrometeorological Institute (CHMI), were assimilated into the model. The CHMI operates two C-band radars at the Skalky and Brdy sites (Novak, 2007; Fig. 1). The measurements are available with a horizontal resolution of 1 km by 1 km with a temporal resolution of 10 min. The measured data were checked and controlled to minimise ground clutter and anomalous propagation artefacts. Ground clutters were removed using the Doppler filter, and vertical profiles of reflectivity were corrected using the method described by Novak and Kracmar (2002). Radar composites were composed from data of both radars, with the maximum value applied to overlapping coverage areas. Optimal locations of radar sites with respect to topography ensure that there is no significant terrain blockage of radar echo in the reflectivity products. The elevation of the lowest radar beam is less than 1500 m for pixels in the Czech Republic (CR). The quality of the CHMI radar data is comparable to other data from European radar