The CMAQ model version 5.2, coupled with the SAPRC07tic chemical mechanism and the AERO6 aerosol module, was applied to simulate OVOC distributions in the YRD during 27 March–30 April (EP1) and 20 October–20 November (EP2) of 2019. Several model updates, including heterogeneous reactions of sulfur dioxide (SO₂), NO₂, glyoxal, and methylglyoxal, were implemented as described in a previous study (Mao et al., 2022). Two nested domains were configured, with the outer domain (d01) covering eastern China and the inner domain (d02) encompassing the YRD. These domains had horizontal resolutions of 12 km × 12 km (127 × 202 grid cells) and 4 km × 4 km (238 × 268 grid cells), respectively (Fig. S1 in the Supplement). Meteorological fields were generated using the Weather Research and Forecasting (WRF) model version 4.2.2, with initial and boundary conditions provided by the fifth generation ECMWF atmospheric reanalysis of the global climate (ERA5) dataset. Anthropogenic emissions in d01 were derived from the Multi-resolution Emission Inventory for China (MEIC) version 1.4 (Geng et al., 2024) for mainland China and the Regional Emission inventory in ASia (REAS) version 3.2.1 (Kurokawa and Ohara, 2020) for other Asian countries and regions. For d02, the 2019 YRD emission inventory incorporating updated VOC profiles (2019 YRD inventory) was applied to the YRD, while the MEICv1.4 was used for the remaining regions. Biogenic emissions were generated using the Model of Emissions of Gases and Aerosols from Nature (MEGAN) version 2.1. Open biomass burning emissions were based on the Fire INventory from NCAR (FINN) version 2.5 (Wiedinmyer et al., 2023). The spatial and temporal allocations of anthropogenic and open biomass burning emissions followed previous work (Hu et al., 2016; An et al., 2021). Notably, the temporal allocation did not account for temporary emission control measures during the China International Import Expo 2019 (CIIE 2019, 27 October–10 November 2019) (Wang et al., 2025). To avoid potential biases and to better represent typical emission conditions in the YRD, model simulation results for this period were excluded from the analysis. In addition, the first five days of each simulation were designated as spin-up and were not considered in subsequent analysis. A complete list of all OVOCs and their precursors used in the model is provided in Table S1 in the Supplement.

The 2019 YRD inventory was developed based on the 2017 version (An et al., 2021), with updated activity data for each source sector. Key improvements included refined VOC speciation, particularly of OVOCs, for major source categories such as transportation (gasoline and diesel vehicles), industrial processes, and residential biomass burning (Gao et al., 2022; Wang et al., 2022a). The VOC composition of emissions from diesel vehicles, industrial processes, and residential biomass burning was characterized based on 160 localized, source-resolved measurements conducted in China. Specifically, twenty in-use heavy-duty diesel vehicles from five major brands were tested, encompassing China VI (n=6), China V (n=10), and China IV (n=4) emission standards. For industrial emissions, a total of 84 samples were collected from priority sectors, including petrochemical industries, chemical raw material production, and other chemical production sectors such as plywood production, coking, pesticides production, ink production, and rubber production. For residential biomass burning, 23 samples of the combustion of four representative biomass fuels (wood, corncob, bean straw, and corn straw) and two common coal types (anthracite and briquette coal) were collected from the stack nozzles of household stoves. Details of the sampling protocols and analytical procedures can be found in a previous study (Gao et al., 2023). VOC profiles for other sources, such as gasoline vehicles, were based on published literature (Wang et al., 2022a; Huang et al., 2024). To develop representative source profiles, a two-step aggregation method was employed. First, sub-category average profiles were derived by averaging individual samples within each specific emission standard, industrial stage, and types of raw materials, products, and fuels. Second, the integrated source profiles were synthesized by weighting these sub-category profiles according to their corresponding total VOC emissions. This method aligns with the national technical guidelines for integrated air pollutant emission inventories (Ministry of Ecology and Environment, 2024).

As a result of these refinements, OVOCs accounted for 56.9 % of total VOC emissions from diesel vehicles (DVE), followed by 44.6 % from residential biomass burning (RBB) and 39.2 % from industrial processes (INP) (Table S2). Among all OVOCs, carbonyl species showed significant enhancements in these three source categories. Aldehyde contributions nearly doubled in DVE and INP, with a particularly large increase in HCHO. In contrast, the emissions of alcohols, such as methanol (MEOH) and ethanol (ETOH) from INP, decreased. Contributions from other OVOCs, including acetic acid (CCOOH) as well as phenols and cresols (CRES), also increased in INP and DVE. Although the overall OVOC fraction from gasoline vehicle emissions slightly decreased, the chemical composition exhibited a marked change, with increases in carbonyls and decreases in alcohols and esters. In this study, primary OVOCs are defined as those directly emitted from anthropogenic and biogenic sources, whereas secondary OVOCs refer to those formed via the photooxidation of VOC precursors.

A suite of pollutant and meteorological observations from ground monitoring stations (Fig. S1) was collected for model evaluation. High time-resolution measurements of 77 OVOCs, including 14 aldehydes and ketones, 23 organic acids and esters, 10 furan compounds, and 30 oxygenated aromatic compounds, were recorded at a 10 s interval using a Proton Transfer Reaction-Quadruple interface Time-of-Flight Mass Spectrometer (PTR-QiTOF) at the Shanghai Academy of Environmental Sciences (SAES). Species were identified by jointly applying the Tofware software package v3.2.3 (Tofwerk Inc.), the proton transfer reaction mass spectrometry (PTR-MS) spectral library (Pagonis et al., 2019), the PubChem database, and source-specific emission profiles reported in literature (Hatch et al., 2017; Koss et al., 2018; Stockwell et al., 2021; Tanzer-Gruener et al., 2022; Coggon et al., 2021). Sensitivities for species with authentic standards were determined through calibration using multi-gradient known concentrations of given VOCs. For identified species lacking standards, their theoretical sensitivities were estimated based on correlations with kinetic rate constants of VOCs (Sekimoto et al., 2017). The raw data were screened to remove outliers (values below background levels) and missing data, and then averaged to hourly means.

In addition, C₂–C₁₂ hydrocarbons, C₂–C₅ carbonyls, and C₁–C₄ alcohols were measured using an online gas chromatography system equipped with a mass spectrometer and a flame ionization detector (GC-MS/FID, TH-300, Wuhan Tianhong Instruments, China) at an hourly resolution. For major aromatic hydrocarbons and carbonyl compounds, good agreement between measurements by PTR-QiTOF and GC-MS/FID was observed (Gao et al., 2022). Formaldehyde and peroxyacetyl nitrate (PAN) were measured using a commercial Hantzsch monitor (AL4021, Aero-Laser GmbH, Germany) and an online gas chromatography-electron capture detector (PANs-1000, Focused Photonics, China), respectively. Details on the instrumentation, analytical procedures, and data quality assurance have been described in previous work (Gao et al., 2022; Liu et al., 2019; Du et al., 2025; Gao et al., 2023).

Hourly O₃ concentrations in 14 cities across the YRD were obtained from the China National Environmental Monitoring Center (https://air.cnemc.cn:18007; last access: 1 November 2024). Meteorological parameters, including temperature, relative humidity, wind speed, and wind direction, were obtained for four representative cities (Shanghai, Nanjing, Hefei, and Hangzhou) from the China Meteorological Administration (http://data.cma.cn/; last access: 1 November 2024). To evaluate model performance, statistical metrics such as Pearson correlation coefficient (R), mean bias (MB), mean error (ME), normalized mean bias (NMB), normalized mean error (NME), root mean square error (RMSE), and index of agreement (IOA) were calculated and compared against benchmark values where available (Tables S3, S4). Additional details are provided in Sect. S1 in the Supplement, and comparisons between model results and observations can be found in Figs. S2–S5.

Comparisons between observed and predicted OVOC concentrations at SAES are presented in Figs. 1 and S6, with the corresponding statistical metrics summarized in Table S5. Among all the observed OVOCs, alcohols emerged as the dominant species in spring (EP1), with ethanol exhibiting the highest median concentrations (6–12 ppb), which declined to 1.9–4.6 ppb in fall (EP2). The model underestimated ethanol during both episodes, with larger biases in EP1. This discrepancy may be attributed to uncertainties in the emission inventory. The monitoring site is located in a typical urban environment characterized by dense commercial and residential buildings and heavy traffic, with strong influences from transportation and solvent use emissions (Liu et al., 2019, 2021). One likely underestimated source of alcohols is the use of alcohol-containing solvent products such as windshield washer fluid, particularly from electric vehicles (Cliff et al., 2023). Considering the rapid growth in electric vehicle ownership in China in recent years, the relative importance of this alcohol source is expected to increase. Another potential source of alcohols is the use of household cleaning and personal care products, which may not be fully represented in the current emission inventory (Mo et al., 2021; Wang et al., 2024b).

Formic acid and acetic acid were the most abundant organic acids during both episodes, with the average concentrations of 0.15 (1.95) ppb and 1.88 (2.59) ppb in the EP1 (EP2) period, respectively. The model consistently underestimated both species, with larger biases in EP2 (Fig. S6 and Table S5). These discrepancies may be partially attributed to uncertainties in the emission inventory. During EP2, acetic acid exhibited strong correlations with biomass burning tracers, such as guaiacol and furan compounds and their derivatives (Pearson coefficients >0.80, p<0.001). In addition, acetic acid showed consistently high correlations with ethyl acetate, methyl ethyl ketone (MEK), and toluene (Pearson coefficients of 0.75–0.85, p<0.001) in both periods, species that are commonly associated with emissions from volatile chemical products (VCPs) (Wang et al., 2024b; McDonald et al., 2018). The underestimation of emissions from these sources likely contributes to the low bias in simulated acetic acid concentrations. However, MEK and toluene may also originate from other sources, such as fuel combustion, and their model performance varied between episodes, being slightly underestimated in EP1 but overestimated in EP2. Therefore, the underestimation of both formic and acetic acids may partially stem from insufficient representation of secondary production. Observations showed that formic acid exhibited moderate to strong correlations with its precursor methylglyoxal in both periods. Laboratory studies have demonstrated that both formic and acetic acids can be produced via aqueous-phase reactions of glyoxal and methylglyoxal (Yu et al., 2011; Zhang et al., 2021; Sui et al., 2017; Lim et al., 2013), with uptake modulated by interactions with inorganic aerosols (i.e., salting effects) (Waxman et al., 2015; Kampf et al., 2013). In the current model, these heterogeneous processes are represented by first-order reactions, with rate constants determined by fixed uptake coefficients (γ=0.016 during daytime and a constant loss rate of 3.33×10-4 s⁻¹ at night), and all products are lumped into a single surrogate species. The simplified treatment of heterogeneous chemistry likely leads to an underestimation of formic and acetic acids, particularly under high aerosol loadings in urban environments, where salting-out effects reduce their solubility in the aqueous phase (Babaei-Gharehbagh et al., 2025). Other sources of model bias may include missing secondary formation pathways, such as multiphase reactions of HCHO and photochemical aging of aerosols, and underestimated yields of formic and acetic acids from precursor VOCs (Yuan et al., 2015; Millet et al., 2015; Permar et al., 2023; Jiang et al., 2023; Cope et al., 2021; Franco et al., 2021; Malecha and Nizkorodov, 2016; Shen et al., 2026).

Most carbonyl species were well captured by the model. During the warm season (EP1), observed carbonyl concentrations exhibited pronounced diurnal variations with daytime maxima, primarily driven by photochemical production. In contrast, such diurnal patterns were less evident in EP2 (Fig. S6), likely due to reduced secondary formation under lower atmospheric oxidation capacity (Fig. S7). HCHO and acetone (ACET) were the most abundant aldehyde and ketone species, with average observed concentrations of 2.52 (2.58) ppb and 5.75 (1.89) ppb during EP1 (EP2), respectively. HCHO is a common oxidation product of VOCs in the troposphere and can also originate from emissions, including fuel and biomass combustion, industrial processes, and the evaporation of indoor building materials (Wang et al., 2022a; Xing et al., 2024; Liang and Yang, 2013; Permar et al., 2021). The model generally reproduced the observed HCHO concentrations, although slight underestimations persisted in both seasons (Table S5). Model performance for ACET varied seasonally, with good agreement in EP2 but significant underestimation in EP1. Since biogenic sources of acetone are important (Lyu et al., 2024; Hu et al., 2013), the larger discrepancy observed in the warm season may be partially due to the model's inadequate representation of urban green vegetation (Maison et al., 2024; Ma et al., 2022).

Notably, biases in OVOC concentrations are influenced by those of their precursors. For example, model biases in methacrolein (MACR) and methyl vinyl ketone (MVK) exhibited similar temporal patterns to those of isoprene (ISOP), reflecting their common origin via isoprene oxidation. The model failed to reproduce the peak noon concentration of isoprene, likely due to the underestimated biogenic emissions from urban green spaces. Overall, most key OVOC precursors, including reactive alkanes (ALK3/ALK4), aromatic hydrocarbons, and alkenes (e.g., ethylene and propylene), were well captured by the model. The simultaneous overestimation of reactive VOC precursors and underestimation of OVOCs also indicates that uncertainties in their photochemical mechanisms may contribute to the model biases in OVOC concentrations.

At the regional scale, average OVOC concentrations across the YRD ranged from 5 to 25 ppb in both periods, with elevated levels observed in the central and northern areas (Fig. 2a, d). HCHO, MEOH, ACET, and acetaldehyde (CCHO) were the dominant species (Fig. S8), with concentration ranges of 2–8, 2–4, 2–4, and 1–3 ppb, respectively. High levels of HCHO, ACET, and CCHO mainly occurred in areas strongly influenced by anthropogenic emissions, where MEOH was predominantly concentrated in regions affected by biogenic emissions in the southern and southwestern YRD (Fig. S9). The concentrations of ETOH, HCOOH, and CCOOH fell in ranges of 1.0–3.0, 0.1–0.4, and 0.5–2.0 ppb, respectively. Given the substantial underestimations of all three species, their actual concentrations are likely higher than the model predictions. In contrast to observations in Shanghai, the model predicted higher concentrations of most OVOCs in EP2 than in EP1 across the YRD, likely due to biases in the model representation of OVOCs (Table S5). The higher simulated MEOH concentrations in EP1 relative to EP2 may be partially attributed to stronger contributions from biogenic sources during the warm season (Fig. S9d, h).

On average, the model predicted that approximately 30 %–70 % of total OVOCs originated from emissions rather than from VOC oxidation production. The contributions of primary OVOCs showed significant seasonal and spatial variations. During the warm season, secondary OVOCs dominated across the entire domain, while elevated emission contributions were observed along the Yangtze River and in northern Jiangsu, northern Zhejiang, and Shanghai (Fig. 2b). In contrast, during the cold season, primary OVOCs played a larger role in Jiangsu and Anhui Provinces as well as in Shanghai and northern Zhejiang (Fig. 2e), consistent with our previous findings for October 2017 in the YRD (Li et al., 2022a).

CRES accounted for the highest OVOC emission rate across the YRD (22 %), followed by HCHO (16 %), ACET (14.4 %), CCOOH (12.5 %), and ETOH (9.5 %) (Table S6). High emission contributions to ambient CRES and CCOOH were also observed, followed by ACET (Fig. S10). Notably, the emission contributions of CCOOH may be biased due to uncertainties in emission inventories and chemical production in the model. Primary emissions accounted for approximately 20 %–60 % of ambient HCHO, with maxima reaching up to 76 % in EP1 and 87 % in EP2 in Jiangsu Province. The model also estimated a substantial fraction of biacetyl (BACL) originating from primary emissions (more than 60 %), particularly in Jiangsu and Anhui Provinces, as well as Shanghai.

The spatial distributions of OVOCs and ozone were closely aligned, with OVOC hotspots coinciding with regions of elevated daily maximum 8 h average ozone (MDA8 O₃) concentrations (Fig. 2c, f). Further analysis revealed that OVOC photooxidation constituted a significant source of HO₂ radicals in the YRD, serving as a key driver of NO-NO₂ cycling and thereby influencing ozone formation.

To investigate the relationship between OVOCs and O₃ formation at the monitoring site in Shanghai, a “clean period” (1–4 April) and a “pollution period” (5–7 April) were selected. During the clean period, observed MDA8 O₃ concentrations (81.4–139.4 µg m⁻³) complied with China's Ambient Air Quality Standards (GB 3095-2012) threshold of 160 µg m⁻³. In contrast, MDA8 O₃ levels during the pollution period ranged from 161.3 to 189.1 µg m⁻³, exceeding this standard. The model showed good agreement with observations, successfully capturing temporal variations of both MDA8 O₃ and OVOC concentrations during these periods, demonstrating its capability to represent OVOC sources and sinks as well as their role in ozone formation.

The ozone production rate (P(O₃), ppb h⁻¹), defined as the rate at which ozone is produced through the conversion of NO to NO₂ by peroxy radicals, was calculated as follows: 1P(O3)=k0[HO2][NO]+∑iki[RO2i][NO] where k0 and ki are the rate constants for the reactions of NO with HO₂ and RO₂ radicals (unit: molec.⁻¹ cm³ s⁻¹), respectively, and [HO₂], [RO₂], and [NO] denote their respective concentrations (unit: ppb). In RO₂ + NO reactions, alkyl nitrates can form as byproducts, with yields (α) depending on the specific RO₂ species. Accordingly, a NO₂ yield of (1-α) was applied to the corresponding RO₂ + NO reactions. The NO₂ molar yields for reactions producing alkyl nitrate byproducts in the current model are listed in Table S7. The analysis of P(O₃) and radical chemistry is focused on the daytime period (09:00–17:00 local time).

As shown in Fig. 3a and b, the daytime average P(O₃) during the clean period was approximately half that under the polluted conditions, at 5.89 and 12.2 ppb h⁻¹, respectively. HO₂-driven oxidation accounted for 53 %–61 % of daytime P(O₃) during the clean period, increasing slightly to 54 %–63 % under the polluted conditions. Ozone production via the HO₂ pathway peaked around noon, coinciding with the highest P(O₃), and declined gradually to approximately 50 % later in the day. These results indicate that the HO₂ pathway consistently dominates ozone production in Shanghai, regardless of pollution levels.

The dominant role of HO₂-driven oxidation in ozone production was further confirmed in other urban areas of the YRD. During both EP1 and EP2, HO₂ radicals contributed approximately 55 %–80 % of total ozone production in regions with elevated P(O3) (Figs. 4a, d and S11a, d). These regions, including Shanghai, are strongly affected by anthropogenic emissions, with substantial NO_x emission rates (approximately 0.4–2.0 mol s⁻¹; Fig. S9) and HO₂ concentrations exceeding those of RO₂ radicals (Fig. S11). In these urban areas, primary OVOCs from major sources, such as industrial processes, solvent use, residential sources, and transportation, made substantial contributions to P(O₃) and peroxy radical levels (particularly HO2), ranging 5 %–40 % and 4 %–47 %, respectively (Figs. S12–S15). In contrast, the RO₂ pathway dominated ozone production in the southern and southwestern parts of the domain, where high emissions of biogenic VOCs and relatively low emissions of NO_x prevail (Fig. S9). The enhanced role of RO₂ radicals in these regions is attributed to their higher concentrations and comparable reactivities to HO₂ radicals in converting NO to NO₂.

To elucidate the mechanism of daytime HO₂ radical production, the contributions of OVOCs to the formation of primary HO₂, RO₂, and RO_x radicals (RO_x = HO₂ + OH + RO₂) are assessed. Primary radicals refer to those generated through the photolysis of OVOCs, nitrous acid (HONO), and O₃ and the ozonolysis of OVOCs and unsaturated VOCs, while contributions from radical interconversion and cycling are excluded. The contribution of HNO₄ to HO₂ production is not considered, as HNO₄ is rapidly formed and decomposed through the reversible reaction between NO₂ and HO₂. RO₂ production from peroxy organic nitrates (PNs) is also excluded from the analysis because peroxyacyl radicals rapidly react with NO₂ to form PNs, leading to strong coupling between RO₂ and PNs (Zeng et al., 2019; Zhang et al., 2015; Lin et al., 2024). Generally, OVOC photooxidation dominated primary HO₂ radical production in Shanghai, with the contributions of 95.3 % and 97.0 % under clean and polluted conditions (Fig. 3c). Similar findings have been reported in other urban areas, where OVOC photolysis accounted for over 80 % of primary HO₂ radicals, highlighting its importance in NO-NO₂ cycling and ozone production (Qu et al., 2021; Xue et al., 2016; Yang et al., 2018). Notably, the significance of primary OVOCs in producing HO₂ radicals is suppressed under polluted conditions, which was only half of that under clean conditions. This is associated with increases in both secondary OVOC concentrations and their relative contributions to total OVOCs.

As HO₂ can be produced from RO₂ and RO_x radical cycling, the influences of OVOCs in the primary RO₂ and RO_x production are further examined. Similar to HO₂, substantial contributions were observed from OVOCs, with a more significant contribution to RO₂ by 64.0 % and 66.5 % during clean and polluted periods, respectively. HO₂ can also be formed via RO₂ + NO and RO₂ + RO₂ reactions. The high contribution to primary RO₂ also conveys the crucial role of OVOCs in HO₂ production. For RO_x radicals, OVOCs still exhibited substantial impacts, accounting for 50.2 % and 43.2 % of their primary production rates under clean and polluted conditions. The contributions of primary OVOCs to RO₂ and RO_x are comparable to those to HO₂ radicals, falling within the ranges of 8.69 %–18.4 % and 4.8 %–11.0 %, respectively. This suggests that secondary OVOCs are important contributors to HO₂ production through radical chain propagation.

In Shanghai, HCHO was the most significant contributor to primary HO₂ (∼ 68 %) under both clean and polluted conditions, followed by photoreactive monounsaturated dicarbonyls from aromatic fragmentation (AFG1, ∼ 13 %). Methylglyoxal (MGLY) and glyoxal (GLY) also played an important role in primary HO₂ production, accounting for 6 %–7 % and 5 % among all OVOCs. These findings align with a modeling study in Beijing, which showed that HCHO, MGLY, and aromatic oxidation products played an important role in HO₂ production (Qu et al., 2021). A substantial fraction of RO₂ was attributed to the oxidation of BACL, AFG1, and MGLY, contributing 23.7 %–34.7 %, 32.5 %–38.8 %, and 17.9 %–19.5 % of OVOC-derived RO₂, respectively. As a result, HCHO, AFG1, MGLY, and BACL are major OVOC species generating primary RO_x.

Across the YRD, OVOC photooxidation contributed 90 %–98 % of primary HO₂ when the HNO₄ pathway was excluded (Fig. 4b, e). Primary OVOCs accounted for 20 %–50 % (Fig. 4c, f), with more significant impacts during the cold season, due to a higher contribution from emissions to their ambient levels (Fig. 2e). A high proportion of primary RO₂ was sourced from OVOCs by 50 %–98 % (Fig. S16a, c). However, secondary OVOCs generally made more pronounced contributions to primary RO₂ production at the regional scale, except at a few hotspots where their contributions were comparable to those from primary OVOCs (Fig. S17a, c). The overall contribution of OVOCs to primary RO_x production ranged from 40 % to 70 % in the YRD (Fig. S16b, d). It is evident that OVOCs play an essential role in HO₂ production and could significantly affect ozone formation in this region.

Given that a substantial fraction of OVOCs originates from emissions, the sensitivity of ozone formation to anthropogenic emissions of OVOCs and their precursors was systematically examined using the CMAQ model coupled with the High-Order Decoupled Direct Method (HDDM). The model configuration followed Li et al. (2022b), with the same inputs described in Sect. 2.1. The daytime average ozone responses to the complete removal of various VOCs (ΔO3) at SAES, along with their normalized contributions relative to total anthropogenic emissions across the domain (ΔO₃/ΔEi) under different polluted conditions, are discussed below.

Among all primary OVOCs, HCHO exhibited the most significant impact, contributing 0.90 and 1.01 ppb to ozone formation under clean and polluted conditions, respectively (Fig. 5a). These ozone changes were comparable to those of most VOC precursors except ethylene (ETHE). BACL was also a key OVOC precursor, contributing 0.35 and 0.50 ppb of ozone under clean and polluted conditions, respectively. BACL undergoes photolysis to produce acetyl peroxy radicals (CH₃CO₃), which facilitates NO-to-NO₂ conversion at a rate significantly higher than most other peroxy radicals represented in the model. On a per-unit-mass basis, OVOCs demonstrated substantially higher ozone formation efficiencies than other VOC precursors. Notably, BACL emerged as the most potential contributor in Shanghai, generating 7.78×10-4 and 1.49×10-3 ppb of ozone per gram emitted under clean and polluted conditions, respectively. HCHO, MVK, MACR, and MGLY also exhibited relatively high ozone formation efficiencies, contributing (1.57–3.77) × 10⁻⁴ ppb per gram emitted across both conditions.

Similar sensitivities of O₃ formation to OVOC and VOC precursor emission controls were also observed across the YRD. Among all OVOCs, HCHO demonstrated the most pronounced impact, with emission reductions leading to ozone decreases of up to 2.0 ppb in Jiangsu Province (Fig. S18). This effect was comparable to that of aromatic hydrocarbons (ARO2MN and TOLU) and reactive alkenes (OLE2), and exceeded the influence of most other VOC precursors, except ETHE and propylene (PROP). BACL also exhibited considerable effects, causing up to 1.0 ppb decreases in O₃, similar to reactive alkanes (ALK3-5), alkenes (OLE1), and most aromatic hydrocarbons. Notably, emission reductions of certain OVOCs (e.g., CCHO, RCHO, ACRO, and CRES) can lead to spatially divergent ozone responses, with both increases and decreases in the YRD. This likely arises from differences in the radical pathways associated with OVOC photooxidation: some pathways involve reactions with NO₂ to form relatively stable NO_x reservoir species, whereas others promote NO-to-NO₂ conversion. The competition between these pathways varies regionally with the abundances of OVOCs and NO_x, resulting in spatial heterogeneity in ozone sensitivities. Nonetheless, controlling emissions of OVOCs, particularly HCHO and BACL, represents an efficient strategy for reducing O₃ in the YRD.