Insight into PM<sub>2.5</sub> Sources by Applying Positive Matrix Factorization (PMF) at an Urban and Rural Site of Beijing

Abstract

Abstract. This study presents the source apportionment of PM2.5 performed by PMF on data collected at an urban (Institute of Atmospheric Physics – IAP) and a rural site (Pinggu-PG) in Beijing as part of the Atmospheric Pollution and Human Health in a Chinese megacity (APHH-Beijing) field campaigns. The campaigns were carried out from 9th November to 11th December 2016 and 22nd May to 24th June 2017. The PMF included both organic and inorganic species, and a seven-factor output provided the most reasonable solution for the PM2.5 source apportionment. These factors are interpreted to be traffic emissions, biomass burning, road dust, soil dust, coal combustion, oil combustion and secondary inorganics. Major contributors to PM2.5 mass were secondary inorganics (22–24 %), biomass burning (30–36 %), and coal combustion (20–21 %) sources during the winter period at both sites. Secondary inorganics (48 %), road dust (20 %) and coal combustion (17 %) showed the highest contribution during summer at PG, while PM2.5 particles were mainly composed of soil dust (35 %) and secondary inorganics (40 %) at IAP. Despite this, factors that were resolved based on metal signatures were not fully resolved and indicate a mixing of two or more sources. PMF results were also compared with sources resolved from another receptor model (i.e. CMB) and PMF performed on other measurements (i.e. online and offline aerosol mass spectrometry (AMS)) and showed a good agreement for some but not all sources. The biomass burning factor in PMF may contain aged aerosols as a good correlation was observed between biomass burning and oxygenated fractions (r2 = 0.6–0.7) from AMS. The PMF failed to resolve some sources identified by the CMB and AMS, and appears to overestimate the dust sources. A comparison with earlier PMF source apportionment studies from the Beijing area highlights the very divergent findings from application of this method.

Source apportionment of fine organic carbon at an urban site of Beijing using a chemical mass balance model

“…In addition, the concentrations of OOA1 and OOA3 were higher in daytime, implying the effect of photochemical processing. The variations in OOA2 tracked well with C 2 H 2 O + 2 (R = 0.89), an aqueousprocessing-related fragment ion (Sun et al, 2016), indicating that OOA2 was an OA factor associated with aqueousphase processing. Previous studies suggested that aqueousphase processing plays an important role in the formation of nitrogen-containing compounds (Xu et al, 2017).…”

“…In order to be compared with the source apportionment results of OC in this study from the CMB model, the OA concentrations from the AMS-PMF were converted to OC based on various OA/OC ratios measured in Beijing: 1.35 for CCOA/CCOC (coal combustion organic carbon), 1.31 for HOA/HOC (hydrocarbon-like organic carbon) (Sun et al, 2016), 1.38 for COA/COC (cooking organic carbon), 1.58 for BBOA/BBOC (biomass burning organic carbon) (W. , and 1.78 for OOA/OOC (oxidized organic carbon; Huang et al, 2010). The concentrations of OA and corresponding OC from AMS-PMF analysis are presented in Table 3.…”

“…The CCOA-AMS factor was mainly characterized by m/z of 44, 73, and 115 (Sun et al, 2016). In winter, CCOA-AMS was 6.2 ± 4.4 µg m −3 , contributing 16.9 % of OM.…”

“…COA-AMS is as a common factor identified in both winter and summer results. It is characterized by high m/z of 55 and 57 in the mass spectrum (Sun et al, 2016). COA-AMS was 5.9 ± 4.1 and 1.8 ± 1.0 µg m −3 in winter and summer, respectively, contributing 16.1 % and 17.8 % of OM.…”