Regional ozone data assimilation experiment based on ensemble Kalman filter
基于集合卡尔曼滤波的区域臭氧资料同化试验

A regional air quality data assimilation system (RAQDAS) was established based on ensemble Kalman filter and Nested Air Quality Prediction Model System. This system was employed to assimilate surface ozone observation of Beijing-Tianjin-Hebei areas during the 2008 Beijing Olympics period and to optimize ozone initial conditions. The effects of data assimilation on 24 h ozone forecast were investigated. The results show that the assimilation with 50 ensemble members can improve the ozone forecast not only over observational areas, but also over non-observed areas. On average, the data assimilation can decrease the root mean square error (RMSE) of 24 h ozone forecast by 15%. Furthermore, the ensemble size can be reduced to 20 with similar improvement on forecast capability. In order to solve the problem of filter divergence, inflating ensemble spread and perturbing model error sources were employed. Inflating ensemble spread can solve the problem of filter divergence, but it can hardly improve ozone forecast and lead to an increase of ozone forecast error; perturbing model error sources can avoid filter divergence and also bring improvement of 24 h ozone forecast with the RMSE decreased by 20%.