Atmospheric reactivity and oxidation capacity during summer at a suburban site between Beijing and Tianjin

. Hydroxyl

Abstract. Hydroxyl (OH) radicals, nitrate (NO 3 ) radicals and ozone (O 3 ) play central roles in the troposphere because they control the lifetimes of many trace gases that result from anthropogenic and biogenic origins. To estimate the air chemistry, the atmospheric reactivity and oxidation capacity were comprehensively analyzed based on a parameterization method at a suburban site in Xianghe in the North China Plain from 6 July 2018 to 6 August 2018. The total OH, NO 3 and O 3 reactivities at the site varied from 9.2 to 69.6, 0.7 to 27.5 and 3.3 × 10 −4 to 1.8×10 −2 s −1 with campaign-averaged values of 27.5±9.7, 2.2 ± 2.6 and 1.2 ± 1.7 × 10 −3 s −1 (± standard deviation), respectively. NO x (NO + NO 2 ) was by far the main contributor to the reactivities of the three oxidants, with average values of 43 %-99 %. Alkenes dominated the OH, NO 3 and O 3 reactivities towards total nonmethane volatile organic compounds (NMVOCs), accounting for 42.9 %, 77.8 % and 94.0 %, respectively. The total OH, NO 3 and O 3 reactivi-ties displayed similar diurnal variations with the lowest values during the afternoon but the highest values during rush hours, and the diurnal profile of NO x appears to be the major driver for the diurnal profiles of the reactivities of the three oxidants. A box model (a model to Simulate the concentrations of Organic vapors, Sulfuric Acid and Aerosols; SOSAA) derived from a column chemical transport model was used to simulate OH and NO 3 concentrations during the observation period. The calculated atmospheric oxidation capacity (AOC) reached 4.5 × 10 8 molecules cm −3 s −1 , with a campaign-averaged value of 7.8 × 10 7 molecules cm −3 s −1 dominated by OH (7.7 × 10 7 molecules cm −3 s −1 , 98.2 %),

Introduction
In the planetary boundary layer, high concentrations of primary pollutants, such as carbon monoxide (CO), nitrogen oxides (NO x = NO + NO 2 ), and volatile organic compounds (VOCs) from both biogenic and anthropogenic origins, are transformed by reactions with atmospheric oxidants, such as hydroxyl (OH) radicals, nitrate (NO 3 ) radicals, chlorine atoms and ozone (O 3 ) on local to global scales (Atkinson and Arey, 2003;Heard and Pilling, 2003;Lu et al., 2018;Wang et al., 2020), with the dominant reaction depending on the time of day and specific trace gases. Ultimately, these processes lead to the formation of a series of important secondary pollutants, including tropospheric O 3 and secondary organic aerosols (SOAs) (Goldstein and Galbally, 2007).
OH radicals control the daytime oxidation capacity of the atmosphere (Heard and Pilling, 2003), initiating and participating in many oxidation reaction processes. OH can react by adding OH groups to or abstracting H from trace gases, such as CO, NO x , methane (CH 4 ) and nonmethane volatile organic compounds (NMVOCs) (Kovacs et al., 2003;Sadanaga et al., 2005). The total OH reactivity, which is equivalent to the inverse chemical OH lifetime, is the sum of the products of the concentrations and respective reaction rate coefficients for all gases that react with OH. The online techniques used to determine OH reactivity include a flow tube with sliding injector method (Kovacs et al., 2003), a comparative rate method (Sinha et al., 2008) and a laser flash photolysis pump probe technique (Whalley et al., 2016). Based on these online methods, total OH reactivity values have been measured in urban, suburban, remote and forest areas during the last decade. The urban areas investigated include Nashville, USA (SOS) (Kovacs et al., 2003), New York, USA (PMTACS-NY2004) (Ren et al., 2006a), Mexico City, Mexico (MCMA-2003) (Shirley et al., 2006), Houston, USA (TRAMP2006) (Mao et al., 2010), Paris, France (MEGAPOLI) (Dolgorouky et al., 2012), London, UK (ClearfLo) (Whalley et al., 2016), Helsinki, Finland (Praplan et al., 2017), Seoul, South Korea (Kim et al., 2016), and Beijing, China (Yang et al., 2017). The total OH reactivity in these urban areas ranged from 1 s −1 in clean air to 200 s −1 in extremely polluted air, and NO x , CO, formaldehyde (HCHO) and nonmethane hydrocarbons (NMHCs) were the main contributors (Ferracci et al., 2018). The suburban areas investigated include Whiteface Mountain, USA (PMTACS-NY2002) (Ren et al., 2006b), Weybourne, UK (TORCH-2) (Lee et al., 2010), Yufa, China (CAREBeijing-2006)  , Backgarden, China (PRIDE-PRD) (Lou et al., 2010), Jülich, Germany (HOxComp) (Elshorbany et al., 2012), Ersa, Corsica (CARBOSOR-ChArMeX) (Zannoni et al., 2017), Po Val-ley, Italy (Kaiser et al., 2015), the Indo-Gangetic Plain, India (Kumar et al., 2018), and Heshan, China (Yang et al., 2017). The total OH reactivity in these suburban areas ranged from 4.6 to 64 s −1 . OH reactivity was also modeled by a global model by Ferracci et al. (2018) and by a box model based on the Master Chemical Mechanism (MCM) (Whalley et al., 2016). The calculated total OH reactivity is the sum of the OH reactivities that are attributed to measured trace gases. The concentrations (in molecules cm −3 ) of trace gases and the reaction rate constants (in cm 3 molecule −1 s −1 ) of these trace gases with the OH radical are the key factors for computing OH reactivity (Mogensen et al., 2011(Mogensen et al., , 2015. In general, the trace gases considered in calculating OH reactivity include NMVOCs, CH 4 , CO, NO x , SO 2 and O 3 . As reported, the contribution from NO x exceeds 50 % for the cities of Paris, Tokyo, New York and Beijing, showing the large influence of traffic-related emissions on OH reactivity (Dolgorouky et al., 2012;Yang et al., 2017;Yoshino et al., 2006), but the contribution from NMVOCs reaches 50 % in Mexico and Houston due to the large quantity of biomass fuel being burned and high industrial solvent emissions (Mao et al., 2010;Shirley et al., 2006).
As OH levels are vastly reduced during the nighttime due to the absence of photolysis, NO 3 formed by the slow reaction NO 2 +O 3 → NO 3 +O 2 is the main initiator of nighttime oxidation chemistry in the troposphere (Asaf et al., 2009;Geyer et al., 2001). NO 3 reacts effectively with unsaturated NMVOCs, such as certain alkenes or aromatics via additions to > C = C < double bonds, which can initiate the formation of peroxyl radicals (HO 2 and RO 2 ) and even OH (Geyer et al., 2001). High NO 3 mixing ratios and large reaction rate constants with several unsaturated NMVOCs result in NO 3 being the dominant sink of many unsaturated NMVOCs during the nighttime. The role of NO 3 as an oxidizing agent can be assessed via its total reactivity towards trace gases. The total NO 3 reactivity is an indication of nighttime oxidation rates of trace gases with direct impacts on NO x levels and indirect impacts on heterogeneous NO x losses and ClNO 2 formation (Liebmann et al., 2017). As frequently reported for total OH reactivity, total NO 3 reactivity can be measured online or calculated by summing the loss rates for a set of reactive trace gases. Previous works on measured total NO 3 reactivity have revealed strong diel variation. For instance, the total NO 3 reactivity obtained in Hyytiälä, Finland, displayed strong diel variation, with a campaign-averaged nighttime value of 0.11 s −1 and daytime value of 0.04 s −1 (Liebmann et al., 2018a), but values varied from 0.005 to 0.1 s −1 during the nighttime and reached values as high as 1.4 s −1 in the daytime in Taunus, Germany (Liebmann et al., 2017).
Along with reactions with OH and NO 3 radicals, trace gases are also oxidized in the troposphere by reactions with O 3 . Although most NMVOCs have reaction rates with O 3 that are much lower than those with either OH or NO 3 , O 3 is very important because it is present at elevated mixing ratios in clean or contaminated atmospheres . The rate constants of the reactions for some alkenes with O 3 are even comparable to those with NO 3 (Atkinson and Arey, 2003). The total reactivity of O 3 with trace gases can reflect the role of O 3 as an oxidizing agent. Direct measurements of total O 3 reactivity were not available until very recently (Geyer, 2003); hence, the reactivity of O 3 has traditionally been calculated by summing the reactivities due to individual reactive trace gases. The calculated O 3 reactivity obtained in Pabstthum, Germany, revealed that terpenes (20 %), isoprene (20 %) and other alkenes (60 %) were the dominant contributors during the night of 20 and 21 July but arose mainly (83 %) from nonbiogenic alkenes during the night of 4 and 5 August (Geyer, 2003).
As mentioned above OH radicals, NO 3 radicals and O 3 react with trace gases via different rate coefficients and mechanisms, resulting in profoundly different reactivities. Therefore, comprehensive evaluations of OH, NO 3 and O 3 reactivities are key to understanding atmospheric oxidation capacity and identifying the controlling active species of secondary pollution in the atmosphere. However, comprehensive evaluations of the total calculated OH, NO 3 and O 3 reactivities are scarce in China. In this study, we calculated the OH, O 3 and NO 3 reactivities at a suburban site (Xianghe) in the North China Plain during an intensive measurement campaign in the summer of 2018. By combining simulated OH and NO 3 concentrations using a box model (a model to Simulate the concentrations of Organic vapors, Sulfuric Acid and Aerosols; SOSAA), we calculated the oxidation capacities of OH, NO 3 and O 3 and estimated their relative contributions.

Site description
The sampling site is located at the Xianghe Atmospheric Observatory (39.798 • N, 116.958 • E; 15 m above sea level), which is operated by the Institute of Atmospheric Physics (IAP)/Chinese Academy of Sciences (CAS). The sampling site is a typical suburban site in the seriously polluted Beijing-Tianjin-Hebei large urban region, which is approximately 50 km southeast of Beijing, 75 km northwest of Tianjin and 35 km northeast of Langfang in Hebei Province. The sampling site is approximately 4 km west of the downtown center and is surrounded by residential areas and agricultural land (see Fig. 1).

Experimental method
Ambient NMVOCs were collected and analyzed continuously and automatically with a time resolution of 1 h using a custom-built gas chromatography-mass spectrometry/flame ionization detection (GC-MS/FID) instrument. The suitability of this system for NMVOC measurements is well verified, and it has been used in several large field campaigns (Chen et al., 2014;Yuan et al., 2013;Wu et al., 2016). De- tailed descriptions of the configuration of the GC-MS/FID system, the detection limits and the precision of NMVOC measurements can be found in our previous paper (Yang et al., 2019). CH 4 was analyzed by an Agilent 7890A gas chromatography (GC) instrument with a flame ionization detector (FID). HCHO was measured by Hantzsch fluorimetry with a commercial instrument (AL4021, Aero-Laser GmbH, Germany) (Lu et al., 2019). Air-quality-related trace gases, including O 3 , NO-NO 2 -NO x , SO 2 and CO were measured by analyzers from Thermo Fisher Scientific, United States. High-resolution (5 min averages) data sets of O 3 , NO-NO 2 -NO x , SO 2 and CO were obtained, and hourly averaged data were used after applying strict data quality control measures. HONO mixing ratios were determined using a custom-made HONO analyzer Tong et al., 2015). The photolysis frequencies, JO 1 D, JNO 2 and JNO 3 , in the atmosphere were measured by a PFS-100 photolysis spectrometer -Juguang Technology (Hangzhou) Co., Ltd, Hangzhou, China. Further details of the measurements of NMVOCs, CH 4 , HCHO, trace gases, HONO and photolysis frequencies can be found in the Supplement. The meteorological parameters, including wind speed, wind direction, temperature and relative humidity, were obtained from the National Meteorological Information Center (http://data.cma.cn/, last access: 20 August 2018). The sensors are approximately 3000 m away from the measurement area.

Atmospheric chemical transport model: SOSAA
SOSAA is a column (or one-dimensional) chemical transport model that was first developed by Boy et al. (2011). A more detailed description of its newest version can be found in Zhou et al. (2017a, b). In this study, a box model version of SOSAA was used, in which the meteorological variables, including air temperature, air pressure, relative humid-ity and incoming global radiation, were directly read from the measurement data. The chemistry scheme was generated by MCM v3.3.1 (http://mcm.leeds.ac.uk/MCMv3.3.1, last access: 25 March 2020) (Jenkin et al., 1997(Jenkin et al., , 2015Saunders et al., 2003) and then converted to Fortran code with kinetic preprocessor (KPP) (Damian et al., 2002). The mixing ratios of chemical species included in the chemistry scheme, e.g., O 3 , NO, NO 2 , SO 2 , CO, HONO, HCHO, isoprene and acetone, were read from the measured data when available. Ten oxygenated VOCs (OVOCs) -acrolein (ACR), C 2 H 5 CHO, methacrolein (MACR), C 3 H 7 CHO, methylvinylketone (MVK), methylethylketone (MEK), 2pentanone (MPRK), C 4 H 9 CHO, diethylketone (DIEK) and C 5 H 11 CHO -were excluded from the input list, despite also being measured, because their simulated concentrations were compared with the measurement data to validate the model performance. Seven photolysis rates (JO 1 D, JH-CHO_M, JNO 2 , JH 2 O 2 , JHONO, JNO 3 _M, JNO 3 _R) were also read from the measurement data, and the related photochemical reactions are shown below.
The other photolysis rates were calculated using the incoming global radiation. The deposition velocities of all noninput species were set to 0.01 m s −1 , and the boundary layer height was assumed to be 1 km Zhu et al., 2020). The simulated OVOCs were also assumed to condense onto pre-existing aerosols. Their condensation sinks were set to make their simulated concentrations approach the measurement data. The model time step was set to 10 s, and the data were output every 0.5 h. All the input data were interpolated to the model time step.

Speciated oxidant reactivity
Atmospheric oxidant reactivity is a measure of the strength of the reaction of trace gases with an oxidant (OH, O 3 or NO 3 ) (Kovacs et al., 2003;Mogensen et al., 2015). High oxidant reactivity values correspond to short lifetimes, and long-lived species have low reactivities. The total OH, NO 3 and O 3 reactivities can be calculated by Eqs.

Atmospheric oxidation capacity (AOC)
The term "oxidation capacity" of an oxidant X (NO 3 , OH and O 3 ) is defined as the sum of the respective oxidation rates of the molecules Y i (NMVOCs, CH 4 and CO) (Geyer et al., 2001).
Here, [Y i ] and [X] are number concentrations of molecule 3 Results and discussion

Overview of measurements
For the data evaluation, all measurements were averaged over 1 h time intervals. The measured concentrations of major pollutants and meteorological parameters at Xianghe are depicted in Fig. 2, while the mean diurnal profiles are shown in Fig. S1 in the Supplement. During the campaign, sunny weather conditions prevailed with temperatures ranging from 25 to 31 • C during the daytime. Wind data suggested that the prevailing wind was from the eastern sampling site with a mean wind speed of 1.0 m s −1 ranging from 0.3 to 1.4 m s −1 , and the average relative humidity was 85 %, reaching up to 96 % during the night (Fig. 2). Campaign-averaged maximum and minimum SO 2 mixing ratios of 3.6 ppb at approximately 14:00 h (local time, LT) and 2.3 ppb during the nighttime were obtained (Fig. S1a). For JO 1 D, JNO 2 and JNO 3 , a similar maximum at ∼ 14:00 h was observed, with maximum values of 2.1 × 10 −5 , 5.3 × 10 −3 and 1.3 × 10 −1 s −1 , respectively ( Fig. S1k-m). The maximum JO 1 D at this site was comparable with those in Shanghai and Chongqing but higher than that in Guangzhou and lower than that in Beijing Wang et al., 2019). The observed mean daily maxima of JNO 2 at this site were higher than those observed in the eastern Mediterranean (Gerasopoulos et al., 2012) but lower than those in Beijing . The diurnal maximum O 3 concentration was 72 ppb at this site ( Fig. S1d), which was in line with that observed in Beijing (72 ppb) but higher than those measured in Guangzhou (65 ppb) and Chongqing (56 ppb) and lower than that observed in Shanghai (80 ppb) . The O 3 precursors, CO, NO x , CH 4 and NMVOCs, are shown in Figs. 2 and S1. The trend of NMVOCs was inversely related to that of O 3 . When the NMVOC concentrations in the atmosphere accumulates to a certain level, as photochemical reactions progress, the O 3 concentration gradually increases, and the NMVOC concentrations gradually decrease (Kansal, 2009;Song et al., 2018). CO and NO x showed a similar diurnal profile with a maximum during rush hour and a minimum in the afternoon ( Fig. S1b and c), suggesting that both CO and NO x originated from the same source (enhanced traffic emission) and/or were manipulated by the same factor (e.g., poor dilution conditions). During the campaign, the average mixing ratio of total NMVOC was 32.4 ppb, with the highest contributions from alkanes (13.2 ppb, 40.6 %), followed by OVOCs (12.0 ppb, 37.0 %), aromatics (4.3 ppb, 13.2 %) and alkenes (3.0 ppb, 9.2 %). The top 10 NMVOC species (Fig. 3a), in terms of emissions, consisted of HCHO (7.0 ppb), propane (3.7 ppb), acetone (3.2 ppb), ethane (3.2 ppb), n-butane (1.9 ppb), m/pxylene (1.6 ppb), iso-pentane (1.3 ppb), ethylene (1.3 ppb), iso-butane (1.1 ppb) and isoprene (1.0 ppb), accounting for 78.4 % of the total NMVOC concentration. As typical tracers of vehicle-related emissions, propane, ethane, ethene, butanes and pentanes were present in high concentrations, suggesting that vehicle-related emissions were likely to be the dominant source of NMVOCs at this site. In addition, the shape of the diurnal variations in total NMVOCs backed the presence of vehicle-related emissions, presenting higher mixing ratios during the early morning and from evening to midnight, which may be related to enhanced traffic emissions during rush hours and poor dilution conditions (Yuan et al., 2009;He et al., 2019;Tan et al., 2019). On the other hand, the mixing ratios of total NMVOCs began to decrease at 10:00 h and maintained a broad trough during daytime hours, probably due to increased photochemical removal processes favoring the destruction of NMVOCs, the elevated planetary boundary layer (PBL) promoting the dispersion of NMVOCs and/or reduced NMVOC emissions reducing the levels of NMVOCs Zheng et al., 2018). In contrast, the OVOC concentrations (Fig. S1i) increased from a minimum near sunrise and reached a maximum in the late afternoon, reflecting the accumulation of OVOCs during the photochemically active period of the day and illustrating the time profile of the formation of secondary species (Yuan et al., 2012).

OH reactivity
The OH reactivity of trace gases was categorized into SO 2 , CO, O 3 , NO x , CH 4 and total NMVOCs, which were grouped into alkanes, alkenes, aromatics and OVOCs (Table S1 lists the NMVOCs included in each group), as shown in Fig. 4a and b. The total OH reactivity was between 9.2 and 69.6 s −1 , with an average of 27.5±9.7 s −1 (± standard deviation). Statistically, the average total OH reactivity was much higher than those determined in Beijing (16.4 s −1 and 20 ± 11 s −1 ) Yang et al., 2017), Shanghai (13.5 s −1 ) ( Tan (Whalley et al., 2016) and Nashville (11.3 ± 4.8 s −1 ) (Kovacs et al., 2003) but was comparable to or lower than those in Heshan (31 ± 20 s −1 ) (Yang et al., 2017), Backgarden (mean maximum value of 50 s −1 ) (Lou et al., 2010) and New York (25 s −1 ) (Ren et al., 2006b). The OH reactivity towards SO 2 , CO and NO x was higher than the values reported in various Chinese cities (Xu et al., 2011;Zhu et al., 2020;Liu et al., 2009) (Table 1). It should be noted that the OH reactivity in this study was calculated from the sum of the products of measured species and their rate coefficients for reactions with OH and does not involve species that were not measured, such as monoterpenes and alcohols. Previous studies have shown that there are some discrepancies between the actual measured values and the calculated values of OH reactivity, which may be attributed to missing OH reactivity that originates from VOC oxidation products of both biogenic and anthropogenic origin (Di Carlo et al., 2004;Dolgorouky et al., 2012;Yoshino et al., 2006;Zhu et al., 2020). Therefore, the OH reactivity calculated in this study is somewhat underestimated.
The total OH reactivity was mainly contributed by NO x (12.0 ± 7.1 s −1 , 43.7 %), followed by NMVOCs (7.9 ± 4.8 s −1 , 28.5 %), CO (7.2 ± 2.6 s −1 , 26.0 %) and CH 4 (0.3 ± 0.1 s −1 , 1.3 %) and to a lesser extent by SO 2 and O 3 (0.2 ± 0.1 s −1 , 0.6 %), indicating the strong influence of anthropogenic emissions in Xianghe. The majority of total OH reactivity values were below 30 s −1 , as seen in the frequency distribution, which was dominated by the sum of low-OHreactivity contributions and less influenced by single compounds with high OH reactivity ( Fig. S2a-f). Specifically, the cumulative frequency distribution (Fig. S3a) clearly showed that the OH reactivity at values > 40 s −1 was dominated entirely by OH reactivity towards NO x , and the OH reactivity at values from 20-40 s −1 was nearly completely dominated by OH reactivity towards NO x and total NMVOCs. In general, the frequency distributions and cumulative frequency distributions of OH reactivity highlighted the necessity of considering a large number of species to obtain a better understanding of OH reactivity.  The OH reactivity towards total NMVOCs was 7.9 ± 4.8 s −1 , which was much lower than those in Beijing (11.2 s −1 ) and Heshan (18.3 s −1 ) (Yang et al., 2017) due to the higher content of reactive hydrocarbons (e.g., alkenes and aromatics) in Beijing and Heshan and due to the unmeasured species (e.g., acetaldehyde) in this study. Alkenes (3.4±3.7 s −1 , 42.9 %) dominated over OVOCs (2.4±1.5 s −1 , 30.2 %), aromatics (1.5±1.7 s −1 , 18.6 %) and alkanes (0.7± 0.5 s −1 , 8.3 %) in the OH reactivity towards total NMVOCs. The majority of the values of OH reactivity towards total NMVOCs were below 13 s −1 (Fig. S4a-d). The cumulative frequency distribution showed that the OH reactivity towards total NMVOCs at values of > 6 s −1 was dominated by OH reactivity towards alkenes, aromatics and OVOCs and that the OH reactivity towards total NMVOCs at values of < 6 s −1 was dominated by OH reactivity towards alkanes (Fig. S5). Alkanes accounted for > 50 % of the mixing ratio of NMVOCs but only 8.3 % of the OH reactivity towards total NMVOCs. In contrast, aromatics, alkenes and OVOCs accounted for 44.6 % of the mixing ratio of NMVOCs, providing 91.7 % of the OH reactivity towards total NMVOCs. Significantly, isoprene accounted for only 4 % of the mixing ratio of NMVOCs but provided 31.2 % of the OH reactivity towards total NMVOCs. This result was explained by (1) the relatively low concentration of aromatics, alkenes, and OVOCs measured during the campaign; (2) the relatively high concentration of isoprene; and (3) the generally large isoprene reaction rate coefficient with OH (101 × 10 −12 cm 3 molecule −1 s −1 ) (Atkinson et al., 2006). The top 10 species, in terms of OH reactivity towards total NMVOCs, consisted of isoprene, HCHO, m/p-xylene, ethylene, hexanal, o-xylene, propylene, styrene, MACR and cis-2-butene (Fig. 3b). These species contributed only 39.1 % to NMVOC emissions but accounted for 80.3 % of OH reactivity towards total NMVOCs. As shown in Table 1, the OH reactivity towards the speciated NMVOCs in this study was basically within the values reported in various Chinese cities Xu et al., 2011;Yang et al., 2017;Zhu et al., 2020).
The mean diurnal profiles of the OH reactivity of trace gases and NMVOCs are presented in Figs. 5a-f and 6a-d, respectively. In general, the total OH reactivity was the lowest in the afternoon and the highest during rush hours, reaching a maximum of 33.0 s −1 during the morning rush hour and a nighttime peak of 30.5 s −1 (Fig. 5a). Most campaigns have also reported slightly higher OH reactivity in the morning traffic rush hour, which can be explained by higher levels of reactive gases such as NO and NMVOCs due to heavy traffic, as well as slower reactions Yang et al., 2016). A similar diurnal profile was also observed for contributions from NO x , CO, alkane and aromatic species, which are typically connected to emissions from anthropogenic activities. The shape of the total OH reactivity diurnal pattern was slightly shifted in the direction of OH reactivity towards NO x , strengthening the idea that the local pollution in Xi-anghe was possibly impacted by traffic emissions. However, a different diurnal behavior to that of the above species was observed for alkenes (Fig. 6b) and OVOCs (Fig. 6d), which are emitted by plants or produced photochemically. The OH reactivity from OVOCs increased by a factor of approximately 2 from nighttime to daytime, suggesting that during the daytime, dilution or chemical removal had a weaker influence on the observed OVOCs than fresh production by photochemistry. The opposite diurnal variation was reported in Wangdu, which showed a weak diurnal variation with a decrease by a factor of approximately 2 from the morning to the evening . The diurnal profile of OH reactivity towards isoprene appears to be the major driver for the diurnal profile of OH reactivity towards alkenes. Biogenic isoprene is dependent on temperature and light intensity (Chang et al., 2014), and anthropogenic isoprene is predominantly emitted by road traffic (Ye et al., 1997); hence, the OH reactivity from alkenes increased during the daytime, with a morning peak of 4.1 s −1 at 09:00 h and a nighttime peak of 7.4 s −1 at 18:00 h. Many rainforest campaigns have also reported a significant diurnal pattern with higher OH reactivity from alkenes and OVOCs at noontime or a maximum at the beginning of the night (Yang et al., 2016). Notably, the large amplitude of the standard deviation bars highlighted the large diel variability.

NO 3 reactivity
The NO 3 reactivity of trace gases was categorized into SO 2 , NO x , CH 4 and NMVOCs, as shown in Fig. 4c and d. The campaign-averaged values of total NO 3 reactivity were 2.2± 2.6 s −1 , ranging from 0.7 to 27.5 s −1 . The average total NO 3 reactivity was much higher than those determined during the IBAIRN campaign (Liebmann et al., 2018a) and at a rural mountain site (988 m a.s.l.) in southern Germany in 2017 (Liebmann et al., 2018b) due to higher contributions from NO x . We noted that NO x was by far the main contributor to the total NO 3 reactivity, representing 99 % of the total NO 3 reactivity on average. NO exhibited the most prominent contribution to the total NO 3 reactivity and represented an average of 78.0 % of the total NO 3 reactivity. In contrast to NO, NO 2 had a maximum contribution during the nighttime and represented, on average, 27 % of the total NO 3 reactivity. The NO 3 reactivity towards CH 4 , NMVOCs and SO 2 was very minor, accounting for no more than 1 % of the total NO 3 reactivity over the whole campaign. The majority of the total NO 3 reactivity values were below 3 s −1 , but values below 5.5×10 −5 , 0.1, 3 and 1×10 −8 s −1 were observed for NO 3 reactivity towards CH 4 , total NMVOCs, NO x and SO 2 , respectively, as seen in the frequency distribution ( Fig. S2g-k). The cumulative frequency distribution clearly showed that the total NO 3 reactivity at low and high values was entirely dominated by NO 3 reactivity towards SO 2 and NO x , respectively (Fig. S3b). In total, the frequency distributions and cumulative frequency distributions of NO 3 reactivity highlighted the necessity of considering a large number of species to obtain a complete picture of NO 3 reactivity. The NO 3 reactivity towards total NMVOCs was 2.4 ± 3.0×10 −2 s −1 on average, with a minimum of 1.1×10 −3 s −1 and a maximum of 0.3 s −1 . The largest fraction of attributed NO 3 reactivity towards total NMVOCs was provided by alkenes (77.8 %), followed by aromatics (20.7 %) and OVOCs (1.3 %). The measured alkanes played virtually no role in NO 3 reactivity towards total NMVOCs, although they accounted for more than 50 % of the mixing ratio of NMVOCs. This result can be largely explained by the fact that the reaction rate coefficients of alkenes, aromatics and OVOCs with NO 3 are 1-5 orders of magnitude higher than the alkane reaction rate coefficients with NO 3 (Atkinson and Arey, 2003;Atkinson et al., 2006). The majority of the NO 3 reactivity values towards alkanes, alkenes, aromatics and OVOCs were below 5.0 × 10 −5 , 0.1, 1.0 × 10 −2 and 1.0 × 10 −3 s −1 , respectively (Fig. S4e-f). The cumulative frequency distribution showed that the NO 3 reactivity towards total NMVOCs at values of > 0.1 s −1 was entirely dominated by NO 3 reactivity towards alkenes, the NO 3 reactivity towards total NMVOCs at values from 0.01 to 0.1 s −1 was dominated by NO 3 reactivity towards alkenes and aromatics, and the NO 3 reactivity towards total NMVOCs at values of < 1.0 × 10 −5 s −1 was entirely dominated by NO 3 reactivity towards alkanes (Fig. S6). The top 10 species in terms of NO 3 reactivity towards total NMVOCs consisted of isoprene, styrene, cis-2-butene, trans-2-butene, cis-2-pentene, hexanal, HCHO, propylene, 1,3-butadiene and trans-2-pentene (Fig. 3c). These species contributed only 27.7 % to NMVOC emissions but accounted for 99.2 % of the NO 3 reactivity towards total NMVOCs. Total NO 3 reactivity displayed a weak diel variation, with a campaign-averaged morning peak value of 4.0 s −1 at 06:00-07:00 h (Fig. 5g). The diurnal profile of NO 3 reactivity towards NO x (Fig. 5i) appears to be the major driver for the diurnal profile of total NO 3 reactivity. The morning peak value of total NO 3 reactivity could be explained by the accumulation of NO x due to traffic emissions that are released into the shallow nocturnal boundary layer during the morning rush hours. In contrast, the average diurnal profile of NO 3 reactivity towards total NMVOCs (Fig. 5k) had a maximum at 18:00 h, which was slightly shifted in the direction of NO 3 reactivity towards alkenes (Fig. 6j). The evening peak value of NO 3 reactivity towards total NMVOCs could be accounted for by the accumulation of alkenes due to vegetation emissions and traffic emissions that are released into the shallow nocturnal boundary layer. NO 3 reactivity towards alkanes (Fig. 6e), alkenes (Fig. 6f), aromatics (Fig. 6g), OVOCs (Fig. 6h) and SO 2 (Fig. 5h) played virtually no role in the diurnal variations in total NO 3 reactivity and NO 3 reactivity towards total NMVOCs but exhibited a more distinct diurnal profile.

O 3 reactivity
The O 3 reactivity of trace gases was categorized into NO x , CH 4 and total NMVOCs, as shown in Fig. 4e and f. The total O 3 reactivity at the site varied between a minimum of 3.3 × 10 −4 s −1 and a maximum of 1.8 × 10 −2 s −1 and was 1.2 ± 1.7 × 10 −3 s −1 on average. NO exhibited the most prominent contribution to the total O 3 reactivity and represented > 99 % of the total O 3 reactivity on average, whereas nearly all other contributions were < 1 %. This result can be largely accounted for by the generally large NO reaction rate coefficients with O 3 (1.8 × 10 −14 cm 3 molecule −1 s −1 ) (Atkinson et al., 2006), which are several orders of magnitude higher than the reaction rate coefficients of NO 2 , alkanes, alkenes, aromatics and OVOCs with NO 3 (Atkinson et al., 2006;Atkinson and Arey, 2003;Yuan et al., 2013;Ferracci et al., 2018;Jenkin et al., 2015). The majority of the total O 3 reactivity values were below 2 × 10 −3 s −1 , but values below 5.5 × 10 −10 , 2 × 10 −6 and 2 × 10 −3 s −1 were observed for the O 3 reactivity towards CH 4 , total NMVOCs and NO x , respectively, as seen in the frequency distribution ( Fig. S2l-o). The cumulative frequency distribution clearly showed that the total O 3 reactivity at low and high values was entirely dominated by O 3 reactivity towards CH 4 and NO x , respectively (Fig. S3c). Generally, the frequency distributions and cumulative frequency distributions of O 3 reactivity highlight the necessity of considering a large number of species to obtain a complete picture of O 3 reactivity.
Compared with the OH and NO 3 reactivities, O 3 reactivity displayed a much weaker diel variation, especially the O 3 reactivity towards alkenes and aromatics, as shown in Figs. 5 and 6. This weakness can be explained by the following reasons. First, for a given species, the reaction rate coefficients with O 3 were much smaller than the corresponding reaction rate coefficients with OH and NO 3 . For example, the ethylene reaction rate coefficients with OH (8.52 × 10 −12 cm 3 molecule −1 s −1 ) and NO 3 (2.05 × 10 −16 cm 3 molecule −1 s −1 ) are 6 and 2 orders of magnitude higher, respectively, than the ethylene reaction rate coefficient with O 3 (1.59 × 10 −18 cm 3 molecule −1 s −1 ) (Atkinson and Arey, 2003; Atkinson et al., 2006). Second, the high-emission species reaction rate coefficients with O 3 are smaller than the low-emission species reaction rate coefficients with O 3 . For instance, the m/p-xylene (one of the top five species in terms of emissions) reaction rate coefficient with O 3 (< 1.0 × 10 −20 cm 3 molecule −1 s −1 ) are much smaller than the 1-hexene (one of the bottom five emissions species) reaction rate coefficients with O 3 (1.13 × 10 −17 cm 3 molecule −1 s −1 ) (Atkinson and Arey, 2003;Atkinson et al., 2006). The above two factors largely weaken the diurnal variation in O 3 reactivity. Table 2 lists the top 10 NMVOC species (excluding isoprene) in terms of concentration and OH, NO 3 , and O 3 re- activities and their corresponding contributions to concentrations and OH, NO 3 , and O 3 reactivities. The order of the major OH, NO 3 and O 3 reactivity-contributing species differed significantly from that of concentration-contributing species. Therefore, NMVOC control strategies based on OH, NO 3 and O 3 reactivities differ significantly from those based on concentrations.

Implications for OH, NO 3 and O 3 reactivity-based NMVOC control strategies
From the perspective of concentration, HCHO, propane, acetone, ethane, n-butane, m/p-xylene, iso-pentane, ethylene, iso-butane and n-pentane should be targeted. If these 10 species were fully controlled, it would lead to an NMVOC concentration reduction of 79.9 % with OH, NO 3 and O 3 reactivity reductions of only 58.4 %, 2.1 % and 6.4 %, respectively. These species are mainly from fuel combustion and vehicle exhaust (Song et al., 2018;Liu et al., 2017); hence, from the perspective of the current emission-based limits, we recommend that the priorities for the control of NMVOC sources include fuel combustion and vehicle exhaust.
From the perspective of OH reactivity, HCHO, m/pxylene, ethylene, hexanal, o-xylene, propylene, styrene, MACR, cis-butene and MVK were the key species. If the releases of these compounds were reduced to zero without any offset, OH reactivity would be reduced by 73.3 % with an NMVOC concentration reduction of 38.1 %, a NO 3 re-activity reduction of 86.4 % and O 3 reactivity reduction of 55.7 %. From the perspective of NO 3 reactivity, the top 10 NMVOC species consisted of styrene, cis-2-butene, trans-2-butene, cis-2-pentene, hexanal, HCHO, propylene, 1,3butadiene, trans-2-pentene and 1-butene. If the concentrations of these species were completely eliminated, it would reduce NO 3 reactivity by 97.8 % with an NMVOC concentration reduction of 25.8 %, an OH reactivity reduction of 49.7 % and an O 3 reactivity reduction of 91.8 %. From the perspective of O 3 reactivity, cis-2-butene, trans-2-butene, cis-2-pentene, propylene, styrene, ethylene, 1-butene, trans-2-pentene, 1-pentene and MACR should be the key targets for control. If the concentrations of these compounds were reduced to zero without any offset, it would lead to an O 3 reactivity reduction of 98.9 % with an NMVOC concentration reduction of 7.3 %, an OH reactivity reduction of 22.3 % and a NO 3 reactivity reduction of 94.2 %. The top 10 species associated with OH, NO 3 and O 3 reactivities are mainly from traffic-related emissions, industry and solvent usage (Song et al., 2018;Liu et al., 2017;Chen et al., 2014). Therefore, in terms of reactivity-based limits, we recommend that the priorities for the control of NMVOC sources include trafficrelated emissions, industry and solvent usage. Clearly, species with large concentrations do not necessarily have high OH, NO 3 and O 3 reactivities, and a small concentration reduction can result in a maximum reduction in reactivity. The key NMVOC species in terms of OH, NO 3 and O 3 reactivities also differed from each other. However, reducing the concentrations of propylene, styrene and cis-2butene may likely achieve a win-win-win situation. Although the above comparisons were made under the assumption that concentrations would be significantly reduced, it is obvious that OH, NO 3 and O 3 reactivity-based control strategies are more efficient than concentration-based policies in terms of reducing NMVOC pollution. Overall, the combined integration and comparison of OH, NO 3 and O 3 reactivities towards NMVOCs could provide useful suggestions for VOC pollution control in the North China Plain.

AOC
3.4.1 Modeling OVOCs, OH, HO 2 , RO 2 and NO 3 by SOSAA With the appropriate setup of the condensation sinks for the 10 calculated OVOCs (ACR, C 2 H 5 CHO, MACR, C 3 H 7 CHO, MVK, MEK, MPRK, C 4 H 9 CHO, DIEK and C 5 H 11 CHO), the modeled diurnal mean pattern generally followed the measured pattern within 1 standard deviation of the measurement data, although the model underestimated measurements, predicting values of less than 1 ppb from 19:00 to 24:00 h (Fig. S8a). With the inclusion of input MTBE (methyl tert-butyl ether) and CH 3 COCH 3 (acetone), which constituted more than 50 % of the total OVOCs, the modeled total OVOC concentration agreed better with the measurements than expected (Fig. S8b). The modeled diurnal median number concentrations of OH, HO 2 and RO 2 showed an apparent diurnal pattern with peaks during midday and values approaching zero during night, which resulted from the dependence of their chemical production reactions on incoming solar radiation (Fig. S9a, b and c). The midday time (12:00-16:00 h) median number concentrations of OH, HO 2 and RO 2 were 1.2 × 10 7 , 5.9 × 10 8 and 3.7 × 10 8 molecules cm −3 , respectively, which were comparable to previous studies . The diurnal variation of in the hourly median NO 3 concentration showed two peaks which were consistent with the high chemical production from NO 2 + O 3 (Fig. S9d). Figure S10 shows the relationship between the modeled OH number concentration and the measured JO 1 D. The coefficient of determination (R 2 ) was 0.86, and the linear regression fit showed that the slope was 6.1 × 10 11 cm −3 s −1 and the intercept was 0.9 × 10 6 cm −3 . These values were comparable to , except that the slope here was approximately 36 % higher than the observed fit in Tan et al. (2017).

Overall characteristics of AOC
The loss rates of NMVOCs, CH 4 and CO via reactions with OH, O 3 and NO 3 were calculated. The calculated AOC was up to 4.5 × 10 8 molecules cm −3 s −1 , with a campaignaveraged value of 7.8×10 7 molecules cm −3 s −1 , daytime average (06:00-18:00 h) of 1.4 × 10 8 molecules cm −3 s −1 and nighttime average of 6.7 × 10 6 molecules cm −3 s −1 . As such, the total number of NMVOC, CH 4 and CO molecules depleted during the daytime and nighttime were, respectively, 6.0 × 10 12 and 2.9 × 10 11 per cm −3 of air. These AOC levels were higher than those determined at the Tung Chung air quality monitoring station (Xue et al., 2016), a polluted area in Santiago, Chile (Elshorbany et al., 2009), and Hong Kong Polytechnic University's air monitoring station at Hok Tsui .
Comparisons of the AOC values calculated from OH, O 3 , and NO 3 and the corresponding oxidation concentrations are shown in Fig. 7. The OH and NO 3 radical concentrations were simulated by the SOSAA box model. The AOC calculated from OH, O 3 and NO 3 correlated well with the corresponding oxidation concentrations, with correlation coefficients (r) of 0.91, 0.83 and 0.57, respectively, suggesting that the parameterized AOC here was consistent with that obtained using radical concentration to indicate AOC. Specifically, the average oxidation capacities of OH, O 3 and NO 3 radicals throughout the entire campaign were 7.7×10 7 , 1.2× 10 6 and 1.8 × 10 5 molecules cm −3 s −1 , representing 98.2 %, 1.5 % and 0.3 % of the total oxidation capacity, respectively. The total number of depleted molecules per day due to oxidation by OH, O 3 and NO 3 was 6.6 × 10 12 , 1.0 × 10 11 and 1.5 × 10 10 molecules cm −3 , respectively; these values were slightly higher than those assessed in a polluted area in Santiago, Chile (Elshorbany et al., 2009). Accordingly, OH radicals are the driving force of AOC in Xianghe, especially during the daytime. Figure 8 shows a comparison of the oxidation capacities of OH, O 3 and NO 3 . On average, the relative contribution of O 3 and NO 3 oxidation capacities when integrated over 24 h was less than 4 % (Fig. 8a-c). OH is the only oxidant of CO in the troposphere. As expected, OH was responsible for 99 % of the oxidation capacity regarding NMVOCs, CH 4 and CO during the daytime (Fig. 8d). The relative contribution of OH to oxidation capacity decreased to 98 % when restricting the calculation to NMVOC families alone (Fig. 8e). Focusing on the oxidation of unsaturated NMVOCs, OH was the dominant oxidant with a relative proportion of approximately 97 % (Fig. 8f). Note that the influence of NO 3 and O 3 on the oxidation of CO and VOCs can be neglected during the daytime. However, elevated relative contributions of O 3 and NO 3 to oxidation capacity can be observed during the nighttime. As expected, O 3 and NO 3 accounted for 10 % and 2 %, respectively, of the oxidation capacity with respect to NMVOCs, CH 4 and CO (Fig. 9g), but 19 % and 3 % of NMVOC families alone (Fig. 8h) occurred at night. Focusing on the oxidation of unsaturated NMVOCs, O 3 and NO 3 accounted for 20 % and 4 %, respectively, of the oxidation capacity (Fig. 8i). This quantitative intercomparison of the oxidation capacities of OH, O 3 and NO 3 confirms the important role of OH in the degradation of NMVOCs, CH 4 and CO. Compared with OH and O 3 , NO 3 had a lower contribution during both the daytime and nighttime, which was mainly caused by high NO concentrations (Liebmann et al., 2018b).

Summary and conclusions
In the summer of 2018, a comprehensive field campaign was conducted at a suburban site in the North China Plain. Based on simultaneous measurements of O 3 , CO, SO 2 , NO, NO 2 , JO 1 D, JNO 2 , JNO 3 , HONO, HCHO, CH 4 and 65 NMVOCs, the reactivities (OH, NO 3 and O 3 reactivities) towards trace gases and AOC were comprehensively analyzed. The main findings are summarized as follows.
The total OH reactivity was between 9.2 and 69.6 s −1 with an average of 27.5 ± 9.7 s −1 , which was mainly contributed by NO x (43.7 %), followed by NMVOCs (28.5 %), CO (26.0 %), CH 4 (1.3 %), and SO 2 and O 3 (0.5 %). OH reactivity towards total NMVOCs was 7.9 ± 4.8 s −1 and dominated by alkenes (42.9 %). The campaign-averaged value of total NO 3 reactivity was 2.2 ± 2.6 s −1 , ranging from 0.7 to 27.5 s −1 . NO x was the main contributor to the total NO 3 reactivity, representing 99 % of the total NO 3 reactivity on average. NO 3 reactivity towards total NMVOCs was 2.4 ± 3.0 × 10 −2 s −1 , on average, and it was dominated by alkenes (77.8 %). The total O 3 reactivity varied between a minimum of 3.3 × 10 −4 s −1 and a maximum of 1.8 × 10 −2 s −1 , with an average of 1.2 ± 1.7 × 10 −3 s −1 . NO exhibited the most prominent contribution to the total O 3 reactivity and represented an average of > 99 % of the total O 3 reactivity. The O 3 reactivity towards total NMVOCs was 1.1 ± 0.8 × 10 −6 s −1 on average, ranging from 2.5 × 10 −7 to 1.0 × 10 −5 s −1 and dominated by alkenes (94.0 %). The total OH, NO 3 and O 3 reactivities displayed a similar diel variation with the lowest value in the afternoon and the highest value during rush hours, and the diurnal profile of NO x appears to be the major driver for the diurnal profiles of total OH, NO 3 and O 3 reactivities. Compared with the OH and NO 3 reactivities, O 3 reactivity displayed a much weaker diel variation, especially the O 3 reactivity towards alkenes and aromatics due to (1) the rate coefficients with O 3 being much smaller than the corresponding reaction rate coefficients with OH and NO 3 for the same species and (2) the high-emission species reaction rate coefficients with O 3 being smaller than the low-emission species reaction rate coefficients with O 3 . OH, NO 3 and O 3 reactivity-based control strategies are more efficient than concentration-based policies in terms of reducing NMVOC pollution. We suggest that policy makers shift the current concentration-based limits to reactivitybased policies.
Our study provides useful insights for VOC pollution control in a typical suburban site in the North China Plain. Further studies, especially direct observations of OH and NO 3 radicals, OH and NO 3 reactivity measurements and speciated measurements, are required to further explore trace gas reactivity and AOC.
Data availability. The underlying research data can be accessed upon contact with the corresponding author (Yuesi Wang: wys@mail.iap.ac.cn; Yonghong Wang: yonghong.wang@helsinki.fi).