Assimilation of the radar reflectivities in the Arome model at Météo-France

These last years, Météo-France has developed a Numerical Weather Prediction (NWP) system at convective scale that has been running operationally since the 18th December 2008. This system, called AROME, covers the French territory with a 2.5 km horizontal resolution. Its main goals are to improve the local meteorological forecasts of potentially dangerous convective events (storms, unexpected floods, wind bursts, etc.) and lower tropospheric phenomena (wind, temperature, visiblity, etc.). Volumes of radar velocities and reflectivities observed by the National Doppler radar network play a key role by providing information about the horizontal wind circulation and the 3D precipitation patterns within precipitating systems over a wide part of France with high horizontal and temporal resolutions. It will be remembered and shown that, when some favourable sampling conditions are present, short term forecasts of precipitations are improved thanks a more realistic analysis of convergence structures in the boundary layer. Doppler winds of sixteen radars have been included in the first operational AROME NWP system.

But a important evolution of the system will be the assimilation of radar reflectivities, which is routinely evaluated since the end of 2008. The direct observation operator (needed to simulate the model counterpart of the observation at the same location) requires complete warm and cold microphysical parameterization which considers nonlinear moit processes and thresholds (convection regimes and saturation in particular). To avoid problems in the minimization algorithm, an original "1D+3DVar method" to assimilate these radar reflectivities has been introduced. We'll describe in details the results of the 1D algorithm. It consists of a Baysian statistic inversion which allows to retrieve relative humidity profiles from the observed columns of reflectivities, the different hydrometeor types being not analysed. For this, the model state in the vicinity of the observation is used as source of information, in order to constrain the solution. The impact of this assimilation on 3DVar analysis and on short term forecasts will be detailed. The method has proved the capability to create proper humidity and wind increments to adjust the model reflectivities towards the observations, even if there is no rain (at the same place) in the model backgrounds fields. It has been shown that the assimilation of "no-rain" signal was very useful to dry and shift the misplaced precipitation patterns. It also avoids the excessive spreading of positive increments of humidity. Sensitivities studies on analysis of the detection threshold of each radar (linked to signal-noise-ratio) have allowed to define the best careful using of the "no-rain" signal. An evaluation of several months of these cycled assimilations of radar reflectivities will be shown.