Comparison of receptor models for source apportionment of volatile organic compounds in Beijing, China Yu Song a , Wei Dai a , Min Shao b, * , Ying Liu b , Sihua Lu b , William Kuster c , Paul Goldan c a Department of Environmental Sciences, Peking University, Beijing 100871, China b State Joint Key Laboratory of Environmental Simulation and Pollution Control, Peking University, Beijing 100871, China c Chemical Sciences Division, NOAA Earth System Research Laboratory, Boulder, CO 80305, USA Received 13 June 2007; received in revised form 28 September 2007; accepted 12 December 2007 VOCs sources were similar for three models with CMB showing a higher estimate for vehicles. Abstract Identifying the sources of volatile organic compounds (VOCs) is key to reducing ground-level ozone and secondary organic aerosols (SOAs). Several receptor models have been developed to apportion sources, but an intercomparison of these models had not been performed for VOCs in China. In the present study, we compared VOC sources based on chemical mass balance (CMB), UNMIX, and positive matrix factorization (PMF) models. Gasoline-related sources, petrochemical production, and liquefied petroleum gas (LPG) were identified by all three models as the major contributors, with UNMIX and PMF producing quite similar results. The contributions of gasoline-related sources and LPG esti- mated by the CMB model were higher, and petrochemical emissions were lower than in the UNMIX and PMF results, possibly because the VOC profiles used in the CMB model were for fresh emissions and the profiles extracted from ambient measurements by the two-factor analysis models were ‘‘aged’’. Ó 2007 Elsevier Ltd. All rights reserved. Keywords: VOC; Source apportionment; UNMIX; PMF; CMB; Beijing 1. Introduction Many ambient volatile organic compounds (VOCs) have been found to have adverse effects on public health (1990 Clean Air Act Amendments, Section 112). Chemical reactions of VOCs with nitrogen oxides under sunlight lead to the pro- duction of secondary air pollutants, resulting in tropospheric ozone and secondary organic aerosols (SOAs) (Seinfeld and Pandis, 1998). The air quality in Beijing is of great concern as the 2008 Olympic Games are approaching. Episodes of high surface ozone concentrations can occur frequently there during the summer and autumn. From 1999 to 2004, the national ozone standard was exceeded on 110, 109, 77, 45, 57, and 67 days, respectively (Beijing Environmental Protection Bureau, 1999e2004). Monitoring data also indicate that PM10 pollu- tion is a serious problem in Beijing. Moreover, recent research has found that in summer, secondary organic carbon (SOC) accounted for more than 50% of the total organic carbon in PM10 pollution in Beijing (Duan et al., 2005). Surface ozone and particulate matter abatement are needed urgently in Bei- jing, and because VOCs are important precursors to ozone and SOA, their sources must be identified and quantified to de- velop effective control measures. Although emission sources can be quantified by means of source inventory establishment, dispersion models, and receptor models, establishing an emis- sions inventory is often too time- and resource-consuming. Moreover, the performance of dispersion models depends Abbreviations: 2M-propene, 2-methylpropene; 2M-1-butene, 2-methyl-1- butene; 3M-1-butene, 3-methyl-1-butene; 2M-2-butene, 2-methyl-2-butene; MTBE, methyl tert-butyl ether. * Corresponding author: Tel.: þ86 10 62757973; fax: þ86 10 62751927. E-mail address: mshao@pku.edu.cn (M. Shao). 0269-7491/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.envpol.2007.12.014 Available online at www.sciencedirect.com Environmental Pollution 156 (2008) 174e183 www.elsevier.com/locate/envpol