Using the data of CHAP PM2.5 and MODIS MCD19A2, the ERA5 reanalysis data and the SNPP/VIIRS satellite monitoring fire point data as well as the least squares method and other analytical methods, this paper explores the spatio-temporal distribution pattern of PM2.5 concentrations in southwest Yunnan and the causes of seasonal bursts from the aspect of backward trajectory of pollutants, circulation situation, high and low air dynamic structure configuration, etc.. The results show that the spatial distribution of PM2.5 concentration and aerosol optical depth (AOD) in the study area are in the pattern of high values in the south and west but low values in the north and east. The PM2.5 concentration is the lowest in July but is the highest in March in a year. The stable transport of pollution sources from February to April causes the spring PM2.5 concentration value in the study area to be higher and the fluctuation to be smaller than in the other three seasons. However, the spatial distribution of the percentage change is more obvious. The percentage reduction in PM2.5 concentration in the past 20 years was mostly by from -30% to -20%, and the percentage change in summer, autumn and winter was mainly less than -30%. More than 90% of the fire points in eastern Myanmar and northern Laos, which are adjacent to the study area, occur from February to April. Under the guidance of the westerly air flow, the westward (southwest-oriented) pollutants are transported over a short distance through the high-fire point area, causing subsidence under the dynamic action of low-layer radiation dispersion in the upper air, and resulting in a seasonal increase in PM2.5 concentration in the study area. A large range of meteorological conditions conducive to the spread of pollutants and the cleaning of more precipitation can reduce the contribution of overseas air pollutants to PM2.5 pollution in southwest Yunnan. The effects of changes in fire points outside the key areas on PM2.5 concentration and AOD show a significant positive correlation. Their effects on PM2.5 lagged behind the change in fire points by about 2 days, and the correlation between the two gradually weakens from south to north.