Influence of aromatics on tropospheric gas-phase composition

Aromatics contribute a significant fraction to organic compounds in the troposphere and are mainly emitted by anthropogenic activities and biomass burning. Their oxidation in lab experiments is known to lead to the formation of ozone and aerosol precursors. However, their overall impact on tropospheric composition is uncertain as it depends on transport, multiphase chemistry, and removal processes of the oxidation intermediates. Representation of aromatics in global atmospheric models 5 has been either neglected or highly simplified. Here, we present an assessment of their impact on the gas-phase chemistry, using the general circulation model EMAC (ECHAM5/MESSy Atmospheric Chemistry). We employ a comprehensive kinetic model to represent the oxidation of the following monocyclic aromatics: benzene, toluene, xylenes, phenol, styrene, ethylbenzene, trimethylbenzenes, benzaldehyde, and lumped higher aromatics that contain more than 9 C atoms. Significant regional changes are identified for several species. For instance, glyoxal increases by 130 % in Europe and 260 % 10 in East Asia, respectively. Large increases in HCHO are also predicted in these regions. In general, the influence of aromatics is particularly evident in areas with high concentrations of NOx, with increases up to 12 % in O3 and 17 % in OH. On a global scale, the estimated net changes are minor when aromatic compounds are included in our model. For instance, the tropospheric burden of CO increases by about 6 %, while the burdens of OH, O3, and NOx (NO + NO2) decrease between 3 % and 9 %. The global mean changes are small, partially because of compensating effects between highand low-NOx 15 regions. The largest change is predicted for the important aerosol precursor glyoxal, which increases globally by 36 %. In contrast to other studies, the net change in tropospheric ozone is predicted to be negative, -3 % globally. This change is larger in the northern hemisphere where global models usually show positive biases. We find that the reaction with phenoxy radicals is a significant loss for ozone, of the order of 200-300 Tg/yr, which is similar to the estimated ozone loss due to bromine chemistry. 20 Although the net global impact of aromatics is limited, our results indicate that aromatics can strongly influence tropospheric chemistry on a regional scale, most significantly in East Asia. An analysis of the main model uncertainties related to oxidation and emissions suggests that the impact of aromatics may even be significantly larger. 1 https://doi.org/10.5194/acp-2020-461 Preprint. Discussion started: 2 June 2020 c © Author(s) 2020. CC BY 4.0 License.

and the set-up of the simulations are described. Section 3 analyzes the calculated impact on selected chemical species both on the global and on the regional scales.
2 Model description 60 We used the ECHAM5/MESSy Atmospheric Chemistry (EMAC) model, which is a numerical chemistry and climate simulation system that includes submodels describing tropospheric and middle atmosphere processes . EMAC uses the second version of the Modular Earth Submodel System (MESSy2) to link multi-institutional computer codes. The core atmospheric model is the 5th generation European Centre Hamburg general circulation model (ECHAM5, Roeckner et al., 2006).
For the present study we performed simulations with EMAC (ECHAM5 version 5.3.02, MESSy version 2.53) in the T63L31ECMWF resolution, which corresponds to a grid with a horizontal cell size of approximately 1.875 • × 1.875 • and 31 vertical hybrid pressure levels, extending from the surface up to about 10 hPa.
Emission rates of the individual aromatics are shown in Tab. 1. The sum of all sources is 29.4 TgC/a. For anthropogenic emissions, we used EDGAR 4.3.2 (Huang et al., 2017), distributed vertically as in Pozzer et al. (2009). The MESSy submodel 70 MEGAN calculates biogenic emissions (Guenther et al., 2012). For biomass burning, the submodel BIOBURN was used, which integrates the Global Fire Assimilation System (GFAS) inventory (Kaiser et al., 2012).
Atmospheric chemistry was calculated with the MECCA submodel, which has been evaluated by Pozzer et al. (2007) and Pozzer et al. (2010). The most recent model version has been described by Sander et al. (2019). The mechanism for aromatic species is a reduced version of the MCM (Bloss et al., 2005b), as described in detail by Cabrera-Perez et al. (2016). In this 75 study, we consider several additions to the MCM reactions: -For several nitrophenols (MCM names: HOC6H4NO2, DNPHEN, TOL1OHNO2, MNCATECH, DNCRES), their photolysis reactions were added (Bejan et al., 2006), e.g.: -For the photolysis of benzaldehyde, the MCM uses the rate constant (j-value) of methacrolein as a proxy. We have 80 calculated the j-value based on the UV/VIS spectrum of benzaldehyde recommended by Wallington et al. (2018). In our code, the photolysis of benzaldehyde produces C 6 H 5 O 2 , HO 2 and CO.

85
-For the reaction of HO 2 with the peroxy radical C6H5CO3 (resulting from the oxidation of benzaldehyde), we use the yields provided by Roth et al. (2010).
-Alkyl nitrate yields are calculated as a function of temperature and pressure, as described by Sander et al. (2019).
-Bicyclic peroxy radicals in the oxidation mechanism of toluene produce some glyoxal and methyl glyoxal as suggested by Birdsall et al. (2010). Benzene is treated analogously.

90
The aerosol calculations follow the approach of Pringle et al. (2010), with the notable difference of the inclusion of the explicit organic aerosol submodel ORACLEv1.0 by Tsimpidi et al. (2014). Although, similar to Tsimpidi et al. (2014), lowand intermediate volatiles are parameterized as lumped species, the equilibrium with their equivalent aerosol phase is explicitly calculated for 600 volatile organic carbon tracers via ORACLE. The volatility and the enthalpy of vaporization of each tracer is estimated with the approaches of Li et al. (2016) andEpstein et al. (2010), respectively.

95
The simulated period covers the years 2009-2010, with the first year as spin-up, and the year 2010 being used for the analysis. The feedback between radiation and chemistry was decoupled to avoid any influence of chemistry on the dynamics (QCTM mode by Deckert et al. (2011)). As a consequence, every simulation discussed here has the same meteorology, i.e., binary identical transport.
To analyze the influence of the aromatic compounds on atmospheric chemistry and composition, we performed three model 100 simulations, as listed in Tab. 2. The AROM simulation includes all chemical reactions and emissions of the following monocyclic aromatic compounds: benzene, toluene, xylenes (lumped), phenol, styrene, ethylbenzene, trimethylbenzenes (lumped), benzaldehydes, and higher aromatics (as representative of aromatics with more than 9 carbon atoms). The reference simulation (NOAROM) is identical to AROM, except that it excludes aromatic compounds. In the ONLYMCM run, we reverted the additions and changes to the MCM that have been described above. Our focus is to compare AROM with NOAROM. Results

105
of ONLYMCM are mainly interesting for benzaldehyde and HONO. As EMAC uses terrain-following vertical hybrid pressure coordinates, we will refer to "surface" as the lowest model level, with an average thickness of roughly 60 m.
Globally averaged surface mixing ratios obtained from all model simulations (AROM, NOAROM, and ONLYMCM) are listed in Tab. 3. Figure 1 shows the annual average mixing ratios of the sum of all aromatic compounds included in the simulation AROM.  Figure 3 shows the model-calculated OH in the AROM and NOAROM simulations. When aromatics are introduced to the model, the global average concentration of OH decreases for two reasons: first, the direct reaction with aromatics consumes OH, and second, additional CO resulting from the degradation of aromatics represents an increased sink for OH. However, 120 in eastern Asia, Europe, and the east coast of the US, where NO x concentrations are high, an increase of OH can be seen.

Hydroxyl radical (OH)
Although the aromatics decrease NO x in these areas (see below), the chemical system remains in the high-NO x regime.
We find a positive correlation between OH and anthropogenic emissions in these regions but a negative correlation in the low-NO x CAF region. The increased OH in the high-NO x regions is mainly caused by the reaction of NO with HO 2 . Figure 4 shows the seasonal cycle of the OH mixing ratio in the planetary boundary layer for the NH and SH. Inclusion of 125 the aromatics leads to a relative decrease between 2.5 % and 5.5 %. Higher OH concentrations are identified over continental areas during the NH autumn, winter and spring than in summer (Fig. 3). In summer, OH concentrations increase only at a few locations when aromatics are included. Figure 5 shows the annual zonal mean changes of the OH mixing ratio. The changes are most pronounced in the NH upper troposphere where reductions range from 7 % to 20 %. This helps bringing the model-simulated inter-hemispheric OH 130 asymmetry closer to that derived from observations (Lelieveld et al., 2016). Globally, aromatics oxidation reduces OH by 7.7 % and consequently increases methane lifetime.

Ozone (O 3 )
In most areas of the globe, surface ozone is slightly lower in AROM than in NOAROM (Fig. 6). The O 3 reduction is due to (i) the decrease in NO x concentrations (limiting ozone formation) and (ii) increasing radical production (OH, HO 2 , and 135 RO 2 ) in ozone-depleting regimes, which enhances reactions of O 3 with HO 2 and OH. Only a few high-NO x regions, where hydrocarbons are the limiting factor for ozone formation, show increased ozone concentrations: mainly East China (EAS), but also the eastern US (EUS) and Europe (EUR). The increases in these areas correlate with anthropogenic emissions of aromatics, which have significant ozone formation potentials. We find a positive correlation between O 3 and anthropogenic emissions in the EAS and EUR regions but a negative correlation in the low-NO x CAF region.

140
The seasonal cycles of the relative differences show lower amplitude than for OH, but similar patterns (Fig. 9). The impact of aromatics is smallest in summer.
The zonal mean changes of O 3 mixing ratio in the troposphere are uniformly negative (Fig. 7). Similar to surface ozone, the annual mean changes for ONLYMCM and AROM are −2.3 % and −3.0 %, respectively. The hemispheric changes are shown in Tab. 4. It is well known that MCM for aromatics overestimates ozone production in chamber experiments (Bloss et al.,145 2005b). The issue has been analysed in the companion paper (Bloss et al., 2005a) where the best mechanism improvement was found to be an early OH source during oxidation. Cabrera-Perez et al. (2016)  However, they only considered benzene, toluene and xylenes. Our results, obtained with a more comprehensive setup, suggest 150 that aromatics could slightly ameliorate the model overestimate in the NH (Jöckel et al., 2016;Young et al., 2018). The overall tropospheric ozone burden decreases from 381 to 369 Tg for the AROM simulation. These estimated changes are robust against the tropopause definition and are about -3.5 and -2.3 % for the Northern and Southern Hemispheres, respectively (Table 4). These changes are associated with the enhanced direct ozone loss by reactions with organic compounds. It is widely acknowledged that this direct loss is only due to the ozonolysis of unsaturated VOCs and is estimated to be about 100 Tg/yr, 155 less than 2 % of the tropospheric ozone budget (Tilmes et al., 2016). However, with aromatics a new direct ozone loss process involving organic radicals comes in place. In Figure 8 the change in tropospheric ozone burden is shown against the change in ozone loss with organic compounds. This change is estimated to be globally in the 200-300 Tg/yr range depending on the mechanism used and is comparable to the loss by bromine chemistry in the troposphere (Sherwen et al., 2016)). Ozone is known to react with organic radicals like methyl peroxy radical although this loss is an insignificant sink (Tyndall et al., 1998). 160 We find that phenoxy radicals from aromatics are a significant sink term of ozone (>200 Tg/yr). These radicals are unique to aromatics oxidation and they also react with NO and NO 2 . When the concentrations of NO x are relatively low, C 6 H 5 O has sufficiently long lifetime to react with O 3 . This ozone loss is modelled based on the results by Tao and Li (1999)  This ozone loss is enhanced by phenoxy radical production in the R2 reaction and the subsequent loss of odd oxygen by NO 3 photolysis and N 2 O 5 heterogeneous loss In our chemical kinetics mechanism (also in MCM) the reaction system just described constitutes an effective catalytic destruction cycle of odd oxygen. The strength of this cycle depends on the phenoxy radical levels and is significantly reduced in AROM compared to onlyMCM (Figure 8). We ascribe this difference to MCM not accounting for the photolysis of nitrophenols (R1) as determined by Bejan et al. (2006) preventing reformation of phenoxy radicals.
Our results for ozone differ both in magnitude and sign compared to the global study by Yan et al. (2019). However, the 175 latter used the SAPRC-11 oxidation mechanism (Carter and Heo, 2013) which does not account for the reaction of phenoxy radicals with ozone (R3) and phenylperoxy radicals with NO 2 (R2).

Inorganic nitrogen
The simulated annual mean NO x concentrations at the surface are significantly lower in AROM than in NOAROM (Figs. 10 and 11). One reason is the formation of aromatic species containing nitrogen (e.g., nitrophenols) in AROM, thereby transferring 180 part of the NO x burden to the nitrogenated species. The largest decreases (both absolute and relative) are found in regions with high NO x concentrations. Since the ozone chemistry is not NO x -limited in these regions, the impact on ozone is small. This holds for the free troposphere for which zonal average decreases in NO x can be larger than 20 % (not shown), which in turn significantly influence OH (Fig. 5).
On the one hand, the reaction with aromatics is a sink for NO 3 . On the other hand, NO 3 is produced in the phenylperoxy 185 reaction with NO 2 (R2). Comparing AROM to NOAROM, the global average of the nighttime species NO 3 increases by more than 7 % (Tab. 3). In contrast to the global mean tendency, NO 3 decreases in several regions in Africa, South America, and India (Fig. 12). These decreases correlate well with emissions from biomass burning.
Although the net change of global HONO is small (about 3 % less in AROM than in NOAROM, see Figure 13 and Tab. 3), the regional differences can be large (Tab. 5). A decrease of HONO is seen mainly in polluted areas (EAS, EUR, EUS) in 190 the winter. In contrast, HONO increases in the regions with emissions from biomass burning (AMA, CAF). Here, HONO is formed by the photolysis of nitrophenols (R1). Since these reactions are not included in the MCM, we do not see any HONO increase in the ONLYMCM simulation (Fig. 14).
On a global average level, HNO 3 is not affected much by aromatics. However, an increase can be seen in the regions where ozone increases (EAS) or where biomass burning decreases NO 3 and N 2 O 5 (CAF), see Figure 15 and Tab. 5. An average zonal 195 mean change of up to 5% throughout the UT/LS is linked to the enhanced NO 3 production by R2.
of ozone loss via R3 and analogous reactions is to be expected. Moreover, the ozone loss is likely underestimated because of 230 the model not accounting for the photolysis of nitrophenols reforming phenoxy radicals. Different from the HONO-formation channel, which destroys the aromatic ring, channels yielding substituted phenoxy radicals may dominate (Cheng et al., 2009;Vereecken et al., 2016) and thus enhance ozone loss.
Cloud chemistry of organic compounds is known to suppress gas-phase HO x -production and directly consume ozone (Lelieveld and Crutzen, 1990). The overall effect on ozone depends on the local chemical regime. In our study water-soluble 235 products are set to only undergo wet deposition (dissolution and removal by precipitation). Their aqueous-phase chemistry might however have a non-negligible effect on ozone and other oxidants. For instance, phenol is known to react very quickly with OH in the aqueous-phase (Field et al., 1982). Moreover, phenoxide anions from phenols react quickly with ozone (Hoigné and Bader, 1983). In particular, nitrophenols might be efficient ozone scavengers as they are stronger acids than unsubstituted phenols. A global assessment of cloud chemistry involving aromatics oxidation products is possible with the modelling system 240 used here (Tost et al., 2006. However, considering the complexity of aqueous-phase oxidation of organic compounds, such an assessment is outside the scope of this study and deserves a dedicated model study.
In our study, biomass burning emissions of aromatics are potentially underestimated. In fact, based on the recent update by Andreae (2019), we estimate that emissions might be up to 5 Tg/yr (65%) higher than what is implemented in our model. an overestimate of the chemical sink in the troposphere by reaction with hydroxyl radical. However, the annual global mean concentration of hydroxyl radicals is potentially 10% too high (Lelieveld et al., 2016), which cannot account for model concentration biases that are larger than 20%. Therefore, we surmise that the impact of aromatics on the trace gas composition may be larger than estimated in this study. Of the nitrogen compounds, mainly NO 3 and HONO are affected by the aromatics chemistry. 265 We conclude that, although the impact of aromatics is relatively minor on the global scale, it is important on regional scales, notably in the anthropogenic source regions, and especially in those where NO x emissions are strongest. Given the uncertainties in the oxidation mechanisms and emissions, the results of our model calculations may underestimate the impact of aromatics on the tropospheric gas-phase composition.

Competing interests. The authors have no competing interests
Acknowledgements. The authors want to acknowledge the use of the Ferret program for analysis and graphics in this paper. Ferret is a product of NOAA's Pacific Marine Environmental Laboratory (information is available at http://www.ferret.noaa.gov).  Table 3. Globally averaged mixing ratios at the surface (annual averages for 2010). "ABSDIFF" denotes the absolute difference, (e.g., AROM-NOAROM), and "RELDIFF" the relative difference, (e.g., AROM/NOAROM-1).