Measurement Report: Aircraft Observations of Ozone, Nitrogen Oxides, and Volatile Organic Compounds over Hebei Province, China

To provide insight into the planetary boundary layer (PBL) production of ozone (O3) over the North China Plain, 15 the Air chemistry Research in Asia (ARIAs) campaign conducted aircraft measurements of air pollutants over Hebei Province, China between May and June 2016. We evaluate vertical profiles of trace gas species including O3, nitrogen oxides (NOx), carbon monoxide (CO), and volatile organic compounds (VOCs) and relate to rates of O3 production. This analysis shows measured O3 levels ranged from 52 to 142 ppbv, with the peak median concentration (~94 ppbv) occurring between 1000 and 1500 m. The NOx concentrations exhibited strong spatial and altitudinal variations, ranging from 0.15 to 49 ppbv. 20 Ratios of CO/NOy and CO/CO2 indicate the prevalence of low efficiency combustion from biomass burning and residential coal burning. Concentrations of total measured VOCs from 26 whole air canisters reveals alkanes dominate the total measured volume mixing ratio of VOCs (68%) and we see evidence of vehicular emissions, fuel and solvent evaporation, and biomass burning sources. Alkanes and alkenes/alkynes are responsible for 74% of the total VOC reactivity assessed by calculating the OH loss rates, while aromatics contribute the most to the total Ozone Formation Potential (OFP) (43%) with 25 toluene, m/p-xylene, ethylene, propylene, and i-pentane playing significant roles in the aloft production of O3 in this region. In the PBL below 500 m, box model calculations constrained by measured precursors indicate the peak rate of mean O3 production was ~7 ppbv/hour. Pollution frequently extended above the PBL into the lower free troposphere around 3000 m, where NO2 mixing ratios (~400 pptv) led to net production rates of O3 up to ~3 ppbv/hour; this pollution can travel substantial distances downwind. The O3 sensitivity regime is determined to be NOx-limited throughout the PBL, while more 30 VOC-limited at low altitudes near urban centres, demonstrating both VOCs and NOx need further control to reduce aloft O3 over Hebei.

and to photosynthetic processes by vegetation (Avnery et al., 2011;Reich and Amundson, 1985), while 40 some VOCs, such as benzene and chloroform, are known to be hemotoxic and carcinogenic (Environmental Protection Agency -Integrated Risk Information System, 2003;Lan et al., 2004). Several studies using the NASA Ozone Monitoring Instrument (OMI) have found reductions of some pollutants like sulfur dioxide (SO2) over the NCP (He et al., 2012;Krotkov et al., 2016;Li et al., 2010, but NO2 pollution still remains severe in China (Figure 1a).
Ozone is created through the oxidation of NO by hydroperoxyl radicals (HO2) and organic peroxy radicals (RO2), products of 45 carbon monoxide (CO) and VOC oxidation. When one of these precursors is the limiting reactant, the rate of O3 production is considered VOC-or NOx-sensitive (Finlayson-Pitts and Pitts, 1999;Sillman et al., 1990). The role of VOCs on the formation of O3 depends on the characteristics of the environment, including the main emission sources of primary pollutants and ambient temperature (Pusede et al., 2014). Quantifying the abundance of NOx and the suite of VOC chemicals throughout the lower troposphere is urgently needed to better understand the photochemistry of O3 production in the NCP to develop successful 50 mitigation strategies.
In-situ airborne measurements provide valuable information regarding the horizontal and vertical distributions of air pollutants over a large spatial area. Airborne measurements are necessary to characterize air pollution over large cities, as well as surrounding areas. Ozone and PM are produced throughout the planetary boundary layer (PBL), so aircraft observations can lead to a more complete picture of pollution formation and transport than is available only from surface observations. While 55 several airborne campaigns have deployed to investigate the regionally transported pollution problem in East Asia, including the NASA Korea-United States Air Quality Study (KORUS-AQ) (Al-Saadi et al., (2015), see https://wwwair.larc.nasa.gov/missions/korus-aq/docs/White_paper_NASA_KORUS-AQ.pdf), that occurred at the same time as our measurements, few airborne studies characterize the source region of severe smog within the Hebei Province region of China.
Through Chinese/American partnerships with Peking University, Beijing Normal University, and the University of Maryland, 60 humidity, weak winds, and shallow PBL height (Tang et al., 2012). The Y-12 aircraft was based at Luancheng Airport (114.59°E, 37.91°N, 58 m above sea level (ASL)), located in southeast Shijiazhuang (population around 10 million), a major economic centre in Hebei, including pharmaceutical and textile industries, machinery and chemical manufacturing, construction, and electronics production. Flight sampling occurred east of the Taihang Mountains and the Y-12 flew vertical 75 spirals from ~300 m to ~3500 m over Shijiazhuang as well as three other locations: Julu (115.02°E, 37.22°N, 30 m ASL), Quzhou (114.96°E,36.76°N,40 m ASL), and Xingtai (114.36°E, 37.18°N, 182 m ASL) (see Table S1 for a description of flight paths, weather conditions, and statistics of measured trace gases).
Various instruments aboard the Y-12 aircraft collected trace gas, aerosol, and meteorological data. The aircraft instrumentation (Table 1)  Power constraints and a converter issue led to limited NOy, NOx, and CO measurements during the campaign, particularly in the lower free troposphere (LFT). Negative values indicate readings around the detection limit, usually at high altitudes. The aircraft was also equipped with an inlet to measure aerosols up to ~5.0 µm diameter and aerosol optical properties, including a nephelometer (TSI Model 3563) to measure aerosol scattering, a particle soot absorption photometer (PSAP) to measure aerosol absorption, and an aethalometer (Magee Model AE31) and a Single-Particle Soot Photometer (SP2, Droplet 90 Measurement Technologies) to measure black carbon. Observed aerosol optical properties have been summarized by F. Wang et al. (2018); further details on aircraft instrumentation are given by Ren et al. (2018).
Twenty-six whole air samples (WAS) were collected directly into 3.2 L fused silica lined electropolished stainless steel canisters (Entech Instrument Inc., Simi Valley, CA) at a variety of pressure altitudes from 400 m to 3500 m between 1:30 and 9:00 UTC (9:30 and 17:00 local time). The sampling period for the WAS canisters was approximately 1-2 minutes during the 95 spirals. Samples were analysed for 54 VOCs and 22 halocarbons. Since the halogenated species have negligible effects on O3 production, we exclude these species from the analysis presented here. We also exclude 2 WAS canisters from this analysis due to evidence of contamination after collection (Text S1). Limited samples collected over one province in one season may not be able to represent O3 chemistry for all of China, but the scarcity of airborne VOC measurements in this region makes these data valuable for characterizing the composition of air throughout and above the PBL, demonstrating how the production 100 of O 3 aloft differs from that at the surface.
The VOC analytical techniques used by the College of Environmental Sciences and Engineering at Peking University (PKU) in Beijing have been summarized in the past (Mo et al., 2015;Wang et al., 2010a), and we briefly describe the method here.
The WAS canisters were cleaned following a standard sampling procedure, pressured with nitrogen and vacuumed three times to 2.6 Pa. The hydrocarbons were quantified using a gas chromatograph equipped with a mass selective detector (GC-MSD, concentrations of VOCs, NO2, CO, and O3. Periodic missing Y-12 NO2 data due to internal auto-zeroing are linearly interpolated since gaps were short (~2 minutes). A physical loss lifetime is set to 24 hours to mitigate build-up of long-lived 140 oxidation products over multiple days of integration. Ground observations are only used for days that a flight occurred. For May 17, surface data for NO2 is filled with 1-hour average data collected for other days of the month, due to missing surface measurements on this day. For every 5-minute interval of ground data, the model is run with a 1-hour time step for 3 days with changing solar zenith angle calculated for the location of ground observations. For each 1-minute interval of flight data, the model is run with a 1-hour time step for 5 days in "solar cycle" mode, allowing photolysis frequencies to evolve over the 145 course of a model step and for calculated reactive intermediates to achieve diel steady state. Photolysis frequencies (not measured during ARIAs) are calculated by combining cross sections and quantum yields with solar spectra derived from the NCAR Tropospheric Ultraviolet and Visible (TUV) version 5.2 radiation model. At the start of the model run, input solar zenith angle, altitude, O3 column, and surface albedo are used for linear interpolation across TUV lookup tables (F0AM's "hybrid" method). A correction factor is used to scale j-values to better agree with the observed NO/NO2 ratio. Optical depth, 150 single scattering albedo, and angstrom exponent during ARIAs (see Wang et al., (2018a)) are used in the TUV online calculator (https://cprm.acom.ucar.edu/Models/TUV/Interactive_TUV/) to assess the impact of aerosols on photolysis frequencies. The impact of aerosol optical properties measured during ARIAs on photolysis frequencies is small compared to the default setting, so no additional adjustments are made to the model.
Due to the limited number of grab canisters per flight, VOCs are constrained based on the altitude of the sampling relative to 155 the height of the PBL, which is determined using potential temperature and water vapor vertical profiles for each flight. All WAS canisters collected below the top of the PBL during a flight are averaged. All of the WAS canisters for the entire campaign collected above the research flight's PBL are averaged for that flight. The average concentration of the samples below 500 m is used as ground concentrations since A 2 BC did not measure VOCs at the surface. This method to constrain VOCs introduces large uncertainty due to the sparsity of measurements obtained over a large area that potentially consists of 160 a wide variety of chemical compositions. However, the production of O3 aloft is not well characterized over Hebei, so our observations may help improve the understanding of air pollution for this region, despite this limitation.
Unlike a 3-dimensional chemical transport model, the box model simulations do not include advection or emissions. These processes, while important, are not included in the box model since O3 precursors were measured and used to constrain the model calculations. Box modelling is used to gain an understanding of O3 production and its sensitivity to ambient levels of 165 NOx and VOCs based upon measured meteorological parameters and the concentration of a wide variety of chemical species.

Ozone Production and Sensitivity Calculations
The photochemical production of O3 during the daytime is determined by the production rate of NO2 molecules from the HO2+NO and RO2+NO reactions minus the loss mechanisms (Finlayson-Pitts and Pitts, 1999). Thus, the net O3 production rate, net(PO3) can be estimated following Equation 1: where k denotes the different reaction rate constants and RO2i is the concentration of individual organic perxoy radicals. The terms subtracted from the production of O3 are the loss mechanisms: the formation of nitrates, P(RONO2), the reaction of OH and NO2 to form nitric acid, the reactions of OH and HO2 with O3, the reaction of O( 1 D) with H2O, and the reactions of O3 175 with alkenes. Additional terms not included here are the rate of O3 loss by dry deposition and direct loss on aerosol surfaces (dilution is the only physical loss in the current F0AM setup).
We evaluate the sensitivity of O3 production to NOx and VOCs using the ratio of LN/Q, where LN is the radical loss through reactions with NO and Q is the primary radical production (Kleinman, 2005a). When LN/Q is much less than 0.5, the O3 production regime is NOx-limited; when LN/Q ratio is much higher than 0.5, the regime is VOC-limited. Different 180 environments can have varying amounts of organic nitrates that impact the cut-off value of LN/Q, so this value could vary around 0.5 (Kleinman, 2005b).

Observations of nitrogen oxides, carbon monoxide, and ozone
Our observations confirm heavy loadings of air pollution over Hebei. Vertical profiles show peak median concentrations of 185 NO (1.6 ppbv), NO2 (4.4 ppbv) and NOy (25.7 ppbv) below 500 m with large variability (Fig. 2). Median concentrations of NO and NO2 drop off gradually with altitude, while median NOy remains close to ~15 ppbv throughout most of the profile.
Between 500 and 1000 m, sufficient levels of NOx are observed (median=3.8 ppbv), indicating continued production of O3 in the PBL. Above 3000 m, median concentrations of NO and NO2 fall to 350 pptv and 106 pptv, respectively (not measured simultaneously), still sufficient to produce O3 as air parcels travels downwind. Median mixing ratios of O3 and CO remain 190 high (~80 ppbv and ~120 ppbv, respectively) throughout the altitudes sampled by the Y-12.
Unlike previous airborne studies over Beijing from 1994Beijing from -2005 that found increased O3 concentrations below 1 km with constant levels (~52 ppbv) between 1 and 2 km (Ding et al., 2008), our O3 concentrations peaked between 1000 and 1500 m (median = 91.6 ppbv). Low ratios of NOx/NOy (<0.30) indicate significant O3 production had already occurred, but the strong correlation (R=0.77, Fig. S2) between 1-minute NOz (NOy-NOx) and Ox (O3+NO2), an empirical estimate of the O3 production 195 efficiency (OPE), below 1500 m demonstrates moderate production of O3 continued during sampling. The OPE of ~4 during ARIAs is smaller than the average OPE value of ~8 obtained during 2013 DISCOVER-AQ flights in Houston (Mazzuca et al., 2016), likely due to the higher NOx concentrations observed in Hebei than Texas.  (Table 2). Comparable to our range of NOx levels from 0.15 to 49.3 ppbv, autumn flights in the Yangtze River 225 Delta in 2007 documented large variability in NOx concentrations, ranging from 3 to 40 ppbv , while April 2006 observations in northern China similarly find a mean concentration ~5 ppbv . The minimum CO concentration during ARIAs (90.6 ppbv) was measured in the lower free troposphere, which is a much smaller minimum concentration than reported by earlier studies. The warm-sector PBL air ahead of a cold front in April 2007 in Shenyang Province in northeast China found ~300 ppbv CO between 1000 and 4000 m (Dickerson et al., 2007), generally larger than 230 most ARIAs profiles (except for Julu). The maximum value of CO during ARIAs (almost 2 ppmv) agrees better with the literature, although there are few reported aircraft measurements of CO in Northeast Asia. Average and maximum O3 concentrations during ARIAs were much higher than in other studies, but comparable to KORUS-AQ measurements from May 24-29 when the flow of air was direct from China. Since the majority of past airborne studies occurred over the sea areas during other seasons, it is not surprising that an urbanized environment like Hebei experienced much larger amounts of O3 235 than previously reported.
The ratios between combustion tracers can be used to understand the source and efficiency. During high-efficiency combustion in modern power plants, fuel carbon is converted to CO2 with near unit efficiency, resulting in low CO/CO2 (<0.10%), while low-efficiency combustion (cold or smoldering processes or low-technology combustion) yields larger ratios. The regression of 1-minute CO against CO2 (Fig. 5a) shows high linear correlation (R=0.90) and high ratios of CO/CO2 (2.4%) together with 240 large amounts of SO2. These measurements are indicative of low-efficiency fossil fuel combustion, likely from residential coal burning as these observations were all collected at ~500 m. These results agree with airborne data from the KORUS-AQ campaign during Chinese-sourced inflow indicating 1-4% CO/CO2 (attributing to low-efficiency combustion) (Halliday et al., 2019). Similarly, Xia et al. (2020) found higher CO/CO2 ratios ~2% in December 2017 at Jingdezhen station in central China when air mass transport was from northern China, although they note some high combustion efficiency may be found in Jiangxi 245 and Hunan provinces in Central China. Higher CO/CO2 ratios (~10%) with less SO2, as occasionally seen during ARIAs, are more in line with emissions from biomass burning (Andreae and Merlet, 2001;Wang et al., 2010b).
The ∆CO/∆NOy ratio (equivalent to the slope in a CO vs. NOy plot) (Fig. 5b) is an indicator for distinguishing plumes with efficient O3 formation, Typical values of this ratio are ~40 in background air and between ~4-7 in fresh emissions plumes in Houston (Neuman et al., 2009). The ∆CO/∆NOy ratio of 14.85 measured during ARIAs indicates some photochemical aging 250 and contributions from fossil fuel or biomass burning, but high values of CO, NOy, and SO2 suggests sampling of air parcels heavily influenced by power plants. The CO/NOx emission ratio (Fig. 5c) from ARIAs agrees with higher emission ratios of gasoline vehicles, while higher amounts of CO, NOx, and SO2 indicate coal burning from the residential sector or inefficient electric generating units. While most of these observations are reflective of the prevalence of low efficiency fossil fuel combustion, the aircraft sampled a plume on June 6 while flying spirals over Julu containing 0.9% CO/CO2 and 0.4% SO2/CO2 255 (Fig. S3), likely due to a coal-burning power plant operating at high combustion efficiency, either using a sulfur scrubber or burning low sulfur fuel.

Observations and sources of VOCs
The total measured VOC mixing ratios ranged from 4 to 23 ppbv, largely dependent upon the altitude of collection, and was mostly dominated by alkanes (Fig. 6). The samples associated with the largest concentrations of O3 were all collected at 260 altitudes ~500 m during a period with stagnant high pressure. Generally, the samples collected below 500 m showed larger amounts of alkenes/alkynes and aromatics than canisters collected elsewhere in the PBL. The top VOCs ranked by mean volume mixing ratio (Table 3) shows that alkanes dominate the total measured VOC mixing ratio during ARIAs (68%), followed by alkenes/alkynes (17%), and aromatics (15%). The top 10 VOC species are C2-C5 alkanes, C2-C3 alkenes/alkynes, benzene, and toluene. The observed mixing ratios of ethane and propane are 2.65 ppbv and 1.39 ppbv, respectively, which 265 together accounts for ~52% of the total alkane mixing ratio.
The levels of ambient VOCs during ARIAs are generally lower than prior surface observations since measurements were taken in the PBL away from primary sources. Prior ground-based studies have similarly found alkanes to contribute the majority (>50%) of the total VOC concentration in late spring in the Beijing-Tianjin-Hebei region Tang et al., 2009;Yuan et al., 2013). The most abundant species during ARIAs are comparable to previous studies finding ethane, propane, and 270 acetylene among the most prevalent, but likely have different sources based on the study location (Jia et al., 2016;Li et al., 2015;Mo et al., 2015;Tang et al., 2009). In the Beijing-Tianjin-Hebei region, ambient acetylene, ethylene, and other light alkanes have been attributed to emissions from gasoline vehicles , while in Guangzhou, the widespread use of LPG has resulted in high levels of propane (Guo et al., 2011). Additionally, our observations have higher amounts of branched alkanes, such as 2,2,4-trimethylpentane and 2-methylheptane, both components of gasoline. Next, we examine the potential 275 sources contributing to observations of VOCs by comparing with ratios and correlations from known sources.
Ethane is the most abundant VOC in this study and correlates well with indicators for biomass and coal burning (R>0.81), such as acetylene, ethylene, benzene, and SO2. The ratio of acetylene to ethane (Fig. 7a) during ARIAs is 0.59, comparable to the ratio found in a plume of fresh biomass burning in Canada (Blake et al., 1994) and within the range of crop residue burning (~0.2-0.6) found in other studies in China . High ratios of benzene/propane (1.12) are comparable to dry 280 grass combustion samples collected in the central Pearl River Delta (PRD) (1.6)  and further confirm the presence of VOCs due to biomass burning.
The C3 and C4 alkanes, including propane and the butanes, are the three main components of LPG and their correlation acts as an indicator for LPG leakage. In this study, a moderate correlation (R~0.50) is found between n-butane and propane and ibutane with n-butane. The ratio of n-butane/propane during ARIAs is 0.60, which agrees well with ratios from vehicle 285 emissions (Liu et al., 2008a), but is lower than slopes measured in the PRD (2.1) (Lai et al., 2009), where VOCs originated from LPG leakage. Additionally, propane correlates well with acetylene and ethylene (Figure 7a), two well-known vehicular emission tracers.
Since acetylene and propane have comparable photochemical lifetimes with respect to OH attack, the ratio can be used to assess the relative importance of fossil fuel combustion and LPG leakage (Goldan et al., 2000). LPG contains propane but not 290 acetylene (acetylene/propane<1) while combustion of fossil fuels commonly produces small amounts of propane relative to acetylene (acetylene/propane>1) (Conner et al., 1995;Gilman et al., 2013;Russo et al., 2010;Watson et al., 2001). In this study, the acetylene/propane ratio (Fig. 7a) is greater than 1, indicating emissions from vehicles (Fraser et al., 1998). These results suggest vehicles are largely responsible for the C3 and C4 alkanes as well as the C2 alkenes/alkynes observations. The C5 alkanes and some C6 alkanes like 2,3-dimethylbutane and 2-methylpentane are found in vehicular exhaust and in 295 gasoline vapor . The i-pentane to n-pentane ratio is commonly used to identify the contributions of natural gas, vehicular emissions, and fuel evaporation since these alkanes have similar boiling points, vapor pressures, and reaction rate coefficients with OH. In areas heavily dominated by natural gas drilling, ratios lie between 0.82-0.89 (Gilman et al., 2013), while higher ratios are associated with vehicle emissions (2.2-3.8) and fuel evaporation (1.8-4.6) (Jobson et al., 2004;McGaughey et al., 2004;Russo et al., 2010;Wang et al., 2013). In this study, i-pentane and n-pentane are highly correlated 300 (R=0.93), indicating a common source of these compounds. The slope is 10.3, higher than reported in previous studies in China , and the large i-pentane concentrations are likely reflective of gasoline evaporation due to the extremely volatile nature of i-pentane. The influence of fuel evaporative emissions is further identified by strong correlations between C4-C7 alkanes and alkenes typical of fuel evaporative emissions. Strong correlations of many long-chain alkanes (C6-C7 and octane) with i-pentane (R>0.73 except for cyclohexane) but absence of correlations with acetylene indicates solvent 305 evaporation may be another source of long-chain alkanes.
Typically, the ratio of cis-2-butene/trans-2-butene is used to determine the source of C4 alkenes Velasco et al., 2007). However, in this study, all measurements of cis-2-butene and trans-2-butene are below the detection limit, so assessing the ratio and correlation is not possible. Previous studies in this region in China have attributed C4 alkenes to vehicular emissions . 310 The correlation between the C7-C8 aromatics is strong (R>0.76) and revealing of typical signatures from incomplete combustion. The toluene/ethylbenzene ratio (10.7) is higher than traffic and urban emission ratios (~5-8), but closer to ratios associated with biomass burning (9.41) (Monod et al., 2001;Parrish et al., 1998). Toluene also correlates with all C7-C9 alkanes (R>0.64) and with i-pentane (R=0.85), compounds from diesel and gasoline evaporation. High levels of toluene There is an excellent correlation (R>0.99) between o-xylene and m/p-xylene (Fig. 7b) and the slope (0.33) is comparable to the emission ratio found in a tunnel study (0.35) (Liu et al., 2008a). The o-xylene/ethylbenzene (0.60, Fig. 7b) slope is lower than vehicle exhaust emission ratios (1.2-1.8) (Conner et al., 1995;Jobson et al., 2004;Kirchstetter et al., 1996;Rogak et al., 1998;Sagebiel et al., 1996), but the correlation is extremely strong, suggesting the preferential loss of xylenes during transport 320 due to their higher reactivity. These correlations and ratios suggest incomplete combustion from vehicular emissions and biomass burning are an important source of C7 and C8 aromatics.
The ratio between benzene/toluene (B/T) is a useful indicator to distinguish between vehicular emissions and other combustion sources. A ratio ~0.5 is often attributed to vehicular sources (Brocco et al., 1997;Perry and Gee, 1995), while ratios larger than 1 have been reported for coal or charcoal burning (Andreae and Merlet, 2001;Moreira Dos Santos et al., 2004). Benzene 325 was observed at high mean ratios over Hebei (0.51 ppbv) and the average B/T ratio is 1.8±1.6 ppbv/ppbv. The correlation of some hydrocarbons can highlight the differences between B/T>1 (N=17) and B/T<1 (N=9). The correlation found between benzene and acetylene when all samples are grouped together (Fig. 7a) substantially improves just considering "traffic-related" samples (B/T<1) (R=0.93), suggesting a contribution of vehicular sources to benzene and acetylene measurements.

The effect of VOCs on ozone formation 330
In order to effectively reduce O3 concentrations, it is crucial to understand the relative importance of individual VOCs in terms of the production of O3 because each VOC exhibits different chemical reactivities. In this section, we present results using the loss rate of each VOC species with OH and ozone formation potential (OFP).

OH loss rate of VOC species
The calculation of the first-order loss rate of OH with different VOCs, termed OH reactivity, provides a measure of the potential 335 to produce HO2 and RO2, key intermediate species in the production of O3 (Stroud et al., 2008). Since the reaction with OH accounts for the majority of loss of most VOCs, the rate constant (obtained from the Master Chemical Mechanism version 3.3.1 (MCM3.3.1) and the National Institute of Standards and Technology (NIST) Chemical Kinetics database (www.kinetics.nist.gov/)) for the reaction between OH and various hydrocarbons reflects the overall reactivity of that hydrocarbon (Finlayson-Pitts and Pitts, 1999). OH reactivity for each VOC species (VOCi) is defined by Equation 2: 340 Where ,+.P,Q 8 is the reaction rate constant between OH and VOCi. Among the VOC groups, alkanes and alkenes/alkynes both contribute the most to the total VOC reactivity, accounting for 37% each. Aromatics accounted for 26% of the total VOC reactivity. The relative contribution of the top 10 VOCs ranked by mean OH reactivity (Table 4) shows ethylene, propylene, and isoprene among the top measured alkene species, together contributing ~33% to total OH reactivity. Among the alkanes, 345 2-methylpentane and i-pentane contribute the most (13%) to total OH reactivity, followed by the branched pentanes and propane. Aromatic compounds such as toluene and m/p-xylene constitute 13% to total OH reactivity. Previous ground-based summer studies in China have found larger contributions of isoprene to OH reactivity, ranging from ~10-30% Xue et al., 2017). Since isoprene is mostly emitted by biogenic sources during the warmer summer months and when soil moisture is sufficient for plant growth, we expect isoprene to have a larger impact on O3 production in the summer than our 350 study in spring.

Ozone formation potential of NMHCs
Since OH reactivity only provides a qualitative identification of the most reactive species and does not reflect products and their production of further free radicals, we next consider the contribution to the formation of O3 using ozone formation potential (OFP). The OFP of a VOC relies on the quantity maximum incremental reactivity (MIR), which represents the amount 355 of O3 formed from the addition of a small amount of the VOC species in interest under high NOx conditions. Values of MIR (unit: g O3 formed/ g VOC) have been calculated based on model simulations evaluated with smog chamber measurements (Carter, 2010(Carter, , 1994. The OFP is calculated according to Equation 3: This method gives an estimate of only the first 24 hours after initial release. The median measured VOC/NOx ratio for all WAS 360 canisters was 4.9 ppbv/ppbv. In comparison, the ratio of reactive organic gas to NOx (ROG/NOx) in Los Angeles is 7.6 ppbv/ppbv (Carter, 1994). VOCs experience photochemical loss from emission sources near the surface to measured aloft concentrations. Estimation of OFP from aircraft observations throughout the PBL indicates how formation of O3 may be different from previous surface studies.
To identify the major contributors to O3 formation in this region, the 10 species with the highest mean OFP are listed in Table  365 4. Aromatic compounds are the largest contributor to total OFP (43%), followed by alkanes (30%) and alkenes/alkynes (27%).
Toluene and ethylene make the largest contributions (19.6% and 15.7%, respectively) to total OFP. The high MIR of these compounds (MIR=4.0 g O3/g VOC and 9.00 g O3/g VOC, respectively) and large mixing ratios (4.9% and 5.7% of the total measured VOC volume mixing ratio) drives their important contribution to O3 formation. The relatively short lifetime of ethylene (~1.4 days) combined with the large range of measured mixing ratios (0.18 to 3.54 ppbv) suggests sampling of air 370 masses with little to moderate photochemical processing, indicating the large range of influence on OFP. The most reactive compound in terms of OFP is trans-2-butene (MIR=15.16 g O3/g VOC), but its low concentration results in only 0.2% to total OFP. At the other extreme, ethane accounts for a relatively high percentage of total measured VOC volume mixing ratio (17.0%) yet only contributes 2.1% to OFP due to its low reactivity (MIR=0.49 g O3/g VOC).
Previous studies in China report aromatics and alkenes account for the most OFP (Cai et al., 2010;Cheng et al., 2010;Jia et 375 al., 2016;Liang et al., 2017;Wang et al., 2010aWang et al., , 2016Xie et al., 2008;Zheng et al., 2009). In particular, ethylene, isoprene, toluene, propylene, and m/p-xylene are most influential to OFP at a surface site in Quzhou in June and July 2014 . J. H. Tang et al. (2007) concluded ethylene, toluene, and m/p-xylene are the main contributors to OFP during spring 2005 at the surface in the PRD, citing emissions from industry and vehicular exhaust. Our study agrees with past research in urban areas in China identifying the most reactive VOCs in terms of OFP; O3 appears to be formed more slowly above the 380 surface and in nonurban areas, but production is still substantial.
The current VOC abatement policy in China mainly focuses on the reduction of anthropogenic VOCs from sources in the petrochemical industry, organic chemical industry, packaging printing, and industrial coating, not considering reactivity or chemical speciation (Li et al., 2018). A 2010 VOC emission inventory study concluded the top 15 OFP species (including m/pxylene, toluene, propylene, o-xylene, and ethylbenzene) contributed 69% of total OFP, but only accounted for 30% of the total 385 emission of VOCs by mass (Liang et al., 2017). Our analysis of the top 10 species ranked by mean OFP shows these compounds contribute 68% to total OFP but only represent 37% of the total volume mixing ratio. Li et al., (2018) classifies industrial coal burning, biomass burning, and motorcycles to the top three VOC emission sources in Shijiazhuang, but OFP is highest for furniture coating, automobile coating, diesel vehicles, fuel evaporation, and gasoline vehicles. These results confirm that reactivity scales and emissions rates should be considered together when formulating control strategies for O3.

Photochemical ozone production rate and sensitivity
In this section, we describe calculated net photochemical production rates of O3 using the box model constrained by aircraft observations. Ozone production rates calculated from the box model are high in major urban centres, particularly Shijiazhuang and Xingtai, but also between these cities (Fig. 8a). The highest rates (>10 ppbv/hour) are generally found closer to the surface, but in some instances upwards of 2000 m. The largest net production rate of O3 (over 16 ppbv/hour) was located along the 395 Taihang Mountains between Shijiazhuang and Xingtai. This large net production rate occurred ~2000 m on June 11, 2016 when NO, NOy, NO2, and O3 were ~2 ppbv, ~18 ppbv, ~3 ppbv, and ~75 ppbv, respectively. Vertical profiles of production, loss, and net rates of O 3 (Fig. 9) show that HO 2 +NO made more O 3 than RO 2 +NO during the campaign. The major loss of O 3 was due to the termination of NO 2 through its reaction with OH below 2500 m. Reaction with O( 1 D) is the main loss of O3 above 2500 m. A maximum of net O 3 production for the mean profile was observed in the lowest 400 500 m of ~7 ppbv/hour. In the PBL between 1500-2000 m, where median NO and NO2 were 534 and 625 pptv, respectively, O3 production rates were ~4 ppbv/hour. In the lower FT from 2500 to 3000 m, peak net O 3 production rates still reached ~3ppbv/hour and were conducive to long-range transport.
Values of LN/Q (Fig. 8b) indicate production rates of O3 are mostly NOx-sensitive (i.e., LN/Q < 0.5) in the PBL over Hebei and some of the largest net production rates of O3 are associated with NOx-sensitivity. In order to control aloft O3 production that 405 has the potential to be transported downwind, NOx is the most important precursor to control. However, at low altitudes near urban centres, the production rate of O3 tends to be more VOC-sensitive (i.e., LN/Q > 0.5), particularly during morning flights.
In Beijing, P. Chen et al., (2013) found a transition of O3 formation from VOC-limited to NOx-limited at ~1 km and many studies conclude O3 production in urban areas of China is VOC-sensitive in spring, while likely more NOx-sensitive in more rural areas (Ran et al., 2011;Xue et al., 2013). Pusede et al., (2014) assessed the temperature dependence of emission control 410 scenarios to lower O3 in San Joaquin Valley, California and concluded reducing organic emissions at moderate and high temperatures with co-occurring NOx decreases will further diminish the number of O3 violations. Thus, the control of NOx as well as VOCs may be necessary to control both aloft and near-ground O3 production in the NCP. quantify the composition and photochemical nature of the lower troposphere associated with smog events. Measurements of trace gases including O3, CO, NOx, NOy, and of aerosol optical properties were acquired in May and June 2016. Twenty-six samples analysed for 54 VOCs were taken aboard a Y-12 research aircraft mostly in the PBL. Our observations confirm heavy loadings of pollution over Hebei.
The major conclusions of our study are: 1. We observed high amounts O3, ranging from 52 ppbv to 142 ppbv, with the highest values found over Shijiazhuang.
The highest NOx concentrations were observed over Xingtai below 500 m. The NOx and CO concentrations ranged from 0.15 ppbv to 49 ppbv and from 91 ppbv to about 2,000 ppbv, respectively. Ratios of CO/CO2 and CO/NOy indicate inefficient combustion from residential coal and biomass burning throughout the region.
2. Concentrations of total measured VOCs reveals alkanes contribute the most by volume mixing ratio (68%), while 425 alkenes/alkynes and aromatics together supply the most (74%) to the calculated OH loss rate. Aromatics constitute most (43%) to the total calculated OFP and toluene, ethylene, m/p-xylene, propylene, and i-pentane play significant roles in the aloft formation of O3 in this region. Sources of VOCs include vehicular emissions, biomass burning, and fuel and solvent evaporation.
3. High amounts of NOx and VOCs throughout the PBL over nonurban parts of Hebei Province were found to generate 430 O3 at a peak mean rate of ~7 ppbv/hour below 500 m. The lower free troposphere (from ~2500 to ~3000 m) was also frequently polluted with CO and NO2 averaging ~125 ppbv and ~140 pptv with peak net production rates of O3 ~3 ppbv/hour, allowing for continued formation of O3 as the air mass travels downwind. The O3 production regime is found to be NOx-limited throughout the PBL over Hebei, while more VOC-limited at low altitudes near urban centres.
Our measurements in spring 2016 over Hebei cannot represent all of China or the seasonal variation of O3 photochemistry, but 435 measurements from an airborne platform make a valuable addition to the understanding of one of the most polluted regions in China, and indeed the world. We show that to improve air quality in Hebei Province, both NOx and VOCs from vehicles and fuel evaporation should be targeted. While VOCs are already targeted for emission reduction in China, the egregious concentrations of O3 observed in this study further confirm the formation of a reactivity-oriented control strategy is urgent.

Author contributions 440
The ARIAs campaign was supervised by RD, ZQ, and XR. XR and HH conducted the measurements on board the research aircraft and VOCs were analyzed by MS and SL. A 2 BC observations were collected by ZL, FW, YW, and FZ. SR helped set up the box model. SB carried out the scientific analysis of the aircraft data and drafted the manuscript with contributions from all co-authors.