Spatiotemporal Variability of Remotely Sensed PM(2.5) Concentrations in China from 1998 to 2014 Based on a Bayesian Hierarchy Model

With the rapid industrial development and urbanization in China over the past three decades, PM(2.5) pollution has become a severe environmental problem that threatens public health. Due to its unbalanced development and intrinsic topography features, the distribution of PM(2.5) concentrations over China is spatially heterogeneous. In this study, we explore the spatiotemporal variations of PM(2.5) pollution in China and four great urban areas from 1998 to 2014. A space-time Bayesian hierarchy model is employed to analyse PM(2.5) pollution. The results show that a stable “3-Clusters” spatial PM(2.5) pollution pattern has formed. The mean and 90% quantile of the PM(2.5) concentrations in China have increased significantly, with annual increases of 0.279 μg/m(3) (95% CI: 0.083−0.475) and 0.735 μg/m(3) (95% CI: 0.261−1.210), respectively. The area with a PM(2.5) pollution level of more than 70 μg/m(3) has increased significantly, with an annual increase of 0.26 percentage points. Two regions in particular, the North China Plain and Sichuan Basin, are experiencing the largest amounts of PM(2.5) pollution. The polluted areas, with a high local magnitude of more than 1.0 relative to the overall PM(2.5) concentration, affect an area with a human population of 949 million, which corresponded to 69.3% of the total population in 2010. North and south differentiation occurs in the urban areas of the Jingjinji and Yangtze Delta, and circular and radial gradient differentiation occur in the urban areas of the Cheng-Yu and Pearl Deltas. The spatial heterogeneity of the urban Jingjinji group is the strongest. Eighteen cities located in the Yangtze Delta urban group, including Shanghai and Nanjing, have experienced high PM(2.5) concentrations and faster local trends of increasing PM(2.5). The percentage of exposure to PM(2.5) concentrations greater than 70 μg/m(3) and 100 μg/m(3) is increasing significantly.