Optimal use of radar radial winds in the HARMONIE numerical weather prediction system

The use of radial velocity information from the European weather radar network is a challenging task, due to a rather heterogeneous radar network and the different ways of providing the Doppler velocity information. A preprocessing is therefore needed to harmonize the data. Radar observations consist of a very high resolution dataset which means that it is both demanding to process as well as that the inherent resolution is much higher than the model resolution. One way of reducing the amount of data is to create super observations (SO) by averaging observations in a predefined area. This paper describes the preprocessing necessary to use radar radial velocities in the data assimilation where the SO construction is included. The main focus is to optimize the use of radial velocities in the HARMONIE-AROME numerical weather model. Several experiments were run to find the best settings for first-guess check limits as well as a tuning of the observation error value. The optimal size of the SO and the corresponding thinning distance for radar radial velocities was also studied. It was found that the radial velocity information and the reflectivity from weather radars can be treated differently when it comes to the size of the SO and the thinning. A positive impact was found when adding the velocities together with the reflectivity using the same SO size and thinning distance, but the best results were found when the SO and thinning distance for the radial velocities are smaller compared to the corresponding values for reflectivity