ROx Budgets and O3 Formation during Summertime at Xianghe Suburban Site in the North China Plain

Min XUE, Jianzhong MA, Guiqian TANG, Shengrui TONG, Bo HU,Xinran ZHANG, Xinru LI, and Yuesi WANG*

1State Key Laboratory of Severe Weather & China Meteorological Administration Key Laboratory of Atmospheric Chemistry, Chinese Academy of Meteorological Sciences, Beijing 100081, China

2State Key Laboratory of Atmospheric Boundary Layer Physics and Atmospheric Chemistry, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, China

3State Key Laboratory for Structural Chemistry of Unstable and Stable Species, Beijing National Laboratory for Molecular Sciences, Chinese Academy of Sciences Research/Education Center for Excellence in Molecular Sciences,Institute of Chemistry, Chinese Academy of Sciences, Beijing 100190, China

4Department of Chemistry, Capital Normal University, Beijing 100081, China

ABSTRACT Photochemical smog characterized by high concentrations of ozone (O3) is a serious air pollution issue in the North China Plain (NCP) region, especially in summer and autumn. For this study, measurements of O3, nitrogen oxides (NOx),volatile organic compounds (VOCs), carbon monoxide (CO), nitrous acid (HONO), and a number of key physical parameters were taken at a suburban site, Xianghe, in the NCP region during the summer of 2018 in order to better understand the photochemical processes leading to O3 formation and find an optimal way to control O3 pollution. Here, the radical chemistry and O3 photochemical budget based on measurement data from 1−23 July using a chemical box model is investigated. The daytime (0600−1800 LST) average production rate of the primary radicals referred to as ROx (OH + HO2+ RO2) is 3.9 ppbv h−1. HONO photolysis is the largest primary ROx source (41%). Reaction of NO2 + OH is the largest contributor to radical termination (41%), followed by reactions of RO2 + NO2 (26%). The average diurnal maximum O3 production and loss rates are 32.9 ppbv h−1 and 4.3 ppbv h−1, respectively. Sensitivity tests without the HONO constraint lead to decreases in daytime average primary ROx production by 55% and O3 photochemical production by 42%,highlighting the importance of accurate HONO measurements when quantifying the ROx budget and O3 photochemical production. Considering heterogeneous reactions of trace gases and radicals on aerosols, aerosol uptake of HO2 contributes 11% to ROx sink, and the daytime average O3 photochemical production decreases by 14%. The O3-NOx-VOCs sensitivity shows that the O3 production at Xianghe during the investigation period is mainly controlled by VOCs.

Key words: photochemical smog, ROx budgets, O3 production rate, O3 chemical regime

1. Introduction

Photochemical smog has become a serious air pollution issue in urban and industrial centers in recent decades(Molina and Molina, 2004; Zhang et al., 2008; Monks et al.,2009; Wang et al., 2017). Ois one of the major contributors to photochemical smog and is produced via photochemical reactions involving nitrogen oxides (NO= NO + NO),carbon monoxide (CO) and volatile organic compounds(VOCs). High concentrations of tropospheric Oaffect human health, ecosystems, and climate (Molina et al., 2010;Monks et al., 2015). However, Opollution is difficult to control, mainly because the photochemical production of Ohas a nonlinear dependence on its precursors, NOand VOCs. The sensitivity of Oproduction to its precursors varies greatly from one region to another mainly due to diverse source types and emission rates of NOand VOCs. Therefore, understanding the photochemical processes leading to Oformation is critical for efficient ozone pollution control.

The hydroxyl radical (OH), hydroperoxy radical (HO),and organic peroxy radical (RO), collectively referred to as RO, control the removal of primary pollutants and formation of secondary pollutants such as Oand secondary aerosols (Hofzumahaus et al., 2009; Stone et al., 2012). OH is primarily produced by photolysis of O, nitrous acid(HONO), and hydrogen peroxide (HO), as well as alkene ozonolysis reactions (Levy, 1971; Ehhalt, 1999; Ma et al.,2012). As the most reactive oxidant, OH initiates the oxidation reactions of most primary pollutants in the atmosphere.OH reacts with CO and VOCs to produce HOand ROradicals, respectively. ROis converted to HOby reacting with NO. HOfurther reacts with NO to recycle OH. The recycling reactions between OH, HO, and ROin the presence of VOCs and NOlead to the catalytic cycle between NO and NOand eventually result in net production of O(Kuhn et al., 2010). ROradicals are ultimately removed by reacting with NOunder high NOlevels to produce nitric acid (HNO) and organic nitrates (RO, NO, and RONO),or by self-reactions under low NOlevels to produce peroxides (HOand ROOH). ROchemistry has been a focus in understanding Ophotochemical processes. A number of studies have investigated RObudgets for areas all over the world with varying characteristics, such as polluted urban and suburban areas (Ren et al., 2003; Emmerson et al.,2005a; Sheehy et al., 2010; Volkamer et al., 2010; Ma et al.,2012; Xue et al., 2016), rural areas (Hofzumahaus et al.,2009; Lu et al., 2012; Tan et al., 2017), and remote areas(Kubistin et al., 2010; Lelieveld, 2010; Mao et al., 2010).

The North China Plain (NCP) is one of the most industrialized and urbanized regions in eastern China. Photochemical smog characterized by high concentrations of Ohas been regarded as the most prevalent air pollution issue in the NCP on the regional scale (Ma et al., 2012; Wang et al.,2017). With increased emissions mitigation since the 2000s,primary pollutant levels have declined, but Oconcentrations have experienced an increasing trend in both urban,background and mountainous areas (Tang et al., 2009;Zhang et al., 2014; Ma et al., 2016; Sun et al., 2016; Lu et al., 2018; Li et al., 2019). The formation mechanism of Opollution in the NCP has been investigated by numerous studies in last two decades. Most studies have focused on the impact of urban plumes on Opollution at the regional scale(Wang et al., 2006; Lin et al., 2008; Ma et al., 2013; Xue et al., 2013), local production and regional transport contributions (Streets et al., 2007; Wang et al., 2009; Wu et al.,2017), precursor-Orelationships (Shao et al., 2009; Lu et al., 2010; Tang et al., 2012), and Osource attributions (Qu et al., 2014; Li et al., 2017; Lu et al., 2019). Some studies have investigated radical sources and atmospheric oxidizing capacity to understand ozone photochemical processes using an observation-based modeling approach (Liu et al.,2012; Ma et al., 2012; Tan et al., 2019). Liu et al. (2012) analyzed RObudgets and Ophotochemistry for summertime in Beijing and found that photolysis of oxygenated VOC(OVOC) and HONO was the dominant primary ROsource,which resulted in high production of O. Ma et al. (2012)investigated the HOand Ophotochemical process and concluded that the lower atmosphere over the NCP was highly polluted and served as an oxidation pool, under which primary pollutants were quickly oxidized. Tan et al. (2019)reported that HONO and HCHO photolysis contributed about 50% of the total primary ROsource in Beijing and high atmospheric oxidation capacity led to Opollution in Chinese megacities. Most studies concentrated on Beijing urban areas, and only a few focused on its surrounding suburban areas (Lu et al., 2013; Tan et al., 2017). In addition to Beijing, Opollution is serious at its surrounding area, such as Xianghe site, with the highest frequency of exceeding national air quality thresholds (Tang et al., 2012). However,the radical chemistry and photochemical processes leading to Oformation for the Xianghe site have not yet been documented, and therefore must be investigated in order to illustrate the underlying chemical mechanism of Opollution on a regional scale.

For this study, measurements of Oand its precursors were taken at the Xianghe suburban site in the NCP region during the summer of 2018 and used to constrain a chemical box model in order to investigate the radical budget and Ophotochemistry of the region. A description of the observation site and measurements and an introduction of the chemical box model are given in section 2. Then, an analysis of the observation data taken over the period of 1−23 July is given, and the modeling analysis results, including the ROradical budgets and Oproduction/loss rates, as well as their sensitivities to NO, VOCs, HONO, and heterogeneous reactions on aerosols are presented in section 3. Finally, conclusions are provided in section 4.

2. Methods

2.1. Sampling site and measurements

The field campaign was conducted at the Xianghe Atmospheric Observatory (39.798°N, 116.958°E; 35 m above sea level) during the summer of 2018. Xianghe is a typical suburban site, about 50 km, 75 km, and 35 km away from Beijing, Tianjin, and Langfang in Hebei province, respectively. The site is surrounded by agricultural land and residential areas. The downtown center (population of about 30 000)is about 4 km east of the site.

Trace gases (including O, NO, and CO) were measured by commercial instruments from Thermo Fisher Scientific, USA. The instruments were located on the roof of a single-story building. O, NO, and CO were continuously measured using the Oanalyzer (Model 49i), the NO-NO-NOchemiluminescence analyzer (Model 42i), and the CO analyzer (Model 48i), respectively. The precision and accuracy, operation and maintenance, and calibration methods for these instruments have previously been described in detail(Ji et al., 2014). Ambient VOC samples were analyzed continuously using a gas chromatography (GC) instrument(7820A, Agilent Technologies, Santa Clara, CA, USA)equipped with a mass spectrometer (MS) and a flame ionization detector (FID) (5977E, Agilent Technology, Santa Clara, CA, USA) with a time resolution of 1 h. Detailed description of the GC-MS/FID analysis method can be found in Wang et al. (2014). Dual columns and detectors were applied for the simultaneous analysis of C2−C12 hydrocarbons. The VOCs measurements include C2−C12 alkanes, C2−C6 alkenes, and C6−C10 aromatics. HONO was measured using a custom-made HONO analyzer, which was described in detail by Hou et al. (2016). Briefly, the principle of a HONO analyzer is similar to a long path absorption photometer (LOPAP) (Kleffmann et al., 2002). Information about detection limit and data accuracy are presented in Table S1 in the electronic supplementary material (ESM). Further details about the instruments including their operation and calibration procedures are available in the supplemental material.

The photolysis frequencies (JOD, JNO, JHONO,JHO, JHCHO, and JNO) were measured using a PFS-100 spectroradiometer (Focused Photonics Inc., Hangzhou,China), which was described in detail by Zhao et al. (2021).The mixing layer height was observed by ceilometer (CL31,Vaisala, Finland), described in detail by Tang et al. (2016).Other meteorological parameters, such as the ambient temperature, pressure, and relative humidity were simultaneously measured at the Xianghe Atmospheric Observatory. Measurement results from 1−23 July, the period during the experiment when the photochemical pollution was the most severe, are presented in this paper. The time series of meteorological parameters are shown in Fig. S1 in the ESM.

2.2. Chemical box model

A box model was used to explore the radical budget and Ophotochemical formation and loss. The model is built on the NCAR Master Mechanism (Madronich and Calvert, 1990; Ma et al., 2002, 2012), which is an explicit gas phase chemical mechanism and calculates the local photochemical equilibrium with a time scale of several minutes.In addition, heterogeneous reactions of trace gases and radicals on aerosol surfaces were also incorporated in the model(see section 2.2 of Ma et al., 2012). Meanwhile, dry deposition and entrainment losses were also included in the model to account for the physical loss processes. The model was constrained by the concentrations of measured trace gases, including inorganic species (HO, NO, NO, O, HONO, and CO)and VOCs. The observed individual VOC species (Table S2 in the ESM) were applied to the model simulations in the same way as used by Ma et al. (2012). Measurements of photolysis frequencies, mixing layer height, and meteorological parameters were also used to constrain the model simulations.

The measured data were averaged or interpolated with a time resolution of 1 hour for the model constraints. Simulations were performed for the time period from 1−23 July starting at 0000 LST (LST = UTC + 8), with a five-day pre-run in advance to let the unconstrained compounds to reach steady state, and then the results after that were used for analyses. The model calculates radical concentrations and chemical reaction rates, which can be used to further quantify radical budget and Ophotochemical production. The uncertainty of model calculation consists of uncertainty of the measurements used to constrain the model and uncertainty of the kinetic reaction rate. OVOCs concentrations were simulated by the model. The mean observed concentration of HCHO was 1.2 ± 0.9 ppbv for the period from Nov 2017 to Jan 2018 at Xianghe (Tang et al., 2019) and 4.4 ± 1.6 ppbv in July 2018 (X. R. LI, personal communication). The mean simulated HCHO concentrations in our study was 3.7 ± 1.3 ppbv, comparable to the other observed results at the same site.

3. Results and discussion

3.1. Observed O3 and precursor analysis

Figure 1 presents the mean diurnal variations of measured NO, CO, HONO, AVOC (anthropogenic VOC), isoprene, O, and O(O+NO). The primary anthropogenic pollutants including NO, CO, and AVOC have similar diurnal variations and all reach a maximum in the early morning and a minimum in the afternoon. The maximum in the morning is related to gasoline and diesel vehicle exhaust (Tang et al., 2019) during rush hour coinciding with a shallow planetary boundary layer (PBL). In the afternoon, the pollutants decrease to minimum due to strong dilution with the development of the PBL. During the nighttime, the pollutants tend to accumulate when the PBL decreases. HONO has a similar variation pattern as NO. Heterogeneous reactions of NOare expected to be the most important source of HONO(Liu et al., 2021) and a good correlation between HONO and NOduring different field studies has been reported by Elshorbany et al. (2012). In this study, the averaged daytime HONO concentration is about 0.9 ppbv, comparable to those observed in Beijing (Lu et al., 2010) and Hong Kong(Xue et al., 2016), and higher than observations reported in New York (Ren et al., 2003) and Birmingham (Emmerson et al., 2005b). Isoprene emissions include both biogenic and anthropogenic sources. Biogenic emissions have a good correlation with temperature. Different to the diurnal variation of AVOC, the diurnal maximum of isoprene occurrs in the afternoon (1500 LST), indicating that isoprene is mostly from biogenic rather than anthropogenic emissions at the Xianghe site.

Fig. 1. Averaged diurnal variations of measured NOx, CO, HONO, AVOC, isoprene, O3, and Ox at the Xianghe site.The vertical bars show the hourly standard deviation.

Ohas a variation pattern that is typical for photochemical pollution with a maximum around 76 ppbv in the afternoon. The hourly maximum Ois 136 ppbv, much higher than the Chinese National Air Quality Standard Grade II,indicating serious Opollution at the Xianghe site. During the daytime, with incoming solar radiation, the rapid chemical cycle between NO, NO, and Oconverts some Oto NO. Therefore, total oxidant O, the sum of Oand NO, is used to indicate Opollution, especially in the urban and suburban areas where NO concentrations are high (Liu et al.,1987; Lu et al., 2010). The mean Odiurnal profile shows a maximum value of 83 ppbv. Mean concentrations plus/minus standard deviations of O, O, NO, NO, CO, HONO,isoprene, and AVOC were 46.1 ± 28.8 ppbv, 57.2 ± 26.5 ppbv, 1.9 ± 1.8 ppbv, 13.0 ± 6.0 ppbv, 694.6 ± 478.1 ppbv,1.0 ± 0.5 ppbv, 0.7 ± 0.9 ppbv, and 20.0 ± 11.0 ppbv, respectively.

From field campaign data, 56 VOC species including alkanes, alkenes, aromatics, and isoprene are quantitatively identified. The reactivity of individual VOC species is calculated using observed VOC concentration multiplied by its corresponding reaction rate constant with OH at 298 K (Atkinson and Arey, 2003). Figure 2 shows the percentage contributions of alkanes, alkenes, aromatics, and isoprene to measured total VOC concentration and calculated total VOC reactivity. The average mixing ratio and reactivity of total VOC at Xianghe are 20.7 ppbv and 4.3 s, respectively, comparable to the values observed at Yufa suburban site(39.514°N, 116.305°E) in Beijing in summer 2006 (Shao et al., 2009; Lu et al., 2010), and much lower than results(41.8 ppbv and 15.8 s) measured at the Shangdianzi(SDZ) regional background station (40.650°N, 117.117°E)for summertime of 2007 (Xu et al., 2011). The difference between Xianghe and SDZ could be caused by different sampling times (samples on each Tuesday at 0800−0830LST and 1400−1430 LST were measured at SDZ, see section 2.2 of Xu et al., 2011). As shown in Fig. 2a, alkanes contribute 62.2% to the total VOC mixing ratio. The contributions of alkenes and aromatics are the same with the value of 17.3%, and isoprene only contributes 3.2%. According to Lu et al. (2010), the contributions of alkanes, alkenes, aromatics and isoprene to the total VOC mixing ratio at the Yufa site were 53%, 24%, 21%, and 2%, respectively, with higher contributions of alkenes and aromatics than results at Xianghe. With respect to the reactivity, however, isoprene accounts for 38.6% of the measured VOC reactivity, higher than that of alkenes, aromatics, and alkanes. The contributions of alkenes and aromatics groups are comparable, with a slightly different portion of 21.1% and 24.7%, respectively, and alkanes only account for 15.6%. The above results indicate the high reactivity of isoprene at this suburban site.

Fig. 2. Percentage contributions of alkane, alkene, aromatic, and isoprene to measured VOC concentration (a) and reactivity (b) at the Xianghe site.

3.2. Modeled OH reactivity

The total OH reactivity (K), equivalent to the inverse of the OH lifetime, is defined as the pseudo-first-order rate coefficient of OH radical in ambient atmosphere. The Kis calculated by the following equation (Lou et al., 2010;Mao et al., 2010):

where [X] is the ambient concentration of individual reactive species (NO, CO, VOCs, and OVOC, etc.),

K

is the rate coefficient of the reaction between Xand OH radical, and

K

[X] represents the reactivity of X. Figure 3a shows the average diurnal variation of Kcalculated with the box model, including the reactivity of measured CO,NO, alkanes, alkenes, aromatics, isoprene, and model simulated other secondary species (HCHO, acetaldehyde, acetone, etc.). The Kranges between 12 sand 18 s, with the maximum value occurring around 0800 LST (the morning rush hour), which is similar to the diurnal variation pattern shown in previous studies (Lou et al., 2010; Fuchs et al., 2017). The diurnal profile of OH reactivity is relatively flat, with less decrease than reactive gas concentrations during daytime, because of the high contribution of isoprene in the afternoon.

Fig. 3. Averaged diurnal variations of the contributions from all measured species and model generated species(denoted as Others) (a). Percent contributions of all measured species and model generated species to the total OH reactivity (b).

Over the past two decades, direct measurements of Khave been performed in different regions, such as New York (Ren et al., 2003), Mexico (Shirley et al., 2006), Paris(Dolgorouky et al., 2012), Guangzhou (Lou et al., 2010;Yang et al., 2017), and Beijing (Lu et al., 2013; Williams et al., 2016). According to the observed results from Guangzhou (Lou et al., 2010; Yang et al., 2017), the Kvalues at a rural site, Backgarden (23.49°N, 113.03°E), in summer 2006 and a suburban site, Heshan (22.73°N, 112.93°E), in autumn 2014 were in the range of 20 to 50 s. The Kwas also measured at the Yufa site in summer 2006, with the average values falling in the range of 10−30 s(Lu et al.,2013). At the rural Wangdu site (39.514°N, 116.305°E) in the NCP, the measured Kranged between 10−20 sin summer 2014 (Fuchs et al., 2017). The calculated OH reactivity at the Xianghe site is comparable to that measured at Wangdu in the NCP and lower than those observed in Guangzhou.

Figure 3b shows the percentage contributions of different species to the total OH reactivity. The sum of CO and NOaccounts for 49.1% of total OH reactivity, with CO contributing 27.5% (4.1 s) and NOcontributing 21.6%(3.2 s), respectively. The AVOCs, including alkanes,alkenes, and aromatics, contribute 17.6% to the total OH reactivity. The isoprene reactivity accounts for 11.1%.Model calculated other secondary species (HCHO, acetaldehyde, acetone, etc.) account for the remaining 22.2%. Our results show that the sum of VOCs and secondary species contributes to 50% of the total OH reactivity.

3.3. Simulated ROx radical concentrations

Measurements of ROradicals in China are relatively rare. In this study, ROradical concentrations are calculated using the NCAR-MM chemical box model, constrained by the observed hourly gas concentrations and physical parameters. Figure 4 presents the mean diurnal variations of OH, HO, and ROradicals at the Xianghe site. Radical concentrations show typical diurnal variations with a maximum occurring at noon. The maximum concentrations of OH, HO, and ROin the averaged diurnal cycle at Xianghe are 10.6 × 10, 5.5 × 10, and 3.7 × 10molecules cm,respectively, comparable to those simulated at the Peking University campus (PKU) site in August 2007, with corresponding values of 9 × 10, 6.8 × 10, and 4.5 × 10molecules cm(Liu et al., 2012). According to the simulated hourly radical concentrations (Fig. S2 in the ESM), the daily maximums of OH, HO, and ROat Xianghe have ranges of(4−17) × 10, (1−13) × 10, and (1−11) × 10molecules cm, respectively. Observed daily maximum concentrations of OH and HOhad ranges of (4−17) × 10and (2−24)× 10molecules cmat the Yufa suburban site in August 2006 (Lu et al., 2013). Daily maximum OH, HO, and ROconcentrations were also observed with ranges of (5−15) ×10, (3−14) × 10and (3−15) × 10molecules cm, respectively, at the Wangdu rural site in the NCP in summer 2014(Tan et al., 2017). In contrast, higher daily maximum values of (15−26) × 10molecules cmfor OH and (3−25) ×10molecules cmfor HOwere observed at the Backgarden rural site in Guangzhou during July 2006 (Lu et al.,2012). The simulated ROconcentrations in this study are comparable to observed values at Wangdu (Tan et al., 2017)and much lower than those at Backgarden (Lu et al., 2012).

Using the NCAR-MM chemical box model, HONO concentrations are calculated with the daytime average value being 0.08 ppbv, only about 10% of the measured result(0.9 ppbv). The large underestimation proportion of HONO in this study is consistent with previously reported results(Su et al., 2011; Tong et al., 2016; Liu et al., 2021), due to only considering homogeneous reactions of NO + OH and a lack of large sources such as heterogeneous reactions of NOin the current chemical models. In order to estimate the impact of the missing HONO source on the radical chemistry at Xianghe, we calculate the ROconcentrations without the HONO constraint. The sensitivity simulation results show that the maximum concentrations of OH, HO,and ROdecrease to 7.2 × 10, 3.8 × 10, and 2.6 × 10molecules cm, respectively, as presented in Fig. 4. For daytime average (0600 to 1800 LST), the OH decreases by 41%. Decreasing OH leads to less active photochemistry and thus 38% and 37% decrease of HOand ROconcentrations, respectively, in agreement with previous results (Su et al., 2011; Liu et al., 2012; Li et al., 2014; Tong et al., 2016).Therefore, the measured HONO data should be included to properly simulate the ROchemistry.

The aerosol size distributions, a key parameter for the calculation of heterogeneous reaction rates, are not measured at Xianghe. Regardless of the uptake coefficients and aerosol particle size distributions, the heterogeneous reaction rates are ultimately determined by the calculated pseudo first-order rate coefficients. So, we adopt the pseudo firstorder rate coefficients from our previous study, which were calculated with the particle size distributions observed at the Xin’an rural site, about 60 km east of Xianghe (Ma et al.,2012). According to Ma et al. (2012), the calculated pseudo first-order rate coefficient of HOwas 0.025 s, which could represent an upper limit since the observed PMconcentrations in that field campaign were more than twice those at Xianghe. Therefore, we use the scaling factor of 0.5 with the corresponding value 0.012 sto represent the pseudo first-order rate coefficients of HOat Xianghe. The sensitivity simulations considering heterogeneous reactions of trace gases and radicals on aerosols show that the daytime average HOconcentration drops by 17%, OH drops by 11%, and ROdrops by 14%, which indicates that aerosol uptake of trace gases and radicals are not negligible and the measured aerosol size distributions are needed to accurately calculate heterogeneous reaction rates.

Fig. 4. Averaged diurnal variations of OH, HO2, and RO2 radical concentration modeled for the Xianghe site. Black lines represent the calculated concentrations with all measured gases as model constraint, and the vertical bars show the hourly standard deviation; Blue lines represent the sensitivity results without the HONO constraint; Red lines represent the sensitivity results considering heterogeneous reactions of gases and radicals on aerosols.

3.4. ROx budget analysis based on model results

Figure 5 shows the average diurnal variation of primary sources and termination sinks of RO. The maximum ROprimary production rate, P(RO), is 6.9 ppbv h. HONO photolysis is the predominant source of ROin the early morning and also a major source throughout the daytime. Ophotolysis followed by reaction with water vapor, together called the effective photolysis of O, and photolysis of HCHO and other carbonyls are the other important ROsources at midday. In contrast, alkene ozonolysis dominates the ROsource during nighttime. In the following, the daytime average (0600 to 1800 LST) will be the focus for the radical budget analysis. The photolysis of HONO and the effective photolysis of Ocontribute 41% and 20% to P(RO), with average production rates of 1.6 ppbv hand 0.8 ppbv h,respectively, which are also the predominant sources of OH radical, while other OH sources from photolysis of HOand HNOare negligible in this study. The photolysis of HCHO and other OVOCs contributes 26% to P(RO), with an average production rate of 1.0 ppbv h, among which HCHO photolysis produces HO(0.4 ppbv h) and other OVOCs photolysis produces HO(0.3 ppbv h) and RO(0.3 ppbv h). Alkene ozonolysis contributes 13%, with an average contribution of 0.5 ppbv h(0.2 ppbv hfor OH,0.2 ppbv hfor HO, and 0.1 ppbv hfor RO). Overall,the average daytime P(RO) is 3.9 ppbv h, and photolysis reactions dominate 87% of the ROsources.

Primary ROproduction depends on the intensity of solar radiation as well as NOand VOCs levels. Therefore,partitioning of primary ROsources is unique in different areas (Xue et al., 2016 and references therein). At Yufa,HCHO photolysis dominates the P(RO) due to high concentration of HCHO (on average 10−15 ppbv at daytime, Lu et al., 2013). In Beijing, Hong Kong, and Mexico City, OVOC photolysis (excluding HCHO) is the dominant source because of high OVOC levels (Volkamer et al., 2010; Liu et al., 2012; Xue et al., 2016). In this study, HONO photolysis dominates the P(RO) due to high HONO concentration, similar to New York city (Ren et al., 2003) and Wangdu (Tan et al., 2017). Sensitivity results without the HONO constraint show that the daytime average P(RO) decreases by 55% from 3.9 to 1.7 ppbv h(Fig. S3 in the ESM), highlighting the importance of measuring HONO for RObudget analysis. In addition to HONO, other radical precursors including HCHO and OVOCs should be measured in future studies since the contributions of their photolysis to the ROproduction are also important in some areas. Sensitivity results considering heterogeneous reactions of trace gases and radicals on aerosols show that there is close to no change in the strength and partition of P(RO) (Fig. S4 in the ESM).

Fig. 5. Averaged diurnal variations of primary sources (a) and termination sinks (b) of ROx radicals derived from model calculations. The colored lines show the speciated ROx primary sources and termination sinks. o-OVOCs represent the other OVOC species excluding HCHO. OH + others denotes loss rates by OH with other species (e.g.HNO3, HNO2, and HNO4).

Due to the short lifetime of ROradicals, the total ROloss rate, L(RO), is equal to the total ROprimary production rate, P(RO), as shown in Fig. 5b. At the Xianghe site,the radical losses are dominanted by their reactions with NOforming nitrogen containing compounds (L), including HNO, HONO, and PAN-type and RONOspecies, a typical feature of chemistry under high NOconcentrations. During daytime, the reaction of NO+ OH producing HNOis the largest contributor to radical termination (1.6 ppbv h,41%), followed by reactions of RO+ NOand RO+ NO,producing net PAN-type species (1.0 ppbv h, 26%) and RONOspecies (0.5 ppbv h, 13%), respectively. In comparison, the radical losses via HO+ HO/ROreactions forming peroxides (L) are much lower (0.3 ppbv h, 8%). Contribution of PAN-type species to radical loss is important in urban and suburban areas (Whalley et al., 2018; Tan et al.,2019). According to results found by Whalley et al. (2018),the formation of PAN-type species contributed 30% to the total radical loss in downtown London, similar to the result in this study. Taking into account heterogeneous reactions,aerosol uptake of HOcontributes 11% to L(RO) for the daytime average (Fig. S4).

The ratio of Lto L(RO) can be used to evaluate the Oproduction sensitivity, as suggested by Kleinman et al.(1997). The threshold of the L/L(RO) ratio is 0.5. When L/L(RO) is greater than 0.5, Ldominates the radical loss,indicating the Oproduction is limited by the VOC abundance. On the other hand, when L/L(RO) is less than 0.5,the radical loss is dominated by Land the Oproduction is controlled by NO. The radical budget analysis shows that Lcontributes about 90% to L(RO) at Xianghe, which indicates that the Oproduction is in the VOC-sensitive regime.

Figure 6 shows the conversion rates between OH, HO,and ROradicals. As shown in Fig. 6a, the conversion of OH to HOis mainly caused by reactions of CO and HCHO with OH. The reactions of speciated VOCs and OVOCs with OH lead to the ROperoxy radicals (Fig. 6b). The conversion of HOfrom ROis dominated by reactions of isoprene-derived radical (ISOP) and methyl peroxy radical with NO (Fig. 6c). The recycling of OH from HOis mainly generated by the reaction of HOwith NO (Fig. 6d). Evidently, the recycling processes between OH, HO, and ROare fast at the Xianghe site. For the daytime average, oxidation production rates of HOand ROby OH are 5.6 and 7.2 ppbv h, respectively. The reactions of RO+ NO and HO+NO further recycle HOand OH, with the corresponding production rates of 6.5 ppbv hand 12.5 ppbv h, respectively. In polluted areas under the coexistence of abundant VOCs and NO, efficient radical recycling can amplify the effect of the initially produced radicals (Ma et al., 2012;Xue et al., 2016).

Fig. 6. Averaged diurnal variations of radical conversion rates of OH→HO2 (a), OH→RO2 (b), RO2→HO2 (c), and HO2→OH (d). (a): OH + others denotes conversion rates by the reaction of OH with other species (e.g. H2, O3, and C6H6).

3.5. O3 photochemical production and loss

The photolysis of NOproduces Oand NO. NO titrates Oto regenerate NO. These reactions represent a null cycle for Oproduction in sunlight because there is no change in Oabundance at steady state. The removal of NO without consuming Oleads to net Oproduction. In the photochemical system, HOand ROradicals are the major NO consumers and Ois produced via the reactions of NO with HOand ROradicals. On the other hand, Ochemical loss occurs via Ophotolysis (production of OD, followed by reaction with water vapor) and reactions with OH, HO, and alkenes. The reaction of NOwith OH should also be considered as Ochemical loss because the conversion of NOto HNOterminates the catalytic cycle between NO and NO. As in previous studies (Liu et al., 1987; Lu et al.,2010; Xue et al., 2013), the total oxidant (O= O+ NO) is used to calculate instantaneous Ophotochemical formation rate F(O) and loss rate L(O), as denoted in Eqs. (2)and (3),

where [OH], [HO], [RO], [OD], [NO], [O], and[Alkene] are concentrations, and

K

represents the rate coefficient of the corresponding reaction. The net Ophotochemical production rate P(O) is the difference between the F(O) and L(O). Figure 7 shows the average diurnal variations of F(O) and L(O) calculated with the model simulated radical concentrations. The average diurnal maximum F(O) is 32.9 ppbv hat 1100 LST. The reaction of HO+NO is the dominant process resulting in Oproduction. For the daytime average (0600−1800 LST), the reaction of HO+NO contributes about 62% to the total Oproduction, which is consistent with previous results (Liu et al., 2012; Ma et al., 2012; Tan et al., 2019). The reaction of ISOP with NO accounts for about 11% of the total Oproduction due to relatively high isoprene and ISOP peroxy radical concentration at the Xianghe site, followed by the reaction of CHOwith NO (7%).

Using an observation-based model, Lu et al (2010) calculated the hourly F(O), varying from zero to 120 and 50 ppbv hat PKU and Yufa sites, respectively, during summer 2006 in Beijing. Liu et al (2012) simulated the average diurnal maximum F(O) of 60 ppbv hat PKU during summer 2007. Tan et al (2018) reported the average diurnal maximum F(O) was 17 ppbv hat Wangdu during summer 2014. The maximum F(O) at Xianghe lies between the simulated results at rural site Wangdu and urban site PKU. The ROradical concentrations at Xianghe are comparable to Wangdu and PKU as mentioned in section 3.3, while the corresponding noontime NO concentrations are 2.3 ppbv, 0.16 ppbv, and 5.0 ppbv at Xianghe, Wangdu, and PKU, respectively, resulting in the difference in the F(O) which depends on both ROand NO as shown in Eq. 2.

The average diurnal maximum L(O) is 4.3 ppbv h,appearing at 1100 LST. The reaction of NO+ OH is the main process leading to Ochemical loss, with a contribution of 63% toward the daytime average. The effective photolysis of Oand alkene ozonolysis contribute to 15% and 12%, respectively. By integrating the difference between the photochemical formation and loss rates over the whole day, the P(O) approaches 237 ppbv d, indicating that the photochemical production of Ois active at the suburban site in the NCP region.

The sensitivity results without the HONO constraint show that average diurnal maximum F(O) decreases from 32.9 to 22.6 ppbv h(Fig. S5 in the ESM). The daytime average F(O) decreases by 42%, consistent with the results reported by Liu et al (2012). Therefore, it is necessary to constrain the photochemical models with the measured HONO data. The sensitivity results considering heterogeneous reactions show that average diurnal maximum F(O) decreases to 28.4 ppbv h(Fig. S5 in the ESM). The daytime average F(O) decreases by 14%, about one third of the results without the HONO constraint, indicating that aerosol size distributions should be measured to accurately calculate heterogeneous reaction rates and Ophotochemical formation.

Fig. 7. Averaged diurnal variations of O3 production (a) and loss rate (b) derived from model calculations. The left panel represents the speciated O3 formation rate from different RO2 species. The right panel represents the speciated O3 destruction rate.

3.6. Chemical regimes of O3 production

The O-NO-VOCs sensitivity is investigated using the Empirical Kinetic Modelling Approach (EKMA). The base case refers to the averaged conditions over the whole campaign during the daytime. The averaged chemical and physical parameters are summarized in Table S2 in the ESM. By increasing or decreasing the concentrations of NOand AVOCs up to a factor of 3 with 30 equal-distance steps and fixing all other parameters as in the base case, the corresponding Oproduction rates are calculated and used to generate an isopleth plot illustrating the P(O) on the concentration of NOand AVOCs. The change of VOCs concentrations only applies to AVOCs to investigate the impact of anthropogenic emission changes on P(O).

Figure 8 shows the response of P(O) to precursors of NOand VOCs at the Xianghe site. The Ochemical regimes between VOC-sensitive and NO-sensitive are separated by the black line, also called the ridge line. The ridge line connects the turning points of the P(O) isopleths and represents the optimum initial NOand VOC concentrations for Oproduction. Above the ridge line, a reduction in initial NOresults in an increase in the P(O), whereas a reduction in initial VOC results in a decrease in the P(O), and this condition is considered VOC-sensitive. Below the ridge line, a reduction in initial VOC has only a very small effect on the P(O), whereas P(O) is reduced by reduction in the initial NO, and this condition is considered NO-sensitive.The black filled square in the isopleth diagrams represents the daytime averaged NOand AVOCs concentrations for the time period of 1−23 July, and blue circles stand for the daytime averaged NOand AVOCs concentrations for each day. The filled square and most of the circles are above the ridge line, indicating that the Oproduction is mainly controlled by VOCs. This conclusion is consistent with the result from radical loss analysis in section 3.4. Compared to Xianghe, Oformation at Wangdu was NO-sensitive due to lower NOlevels during daytime (Tan et al., 2018). This radical loss process was further verified by the fact that the radical loss was dominated by HO+ HO/ROreactions at Wangdu, representing a NO-sensitive regime (Tan et al.,2017). When the NOconcentration is larger than 10 ppbv,the calculated P(O) decreases with increasing NO, demonstrating that Oproduction enters the NOtitration regime.

The impact of HONO and aerosol uptake of trace gases and radicals on P(O) chemical regimes is also investigated as part of this study. As shown in Fig. 8, without the HONO constraint, the ridge moves down and all circles are above the ridge line, which means that the Oformation is more VOC-sensitive. The aerosol uptake of trace gases and radicals makes the ridge move up as shown in Fig. 8, whereas the filled square and most of the circles are still above the ridge line, which means that the Oproduction is unchanged but has a tendency shifting toward NO-sensitive. Therefore, VOCs control should be the optimal way to mitigate the Opollution at Xianghe.

Fig. 8. Isopleth diagrams of the net O3 production rate [P(O3), units: ppb h−1] as functions of the NOx and AVOCs derived from an empirical kinetic modeling approach. The black filled squares represent the daytime averaged NOx and AVOCs concentrations measured for the time period of 1−23 July. The blue circles represent daytime averaged AVOCs and NOx concentrations measured for each day. The left panel represents the EKMA diagram with all measured gases as model constraints; the middle panel represents the EKMA diagram without the HONO constraint; the right panel represents the EKMA diagram considering heterogeneous reactions of gases and radicals on aerosols.

4. Conclusions

We present the measurements of O, NO, CO, VOCs,and HONO and major physical parameters at the Xianghe suburban site located in the NCP region during summer 2018.Using a chemical box model constrained by these measurement data, we investigate the radical budget, Ophotochemical production and loss, and the O-VOC-NOsensitivity based on the EKMA approach. We further discuss the sensitivity of HONO and heterogeneous reactions of radicals on the radical budget and Ophotochemical production.

Results show that mean concentrations plus/minus standard deviations of O, O, NO, NO, CO, AVOC, Isoprene,and HONO are 46.1 ± 28.8 ppbv, 57.2 ± 26.5 ppbv, 1.9 ±1.8 ppbv, 13.0 ± 6.0 ppbv, 694.6 ± 478.1 ppbv, 20.1 ± 11.0ppbv, 0.7 ± 0.9 ppbv, and 1.0 ± 0.5 ppbv, respectively. Additionally, the OH reactivity values vary between 12 sand 18 s, with average Kof 15.0 s.

The average diurnal maximum ROprimary production rate, P(RO), is 6.9 ppbv h. For the daytime average,the total ROprimary production rate is 3.9 ppbv h. Photolysis reactions dominate 87% (41% for HONO photolysis,20% for Ophotolysis, and 26% for OVOC photolysis) of the primary radical sources, and alkene ozonolysis contributes 13%. The radical termination is dominated by the reaction of OH with NO, representing the VOC-sensitive regime at Xianghe site.

The average diurnal maximum chemical production of Ois 32.9 ppbv h, and the maximum chemical loss of Ois 4.3 ppbv h. The reaction of HOwith NO is the main process for Oproduction, contributing about 62% to the total Oproduction for the daytime average. Oloss is dominated by reaction of NOwith OH, with a contribution of 63%.

Sensitivity tests without the HONO constraint point to decreases in the daytime average primary ROproduction rate by 55%, and Ophotochemical production by 42%, highlighting the importance of measuring HONO for quantifying RObudget and Ophotochemical production. Taking into account aerosol uptake of trace gases and radicals, aerosol uptake of HOcontributes 11% to ROsink, and the daytime average F(O) decreases by 14%, indicating that aerosol size distributions should be measured to accurately calculate heterogeneous reaction rates and Ophotochemical formation. EKMA contour suggests that the Oproduction in suburban Beijing lies in the VOC-sensitive regime, where reduction of VOCs could lead to reduced Oproduction. Sensitivity test results without the HONO constraint over-emphasize the VOC-sensitive effect on the Oproduction. Considering aerosol uptake of trace gases and radicals, the Oproduction regime remains unchanged but has a tendency shifting toward being NO-sensitive.

Acknowledgements

. This work was supported by grants from the National Key Research and Development Program of China (Grant No. 2017YFC0210003). The authors are grateful to all staff and workers from the Xianghe Atmospheric Observatory of Institute of Atmospheric Physics (IAP) of the Chinese Academy of Sciences for their support during the sampling campaign.

Electronic supplementary material:

Supplementary material is available in the online version of this article at https://doi.org/10.1007/s00376-021-0327-4.