Evaluation of the Direct and Indirect Assimilation of Radar Reflectivity using the WRFDA 3D-Var

It is known that developments have being made in the concern of radar data assimilation and many studies point out some improvements on the quantitative precipitation forecast. However, it is not clear whether the most common way of assimilating reflectivity, i.e., directly as a control variable, is the best choice. We have made experiments using the Weather Research Forecasting model three-dimensional data assimilation system (WRFDA 3D-Var) over Brazil and we have found that the improvement achieved when using the direct assimilation is short-lived. Besides that, there are some works which point out this problem and suggest that the best way of assimilating reflectivity is indirectly through the assimilation of rainwater mixing ratio obtained from reflectivity. It would avoid problems related to the linearization of the reflectivity-rainwater mixing ratio relationship (Z-qr) that is needed in the incremental formulation used in WRFDA 3D-Var. Therefore, the aim of this work was to perform experiments over a specific region in Brazil to evaluate whether the indirect assimilation of reflectivity outperform the results obtained using the direct assimilation. It was chosen six cases of precipitation and 3 experiments for each case were performed. The three experiments were: i) a control without any radar data assimilation; ii) using direct assimilation and; iii) indirect assimilation of reflectivity. Radial velocity was assimilated in both cases. The Fractional Skill Score (FSS) and the Root Mean Square Error (RMSE) were used to compare quantitatively the performance of each experiment against observations. The results have shown that the indirect assimilation can produce better QPF than the one where reflectivity is assimilated directly and the improvement tends to last longer.