Field observational constraints on the controllers in glyoxal (CHOCHO) loss to aerosol

Glyoxal (CHOCHO), the simplest dicarbonyl in the troposphere, is an important precursor for secondary organic aerosol (SOA) and brown carbon (BrC) affecting air-quality and climate. The airborne measurement of CHOCHO concentrations during the KORUS-AQ (KORea-US Air Quality study) campaign in 2016 enables detailed quantification of 30 loss mechanisms, pertaining to SOA formation in the real atmosphere. The production of this molecule was mainly from oxidation of aromatics (59%) initiated by hydroxyl radical (OH), of which glyoxal forming mechanisms are relatively well constrained. CHOCHO loss to aerosol was found to be the most important removal path (69 %) and contributed to roughly ~20 % (3.7 μg sm ppmv hr, normalized with excess CO) of SOA growth in the first 6 hours in Seoul Metropolitan Area. To our knowledge, we show the first field observation of aerosol surface-area (Asurf)-dependent CHOCHO uptake, which 35 diverges from the simple surface uptake assumption as Asurf increases in ambient condition. Specifically, under the low (high) aerosol loading, the CHOCHO effective uptake rate coefficient, keff,uptake, linearly increases (levels off) with Asurf, thus, the irreversible surface uptake is a reasonable (unreasonable) approximation for simulating CHOCHO loss to aerosol. https://doi.org/10.5194/acp-2021-672 Preprint. Discussion started: 31 August 2021 c © Author(s) 2021. CC BY 4.0 License.

In the presence of inorganic salts such as ammonium sulfate (AS) and ammonium nitrate (AN), CHOCHO's 65 solubility in aerosol increases several orders of magnitude compared to solubility in pure water, which is known as "saltingin" effect (Kampf et al., 2013). This can lead to a large fraction of CHOCHO being oxidized in the aqueous phase, thus resulting in less volatile products (Knote et al., 2014;Waxman et al., 2015;Sareen et al., 2017;Ling et al., 2020). In addition, CHOCHO was observed to have high solubility to some water soluble organic acid aerosol seeds such as fulvic acid, l-tartaric acid and dl-malic acid (Corrigan et al., 2008;Volkamer et al., 2009;Kampf et al., 2013). More recently, more 70 complex controls such as the acceleration of CHOCHO uptake processes in the presence of other gaseous organic species have been reported (Qin et al., 2020). Overall, most of these studies have concluded that CHOCHO has a significant role in SOA formation. However, the solubility of CHOCHO in ambient condition is highly uncertain since the studies that constrained CHOCHO solubility in ambient conditions are rare. Many recent regional and global modelling studies (Fu et al., 2008;Knote et al., 2014;Hu et al., 2017;Stadtler et 75 al., 2018;Bates and Jacob, 2019;Qiu et al., 2020;Xu et al., 2020), as well as several studies constrained by in-situ field observations (Volkamer et al., 2007;Washenfelder et al., 2011;Li et al., 2016;Ling et al., 2020), have reported that CHOCHO contributes 0 to 25 % to SOA formation. This large variability resulted from the different conditions of the individual studies, focusing on different VOC precursors, biogenic (Li et al., 2016;Xu et al., 2020) vs anthropogenic (Volkamer et al., 2007;Qiu et al., 2020), and using different simulation tools, such as, 0-D box (Volkamer 80 et al., 2007), Lagrangian (Washenfelder et al., 2011), and 3-D transport models (Fu et al., 2008;Knote et al., 2014;Chan Miller et al., 2016;Li et al., 2016;Chan Miller et al., 2017;Sareen et al., 2017;Hu et al., 2017;Qiu et al., 2020;Xu et al., 2020). These model studies were run with (Washenfelder et al., 2011) or without (Volkamer et al., 2007;Knote et al., 2014) steady-state assumption for CHOCHO, and employed different aerosol uptake treatments such as surface uptake only (Washenfelder et al., 2011;Li et al., 2016) or allowing simultaneous volume control processes (Knote et al., 85 2014;Ling et al., 2020).
To add more constraints on the CHOCHO contribution to SOA formation, we investigated CHOCHO partitioning to aerosols from an in-situ airborne campaign over the Korean Peninsula, where the emissions of CHOCHO-producing VOCs are highly variable. Two independent methods -a steady-state approach using a 0-D box model and a hybrid method (Knote et al., 2014) using semi-explicit aqueous chemistry together with surface uptake were used and compared for 90 investigating more details in uptake processes of CHOCHO. From this work, we found that the simple surface uptake approximation is only applicable for low aerosol loading circumstance during the KORUS-AQ campaign. The dependence of these processes on photochemical activeness, together with evidence of insufficient understanding on CHOCHO solubility, especially in low inorganic salts condition, are discussed.

Materials and methods 95
We investigated the CHOCHO contribution to SOA formation (hereinafter, glySOA) via comparison of direct observations with an estimate from a diurnal steady-state 0-D box model, constrained by in-situ airborne precursor measurements. The magnitude of the evaluated aerosol enhancement owing to CHOCHO uptake is then compared to the results from the hybrid scheme similar to Knote et al. (2014). Details on the data treatment, model settings as well as the analysis methods follow.
In addition, aerosol data reported here are at standard temperature and pressure (273 K and 1013 hPa) condition, 100 leading to the unit of µg sm -3 where sm 3 refers to the standard volume in cubic meters. Moreover, the median and interquartile range (25-75 %) were used to represent the variabilities in distributions, except as otherwise indicated.

Data description
We used airborne measurements of CHOCHO, taken as part of the KORUS-AQ 2016 (KORea and United States -Air Quality study 2016), conducted from 1 May to 10 June, 2016, over South Korea, under the close collaboration among 105 scientists from South Korea, U.S.A, and other countries to understand the local, regional, and hemispheric influences on emissions, chemistry, and air quality (Crawford et al., 2021). The entire KORUS-AQ 2016 dataset acquired from various measurement platforms (i.e., aircraft, research vessel, satellite and ground site) is available from the NASA data archive (https://www-air.larc.nasa.gov/missions/korus-aq/). The observational data used in this analysis were taken aboard the NASA DC-8 for research flights (RFs) where CHOCHO measurements are available (18 out of 20 RFs except RF # 01 and # 110 03, details are in Table S1 in supporting information, SI). We only focused on data that fall in the geographically confined area of the Korean peninsula and its coastal region (latitude: 34.3-38°N, longitude: 125.8-129°E), as shown in Fig. 1a, to investigate production and loss mechanisms of CHOCHO under the influence of inland sources in Korea (Peterson et al., 2019).
The list of chemical species as well as physical parameters used in our analysis, together with their measurement 115 techniques, are summarized in Table 1. Here, we only provide brief descriptions of the CHOCHO measuring system, CAvity Enhanced Spectrometer for Atmospheric Researches (CAESAR), based on the same measurement principle and layout as the CHOCHO instrument at NOAA (National Oceanic and Atmospheric Administration) . Air is drawn into the system via a coaxial overflow inlet (1/2" and 1/4" O.D. PFA tubing, total length: < 1.5 m) through a Teflon membrane filter (2 μm pore) to keep the optics clean and to avoid light attenuation by aerosol scattering. The 455 nm centered cavity 120 consists of an LED as a light source, collimating optics, and high-reflection mirrors (reflectivity: 99.9973%, effective light path length: up to 12 km) enable us to quantify CHOCHO and NO 2 with zeroing error of 34 and 80 pptv for 5-second averages (2σ) and an accuracy of 5.8, and 5.0 %, respectively, using the custom-built retrieving algorithm of the DOASIS software (Kraus, 2006) (fitting range: 438 -468 nm). To ensure the best instrumental performance, cavity calibrations were performed every 5 minutes by sequential overflows of He and zero air (30-second injection for each). 125 To analyze the magnitude of glySOA, we only focused on measurements below 2 km altitude which include the boundary layer in most cases, where the CHOCHO abundance was higher than CAESAR's detection limits. Also, CHOCHO data lower than zeroing error were removed to reduce uncertainty in the calculation of glySOA. In addition, to constrain the 0-D box model with DC-8 observations, 1-minute merged data (version R6, DOI: 10.5067/Suborbital/KORUSAQ/DATA01) were used for the measurement parameters with high time resolution. Meanwhile, for the species with coarser time 130 resolution than 1-minute, (i.e., WAS data), a pseudo-1-minute data generation scheme, similar to that of Schroeder et al. (2020), was applied based on the measurements of the PTR-TOF-MS (for aromatics and isoprene) and CAMS (for other hydrocarbon compounds such as alkanes, alkenes, and alkyne) to match the timestamps; the detailed interpolation method is described in section S2, Fig. S1, and Table S2 in the SI. For the comparison of oxalate and OA, synchronized timestamps with SAGA filter measurements were used. 135

Model description
The Framework for 0-Dimensional Atmospheric Modeling (F0AM v3.2)  was used as our tool to simulate CHOCHO with the Master Chemical Mechanism v3.3.1 (Jenkin et al., 2015 and references there in). We constrained our model with 50 chemical species and 10 physical parameters as listed in Table S2 in SI. The total number of chemical reactions considered in our model is over 11,325 with 3,742 chemical species tracked. 140 The model was initialized with observations from the aircraft, and the oVOCs (oxidized VOCs), including CHOCHO, were produced under co-evolving irradiation with fixed parent VOCs concentrations for a day at 1-minute time resolution. The photolysis frequencies evolve in real-time over the course of model steps, similar to other aircraft observations constrained box modeling works (Olson et al., 2006;Kaiser et al., 2016;Anderson et al., 2017;Marvin et al., 2017;Brune et al., 2018;Haskins et al., 2019;Brune et al., 2020;Souri et al., 2020). Specifically, the model was run for one 145 full day, with updated photolysis frequencies (js) every 1 hour, by implementing calculated solar zenith angle (SZA) and scaled with measured photolysis frequencies from the CAFS (CCD Actinic Flux Spectrometers).

Parametrizations of CHOCHO loss by SOA formation
From the F0AM results, the evolution of CHOCHO was calculated from the balance of production (P Gly ) and loss (L Gly ) rates, as described in Eq. (1). 150 =P Gly,mod -L Gly,mod (1) L Gly,mod =L Gly+OH +L Gly+hv +L met = (k Gly+OH [OH] mod +j Gly +k dil ) [Gly] mod (2) L Gly,mod only includes the losses of OH oxidation (L Gly+OH ), photolysis (L Gly+hν ), and physical processes to mimic deposition and dilution (L met ), as expressed in Eq.
(2). Thus the [Gly] mod stands for the simulated CHOCHO concentration without aerosol loss. The loss rates of individual paths in the model were calculated as the sum of the corresponding reaction rate 155 coefficient k Gly+OH and photolysis frequency j Gly , with physical mixing rate representing the dilution, deposition, and transport (k dil , set as a day by following Wolfe et al., 2016).
The loss rate of CHOCHO via aerosol formation (L Gly,aerosol ) was then analytically quantified under steady-state assumptions using modeled production and loss rates of CHOCHO. CHOCHO is a short-lived intermediate oVOC with a lifetime of a few hours (even without considering the aerosol loss path, see Sect. 3.3), thus, the steady-state assumption is 160 valid except for measurements near sources. Under the steady-state assumption, the production rates of CHOCHO (P Gly,mod ), directly extracted from F0AM, need to be balanced with real loss rates of CHOCHO (L Gly,real ), accounting for L Gly,aerosol , as Eq. (3) shows. P Gly,mod ≈L Gly,real = (k Gly+OH [OH] mod +j Gly +k dil ) [Gly] obs +L Gly,aerosol Here, the instantaneous OH oxidation, photolysis, and meteorological losses (k Gly+OH [OH] mod +j Gly +k dil ) were inferred from 165 modeled outputs as L Gly,mod /[Gly] mod .
[Gly] obs refers to the measured CHOCHO concentration. L Gly,aerosol =P glySOA, eff. Steady State= P Gly,mod -(k Gly+OH [OH] mod +j Gly +k dil ) [Gly] obs =k eff,uptake [Gly] obs (4) One should note that the CHOCHO uptake rate coefficient assessed as in Eq. (4) is regarded as a pseudo-first order, effective uptake rate coefficient (k eff,uptake ), which is a net result of competition between reversible and irreversible processes in real atmosphere. This stems from the underlying mechanisms of CHOCHO uptake on the aerosol either being reversible or not, 170 which depends on various factors, such as aerosol acidity, seed type, RH and radiative flux, etc (Liggio et al., 2005a(Liggio et al., , 2005bKroll et al., 2005;Corrigan et al., 2008;Kampf et al., 2013;Curry et al., 2018).
Then, the SOA production rate, owing to CHOCHO uptake (P glySOA ), was estimated from the inferred k eff,uptake and observed CHOCHO concentration. In Sect. 3.6, we tracked the importance of P glySOA in organic aerosol (OA) growth as the air mass evolves with the photochemical age, (Photo age , hr), based on the NO x /NO y chemical clock, by following Kleinman 175 et al. (2003), via Eq. (5).
To reconcile the deviation of diurnal steady-state assumption against the reality, photochemical age-dependent adjustment analysis was also carried out to account for the potential underestimation or overestimation in oVOCs production with respect to the plume age. Fresh plumes close to the sources tend to have not enough time for oxidation products to build up as the simulation of a full-day evolution. Meanwhile, old plumes tend to be over-processed than just one day. For that, an 185 adjustment factor was introduced from the relationship between the ratio of modelled and measured HCHO against Photo age .
This adjustment was applied to the CHOCHO production rate; more details on the discussions and results are in section S3 and Fig. S2 to S4. The main results with this adjustment are presented together with that of the original simulation without this adjustment.

Semi-explicit estimation of CHOCHO loss to aerosol 190
To provide more constraints on aqueous phase reactions once CHOCHO is taken up by aerosol as well as to test the validity of our steady-state assumption in glySOA estimation, a hybrid approach, via semi-explicit aerosol loss treatment of volume processes together with surface uptake treatment, was conducted similar to Knote et al. (2014) (section S9). Details on the backgrounds and parameterizations are described in Knote et al. (2014). Briefly, for the surface processes, first-order reactive uptake treatment was used to compare with F0AM results, under the assumption of irreversible surface reaction in 195 free molecular regime without Fuch-Sutugin correction (Tang et al., 2015;Seinfeld and Pandis, 2016) which corrects for slip correction in the transition regime. This is not only due to its negligible influence (< 5 % difference in k surf,uptake ) but also due to retain the linear relation between k surf,uptake and reactive CHOCHO uptake coefficient (γ Gly,uptake ), as described in Eq. (6).
Here, A surf refers to the aerosol surface area density (cm 2 cm -3 ), v Gly stands for the mean molecular speed (cm s -1 ), and 200 γ Gly,uptake (unitless) indicates the uptake probability of a CHOCHO molecule colliding and reacting on an aerosol surface.
For simulating volume-controlled processes through aqueous-phase reactions, it was assumed that gas-phase and dissolvedphase CHOCHO (CHOCHO monomer, hydrates, and oligomers) are in steady-state (more details in section S4 and Fig. S6 in SI). Briefly, the equilibrium concentration of aqueous phase CHOCHO, including both hydrates and oligomers, was calculated using effective Henry's law coefficient, K H,eff_Gly , as in Eq. (7) from Kampf et al. (2013), and reflecting the 205 parameterization from Waxman et al. (2015) with the kinetic limit of "salting-in" effect at high salt concentration.
K H,eff_Gly = K H,water Gly 10 -0.24× min(12,C AS )-0.07×C AN (7) Here, the parameters in parenthesis stand for the adjustment factor for K H,water_Gly , Henry's law coefficient of CHOCHO in dilute water. C AS and C AN denote the molarities (mol kg -1 , m) of AS and AN in aerosol, respectively. In this study, salting constants of 0.24 and 0.07 were applied for AS and AN, respectively, by following Waxman et al. (2015). In addition, the 210 kinetic limit of 12 m was set only for the concentrations of AS (Kampf et al., 2013), since that of AN has not been experimentally demonstrated yet.
For the instantaneous equilibrium partitioning between CHOCHO monomers and oligomers, an oligomerization constant (K olig =[Gly] oligomer /[Gly] monomer+hyrate ) of 0.5 (1) was used for the cases where C AS is larger (smaller) than the limit of salt concentration, 12 m (Knote et al., 2014). In addition, aqueous phase OH oxidation path was treated as an effective 215 photochemical reaction with the rate of CHOCHO monomer and hydrates (k photochem , 2 s -1 ) adopted from Ervens and Volkamer (2010), since [OH] aq inferred from measured gas-phase OH under the assumption of equilibrium state (K H,OH = 25 M atm -1 ) (Klaning et al., 1985) is highly uncertain; estimated [OH] aq with K H,OH does not only account for losses of OH in aerosols (Herrmann et al., 2010;Waxman et al., 2013) but also for additional radical generation from photochemical reactions of organics (Volkamer et al., 2009;Monge et al., 2012;Tong et al., 2016) and inorganics (Paulson et al., 2019). 220 Aerosol physical sizes measured from LARGE (Langley Aerosol Research Group Experiment) are dried aerosol sizes. To estimate the ambient diameter of an aerosol, and thus estimate more realistic aerosol surface area density, the hygroscopic growth factor (gf diam ) following Brock et al. (2016) was used as described in Eq. (8).
Here, the hygroscopicity parameter (κ chem ) for a mixed particle was calculated from the volume-weighted average of its 225 individual chemical components (OA, NH 4 NO 3 , (NH 4 )HSO 4 and (NH 4 ) 2 SO 4 ) and most of the hygroscopicity were taken from Brock et al. (2016) and references therein, but OA. The hygroscopicity parameter of OA (κ OA ) was parameterized with O/C as 0.190×(O/C)-0.0048 by following Rickards et al. (2013).
C AS and C AN in Eq. (8) are calculated from the sulfate, nitrate and aerosol liquid water content (ALWC) calculated from Extended Aerosol Inorganic Model (E-AIM) (Clegg and Brimblecombe, 1990;Clegg et al., 1998;Massucci et al., 230 1999;Wexler, 2002;Friese and Ebel, 2010). In the model, the amount of each chemical component in aerosol was calculated from the measurements of gas-phase HNO 3 , particle-phase ammonium, nitrate, and sulfate as well as local environmental parameters (e.g., RH and temperature). Inorganic aerosol mass from the AMS and gas-phase HNO 3 from CIT-CIMS were used to run the E-AIM model (specifically, Model IV, in forward mode with ammonia being estimated recursively) (Clegg et al., 1998;Friese and Ebel, 2010). ALWC associated with inorganic compounds was also calculated from the E-AIM model. 235 When calculating ALWC, organic portion was added if O/C (oxygen to carbon ratio of OA) was higher than 0.7, assuming organic phase separation from inorganics at O/C less than 0.7 (Song et al., 2018(Song et al., , 2019Gorkowski et al., 2020).

Observed spatial distribution of CHOCHO and relevant species
The average CHOCHO concentration over the whole campaign period was 107 ± 58 ppt, with the highest concentration of States (up to 1.0 ppb within boundary layer) (Li et al., 2016). The spatial distribution of CHOCHO (Fig. 1b) is somewhat different from other oVOCs, specifically, HCHO, another ubiquitous oVOC (Fig. 1c). The CHOCHO enhancement over the WCIA is less comparable to CO than that of HCHO, where VOCs such as ethane, propane, ethene, and n-butane emissions are dominant in this region (Simpson et al., 2020). Meanwhile, aromatic distributions have more similarities with CHOCHO, such as toluene (Fig. 1f), xylene, and trimethylbenzenes near SMA (Fig. S7). 250 Figure 2 shows the direct comparisons of chemical species estimated from the F0AM to the measurements taken on board.

0-D box model results
Unless otherwise specified, the least-orthogonal-distance regression (ODR) forced through zero intercept method is used for linear regression fits and Pearson R is for the goodness of the fits. In general, the diurnal steady-state box model simulates the gas-phase oxidation processes reasonably well considering the measurement uncertainties as shown for OH, HO 2 , and 255 HCHO (Fig. 2a-c). In contrast to HCHO, where the model shows broad agreement (an overall slope of 1.23 ± 0.02), the CHOCHO model overestimates by 3.24 (± 0.06) times (Fig. 2d). For the model run without photochemical age dependent adjustment, HCHO and CHOCHO regression slopes are 1.08 (± 0.01, Fig. S2c-d) and 2.89 (± 0.05). This model-observation discrepancy of CHOCHO was also observed in other urban environments such as Mexico City (Volkamer et al., 2007) and the Pearl River Delta (Ling et al., 2020). 260 We presume that our model overestimation is due to underestimation in CHOCHO loss, since the CHOCHO instrumental loss along the airway had been identified as negligible (0.0001±0.005 cm 2 s -1 %)  and oVOC production in the model is comparable to that of reality; the modelled OH reactivity is only 8 % (on average) higher than measurements (Fig. S5d), in addition to the fact that the constrained parent VOCs from the combined observations of WAS, PTR-TOF-MS, and CAMS were quite comprehensive. 265 Moreover, the direct comparison between modelled and measured CHOCHO neglects the possible contribution of primary CHOCHO emission in the measurements. Qiu et al. (2020) showed the importance of primarily emitted CHOCHO in Beijing in a wintertime study, finding direct CHOCHO emissions from vehicles and industrial activities. Since we assumed that all measured CHOCHO was secondarily formed, we cannot rule out the possibility of an even larger discrepancy in model and observation, especially for the SMA and industrial area where direct CHOCHO emission is 270 suspected to be not negligible. We prefer to provide a conservative glySOA by assuming all measured CHOCHO to be secondary, since the portion of primary CHOCHO in aircraft measurements is largely uncertain. However, considering vertical transport time with respect to the short lifetime of CHOCHO, most of the measured CHOCHO aboard is expected to be secondary.
To test the hypothesis that the gap between modelled and observed CHOCHO can be explained by uptake to 275 aerosol, a steady-state approach was used (as described in Sec. 2.3) to infer k eff,uptake . As a quality check, we plugged k eff,uptake parameter back into the model to simulate the loss of CHOCHO via glySOA path. The CHOCHO prediction with updated loss to aerosol reproduced measured CHOCHO well, as shown in Fig. 2d, thus we presume that the possible bias of primary CHOCHO contribution is not significant. Possible bias in k eff,uptake due to the measurement errors was quantified by varying input concentrations with their uncertainties. As shown in Fig. S8 in section S6, the largest variability in VOCs input results 280 in ±10 % variation of k eff,uptake . In addition, the possible errors in photochemical age dependent adjustment are about 27 %, as shown in Fig. S4. More about the controlling factors in k eff,uptake follows in section 3.4 after the sources and sinks discussion.

CHOCHO sources and sinks apportionment
Roughly 59 % of CHOCHO in the model originates from aromatics, with the largest contribution from toluene (41 %, Fig.   3a). We estimate that 20 % of CHOCHO is produced from biogenic VOCs (BVOCs), specifically isoprene and α-, β-pinene. 285 However, we cannot rule out possible underestimation in fast oxidized BVOCs contribution since we initialized our model with measurement aboard than emission rates or surface measurements.
The OH oxidation and photolysis loss are responsible only for 13 and 19 % of CHOCHO losses, respectively, while roughly 69 % of CHOCHO must be lost via aerosol uptake to reconcile the imbalance between modelled and measured concentration (Fig. 3b). This large importance of the aerosol uptake pathway is consistent with previous works in Mexico 290 City (75-95 %) (Volkamer et al., 2007) and in the Pearl River Delta measurements (~ 62 %) (Ling et al., 2020), while much larger than that in Los Angeles (0-30 %) (Washenfelder et al., 2011). The model-estimated lifetime of CHOCHO, 1.57 ± 0.23 hours, without aerosol uptake (in Fig. S9), is consistent with the general estimation of a few hours. However, when aerosol uptake is included, the CHOCHO lifetime decreases to 0.50 ± 0.26 hours during the typical flight duration (8:00 -17:00 LT), which is also consistent with previous results (Volkamer et 295 al., 2007).
The importance of aerosol uptake loss has clear dependence on A surf . Figure 3c shows the ratio of k eff,uptake /(k OH+Gly [OH]+j Gly ) against A surf . This ratio levels off around 2.5 under high A surf condition but it converges to 0 as A surf decreases since the k eff,uptake becomes close to 0 under no aerosol condition (no significant positive intercept) as shown in Fig. 4a. This confirms that our inferred k eff,uptake is legitimate, since the CHOCHO budget is closed only with photolysis and 300 OH oxidation loss in the absence of aerosol. Meanwhile, our steady-state model diverges more, and thus required higher k eff,uptake to reconcile the gap with observed CHOCHO as A surf increases. This is a clear evidence of required additional sink of CHOCHO which is linked with aerosol abundance.
The linearity between k eff,uptake and A surf in Regime I indicates that irreversible uptake on the aerosol surface is a reasonable approximation to mimic the behaviour of CHOCHO uptake for the conditions with low aerosol loading, since averaged aerosol modes across the regimes are found to be similar with each other over varying A surf (Fig. S10), thus the 315 number of particles matters more than aerosol size.
As shown in Fig. 4b, the k eff,uptake dependence on j NO2 (a proxy for solar irradiation), exists in both regimes (more apparent in Regime II, where k eff,uptake enhanced by ~ 40 % at higher j NO2 , compared to lower j NO2 ) and suggests the possibility of accelerated aqueous phase chemistry in SOA under active photochemistry. This light dependence of k eff,uptake is consistent with findings from previous laboratory studies as significant acceleration in CHOCHO aerosol uptake under 320 irradiated (compared to dark) conditions (Volkamer et al., 2009;Ervens and Volkamer, 2010). We speculate the effect of low volatile compounds' coatings on aerosols seed may not have played a significant role in our environment . The increase in k eff,uptake (decrease in CHOCHO uptake time scale) under high j NO2 may reflect faster production of low volatility products via irreversible pathways, likely due to enhanced organic photochemistry (Volkamer et al., 2009;Lee et al., 2011). 325 In Regime II, k eff,uptake plateaus in a range of 3-6×10 -4 s -1 and becomes relatively insensitive to A surf . One possible explanation of independent behavior of k eff,uptake in Regime II is likely related to aerosol viscosity. Figure 4c shows C AS and C AN change with respect to A surf ; C AN increases and levels off near the boundary of Regime I and II. An increase in salt content will lead to highly viscous aerosol, and thus retards mass transfer into the aerosol. Kampf et al. (2013) previously observed slower CHOCHO mass transfer for C AS larger than 12 m condition. Similar behavior is expected for C AN , thus, 330 more studies on C AN effect on the time scale of CHOCHO mass transfer into particle are desired.

glySOA production rate comparison
Figure 5 (a) shows the results of glySOA production rate (P glySOA ) comparison between 0-D box model (P glySOA, eff. Steady State , Eq. 4) and hybrid treatment (P glySOA, hybrid , Sect. S9) in Regime I. Calculated P glySOA without surface uptake (volume process only) is 2-3 orders of magnitude slower than that of the steady-state result, suggesting the importance of surface uptake 335 process in Regime I, which is consistent with previous model studies (Ervens and Volkamer, 2010;Knote et al., 2014). By adding surface uptake process, using the median γ Gly,uptake in Regime I (0.98 ×10 -2 ), the hybrid method matches within an order of magnitude with steady-state box model analysis.
In Regime II, where reactive surface uptake process is unlikely due to the lack of linearity between k eff,uptake and A surf ( Fig. 4a-b), inferred P glySOA from constrained K H,eff_Gly with measured C AS and C AN through Eq. (7) without surface uptake 340 process (P glySOA, eff. photochem , Sect. S4) is 2 orders of magnitude lower than P glySOA, eff. Steady State as shown in colored circles in Fig. 5b. One fixed K H,eff_Gly of 7.0×10 8 M atm -1 , close to the K H,eff_Gly at AS seed kinetic limit (~3×10 8 M atm -1 , Kampf et al., 2013) provides better agreement. It indicates that the K H,eff_Gly driven by AS and AN is not sufficient enough to explain the required K H,eff_Gly of CHOCHO. Although the increase in CHOCHO solubility by some of the inorganics other than AS and AN not included in K H,eff_Gly calculation (e.g., NaCl and NH 4 HSO 4 ) cannot be ruled out, due to their low concentrations, the 345 K H,eff_Gly contribution from these species is not significant enough (less than 5 %) to explain the required K H,eff_Gly . Even higher K H,eff_Gly (3×10 9 M atm -1 ) is required for Regime I (Fig. S11).
One possible explanation for high K H,eff_Gly (~10 8 M atm -1 ) even in low salt conditions (i.e., C AS <12 m) is the influence of organic compounds by various oligomerization pathways. Moreover, aerosol seeds consist of water-soluble organic carbons (e.g. fulvic acid) can enhance the solubility of CHOCHO (up to ~6.0×10 8 M atm -1 of K H,eff_Gly , Volkamer et 350 al., 2009). Corrigan et al. (2008) observed K H,eff_Gly > 10 8 M atm -1 with various organic acid aerosol seeds, including l-tartaric acid, dl-malic acid, and other liquid-phase aerosol particles containing amine functional groups. In addition, Qin et al. (2020) found the synergetic effect of CHOCHO uptake by coexistence with pinanediol and proposed acid-catalyzed cross-reactions which enhance the reactivity of CHOCHO in the aqueous phase.

CHOCHO contribution to SOA in SMA 355
To evaluate CHOCHO contribution to SOA formation in SMA, the relationship between P glySOA and organic aerosol enhancement (∆OA) normalized with ∆CO for minimizing dilution effect, was analyzed along with the air mass evolution https://doi.org/10.5194/acp-2021-672 Preprint. Discussion started: 31 August 2021 c Author(s) 2021. CC BY 4.0 License. (Fig. 6), where the background concentrations of OA and CO are 1 μg sm -3 and 140 ppb, respectively, as for Yellow Sea case in Nault et al. (2018). The P glySOA /∆CO increases as an average of 3.67 μg sm -3 ppm -1 hr -1 for the first 6 equivalent hours under 4.84×10 6 molecules cm -3 OH (averaged OH concentration over SMA) and the rate of ∆OA/∆CO growth over that time 360 window is 17.3 μg sm -3 ppmv -1 hr -1 . Thus, the CHOCHO contribution to SOA formation during the early stage of photochemical processing is estimated to be ~20 %. During the KORUS-AQ, the observationally constrained P glySOA ranges from 0-0.8 μg sm -3 hr -1 for individual RFs (shown in Fig. S12 in SI), which is comparable to Mexico City (> 15 %, Volkamer et al., 2007), Pearl River Delta (11.3 %, Ling et al., 2020) and Los Angeles basin (1-15 %, Knote et al., 2014) but higher than Pasadena (0-4 %, Washenfelder et al., 2011). 365 We also estimate the upper limit of oxalate mass yield from CHOCHO over SMA as ~ 10 % based on the oxalate fraction (~2 %) in OA (Fig. 6c). Our analysis shows lower oxalate yield than previously reports from the OH oxidation in cloud process (Tan et al., 2009, Lim et al., 2013Ortiz-Montalvo et al., 2014). We presume our lower yield has likely originated from the existence of many fates in uptaken CHOCHO, such as oligomerization (Lim et al., 2013), reaction with sulfate (Lim et al., 2016) and ammonium (Yu et al., 2011;Ortiz-Montalvo et al., 2014), etc., since our analysis is over SMA 370 area. From our dataset, in cloud process case, analysis was impossible due to low abundance in gas-phase CHOCHO concentration (below detection limit).

Conclusions
In the present study, production rates and loss rates of CHOCHO were simulated using a 0-D box model constrained by high-quality gas, aerosol and meteorological measurements on board the NASA DC-8 during the KORUS-375 AQ mission, to elucidate the controllers of CHOCHO uptake to aerosols. High CHOCHO concentrations were observed over highly populated cities (i.e., Seoul) and industrial area, with peak concentration of 1.05 (2.10) ppb for 1 minute (10 seconds) average. Aromatics were the most important precursors of CHOCHO production (59 %), toluene being the highest contributor (41 %). We found the importance of the loss path to aerosol (69 %) followed by photolysis (19 %) and OH oxidation (12 %). Comparison of dilution-corrected P glySOA to that of OA growth over the first 6 photochemical hours shows 380 that glySOA contributes to ~20 % of SOA formation and oxalate yield from glySOA was estimated to be less than 10 % in Seoul. The k eff,uptake linearly increase with A surf in Regime I (A surf < 4.9×10 -6 cm 2 cm -3 , γ Gly,uptake = 9.8×10 -3 ) but plateaus in Regime II (A surf > 4.9×10 -6 cm 2 cm -3 ) which suggests the limitation in surface uptake approximation. This slower uptake under high aerosol loading condition can be attributed to the increased AN molality which likely induced high aerosol viscosity and thus, slow down CHOCHO mass transfer to aqueous phase. We also found light dependence of k eff,uptake 385 suggesting the importance of photochemistry in ambient condition.
Finally, our work highlights the lacking knowledge to explain the CHOCHO solubility in real atmospheric circumstance. K H,eff_Gly (~ 10 6 M atm -1 ) derived from salting-in by AS and AN was not enough to describe CHOCHO loss by heterogeneous processes. This urges more attention to other various factors in CHOCHO solubility control in addition to the abundance of inorganic salts. Adding more constraints on these factors are important not only in SOA forming but also in air 390 mass source tracking from satellite since CHOCHO is one of the two VOC proxies measured from space.

Author contribution
DK and KEM designed and executed the study and led the writing of the paper. DK, CC, SJ, KEM, SL, BAN, PCJ, DAD, 395 JCS, JLJ, DRB, AW, AF, JPD, GSD, SEP, SRH, KU, LGH, DJT and JD provided observational data. RV and CJK contributed to aqueous phase mechanism model setups and data interpretation. BAN and PCJ ran E-AIM model. All coauthors contributed with feedback during the development and writing of the study.

Competing interests
The authors declare that they have no conflict of interest. 400 Müller, is acknowledged for their support with field work and data processing. Ionicon Analytik is acknowledged for instrumental support.