Assimilation of radar dual-polarization observations in the AROME model

This article describes the results of assimilation experiments of dual-polarization (DPOL) radar observations in the AROME model with its 3D-Var assimilation system. The assimilation of dual-polarization observations is derived from the current operational assimilation method for radar reflectivities. It has two steps (1D+3D-Var). Relative humidity pseudo-observations are first retrieved from observed profiles of reflectivity (Z(hh)), with or without attenuation correction, from differential reflectivity (Z(dr)), from specific differential phase (K-dp), or from a combination of Z(hh), Z(dr) and K-dp, through a unidimensional (1D) Bayesian inversion. The retrievals are then assimilated in the AROME 3D-Var system. In the case of attenuation due to precipitation, the values of the relative humidity profiles retrieved with uncorrected reflectivity were found lower on average than with attenuation-corrected reflectivity or with K-dp. In regions with large values of K-dp, both methods using K-dp produced the profiles with the largest values of relative humidity. Assimilation experiments were conducted for two convective cases. They demonstrated the benefit of the assimilation of K-dp and of attenuation-corrected Zhh in the model analysis. The humidity increments were found higher in convective regions when assimilating either K-dp or attenuation-corrected Zhh. Furthermore, the impact of DPOL observations on precipitation forecasts was found slightly positive for the periods of intense convection.