资料同化方法在空气污染数值预报中的应用研究

基于第5代中尺度非静力气象模式MM5以及区域气溶胶和沉积模式REMSAD耦合的空气污染数值预报模型系统,分别采用最优插值法和集合卡尔曼滤波法对南京2002-08～2002-09 NOx和SO2模型预报结果进行了资料同化试验,结果表明,NOx和SO2经最优插值法同化后偏差平均值的改进率分别为34.20%、47.53%,均方根误差的改进率分别为31.95%、42.04%;NOx和SO2经集合个数为30的集合卡尔曼滤波法同化后偏差平均值的改进率分别为26.73%、60.75%,均方根误差的改进率分别为25.20%、55.16%;说明最优插值法和集合卡尔曼滤波法都具有改善空气污染数值预报中污染物浓度初始场的作用.进行了集合卡尔曼滤波法中集合个数为61时2种同化方法同化效果比较的试验,结果表明,随着集合卡尔曼滤波法集合个数的增加,NOx和SO2的同化效果都较集合个数为30时有所改善,并且,集合卡尔曼滤波法对NOx和SO2模式预报结果的改善效果将好于最优插值法.

Application Research of Data Assimilation in Air Pollution Numerical Prediction

Based on an air pollution modeling system coupling with the non-hydrostatic fifth generation mesoscale meteorological model (MM5) and the regional modeling system for aerosols and deposition (REMSAD), the forecast results of NO_x and SO₂ in August and September 2002 in Nanjing were assimilated with the optimal interpolation method and the ensemble Kalman filter. The results show that the improvement rates of deviation mean value of NO_x and SO₂ after assimilated with the optimal interpolation method are 34.20% and 47.53%, and the improvement rates of root mean square errors are 31.95% and 42.04% respectively. It is also demonstrated that the improvement rates of deviation mean value of NO_x and SO₂ after assimilated with the ensemble Kalman filter with 30 ensemble members are 26.73% and 60.75%, and the improvement rates of root mean square errors are 25.20% and 55.16% respectively. So, the optimal interpolation method and the ensemble Kalman filter both can improve the quality of the initial state from the air pollution numerical prediction model. The comparative experiments on the assimilation performance with the optimal interpolation method and the ensemble Kalman filter with 61 ensemble members were performed, and the experiments demonstrate that the assimilation performance of the ensemble Kalman filter with 61 ensemble members were improved compared with 30 ensemble members, and with the increase of the ensemble members, the improvement to the initial state of NO_x and SO₂ with the ensemble Kalman filter will be better than the optimal interpolation method.