Significant influence of oxygenated volatile organic compounds on atmospheric chemistry: a case study in a typical industrial city in China

Abstract

Oxygenated volatile organic compounds (OVOCs), an important subgroup of volatile organic compounds (VOCs), are emitted directly or formed secondarily through photochemical processes. They play a crucial role in tropospheric chemistry as ozone (O₃) precursors. However, due to measurement limitations, the influence of OVOCs on O₃ formation has often been underestimated. In this study, 74 VOCs (including 18 OVOCs) were measured at five representative stations (urban, suburban, industrial, upwind, and downwind stations) in Zibo, an industrial city in the North China Plain. The VOCs level in Zibo (44.6 $\pm$ 20.9 $\times$ 10${}^{-\mathrm{9}})$ is in the upper-middle range ($>$ 32 $\times$ 10${}^{-\mathrm{9}})$ compared to previous studies conducted in most Chinese cities, with OVOCs contributing for 30.0 %–37.8 %. The average O₃ formation potential in Zibo is 410.4 $\pm$ 197.2 $\mathrm{\mu }$g m⁻³, with OVOCs being the dominant contributor (31.5 %–55.9 %). An observation-based model (OBM) was used to access the contributions of chemical production ($R_{\mathrm{NetProd}}$) and emissions/transport ($R_{\mathrm{Emis}\mathrm{\&}\mathrm{Trans}}$) to individual OVOCs. Daytime (08:00–18:00 LT) $R_{\mathrm{NetProd}}$ is highest at the urban site (5.9 $\times$ 10⁻⁹ h⁻¹), while nighttime $R_{\mathrm{Emis}\mathrm{\&}\mathrm{Trans}}$ is most negative at the industrial site (0.76 $\times$ 10⁻⁹ h⁻¹). Simulations without OVOC constraint overestimate OVOCs (42.1 %–126.5 %) and key free radicals (e.g., hydroperoxy radicals (HO₂, 5.3 %–20.4 %) and organic peroxy radicals (RO₂, 6.6 %–35.1 %)), leading to a 1.8 %–11.9 % O₃ overestimation. This overestimation causes an underestimation of hydroxyl radicals (OH) (1.8 %–20.9 %) and atmospheric oxidizing capacity (3.5 %–12.5 %). These findings emphasize the importance of comprehensive OVOC measurements to constrain numerical models, especially in regions with dense anthropogenic emissions, to better reproduce atmospheric photochemistry, and to formulate more effective air pollution control strategies.

1 Introduction

Oxygenated volatile organic compounds (OVOCs), contributing 20.1 %–73.5 % of total volatile organic compounds (VOCs) (Han et al., 2019; Huang et al., 2020; Li et al., 2022a; Liu et al., 2024; Song et al., 2024), are critical components of tropospheric photochemistry (Yang et al., 2014). Photolysis of OVOCs has been proved to be the most significant primary source of RO_x (OH $+$ HO${}_{\mathrm{2}}+$ RO₂) in Guangzhou, Beijing, and Xi'an in China (Wang et al., 2022c; Yang et al., 2018; Zhang et al., 2021b), and thereby accelerating the recycling of radicals to promote ozone (O₃) formation (Qu et al., 2021; Wang et al., 2022c). In addition, previous studies have shown that sufficient free radicals produced by photolysis of OVOCs are the dominating contributors to O₃ pollution during winter (Edwards et al., 2014; Emmerson et al., 2005). The study of Li et al. (2021b) indicates that the fast generation of O₃ during winter haze in the North China Plain is mainly driven by the photolysis of formaldehyde (HCHO), which leads to a large production of HO_x radical and offsets the radical titration induced by NO_x emissions. In addition, HCHO and other OVOCs dominated the OH loss with VOCs (Goldan et al., 2004), resulting in a predominant role in OH reactivity (Ling et al., 2014; Yang et al., 2018). Therefore, OVOCs play a significant role in the atmospheric chemistry.

OVOCs have complex and diverse sources, including primary emissions from anthropogenic, e.g., vehicle exhausts (Gentner et al., 2013; Legreid et al., 2007; Wang et al., 2022b), volatile chemical product use (Ou et al., 2015), industries (Wang et al., 2023), biomass combustion (Gilman et al., 2015; Karl et al., 2007; Li et al., 2014a; Yokelson et al., 2007), and biogenic sources (Ou et al., 2015; Rieksta et al., 2023). They are also formed secondarily through photochemical reactions (Huang et al., 2020; Song et al., 2024; Xia et al., 2021). Mo et al. (2016) estimated that OVOCs from heavy-duty diesel vehicle emissions accounted for 53.8 % of total VOCs in China, and OVOCs account for 12.4 %–46.3 % of VOCs emission from biomass and residential coal combustion, which demonstrates the importance of combustion-related sources of OVOCs. In addition, measurement of VOC fluxes based on the airborne eddy covariance technique showed that urban emission sources comprise a surprisingly large proportion of OVOCs (29 %–56 %) (Karl et al., 2018; Pfannerstill et al., 2023). Due to the high share of OVOCs in VOCs, previous studies have reported that OVOCs could contribute 38 %–60 % of ozone formation potential (OFP) (Liu et al., 2024; Mo et al., 2022; Wang et al., 2022a, 2024). The loss of OVOCs occurs through photolysis, reactions with oxidants (e.g., OH, NO₃, and O₃), dilution mixing and deposition (Atkinson, 2000; Atkinson and Arey, 2003). Moreover, air mass transport also can significantly affect the mixing ratio of OVOCs.

Chemical transport models (CTMs) have been widely used for the study of formation mechanism of OVOCs and their influence on air quality (Chen et al., 2022; de Gouw et al., 2018; Luecken et al., 2012; Steiner et al., 2008; Yang et al., 2023). However, due to the deviation of the meteorological field, uncertainty of the emission inventory (Li et al., 2019; McDonald et al., 2018; Shen et al., 2019), and defects of lumped chemical mechanism (Li et al., 2014b; Sarwar et al., 2008; Stockwell et al., 1997a; Venecek et al., 2018), there is a large uncertainty in the OVOCs simulated by CTMs, which in turn leads to large deviations in the simulated atmospheric photochemistry. The observation-based model (OBM) can avoid these biases to a certain extent by constraining meteorological parameters and chemical species, and leveraging detailed chemical mechanism (e.g., Master chemical mechanism (MCM)). Nevertheless, due to the limited observations of OVOCs (e.g., Pfannerstill et al., 2023), many existing studies use OBM without the observed OVOC data, or only with limited inputs for certain OVOC species (formaldehyde, acetaldehyde, acetone), which can greatly bias the assessment of O₃ generation mechanism, free radical chemistry, and atmospheric oxidation. Wang et al. (2022a) showed that the box model without the constraint of OVOCs underestimates OVOC concentrations, which in turn lead to the underestimation of RO_x and O₃ formation. Thus, it is meaningful to couple OVOC observation with OBM to investigate how OVOCs affect radical chemistry, atmospheric oxidization capability, and O₃ formation mechanism.

Zibo, a typical industrial cluster city in China, has been suffering from O₃ pollution for years (Li et al., 2021a; Qin et al., 2023). However, comprehensive studies involving the observation of VOCs, particularly OVOCs such as HCHO, are rare. Qin et al. (2023) used observations of 98 VOCs (without HCHO) in Zibo to constrain OBM, but the absence of HCHO in their simulation could result in underestimation RO_x, thus disturbing the investigation of OH budget (Fuchs et al., 2017; Guo et al., 2021; Ling et al., 2014; Qu et al., 2021; Tan et al., 2017). This study hypothesizes that incorporating observational constraints on OVOCs significantly influences the OBM simulations. To evaluate this, a 5 day field campaign was conducted across five representative sites in Zibo. Concentrations of 74 VOC species, including 29 alkanes, 16 aromatics, 9 alkenes, 18 OVOCs, acetylene and isoprene, were obtained. The contributions of secondary formation, emissions/transport to OVOCs level were analyzed by the OBM. Additionally, the effect of OVOCs on radical chemistry, atmospheric oxidation capability, and consequently O₃ production were quantified.

2 Methodology

2.1 Sampling sites and measurements

To capture a typical ozone pollution case, the field campaign was conducted from 8 to 12 August 2021, at five monitoring sites (Zhonglou (ZL), Chengdong (CD), Chengqu (CQ), Tianzhen (TZ), and Xindian (XD)) in Zibo (Fig. 1, Table S1 in the Supplement). Among the five sites, the ZL site (117°54^′ E, 36°39^′ N) is an urban site, which is located in the central area of Zibo, and is mainly surrounded by residential areas and factory buildings. According to the prevailing wind direction (northeast, Fig. S1a in the Supplement), CQ (118°60^′ E, 36°57^′ N) the location is an upwind site, while CD (117°53^′ E, 36°31^′ N) is a downwind site. CD is located on a hillside in the southern part of Zibo, with a small number of ceramic and refractory factories nearby. TZ (117°48^′ E, 37°10^′ N) is close to the Shengli Oil field on the west, and is surrounded by farmland. This site is regarded as a suburban site and is affected by residential emissions in the north of Zibo, as well as nearby oil production operations. XD (118°19^′ E, 36°48^′ N) is close to a chemical industrial park and serves as an industrial site. More detailed information about these sites can be found listed in Table S1.

Site-scale wind patterns can affect the levels and spatial distribution of OVOCs and PAMS (target VOC species from the photochemical assessment monitoring stations include 29 alkanes, 16 aromatics, 9 alkenes, isoprene, and acetylene) across sites. Urban (ZL) and downwind (CD) sites are affected by OVOC pollution from northeasterly (NNE, NE, ENE) winds, while the upwind (CQ) site experiences higher OVOC and VOC pollution from both northeasterly and northwesterly (WNW, NW, NNW) winds (Fig. S2a, b). Suburban (TZ) and industrial (XD) sites exhibited higher OVOC and VOC levels under southeasterly (SE) and southwesterly (WSW) winds, respectively, probably due to upwind emissions from nearby industrial sources. Wind speeds (WSs) between 1 and 2 m s⁻¹ were most common (40.4 %) during the observation period. At suburban (TZ) and industrial (XD) sites, OVOC and VOC levels were higher at low WSs (WS $<\mathrm{2}$ m s⁻¹) than that at higher WS, reflecting the influence of local emissions (Fig. S2c, d). In contrast, at the downwind (CD) site, higher OVOC and VOC levels were observed at WS $>$ 2 m s⁻¹, indicating the effect of regional transport. At the urban (ZL) site, higher WS is associated with lower VOC levels and higher OVOC levels, indicating the influence of aging air masses transported from upwind regions.

Figure 1

(a) Map of the North China Plain and (b) map of Zibo with locations of VOCs monitoring stations (red triangles), and Shengli Oil field (yellow star), and prevailing wind direction (orange arrow).

[Figure omitted. See PDF]

During the campaign, two online gas chromatography-flame ionization detector (GC-FID, Thermo Scientific GC5900) systems were deployed at suburban (TZ) and upwind (CQ) stations, and three online gas chromatography-flame ionization detector (GC-FID, Syntech Spectras GC 955–615/815) systems were deployed at downwind (CD), industrial (XD) and urban (ZL) stations, respectively. A mixture of 56 PAMS target species (Spectra Gases Inc., USA, Table S2) was used for the calibration of the GC-FID system. Each VOC analyzer provided measurements with a 1 h temporal resolution. More detailed descriptions of these instruments can be found in previous studies (Li et al., 2023; Wang et al., 2014; Yang et al., 2022a; Zheng et al., 2023). Overall, the detection limits for most VOC species are below 0.1 $\times$ 10⁻⁹. Additionally, 18 oxygenated VOC species were collected by 2,4-dinitrophenylhydrazine (DNPH) sorbent tubes in conjunction with an automated sampler for a period of 1 or 3 h per sample. A 47 mm quartz filter membrane is attached to the front of the sampling tube to filter particulate matter. An ozone scrubber (silica gel column tubing coated with potassium iodide) was placed at the front of the air inlet to avoid ozone interference. OVOCs were derivatized in cartridges to hydrazones during sampling. The cartridges were eluted with 3 mL of acetonitrile and stored at 0–4 °C immediately. Then the eluants were analyzed using an Agilent HPLC, equipped with an ultraviolet absorption detector (UVD), quadruple pump, and an Agilent TM C18 reversed column (250 mm $\times$ 4.6 mm, 5.0 $\mathrm{\mu }$ m). A gradient elution was used, and the mobile phase was a mixture of acetonitrile, tetrahydrofuran, and water. The analysis was carried out using a ternary gradient elution program at a flow rate of 1.2 mL min⁻¹, with a detection wavelength of 360 nm, and sample volume of 10 $\mathrm{\mu }$L at a column temperature of 45 °C. More details about OVOC samplings and analysis can be found in Peng et al. (2023). The lower limit of detection for OVOC were $<$ 0.1 $\times$ 10⁻⁹ (Peng et al., 2023). A total of 271 valid OVOCs samples were collected during the campaign. At the industrial (XD) and suburban (TZ) stations, eight samples were collected per day at 3 h intervals. At the urban (ZL), upwind (CQ), and downwind (CD) stations, 10 samples were collected per day, with seven samples collected at 1 h intervals during 07:00–21:00 LT, and three samples collected at 3 h intervals during the night (01:00–06:00 and 22:00–01:00⁺¹ LT); totaling 59 valid samples per station. Finally, a total of 74 VOCs (56 PAMS and 18 OVOCs) were combined to conduct data analysis in this study (Table S2). Conventional gas phase pollutants (e.g., O₃, CO, and NO_x (NO and NO₂)) were measured using commercial online analyzers (Thermo Scientific 49i, 48i, and 42i, USA) at each site. NO_x was measured by NO chemiluminescence and chemical conversion with a molybdenum convertor, which is known to be affected by NO_z species (Tan et al., 2017, 2019a). Meteorological parameters, including temperature ($T$), relative humidity (RH), WS, wind direction (WD), and ambient pressure ($P$) were obtained synchronously by Chinese ground-based meteorological stations (Boshan, Huantai, Gaoqing, Linzi, and Zichuan sites) (http://data.cma.cn/, last access: 26 March 2024).

2.2 Observation-based model

A box model (F0AM) coupled with MCM v3.3.1 was utilized to simulate the in situ atmospheric chemical process at the five sites (Jenkin et al., 2015; Wolfe et al., 2016). The MCMv3.3.1, as a nearly explicit mechanism with more than 5800 species and 17 000 reactions, provides a more detailed gas chemistry than other lumped mechanisms, such as the Carbon Bond Mechanism (CB) (Yarwood et al., 2005, 2010), Regional Atmospheric Chemistry Mechanism (RACM) (Goliff et al., 2013; Stockwell et al., 1997b), and SAPRC (Carter, 1990, 2010b; Carter and Heo, 2013). The box model calculations were constrained by comprehensive measurements of trace gases (NO, NO₂, CO, and SO₂) and 45 speciated VOCs, encompassing 20 alkanes, 9 alkenes, 14 aromatics, 15 OVOCs, isoprene and acetylene, as well as meteorological parameters ($T$, RH and $P$). To address potential NO₂ measurement artifacts, several adjustments were implemented. Considering that PKU-Mo as a catalytic converter for NO₂ measurement can cause interferences from other nitrogen–oxygen compounds (e.g., PAN, HNO₃), potentially overestimating NO₂ by 30 %–50 % (Kim et al., 2015; Tan et al., 2017, 2019a; Xu et al., 2013). In this study, the observed NO₂ concentrations at the five sites were reduced by 30 %–40 % (40 % for ZL and CQ, 30 % for CD, TZ and XD) to compensate for catalytic converter interferences (Xu et al., 2013). Additionally, strong anthropogenic emissions (e.g., vehicle emissions) near the sites may prevent the model from reaching steady state, leading to positive deviations (Li et al., 2014c). Therefore, NO steady-state approximations (NO_ss), calculated according to the equations proposed by Del Negro et al. (1999) (Eq. S1 in the Supplement), were used to constrain the simulated NO. The uncertainties derived from the NO_x settings are listed in Table S3 and analyzed in Sect. 3.4. HONO was fixed to 2 % of the corrected NO₂ mixing ratio (Elshorbany et al., 2012; Tan et al., 2019a), and the corresponding uncertainty is summarized in Sect. 3.4. In addition, the boundary-layer height (BLH), and surface net solar radiation (SSR) were obtained from the fifth generation of the European Centre for Medium-Range Weather Forecasts (ECMWF) reanalysis for the global climate and weather (https://cds.climate.copernicus.eu, last access: 1 March 2024). The photolysis frequency correction factor ($J_{\mathrm{corr}}$) of the model input was adjusted by SSR. BLH was also included in the model to control the deposition process (Xuan et al., 2023; Zhu et al., 2020).

The model ran with a continuous time-series profile for the campaign period (8–12 August) with a 1 h time step. A sensitivity analysis was performed for the time step and the results are summarized in Sect. 3.4. Each simulation started with a 10 d spin up to reach the steady-state condition. Missing observation data were filled with linear interpolation, and the mixing ratios of OVOCs were also linearly interpolated to 1 h resolution for modeling. An artificial loss process corresponding to an atmospheric lifetime of 24 h or a first-order dilution rate ($k_{\mathrm{dil}}$) of 1/86400 s⁻¹ was introduced for all simulated species, including secondary species and radicals, to approximately simulate dry deposition and other losses (Lou et al., 2010; Tan et al., 2018b; Wang et al., 2022c). The model cases that run with the above settings with 15 constrained OVOC species are called the Base scenario. To investigate the effects of constraints of OVOCs on atmospheric chemistry, the Free scenario was conducted, with all the setting of the Base scenario except for the OVOC constraint.

2.3 Budgets of OVOCs and O₃

At a given site, variations in OVOC mixing ratios are mainly influenced by in situ photochemical production and chemical loss, emissions, regional transport, and deposition (Tan et al., 2018a; Xue et al., 2014a; Zhang et al., 2021b). The change rate of observed OVOCs ($R_{\mathrm{Meas}}$) is calculated by Eq. (1). The in situ photochemical production of OVOC ($R_{\mathrm{ChemProd}}$) is mainly caused by the oxidation of VOCs, while their in situ chemical loss ($R_{\mathrm{ChemLoss}}$) includes photolysis and reactions with oxidants (OH, NO₃, and O₃) (https://mcm.york.ac.uk/MCM/, last access: 13 January 2025) (Atkinson, 2000; Atkinson and Arey, 2003; Jenkin et al., 2015; Saunders et al., 2003). The in situ net OVOC chemical production ($R_{\mathrm{NetProd}}$) (Eq. 2) and their removal by deposition ($R_{\mathrm{Deps}}$) are calculated hourly according to the OBM simulation. The OBM primarily accounts for atmospheric photochemical reactions, and deposition within the boundary layer. However, previous studies have reported that the OBM lacks an explicit representation of transport processes and emissions (Wolfe et al., 2016; Zhang et al., 2021b), making it challenging to disentangle their respective contributions. Therefore, emissions and transport are combined to a single term ($R_{\mathrm{Emis}\mathrm{\&}\mathrm{Trans}}$) to represent their contributions collectively. If the $R_{\mathrm{Emis}\mathrm{\&}\mathrm{Trans}}$ is positive, it is considered a net import of emissions/transport, whereas a negative suggests a net export. The emissions and regional transport of OVOCs ($R_{\mathrm{Emis}\mathrm{\&}\mathrm{Trans}}$) are computed as Eq. (3). $$\begin{array}{}\text{1}& {\displaystyle }{R}_{\mathrm{Meas}}=\sum _i{\displaystyle \frac{\mathrm{d}\left(\right[\mathrm{OVOC}{]}_i)}{\mathrm{d}t}},\text{2}& {\displaystyle }{R}_{\mathrm{NetProd}}=\sum _i(R_{\mathrm{ChemProd},i}-R_{\mathrm{ChemLoss},i}),\text{3}& {\displaystyle }{R}_{\mathrm{Emis}\mathrm{\&}\mathrm{Trans}}=(R_{\mathrm{Meas}}-R_{\mathrm{NetProd}}-\sum _i{R}_{\mathrm{Deps},i}),\end{array}$$ where [OVOC]_i represents the mixing ratios of OVOC species $i$ constrained in OBM, 15 in total (Table S2); dt is the time step of modeling, and d[OVOC]_i refers to the change in mixing ratio of OVOC species $i$.

Considering the oxidation of NO to NO₂ by peroxyl radicals, the total oxidant (O${}_x=$ O${}_{\mathrm{3}}+$ NO₂) is generally used to characterize the chemical budget of O₃ (Kanaya et al., 2009; Xue et al., 2014b). The total chemical production of O_x through oxidations of NO by HO₂ and RO₂ radicals (Tan et al., 2018b), is defined as the production of O₃ ($P$(O₃)), which is calculated according to Eq. (4):

4$$P\left({\mathrm{O}}_{\mathrm{3}}\right)=k_{{\mathrm{HO}}_{\mathrm{2}}+\mathrm{NO}}\left[{\mathrm{HO}}_{\mathrm{2}}\right]\left[\mathrm{NO}\right]+\sum k_{{\mathrm{RO}}_{\mathrm{2}},j+\mathrm{NO}}\left[{\mathrm{RO}}_{\mathrm{2}}{]}_j\right[\mathrm{NO}].$$ The chemical loss rate ($L$(O₃)) of O₃ is equal to the sum of loss rates of O₃ and NO₂, including O₃ photolysis, reactions of O₃ with OH, HO₂ and alkenes, as well as reactions of NO₂ with OH and RO₂, and the reaction of NO₃ with unsaturated VOCs (Chen et al., 2020; Liu et al., 2022; Xue et al., 2016, 2014b). 5$$\begin{array}{rl}L\left({\mathrm{O}}_{\mathrm{3}}\right)& =k_{\mathrm{O}\mathrm{1}\mathrm{D}+{\mathrm{H}}_{\mathrm{2}}\mathrm{O}}\left[\mathrm{O}\mathrm{1}\mathrm{D}\right]\left[{\mathrm{H}}_{\mathrm{2}}\mathrm{O}\right]+k_{{\mathrm{O}}_{\mathrm{3}}+\mathrm{OH}}\left[{\mathrm{O}}_{\mathrm{3}}\right]\left[\mathrm{OH}\right]\\ & +k_{{\mathrm{O}}_{\mathrm{3}}+{\mathrm{HO}}_{\mathrm{2}}}\left[{\mathrm{O}}_{\mathrm{3}}\right]\left[{\mathrm{HO}}_{\mathrm{2}}\right]+k_{{\mathrm{O}}_{\mathrm{3}}+\mathrm{alkenes}}\left[{\mathrm{O}}_{\mathrm{3}}\right]\left[\mathrm{alkenes}\right]\\ & +k_{{\mathrm{NO}}_{\mathrm{2}}+\mathrm{OH}}\left[{\mathrm{NO}}_{\mathrm{2}}\right]\left[\mathrm{OH}\right]\\ & +\sum k_{{\mathrm{NO}}_{\mathrm{2}}+{\mathrm{RO}}_{\mathrm{2}},j}\left[{\mathrm{NO}}_{\mathrm{2}}\right][{\mathrm{RO}}_{\mathrm{2}}{]}_j\\ & +\sum k_{{\mathrm{NO}}_{\mathrm{3}}+\mathrm{VOC},i}\left[{\mathrm{NO}}_{\mathrm{3}}\right][\mathrm{VOC}{]}_i.\end{array}$$ The concentrations of radicals and intermediates are derived from the outputs of the OBM. The $k$ values in Eqs. (4) and (5) rate constants of the corresponding reactions, which can be found from https://mcm.york.ac.uk/MCM/ (last access: 13 January 2025) or the study by Liu et al. (2022). The subscript “$j$” in Eqs. (4) and (5) denotes individual RO₂ species. The subscript “$i$” in Eq. (5) represents individual VOC species. The net O₃ production rate can be obtained from the difference between $P$(O₃) and $L$(O₃).

2.4 Evaluation of ozone formation potential and atmospheric oxidation capacity

Different VOC species vary in their capability to form ozone, and their potential to produce O₃ can be evaluated by the maximum incremental reactivity (MIR) (Carter, 2010a). The ozone formation potential (OFP) calculated for each VOC species represents its maximum contribution to ozone production (Bufalini and Dodge, 1983). The OFP of VOCs is calculated as follows:

6$${\mathrm{OFP}}_i=[\mathrm{VOC}{]}_i\times {\mathrm{MIR}}_i,$$ where OFP_i is the OFP of VOC species $i$ ($\mathrm{\mu }$g m⁻³), [VOC]_i is the atmospheric concentration of VOC species $i$ ($\mathrm{\mu }$g m⁻³), and MIR_i is the maximum incremental reactivity coefficient of the VOC species $i$ (g O₃ (g VOCs)⁻¹) (Table S2) from Carter, 2010a.

Atmospheric oxidation capacity (AOC) is the core driving force of complex air pollution, influencing the removal rate of trace gases and the production rates of secondary pollutants (Liu et al., 2021). AOC is calculated based on the sum of oxidation rates of oxidants (OH, O₃, and NO₃) with primary pollutants (VOCs, CO, and CH₄) (Elshorbany et al., 2009; Geyer et al., 2001; Yang et al., 2022b). The formula is as follows: 7$$\mathrm{AOC}=\sum _i{k}_{Yi}\left[Y_i\right]\left[X\right],$$ where $Y_i$ represents primary VOCs (excluding OVOCs), CO and CH₄, $X$ represents oxidants (OH, O₃ and NO₃), and $k_{Yi}$ is the bimolecular rate constant for the reaction of $Y_i$ with $X$. Atmospheric oxidation capacity determines the rate of $Y_i$ removal.

3 Results and discussion

3.1 Meteorological and chemical conditions

The field campaign is characterized by consistent hot and sunny conditions, with an average daily maximum temperature and SSR of 32.2 $\pm$ 1.4 °C (peak at 34.1 °C) and 2.1 $\pm$ 0.4 $\times$ 10⁶ J m⁻² (Figs. 2, S2a), respectively, which favors the photochemical formation of O₃. A typical O₃ episode was observed, with an average maximum daily 8 h average O₃ (MDA8-O₃) of 89.8 $\times$ 10⁻⁹ in Zibo city. According to the Chinese National Ambient Air Quality Standard Grade II (approximately 93.3 $\times$ 10⁻⁹ for 1 h average, or 74.7 $\times$ 10⁻⁹ for MDA8-O₃), there are four O₃ pollution days (8 to 11 August) during the campaign. The average mixing ratios of SO₂, NO₂, NO, and CO in Zibo are 2.8 $\pm$ 1.6, 12.0 $\pm$ 6.9, 2.8 $\pm$ 4.8, and 897 $\pm$ 670 $\times$ 10⁻⁹, respectively (Figs. 2, S3b). The mean VOC mixing ratio in this study is 44.6 $\pm$ 20.9 $\times$ 10⁻⁹, which is overall higher than that in Beijing (18.3 $\pm$ 8.9 $\times$ 10⁻⁹) from 23 July to 31 August in 2016 (Wu et al., 2023), Rizhao (9.83 $\times$ 10⁻⁹) in summer in 2022 (Zhang et al., 2023), and Xi'an (29.1 $\pm$ 8.4 $\times$ 10⁻⁹) from 20 June to July 2019 (Song et al., 2021). Compared with the median VOC levels ($\sim \mathrm{32}$ $\times$ 10⁻⁹) in other cities in China (Fig. 3, Table S4), VOC levels in Zibo are in the upper-middle range. Previous studies have demonstrated that industrial processes account for approximately 49 % of total VOC emissions in Shandong Province (Jiang et al., 2020; Li et al., 2017; Ren, 2011; Zheng et al., 2021). This indicates strong anthropogenic VOC emission in Zibo. Notably, VOC emission intensity in Zibo was among the highest in Shandong Province, with values $>$ 90 t km⁻² yr⁻¹, even $>$ 108 t km⁻² yr⁻¹ in some areas in 2016 (Jiang et al., 2020; Zhou et al., 2021). In terms of the VOC groups (Figs. 5c, S4a), alkanes and OVOCs were the two predominant groups at each site, accounting for 33.3 %–51.5 % and 30.0 %–37.8 % to the total VOCs, respectively, followed by aromatics (3.8 %–16.5 %) and alkenes (5.0 %–13.8 %). In addition, the difference between peak and valley NO₂ mixing ratios was 14.4 $\pm$ 3.2 $\times$ 10⁻⁹, indicating that substantial NO_x was converted to O₃.

Figure 2

Time profiles of pollutant mixing ratios and meteorological parameters in Zibo from 8 to 12 August 2021. The meteorological data were from ZL, the central site of Zibo, and the pollutants data were the average of the five sites. The hourly PAMS (including alkanes, alkenes, aromatics, acetylene, and isoprene) data were aligned with the $\mathrm{1}/\mathrm{3}$ h sampling intervals of the OVOCs data to ensure comparability between the two datasets.

[Figure omitted. See PDF]

Figure 3

Comparison of VOCs mixing ratios and compositions in this study with former studies based on Table S4. The red dash line represents the median levels ($\sim \mathrm{32}$ $\times$ 10⁻⁹) of VOCs.

[Figure omitted. See PDF]

Figure 4

Distribution of O₃ and its precursors (OVOCs, PAMS, NO₂, and NO) and meteorological parameters (WS, $T$, and RH) at five sites, with the green triangle being the mean value and the horizontal line in the box being the median.

[Figure omitted. See PDF]

Across the five sites, the O₃ mixing ratios are comparable across all the sites (Tables S6, S11) ($p>$ 0.05). However, TVOCs at the suburban site (TZ, 58.5 $\pm$ 35.0 $\times$ 10⁻⁹) are the highest (Figs. 5c, S4b), which is attributed to oil refineries near this site. The downwind site (CD) has slightly lower NO₂ levels (10.8 $\pm$ 5.1 $\times$ 10⁻⁹) and lower TVOC mixing ratios (35.7 $\pm$ 12.5 $\times$ 10⁻⁹) than the urban site (ZL, 14.8 $\pm$ 6.5 and 40.6 $\pm$ 10.3 $\times$ 10⁻⁹, respectively) and upwind site (CQ, 12.7 $\pm$ 8.1 and 42.3 $\pm$ 15.4 $\times$ 10⁻⁹, respectively), and has higher O₃ mixing ratio (58.6 $\pm$ 30.0 $\times$ 10⁻⁹) than the CQ and ZL stations. This may be attributed to the sequential transport of O₃ and its precursors from the upwind station (CQ) to the urban station (ZL), and subsequently to the downwind station (CQ), driven by the dominant northeasterly winds (Figs. 1b, S1).

According to the time series of individual pollutant (Fig. S3b), CQ showed obvious peak mixing ratios of O₃, NO₂, NO, and CO than the other sites during 8–9 August, with stagnant conditions (WS $<$ 2 m s⁻¹), indicating stronger emissions from combustion sources and possibly fast photochemical process near CQ. In addition, XD showed high mixing ratios of CO during 8–9 August, and high daytime TVOCs levels on 9 August (09:00–14:00 LT, 90–110 $\times$ 10⁻⁹). Given CO's relatively inert nature and the absence of similar CO peaks at the other four sites, the abnormal CO peak at XD is related to strong emissions from nearby factories in the industrial park. TZ showed distinct morning and evening peaks of TVOCs at 06:00 LT (163.0 $\times$ 10⁻⁹) and 21:00 LT (120.0 $\times$ 10⁻⁹) on 8 August, and a night peak at 01:00 LT on 10 August (130.3 $\times$ 10⁻⁹), which were attributed to emissions from the neighboring oil field operations. From 10 to 12 August, as WS increased, pollutants levels at all sites decreased to similar levels. Overall, local anthropogenic emissions in Zibo were more prominent under weak wind conditions.

To compare the secondary O₃ formation in each site, the ozone formation potential (OFP) of each VOCs is calculated (Eq. 6). The mean OFP in Zibo during the observation is 410.4 $\pm$ 197.2 $\mathrm{\mu }$g m⁻³, with OVOCs accounting for the largest proportion (31.5 %–55.9 %), followed by aromatics (10.2 %–41.2 %). Alkanes (10.3 %–24.6 %) and alkenes (11.4 %–23.1 %) make comparable proportions, while BVOCs accounted for only 2.1 %–7.6 % of the total OFP (Fig. 5b). The one-way analysis of variance (ANOVA) results ($p<$ 0.05) indicate significant differences (Armstrong et al., 2000) in VOC subclass contribution to OFP across the five sites. Alkanes and aromatics show larger $F$-values (Table S11), reflecting greater variations in the contributions to OFP across the five sites, whereas OVOC and BVOC (isoprene) exhibited lower variability. Post hoc Tukey honestly significant difference (HSD) tests were performed following ANOVA tests to further identify specific significant differences in VOC subcategories between sites (Fig. S14). The OFP of TVOCs is generally similar across stations, except for the downwind station (CD). The upwind station (CQ, 464.2 $\pm$ 162.3 $\mathrm{\mu }$g m⁻³) has the highest OFP, followed by the suburban site (TZ, 456.3 $\pm$ 295.3 $\mathrm{\mu }$g m⁻³), the urban site (ZL, 441.1 $\pm$ 174.5 $\mathrm{\mu }$g m⁻³), the industrial site (XD, 422.9 $\pm$ 166.9 $\mathrm{\mu }$g m⁻³), and the downwind site (CD, 279.4 $\pm$ 101.2 $\mathrm{\mu }$g m⁻³) (Table S5). Differences in OFP levels of aromatics and alkanes at downwind station (CD), suburban station (TZ), and industrial station (XD) are minimal (Fig. S14b, c). However, significant differences in OFP contributed by OVOC at downwind station (CD) compared to suburban station (TZ) and industrial station (XD) are attributed to OVOC emission sources, regional transport, and secondary formation (Fig. S14a). Apart from CQ, OVOCs are the dominant contributors to OFP at each site, especially TZ and XD, with mean OFP of 254.9 $\pm$ 276.1 (55.9 %) and 194.7 $\pm$ 101.0 $\mathrm{\mu }$g m⁻³ (46.0 %) from OVOCs, respectively. This indicates the key role of OVOCs in the formation of O₃ at our observational sites. Among OVOC species, HCHO is the dominant contributor to OFP across the five sites (56.6–202.0 $\mathrm{\mu }$g m⁻³). This is consistent with previous studies (Duan et al., 2008; Huang et al., 2020; Zhou et al., 2024). The top four OVOC species are formaldehyde, acetaldehyde, propionaldehyde, and butyraldehyde, which cumulatively contributed 91 %–95 % of the OFP from OVOCs (Table S5). Additionally, the variety of VOC sources, meteorological conditions, and photochemical conditions at each site lead to differences in key species of OFP at each site. At the suburban site (TZ), isoprene (34.9 $\mathrm{\mu }$g m⁻³) ranks second in terms of OFP after formaldehyde (202.0 $\mathrm{\mu }$g m⁻³), indicating a high effect of biogenic emissions (Mo et al., 2018; Sindelarova et al., 2014). At the industrial site (XD), the contribution of isopentane, marker of oil and gas emissions, to OFP is more prominent (as high as 66.2 $\mathrm{\mu }$g m⁻³) than other sites. OFP of highly reactive aromatic hydrocarbon species, such as m/p-xylene (53.8 $\mathrm{\mu }$g m⁻³) and o-xylene (23.6 $\mathrm{\mu }$g m⁻³) are predominant at the upwind site (CQ), indicating the outstanding contribution of solvent-using sources. OFP contributed by alkenes is the highest at the urban site (101.8 $\pm$ 56.8 $\mathrm{\mu }$g m⁻³) (Fig. 5a), with ethylene and propylene being the most key species, which is consistent with the dense vehicle emission near this site.

Figure 5

(a) Ozone formation potential (OFP) and (b) proportions of OFP contributed by VOCs subgroups, along with (c) mixing ratios of VOCs subgroups at five sites.

[Figure omitted. See PDF]

3.2 Contribution of chemical generation, emissions/transport to OVOCs

OBM simulation results were used to analyze the contributions of chemical processes, and emissions/transport to OVOCs. Overall, the modeled O₃ in the Base scenario exhibited good model performance at the five sites, with $R$ values exceeding 0.85 and index of agreement values greater than 0.80 (Table S7). These metrics indicate a high level of agreement between observed and modeled data, comparable to results reported in previous studies (Qin et al., 2023; Zheng et al., 2023). The contributions of $R_{\mathrm{NetProd}}$ predominantly occur during the daytime (Fig. 6). The maximum average daytime $R_{\mathrm{NetProd}}$ of OVOCs was observed at ZL (5.9 $\pm$ 3.5 $\times$ 10⁻⁹ h⁻¹), followed by CQ (4.11 $\pm$ 11.9 $\times$ 10⁻⁹ h⁻¹), XD (3.6 $\pm$ 2.4 $\times$ 10⁻⁹ h⁻¹), CD (3.5 $\pm$ 2.0 $\times$ 10⁻⁹ h⁻¹), and TZ (1.9 $\pm$ 3.6 $\times$ 10⁻⁹ h⁻¹) sites. This suggests that abundant reactive VOC emissions at urban areas as well as in the industrial areas could lead to faster generation of OVOCs. Generally, the $R_{\mathrm{NetProd}}$ varied with a single peak due to photochemical formation and export transport, with the maximum value at 12:00–14:00 LT. The mean peak of $R_{\mathrm{NetProd}}$ at ZL was 8.8 $\times$ 10⁻⁹ h⁻¹, followed by XD (5.6 $\times$ 10⁻⁹ h⁻¹), CQ (5.5 $\times$ 10⁻⁹ h⁻¹), CD (5.1 $\times$ 10⁻⁹ h⁻¹), and TZ (3.0 $\times$ 10⁻⁹ h⁻¹). Generally, in the early morning hours (04:00–10:00 LT), positive $R_{\mathrm{Meas}}$ at CD, CQ, and XD sites are driven by $R_{\mathrm{Emis}\mathrm{\&}\mathrm{Tran}}$ import. During this period, OVOC mixing ratios show a significant upward trend, peaking between 08:00 and 10:00 LT.

Overall, OVOC mixing ratios at CD, CQ, and ZL sites were typically lower at night but higher during the daytime (Fig. 6), attributing to strong daytime photochemical generation, especially at around 07:00–10:00 LT. In contrast, TZ and XD showed higher nighttime OVOCs than that at daytime, which is due to stronger emission import during night. In addition, though $R_{\mathrm{NetProd}}$ at ZL was the fastest during the daytime, the air mass transport can efficiently export OVOCs to downwind areas, leading to relatively lower OVOC mixing ratios. While at TZ, $R_{\mathrm{NetProd}}$ was the lowest, but daytime OVOCs were the highest due to the predominant daytime import, especially the southwestward import on 8 August (Fig. S5f).

Figure 6

Diurnal profiles of OVOCs contributions from local photochemical production ($R_{\mathrm{NetProd}}$), deposition ($R_{\mathrm{Deps}}$), emission/transport ($R_{\mathrm{Emis}\mathrm{\&}\mathrm{Trans}}$), and measured OVOCs formation rate ($R_{\mathrm{Meas}}$) of the five sites.

[Figure omitted. See PDF]

3.3 Importance of OVOC observational constraint in OBM

To investigate the effect of the constrain setting of OVOCs on OBM performance, the simulated OVOC, O₃, radicals in the Free and Base scenarios are compared (Fig. 7). It has been shown that the box model, which did not take into account transport (including horizontal and vertical diffusion) and emissions, will result in overestimations of OVOCs, peroxyl radical, and PAN (Qu et al., 2021). In this study, OVOCs are overestimated by 42.1 %–126.5 % in the Free scenario compared with the Base scenario (Figs. 7a, S6c), especially HCHO (76.3 %) and benzaldehyde (737.5 %). The daytime RO_x was overestimated by 6.5 %–23.3 %, with RO₂ and HO₂ being overestimated by 6.6 %–35.1 % and 5.3 %–20.4 %, respectively, while OH was underestimated by 1.8 %–20.9 % (Figs. 7d–f, S7b). As shown in Fig. 8a, photolysis of OVOCs (including HCHO) is the predominant source of RO_x radicals ($P$(RO_x)) in the daytime, which is consistent with the findings in Beijing (Liu et al., 2012), Shanghai (Zhang et al., 2021a), Hong Kong (Xue et al., 2016), and Mexico (Volkamer et al., 2010). To assess the effect of OVOCs on the simulation of RO_x species (RO₂, HO₂, and OH), the chemical budgets of these species, as influenced by OVOCs, are quantified according to Liu et al. (2012) and Xue et al. (2016) (Figs. 8b, S8a).

In the Free scenario, the daytime net production of RO_x ($P$(RO_x)) was estimated to range from 0.03 to 0.14 $\times$ 10⁻⁹ h⁻¹ compared to the Base scenario across four sites (excluding TZ), indicating an overestimation of RO_x. Notably, the TZ site exhibited negative $P$(RO_x) values, suggesting the potential existence of unaccounted RO_x sources in this region. The mean daytime $P$(RO_x) in the Free scenario was calculated as 4.8 $\pm$ 2.7 $\times$ 10⁻⁹ h⁻¹, 18.8 % higher than that in the Base scenario (4.0 $\pm$ 2.3 $\times$ 10⁻⁹ h⁻¹). As illustrated in Figs. 8 and S8a, the photolysis of OVOCs (including HCHO) dominates $P$(RO_x), with a mean rate of 2.9 $\pm$ 1.9 $\times$ 10⁻⁹ h⁻¹ in the Free scenario, 27.4 % higher than that in Base scenario (2.3 $\pm$ 1.5 $\times$ 10⁻⁹ h⁻¹). This substantial increase in OVOC photolysis consequently amplified the formation of peroxyl radicals (RO₂ and HO₂). Among the production pathways, the photolysis of HCHO demonstrated the most pronounced effect on HO₂ production in the Free scenario (0.1–1.9 $\times$ 10⁻⁹ h⁻¹), with an increase of 7.8 %–151.2 % (0.1–1.2 $\times$ 10⁻⁹ h⁻¹) in the Base scenario (0.5–1.1 $\times$ 10⁻⁹ h⁻¹) (Fig. 8b).

The interference of OVOCs on OH is comprehensive. On the one hand, increased OVOCs tend to elevate the generation of HO₂, which can directly or indirectly boost OH generation via the reaction of NO (Fig. S9). On the other hand, higher OVOCs levels can decrease OH via the reaction of OH$+$OVOCs (Qu et al., 2021; Tan et al., 2019b). In the Free scenario, total OH sources (including H₂O${}_{\mathrm{2}}+$hv, HONO$+$hv, O${}_{\mathrm{3}}+$hv, and HO${}_{\mathrm{2}}+$NO) is 7.5–12.1 $\times$ 10⁻⁹ h⁻¹, which is 0.3–1.1 $\times$ 10⁻⁹ h⁻¹ higher than that in the Base scenario (10.2–11.0 $\times$ 10⁻⁹ h⁻¹) (Fig. S9). Conversely, OH destruction to peroxyl radicals in the Free scenario (7.1–11.8 $\times$ 10⁻⁹ h⁻¹) is 0.3–2.1 $\times$ 10⁻⁹ h⁻¹ higher than that in Base scenario (6.1–9.7 $\times$ 10⁻⁹ h⁻¹), leading to a net OH loss of 0.1–0.9 $\times$ 10⁻⁹ h⁻¹. This underestimation of OH without OVOCs constraint biases atmospheric oxidation capacity (AOC) by 0.1 %–10.0 % (excluding XD) (Fig. S10), affecting the evaluation of VOCs decay via OH oxidation (Li et al., 2022b).

Figure 7

Comparison of average (a) OVOC, (b) daytime O₃, (c) daytime $|\mathrm{\Delta }$O${}_{\mathrm{3}}|$ (the gap between simulated and observed daytime (08:00–18:00 LT) O₃ mixing ratios), and RO_x (d RO₂, e HO₂, and f OH) between the Base and Free scenario simulations.

[Figure omitted. See PDF]

Figure 8

(a) Average daytime (08:00–18:00 LT) sources and sinks of RO_x in Base case and (b) the impact of OVOCs observationally constraints on RO_x budget, calculated by the difference between the Free and Base scenarios.

[Figure omitted. See PDF]

Figure 9

(a) Average daytime (08:00–18:00 LT) O₃ production and loss rates ($\times$ 10⁻⁹ h⁻¹) in Base scenario and (b) the impact of OVOCs observationally constraints on O₃ budget, calculated by the difference between the Free and Base scenarios.

[Figure omitted. See PDF]

Overall, the gap between simulated and observed daytime O₃ mixing ratios ($|\mathrm{\Delta }$O${}_{\mathrm{3}}|$) in the Free case increased by 7.5 % compared to that in the Base scenario. To better understand the influence of OVOCs on O₃ formation, the rates of the main production and sink pathways of O₃ in the Base and Free scenarios are summarized (Figs. 9, S8b). Compared to the Base scenario, the diurnal $P$(O₃) in the Free scenario increases by 3.7 %–13.9 %, with the reaction rates of RO${}_{\mathrm{2}}+$NO and HO${}_{\mathrm{2}}+$NO increases by 1.6 %–25.9 % (0.1–1.6 $\times$ 10⁻⁹ h⁻¹) and 2.5 %–10.0 % (0.2–1.0 $\times$ 10⁻⁹ h⁻¹), respectively. This is attributed to the elevated RO₂ and HO₂ radical concentrations simulated without the constraints of OVOC observations (Fig. 7d–e). In contrast, the lower simulated concentration of OH radicals in the Free scenario results in a 0.5 %–14.8 % (0–0.1 $\times$ 10⁻⁹ h⁻¹) decrease in the OH$+$O₃/NO₂ reaction rate. Although the $L$(O${}_{\mathrm{3}})$ in the Free scenario is 0.4–1.7 $\times$ 10⁻⁹ h⁻¹ higher than that in the Base scenario, it cannot offset the increase of $P$(O₃), leading to higher net product of O₃. Thus, no constraints of OVOCs could lead to overestimates of peroxyl radicals in the OBM, which in turn significantly overestimates the deviation of O₃ due to the formation pathway of RO₂/HO${}_{\mathrm{2}}+$NO. Of course, the effect of OVOCs varies considerably in different emission and functional regions. The simulated mean daytime O₃ in the Free scenario is 106.3, 86.7, 84.1, 68.8, and 68.2 $\times$ 10⁻⁹ h⁻¹ at ZL, XD, CQ, TZ, and CD, respectively, which is 9.8 %, 1.8 %, 2.0 %, 3.9 %, and 11.9 % higher than that in the Base scenario (Fig. 7b, Table S8). The mean diurnal $|\mathrm{\Delta }$O${}_{\mathrm{3}}|$ at ZL site (29.1 $\times$ 10⁻⁹, 38.2 %) in the Free scenario was significantly higher compared to that of the Base scenario, followed by XD (19.8 $\times$ 10⁻⁹, 6.4 %), TZ (13.1 $\times$ 10⁻⁹, 6.2 %) and CQ (18.1 $\times$ 10⁻⁹, 0.6 %) (Fig. 7c, Table S8), suggesting that the absence of OVOCs constraint in OBM can significantly bias the O₃ formation analysis in urban and industrial areas with complex emissions.

3.4 Uncertainty analysis

The uncertainty of the model comes mainly from the setting of NO_x (Sect. S1 in the Supplement). Sensitive model runs are performed with different NO_x settings to show the corresponding uncertainty. First, NO_ss, VOCs, OVOCs, $T$, RH, and BLH are constrained by the observed data, and sensitivity simulations were carried out for gradient NO₂ mixing ratios (50 %, 60 %, 70 %, and 100 % of observed NO₂ mixing ratios). At the four sites (except TZ), different NO₂ settings produce uncertainties of $-\mathrm{12}$ %–25 % for daytime O₃ and $-\mathrm{17}$ %–36 % for the daytime $R_{\mathrm{NetProd}}$ of OVOCs. In addition, different NO₂ settings produce uncertainties of $-\mathrm{27}$ %–51 % for the $R_{\mathrm{Emis}\mathrm{\&}\mathrm{Trans}}$ of OVOCs. This indicates that reducing the uncertainty of NO₂ observations is important for further atmospheric chemistry modeling. Second, the NO_ss and observed NO mixing ratio were, respectively, used in the model to investigate the influence of nearby NO emissions on the OBM result (Table S3). Sensitivity tests show that directly using the observed NO data may result in an uncertainty of 3 %–20 %, 10 %–27 %, and 7.1 %–38 % for O₃, $R_{\mathrm{NetProd}}$ and $R_{\mathrm{Emis}\mathrm{\&}\mathrm{Trans}}$, respectively, with significantly higher uncertainties at TZ than at other stations, which may be related to the strong transient emissions of NO_x.

The OH radicals are the main oxidant in the atmospheric troposphere. Most trace gases (CO, CH₄, and VOCs) react with OH to produce peroxyl radicals, including HO₂, and RO₂. Peroxyl radicals then react with NO to promote O₃. Photolysis of HONO is one of the important sources of OH radicals. However, direct measurement of HONO was not conducted in this study. The input of HONO in this study used the value of linear relationship with NO₂ mixing ratios, which may cause some uncertainty. Sensitivity analyses were conducted based on the typical urban site (ZL) to quantify the effect of different HONO $/$ NO₂ ratios on free radicals and O₃. As shown in Fig. S12, as the ratio of HONO $/$ NO₂ increased, the mixing ratio of HONO increased, and the concentration of OH produced by photolysis of HONO increased. At the same time, the peroxyl radical and O₃ concentrations also increased (Table S9). At the lowest HONO $/$ NO₂ ratio (0.005), the daytime OH, HO₂, RO₂, and O₃ levels decreased by 4.6 %, 4.8 %, 6.0 %, and 3.4 %, respectively, compared to the HONO $/$ NO₂ ratio of 0.02. In contrast, at the highest HONO $/$ NO₂ of 0.04, the daytime OH, HO₂, RO₂, and O₃ levels increased by 5.4 %, 5.6 %, 7.2 %, and 3.8 %, respectively, compared to the Base case.

When analyzing the contribution of chemical pathways, emission/transport and deposition to OVOCs according to Eqs. (1)–(3), the OVOC budget may be affected by the modeling time step. A sensitivity analysis was conducted at ZL, with simulations at different time steps (30, 10, and 5 min). As shown in Fig. S13, the diurnal trends of the chemical contributions ($R_{\mathrm{NetProd}}$), emission/transport ($R_{\mathrm{Emis}\mathrm{\&}\mathrm{Trans}}$) and deposition ($R_{\mathrm{Deps}}$) to OVOC are similar. The magnitude of the instantaneous change in $R_{\mathrm{Emis}\mathrm{\&}\mathrm{Trans}}$ decreases when the time step is shortened. Specifically, the contributions of $R_{\mathrm{NetProd}}$ and $R_{\mathrm{Emis}\mathrm{\&}\mathrm{Trans}}$ to OVOC increased, while the contribution of $R_{\mathrm{Deps}}$ decreased with shorter time steps (Table S10). When the time step was reduced to 5 min, the contributions of $R_{\mathrm{NetProd}}$ and $R_{\mathrm{Emis}\mathrm{\&}\mathrm{Trans}}$ to OVOC increased by 4.3 % and 5.0 %, respectively, while the contribution of $R_{\mathrm{Deps}}$ decreased by 0.2 %. Therefore, shortening the time step in the model simulation may result in limited increased contribution from $R_{\mathrm{NetProd}}$ and $R_{\mathrm{Emis}\mathrm{\&}\mathrm{Trans}}$ to OVOC.

The OBM, while advantageous for its use of the near-explicit MCM and its ability to reproduce atmospheric chemistry based on observations, has several inherent limitations. First, it inadequately considers emission and transport processes (Wolfe et al., 2016), leading to uncertainties and potential biases in atmospheric chemical analysis. Future studies could differentiate the two processes using CTMs. Second, the steady-state assumption, which underpins OBM calculations, might become invalid in areas strongly influenced by nearby emissions (e.g., high-traffic and industrial intensive regions) (Wolfe et al., 2016). This may potentially cause overestimations of secondary products such as O₃, RO_x radicals, and OVOCs (Li et al., 2021a, 2014c). In this study, simulations use NO_ss as inputs, acknowledging that traffic-related NO emissions may prevent the system from reaching an approximate steady state. The associated uncertainties have been discussed above. Finally, the model may not fully capture all relevant chemical processes, such as heterogeneous reactions or the formation and the fate of intermediate species. For instance, the exclusion of ClNO₂ chemistry could result in underestimating RO_x and O₃ production in certain environments (Riedel et al., 2014; Xia et al., 2020). Future research should focus on refining the chemical mechanisms to better approximate real atmospheric conditions.

4 Conclusions

Compared with previous studies conducted in most Chinese cities, the level of VOCs in Zibo is in the upper-middle range ($>$ 32 $\times$ 10⁻⁹), with OVOCs being the second-largest contributor (29.4 %–36.1 %) after alkanes (34.8 %–53.3 %). Higher levels of OVOCs were observed at sites with more prominent emissions, with mixing ratios ranked as suburban (TZ, 19.7 $\times$ 10⁻⁹) $>$ industrial (XD, 16.8 $\times$ 10⁻⁹) $>$ urban (ZL, 14.9 $\times$ 10⁻⁹) $>$ upwind (CQ, 13.9 $\times$ 10⁻⁹) $>$ downwind (CD, 10.4 $\times$ 10⁻⁹). The OFP in Zibo is 410.4 $\pm$ 197.2 $\mathrm{\mu }$g m⁻³, with OVOCs accounting for the largest proportion (31.5 %–55.9 %). The upwind site (CQ, 464.2 $\pm$ 162.3 $\mathrm{\mu }$g m⁻³) has the highest OFP, followed by the suburban site (TZ, 456.3 $\pm$ 295.3 $\mathrm{\mu }$g m⁻³), the urban site (ZL, 441.1 $\pm$ 174.5 $\mathrm{\mu }$g m⁻³), the industrial site (XD, 422.9 $\pm$ 166.9 $\mathrm{\mu }$g m⁻³), and the downwind site (CD, 279.4 $\pm$ 101.2 $\mathrm{\mu }$g m⁻³) (Table S5). OFP contributed by OVOCs is most dominant at suburban (TZ, 254.9 $\mathrm{\mu }$g m⁻³) and industrial (XD, 194.7 $\mathrm{\mu }$g m⁻³) sites, followed by urban (ZL, 148.9 $\mathrm{\mu }$g m⁻³), upwind (CQ, 146.2 $\mathrm{\mu }$g m⁻³), and downwind (CD, 102.3 $\mathrm{\mu }$g m⁻³) sites. The high levels of OVOCs and their significant contribution to OFP highlights their crucial role in O₃ production across the observing stations.

Based on the OBM simulation, daytime OVOCs primarily originate from photochemical generation, while at nighttime, emissions/transport are the main sources. This diurnal pattern is closely related to the cyclical nature of human activities in urban areas (ZL), where stronger human activities such as vehicle emissions in the daytime enhance the secondary generation of OVOCs. Conversely, in industrial and suburban areas (XD, CQ, and TZ), emissions/transports dominate nighttime OVOC levels, leading to higher mixing ratios at night compared to the daytime.

Without OVOC constraint in the OBM, OVOCs are overestimated by 42.1 %–126.5 % in the Free scenario. The effect of OVOC constraint on $P$(RO_x) is most significant at the urban site (ZL) (29.4 %), comparable to the downwind site (CD) (27.6 %), and higher than the industrial site (XD) (17.8 %), upwind site (CQ) (15.8 %), and suburban site (TZ) (4.7 %). In addition, this overestimation of OVOCs in the Free scenario accelerates the reaction of OH with OVOCs and the photolysis of OVOCs, promoting increased production of RO₂ (6.6 %–35.1 %) and HO₂ (5.3 %–20.4 %), which in turn leads to an overestimation of O₃ (1.8 %–11.9 %) during the daytime. However, the reaction rates of OH with OVOCs are overestimated without OVOC constraint, which leads to underestimation of OH (3.4 %–12.7 %) and AOC (0.1 %–10.0 %). Therefore, to minimize the bias of numerical models, particularly in areas with complex anthropogenic activities, it is essential to intensify OVOC observations and integrate them into numerical models. These efforts are crucial for refining atmospheric photochemistry simulation, improving the accuracy of O₃ formation predictions, and formulating more effective air quality management strategies for regions experiencing similar pollution challenges.

Data availability

The model output is available upon request by contacting the corresponding author.

The supplement related to this article is available online at https://doi.org/10.5194/acp-25-7467-2025-supplement.

Author contributions

KZ, XY, RL, JX, QL, LS, JL, YY, FW, and LY conducted the field measurements. JD and KZ performed the data analysis and prepared the manuscript with contributions from all co-authors. KZ, YF, HC, LH, JT, YW and LL reviewed and edited the manuscript. All authors contributed to data interpretation and discussions.

Competing interests

The contact author has declared that none of the authors has any competing interests.

Disclaimer

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors.

Acknowledgements

This study was financially supported by the National Research Program for Key Issues in Air Pollution Control (DQGG202119). This work is supported by Shanghai Technical Service Center of Science and Engineering Computing, Shanghai University.

Financial support

This research has been supported by the National Natural Science Foundation of China (grant no. 42375102), the Shanghai International Science and Technology Cooperation Fund (24230740200), and the National Research Program for Key Issues in Air Pollution Control (DQGG202119). This work is supported by Shanghai Technical Service Center of Science and Engineering Computing, Shanghai University.

Review statement

This paper was edited by Andrea Pozzer and reviewed by two anonymous referees.