Abstract: PM_2.5 is an important pollutant forming haze weather. Exploring the temporal and spatial evolution law and spatial heterogeneity of PM_2.5 pollution is of great significance to the accurate treatment of air pollution. Based on the county PM_2.5 concentration data of Shaanxi province from 2000 to 2018, this paper systematically studies the temporal and spatial evolution trend of PM_2.5 pollution by using the methods of gravity center model and spatial autocorrelation analysis. The results showed that from 2000 to 2018, the annual mean value of PM_2.5 concentration in Shaanxi province experienced an inverted “N-type” of small fluctuation decline in the early stage, short sharp rised in the middle stage and large fluctuation decline in the later stage. 2011 was an important “inflection point” for the fluctuation of the annual mean value of PM_2.5 concentration. The annual mean value of PM_2.5 concentration was lower than the primary concentration limit （15 μg/m³） of low pollution counties （districts） account for relatively small and unstable changes, the proportion of 15~35 μg/m³ （secondary concentration limit） had increased, and the proportion between 35~70 μg/m³ continues to decrease, reflecting that the PM_2.5 pollution in most counties （districts） will be gradually controlled below the secondary concentration limit in the future. In terms of spatial distribution, the average annual value of PM_2.5 concentration had obvious regional differentiation. The annual mean value of PM_2.5 concentration had eight “Rise-Decline” time series evolution types in the four stages of 2000—2005, 2005—2010, 2010—2015 and 2015—2018, and the counties （districts） of “D-R-D-D” type account for the absolute proportion （59.81%）. The gravity center analysis showed that the PM_2.5 pollution gravity center in Shaanxi province moves from southwest to northeast, which roughly presents the spatial pattern and trend of prominent pollution in Guanzhong, high in the west and low in the east, and light pollution in northern and southern Shaanxi. The annual average value of PM_2.5 concentration showed positive spatial autocorrelation in space, and the spatial agglomeration generally showed the time evolution trend of first falling, then leveling and then rising. Most counties （districts） were “High-High” type agglomeration and “Low-Low” type agglomeration, with obvious convergence trend and serious polarization.

Key words: PM_2.5, center of gravity model, spatial autocorrelation, spatiotemporal evolution, Shaanxi province