Data assimilation of high-density observations. II: Impact on the forecast of the precipitation for the MAP/SOP IOP2b

The impact of the data assimilation of high-density (space and time) data on the precipitation forecast is evaluated by improving the initial conditions of a mesoscale model. The high-frequency data allow for improving the three-hourly initial and boundary conditions as well. The data assimilation is performed using initial objective analysis (Cressman and multiquadric schemes) and 3D-Var. The MM5 (version 3) mesoscale model from Penn State University/National Center for Atmospheric Research is used to evaluate the impact of the improved initial and boundary conditions on the model simulations. The comparison of model results with observations shows: (i) the forecast of the precipitation at high resolution produces better results than those without data assimilation only if thrce-hourly data are assimilated by multiquadric; (ii) the mean error of the model rainfall largely decreases only if 3D-Var is used, but no comparable improvement in the spatial distribution of the precipitation is found; (iii) the improvement for the rainfall is not as good as it is for the initial conditions for all experiments. Moreover, the observations ingested by objective analysis modify both the amount and the timing of the precipitation on the Po valley. On the other hand, 3D-Var modifies only the amount of the precipitation, but both techniques barely recover large-model failure.