Process-based and Observation-constrained SOA Simulations in China: The Role of Semivolatile and Intermediate-Volatility Organic Compounds and OH Levels

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 to reproduce the variability of OA concentrations in polluted areas, hindering understanding of the OA budget. Herein, we applied both process-based and observation-constrained schemes to simulate OA in GEOS-Chem. Comprehensive data sets of surface OA, OA components, 20 secondary organic aerosol (SOA) precursors, and oxidants were used for model-observation comparisons. In the revised schemes, updates of the emissions, volatility distributions, and SOA yields of semivolatile and intermediate volatility organic compounds (S/IVOCs) were made. These updates are however insufficient to reproduce the SOA concentrations in observations. The addition of nitrous acid sources is an important model modification, which improves the simulation of surface concentrations of hydroxyl radical (OH) in winter in northern China. The increased surface OH concentrations enhance 25 the SOA formation and lead to greater SOA mass concentrations by over 30%, highlighting the importance of having good OH simulations in air quality models. There is a greater sensitivity of the SOA formation to the oxidant levels in winter than in summer in China. 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 S/IVOCs are the dominant source of SOA in China with a contribution of over 30 50%. The residential sector may be the predominant source of S/IVOCs in winter, despite large uncertainty remains 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 S/IVOC measurements are needed to constrain their emissions. https://doi.org/10.5194/acp-2021-628 Preprint. Discussion started: 28 July 2021 c © Author(s) 2021. CC BY 4.0 License.


Introduction
Organic aerosol (OA) is a major component of tropospheric submicron aerosol, which can be directly emitted as primary 35 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 Gani et al., 2019). However, atmospheric chemical transport models (CTMs) have difficulties in reproducing the magnitude and the variability of OA mass in polluted environments, mainly resulting from the underestimation 40 of SOA (Park et al., 2021;Miao et al., 2020;Jiang et al., 2019).
SOA is generally simulated in CTMs by process-based schemes, 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 Tg yr -1 in different models, which has been attributed largely to the uncertain contribution from semivolatile and intermediate volatility organic 45 compounds (S/IVOCs) (Spracklen et al., 2011;Hodzic et al., 2016;Pai et al., 2020). The S/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 etc. are all important sources of S/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 S/IVOCs are estimated by applying empirical scale factors to different proxies such as POA, non-methane 50 volatility organic compounds (NMVOCs), and speciated IVOCs 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 Lu et al., 2020). For IVOCs, some CTMs use one lumped precursor with specific SOA yields 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 volatility of oxidation products and alters gas-to-particle partitioning (Li et al., 55 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 , there is still a lack of source-dependent model frameworks.
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, 60 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). The parameterization is however too generalized to differentiate specific anthropogenic source contributions. In addition, the model performance on atmospheric oxidation capacity may affect the simulation of SOA production and lead to uncertain budget analysis and source https://doi.org/10.5194/acp-2021-628 Preprint. Discussion started: 28 July 2021 c Author(s) 2021. CC BY 4.0 License.

Ambient observations
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 (AMS) and aerosol chemical speciation monitors (ACSM) and 80 covered mains regions in China, including 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), biomassburning-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. 85 Unlike our previous study (Miao et al., 2020), we did not divide the measured concentrations by the empirical submicron-tofine mass ratio because of the lack of such information for different regions and seasons (Y. Sun et al., 2020a). Moreover, we synthesized a dataset 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 flame ionization detector (FID) and/or mass spectrometer (MS). Table S2 in the Supplement lists the sampling information and the 90 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°17′ N, 121°44′ E) from 5 December 2016 to 3 January 2017 and from 16 July to 8 August 2017 was used for comparisons in this study . We also included 28 measurements of HONO from 2011 to 2019 and 10 measurements of OH and hydroperoxy radical (HO2) from 2014 to 2019 in China in the analysis (Tables S3 and S4 in the Supplement)
The model was set for 47 vertical levels from the surface to 0.01 hPa and was driven by the MERRA2 reanalysis assimilated meteorological data. The boundary conditions were generated by global simulations under a horizontal resolution of 2°×2.5°. 100 For computation efficiency, the model simulations for the base year of 2014 were compared with the observations. A recent study shows that the long-term trend of particulate matter is mainly driven by the change of anthropogenic emissions (Zhai et al., 2019). The emissions of nitrogen oxides (NOx), NMVOCs, and organic carbon (OC) changed by −17%, +11%, and −35% during 2011 to 2017 in China (Zheng et al., 2018), suggesting perhaps a minor impact of the inter-annual variability on the model evaluations herein. Common model parameters and emission inventories are described in detail elsewhere (Miao et al., 105 2020).
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 scheme represents the default Complex SOA configuration, in which SOA is produced by the oxidation of lumped biogenic, aromatic, and S/IVOC precursors, heterogeneous uptake of glyoxal and methylglyoxal, and isoprene multi-phase chemistry (Marais et al., 2016;Fisher et al., 2016;. The emissions of 110 SVOCs are treated as 1.27 times of primary OC emissions, and the emissions of IVOCs are set as 66 times of naphthalene emissions . Primary SVOCs are emitted as two tracers with saturation concentrations (C*) of 1646 and 20 μg m -3 (Shrivastava et al., 2006). Once emitted, SVOCs partition to the particle phase to form POA. The remaining gas-phase SVOCs are oxidized by OH with a reaction rate constant of 2×10 -11 cm 3 molec -1 s -1 , which produces two SOA surrogates that have two orders of magnitude lower volatilities compared to their precursors (Grieshop et al., 2009). The 115 organic matter to OC ratios for POA and SOA are 1.4 and 2.1, respectively (Turpin and Lim, 2001). SOA produced by the oxidation of monoterpenes, sesquiterpenes, aromatics, and IVOCs is parameterized by using a VBS approach with NOxdependent SOA yields. Naphthalene is used as a surrogate of IVOCs (Chan et al., 2009). Only photooxidation is considered for aromatics and IVOCs, whereas the oxidations by OH, O3, and nitrate radical (NO3) are all included for monoterpenes and sesquiterpenes . For isoprene, SOA is simulated by the heterogeneous uptake of isoprene oxidation products 120 that are produced under low or high NOx conditions (Marais et al., 2016;Pai et al., 2020). The Sp_base scheme represents the default Simple SOA configuration. Primary OC emissions from the MEIC inventory are treated as non-volatile. The ratios of the emissions of anthropogenic and biomass burning surrogate precursors to CO (EFSOAP/EFCO) are fixed to 0.069 and 0.013, respectively. The SOA yields for isoprene and terpenes are set to be 0.03 and 0.10, respectively. SOA precursors are converted to SOA with a fixed lifetime of one day (Miao et al., 2020;Pai et al., 2020). 125 Modifications on the SOA schemes are listed in Table 1. The Cp_R1 and Sp_R1 schemes have updates on precursor emissions, SOA yields, or parameters related to the production and removal. Specifically, the Cp_R1 scheme applies a more reasonable https://doi.org/10.5194/acp-2021-628 Preprint. Discussion started: 28 July 2021 c Author(s) 2021. CC BY 4.0 License. scale factor of 1.0 for SVOC emissions instead of 1.27 that is used in the Cp_base scheme (Lu et al., 2018). Instead of using two bins for all sources, the volatility distributions of SVOCs emissions are specified for transportation, other anthropogenic sources, and biomass burning and contain five bins with C* of 10 -2 to 10 2 μg m -3 ( Figure S1 in the Supplement), which have 130 lower volatilities compared with the default distribution in the Cp_base scheme (Zhao et al., 2015;May et al., 2013b;May et al., 2013a). The updates on the emissions and SOA yields of IVOCs are described in detail in Sect. 2.3. Additionally, the scavenging efficiency of POA in wet deposition is set to be 50% instead of 0% (Shah et al., 2019). In the Sp_R1 scheme, an OH-dependent oxidation rate of SOA precursors is used for the daytime simulations, which applies a rate constant of 1.25×10 -11 cm 3 molec -1 s -1 instead of a fixed rate of 1.2×10 -5 s -1 (Hodzic and Jimenez, 2011). For the nighttime simulations, a fixed 135 oxidation rate of 2.5×10 -6 s -1 is used instead of 1.2×10 -5 s -1 .
The Cp_R1+2 and Sp_R1+2 schemes aim at improving the OH simulation upon the Cp_R1 and Sp_R1 configurations. The GEOS-Chem model underestimates daytime surface OH concentrations in Beijing (Miao et al., 2020), which is partially driven by inadequate HONO sources. In the default model, HONO is produced by the gas-phase reaction of NO with OH as well as the heterogeneous reaction of nitrogen dioxide (NO2) on aerosols. We first revised the heterogeneous uptake coefficient of 140 HO2 (γHO2) on aerosols from 0.2 to 0.08 as suggested by Tan et al. (2020), and then added additional HONO sources in the model ( Table S5 in the Supplement). Specifically, the HONO emissions from traffic sources (EHONO, traffic) are estimated as 1.7% of the traffic NOx emissions (ENOx, traffic) (Rappengluck et al., 2013), which can reproduce well the diurnal cycle of HONO concentrations in urban environments (Czader et al., 2015). The emissions from soil (EHONO, soil) are estimated from the soil NOx emissions (ENOx, soil) by applying scale factors that depend on biomes and soil water content (Hudman et al., 2012;Oswald 145 et al., 2013;Rasool et al., 2019). The HONO emissions from biomass burning are calculated on the basis of the burned areas provided by the Global Fire Emission Database (GFED4) and combustion-type dependent emission factors (Giglio et al., 2013;Andreae, 2019). Moreover, the heterogeneous reaction of NO2 on the ground is added to the surface layer of the model. The reaction rate (kg) depends on the mean molecular speed of NO2 (υNO2), the ground surface-to-volume ratio (Sg/V), and the uptake coefficient of NO2 on the ground (γg-NO2) (Li et al., 2010). The Sg/V is set to be 0.1 m -1 for urban areas (Vogel et al., 150 2003) but varies by the leaf area index and the height of the boundary layer in non-urban areas (Sarwar et al., 2008). The γg-NO2 value is set to be 10 -6 for nighttime (Kurtenbach et al., 2001) and 2×10 -5 multiplied by a photo-enhancement scale factor associated with the photolysis rate of NO2 (JNO2) for daytime (J. . In addition, the photolysis of nitrate is considered. The photolysis rate (Jnitrate) is set to be 100 times the photolysis rate of HNO3 (JHNO3) with a HONO molar yield of 0.67 (Kasibhatla et al., 2018). Finally, in the Cp_R1+2+3 and Sp_R1+2+3 schemes, we tested the impacts of potentially 155 underrepresented heating-season emissions of SOA precursors from the residential sector upon the previous modifications.
The IVOC emissions from the residential sector during November to March are multiplied by 7 in the Cp_R1+2+3 scheme according to the observed IVOC concentrations. In the Sp_R1+2+3 scheme, the value of EFSOAP/EFCO is updated from 0.069 to 0.080 for anthropogenic emissions during November to March. The factor of 0.080 has been used in other model studies for urban plumes (Shah et al., 2019). 160 https://doi.org/10.5194/acp-2021-628 Preprint. Discussion started: 28 July 2021 c Author(s) 2021. CC BY 4.0 License.

Emissions and SOA yields of IVOCs
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).  Table S7 in the Supplement lists the annual emissions of IVOC6, IVOC5, and IVOC4 in 2014. Industry and residential sectors are the major sources of IVOCs in China. The reaction rate constant with OH for these IVOC species used in the model is 2.3×10 -11 cm 3 molecule -1 s -1 at 298 K which is the same with the rate constant of naphthalene photooxidation (Chan et al., 2009). Table S8 in the Supplement lists the SOA yield parametrizations of IVOCs used in this study. For high-NOx condition, 175 mass-weighted yields of the photooxidation of C12-C14, C15-C16, and C≥17 n-alkanes are used for IVOC6, IVOC5, and IVOC4, respectively (Presto et al., 2010;Zhao et al., 2015). For low-NOx condition, a fixed yield of 0.73 obtained from naphthalene photo-oxidation is applied to all IVOCs because of the lack of low NOx yields for n-alkanes (Chan et al., 2009). The corresponding IVOC yields for 10 μg m -3 OA range from 0.19 to 0.44, which are greater than the yields in the Cp_base scheme but within the range of the yields used in other studies Koo et al., 2014;Jathar et al., 2014;Lu et al., 180 2020). Table 2 lists the total IVOC emissions estimated in various studies. Globally, the IVOC emissions range from 16.0 to 234 Tg yr -1 for which the POA-based methods have the highest estimates and the naphthalene-based methods have the lowest Jathar et al., 2011;Shrivastava et al., 2015;Hodzic et al., 2016). Our new NMVOC-based method suggests a global emission of 32.2 Tg yr -1 and an emission of 6.6 Tg yr -1 in China that is similar to the POA-based estimate made by 185 Wu et al. (2021). The spatial distribution of IVOC emissions shows that the most increase of the new NMVOC-based emission occurs in urban areas compared with the naphthalene×66 and the POA×1.5 estimates of other models ( Figure S3 in the Supplement). The POA×1.5 estimate of IVOC emissions has a greater winter-summer emission difference compared with the naphthalene×66 and the new NMVOCs-based emissions ( Figure S4 in the Supplement). The additional increase of IVOC emissions in the Cp_R1+2+3 scheme (i.e., 7 times of the residential IVOC emissions during the heating season) leads to a 190 large emission enhancement in northern China ( Figure S3) and a greater winter-summer emission difference than that in the Cp_R1+2 scheme ( Figure S4), which agrees better with the PMF-derived SOA results (Sect. 3).

Results and discussion
The model performance on meteorological parameters (e.g., temperature, relative humidity, wind speed and direction, and boundary layer height), oxidants (e.g., OH, O3 and NO3), and aerosol precursors in GEOS-Chem have been evaluated 195 elsewhere (Miao et al., 2020) (Zhang et al., 2007;Li et al., 2017). The highest OA concentrations occurred in winter in northern China, corresponding to high POA fractions that may go over 50% at some urban sites. In particular, residential solid fuel consumption emits a large amount of POA and SOA precursors and stagnant meteorological 205 conditions often happen in winter, leading to severe haze in northern China Peng et al., 2019). The OA concentrations are typically low in summer when meteorological conditions favor particles dilution and deposition and in southern China where primary contributions are less than in northern China. The SOA fractions are generally high in southern China (above 65%), which may be explained by low primary emissions and high oxidation capacity that leads to fast conversion of organic vapors to SOA . The lowest OA concentrations were observed in remote regions (e.g., 210 Tibetan Plateau) in summer, representing natural background conditions in China.
The statistical values such as normalized mean bias (NMB), normalized mean error (NME), root mean square error (RMSE), and Pearson's correlation coefficient (R) for the model-observation comparisons of campaign-average concentrations of OA, POA, and SOA are listed in Table 3. As is consistent with our previous results (Miao et al., 2020) Because a fraction of aerosol particles may present in the supermicron domain that cannot be detected efficiently by AMS or ACSM (Sun et al., 2020a), such model underestimation of OA can be greater in certain circumstances, e.g., in northern China under winter-haze conditions. By contrast, the Sp_base simulation may reproduce the OA loadings (NMB = −0.14). The POA simulations are improved in the Sp_base (NMB = −0.18) and Cp_R1 cases (NMB = −0.11). The Sp_base scheme considers primary OC as non-volatile, for which the model results agree with the PMF-derived POA results at urban sites but significantly 220 overestimate the POA concentrations in suburban and remotes regions ( Figure S5 in the Supplement). The Cp_R1 scheme considers primary OC as semivolatile with lower volatility distributions compared with the Cp_base scheme, leading to more OC mass in the particle phase as emitted (i.e., "POA"). This scheme is slightly better than the Sp_base scheme for POA but still overestimates its concentrations in suburban and remotes regions.    (Table   S3) and improve NMB from −0.58 in the base simulations to −0.14 ( Figure 3). The addition of HONO sources can increase the surface mean OH concentrations by a factor of 2 to 4, especially in winter in northern China when the photolysis of HONO contributes predominantly to the primary production of OH (Tan et al., 2018;Slater et al., 2020). Table S4 lists the observed and modeled surface OH and HO2 concentrations in China. The modified HONO sources significantly improve the simulations 250 of peak concentrations of OH and HO2 in winter, which improves the SOA simulations. As shown in Figure 2b, the increased OH concentrations lead to greater SOA concentrations nationwide. In particular, the increase can be over 30% in northern China, suggesting that the SOA simulation is more sensitive to the OH simulation in northern China than in southern China.
Consistently, a recent study suggests that enhanced OH levels likely promote fresh SOA formation in northern China but increases the oxidation state of OA in southern China (J. Li et al., 2019b). In summer, the SOA mass enhancements mainly 255 occur in the near-source regions in the Sp_R1+2 simulations. Although the OH and HO2 concentrations in summer in southern and southwestern China are overestimated in the Sp_R1+2 and Cp_R1+2 simulations (Table S4) impact on the model-observation comparisons of SOA herein. Among the added HONO sources, the heterogeneous reaction of NO2 on the ground contributes predominantly to the enhancements of surface HONO and OH concentrations, which is consistent with the results from budget analysis of ambient observations (Xue et al., 2020;Liu et al., 2019;Huang et al., 2017). 260 The greatest enhancements of OH concentrations therefore occur in urban areas where high NOx emissions and large Sg/V facilitate the heterogeneous formation of HONO. The model parameters such as γg-NO2 and the HONO yield vary significantly by relative humidity, light intensity, and NO2 concentrations and are associated with large uncertainties (C. Han et al., 2016;, which requires more future observations to constrain. Updates in Cp_R1+2+3 and Sp_R1+2+3 aim at increasing the emissions of anthropogenic SOA precursors. For the process-265 based schemes, anthropogenic aromatics, IVOCs, and SVOCs are uncertain precursors in the model. Figure  Measurements of SVOCs and IVOCs are rare (Y. . Table S9 in the Supplement lists the observed and simulated 270 campaign-average concentrations of primary IVOCs in China. The Cp_R1+2 simulation largely underestimates the IVOC concentrations, especially in winter. The underestimated IVOC emissions are likely from the residential sector that has highly uncertain emission activity (Tao et al., 2018;Peng et al., 2019;J. Li et al., 2019a). The emission factors of IVOCs from residential combustion vary in a wide range and are sensitive to the fuel types and combustion conditions (Cai et al., 2019;Qian et al., 2021). We tested seven-fold IVOC emissions from the residential sector in the Cp_R1+2+3 simulation to eliminate 275 the potential seasonal bias of IVOC emissions. The simulation-to-observation ratio of primary IVOC concentrations in winter became 0.44 that is similar to the ratio in summer. The Cp_R1+2+3 simulation indeed improves the winter SOA simulations at urban sites significantly and reduce NMB from −0.55 in Cp_R1+2 to −0.28 ( Figure S6a).
For the observation-constrained schemes, the emissions of SOA precursors depend on the emissions of CO. The observed CO concentrations in China are generally greater than the modeled surface concentrations, indicating possibly underestimated CO 280 emissions especially in winter (Kong et al., 2020). Consistently, top-down estimates suggest greater CO emissions than those in the MEIC inventory (X. Feng et al., 2020;Gaubert et al., 2020). On the other hand, recent measurements of SOA formation potential show a wide range of reference values for EFSOAP/EFCO (Table S10 in the Supplement). The fixed EFSOAP/EFCO ratios used in the model may not fully represent ambient conditions . In the Sp_R1+2+3 simulation, we applied a higher value of EFSOAP/EFCO (i.e., 0.08 instead of 0.069) for all anthropogenic sources during the 285 heating season. This modification increases the SOA concentrations in winter at urban sites and reduces NMB from −0.26 in Sp_R1+2 to −0.15 ( Figure S6a). Some overestimation occurs at suburban sites ( Figure S6b improving the estimation of IVOC emissions in China for modeling SOA. Overall, the revised process-based schemes have greater biases than the revised observation-constrained ones at urban sites ( Figure S6). Figure 6 shows the concentrations of OA, POA, and SOA as well as the mass fractions of POA and SOA simulated in the Cp_R1+2+3 simulation. The POA concentrations are several times greater in winter than in summer because of higher emissions of S/IVOCs as well as low temperature and high OA concentrations that favor the gas-to-particle partitioning of 305 organic vapors. The seasonal difference of SOA is smaller than that of POA. One explanation is the enhanced formation of biogenic SOA (BSOA) in summer. SOA is the dominant component of OA in summer that contributes over 60% of OA nationally, whereas POA contributes more than SOA in winter in northern China. The Cp_R1+2+3 scheme represents our best-estimate scenario for process-based simulations that capture well the seasonal and spatial patterns of OA and the split of POA and SOA in the observations (Figure 1a,b). Figure 7 shows the corresponding OA compositions in different regions as 310 well as their sources. The SOA mass concentrations are dominated by anthropogenic sources, among which S/IVOCs contribute over 50% in the three regions. The contribution of SVOCs to SOA depends on season. In summer, SVOC-related SOA (SVOCs-SOA) is the largest OA component in all regions, for which residential and industry sectors are the main sources.
The contributions of IVOCs to OA are generally over 15% in which industry is the predominant contributor. In winter, the residential sector is the major source of S/IVOCs. SVOCs-SOA contributes less to OA than in summer because SVOCs 315 favorably form POA at low temperatures.
Other model studies that considered the contributions of S/IVOCs in China also show S/IVOCs contribute greatly to the simulated SOA (B. Yang et al., 2019;Li et al., 2020). the multi-generation oxidation of IVOCs for which the mechanism and parametrization remain unclear. 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 (9-325 13%) because high emissions of isoprene enhance the formation of 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  in NCP but is much lower than the 330 estimations made by Qiu et al. (2020) in Beijing and Ling et al. (2020) in PRD.

Conclusions
In this study, we applied both process-based and observation-constrained schemes to simulate OA in China. sector however remain largely uncertain, which needs future efforts to constrain. The control of residential emissions may reduce POA and SOA simultaneously besides the reduction of other primary aerosols and secondary inorganic aerosols (Meng et al., 2020). Control measures of residential emissions have already been taken since 2017 Duan et al., 2020).
High OA concentrations 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), indicating the need for further reduction of 350 residential emissions. In summer, the industry sector becomes the predominant source of S/IVOCs-SOA, which has not yet been effectively controlled in China.
Data availability. Data presented in this manuscript are available upon request to the corresponding author.