Organic aerosol (OA) is a major component of tropospheric submicron aerosol that contributes to air pollution and causes adverse effects on human health. Chemical transport models have difficulties in reproducing the variability in OA concentrations in polluted areas, hindering understanding of the OA budget and sources. Herein, we apply both process-based and observation-constrained schemes to simulate OA in GEOS-Chem. Comprehensive data sets of surface OA, OA components, secondary organic aerosol (SOA) precursors, and oxidants were used for model–observation comparisons. The base models generally underestimate the SOA concentrations in China. In the revised schemes, updates were made on the emissions, volatility distributions, and SOA yields of semivolatile and intermediate-volatility organic compounds (SVOCs and IVOCs) and additional nitrous acid sources. With all the model improvements, both the process-based and observation-constrained SOA schemes can reproduce the observed mass concentrations of SOA and show spatial and seasonal consistency with each other. Our best model simulations suggest that anthropogenic SVOCs and IVOCs are the dominant source of SOA, with a contribution of over 50 % in most of China, which should be considered for pollution mitigation in the future. The residential sector may be the predominant source of SVOCs and IVOCs in winter, despite large uncertainty remaining in the emissions of IVOCs from the residential sector in northern China. The industry sector is also an important source of IVOCs, especially in summer. More SVOC and IVOC measurements are needed to constrain their emissions. Besides, the results highlight the sensitivity of SOA to hydroxyl radical (OH) levels in winter in polluted environments. The addition of nitrous acid sources can lead to over 30 % greater SOA mass concentrations in winter in northern China. It is important to have good OH simulations in air quality models.
Organic aerosol (OA) is a major component of tropospheric submicron aerosol, which can be directly emitted as primary organic aerosol (POA) or formed from atmospheric oxidation processes as secondary organic aerosol (SOA) (Zhang et al., 2007). Accurate OA simulation is important for understanding the aerosol budget as well as evaluating the impacts of fine particles on air quality and human health. High OA concentrations occur in populated and polluted areas, especially in China and India (Li et al., 2017; Gani et al., 2019). However, atmospheric chemical transport models (CTMs) have difficulties in reproducing the magnitude and the variability in OA mass in polluted environments, mainly resulting from the underestimation of SOA (Park et al., 2021; Miao et al., 2020; Jiang et al., 2019).
SOA is generally simulated in CTMs by the process-based scheme, for which the oxidation of each category of lumped SOA precursors is parameterized with specific SOA yields (Chung and Seinfeld, 2002; Hodzic et al., 2016). Some of the SOA sources are uncertain. For example, the estimated annual production of anthropogenic SOA varied by tens of teragrams per year in different models, which has been attributed largely to the uncertain contribution from semivolatile and intermediate-volatility organic compounds (SVOCs and IVOCs) (Spracklen et al., 2011; Hodzic et al., 2016; Pai et al., 2020). The SVOCs and IVOCs have been recognized as key SOA precursors in polluted areas for over a decade (Robinson et al., 2007; Grieshop et al., 2009). Transportation, industry, and residential use of solid fuel, for example, are all important sources of SVOCs and IVOCs. Although tremendous efforts have been made to characterize their SOA production, CTMs treat their emissions, volatility distributions, reactivities, and SOA yields differently. The emissions of SVOCs and IVOCs are estimated by applying empirical scale factors to different proxies such as POA, non-methane volatility organic compounds (NMVOCs), and speciated IVOCs (Pye and Seinfeld, 2010; Jathar et al., 2011; Shrivastava et al., 2015; Hodzic et al., 2016). The uncertainties can be over 200 % for individual emission sectors, especially at a regional scale (Wu et al., 2021; Lu et al., 2020). For IVOCs, some CTMs use one lumped precursor with specific SOA yields (Pye and Seinfeld, 2010; Hodzic et al., 2016; Ots et al., 2016). Some CTMs use a volatility basis set (VBS) approach for which continuous oxidation occurs to decrease the volatilities of oxidation products and alters gas-to-particle partitioning (Li et al., 2020; Chrit et al., 2018; Shrivastava et al., 2015). Although a recent study categorizes IVOCs into six groups based on volatility and molecular structure for which SOA yield parameters of each group are derived from laboratory experiments of mobile emissions (Lu et al., 2020), there is still a lack of a source-dependent model framework for IVOC-related SOA simulations. A new observation-constrained scheme has been developed in CTMs to improve the simulation of SOA mass in polluted areas, which estimates anthropogenic SOA formation potential based on the emission of carbon monoxide (CO) (Hodzic and Jimenez, 2011). This SOA scheme was able to reproduce the OA mass concentrations in the Mexico City metropolitan area, the United States, and China (Hodzic and Jimenez, 2011; Kim et al., 2015; Woody et al., 2016; Miao et al., 2020). However, its parameterization is too generalized to differentiate specific source contributions. The contribution of SVOCs and IVOCs to SOA in polluted environments remains unclear.
On the other hand, the model performance on atmospheric oxidation capacity may affect the simulation of SOA production. The measured concentrations of
hydroxyl radicals (OHs) show high values in polluted environments in China, caused by strong production from ozone (
Herein, we conduct the OA simulations in China with both of the process-based and observation-constrained schemes in the atmospheric chemical transport model GEOS-Chem. Model improvements are made on the emissions, volatility distributions, and SOA yields of SVOCs and IVOCs as well as the HONO sources. The model simulations are evaluated against nationwide measurements and the positive matrix factorization (PMF)-based source apportionment results of OA. The improved model simulations provide insights into the budget and sources of SOA in China and hence assist in developing control strategies for the SOA pollution.
The campaign-average mass concentrations of OA were taken from 68 surface measurements at urban sites, 18 measurements at suburban sites, and
8 measurements at remote sites from 2011 to 2019 (Table S1 in the Supplement). These measurements were conducted by Aerodyne aerosol mass
spectrometers (AMSs) and aerosol chemical speciation monitors (ACSMs) and covered mains regions in China, including the North China Plain (NCP), the Yangtze
River Delta (YRD), the Pearl River Delta (PRD), and Northwest China (NW). The campaign-average mass concentrations of OA factors that were resolved by
PMF analysis were also synthesized. These OA factors include hydrocarbon-like OA (HOA), cooking-related OA (COA), biomass-burning-related OA (BBOA),
coal-combustion-related OA (CCOA), and various oxygenated OAs (OOAs). We named the summed concentrations of HOA, COA, BBOA, and CCOA as PMF-derived
POA and those of OOAs as PMF-derived SOA. Unlike our previous study (Miao et al., 2020), we did not divide the measured concentrations by the
empirical submicron-to-fine mass ratio because of the lack of such information for different regions and seasons (Y. Zheng et al., 2020; Sun et al.,
2020a). Because the model simulations consistently underestimate the OA concentrations, taking into account the potential supermicron mass may lead to
greater model–observation gaps but not affect the analysis herein. Moreover, we synthesized a data set of the campaign-average concentrations of
benzene, toluene, and xylene from 49 measurements in China from 2011 to 2018 that were conducted by online gas chromatography (GC) coupled with a flame
ionization detector (FID) and/or mass spectrometer (MS). Table S2 in the Supplement lists the sampling information and the results of these
measurements. In addition, a recent result of primary IVOCs measured by offline sampling with thermal desorption (TD)–GC–MS in urban Shanghai
(31
Model simulations were conducted on an atmospheric chemical transport model GEOS-Chem v12.6.3 (
Descriptions of the OA simulations in this study. Additional HONO sources include direct emissions from traffic, soil, and biomass burning as well as secondary formation from the heterogeneous reaction of
We focus here on the simulations of OA. OA is simulated by so-called complex (i.e., process-based) and simple SOA (i.e., observation-constrained)
schemes (Pai et al., 2020). The Cp_base simulation represents the default complex SOA configuration, in which SOA is produced by the oxidation of
lumped biogenic, aromatic, and SVOC and IVOC precursors; heterogeneous uptake of glyoxal and methylglyoxal; and isoprene multi-phase chemistry (Marais
et al., 2016; Fisher et al., 2016; Pye and Seinfeld, 2010; Pye et al., 2010). The emissions of SVOCs are treated as 1.27 times the primary OC
emissions, and the emissions of IVOCs are set as 66 times the naphthalene emissions (Pye and Seinfeld, 2010). Primary SVOCs are emitted as two tracers
with saturation concentrations (
Modifications to the SOA schemes are listed in Table 1. The Cp_R1 and Sp_R1 simulations have updates on precursor emissions, SOA yields, or
parameters related to the production and removal processes. Specifically, the Cp_R1 simulation applies a more reasonable scale factor of 1.0 for SVOC
emissions instead of 1.27 that is used in the Cp_base simulation (Lu et al., 2018). Instead of using two bins for all sources, the volatility
distributions of SVOC emissions are specified for transportation, other anthropogenic sources, and biomass burning and contain five bins with
The annual emissions of IVOCs derived by different approaches in the literature and in our study. The values in parentheses are the IVOC emissions when the residential emission is multiplied by a factor of 7.
The Cp_R1
We estimated the IVOC emissions from the emissions of NMVOCs instead of naphthalene in the revised model schemes because laboratory experiments show a
better correlation of the total IVOC emissions with NMVOCs than with individual IVOC species (e.g., naphthalene) or POA (Zhao et al., 2015; Y. Zhao
et al., 2016). Global anthropogenic emissions of NMVOCs are provided by CEDS (Hoesly et al., 2018), and the emissions from biomass burning are
provided by GFED4 (Giglio et al., 2013; Andreae, 2019). In China, anthropogenic emissions of NMVOCs are taken from MEIC. The NMVOC emission profiles
of sectors (i.e., power, transportation, industry, and residential) and subsectors (i.e., gasoline, diesel, coal, solvent, and biofuel or biomass
burning) and the IVOC/NMVOC emission ratios of the subsectors as well as the volatility distributions of IVOCs for the subsectors with
Table 2 lists the total IVOC emissions estimated in various studies. Globally, the IVOC emissions range from 16.0 to 234
The statistics of model–observation comparisons of the campaign-average mass concentrations of OA, POA, and SOA in China. “OBS” and “SIM” represent the mean values of the observations and the simulations, respectively. NMB is normalized mean bias, NME is normalized mean error, RMSE is root mean square error, and
Figure 1a shows the observed campaign-average mass concentrations of OA in China, which range from 0.7 to 128.5
The statistical values such as normalized mean bias (NMB), normalized mean error (NME), root mean square error (RMSE), and Pearson's correlation
coefficient (
Figures 1c and S6 in the Supplement show the simulated SOA concentrations compared with the observations at urban, suburban, and remote sites. The
underestimation of the Cp_base simulation of SOA mainly occurs in urban and suburban regions, which is consistent with previous understanding about
the underrepresented sources of anthropogenic SOA in process-based models (B. Zhao et al., 2016; Z. Han et al., 2016). The Sp_base simulation
captures well anthropogenic SOA (NMB
In addition, the model performance of meteorological parameters (e.g., temperature, relative humidity, wind speed and direction, and boundary layer
height), oxidants (e.g., OH,
The ratios of seasonal mean concentrations of
Scatterplots of the observed campaign-average mixing ratios of HONO and those simulated by
Figure 2a shows the ratios of seasonal mean OH concentrations simulated by Sp_R1
The remarkable increase in the SOA concentrations in northern China in the Sp_R1
Further improved model performances in Sp_R1
Comparisons of the observed campaign-average mixing ratios of benzene, toluene, and xylene in China with those simulated by the Cp_base simulation. Note that the mixing ratios of these aromatic compounds in the Cp_base, Cp_R1, Cp_R1
For the process-based scheme, anthropogenic aromatics, IVOCs, and SVOCs are uncertain precursors in the model. Figure 4 shows the model–observation
comparison of campaign-average concentrations of benzene, toluene, and xylene in China. The model simulations in general agree with the observations,
with NMBs of
The differences in seasonal mean mass concentrations of SOA between
The inter-comparisons of the process-based and observation-constrained simulations of season-mean mass concentrations of SOA highlight the importance
of improving the IVOC emissions in winter for modeling SOA in China (Fig. 5). The spatial distributions of the SOA concentrations in the Sp_R1
Seasonal mean mass concentrations of OA, POA, and SOA and the mass fractions of POA and SOA simulated by the Cp_R1
Seasonal mean mass fractions of OA components and the sources of POA, SVOCs-SOA, and IVOCs-SOA in NCP, YRD, and PRD in
The Cp_R1
Other model studies that considered the contributions of SVOCs and IVOCs in China also show that SVOCs and IVOCs contribute greatly to the simulated SOA (B. Zhao et al., 2016; Yang et al., 2019; Li et al., 2020). The mass fractions of each SOA component vary among studies. For example, Li et al. (2020) suggested a lower contribution of IVOC-related SOA (IVOCs-SOA) and a higher contribution of aromatic SOA (ASOA) in NCP in winter compared to our study, explained by lower emissions of IVOCs and enhanced formation of ASOA from the aging process in their study. B. Zhao et al. (2016) and Yang et al. (2019) suggested over 50 % contributions of IVOCs-SOA to SOA in all seasons, which is greater than our estimations. Their studies considered the multi-generation oxidation of IVOCs for which the mechanism and parameterization remain unclear. The formation of SOA from IVOCs is the most uncertain part, calling for more constraints from ambient measurements and experiments.
For BSOA, its contribution to total OA is negligible in winter but can increase to 15 % in PRD in summer because of the enhanced emissions of biogenic precursors. The contribution of SOA formed by aqueous-phase ways (aqSOA) is also much greater in summer (9 %–13 %) because high emissions of isoprene enhance the formation of isoprene epoxydiols (IEPOX), glyoxal, and methylglyoxal (Hu et al., 2017). Field observations suggest an important role of aqSOA in SOA formation during the winter haze periods (Kuang et al., 2020; Wang et al., 2021). The simulated mass fraction of aqSOA is only 3 %–5 % in SOA in winter herein, indicating that more precursors are perhaps involved in the SOA formation related to aerosol liquid water than the model has considered (Gkatzelis et al., 2021). The estimated contribution of aqSOA is similar to the estimation of Li et al. (2021) in NCP but is much lower than the estimations made by Qiu et al. (2020) in Beijing and Ling et al. (2020) in PRD. Further improvements of aqSOA simulation may be important for capturing the variability in SOA during the winter haze episodes.
In this study, we applied both process-based and observation-constrained schemes to simulate OA in China. Compared with the PMF results from observations, the model underestimation of SOA mainly occurs in winter in northern China in the default model. Updates to the emissions, volatility distributions, and SOA yields of SVOCs and IVOCs as well as the addition of HONO sources lead to significant model improvements. The bias of SOA simulation is sensitive to the model performance of OH levels and IVOC emissions. For the former, the addition of HONO sources can significantly improve the simulations of surface OH concentrations in winter and increase the simulated SOA concentrations by over 30 % in northern China, highlighting the importance of HONO chemistry in polluted environments. Greater sensitivity of the SOA formation to the OH levels is more present in winter than in summer. For the latter, the most uncertain part of IVOC emissions is from the residential sector, which needs future efforts to constrain.
With all improvements, the two types of SOA schemes show similar seasonal and spatial variations that reasonably agree with the observations. SVOCs and IVOCs are the main contributors to SOA in most of China, with mass contributions of over 50 %. Control measures of primary OC emissions have already been taken since 2014 and resulted in an effective reduction in the POA concentrations in NCP (Lei et al., 2021; Duan et al., 2020). High concentrations of SOA were still observed in NCP during the COVID-19 lockdown period, when the emissions from industry and transportation were largely reduced (Sun et al., 2020b; Zheng et al., 2021). Further reduction should focus on the SVOC and IVOC emissions. The seasonal variations in the OA composition and the source contribution suggest that the control strategies for SOA pollution should vary by season. The model suggests the residential sector as the major source of POA, SVOCs-SOA, and IVOCs-SOA in winter in polluted areas in China. The control of residential emissions may reduce POA and SOA simultaneously besides the reduction in other primary aerosols and secondary inorganic aerosols (Meng et al., 2020). In summer, the industry sector becomes the predominant source of SVOCs- and IVOCs-SOA, which has not yet been effectively controlled in China.
Data presented in this paper are available upon request to the corresponding author.
The supplement related to this article is available online at:
QC and RM designed the study. RM performed the model simulations and conducted the data analysis. LZ and YC provided the emission inventory of ammonia. QC and RM prepared the manuscript with contributions from all co-authors.
The contact author has declared that neither they nor their co-authors have any competing interests.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This work was supported by the National Natural Science Foundation of China (grant nos. 41961134034, 91544107, 41875165, and 51861135102) and the 111 Project of Urban Air Pollution and Health Effects (B20009). Manish Shrivastava was supported by the US DOE, Office of Science, Office of Biological and Environmental Research, through the Early Career Research Program.
This research has been supported by the National Natural Science Foundation of China (grant nos. 41961134034, 91544107, 41875165, and 51861135102) and the Higher Education Discipline Innovation Project (grant no. B20009).
This paper was edited by Qiang Zhang and reviewed by two anonymous referees.