Abstract: With the acceleration of urbanization, air pollution has become one of the most important urban problems in China, affecting public health seriously. At present, there are few studies to explore the influence of landscape pattern on PM2.5 concentration at the microscopic scale. Therefore, the current study, taking the Chang-Zhu-Tan urban agglomeration as an example, selected six predictors including terrain, pollution, population, road traffic, land use and urban landscape pattern, then used a stepwise linear regression model to explore the effects of urban landscape pattern on PM2.5 concentration. The urban landscape pattern was represented by seven landscape metrics which were ED, CONTIG, LSI, AREA_MN, CONTAG and SHEI. The results showed that: 1) the selected landscapes metrics could explain 73.2% variation of PM2.5 concentration in the study area and the model fitted well; 2) the types of land use that affected PM2.5 concentration included construction land, woodland, grassland and water body. At the microscopic scale, landscape metrics including CONTIG and LSI had significant effects on PM2.5. When the contiguity of construction land was higher, the distribution was more concentrated. Consequently, PM2.5 concentration was higher. If the shape index of water body was smaller, the shape of water body was simpler and more regular. It was easier to reduce PM2.5 concentration; 3) at the overall urban landscape-level, landscape aggregation degree and landscape diversity had important impacts on PM2.5 concentration. Reducing the discrete distribution of various types of patches in the landscape, which could make each landscape type evenly distribute in the overall landscape, could help to reduce PM2.5 concentration. Results of this research could provide reference for future atmosphere control and urban planning.