中文摘要：

PM_（2.5）已对我国的空气质量构成了严重威胁,对其预警、预报具有重要的意义.由于PM_（2.5）数据同时具有时间与空间属性,而目前的研究缺少对其时空属性的探索与挖掘.以2015年10~12月华北地区58个城市的日均PM_（2.5）浓度数据作为实验数据,利用时空自回归移动平均（STARMA）模型及只考虑时间属性的自回归移动平均（ARMA）模型对PM_（2.5）进行预测,并利用时空克里金插值与只考虑空间属性的普通克里金插值对PM_（2.5）进行插值.结果表明,STARMA模型与ARMA模型的预测相比,时空克里金与普通克里金的插值相比,PM_（2.5）预测及插值精度均有所提升,且具有时空灵活性.

英文摘要：

The air quality of China has been seriously polluted by PM_（2.5）.It is necessary to warn and forecast PM_（2.5）in China.The data of PM_（2.5）has time and space dual properties,but currently,there is little study in spatio-temporal data mining of PM2.5.The daily average PM_（2.5）concentration datum of 58 cities in North China during October to December 2015 were acquired as experimental data,then the PM_（2.5） data were respectively predicted by space-time autoregressive moving average（STARMA）and ARMA model which just considered the time property of the data,and the PM_（2.5） were interpolated by space-time Kriging（ST-Kriging）and ordinary Kriging which just considered the space property of the data.Experimental results show that compared with ARMA and ordinary Kriging model,STARMA and ST-Kriging with the flexibility of time and space have considered the space-time correlation of PM_（2.5）data and could improve the prediction and interpolation accuracy.