PM2.5 is one of the important pollutants in the atmosphere, so simulating the spatial distribution of PM2.5 concentration is of great significance to the prevention and control of air pollution. The Land Use Regression (LUR) model was applied to the heavily polluted Northern Anhui region in Anhui Province, China. Taking the monitoring points as the center, the buffer zones with radius of 0.5, 1, 1.5, 2, 3, 4 and 5 km were established respectively. Combined with 105 variables including land use factor, road factor, pollution source factor, meteorological factor, elevation factor and population factor, a four-season and annual average LUR model for this district was established, and the accuracy of the model was verified by leave-one-out cross validation. The results show that the PM2.5 concentration in the study area is greatly affected by grassland, wetland, rainfall, relative humidity, atmospheric pressure, wind speed, secondary roads,tertiary roads, air-polluting enterprise, and population. The adjusted R2 is 0.828 (spring), 0.731 (summer), 0.831 (autumn), 0.775 (winter) and 0.892 (annual average) respectively. The root mean square error (RMSE) is 6.34 μg·m-3 (spring), 7.01 μg·m-3 (summer), 6.28 μg·m-3 (autumn), 6.71 μg·m-3 (winter) and 5.33 μg·m-3 (annual average). The simulation accuracy R2 is 0.825 (spring), 0.730 (summer), 0.834 (autumn), 0.772 (winter) and 0.897 (annual average). The model shows good performance and strong explanatory power. As can be seen from the simulated spatial distribution of PM2.5 concentration, the spatial distribution characteristics in the area are obviously different in different seasons, which is related to a large number of pollution particles from the north, local coal mining, straw burning during autumn tillage and other potential pollution sources.