Abstract: Atmospheric environmental background data,weather conditions and formation mechanisms of four typical air pollution episodes in Beijing in October 2014 were investigated by combining observed data and result from the numerical model CAMx. The results showed that the occurrences of heavy air pollution resulted mainly from stable regional or local atmospheric conditions.Observed heavy pollution episodes were characterized by a stagnant atmospheric structure with average wind speed of 1.52 m/s,high humidity of 80.75% and, large inversion strongth of 2.26 ℃/100 m which were disadvantageous to the dispersion of air pollutants.The air pollution in October 8^th-11^st episode was most serious with daily average PM_2.5 concentration of 264 μg/m³ and highest contribution from regional chemistry transport (63.75%), the second most serious pollution episode was October 24^th-25^th during which there was a strongest inversion (5.94 ℃/100 m). During the October 18^th-20^th heavy pollution episode,high concentrations PM_2.5(above 200 μg/m³) were found mainly in the northwest part of Beijing. The heavy air pollution episode of October 30^th-31^st was actually the lightest among the four episodes, with daily average PM_2.5 concentration of 154 μg/m³.Model simulation analysis showed that regional transport contribution to Beijing's air pollution ranged between 42.36%-69.12%, and also that the formation of secondary inorganic and organic aerosols was significant.