Abstract: Based on the raster dataset of annual average concentration of global atmospheric PM2.5, using the gravity model, coefficient of variation and exploratory spatial data analysis method, the paper systematically analyzes the spatio-temporal evolution of PM2.5 in Yangtze River Economic Belt, at regional, provincial, prefectural, county and grid scales. The results show that: (1) From 1998 to 2016, the average annual concentration of PM2.5 in the Yangtze River Economic Belt is higher than the national average, and the trend of PM2.5 concentration changes in the Yangtze River Economic Belt is generally inverted U-shaped. 2005 is an important turning point for the change of PM2.5 concentration in the Yangtze River Economic Belt.; (2) The spatial and temporal evolution of PM2.5 in the Yangtze River Economic Belt shows significant spatial scale effects. At different spatial scales, the spatial differentiation characteristics of annual average concentration of PM2.5 in the Yangtze River Economic Belt have obvious differences, but the spatial differences get enlarged; (3) There is a significant spatial positive correlation of annual average concentration of PM2.5 in the Yangtze River Economic Belt, and it is mainly characterized by high-value agglomeration.