This study investigated the advantages and limitations of three different methods for the evaluation of a PM2.5 pollution episode over Beijing-Heibei-Tianjin (BHT) region during 19-28 February 2014. The three methods are the Cressman spatial interpolation using observations, numerical simulation with a 3-dimensional chemical transport model, and data assimilation combining both observations and simulations. The main results are as follows. First, the Cressman spatial interpolation using only the sparse surface observations could not well reproduce the spatial-temporal variation of PM2.5 concentration during this episode. It tended to produce artificial high and low PM2.5 concentration centers over some areas where observations were missing. Second, the numerical simulation using a Nested Air Quality Prediction Modeling System (NAQPMS) could provide high spatial-temporal resolution details of PM2.5 concentration over the BHT region. However, this method was limited by large uncertainties in the simulation results and the root mean square error (RMSE) of the simulated surface PM2.5 concentration was higher than 100 μg/m³. Third, the data assimilation using optimal interpolation algorithm performed better than the above two methods in reproducing the evolution of the episode. It could reproduce the spatial pattern of surface PM2.5 concentration during this episode over the BHT region with high accuracy and high spatial and high temporal resolution, and the RMSE of the PM2.5 concentration was lower than the simulated data by about 50%. Finally, the assimilation results of surface PM2.5 were used to evaluate the evolution of the PM2.5 pollution episode. It was found that the PM2.5 concentration over southern Hebei province increased to a high level at the beginning of this episode. The polluted areas then extended to northern Hebei, Beijing, and Tianjin. On 27 February, surface winds over this region systematically turned to be northerlies with enhanced wind speed. Thereby the high concentration of PM2.5 decreased rapidly from north to south. Based on this result with high accuracy, it could be found that the PM2.5 pollution covered a largest area on February 25th, which accounts for about 60.5% of the third model domain.