中文摘要：

空气质量统计预报模型是当前空气质量预报的重要工具之一。该研究选取珠海市4个国控大气自动监测站点（吉大、唐家、前山、斗门）,基于大气环境监测数据和气象数据研究了同期回归、多元回归2种空气质量预报统计模型,并对模型在不同污染物（NO_2、SO_2、CO、O_3、PM_（10）、PM_（2.5））、不同预报时段（24、48、72、96、120、144 h）、不同国控监测站点的情景下进行了预报准确度综合评估。结果表明：（1）PM_（2.5）浓度与风向的皮尔逊相关系数最大,其相关系数为0.403;O_3浓度与气温、湿度均具有较大的相关性,其相关系数分别为0.705、-0.823;（2）同期回归模型对于6项污染物浓度预报的准确度由高至低分别为：CO〉PM_（10）〉PM_（2.5）〉O_3〉NO_2〉SO_2,其准确度分别为84%、75.9%、73.4%、72.3%、66.8%与61.9%;（3）多元回归模型对于6项污染物浓度预报的准确度由高至低分别为：CO〉PM_（2.5）〉PM_（10）〉O_3〉NO_2〉SO_2,其准确度分别为85.6%、73.3%、69.9%、67.6%、67.4%与58.7%。

英文摘要：

The statistical prediction model is one of the important tools for air quality forecast. In this paper, the environment monitoring data and meteorological data are studied in order to establish two air quality forecast statistical models, and the monitoring data were obtained from 4 state controlling air sampling sites in Jida, Tangjia, Qianshan, Doumen of Zhuhai. The accuracy of predicted values were discussed in terms of different pollutants such as NO_2, SO_2, CO, O_3, PM_（10）, PM_（2.5）, different forecasting periods including 24, 48, 72, 96, 120, 144 h, different sites. The results showed that the Pearson correlation coefficient between PM_（2.5） and wind direction is 0.403, which is the highest; O_3 concentration is closely related to both temperature and humidity, with the Pearson correlation coefficient as 0.705 and-0.823 respectively. The accuracy for 6pollutants concentration calculated by corresponding period regression model indicated COPM_（10）PM_（2.5）O_3NO_2SO_2, with the corresponding values as 84%, 75.9%, 73.4%, 72.3%, 66.8% and 61.9% respectively; while the accuracy for 6 pollutants concentration calculated by multiple regression model indicated COPM_（2.5）PM_（10）O_3NO_2SO_2, with the corresponding values as 85.6%, 73.3%, 69.9%, 67.6%, 67.4% and 58.7% respectively.