Substantial contribution of transported emissions to organic aerosol in Beijing

The study addresses the unresolved question of which sources and formation pathways drive secondary organic aerosol (SOA) in Beijing’s haze, a major public health concern. Despite reductions in secondary inorganic aerosol (SIA), health risks persist and may be driven more by organic aerosol (OA), dominated by SOA. The composition-based health effects of PM2.5 are uncertain, and SOA sources and mechanisms are poorly constrained, impeding effective mitigation. The authors propose detailed identification of SOA sources (sector, temporal variability, spatial origin) and formation processes using near-molecular measurements to inform targeted reduction strategies.

Prior work shows high secondary aerosol contributions during Chinese haze and widespread use of aerosol mass spectrometry for source apportionment, but conventional instruments fragment OA, obscuring precursor-specific information. Recent soft-ionization mass spectrometers (e.g., FIGAERO-CIMS, EESI-TOF) enable semi-online molecular characterization with high time resolution, improving SOA source identification. Literature suggests mixed evidence for aqueous-phase SOA formation in Beijing winters, with some estimates indicating major roles for gas-phase oxidation and others highlighting multiphase processes. Studies also document regional transport’s importance to Beijing haze, monsoon-driven seasonality, and interactions between biogenic emissions and urban NOx that enhance SOA formation. These works motivate the combined bulk and near-molecular approach used here.

Study period: November 2019 to July 2020 at an urban residential site on the west campus of Beijing University of Chemical Technology (39°56′31″ N, 116°17′50″ E, ~20 m a.g.l.). Instruments and measurements: (1) ToF-ACSM with PM2.5 lens and standard vaporizer measured non-refractory PM2.5 species (OA, NO3−, SO42−, Cl−, NH4+) with composition-dependent collection efficiency and RIE calibrations; AE33 aethalometer measured eBC. Gaps in ACSM ions were filled with MARGA data. ISORROPIA was used to estimate particle liquid water content (LWC). (2) FIGAERO-CIMS (iodide chemical ionization, long TOF-MS) provided near-molecular particle-phase OA composition via thermal desorption; data were corrected for reagent ion fluctuations, contamination suppression, thermal baselines, background subtraction, and field blanks, yielding levoglucosan-equivalent concentrations; only CxHyOzNwSv iodide clusters retained. Uncertainty propagated from thermogram repeatability and processing steps; poor S/N peaks excluded. (3) NO3−-CIMS characterized gas-phase oxygenated organic molecules (OOM), calibrated to H2SO4, with decision-tree classification of isoprene- and monoterpene-derived OOM. Source apportionment: Independent PMF analyses for ToF-ACSM (rolling ME-2 via SoFi; constrained HOA, COA, BBOA, CCOA profiles; two OOA components) and for FIGAERO-CIMS (unconstrained PMF; selected 8-factor solution for interpretability). Multilinear regression (MLR, Bayesian HMC via Stan) linked FIGAERO-CIMS factor time series to ToF-ACSM OA minus (HOA+COA) to derive factor-specific response factors and mass loadings. HOA and COA were quantified via ToF-ACSM; when ACSM was down, HOA and COA were approximated from eBC and parameterized COA/HOA. Sensitivity analyses: Monte Carlo on RIEs and FIGAERO response factors (3,600 runs); comparisons of MLR-based quantification with direct and levoglucosan-equivalent methods. Transport analysis: 72-hour backward dispersion PES fields from FLEXPART driven by ECMWF data, combined in a concentration-weighted trajectory framework to infer geographic source regions (domain 20–60°N, 95–135°E). Supporting offline analyses: Filter-based quantification of levoglucosan, MBTCA, pinic acid by UHPLC-HRMS for tracer comparison and temperature relationships. Data and code: Dataset on Zenodo; tools include SoFi, Tofware, FLEXPART, MATLAB mapping tools.

- PM2.5 levels and composition: Winter daily mean PM2.5 = 36 µg m−3; summer = 21 µg m−3. Bulk composition similar across seasons: 61–65% secondary inorganic aerosol (SIA), 27–30% organic aerosol (OA), 8–9% eBC. OA to eBC ratio ~3.3–3.5, indicating aged OA and substantial SOA. - OA source split: Even in winter, well over half of OA is secondary. FIGAERO-CIMS directly measures ~61% of OA mass on average (winter 58%, lockdown 66%, summer 59%) when combined with MLR quantification. - Winter OA: Less than half is primary; primary is dominated by solid-fuel OA (SFOA). Secondary components are mainly solid-fuel SOA (sfSOA) and aqueous SOA (SOAaq). Solid-fuel SOA contributes on average 15% of OA in winter (2% in summer) and peaks during cold haze. Aqueous SOA correlates with high particle LWC (R = 0.68), contributes 28% of OA in winter (15% in summer), and accounts for 49% of winter SOA. - Haze episodes (winter): During daily OA >35 µg m−3, OA is predominantly SOA. Solid-fuel SOA contributes ~38–39% of OA and can reach up to ~80% during severe haze. Aqueous SOA contributes ~17–29% of OA; aromatic SOA contributes ~8–9% of OA. Biomass burning dominates primary SFOA during polluted episodes. - Summer OA: OA is dominated by SOA; POA (HOA+COA+SFOA) ~19% during clean conditions. Aromatic-dominated SOA is the main summertime driver (about 61% of SOA), with biogenic SOA a smaller fraction (about 36% of SOA; averages to 27% of total OA). Biogenic SOA increases with temperature (0.6 µg m−3 at 0 °C; 2.1 µg m−3 at 25–30 °C). - Aromatic SOA: Daytime aromSOA shows fingerprints similar to TMB+OH lab SOA and builds up during photochemically active daytime; nighttime aromSOA has strong nitrophenol-like signals (C2H4NO3). Summertime aromSOA originates regionally from the south across the Xi’an–Shanghai–Beijing corridor; air mass ages of 2–3 days and high oxidant loadings support strong oxidation before arrival. - Biogenic SOA: bioSOA correlates better than aromSOA with gas-phase isoprene- and monoterpene-derived OOM. Nighttime bioSOA features nitrate-radical limonene-oxidation markers; daytime bioSOA shows α-pinene ozonolysis-like signatures plus small acids. Under high OA, bioSOAnight exceeds bioSOAday, indicating enhancement via anthropogenic NOx interactions. Main influences are from forested areas south of Beijing. - Aqueous SOA: Associated with regions of high SO2 emissions, high RH/LWC, and sometimes transport over Bohai Sea; contains dimethylnitrobenzoic-acid-like and small di/monocarboxylic acids; weak correlation of aromSOA with LWC suggests distinct non-aqueous formation pathways. - Transported contributions: Elevated sfSOA and SFOA linked to air from the Beijing–Tianjin–Hebei (BTH) region and rural mountainous areas west/northeast of Beijing. Summertime aromSOA and bioSOA originate predominantly from the south within and beyond the Xi’an–Shanghai–Beijing region. - COVID-19 lockdown: Despite reduced traffic and coal usage, secondary PM2.5 remained substantial; FIGAERO showed ~75% of OA was secondary with roughly equal contributions from biogenic, aromatic, and aqueous SOA (Extended Data). - Method validation: FIGAERO-CIMS OA correlates with ToF-ACSM OA minus (HOA+COA) (R = 0.87); nitrate and sulfate tracers show high correlations with ToF-ACSM NO3− (R = 0.91) and SO42− (R = 0.85). Sensitivity analyses confirm robustness of source contributions despite response-factor uncertainties.

The findings demonstrate that SOA in Beijing is largely governed by transported precursor emissions rather than strictly local sources. In winter, solid-fuel-related emissions and aqueous multiphase chemistry dominate SOA, especially during haze, implicating transported contributions from the BTH region and mountainous areas. In summer, anthropogenic aromatic precursors from the regional Xi’an–Shanghai–Beijing corridor dominate SOA, while biogenic contributions, though present and temperature dependent, remain secondary and are modulated by NOx interactions. These seasonal and geographic contrasts explain persistent SOA-driven pollution episodes even during periods of reduced local emissions (e.g., COVID-19 lockdown) and underscore the need to control precursor emissions regionally. The near-molecular source apportionment bridges the gap left by bulk-fragmenting instruments, clarifying precursor classes (solid-fuel, aromatics, biogenic) and formation regimes (aqueous vs. gas-phase oxidation), thereby addressing the initial uncertainty over SOA sources and pathways.

This study provides a near-molecular, high-time-resolution source apportionment of OA in Beijing, revealing that transported emissions substantially drive SOA across seasons. Winter haze is most strongly associated with solid-fuel-derived SOA and aqueous-phase formation, while summer OA is dominated by aromatic SOA with regional southern origins; biogenic SOA is present but less dominant. The combined FIGAERO-CIMS and ToF-ACSM plus PMF/MLR framework identifies SOA source classes and quantifies their contributions, enabling targeted mitigation strategies. Policy implications include the need for coordinated, large-scale regional controls on key organic precursor sectors (solid-fuel combustion, aromatic VOCs) across the Xi’an–Shanghai–Beijing corridor and neighboring regions. Future research should extend this near-molecular framework to other megacities and seasons, integrate toxicity metrics to connect source-specific SOA to health outcomes, refine response-factor calibrations, and better resolve multiphase chemistry and anthropogenic–biogenic interactions under varying NOx and humidity regimes.

- Instrumental detection and quantification: FIGAERO-CIMS does not detect all OA compounds and can thermally decompose labile species; molecular structures are not resolved. Factor-specific response factors vary (~10× range) and introduce uncertainty despite MLR calibration to ToF-ACSM OA minus (HOA+COA). - Source apportionment constraints: HOA and COA are not directly identified by FIGAERO-CIMS and rely on ToF-ACSM estimates; PMF solutions depend on chosen factor numbers and constraints. - Generalizability: Results are from a single urban site and a specific period (Nov 2019–Jul 2020), including an atypical COVID-19 phase that altered emissions. - Aqueous-phase inference: Identification relies on correlations with LWC and transport over SO2-rich regions; causality and microphysical mechanisms are inferred rather than directly observed. - Transport analysis: CWT/FLEXPART-based source regions are subject to meteorological and model uncertainties; 72-hour back dispersion may miss farther sources.