The critical role of oxygenated volatile organic compounds (OVOCs) in shaping photochemical O₃ chemistry and control strategy in a subtropical coastal environment

Photochemical ozone (O₃) pollution remains a persistent environmental challenge, and growing evidence highlights the critical role of oxygenated volatile organic compounds (OVOCs) in photochemical processes. However, comprehensive and quantitative measurements of OVOCs remain limited. This study investigates the impact of OVOCs on O₃ formation mechanisms and radical budgets by intergrating high-resolution field measurements from a subtropical coastal region in South China with observation-based photochemical modeling. 63 OVOC species were measured by a proton-transfer-reaction time-of-flight mass spectrometry (PTR-ToF-MS), and accounted for 72 %–77 % of total VOC concentrations. The O₃-precusor relationship analysis revealed a transition regime for O₃ formation and high sensitivity to OVOCs. OVOC-related reactions, including OVOC photolysis, OVOC oxidation by OH and NO₃ radicals, contributed approximately 36 %–73 % to daytime production rates of HO₂ and RO₂ radicals. Model simulations without comprehensive consideration of OVOCs would significantly underestimate daytime production rates of O₃ and RO_x radicals by 41 %–48 %, and shift the diagnosis of O₃ formation from a transition regime to a VOC-limited regime, leading to biased policy recommendations and potentially ineffective control strategies. These findings underscore the critical role of OVOCs in atmospheric photochemistry and highlight the urgent need for comprehensive OVOC quantification to improve OVOC-inclusive model frameworks. Such improvements are essential for accurately characterizing O₃-precursor relationships and for developing effective and sustainable strategies to mitigate regional O₃ pollution.

1 Introduction

Ground-level ozone (O₃) is a significant secondary air pollutant and a major component of photochemical smog, posing serious threats to human health, ecosystems, and the climate (Feng et al., 2019; Yue and Unger, 2014; Mills et al., 2018). Elevated O₃ levels remain a persistent environmental challenge in many urban regions worldwide, especially a notable upward trend in East Asia (Li et al., 2019b; Li et al., 2020). O₃ formation in the troposphere arises from complex photochemical reactions involving nitrogen oxides (NO_x) and volatile organic compounds (VOCs) under the sunlight (Zhao et al., 2022; Xu et al., 2022). The oxidation of VOCs by hydroxyl (OH) radical plays a central role in producing peroxy radicals, such as hydroperoxyl (HO₂) and alkyl peroxy (RO₂) radicals, which sustain the chain reactions driving photochemical O₃ production (Lyu et al., 2022).

Previous studies have underscored the importance of non-methane hydrocarbons, particularly alkenes and aromatics, as major precursors of O₃ formation, leading to targeted control measures in various regions (Li et al., 2015; Li et al., 2017; Hong et al., 2019). However, recent research highlights the crucial roles of oxygenated VOCs (OVOCs) in regulating atmospheric oxidation capacity and contributing to radical production (Wang et al., 2022a; Shen et al., 2021; Chai et al., 2023; Yang et al., 2023). OVOCs, such as carbonyls and alcohols, can be emitted directly from diverse sources or formed as secondary products from VOC oxidation (Mellouki et al., 2003). These OVOC compounds can further react with OH radical or undergo photolysis processes, serving as significant sources of HO₂ and RO₂ radicals that amplify radical cycling and promote O₃ formation (Xue et al., 2016; Chen et al., 2020; Huang et al., 2020b). Despite their importance, the diversity and high reactivity of OVOCs introduce substantial uncertainties in atmospheric chemistry and air quality models, particularly in regions with limited OVOC measurements.

Traditional analytical techniques such as gas chromatography (GC) coupled with flame ionization or mass spectrometry detection (FID/MS) have been widely used to measure non-methane hydrocarbons, but only a subset of OVOCs (Huang et al., 2015; Yang et al., 2019; Li et al., 2019a; Han et al., 2019). High-performance liquid chromatography (HPLC) can detect several carbonyl compounds, such as formaldehyde, acetaldehyde, acetone, but its reliance on offline sampling limits its temporal resolution, making it less effective for capturing real-time atmospheric variations (Lu et al., 2010; Yang et al., 2017; Zhang et al., 2024). Furthermore, many other key OVOC species, such as larger aldehydes, ketones, carboxylic acids, organic peroxides, and other multifunctional compounds, have been rarely measured and poorly characterized in ambient air, which may result in underestimating the role of OVOCs in atmospheric chemistry (Wang et al., 2022a). While NO_x and VOCs are well-recognized as the primary drivers of O₃ formation, the role of OVOCs in shaping photochemical O₃ chemistry has received comparatively less attention due to limited field observations and insufficient representation in chemical models. In addition, previous model studies tended to considerably underestimate RO_x (OH, HO₂ and RO₂) radicals compared to observation (Hofzumahaus et al., 2009; Ma et al., 2019; Rohrer et al., 2014). Some studies have attempted to address these gaps by simulating unmeasured OVOC species using photochemical box model, but large uncertainties still exist, largely due to missing OVOC primary sources, incomplete or underestimated secondary chemical pathways (Karl et al., 2018; Mo et al., 2016; Bloss et al., 2005; Li et al., 2014). These knowledge gaps hinder an accurate understanding of the O₃-precursor relationship, complicating the development of effective control strategies.

To gain a comprehensive understanding of the role of OVOCs in photochemical O₃ chemistry, a continuous field campaign was conducted at a coastal suburban site in Hong Kong in South China. High-resolution measurements of more than sixty OVOC species were measured using proton-transfer-reaction time-of-flight mass spectrometry (PTR-ToF-MS). By integrating these measurements into a model simulation framework, we quantified the contributions of OVOCs to radical production and O₃ formation, and examined their impacts on O₃-precursor relationship, providing critical insights into the formulation of targeted strategies for mitigating O₃ pollution in this subtropical region and similar environments.

2 Methodology

2.1 Field measurement and instrumentation

Field measurements were conducted at a suburban coastal site (22.33° N, 114.27° E) located at the campus of The Hong Kong University of Science and Technology (HKUST) in eastern Hong Kong. Situated on a cliff overlooking the sea, the site is near a hotel and a construction project, which may introduce influences from construction activities, household activities, and vehicle emissions. The continuous field campaign spanned from 4 September to 20 December in 2021, covering three seasons: summer (4 September–12 October), autumn (13 October–1 December), and early winter (2–20 December). Seasonal classification in this study was based on the occurrence of synoptic events and abrupt changes in key meteorological parameters, including upper-level wind direction, sea-level pressure, and dew point temperature (Fig. S1), as detailed in our previous studies (Feng et al., 2023). The measurement site is generally affected by long-range regional transport of aged air masses from South and East China due to the Asian monsoon, as well as fresh emission plumes from Hong Kong and the Pearl River Delta region (Ding et al., 2013).

A PTR-ToF-MS (Ionicon Analytik GmbH, Innsbruck, Austria) with H₃O⁺ as the primary reaction ion was used to measure the gaseous VOC and OVOC species with high time resolution of 10 s during the whole field campaign. Ambient air was drawn from a $\mathrm{1}/\mathrm{4}$-inch (6.35 mm) stainless-steel sampling manifold with an inert silicon-based coating at a flow rate of 5 L min⁻¹, and a subsample of filtered air via a $\mathrm{1}/\mathrm{16}$-inch (approximately 1.59 mm) polyetheretherketone (PEEK) tube was directed to the PTR at a frow rate of 100 mL min⁻¹. A polytetrafluoroethylene (PTFE) membrane particle filter was installed to prevent particulate matter, dust and debris from entering the instrument. To minimize potential cumulative adsorption effects, the filter was replaced frequently throughout the campaign. The sampling inlet was maintained at 80 °C throughout the measurements to mitigate humidity-related effects, reduce adsorption losses, and ensure gas-phase stability of target compounds prior to ionization and detection. It should be note that this heating may unintentionally promote the volatilization of some organic compounds from aerosols, thus causing positive artifacts. The instrument operated under optimized conditions: drift tube temperature maintained at 80 °C, drift voltage at 520 V, and drift tube pressure at 2.8 mbar, achieving a field density ratio ($E/N$) of 98 Td (1 Td  = 10⁻¹⁷ V cm²). Lower $E/N$ ratios can lead to a higher proportion of primary ions forming water hydronium clusters (Holzinger et al., 2019), and thus the $E/N$ of 98 Td was selected to balance ion fragmentation and water cluster formation, which can effectively suppress water clusters formation, thereby minimize the strong humidity dependence of the target species (Yuan et al., 2017).

Automatic mass calibration was conducted every 100 s using the built-in Ionicon permeation unit (PerMasCal), which releases strong signals of $m/z$ 203.943 (C₆H₄I₂H⁺, fragment) and $m/z$ 330.848 (C₆H₄I₂H⁺). Background measurement and multi-point calibration were conducted periodically during the field campaign using a Liquid Calibration Unit (LCU, Ionicon) with pure nitrogen and multi-component VOC gas standards. 19 $\mathrm{VOC}/\mathrm{OVOC}$ standard gases were used for multi-point calibration, achieving linear correlation coefficients (R²) above 0.99 (Table S1). The limit of detection (LOD) for each species was defined as three times the standard deviation (3σ) of background signal (Zhou et al., 2019), ranged from approximately 0.009 to 0.094 ppbv (Table S1). Transmission correction was applied using a set of reference compounds, including benzene ($m/z$ 79.054), toluene ($m/z$ 93.070), m-xylene ($m/z$ 107.086), 1,2,4-trimethylbenzene ($m/z$ 121.101), dichlorobenzene ($m/z$ 146.976), and trichlorobenzene ($m/z$ 180.937). In total, 117 $\mathrm{VOC}/\mathrm{OVOC}$ species were identified and quantified by attributing the measured ion masses to the most plausible molecular contributors. This attribution was based on molecular mass, elemental combinations consisting of carbon, hydrogen, oxygen, and nitrogen atoms, reflecting plausible atmospheric molecules and functional groups, as well as prior studies utilizing high-resolution mass spectrometry (Yuan et al., 2017; Koss et al., 2018; Wu et al., 2020), as summarized in Table S2. For the remaining 98 $\mathrm{VOC}/\mathrm{OVOC}$ species lacking available calibration standards, concentrations were determined using an assumed proton transfer reaction rate coefficient of 2 × 10⁻⁹ cm³ s⁻¹, combined with mass-dependent transmission correction (Zhang et al., 2022). To reduce uncertainties for uncalibrated compounds, an approach developed by Sekimoto et al. (2017), which estimates proton-transfer rate constants based on molecular properties such as polarizability and dipole moment, provides a scientifically robust method that could be applied in future work.

However, we acknowledge the several caveats associated with the $\mathrm{VOC}/\mathrm{OVOC}$ quantification. First, PTR-ToF-MS is limited in its ability to differentiate between isomeric compounds, accurate quantification of individual compounds remains challenging. Most molecular formulas were therefore evenly distributed among potential isomers (e.g., phenols, nitrophenols), while specific formulas for aldehydes and ketones were identified based on prior studies using GC-PTR-ToF measurement. For example, for C₁₀H₁₆, given that α-pinene and β-pinene are typically the predominant contributors (Kim et al., 2009; Byron et al., 2022; Kammer et al., 2020), an equal 50:50 allocation between the two species was adopted as a modeling assumption for the apportionment. As important oxidation products, C₄H₆O was apportioned as 48 % methyl vinyl ketone (MVK), 19 % methacrolein (MACR), and 33 % crotonaldehyde for model simulations, based on previous studies using PTR-MS combined with GC pre-separation (Koss et al., 2018). Although we attempted to assign signals based on likely contributors informed by literature, this approach introduces uncertainties in the molecular-level identification due to variability in instrument sensitivity, resolution, ambient conditions, and sampling periods across studies. These factors can affect the observed chemical composition and relative contributions of individual species, thereby influencing the accuracy of signal attribution and subsequent model inputs. Second, the ions detected by PTR-ToF-MS can include fragmentation products or hydrated clusters, particularly for highly functionalized OVOCs, which may lead to over- or underestimation of specific compounds if not correctly interpreted. To minimize uncertainties arising from water cluster interferences, ions susceptible to such effects, such as C₄H₆H⁺ ($m/z$ 55.054), which overlaps with the H₃O⁺(H₂O)₂ cluster, were excluded from compound identification and subsequent analysis in this study. Regarding impacts from fragmentation, for example, C₅H₈ may be affected by fragment interferences from higher-carbon aldehydes and cycloalkanes (Coggon et al., 2024; Claflin et al., 2021; Yuan et al., 2017; Zhang et al., 2025), therefore, the attribution of C₅H₈ to isoprene follows the proportion of 63 % reported in previous studies employing PTR-MS coupled with GC pre-separation, which effectively minimizes interference from fragments of higher molecular compounds (Koss et al., 2018). A sensitivity test assuming all C₅H₈H⁺ signals as isoprene was conducted, and it leads to a 2.9 %–6.0 % increase in daytime O₃ and RO_x production rates compared to the above assumption. Given these limitations, our quantification of the 63 OVOCs measured by PTR should be considered as semi-quantitative for 55 uncalibrated species, and as high-confidence for the 8 calibrated species.

In addition, canister samples of VOCs were collected every 3 h from 09:00 to 18:00 LT in three seasons, and nonmethane hydrocarbons and alkyl nitrates were analyzed using gas chromatograph system equipped with mass spectrometry, flame ionization, and electron capture detectors (GC-MS/FID/ECD). PTR-based concentrations of isoprene, benzene, and toluene, assigned using fragmentation assumptions from previous online GC-PTR-ToF, were compared with canister measurements. The two datasets showed overall good agreement (R² = 0.70–0.86), although PTR reported slightly higher concentrations, with slope of 1.25, 1.51 and 1.63 for isoprene, toluene, and benzene, respectively. This comparison indicates that the derived PTR data are reliable within an acceptable uncertainty range. Trace gases including O₃, NO_x (nitric oxide (NO) and nitrogen dioxide (NO₂)) and carbon monoxide (CO) were measured by O₃ analyzer (Thermo Scientific, model 49i), NO_x analyzer (Ecotech Serinus 40) and CO analyzer (Thermo Scientific, model T300), respectively. The meteorological parameters including temperature (T), relative humidity (RH), wind speed (WS), and wind direction (WD) were recorded by a Weather Station. Detailed descriptions of the instruments are available in previous work (Hui et al., 2023; Sun et al., 2024).

2.2 Photochemistry Modeling

A zero-dimensional photochemical box model (the Framework for 0-D Atmospheric Modeling, F0AM v4.2.2) coupled with the Master Chemical Mechanism (MCM) v3.3.1 was applied to simulate the atmospheric photochemistry of observed species. The MCM v3.3.1 is a nearly explicit gas-phase chemical mechanism describing over 17000 reactions and 5800 primary, secondary, and radical species. The model simulation was constrained by observed hourly data of meteorological parameters (T, RH, and pressure), trace gases (O₃, NO, NO₂, and CO), 38 VOC species measured by GC-MS/FID/ECD (Table S3), and 88 $\mathrm{VOC}/\mathrm{OVOC}$ species measured by PTR-ToF-MS (Table S2). For the offline canister VOC samples measured by GC-MS/FID/ECD, daytime data from 09:00 to 18:00 were linearly interpolated to an hourly resolution for the model input (Yang et al., 2018), while nighttime concentrations of unmeasured C₂-C₁₀ hydrocarbons (excluding isoprene and monoterpenes) and alkyl nitrates were estimated using linear regression relationships with continuously measured hydrocarbons (e.g., C₃H₆, C₅H₁₀, C₆H₁₀) and nitrophenols obtained from PTR-ToF-MS measurements. The PTR measured species used in the linear regression calculation were selected based on their strong correlations with corresponding compounds in the canister data to ensure more reliable estimates. These approximations were primarily used to pre-run the model and ensure continuous modeling, and were not expected to significantly affect the daytime simulation results, since photochemical activity is minimal during nighttime, and most hydrocarbons are less reactive in the absence of sunlight (Chen et al., 2020). Photolysis frequencies within the model were calculated as the function of solar zenith angle (Wolfe et al., 2016). Observation-based simulations were performed for consecutive days with high O₃ concentrations during summer (8 September–2 October), autumn (12–30 November), and early winter (4–19 December). Three days from summer (11, 12, and 17 September) were selected as the case study of summer high-O₃ episode, where the maximum O₃ concentration exceeded 110 ppbv. The model was pre-run for four days to stabilize the concentrations of unconstrained species, with results from the 5th day used for further analysis.

Model performance was evaluated using the index of agreement (IOA), as illustrated in Eq. (1) (Huang et al., 2005), with values of 0.81–0.87 for the simulated O₃ across seasons in this study, comparable to previous studies (0.6–0.9), suggesting that the abundance and variation of O₃ were deemed reasonably reproduced (He et al., 2019; Liu et al., 2021; Wang et al., 2018b, 2017, 2015).

$$\begin{array}{}\text{(1)}& \mathrm{IOA}=\mathrm{1}-{\displaystyle \frac{{\sum }_{i=\mathrm{1}}^n{\left(O_i-S_i\right)}^{\mathrm{2}}}{{\sum }_{i=\mathrm{1}}^n{\left(|O_i-\stackrel{\mathrm{‾}}{O}\left|+|S_i-\stackrel{\mathrm{‾}}{O}\right|\right)}^{\mathrm{2}}}}\end{array}$$

Where O_i and S_i represent the measured and simulated O₃ concentration, respectively; $\stackrel{\mathrm{‾}}{O}$ represents the mean measured O₃ concentration; and n represents the number of samples. The index of IOA typically ranges from 0 to 1, with higher values indicating stronger alignment between simulation and observation.

Dominant photochemical production pathways (HO₂ + NO, RO₂ + NO) and destruction pathways (O₃ photolysis, O₃ + OH, O₃ + HO₂, VOCs + O₃, NO₂ + OH) of O₃ were determined using Eqs. (2)–(4) (Tan et al., 2019a; Wang et al., 2018a). The production rates of RO_x radicals (OH, HO₂ and RO₂ radicals) were also calculated, incorporating primary sources such as photolysis reactions (i.e., O₃ photolysis, OVOC photolysis, etc.), VOC reactions with O₃ and NO₃, as well as recycling processes (Xue et al., 2016; Wang et al., 2018a; Tan et al., 2019b).

To evaluate the model uncertainties associated with the presence of multiple isomers in PTR-ToF-MS measurements, we conducted a sensitivity analysis by estimating the lower and upper limits of RO_x radicals and O₃ production rates with possible isomers. For OVOCs measured by the PTR with multiple isomers, each molecular formula was assigned either to the isomer with the minimum or maximum photolysis frequencies and K_OH values among all plausible isomeric structures. Specifically, for OVOCs with isomers containing both aldehydes and ketones, the upper bound was defined by assigning OVOCs to aldehydes with the highest photolysis frequency, while the lower bound assumed ketones with the lowest photolysis frequency (e.g., C₄H₆O, C₄H₈O). For OVOCs whose isomers do not undergo photolysis but are susceptible to OH oxidation, the upper and lower bounds were determined based on OH reactivity, with the upper and lower bounds corresponding to the isomers with highest and lowest K_OH values, respectively (e.g., C₈H₁₀O, C₆H₈O₂). This approach accounts for the variability in chemical reactivity stemming from the unresolved isomer distribution and provides a range of potential impacts on atmospheric RO_x radicals generation and O₃ formation.

$$\begin{array}{}\text{(2)}& {\displaystyle }\begin{array}{rl}P\left({\mathrm{O}}_{\mathrm{3}}\right)& =k_{{\mathrm{HO}}_{\mathrm{2}}+\mathrm{NO}}\left[{\mathrm{HO}}_{\mathrm{2}}\right]\left[\mathrm{NO}\right]\\ & +\sum k_{{\mathrm{RO}}_{\mathrm{2}}+\mathrm{NO}}\left[{\mathrm{RO}}_{\mathrm{2}}\right]\left[\mathrm{NO}\right]\end{array}\text{(3)}& {\displaystyle }\begin{array}{rl}L\left({\mathrm{O}}_{\mathrm{3}}\right)& =k_{\mathrm{O}{(}^{\mathrm{1}}\mathrm{D})+{\mathrm{H}}_{\mathrm{2}}\mathrm{O}}\left[\mathrm{O}{(}^{\mathrm{1}}\mathrm{D})\right]\left[{\mathrm{H}}_{\mathrm{2}}\mathrm{O}\right]\\ & +k_{{\mathrm{O}}_{\mathrm{3}}+\mathrm{OH}}\left[{\mathrm{O}}_{\mathrm{3}}\right]\left[\mathrm{OH}\right]+k_{{\mathrm{O}}_{\mathrm{3}}+{\mathrm{HO}}_{\mathrm{2}}}\left[{\mathrm{O}}_{\mathrm{3}}\right]\left[{\mathrm{HO}}_{\mathrm{2}}\right]\\ & +\sum k_{{\mathrm{O}}_{\mathrm{3}}+\mathrm{VOCs}}\left[{\mathrm{O}}_{\mathrm{3}}\right]\left[\mathrm{VOCs}\right]\\ & +k_{{\mathrm{NO}}_{\mathrm{2}}+\mathrm{OH}}\left[{\mathrm{NO}}_{\mathrm{2}}\right]\left[\mathrm{OH}\right]\end{array}\text{(4)}& {\displaystyle }\mathrm{Net}P\left({\mathrm{O}}_{\mathrm{3}}\right)=P\left({\mathrm{O}}_{\mathrm{3}}\right)-L\left({\mathrm{O}}_{\mathrm{3}}\right)\end{array}$$

Where P(O₃), L(O₃), and Net P(O₃) represents the production rate, loss rate, and net production rate of O₃, respectively. The O₃ photolysis was represented as the reactions of O(¹D) and H₂O (Shen et al., 2021). VOCs here included both constrained VOCs and model simulated VOCs. The constants (k) represent the rate coefficients of each reaction.

The O₃-precursors relationship was characterized using the relative incremental reactivity (RIR) method, calculated by Eq. (5) (Liu et al., 2021). O₃ formation regime was further characterized by O₃ isopleth method, which was derived by scaling precursor concentrations (10 %–200 % of original values) to simulate O₃ concentrations under varying VOCs and NO_x levels (Tan et al., 2018).

$$\begin{array}{}\text{(5)}& \mathrm{RIR}\left(X\right)={\displaystyle \frac{\left[P_{{\mathrm{O}}_{\mathrm{3}}\text{–}\mathrm{NO}}\left(X\right)-P_{{\mathrm{O}}_{\mathrm{3}}\text{–}\mathrm{NO}}(X-\mathrm{\Delta }X)\right]/P_{{\mathrm{O}}_{\mathrm{3}}\text{–}\mathrm{NO}}\left(X\right)}{\frac{\mathrm{\Delta }S\left(X\right)}{S\left(X\right)}}}\end{array}$$

Where RIR represents relative incremental reactivity; X represents specific O₃ precursor (i.e., VOCs, NO_x); S(X) represents observed concentration of precursor X (ppbv); ΔS(X) represents hypothetical change of the concentration of precursor X; $P_{{\mathrm{O}}_{\mathrm{3}}\text{–}\mathrm{NO}}\left(X\right)$ represents net O₃ production in a base run with original observed precursor concentrations, while $P_{{\mathrm{O}}_{\mathrm{3}}-\mathrm{NO}}(X-\mathrm{\Delta }X)$ represents the net O₃ production in a second run with a hypothetical change ΔS(X) of 10 % in this study. The net O₃ production was calculated by Eq. (4). A larger positive RIR value indicates higher sensitivity of O₃ production to this precursor, implying that reducing emissions of this precursor would more effectively suppress O₃ formation. Conversely, a negative RIR value suggests that emission reductions of this precursor could paradoxically increase O₃ production (Wang et al., 2018b).

3 Results and discussion

3.1 Overview of the observations

During the campaign, 117 $\mathrm{VOC}/\mathrm{OVOC}$ species were continuously measured using the PTR-ToF-MS, including 2 biogenic VOCs (BVOCs), 24 anthropogenic VOCs (AVOCs, comprising alkenes, cycloalkanes, and aromatics), 63 OVOCs (categorized as C_xH_yO_1–3), and 28 nitrogen/sulfur containing VOCs ($\mathrm{N}/\mathrm{S}$-containing VOCs). BVOCs here refer specifically to C₅H₈ and C₁₀H₁₆, which primarily represent isoprene and monoterpenes (e.g., pinenes, limonene, camphene, etc.), respectively. It should be noted that C₅H₈ may be affected by fragment interferences from higher-carbon aldehydes and cycloalkanes (Coggon et al., 2024; Claflin et al., 2021; Yuan et al., 2017; Zhang et al., 2025), which may potentially lead to an overestimation of BVOCs, particularly isoprene. The time series of different VOC groups, meteorological parameters, and trace gases are shown in Fig. S2. The campaign witnessed twenty O₃ episode days (maximum O₃ value > 80 ppbv) across three seasons, with three extreme episodes exceeding 110 ppbv (defined as high-O₃ episodes) in summer. The measured $\mathrm{VOCs}/\mathrm{OVOCs}$ showed the highest total concentration in early winter (47.84 ppbv), followed by autumn (44.26 ppbv) and summer (28.83 ppbv), which is consistent with the O₃ seasonal trend (Fig. S5). Furthermore, the total $\mathrm{VOCs}/\mathrm{OVOCs}$ concentration was much higher during O₃ episode days and reached 60.96 ppbv during summer high-O₃ episode days, with 76 % contribution from OVOCs, emphasizing their pivotal role in O₃ production. OVOCs, AVOCs, and $\mathrm{N}/\mathrm{S}$-containing VOCs increased progressively from summer to early winter, with the most pronounced rise observed between summer and autumn (Fig. 1a). In contrast, BVOCs displayed an opposite seasonal pattern, with concentrations peaking in summer. OVOCs were the dominant group across all seasons, accounting for 72 %–77 % of the total concentration, with C_xH_yO₁ and C_xH_yO₂ accounting for 51 %–53 % and 17 %–24 %, respectively (Fig. 1b). CH₄O (methanol) was the most abundant OVOC species, with average concentration ranging from 5.98–10.10 ppbv across seasons, followed by C₂H₄O₂ (1.91–5.75 ppbv), C₃H₆O (4.22-5.67 ppbv) and C₂H₄O (1.85–3.86 ppbv), which primarily correspond to acetic acid, acetone and acetaldehyde, respectively, as shown in Fig. S3. A statistical summary of $\mathrm{VOC}/\mathrm{OVOC}$ concentrations for each season is provided in Table S2.

The diurnal variations of OVOC subgroups across three seasons are shown in Fig. 1c. C_xH_yO_1–3 groups displayed similar diurnal patterns in different seasons, characterized by pronounced daytime enhancements, particularly for species such as C₂H₄O (acetaldehyde), C₃H₆O (acetone), C₄H₆O (MVK and MACR) and C₂H₄O₂ (acetic acid) (Fig. S4). These species increased at about 07:00 local time (LT), peaked during 12:00–17:00 LT in the afternoon, and then gradually decreased, which was aligned closely with the diurnal patterns of O₃ (Fig. S5), indicating their likely formation through photochemical reactions. Notably, C₄H₆O (MVK and MACR), the key oxidation products of isoprene (Li et al., 2021), showed pronounced photochemical daytime peaks but significantly higher concentrations in summer compared to autumn and early winter, consistent with the seasonal trend of its precursor (Fig. S6), which was different from other OVOCs formed by anthropogenic precursors. CH₄O (methanol) exhibited clear daytime enhancements in summer and autumn but showed no distinct diurnal pattern in early winter (Fig. S4). Instead, it maintained relatively high levels throughout the day and night in early winter, likely reflecting the larger contributions from regional background sources, such as anthropogenic emissions and background transport (Brito et al., 2015; Huang et al., 2019). In addition to precursor availability and photochemical activity, diurnal variations in boundary layer height may also influence OVOC concentrations by modulating vertical mixing and accumulation processes. These findings underscore the significance of OVOCs in atmospheric chemistry, given their abundance and complex roles in photochemical reactions.

Figure 1 (a) The concentrations of different VOC groups in summer, autumn and early winter. The box plots show mean values (square), median (line within the box), interquartile range (IQR, 25 %–75 %), and whiskers extending to ±1.5 × IQR. (b) The contributions of different VOC groups to total concentration in summer (inner), autumn (middle), and early winter (outer). (c) Diurnal variations of OVOC subgroups across three seasons.

3.2 O₃-precursor relationships

To further evaluate the contribution of various VOCs to O₃ formation, the RIR values of key O₃ precursors (including BVOCs, AVOCs, OVOCs, and NO_x) were calculated for different seasons based on the model simulations. It should be noted that the subgroups of OVOCs and AVOCs analyzed here differ slightly from those in Sect. 3.1, due to the limitations that MCM does not include all mechanisms for all observed OVOC species. Additionally, C₂–C₁₀ hydrocarbons measure by GC-MS/FID/ECD were also included in the subgroup of AVOCs for this analysis. The species included in RIR calculation comprised 3 BVOC species, 45 AVOC species and 63 OVOC species, with detailed information summarized in Table S4.

The RIR values for all VOCs subgroups and NO_x were positive across three seasons (Fig. 2a), indicating a transition regime of O₃ formation in the study region and that reduction in VOCs and/or NO_x would lead to decreases in O₃ levels. This result was different from previous studies which reported dominantly VOC-limited regime of O₃ formation in Hong Kong (Liu et al., 2021; Zhang et al., 2007; Cheng et al., 2010; Guo et al., 2013). In summer, NO_x exhibited the highest RIR value (0.65), followed by BVOCs (0.21) and OVOCs (0.16). Similar trend was observed during high-O₃ episode days (Fig. S7), indicating that summer O₃ formation is more sensitive to NO_x. By contrast, in autumn and early winter, OVOCs exhibited the highest RIR values (0.42–0.48), followed by NO_x (0.25–0.35) and BVOCs (0.14–0.16), indicating that O₃ formation in these seasons is more sensitive to OVOCs. It should be noted that current photochemical models typically represent monoterpenes using only α- and β-pinenes, neglecting some highly reactive species such as limonene. Moreover, gaps in the MCM, such as the absence of certain highly reactive monoterpenes and associated oxidation pathways, may further introduce uncertainties in assessing the role of BVOCs in atmospheric photochemistry.

The dramatic seasonal increase in the RIR value of OVOCs, coupled with a decline in those of NO_x highlights the need for different control strategies to effectively reduce O₃ levels depending on the seasons. Previous RIR studies of O₃ formation have primarily focused on AVOCs, BVOCs and NO_x, with limited consideration of OVOCs (Wang et al., 2018b; Tan et al., 2018; Wang et al., 2022b; Yu et al., 2020; Zhang et al., 2008; Lin et al., 2020; Zhao et al., 2020; Guo et al., 2022). Recent studies, however, have reported relatively high RIR values for OVOCs when they are included in the simulations, although these findings are confined to a narrow subset of OVOCs, mainly short-chain carbonyl compounds, based on low-resolution offline measurements (Shen et al., 2021; Yang et al., 2018; Wang et al., 2024; Feng et al., 2023). This study integrates a much broader spectrum of OVOCs (including aldehydes, ketones, organic acids, alcohols, phenolic compounds, etc.) supported by high-resolution measurement into the observation-constrained modeling framework. This improved chemical comprehensiveness allows for a more robust characterization of OVOCs reactivity, particularly their contributions to radical production and O₃ sensitivity. Nevertheless, due to the inherent limitations of PTR-ToF-MS, accurate quantification of isomers with distinct chemical reactivities remains challenging, introducing some uncertainties in atmospheric photochemical modeling.

In addition, we have further examined the diurnal variation of RIR values for O₃ precursors in different seasons, as shown in Fig. 2b. Significantly higher positive RIR values were observed for NO_x than other precursors in the afternoon of summer and during the episodes (Fig. S8), indicating a consistently strong sensitivity of O₃ production to NO_x. In autumn and early winter, OVOCs exhibited higher positive RIR values during the morning (09:00–11:00) and midday (11:00–14:00), underscoring their prominent role in O₃ formation. The majority of NO_x RIR values were positive except for the negative values in the morning (09:00–11:00), reflecting a shift from a VOC-limited regime in the morning to a transitional regime at the noon and in the afternoon. By late afternoon, the O₃ formation regime became increasingly sensitive to NO_x. Notably, in autumn, RIR values of NO_x were comparable or even surpassed these of OVOCs in the afternoon. This variation is likely attributed to fresh NO_x emissions from vehicle and/or household combustion activities in the morning, which decrease in the afternoon due to photochemical consumption, atmospheric diffusion, and dry deposition (Liu et al., 2021). These findings indicate a clear diurnal transformation pattern in O₃ formation regimes during the photochemical active seasons (autumn and early winter) in Hong Kong. Specifically, the regime transitions from VOC-limited in the morning to a transitional regime with higher sensitivity to OVOCs at midday, and then to a general transitional regime with sensitivity to both OVOCs and NO_x in the afternoon, and becomes increasingly NO_x sensitive by late afternoon.

Figure 2 (a) Average RIR values of O₃ precursors (BVOCs, AVOCs, OVOCs, and NO_x) during summer, autumn and early winter. (b) Diurnal patterns of RIR values for O₃ precursors across summer, autumn and early winter.

3.3 O₃ formation mechanism and radical budget

The main pathways of daytime O₃ production and destruction in three seasons were further explored using the photochemical box model, as shown in Fig. S9. The daytime average net O₃ production rate (P_net) was 5.8, 6.1, and 6.4 ppbv h⁻¹ in summer, autumn, and early winter, respectively, much lower than that during summer high-O₃ episode (13.1 ppbv h⁻¹, Fig. S9b). These trends were consistent with the diurnal patterns of observed O₃ concentrations (Fig. S10). Daytime O₃ production (P(O₃)) was dominated driven by the reactions of HO₂ + NO and RO₂ + NO, contributing 43.8 %–53.0 % and 47.0 %–56.2 %, respectively. Among the diverse RO₂ + NO reactions (involving over 1000 different RO₂ radicals), the top 10 pathways contributed 51.1 %–54.3 % of the total production rates (Fig. S11). CH₃O₂ + NO was the dominant pathway, accounting for 13.5 %–19.3 % of O₃ production. The CH₃O₂ radical could be generated through various reactions, including the photolysis of OVOCs (e.g., acetaldehyde, acetone) and VOCs reactions with OH radical (e.g., acetic acid), etc. Additionally, reactions of C₂ radicals with NO contributed significantly (10.4 %–18.5 %), with CH₃CO₃ + NO being the most prominent (10.4 %–15.2 %). Seasonal variations in RO₂ + NO pathways were evident. RO₂ radicals from BVOCs (such as isoprene and pinenes) contributed more to summer (20.7 %) than autumn and early winter (6.3 %–6.9 %), reflecting the seasonal patterns of vegetation emissions. Conversely, RO₂ radicals derived from anthropogenic sources, such as aromatics, played a larger role in autumn and early winter (8.6 %–9.1 %). Daytime O₃ destruction also exhibited seasonal differences. In autumn and early winter, the dominant loss pathway was OH + NO₂ (51.1 %–57.0 %), followed by VOCs + O₃ (17.9 %–23.2 %) and O₃ + HO₂ (12.6 %–13.8 %). In summer, however, VOCs + O₃ accounted for a much higher fraction of O₃ loss (43.3 %) compared to other seasons, especially the reactions with BVOCs including α-pinene and isoprene.

The OH, HO₂, and RO₂ radicals play important roles in the initiation and propagation of atmospheric photochemical reactions (Wang et al., 2022b). The daytime production budgets of these radicals were analyzed across three seasons, as shown in Fig. S12, with detailed contributions of each pathway presented in Fig. 3. Daytime OH and HO₂ radical production was highest in early winter, with lower values in summer and autumn. Radical recycling via HO₂ + NO was the dominant source of OH production (86.4 %–94.4 %) across three seasons, while O₃ photolysis (via O(¹D) + H₂O) served as the main primary source for OH production, accounting for around 60 % of primary OH production rate. In addition, the photolysis reactions of OVOCs, HONO and VOCs + O₃ contributed 7.6 %–18.3 %, 2.7 %–20.4 %, and 9.3 %–15.6 % of primary OH production rates, respectively. Similarly, the largest source of HO₂ radicals was radical recycling through RO₂ + NO reactions, contributing 42.5 %–50.2 %. OVOC photolysis accounted for 20.3 %–21.0 % to total HO₂ production rates, but dominated the primary HO₂ production in three seasons (95.2 %–98.6 %). Moreover, the reactions of OVOCs + OH (15.4 %–17.7 %) and CO + OH (10.2 %–18.7 %) also played significantly roles in total HO₂ production. The dominant source of daytime RO₂ production was the reaction of VOCs + OH, accounting for 50.8 %–61.1 %, with OVOCs + OH contributing the majority proportion (41.2 %–51.3 %). Seasonal variations in OVOCs + OH contributions aligned with observed OVOC concentrations, being highest in early winter, followed by autumn and summer. Additional RO₂ sources included RO decomposition (13.9 %–21.3 %), OVOC photolysis (10.0 %–14.2 %), and VOCs + NO₃ (9.6 %–14.1 %). OVOC photolysis was also an important primary source of daytime RO₂ production (36.0 %–57.0 %). During summer high-O₃ episodes, daytime RO_x radical production rates were significantly higher, and O₃ photolysis and VOCs + O₃ reactions contributed more substantially to RO_x production (Fig. S13). Furthermore, the results highlight the importance of VOCs + NO₃, especially OVOCs + NO₃, as a source of RO₂ radicals in the daytime photochemistry, particularly in polluted atmosphere, consistent with prior observations at an urban site in Hong Kong (Xue et al., 2016). In total, OVOCs played a significant role in the formation of HO₂ and RO₂ radicals across all three seasons. OVOC-related reactions, including OVOC photolysis, OVOCs + OH oxidation, and OVOCs + NO₃, accounted for 36.4 %–38.5 % and 59.1 %–73.4 % of daytime HO₂ and RO₂ production, respectively. These results emphasize the critical importance of OVOCs in sustaining radical cycling and driving photochemical O₃ formation, especially in urban and semi-urban atmospheric environments.

Figure 3 Contributions of key pathways for total daytime production rates (left) and primary daytime production rates (right) of (a) OH, (b) HO₂ and (c) RO₂ radicals across seasons. OVOC-related reactions including OVOC photolysis (OVOCs + hv), OVOCs oxidation by OH radicals (OVOCs + OH), and OVOCs oxidation by NO₃ (OVOCs + NO₃) are shaded by oblique lines. The percentage in red represents the contribution of OVOC-related reactions for overall (left) and primary (right) daytime production rates of HO₂ and RO₂ radicals.

3.4 Importance of OVOCs in O₃ and radical formation

To better quantify the critical roles of OVOCs in photochemical O₃ and radical formation, a sensitivity simulation was conducted without constraining the observed OVOC species in the model. The comparison of observed and simulated O₃ concentrations under scenarios with and without OVOCs constraints across three seasons is shown in Fig. S14. Incorporating a broader range of OVOCs improved the simulation of O₃, particularly in autumn and early winter, where daytime concentrations were underestimated by 26.5 % and 35.7 %, respectively, without OVOCs constraints. In contrast, the discrepancy was minimal in summer, likely due to the dominant role of NO_x in O₃ formation and elevated daytime NO levels during high-O₃ episodes. It should be noted that the model considers only in situ photochemical processes and does not include influences such as regional transport. As a result, discrepancies between observed and simulated O₃ remain, especially in autumn and early winter, when periods typically influenced by the Asian monsoon, and during nighttime when photochemical activity is minimal. Given the higher observed O₃ levels in early winter and the substantial underestimation without OVOC constraints, we conducted a focused evaluation of model performance for this period.

Figure 4 Model simulated daytime O₃ production and loss rates of main pathways in early winter (a) with and (b) without observed OVOCs constraints. The black solid line represents the net O₃ production rate (P_net). (c) The daytime reaction of the top two pathways of RO₂ + NO (CH₃O₂ + NO, CH₃CO₃ + NO) towards daytime O₃ formation in early winter with and without OVOCs constraints. (d) The daytime P_net with and without observed OVOCs constraints in early winter. The orange and dark red lines represent the scenarios with the minimum and maximum OVOC contributions to P_net, respectively.

As shown in Fig. 4a and 4b, simulated daytime P(O₃) and P_net without OVOCs constraints in early winter decreased by 44.0 % and 45.1 %, respectively, consistent with the underestimation of O₃ concentrations during the same period. The reduction in RO₂ + NO reaction rates (45.6 %) was slightly larger than that for HO₂ + NO (42.6 %), with the most substantial decreases observed in CH₃O₂ + NO (61.4 %) and CH₃CO₃ + NO (58.6 %) pathways (Fig. 4c). These reductions were primarily attributed to the underestimation of radical precursors without OVOCs constraints. Moreover, the existence of multiple OVOC isomers detected by PTR-ToF-MS, introduces additional uncertainties in quantifying daytime O₃ production. Sensitivity analysis revealed that P_net underestimations without OVOCs constraints ranged from 43.4 % to 52.1 %, depending on the assumed photolysis frequencies and K_OH values of potential isomers (Fig. 4d). These results highlight the critical role of OVOCs in O₃ formation and the potentially large uncertainties in O₃ modeling when their contributions are not adequately represented. The discrepancies are significantly larger than those reported in previous studies (Wang et al., 2022a; Shen et al., 2021), and likely arise from our inclusion of a broader range of OVOCs beyond commonly considered carbonyls, enhancing the chemical completeness of the model and improving the simulation accuracy.

The critical role of OVOCs in modulating atmospheric radical budgets was further quantified through scenario-based simulations. As illustrated in Fig. 5a, without OVOCs constraints led to significant reductions in the production rates of OH (40.2 %–47.4 %), HO₂ (43.2 %–51.0 %) and RO₂ (45.8 %–57.1 %) radicals, with ranges reflecting the minimum and maximum assumptions. These underestimations were amplified in OVOC-related reactions, where HO₂ and RO₂ production were reduced by 58.1 %–65.5 % and 62.0 %–72.2 %, respectively, underscoring the significance of OVOCs in radical cycling. These reductions were primarily attributed to the underestimation of OVOC photolysis (37.4 %–64.5 %) and OVOCs + OH reactions (60.0 %–71.0 %) (Fig. 5b). Furthermore, prominently large underestimations (up to 93.6 %–95.0 %) occurred in RO₂ production from OVOCs + NO₃ reactions (Fig. 5b), revealing a critical gap in modeling daytime NO₃ oxidation of OVOCs. These discrepancies were largely due to the underestimation of multiple OVOC species, for example, methanol, acetaldehyde, and acetone were underestimated by 73 %–99 % in early winter simulations without OVOC constraints (Table S5). Similar underestimation (10 %–100 %) of simulated OVOCs have been reported in previous studies (Wang et al., 2022a). Although some carbonyl compounds are commonly included in the modeling of O₃ and radical formation (Zhao et al., 2020; Yu et al., 2020; Chen et al., 2023; Han et al., 2023; Huang et al., 2020a; Liu et al., 2019; Liu et al., 2022), our results highlight that many other photoreactive OVOCs remain overlooked, contributing to large uncertainties in radical simulations. A sensitivity test including only three commonly measured OVOCs (methanol, acetaldehyde, acetone) (Whalley et al., 2021; Whalley et al., 2018; Yang et al., 2018; Feng et al., 2023; Shen et al., 2021) increased daytime O₃ by 14.1 % in autumn and 17.6 % in early winter compared to the case without any OVOCs constraints, yet still left 16.2 % and 24.4 % of the discrepancies unexplained (Fig. S14). In early winter, these three OVOCs increased daytime O₃ and RO_x production rates by 14.6 %–22.9 %, but a remaining gap of 32.8 %–36.0 % indicates that simple OVOCs alone cannot resolve the underestimation. Several less commonly measured yet reactive OVOC, such as butanedione, glycolaldehyde, cresols, and hydroxyacetone, were found to contribute significantly to daytime O₃ and radical production via photolysis and OH oxidation. Together, these findings suggest that missing or unresolved primary and secondary OVOC sources, as well as incomplete representation of OVOC chemistry in current mechanisms (e.g., MCM) (Karl et al., 2018; Mo et al., 2016; Bloss et al., 2005) are major contributors to model bias. However, residual uncertainties remain due to limitations in isomer-specific quantification, emphasizing the need for improved measurement techniques to better constrain OVOC impacts in atmospheric models.

Figure 5 (a) Model simulated daytime production rates of OH, HO₂ and RO₂ radicals of main pathways in early winter with (left) and without (right) observed OVOCs constraints. The orange and dark red dash lines represent daytime production rates with minimum and maximum scenarios, respectively. (b) The reaction rates of OVOC photolysis (OVOCs + hv), OH oxidation of OVOCs (OVOCs + OH), and NO₃ oxidation of OVOCs (OVOCs + NO₃) towards daytime production of HO₂ and RO₂ radicals in early winter. The dash lines represent the scenario without OVOCs constraints. The areas represent daytime production rates with minimum and maximum scenarios.

3.5 Implication for O₃ pollution control strategies

Given the critical role of OVOCs in O₃ production, EKMA O₃ isopleths were derived to evaluate the dependence of daytime O₃ production on VOCs and NO_x variations. The analysis revealed a critical difference between the two scenarios: suburban Hong Kong was classified in the transition regime with OVOCs constraints (Fig. 6a), whereas it shifted to a VOC-limited regime without OVOCs constraints (Fig. 6b). This highlights the importance of including OVOCs in modeling efforts, as their exclusion may lead to potentially misleading strategies for O₃ pollution control. Figure 6c further illustrates changes in daytime O₃ production in response to VOCs or NO_x reductions (0 % to 90 %) under the two scenarios. With OVOCs constraints, O₃ concentration would decrease with reductions in either VOCs or NO_x, but more strongly with VOCs, consistent with the transition regime. In contrast, without OVOCs constraints, O₃ concentration would initially increase with NO_x reduction of 0 %–50 %, before declining at higher reductions. Similar patterns were observed in changes to daytime production rates of O₃ and RO_x radicals (Fig. S15). These results demonstrate that neglecting OVOCs exaggerates the VOC-limited degree and overestimate the effectiveness of VOCs reduction on O₃ reduction, while underestimating the potential benefits of NO_x control. This could lead to suboptimal or ineffective O₃ mitigation strategies.

For regions in the transition regime, such as suburban Hong Kong, simultaneous reductions in both VOCs and NO_x are necessary for effective O₃ control. As shown in Fig. 6d, when VOC reduction is between 0 % and 40 %, any reduction in NO_x would result in corresponding O₃ reductions. However, at VOC reductions (60 %–90 %), minor NO_x reductions would paradoxically increase O₃ levels unless NO_x reduction is sufficiently large to outweigh VOC reductions. Notably, when NO_x reduction exceeds 90 %, O₃ concentration falls below 25 ppbv, regardless of VOC levels. While such drastic emission reductions are challenging, a dual-control strategy with reduction of VOCs by 0 %–40 % and minimizing NO_x emissions emerges as both feasible and effective, avoiding unintended increases in O₃ levels. It should be noted that due to the absence of a robust mechanism to represent the nonlinear formation and diverse sources of OVOCs, we employed a simplified scaling approach based on precursor VOCs in O₃ isopleth analysis. Despite inherent uncertainties, this provides a practical approximation for assessing OVOC impacts on O₃ formation under varying emission scenarios.

Although this study focused on a representative suburban region in Hong Kong, the methodology and findings are broadly applicable to other urban and suburban regions with diverse emission profiles and photochemical regimes. In such environments, excluding OVOCs can lead to misclassification of O₃ formation regimes, resulting in ineffective or counterproductive control strategies. These insights underscore the critical need for OVOCs-inclusive modeling frameworks to guide effective and science-based air quality management.

Figure 6 Isopleth diagram for average daytime O₃ production depending on NO_x and VOCs changes in early winter (a) with and (b) without observed OVOCs constraints. The “pink star” represents the base scenario. (c) Changes in daytime O₃ production with VOCs or NO_x reductions from 0 % to 90 % with and without observed OVOCs constraints. The daytime average O₃ concentrations were normalized to the corresponding values in the base run. The grey dash line represents the normalized O₃ concentration of 1.0, indicating no O₃ changing. (d) The response of daytime O₃ concentration to VOCs and NO_x under different reduction scenarios with observed OVOCs constraints.

4 Conclusions

This study integrated intensive field measurements with observation-based photochemical modeling to investigate the role of OVOCs in O₃ and radical chemistry at a coastal suburban site in subtropical Hong Kong. High-resolution measurements using PTR-ToF-MS identified and quantified 117 $\mathrm{VOC}/\mathrm{OVOC}$ species, among which 63 OVOCs contributed the majority (72 %–77 %) of total VOC concentrations across three seasons in 2021. RIR analysis revealed a transitional O₃ formation regime in this suburban region, with heightened sensitivity to OVOCs, especially in autumn and early winter. Notably, O₃-precursor relationship also showed diurnal variations, transitioning from a VOC-limited regime in the morning to a transitional regime during midday and afternoon, underscoring the dynamic nature of O₃ chemistry.

Photochemical modeling demonstrated that OVOC-related reactions, including photolysis and oxidation by OH and NO₃ radicals, contributed substantially to radical formation, accounting for 36.4 %–38.5 % of daytime HO₂ and 59.1 %–73.4 % of RO₂ radical production. Importantly, simulations without comprehensive OVOC constraints would significantly underestimate daytime O₃ and RO_x production rates by 41 %–48 % and incorrectly diagnosed the O₃ chemical regime. Such misclassification may lead to misguided control strategies. Compared with previous studies that only focused on a limited set of carbonyls using offline techniques, this study expands the chemical scope by including a broader suite of OVOCs through high-resolution, real-time measurements, providing a more complete assessment of OVOC-driven radical and O₃ formation. The mechanistic insights and modeling framework developed here offer practical value for diagnosing O₃ formation sensitivity and designing more effective air quality management strategies in chemically complex environments globally.

Overall, these findings underscore the critical role of OVOCs in shaping atmospheric oxidation capacity and O₃ formation, and highlight the need for integrating high-resolution, chemically comprehensive OVOC measurements into photochemical models. Doing so will improve the accuracy of O₃ formation regimes classification, reduces uncertainties in radical budgets, and supports the development of targeted, science-based, and sustainable O₃ pollution control strategies at both regional and global scales.