Performance of the air quality forecasting simulations for the summer 2006 during the Second Texas Air Quality Study (TexAQS-II) suffered due to the bad meteorological predictions, serious emissions upset events, or inaccurate specification of initial and/or boundary conditions. We have preformed 12 different meteorological simulations, i.e., a set of ensemble with 12 members, utilizing various multi-stage nested four-dimensional data assimilation (MS-FDDA) methods and MM5 model options. The MS-FDDA is repeatedly applied at multiple nest domains to minimize the prediction errors at the target domains. To prepare meteorological input data for the interpretation of diverse atmospheric measurements and for the air quality assessment study, the best meteorological simulation is selected out of the ensemble simulations for each day. In addition to the statistical measures such as RMSE and IOA, the stagnation potential (SP) of the backward trajectories and synoptic scale flow patterns are utilized to rank the performance of meteorological simulations. The scheme may need to be refined and evaluated further but the preliminary results show the best of ensemble meteorological inputs lead to best air quality simulations subsequently. With the accurate meteorological inputs, uncertainties in the emissions inputs and model initial and boundary conditions are studied.