[Optimization of PM10 monitoring network in Beijing]

PM10 monitoring network in Beijing was classified using a new technique, positive matrix factorization (PMF). And then the removal bias of each cluster was calculated by GIS system and sites with redundant information were identified. The daily average mass concentrations of PM10 from July 2007 to June 2008 were analyzed at 26 sites. The result showed that PM10 monitoring network of Beijing was separated into 10 clusters. Tongzhou, Yanqing, Miyunshuiku, Fangshan, and Pinggu formed five separate clusters. The five clusters with more than one site each were Cluster 4, which included sites Fengtaihuayuan, Fengtaiyungang, Mentougou, Haidianbeibuxinqu, and Shijingshan, located within the west developing urban area; Cluster 7, which included Dongchengdongsi, Dongchengtiantan, Xichengwanshouxigong, Xichengguanyuan, Chaoyangaotizhongxin, Chaoyangnongzhanguan, and Shunyi, located mainly within the developed area and the east developing area; Cluster 8, which included Daxingyizhuang, Daxinghuangcun, and Daxingyufa, located within the southern suburban industrial area; Cluster 9, which included Miyunzhen and Huairou, located within the north remote rural area; and Cluster 10, which included Haidianxiangshan, Changpingdingling, Haidianwanliu, and Changpingzhen, located within the northwest suburban area. All the 10 clusters had unique seasonal variations. According to the removal criteria, two scenarios were constructed. The criterion of scenario 1 was the uncertainty of the PM10 monitoring network, and the optimization result in which 12-18 sites should be retained was equal to the original monitoring network included 26 sites. The criterion of scenario 2 was two times of the uncertainty, and 10-13 sites needed to be retained.