Beijing, which is the capital of China, suffers from severe Fine Particles (PM2.5) pollution during the heating season. In order to take measures to control the PM2.5 pollution and improve the atmospheric environmental quality, daily PM2.5 samples were collected at an urban site from 15 November to 31 December 2016, characteristics of PM2.5 chemical compositions and their effect on atmospheric visibility were analyzed. It was found that the daily average mass concentrations of PM2.5 ranged from 7.64 to 383.00 μg m−3, with an average concentration of 114.17 μg m−3. On average, the Organic Carbon (OC) and Elemental Carbon (EC) contributed 21.39% and 5.21% to PM2.5, respectively. Secondary inorganic ions (SNA: SO42− + NO3− + NH4+) dominated the Water-Soluble Inorganic Ions (WSIIs) and they accounted for 47.09% of PM2.5. The mass concentrations of NH4+, NO3− and SO42− during the highly polluted period were 8.08, 8.88 and 6.85 times greater, respectively, than during the clean period, which contributed most to the serious PM2.5 pollution through the secondary transformation of NO2, SO2 and NH3. During the highly polluted period, NH4NO3 contributed most to the reconstruction extinction coefficient (b′ext), accounting for 35.7%, followed by (NH4)2SO4 (34.44%) and Organic Matter (OM: 15.24%). The acidity of PM2.5 in Beijing was weakly acid. Acidity of PM2.5 and relatively high humidity could aggravate PM2.5 pollution and visibility impairment by promoting the generation of secondary aerosol. Local motor vehicles contributed the most to NO3−, OC, and visibility impairment in urban Beijing. Other sources of pollution in the area surrounding urban Beijing, including coal burning, agricultural sources, and industrial sources in the Hebei, Shandong, and Henan provinces, released large amounts of SO2, NH3, and NO2. These, which were transformed into SO42−, NH4+, and NO3− during the transmission process, respectively, and had a great impact on atmospheric visibility impairment.
An Atmospheric Visibility Grading Method Based on Ensemble Learning and Stochastic Weight Average