Prediction of PM2.5 concentration based on the similarity in air quality monitoring network

Recently particulate matter pollution has becoming more and more serious in China and plenty of equipment has been purchased to detect it in air quality monitoring network. But it is inevitable to make the government to bear a significant financial burden because of expensive equipment. With this consideration, we attempt to explore some practicable methods to estimate the pollutant concentration with available data at surrounding stations instead of measurement. In light of this, the Spearman correlation analysis and cluster analysis are utilized to reveal the similar behavior in Shanghai PM2.5 monitoring network respectively. They coincidentally demonstrate that there exists redundant equipment in monitoring network. Then based on it, the linear method of stepwise regression and the nonlinear method of support vector regression are applied to predict PM2.5 concentration at target station in term of the values at surrounding stations. Both of them show good performance and they are recognized to be practicable to estimate the values measured by redundant equipment. Obviously, these findings give rise to the possibility to remove some equipment in monitoring network. Hence, in order to remove it reasonably, two removing criteria for redundant equipment are suggested finally. It makes use of the similarity in air quality monitoring network and guarantees that the missed values caused by removed equipment can be replaced successfully through prediction, which are advantage for monitoring network advisors to make informed decisions as to whether a redundant equipment must be removed or relocated.