Impact of radar data assimilation using WRF three-dimensional variational system, for the simulation of a heavy rainfall case in Central Italy

Abstract. This work is a first assessment of the role of Doppler Weather radar (DWR) data in a mesoscale model for the prediction of a heavy rainfall. The study analyzes the event occurred during 19–22 May 2008 in the urban area of Rome. The impact of the radar reflectivity and radial velocity acquired from Monte Midia Doppler radar, on the assimilation into the Weather Research Forecasting (WRF) model version 3.2, is discussed. The goal is to improve the WRF high resolution initial condition by assimilating DWR data and using ECMWF analyses as First Guess thus improving the forecast of surface rainfall. Several experiments are performed using different set of Initial Conditions (ECMWF analyses and warm start or cycling) and a different assimilation strategy (3 h-data assimilation cycle). In addition, 3DVAR (three-dimensional variational) sensitivity tests to outer loops are performed for each of the previous experiment to include the non-linearity in the observation operators. In order to identify the best ICs, statistical indicators such as forecast accuracy, frequency bias, false alarm rate and equitable threat score for the accumulated precipitation are used. The results show that the assimilation of DWR data has a positive impact on the prediction of the heavy rainfall of this event, both assimilating reflectivity and radial velocity, together with conventional observations. Finally, warm start results in more accurate experiments as well as the outer loops strategy.