Assessment of the Air Quality Numerical Forecast in the Main District of Beijing (2018)

In order to improve model performance, the impact of the new generation WRF-CMAQ (Weather Research and Forecasting model-Community Multi-scale Air Quality model) air quality model system performance of different resolutions for the main district of Beijing was evaluated in 2018. Based on the data set, with PM2.5 as the primary pollutant, forecast grade accuracy of BJ01 (resolution of 1 km) and BJ03 (resolution of 3 km) domains were found to be better compared to that of the official forecast. More than 50% accuracy rate was achieved with BJ01 and BJ03 domains. Compared with the accuracy rate on the first day of official forecast (59%), accuracy rate using the proposed system reached over 60%. A comprehensive scoring method based on the IAQI (Individual Air Quality Index) accuracy and the grade accuracy is adopted. Results show that BJ03 domain achieved the highest score (75.0 points) followed by BJ01 domain. The official forecast scored 70.6 points while BJ09 (resolution of 9 km) domain achieved the lowest score of 69.1 points. Based on the analysis of the prediction results of 2018 long time series of the model system, the model’s predicted PM2.5 concentration is observed to be consistent with that of the observation trend. In addition, the analysis reveals that the correlation coefficient between the model result of BJ03 domain and that of the observation is 0.76. Good peak value simulation performances are achieved in BJ03 and BJ09 domains where there are large area coverages. Similar error trends in peak value simulation of the three model domains are observed. It is evident that the results from the model with coarse resolution are higher than that of the model with fine resolution, which covers a smaller area. Consistent with the forecast comprehensive score, the statistical analysis results reveal that BJ03 domain prediction has the best performance with an average deviation of 0.83 μg/m³. Compared with the observation forecast, BJ09 domain forecast is generally higher whereas BJ01 domain forecast is observed to be lower. Spatial difference analysis of different resolutions from the same site yields inconsistent results. This study shows that best performance is achieved by BJ01, BJ03, and BJ09 areas for the Nongzhanguan, Wanliu, and Dongsi stations, respectively.