中文摘要：

利用嵌套网格空气质量预报模式，在对上海市环境监测中心提供的观测数据进行必要的质量控制后，采用最优插值方法对可吸入颗粒物（PM10）、二氧化氮（NO2）和二氧化硫（SO2）进行资料同化。选取2004年8月1～20日做作逐日同化试验的结果表明，无论是PM10、NO2还是SO2，其同化偏差平均值均在20μg·m^-3以下，比同化前减少了至少50％；3种污染物的同化偏差小于其未同化偏差的天数均在16天以上。在大气清洁和污染两种情况下，对PM10分别作10天的同化试验表明，同化后的均方根误差均小于同化之前。此同化方法能利用观测数据较好地修正空气质量模式预报场，从而为模式提供与实际更加接近的初始场。

英文摘要：

Observations of PM10, NO2 and SO2 by Shanghai Environmental Monitoring Center with necessary quality control were assimilated into the nested air quality prediction model, using the optimal interpolation approach. The assimilation performed day by day from August 1st to 20th 2004 shows that for all of PM10, NO2 and SO2, the mean assimilating departures are less than 20μg · m^-3 , decreasing to at least 50 percent of those without assimilation. Also, the bias errors of the three pollutants with assimilation are smaller than those without assimilation for more than 16 days. The assimilation of PM10 for 10 days from both of clean air condition and polluted air condition shows that the root mean square （RMS） errors with assimilation are less than those without assimilation. It is illustrated that this assimilation method can improve a large extent the simulation of urban air quality model using observations so as to afford initial condition more close to the true values.