Investigation of Summertime Ozone Formation and Sources of Volatile Organic Compounds in the Suburb Area of Hefei: A Case Study of 2020

Abstract

Ground surface ozone (O₃) is an emerging concern in China due to its complex formation process. In August 2020, field measurements of O₃, NO_x, and volatile organic compounds (VOCs) were carried out in Hefei’s western suburbs. The pollution features of VOCs and O₃ formation were thoroughly analyzed. The total VOC concentration was 42.26 ppb, with the dominant contributor being oxygenated VOCs (OVOCs). Seven emission sources were recognized using the positive matrix factorization (PMF) model, including aged air masses, combustion sources, fuel evaporation, industrial emissions, vehicular emission, solvent utilization, and biogenic emission. Ozone generation mainly occurred under an NO_x-limited regime based on the zero-dimensional box model analysis. According to the scenario analysis, the 13% cut in O₃ might be achieved by the 10% and 30% reduction in NO_x and VOCs, respectively. The O₃ budget analysis demonstrates its high ozone production rate during the pollution period. The influence of regional transport cannot be ignored for high O₃ pollution. This paper provides scientific evidence for O₃ production and the strategies of reducing O₃ by controlling its precursors.

1. Introduction

Tropospheric ozone (O₃) is an important air pollutant that harms human health and plant and animal growth [1,2]. It directly affects the climate and natural ecosystems [3]. In China, ground surface O₃ concentrations have increased rapidly in many cities in recent years [4]. Lu et al. showed that the warm-season daily maximum 8 h average (MDA8) ozone levels from Chinese urban sites increased by 2.4 ppb year⁻¹ between 2013 and 2019 [5]. Nevertheless, the formation mechanisms of O₃ pollution are exceedingly complicated because of the complicated varieties and sources of its precursors [6,7]. Thus, mitigating O₃ pollution has become vital for the promotion of ambient air quality in China.

O₃ generation is affected by multiple factors, such as O₃ precursors, atmospheric oxidation capacity, meteorological factors, and regional transport [8,9,10]. Generally, ground surface O₃ is primarily produced from photochemical processes between volatile organic compounds (VOCs) and atmospheric radicles such as hydroxyl and nitrate in the presence of nitrogen oxides (NO_x) and ammonia (NH₃) [11,12]. In addition, during the night, O₃ is consumed by biogenic VOCs, such as pinene and isoprene [13,14,15]. Over the past two decades, extensive studies have been performed exploring the influences of emissions and atmospheric processes on O₃ pollution in China [16,17,18,19]. Most of them were focused on developed regions, such as the Beijing–Tianjin–Hebei (BTH) [10], Yangtze River Delta (YRD) [7], Pearl River Delta (PRD) [17], and Chengdu–Chongqing regions [18]. Those studies showed that ozone production was more likely to be under VOC-limited regimes in urban regions but NO_x-limited regimes in rural regions in China [16]. However, O₃ generation is much more complicated due to the non-linear relationship between VOCs and NO_x. Therefore, in-depth identification of O₃ and its precursors in the O₃ pollution period is fundamental to fully understanding the O₃ formation process.

Observation-based models (OBMs) and emission-based chemical transport models (EBMs) are widely used to investigate the non-linear relation of O₃ and its precursors [20,21,22,23]. The EBM method is suitable for a large geographical scale [21,22], while the OBM approach mainly utilizes the local scale for in situ O₃ formation chemistry analysis and has the advantage of using local online measurements [23]. For example, Liu et al. (2022) adopted OBM and found that the O₃ episode period was the simultaneous action of regional transport and local photochemical generation in the coastal cities in China [24]. Li et al. (2023) applied the zero-dimensional (0-D) box model (an observation-based method) and revealed that O₃ production regimes covered the VOC-limited, transitional, and NO_x-limited regimes in the nearby oil fields of Shandong Province [25]. Furthermore, VOCs are critical precursors for the formation of ground-level O₃, while there are significant geographical and seasonal variations in the distribution of VOCs in China [19]. Thus, the identification of VOC sources is essential for the implementation of O₃ pollution control. The chemical mass balance method (CMB), principal component analysis (PCA), and positive matrix factorization (PMF) are common source apportionment technologies [26,27]. In these source apportionment technologies, PMF is considered the most efficient quantitative method [26]. PMF can weigh and conduct least squares tests of the data. The distinctive feature of the PMF is that non-negative source component spectra and source contribution can be obtained, and the analysis results are more reliable and meaningful for the source apportionment of environmental pollutants.

Hefei, located in the middle of the YRD, is the capital city of Anhui Province, whose economy and population have been growing rapidly in recent years. Like most other Chinese cities, its air quality has significantly improved in the last five years. The annual concentration of fine particulate matter (PM_2.5) and sulfur dioxide (SO₂) decreased, while the concentration of O₃ increased. Compared with the annual concentrations of O₃ in 2015, it increased by 41.6% in 2020 [28]. It is imperative to reveal the characteristics of ozone pollution. Previous studies have studied VOC and O₃ pollution in the urban atmosphere of Hefei [29,30,31,32]. However, most focused on O₃ spatiotemporal distributions, VOC sources, and the effect of meteorological conditions on ozone formation, with limited studies having been performed on the O₃ precursor relationship in Hefei.

To better understand the chemical interactions of O₃ and its precursors, a field campaign was conducted in the summer of 2020 at a suburban site in Hefei. The PMF model was used to deduce the emission sources of VOCs. A zero-dimensional box model was used to assess the in situ O₃ photochemistry and its sensitivities to O₃ precursors. These findings could help local governments implement effective strategies for the reduction of O₃ and its precursors.

2. Materials and Methods

2.1. Observation Site Location

The observation was performed in the Anhui Institute of Optics and Fine Mechanics (AIOFM), Chinese Academy of Sciences (31°54′19″ N, 117°9′43″ E). This site is located on a peninsula near Shushan Lake in the western suburb of Hefei, approximately 10 km from downtown (Figure 1), and it is surrounded by water on three sides and sheltered by shrubs and trees. The sampling was conducted from 18 August to 2 September 2020.

2.2. Measurement of VOCs and Other Air Pollutants

In this study, VOCs were sampled on the top of the 7th floor of the building in AIOFM, nearly 20 m above the ground. Sixteen VOC samples were collected each day. The specific sampling intervals were 1 hour during the daytime (07:00–19:00) and 3 h at night (19:00–07:00). Each sample was collected in a 3.2 L stainless steel canister (Entech Instrument Inc., Simi Valley, CA, USA) with a flow restrictor (CS1200E series, Entech Instrument Inc., Simi Valley, CA, USA).

The VOC samples were treated using a preconcentrator (Model 7200, Entech Instruments Inc., Simi Valley, CA, USA) combined with a gas chromatography system equipped with a mass spectrometer and a flame ionization detector (GC–MS/FID). The details of the analytical process are described in Text S1. A total of 116 VOC species were measured, including 29 alkanes, 11 alkenes, 1 alkyne, 18 aromatics, 35 halocarbons, 21 OVOCs, and carbon disulfide. VOC species information is listed in Table S1.

Additionally, the meteorological parameters (total solar irradiance, temperature, relative humidity, pressure, wind speed, and wind direction) and trace gas concentrations (SO₂, NO, NO₂, CO, O₃, PM₁₀, and PM_2.5) were collected from the Dongpu Reservoir’s air quality monitoring station (http://sthjj.hefei.gov.cn/, accessed on 1 October 2020), which is installed on the 6th floor of a building in AIOFM.

2.3. Source Identification

PMF version 5.0 software [26,33] was used to determine potential VOC sources. The basic formulas for PMF calculation are as follows:

$$x_{ij}={\displaystyle \sum _{k=1}^p{g}_{ik}{f}_{kj}+e_{ij}}$$

(1)

where x_ij is the concentration of species j measured on sample i, p is the number of factors, f_kj is the concentration of species j in factor profile k, g_ik is the relative contribution of profile k to sample i, and e_ij is the error for the j species measured on sample i. The g_ik and f_kj are modified until a minimum Q value for a specified p is obtained. Q is defined as follows:

$$Q={\displaystyle \sum _{i=1}^n{\displaystyle \sum _{j=1}^m{\left(\frac{e_{ij}}{{\sigma }_{ij}}\right)}^2}}$$

(2)

where σ_ij is the uncertainty of the jth species concentration in sample i, n is the number of samples, and m is the number of species. Details about the PMF model are summarized in the Supplementary Materials (Text S2).

2.4. 0-D Box Model Analysis

To simulate how VOCs contribute to photochemical O₃ formation, a Framework for 0-D Atmospheric Modeling (F0AM) based on the Master Chemical Mechanism (MCM3.3.1) was used in this study [25,34]. The observed meteorological parameters (T, P, and RH) and gaseous pollutants (i.e., O₃, CO, SO₂, NO, NO₂, VOCs) were input into the F0AM as constraints.

Photochemical ozone production in the troposphere can be calculated from the oxidation rate of NO to NO₂ by peroxyl radicals (HO₂, RO₂). Based on the model output, the radical concentrations were used to quantify the generation and destruction pathways of ozone as follows:

$$G\left({\mathrm{O}}_3\right)=k_{{\mathrm{NO}+\mathrm{HO}}_2}\left[\mathrm{NO}\right]\left[{\mathrm{HO}}_2\right]+\left[\mathrm{NO}\right]{{\displaystyle \sum }}_i{k}_i{\left[{\mathrm{RO}}_2\right]}_i$$

(3)

$$D\left({\mathrm{O}}_3\right)=k_{{\mathrm{O}}^1{\mathrm{D}+\mathrm{H}}_2\mathrm{O}}\left[\mathrm{O}{(}^1\mathrm{D}\right)\left]\right[{\mathrm{H}}_2\mathrm{O}]+k_{{\mathrm{O}}_3+\mathrm{Alkene}}\left[{\mathrm{O}}_3\right]\left[\mathrm{Alkene}\right]+k_{{\mathrm{O}}_3+\mathrm{OH}}\left[{\mathrm{O}}_3\right]\left[\mathrm{OH}\right]$$

$${+k}_{{\mathrm{O}}_3{+\mathrm{HO}}_2}\left[{\mathrm{O}}_3\right]\left[{\mathrm{HO}}_2\right]+k_{{\mathrm{NO}}_2+\mathrm{OH}}\left[{\mathrm{NO}}_2\right]\left[\mathrm{OH}\right]$$

(4)

P(O₃) = G(O₃) − D(O₃)

(5)

where G(O₃) is the gross O₃ production rate, D(O₃) is the ozone chemical loss rate, P(O₃) is the net O₃ production rate, and k represents the related reaction rate constant. Detailed descriptions of the chemistry calculation can be found elsewhere [24,25].

The relative incremental reactivity (RIR) was used to calculate the sensitivity of ozone production to its precursors, and it is defined by the following equation:

$$\mathrm{RIR}=\frac{\Delta P({\mathrm{O}}_3)/P({\mathrm{O}}_3)}{\Delta \mathrm{X}/\mathrm{X}}$$

(6)

where X is the specific O₃ precursor, ΔP(O₃)/P(O₃) is the percentage change in the ozone production rate, ∆X/X is the fraction of ozone precursors reduction, in this study, a hypothetical relative change (∆X/X) of 20% was set. A positive RIR value indicates that reducing emissions of the target precursor is effective in reducing O₃ production, and the larger the positive value, the more sensitive to the target precursor for O₃ formation [25].

The Empirical Kinetics Modeling Approach (EKMA) curve is an important tool for diagnosing the sensitivity of zone formation to VOCs and NO_x. The O₃ concentration was simulated by setting different VOCs and NO_x mixing ratios. We set up a 20 × 20 matrix by reducing or increasing the measured VOCs and NO_x concentrations in the model input. The ozone concentration was modeled with intervals of 24 h spanning from 0:00 to 23:00 at a 1 h time resolution. The maximum daily O₃ concentration (06:00–18:00) can be obtained using various concentrations of NO_x and VOCs. Finally, the O₃ maximum concentration was plotted under the corresponding scenarios to generate an EKMA curve. In this study, S(NO_x), S(VOCs), and S(O₃) correspond to concentrations of NO_x, VOCs, and O₃, respectively.

3. Results and Discussion

3.1. Overview of Meteorology and Pollutants Characteristics

3.1.1. Levels of Air Pollutants and Meteorological Factors

Figure 2 shows the meteorological parameters, O₃ concentrations, and precursor time series observed from 18 August to 2 September 2020. During the campaign, the average T, RH, and wind speed were 27.94 °C, 86.1%, and 2.31 m s⁻¹, respectively, with the prevailing wind direction coming from the northeast. The mixing ratio of O₃ changed greatly, and its concentration ranged from 0.5 ppb to 110 ppb, with a mean value of 45.66 ppb. According to the specification for stage II of the Chinese National Ambient Air Quality Standards (NAAQS) [35], the MDA8 of O₃ concentration is 160 µg m⁻³, which is comparable to 75 ppb under the typical summer conditions (25 °C, 101.3 kPa). As shown in Figure 2, two cases of different O₃ pollution levels were chosen to study the characteristics of VOCs and O₃. The dates when the MDA8-O₃ mixing ratio was greater than 75 ppb were classified as O₃ episode days, and the remaining dates were non-O₃ episode days.

The time profiles of the VOC groups in this study are depicted in Figure 2. The average total TVOCs (TVOCs) mixing ratio was 42.26 ppb. The average concentrations of OVOCs, alkanes, halogenated hydrocarbons, alkenes and alkynes, aromatic hydrocarbons, and sulfide were 22.13, 8.99, 5.63, 2.62, 1.61, and 1.3 ppb, respectively. They accounted for 52.4%, 21.3%, 13.3%, 6.2%, 3.8%, and 3% of the total detected VOCs, respectively. A comparison of the results with different cities in China is listed in

Table 1. Mixing ratio and proportion of VOC components in Hefei and other cities in China.

Time	Location	TVOCs (ppb)	Proportion (%)					Reference
			Alkanes	Alkenes	Halocarbons	OVOCs	Aromatics
May 2018	Zhengzhou	29.11	30.1	10.7	20.6	31.1	5.6	[27]
May–November 2019	Ningde	49.1	44.5	5	7.9	23.3	19.3	[19]
May–November 2018	Guilin	23.67	19.1	4.1	5.7	1.8	65.7	[38]
July 2021	Yibin	23	31	7	13	36	13	[36]
September–December 2018	Guangzhou	76.84	25.38	–	–	57	5.7	[40]
May 2017	Shanghai	42.70	35.36	5.62	12.65	31.62	11.94	[39]
June–July 2019	Xi’an	29.1	35.75	10.23	11.11	31.77	6.22	[37]
August 2020	Hefei	42.3	21.3	3.4	13.3	52.4	3.8	This work

. The concentration of TVOCs in Hefei’s suburbs was higher than that in Zhengzhou (29.11 ppb) [27], Yibin (23 ppb) [36], Xi’an (29.1 ppb) [37], and Guilin (23.67 ppb) [38], comparable to Shanghai (42.7 ppb) [39], and lower than Guangzhou (76.84 ppb) [40] and Ningde (49.1 ppb) [19]. OVOCs were the main components of VOCs in this study, in contrast to other cities in China where alkanes were the largest contributor to VOC species. The high mixing ratio of OVOCs in summer Hefei was similar to that observed in Zhengzhou [27] and Guangzhou [40].

3.1.2. Diurnal Variation Characteristics

Diurnal variations for diffident O₃ pollution levels at the AIOFM site are shown in Figure 3 and Figure S1. During the O₃ pollution days, there was a relatively higher total solar irradiance (TSI), temperature and lower relative humidity than during non-O₃ pollution periods, which is consistent with previous observations that high solar radiation and temperature and low relative humidity could promote ozone generation [25,41]. The diurnal profile of O₃ shows an inverse variation trend with NO₂ and VOC groups. The peak values of ozone were 94.2 ppb (O₃ pollution period) and 63.68 ppb (non-O₃ pollution period) in the two periods, respectively. The observational results demonstrate that increased O₃ concentrations are related to stronger solar radiation, high temperature, low relative humidity, and higher precursor concentrations.

Generally, the diurnal profile of VOCs in the atmosphere is characterized by direct emissions, consumption by atmospheric oxidants, and boundary layer height (BLH) variations [42]. Increased anthropogenic activities in the morning resulted in VOC accumulation. The VOCs gradually decreased with the increase in the photochemical reaction and BLH (Figure S1). When the photochemical reaction stops in the evening, the concentration of VOCs rises due to the influence of emissions. As anthropogenic emissions species, benzene, and propane demonstrated consistent daily variation with multiple peaks, which is in agreement with the diurnal profiles of NO₂ (Figure 3). The first peak occurs in the early morning, typical of traffic source emissions. As the typical tracer for biogenic emission, isoprene showed good consistency with the diurnal profiles of temperature and solar radiation (Figure S1). As shown in Figure 3, it was clear that the nighttime and morning VOC mixing ratios during the O₃ pollution days were higher than those during the non-O₃ pollution period. This suggested that the high mixing ratios of VOCs gathered at night and morning might be prime reasons for the increasing O₃ concentration.

Table 2. Average values of air pollutants and meteorological parameters during the entire observation period, non-O₃ episode days, and O₃ episode days.

Species	Entire Observation Period	Non-O3 Episode Days	O3 Episode Days
Temp (℃)	27.94 ± 3.38	27.63 ± 3.25	28.67 ± 3.57
RH (%)	86.12 ± 16.62	87.9 ± 14.78	81.91 ± 16.71
Wind speed (m s−1)	2.31 ± 1.34	2.29 ± 1.35	2.35 ± 1.33
SO2 (ppb)	2.06 ± 0.48	1.99 ± 0.46	2.21 ± 0.48
CO (ppm)	0.49 ± 0.14	0.47 ± 0.14	0.54 ± 0.14
NO2 (ppb)	6.48 ± 3.98	6.29 ± 4.1	6.94 ± 3.63
O3 (ppb)	45.66 ± 23.81	41.57 ± 20.46	55.54 ± 28.04
PM2.5 (µg m−3)	24.78 ± 11.42	21.56 ± 10.87	32.52 ± 8.67
PM10 (µg m−3)	49.33 ± 22.88	44.84 ± 21.81	60.06 ± 21.79
Alkanes (ppb)	8.99 ± 6	8.6 ± 6.53	10.05 ± 4.03
Alkenes & alkynes (ppb)	2.62 ± 1.52	2.46 ± 1.28	3.05 ± 1.98
Halocarbons (ppb)	5.63 ± 1.46	5.44 ± 1.43	6.15 ± 1.42
OVOCs (ppb)	22.13 ± 12.64	20.15 ± 10.02	27.53 ± 6.79
Aromatics (ppb)	1.61 ± 0.94	1.53 ± 0.99	1.84 ± 0.71
Sulfide (ppb)	1.28 ± 1.27	1.3 ± 1.25	1.22 ± 1.33
TVOCs (ppb)	42.26 ± 16.92	39.47 ± 15.27	49.86 ± 18.76

shows that all VOC groups increased continuously during the O₃ period except for carbon disulfide. Alkanes, aromatic hydrocarbons, alkenes and alkynes, halogenated hydrocarbons, and OVOCs increased by 16.9%, 20.1%, 24.1%, 13.1%, and 36.7% on the O₃ episode days, respectively. The concentration of OVOCs increased the most, with an average concentration of 27.53 ppb during the pollution period, accounting for 55% of the total detected VOCs. As with the features of the whole campaign, OVOCs were the most abundant components in both non-O₃ and O₃ pollution days. Except for direct emissions, OVOCs can also be generated by chemical reactions from the oxidation of their precursor hydrocarbons [43]. Moreover, the mean concentrations of NO₂, O₃, PM_2.5, and PM₁₀ on the O₃ episode days were 6.94 ppb, 55.54 ppb, 60.06 µg m⁻³, and 44.84 µg m⁻³, respectively (Table 2). Compared with the non-O₃ pollution period, NO₂, O₃, PM_2.5, and PM₁₀ increased by 10.4%, 33.6%, 50.7%, and 34%, respectively. Thus, high concentrations of precursors and increased atmospheric oxidation may lead to more secondary components, such as OVOCs, O₃, and PM_2.5.

3.2. Source Apportionment of VOCs

3.2.1. Ratio of Specific Compounds

The ratios between specific VOCs can reflect different sources. The most commonly used ratios are propane/ethane (P/E), i-pentane/n-pentane (i/n), m/p-xylene/ ethylbenzene (X/E), and toluene/benzene (T/B) [19,38]. The P/E ratio is frequently used to assess the impacts of liquefied petroleum gas/natural gas (LPG/NG) use on VOCs [44,45]. As shown in Figure 4, the average P/E ratio in this study was 1.66 ± 0.64, lower than that in LPG/NG vehicle exhaust (about 3) [46], indicating that propane was also affected by other emission sources. The i/n ratio is 0.56~0.8 for coal burning, 1.8~4.6 for fuel evaporation, and 2.2~3.8 for vehicle exhaust emission [19,47]. As shown in Figure 4, the i/n ratio during the observation period was mainly distributed between 1.34~1.88, with a mean value of 1.66 ± 0.64, suggesting that alkanes near the sampling site were related to fuel evaporation emission.

The T/B ratio at about 0.3 is for combustion sources, 0.9~2.2 for vehicle exhaust, and 1.4~5.8 for solvent usage sources [37,48,49]. As depicted in Figure 4, the T/B ratio was mainly distributed between 0.46~1.29, with an average value of 1.0 ± 0.77, indicating that the VOCs near the sampling site had the characteristics of combustion and traffic source emissions. However, the correlation between toluene and benzene was poor (Figure S2), with a correlation coefficient (R²) of 0.15, suggesting that the sources of toluene and benzene differed significantly. The X/E ratio is often used to indicate the degree of air mass aging [50,51,52]. The small X/E ratio revealed a high degree of air mass aging, which could be attributed to the influence of air mass transport. Generally, an X/E ratio in the range of 2.5~2.9 is an attribute of typical urban environments. If the ratio is much smaller than 2.9, the VOCs are impacted greatly by long-distance transmission [53,54]. The correlation coefficient between m/p-xylene and ethylbenzene in this study was 0.74 (Figure S2), implying that the VOC sources had a high consistency. As demonstrated in Figure 4, the average value of X/E was 1.11 ± 0.36, much lower than 2.9, highlighting the role of long-distance transport in VOC sources.

3.2.2. PMF Analysis

PMF 5.0 software was used for source analysis in this study. The OVOCs and alkenes with high chemical activities can be greatly affected by the secondary formation. These pollutants were excluded from the PMF analysis [55]. The Supplementary Materials show the factor screening and error analysis for the source analysis model (Text S2). Seven VOC sources were identified: combustion, biogenic emission, fuel evaporation, solvent utilization, industrial process, vehicular emission, and aged air masses. Details characterizations are provided in Figure S1 and Text S3.

Figure 5 shows the proportion of various sources according to the PMF results. The aged air masses made the largest contributions, accounting for 25% of atmospheric VOCs. The combustion sources, fuel evaporation, and industrial emission were also relatively high, contributing 19.2%, 15.3%, and 14.4%, respectively. The vehicular emission, solvent utilization, and biogenic emission contribution were relatively low at 9.4%, 8.7%, and 8.1%, respectively. Furthermore, the contributions of various sources did not differ significantly at different ozone pollution levels. Compared with the non-O₃ pollution period, the contributions of aged air masses and the combustion sources increased only by 3.2% and 4%, respectively. The proportion of industrial and biogenic emissions decreased by 2.1% and 2.2%, respectively. However, the mixing ratio of each source increased significantly, except for biogenic emissions. The mixing ratio of combustion sources, aged air masses, and fuel evaporation increased by 38%, 29.6%, and 21.3%, respectively.

3.3. Photochemical O₃ Formation and Regional Transport

3.3.1. Sensitivity Analysis of Ozone Formation

The ozone generation process was simulated by F0AM–MCM to diagnose the sensitivity of ozone generation to its precursors in this study. The RIR values of the O₃ precursors corresponding to the different ozone pollution levels are shown in Figure 6. The RIR value of NO_x is the greatest, followed by alkenes, OVOCs, aromatics, alkanes, and halocarbons. The RIR values of VOCs and NO_x were both positive during the campaign. This revealed that reducing all precursors could benefit overall O₃ regulation in summer in Hefei’s suburbs. Moreover, the RIR value for OVOCs increased during the ozone episode days, implying that reducing OVOCs might reduce O₃ pollution on high O₃ days. The high fraction of alkenes and OVOCs in RIR values suggests that ozone control in Hefei’s suburbs should be performed in consideration of such VOC species in the summer.

Figure 7 depicts O₃-VOCs-NO_x sensitivity using the EKMA curve. A total of 400 scenarios were performed, assuming that 20 emission scenarios were for NO_x and VOCs. The photolysis rates and meteorological conditions for all the scenarios were based on the average conditions during the whole campaign. The maximum O₃ concentration (06:00–18:00) modeled is depicted with spots of various colors on the corresponding Z-axis. The black line is the ridgeline, formed by the turning point of the O₃ isopleth. Ozone production is NO_x-limited below the ridgeline and VOC-limited above the ridgeline, with a transition region near the ridgeline. As depicted in Figure 7, the slope of the ridgeline is approximately 3 (VOCs/NO_x), which provides a scientific basis for O₃ control policies in reducing VOCs and NO_x emissions [56]. The daily average concentrations of VOCs and NO_x were plotted in the EKMA curve. Most of those were below the ridgeline, indicating that ozone formation is significantly affected by NO_x at the AIOFM site, which is consistent with those found in the other suburban regions in China [57,58]. However, Figure 7 also shows that several spots were scattered close to the ridgeline in the EKMA curve (18 August and 28 August), indicating that reducing NO_x alone was not the most effective strategy for cutting O₃.

3.3.2. O₃ Control Strategies

In order to investigate O₃ control policies during the O₃ episode days, F0AM–MCM was used to simulate different scenarios with reductions of VOCs and NO_x. The EKMA results showed that ozone generation was more sensitive to NO_x during the whole campaign, which is compatible with the results of the RIR analysis. Furthermore, the MDA8h of the O₃ mixing ratio during the O₃ episode days was 85.4 ppb, exceeding stage II of the NAAQS of 75 ppb for O₃ by 13%. As shown in Figure 8, the 13% ozone cut goal was achieved with the 26% mitigation of S(NO_x) and the S(VOCs), maintaining the initial emission. NO_x emission reduction might be efficient for reaching short-term O₃ attainment in local areas. However, considering the results of RIR and EKM analyses, and full compliance with the long-term multi-pollutant air quality improvement strategy in China, VOCs should be properly decreased, since they are important precursors of SOA; to reach the 13% reduction of O₃, a more practical and reasonable way would be to reduce S(VOCs) and S(NO_x) by 30% and 10%, respectively (Figure 8). Moreover, the reduction ratio of VOCs/NO_x follows the ridgeline in the EKMA curve (Figure 7). In general, some conclusions and suggestions were drawn for the short observation period, and longer-term field measurements of air pollutants should be performed in the future. Simultaneously, it is crucial to properly adapt the mitigation of VOCs and NO_x policies when O₃ formation sensitivity changes.

3.3.3. In Situ O₃ Formation Process

Variations in O₃ concentration depend on the local photochemical production and physical transport processes [10,59,60]. Thus, the impact on O₃ concentrations of photochemical and physical transport processes can be expressed as follows:

$$\frac{\mathrm{d}({\mathrm{O}}_x)}{\mathrm{d}t}=P({\mathrm{O}}_3)+R({\mathrm{O}}_3)$$

(7)

where d(O_x)/dt is the observed rate of change in O₃, P(O₃) is the net O₃ formation rate, and R(O₃) is the regional transport rate. P(O₃) can be computed by the F0AM. O₃ was replaced with O_x (O_x = O₃ + NO₂) to correct the titration of O₃ by NO [61]. The positive R(O₃) values indicated the O₃ import of regional transport and vice versa [24,61].

The results of the ozone budget analysis conducted at the AIOFM site are shown in Figure 9. The d(O_x)/dt shows a similar increase in the morning during the O₃ and non-O₃ episodes. Compared with the O₃ pollution period, a positive d(O_x)/dt is observed at 06:00, 1 h earlier than the non-O₃ pollution period. Moreover, the increase rate of O_x is the fastest during the O₃ pollution period in the morning hours, and it remains positive until 14:00, while the d(O_x)/dt sharply decreases to zero at 13:00 during the non-O₃ pollution period. The difference in the d(O_x)/dt results in a high O₃ peak during the O₃ pollution period.

As shown in Figure 9, the maximum values of P(O₃) were 13 ppb h⁻¹ (O₃ pollution period) and 7.31 ppb h⁻¹ (non-O₃ pollution period) at around 09:00, respectively, and the increase in the O_x mixing ratio is close to the local ozone production rate, indicating that the fast O₃ mixing ratio increase in the morning (around 06:00–09:00) is mainly caused by local photochemical production. However, both positive R(O₃) values for the different pollution periods were observed at around 09:00–12:00, indicating that the local ozone budget was O₃ importing. Specifically, during the O₃ pollution period, the R(O₃) values at 10:00 and 11:00 were 8.46 ppb h⁻¹ and 7.31 ppb h⁻¹, respectively. They are responsible for 47.9% and 50.9% of the d(O_x)/dt, respectively. This result indicated that the high O₃ formation was caused by the interaction of physical transport and local photochemical production in the studying area. The O_x is probably emitted from the residual layer, which holds the high O_x load generated on the earlier day and then separated from the surface layer [59].

The O₃ import stops (negative R(O₃)) at around 13:00 during the O₃ pollution period, while the local photochemical production kept producing O₃, it showed that the downwind regions as well as the in situ O₃ concentration were both influenced by the local photochemical production at the studying site. This can be further observed in Figure 9a under high-O₃ conditions. Overall, these results indicate that regional transport and local photochemical production played significant roles during the O₃ pollution period.

4. Conclusions

In this study, comprehensive field observations were conducted in the western suburbs of Hefei from 18 August to 2 September 2020. A total of 116 VOC species were monitored, and the mixing ratio of TVOCs was 42.26 ppb. OVOCs were the most abundant group, followed by alkanes. Compared with the non-O₃ pollution period, enhanced solar radiation, high temperature, and low humidity conditions were observed during the O₃ pollution period. Meanwhile, the mixing ratios of NO_x and VOCs increased significantly. Those conditions led to greater formation of secondary pollutants.

The average X/E ratio during the campaign was 1.11, with a high degree of air mass aging, suggesting that regional transportation had a major impact on VOCs. Seven sources were recognized by the PMF model, including aged air masses (25%), combustion sources (19.2%), fuel evaporation (15.3%), industrial emission (14.4%), vehicular emission (9.4%), solvent utilization (8.7%), and biogenic emission (8.1%). The contributions of each source were comparable under different O₃ episode days. However, the mixing ratio of each source greatly increased, except for biogenic emission during the O₃ pollution period.

Based on the RIR and EKMA analyses, the ozone generation in summer in the western suburbs of Hefei was more sensitive to NO_x, with alkenes and OVOCs as the major contributors to O₃ formation. The F0AM–MCM-simulated results illustrated that a 10% diminution of S(NO_x) and 30% diminution of S(VOCs) may result in a 13% diminution of S(O₃) during the O₃ episode days. The in situ O₃ generation process results demonstrated that regional transport and local photochemical production aggravated local O₃ concentration.