Correlation analysis based on PLS1 between PM_2.5 and air pollutants in Tianjin

The correlation analysis, PLS1 and path analysis were applied to study the direct effect five main air pollutants on the quality of PM_2.5, their indirect effect on other pollutants and also the total effect in Tianjin in different seasons (291 days) of 2014.The results showed that the factors significantly related to PM_2.5 were SO₂, NO₂ and CO in spring, O_{3(8 h)} and CO in summer, SO₂, NO₂ and CO in autumn and SO₂, NO₂, CO, O_{3(8 h)} in winter, respectively. Meanwhile there was a positive multicollinearity between O_{3(1 h)} and O_{3(8 h)} in winter. The goodness of fit R ²of PLS1 model in four seasons was 0.553, 0.323, 0.713, 0.839, respectively, which indicated that the model was reasonable. Path analysis revealed that CO was the greatest pollutant factor to PM_2.5 and showed there existed obvious seasonal variations, with the total effect coefficients of 10.810, 5.587, 6.271, 12.500 in four seasons, respectively. The total effect coefficient of O_{3(8 h)} in summer was 0.897, and that of NO₂ was 0.185, 0.338 and 0.290 in spring, autumn and winter, respectively. Moreover, the direct effect coefficients of the pollutants in four seasons were higher than those of indirect effect.