This study took Shenzhen as study area. PM_2.5 concentration in air quality monitoring stations was used and five kinds of landscape metrics including PLAND (percentage of landscape), ED (edge density) at class-level, and CONTAG (contagion), NP (number of patches), AREA_MN (mean patch area) at landscape-level were applied. Further, other data, such as street length, catering number, elevation and land use types considered as factors influencing PM_2.5, were also obtained. By means of correlation analysis and stepwise multiple regression, the effects of land use and landscape pattern on PM_2.5 concentration were explored. The results showed that among land use as sink landscape for PM_2.5, vegetation had the most obvious influence on PM_2.5 concentration; at class-level metrics, both composition metric (PLAND) and configuration metric (ED) were significantly related with PM_2.5 concentration; at landscape-level, fragment (CONTAG and AREA_MN) of the whole landscape had a significant relationship with PM_2.5 pollution. This study could widen the understanding on relationship between landscape and process in landscape ecology and offer advice for air pollution control and landscape planning. Furthermore, it would also provide an effective method to estimate PM_2.5 concentration in case of no measurement.