Spatio-temporal distribution of PM2.5 and variations in the relationship between PM2.5 and other pollutants are the main components of PM2.5spatio-temporal statistical analysis. Existing methods directly analyze spatio-temporal distribution based on monitoring data; thus, it is difficult to effectively reveal the aggregation structure of PM2.5 concentrations. Geographically weighted regression, commonly used to model the relationships between PM2.5 and other pollutants, cannot accurately describe the spatio-temporal variability of dependency. In this study, the clustering structure of PM2.5 concentrations in Beijing was identified using the spatial clustering algorithm and the seasonal distribution characteristics of PM2.5 were analyzed based on the clustering results. The relationship between PM2.5 and PM10 was modeled by geographically and temporally weighted regression and the spatio-temporal variability of dependency was analyzed according to the regression results. The results showed that PM2.5 pollution levels and spatial variability were lower in spring and summer than those in autumn and winter and the concentration of PM2.5 in each season was characterized by low spatial distribution in the north and high spatial distribution in the south. Geographically and temporally weighted regression showed better performance; the correlations between PM2.5 and PM10 in spring and summer are weaker than those in autumn and winter and the correlation between PM2.5 and PM10 in the northwest is stronger than that in the southeast in each season.
Keywords: PM2.5; fuzzy-c-means(FCM) clustering; geographically temporally weighted regression; spatial clustering analysis; spatio-temporal distribution.