To study the spatial and temporal distributions of the near-ground concentration of fine particulate matter with an aerodynamic diameter smaller than 2.5 μm (PM_2.5) in Anhui Province from 2017 to 2018, a remote sensing PM_2.5 estimating model was constructed based on the generalized additive model (GAM) considering Himawari-8 aerosol optical depth (AOD), meteorological data, population density (POP), the normalized difference vegetation index (NDVI) and gross domestic product (GDP). The nonlinear response relationship between AOD and PM_2.5 concentration was stronger than other factors. The Bayesian maximum entropy (BME) model was applied to analyse the residuals of the GAM estimation results. After the introduction of the BME model, the coefficient of determination (R²) increased from 0.439 to 0.813, the root mean square error (RMSE) decreased from 18.810 to 12.452 μg/m³, and the model accuracy was improved. The case study in Anhui Province from 2017 to 2018 revealed that the high-pollution areas included Huaibei, Suzhou, Fuyang, Huainan, the junction of Tongling and Anqing, and Hefei, whereas the low-pollution areas were mainly located in Huangshan, in the south. In winter, the average concentrations in Fuyang and Huaibei in northern Anhui and north of Huainan and Bozhou exceeded 105 μg/m³. In summer, no obvious spatial difference in pollution was observed, the pollution concentration was generally lower than 15 μg/m³. The model was prone to underestimating high values in winter, and the characteristics of the different regions affected the estimating accuracy of model.