Based on comprehensive analysis of written data, consultation with relevant experts, and thorough on-site investigations, we selected five representative factories and industries from various industrial VCP sources. The selection criteria for these industries were based on relevant emission inventory research conducted in the PRD region of China (Zhong et al., 2018). Sampling methods focused on capturing ROG emissions generated during the main manufacturing processes, such as the spraying of coatings and the use of adhesives in factories. Both online measurements and offline sampling were carried out in semi-open workshops and using ROG treatment devices (i.e., before and after emission treatment, generally located at the front and rear sampling ports of the ROG treatment devices) in factories (Table S1 in the Supplement).

Typically, workshop waste gases are routed through collection devices (e.g., gas-collecting hoods, airtight partitions) and then processed in ROG treatment devices (e.g., ultraviolet (UV)-ray oxidation, activated carbon adsorption, combustion, and biodegradation). These treated gases are then released into the atmosphere through exhaust stacks. ROG treatment devices play a crucial role in reducing ROG emissions by employing recovery and destruction technologies (Wang et al., 2023; Kamal et al., 2016). Recovery processes involve enriching and separating VOCs by means of temperature or pressure changes and selective absorbents, while destruction processes convert VOCs into harmless substances such as CO2 and H2O through combustion (Wang et al., 2023). In this study, we evaluate two types of ROG treatment devices: activated carbon adsorption combined with UV photolysis devices (installed in the shoemaking, plastic surface coating, furniture coating, and printing industries) and catalytic combustion devices (installed in the printing and ship coating industries).

During the campaign, a mobile monitoring vehicle was equipped with online measurement equipment and strategically parked near emissions from the sampling ports of both workshops and ROG treatment devices (Fig. S1 in the Supplement). A CO2/ H2O gas analyzer (LI-840A, LI-COR Inc., USA) was used to measure the concentrations of CO2 and H2O. To ensure continuous sampling, air from various factories was drawn through a length of perfluoroalkoxy (PFA) Teflon tubing, ranging from 10 to 100 m, at a controlled flow rate of 6 L min-1 facilitated by an external pump. The use of long tubing was assessed through laboratory tests, which showed that the tubing had a negligible and minor influence on most ROG species. This confirmed the feasibility of measurement using long PFA tubing; more detail can be found elsewhere (Li et al., 2023).

In this study, ROGs were measured using a proton-transfer-reaction quadrupole-interface time-of-flight mass spectrometer (PTR-QiToF-MS) (IONICON Analytik, Innsbruck, Austria) (Sulzer et al., 2014) and a combination of canister sampling and an offline GC–MS/FID analysis system (canister GC–MS/FID). A more comprehensive speciation of ROGs was achieved by analyzing hydrocarbons by canister GC–MS/FID, quantifying all signals using H3O+ PTR-ToF-MS and supplementing with acyclic, cyclic, and bicyclic alkanes from NO+ ionization of PTR-ToF-MS. The selection of overlapping ROGs was similar to a previous study (Table S2) (Gao et al., 2023).

To capture the real-time emission characteristics of ROGs from industrial VCP sources, the mass spectra of PTR-ToF-MS were recorded every 10 s. Prior to each test, background measurements of the instrument were carried out by passing sampling air through a custom-built platinum catalytical converter that had been preheated to 365 °C for 1 min. Throughout the campaign, the PTR-ToF-MS instrument automatically alternated between two reagent ions (H3O+ and NO+) every 10 min. Detailed setting parameters for H3O+ and NO+ chemical modes in this instrument can be found in previous studies (C. Wu et al., 2020; Wang et al., 2020; He et al., 2022b). The Tofware software package (version 3.0.3, Tofwerk AG, Switzerland) was employed to facilitate accurate data analysis (Stark et al., 2015).

Calibration for ROGs measured by PTR-ToF-MS were carried out both in the laboratory and during the campaign. The PTR-ToF-MS instrument was regularly calibrated using a 23-component gas standard (SPECTRA, Linde) throughout the campaign. During the later period of the campaign, two gas standards (Apel-Riemer Environmental Inc.) were used for the calibration of other ROGs, specifically for acyclic and cyclic alkanes using NO+ chemical ionization (Wang et al., 2020; Chen et al., 2022; Wang et al., 2022). A total of 11 organic acids and nitrogen-containing compounds were calibrated using the liquid calibration unit (LCU; IONICON Analytik, Innsbruck, Austria) (Tables S3–S5). In order to account for the humidity dependence of some ROGs in the PTR-ToF-MS instrument (Yuan et al., 2017; Koss et al., 2018), humidity-dependence curves established in the laboratory were utilized for correction (C. Wu et al., 2020; He et al., 2022b; Wang et al., 2022). Sensitivities of uncalibrated species were determined based on the kinetics of proton-transfer reactions of H3O+ with ROGs (Fig. S2) (Cappellin et al., 2012; Sekimoto et al., 2017), with an associated uncertainty of approximately 50 % for the concentrations of uncalibrated species.

Simultaneously, offline sampling was conducted near the sampling ports of workshops and ROG treatment devices. Whole air samples were collected using canisters for the determination of hydrocarbons in industrial VCP sources and analyzed by an offline GC–MS/FID system. The GC–MS/FID system was calibrated using photochemical assessment monitoring stations (PAMS) and TO-15 standard mixtures, which enabled the identification and quantification of a total of 94 hydrocarbons. More information about this instrument and the dataset for canister sampling and the offline GC–MS/FID system can be found elsewhere (Li et al., 2020).

Since ester species (including acetates and acrylates) play a significant role in industrial VCP sources, it is important to accurately quantify their concentrations (Khare et al., 2022). Previous studies have demonstrated that ethyl acetate exhibits notable fragmentation, resulting in interference at m/z 61 (e.g., C2H4O2H+) and m/z 43 (e.g., C2H2OH+) (Haase et al., 2012; de Gouw and Warneke, 2007; Rogers et al., 2006; Fortner et al., 2009). Therefore, we employed the PTR-ToF-MS instrument to directly measure high-purity ester chemicals and identify the characteristic product ions produced by esters under H3O+ and NO+ chemical ionization. Several common esters, including methyl acetate, ethyl acetate, isopropyl acetate, and vinyl acetate, were selected to investigate instrument fragmentation under different ionizations. As shown in Table S6, it is intriguing to observe that high-molecular-weight acetates tend to exhibit more fragmentation, resulting in interference at m/z 61 (e.g., C2H4O2H+) and m/z 43 (e.g., C2H2OH+). Methyl acetate (95 %) and ethyl acetate (72 %) displayed limited fragmentation in the instrument, while isopropyl acetate accounted for only 13 % of the C5H10O2H+ ions. Additionally, esters with different chemical structures may undergo distinct modes of fragmentation, for example, vinyl acetate primarily fragmented to produce interfering fragments at m/z 43 (e.g., C2H2OH+) with a fraction of 78 %. Furthermore, considering the PTR-ToF-MS mass spectra from various industrial VCP sources, it is conceivable that other ester compounds might also contribute to these mass channels, emphasizing the need for cautious consideration of m/z 61 (e.g., C2H4O2H+) and m/z 43 (e.g., C2H2OH+) signals measured by H3O+ PTR-ToF-MS in industrial VCP sources. The use of NO+ chemical ionization exhibits various reaction pathways with ROGs (Wang et al., 2020; Chen et al., 2022), which can partially mitigate interference from fragment ions (Table S6). The identified results of acetates based on NO+ ionization demonstrated considerable improvements for methyl acetate (83 %) and ethyl acetate (80 %), whereas vinyl acetate exhibited more fragmentation, with the largest contribution (47 %) at m/z 43 (e.g., C2H2OH+).

Additionally, it is challenging to calibrate isopropanol in H3O+ PTR-ToF-MS, since alcohols split off water during ionization (Buhr et al., 2002). To overcome this challenge, we implemented daily calibrations of isopropanol under ambient humidity conditions throughout the campaign (Fig. S3). NO+ PTR-ToF-MS was also employed to calibrate isopropanol by identifying the characteristic product ions produced under NO+ ionization (Table S6). The dominating product ion of isopropanol was observed at m/z 59 (e.g., C3H7O+) (88 %), which corresponds to acetone (C3H6OH+) ions in H3O+ PTR-ToF-MS. Although the dominant product ion for acetone under NO+ ionization was observed at m/z 88 (e.g., C3H6O(NO)+) (77 %), the interfere at m/z 59 (e.g., C3H6OH+) (23 %) was not insignificant. Therefore, the concentration of isopropanol measured by NO+ PTR-ToF-MS in this campaign has eliminated the influence of acetone. Finally, the comparison between PTR-ToF-MS with H3O+ and with NO+ chemical ionization is shown in Figs. S4–S5. Previous studies have shown good agreement between measurements obtained using PTR-ToF-MS with H3O+ and NO+ chemical ionization in ambient measurements (Wang et al., 2020). However, a slightly weaker correlation was observed in industrial VCP sources, potentially due to the large changes for different species between the switch of the two reagent ions. Our results demonstrated that NO+ PTR-ToF-MS can also provide a complementary approach for characterizing ester species and isopropanol in ambient air and emission sources.

We conducted a comprehensive comparison of various ROG emission sources by regarding the entire range of species in mass spectra as dimensions in a vector and calculating the cosine angle (θ) similarity (Humes et al., 2022; Ulbrich et al., 2009; Kostenidou et al., 2009). The angle θ between the two mass spectra (MSa and MSb) is given by the following: 1cos⁡θ=MSaMSbMSaMSb. The θ angle between two mass spectra is divided into four groups, including 0–15, 15–30, 30–50, and > 50°, which correspond to excellent consistency, good consistency, many similarities, and poor consistency, respectively. Due to the distinct ionization methods of the instruments, our classification of angle similarity is not as strict as that reported in previous studies (Kostenidou et al., 2009; Zhu et al., 2021), as these previous studies utilize the similarity analysis on mass spectra of an aerosol mass spectrometer (AMS) obtained from electron ionization, leading to very similar mass spectra for different sources.

Time series of several ROGs measured by H3O+ PTR-ToF-MS from five industrial VCP sources are shown in Figs. 1 and S6. More information on these sources can be found in Sect. S1 in the Supplement. Online measurements were carried out in semi-open workshops (workshop emissions) and from ROG treatment devices (i.e., emissions before and after treatment). As the waste gas was directly discharged into the ambient air from exhaust stacks, the after-treatment emissions can be regarded as stack emissions (Zheng et al., 2013). The average concentrations of eight representative ROGs, including aromatics, ketones, alcohols, and esters, between workshop emissions and stack emissions for all factories is presented in Fig. S7. The evaluation of the ROG treatment efficiency is based on the analysis of emission characteristics before and after treatment in the ROG treatment devices, which is discussed in greater detail in Sect. 3.3. Along with the typical ROGs, the PTR-ToF-MS instrument measured a wide range of ions in abundance in the mass spectra. Figure 2 displays mass spectra representing the average concentrations of stack emissions from five industrial VCP sources for all detected ROGs. These ROGs measured by the PTR-ToF-MS instrument were categorized based on their chemical formula, namely hydrocarbon species (CxHy), OVOCs (CxHyO1, CxHyO2, and CxHyO≥3), species containing nitrogen and/or sulfur atoms (N/S-containing), species containing siloxanes (Si-containing), and other ions (others).

We compared the mass spectra of these industrial VCP sources and calculated the similarity of the θ angles (Fig. 3, Table S7). The ROGs showed a diverse similarity among different types of industrial VCP sources. Only the plastic surface coating industry versus the printing industry demonstrated good consistency (27°), while other mass spectra exhibited poor consistency (θ> 60°). Combined with mass spectra of vehicular emissions (Wang et al., 2022), the θ angle similarities among the mass spectra of industrial VCP sources (62–90°) were worse than those of vehicular emissions (41–75°) (Fig. 3). It is interesting to observe that the θ angle similarity among the mass spectra in different workshops in the printing and ship coating industries ranged from 1.6 to 9.0° (Table S8), similar to the mass spectra in various emission standards for gasoline vehicles (4.9–17°) (Table S9). Conversely, the θ angle similarity among the mass spectra of workshops in other industrial VCP sources ranged from 13 to 60°, indicating significant differences in ROG emissions from industrial VCP sources. These substantial differences indicate that ROG emissions from industrial VCPs are more complex and diverse than vehicular emissions. Consequently, a more accurate classification of industrial VCP emissions is necessary, as they cannot be directly unified as a single class of emission sources.

The combination of PTR-ToF-MS and canister GC–MS/FID measurements allowed for more comprehensive speciation of ROG emissions from industrial VCP sources. This comprehensive approach enabled the determination of the fractions of ROGs in total ROG emissions for various industrial VCP sources (Table S10, Fig. S5; details in Sect. S2). Additionally, the reactivity of ROGs plays a crucial role in characterizing the contributions of different ROGs to atmospheric chemical reactions and the formation of secondary pollutants (C. Wu et al., 2020; Yang et al., 2016). The overall OHR of ROGs was calculated to comprehend the role of ROGs emitted by industrial VCP sources. The calculation only employed ROGs with known reaction rate constants with the OH radical, which were taken from previous studies (Atkinson and Arey, 2003; Atkinson et al., 2004, 2006; Koss et al., 2018; C. Wu et al., 2020; Zhao et al., 2016). The fractions of ROGs in the total OHR of ROGs can be determined for various industrial VCP sources (Table S11). ROGs are grouped into categories, including OVOCs, N/S-containing, and heavy aromatics and monoterpenes measured by H3O+ PTR-ToF-MS; higher alkanes (including C10–C20 acyclic, cyclic, and bicyclic cycloalkanes) measured by NO+ PTR-ToF-MS; and alkanes, alkenes, aromatics, and halohydrocarbons measured by canister GC–MS/FID.

OVOCs contributed significantly to total ROG emissions (Fig. 4a), and fractions of OVOCs in total ROG emissions are comparable to previous studies (Fig. 5). Notably, OVOCs account for 67 % of total ROG emissions from the shoemaking industry, which is slightly lower than findings from other studies in the PRD region (Zheng et al., 2013) but higher than those reported in previous studies (Zhou et al., 2020a; Zhao et al., 2018a). The fractions of OVOCs in total ROG emissions from the plastic surface coating, printing, and furniture coating industries are 96 ± 0.2 %, 85 ± 6.5 %, and 77 %, respectively. Compared to previous studies (Zhong et al., 2017; Zheng et al., 2013; Fang et al., 2019; Zhao et al., 2018a; Wang et al., 2019; Zhou et al., 2020a; Zhao et al., 2021), determined OVOC fractions for these industrial VCP sources are much higher (Fig. 5), which may be related to two reasons: (1) more OVOC species are detected in this study, and (2) water-borne coatings and inks are more widely employed in recent years, which may enhance OVOC fractions. Moreover, OVOCs account for 16 ± 3.5 % of total ROG emissions from the ship coating industry by using the solvent-borne coatings, and the fraction is also higher than findings from a previous study (Zhong et al., 2017). Additionally, OVOCs also contribute to the largest fraction in total OHR of ROGs from all industrial VCP sources (72 %–97 %) except for the ship coating industry (15 ± 3.6 %) (Fig. 4b). In contrast to the important contribution of OVOCs, the fractions of hydrocarbons measured by canister GC–MS/FID only made considerable contributions to specific industrial VCP sources (Fig. 4). For instance, aromatics were found to be the major contributors to both total ROG emissions and OHR in the ship coating industry, making up 74 ± 6.1 % and 79 ± 4.8 %, respectively. Alkanes measured by canister GC–MS/FID only make contributions in the shoemaking industry, comprising 26 % of the total ROG emissions. Overall, the total OHR of ROGs was dominated by OVOCs and aromatics, and the contributions of other species were in the range of 1.8 %–21 % (Fig. 4b). These results stress the importance of measuring a broad range of OVOCs using PTR-ToF-MS in characterizing ROG emissions from industrial VCP sources.

The application of NO+ PTR-ToF-MS provided the opportunity for detecting emissions of higher alkanes from industrial VCP sources. We show that the contribution of higher alkanes can be significant for VCP sources. Specifically, the printing industry demonstrates a noteworthy presence of higher alkanes, accounting for 27 ± 2.7 % and 8.2 ± 2.4 % in workshop and stack emissions, respectively (Table S10). This can be attributed to the use of lubricating oil, a primary component of industrial inks, which contains substantial amounts of alkanes (Liang et al., 2018). Furthermore, emissions from forklifts transporting products in printing workshops also contribute to the emission of higher alkanes (Li et al., 2021), suggesting that non-road vehicles may contribute to the emissions from industrial VCP factories. In addition, the fractions of higher alkanes in stack emissions are lower than in workshops, suggesting that ROG treatment devices effectively reduce emissions of higher alkanes.

To facilitate making controlling strategies of ozone, we determine the OFP from a unity of emissions from different sources for comparison (Yuan et al., 2010; Na and Pyo Kim, 2007), which represent the ability to form ozone from ROG sources on a relative basis (Fig. 6), and calculate this using the following equation: 2OFPi=∑j=1nfjiMIRj, where OFPi is the estimated ozone formation amount when 1 g ROGs is emitted from source i, fji is the mass fraction of species j in source i, and MIRj is the maximum incremental reactivity (MIR) of species j (Carter, 2010). Among the industrial VCP sources considered, the ship coating industry exhibited the highest OFP, reaching as high as 5.5 g O3 g-1 ROGs, followed by the furniture coating industry, with a value of 2.7 g O3 g-1 ROGs. The OFP for other industrial VCP sources ranged from 0.79 to 1.4 g O3 g-1 ROGs. Among all industrial VCP sources, aromatics (ranging from 4.2 % to 91 %) and OVOCs (ranging from 6.7 % to 94 %) were identified as the primary contributors to OFP. Compared to vehicular emissions, the OFP from the ship coating and furniture coating industries is significantly higher (Fig. 6), suggesting that these sources should be controlled as a priority. Given the higher reactivity value for the ship coating industry relative to other sources, it is evident that controlling ROG emissions from solvent-borne industrial chemicals would have a more substantial impact on reducing ozone formation compared to other sources. Moreover, it is important to note that the emissions of solvent-borne chemicals surpass those of vehicles, while water-borne chemicals have lower emissions compared to vehicles. This observation implies that the substitution of solvent-borne chemicals with water-borne chemicals in China holds considerable importance in mitigating and controlling ozone pollution.

We further compare centralization for species among different ROG sources by determining the contribution from the top 10 species in terms of concentrations, OHR, and OFP (Figs. 7 and S10; also in Table S12). We show that the top 10 ROGs account for over 50 % of ROG emissions, OHR, and OFP (Fig. 7). With the exception of the furniture coating industry, the fractions of the top 10 species in total emissions, OHR, and OFP from industrial VCP sources were in range of 89 %–96 %. The lower fractions (ranging from 69 % to 86 %) of the top 10 species in the furniture coating industry may be a result of the wider range of industrial coatings (i.e., both solvent-borne and water-borne coatings) utilized in this industry. ROGs emitted from industrial VCP sources are apparently more centralized compared to vehicular emissions (ranging from 51 % to 87 %). Additionally, the cumulative fractions of the top 100 species in overall ROG emissions, OHR, and OFP in various industrial VCP sources are further indicated by the highly centralization of ROG emissions from various emission sources (Fig. S10). More than half of the top 10 species in ROG emissions, OHR, and OFP from industrial VCP sources were OVOCs (Table S13). Among them, isopropanol made a notable contribution to the printing, plastic surface coating, and shoemaking industries. Other OVOCs such as MEK, acetone, and ethyl acetate contributed to total ROG emissions in each industry, while formaldehyde and acetaldehyde contributed to total OHR and OFP. It should be noted that the proportions of C13, C14, and C15 cycloalkanes from the printing industry (accounting for 6.3 % of ROG emissions) and the proportion of C11 aromatics from the ship coating industry (accounting for 1.0 % of ROG emissions) are not negligible. Additionally, acetylacetone is a common species with broad industrial applications and contributes importantly to the formation of secondary pollutants in polluted environments (Ji et al., 2018). Although it only contributes 8.7 % to total emissions from the furniture coating industry, its fraction in terms of total OHR can be as high as 30 %. These findings demonstrate that previously underreported ROGs should receive greater attention in future research.

The updated measurements of OVOC emissions by using PTR-ToF-MS substantially improve our understanding of the emissions of industrial VCP sources. The effective saturation concentrations (C*) of high-molecular-weight OVOCs were found to be lower, corresponding to intermediate-volatility organic compounds (IVOCs) and semi-volatile organic compounds (SVOCs). Since these S/I-VOCs are crucial precursors for SOAs in urban environments (Zhao et al., 2014), it is important to comprehend their contributions from the emissions of industrial VCP sources across various volatility classes, including volatile organic compounds (VOCs), IVOCs, and SVOCs (Guenther et al., 2012; Li et al., 2016). Figure 8 illustrates the distribution of ROG species in a two-dimensional volatility basis set (2D-VBS) space for various industrial VCP sources, categorized based on volatility bins (Li et al., 2016; Donahue et al., 2011). It is worth noting that the volatility distributions exhibit substantial variation across industrial VCP sources (Fig. 8a). Generally, VOCs constitute the predominant fraction of emissions from industrial VCP sources, accounting for 59 % to 98 % of the total emissions. The fractions of IVOCs are largest in the printing industry (40 %) compared to the range of 2.1 %–9.6 % in other industrial VCP sources. Conversely, the contribution of SVOCs from industrial VCP sources is negligible in our study, accounting for less than 1 %. Considering the importance of S/I-VOCs in SOA formation, particularly with the increasing adoption of improved online mass spectrometry technologies, the S/I-VOC emissions from industrial VCP sources should be paid more attention in future research.

The analysis of the PTR-ToF-MS mass spectra offers valuable insights into the impact of ROG emissions from industrial VCP sources. This comprehensive information provided by PTR-ToF-MS also allows a systematic comparison of emissions before and after the treatment of ROGs. The scatterplot of the concentrations of various ROGs before and after treatment in industrial VCP sources is shown in Figs. 9 and S11. The observed treatment efficiency, represented by a slope gradient of 1, did not reach the desired levels, ranging from -12 % to 68 %. Among the industrial VCP sources investigated, the shoemaking industry exhibited the highest treatment efficiency (slope of 0.32) with the activated carbon adsorption combined with the UV photolysis device. This remarkable efficiency can be attributed to the large-scale nature of the factory and meticulous regulation of the ROG treatment devices. Following closely behind is the printing industry, utilizing catalytic combustion devices, with a slightly higher efficacy (slope of 0.67) than another treatment device in the same factory (slope of 0.80). Nonetheless, it is evident that the treatment efficiency has not reached the desired levels for all ROG groups (Fig. S11), which is possibly due to the challenges associated with effectively removing the majority of ROG emissions using current treatment technologies. Additionally, we also observed that some OVOCs may be generated as byproducts after the implementation of treatment devices. For instance, the concentrations of CH2O2 (e.g., formic acids), C4H6O3 (e.g., propylene carbonate), and C9H18O (e.g., nonanal) were found to be higher after the application of activated carbon adsorption combined with UV photolysis devices (Fig. 9d). Similarly, the concentrations of C3H4O (e.g., acrolein) and C12H18O4 (e.g., dibutyl squarate) were also higher following the utilization of catalytic combustion devices (Fig. 9e). Therefore, it is crucial to consider the potential contribution of these ROGs when assessing the emissions released into the atmosphere. The lowest treatment efficiency of ROGs was obtained in the furniture coating industry (slope of 1.12). This treatment device demonstrates inefficiency for all ROG groups (Fig. S11). The inadequate performance of the ROG treatment devices in this specific facility may be attributed to a number of possible reasons, e.g., delayed replacement of activated carbon and other adsorption materials, and the implementation of the UV photolysis device could potentially result in the generation of more ROGs as byproducts.

Furthermore, the θ angles between the mass spectra of ROG from workshops before and after ROG treatment devices for various industrial VCP sources were calculated and summarized in Fig. S12 (also in Table S13). A comparison between the correlation of mass spectra among workshops versus after treatment devices (ranging from 6.2 to 49°) and workshops versus before treatment devices (ranging from 4.2 to 41°) demonstrated a poorer correlation in the former case. The similarities between workshops and stack emissions in the shoemaking industry were lower compared to other industrial VCP sources. This discrepancy can potentially be attributed to the inclusion of ROG emissions from non-VCP-usage manufacturing processes (e.g., sole injection molding) in the collection process of ROG treatment devices. Additionally, the similarity of θ angles between the mass spectra from before and after ROG treatment devices in various industrial VCP sources also provided insight into the efficacy of the devices in removing ROGs (Fig. 9). The θ angles in ROG treatment devices from five industrial VCP sources were found to range from 1.8 to 27°, indicating good consistency between the mass spectra before and after treatment of ROGs. This alignment suggests that the chemical compositions of ROG emissions remain comparable before and after treatment (R≥ 0.87), implying that the relative proportions of various ROG components are not significantly affected by ROG treatment devices in these industrial VCP sources.

To gain deeper insights into the atmospheric impact of emissions from industrial VCP sources, an in situ measurement was carried out at a monitoring station near the furniture coating industry, located 2 km northeast of the industry site. The measurement was conducted using a PTR-ToF-MS instrument (Kore Inc., UK), which enabled the quantification of various common ROGs. More information about this instrument and dataset for in situ measurement can be found elsewhere (Gonzalez-Mendez et al., 2016; Song et al., 2023). Concordant with expectations, the average concentrations of representative ROGs generally demonstrate a discernible decline from the furniture coating industry (including stack emission and workshops during working and non-working hours) to the ambient measurement (Fig. 10). Considerable emissions of C8 aromatics and MEK are generated from the furniture coating industry, while their concentrations in the ambient air are orders of magnitude lower than those observed in the industry. However, ambient concentrations of OVOCs, specifically MEK (6.8 ± 8.2 ppb) and ethyl acetate (7.5 ± 5.9vppb), are still significantly higher than in clean environments and are among the highest measured concentrations in the literature (C. Wu et al., 2020; He et al., 2022b; Khare et al., 2022; Yang et al., 2022). It is confirmed that OVOCs should be paid attention to in industrialized urban areas, thereby further substantiating the significance of OVOC emissions from industrial VCP sources in atmospheric pollution. These results stress the invaluable insights provided by PTR-ToF-MS in comprehensively characterizing ROG compositions in both emission sources and urban air.

The preceding discussions illustrate that the emission characteristics of ROGs significantly vary among industrial VCP sources. As a result, the ratio of ROG pairs can be used to distinguish emissions of industrial VCP sources. MEK and C8 aromatics emerge as key species in industrial VCP emissions, and the reaction rate constants of C8 aromatics with OH radicals (kOH=(1.4-2.3)×10-11 cm3 molec.-1 s-1) are higher than MEK (kOH=5.5×10-12cm3 molec.-1 s-1) (Atkinson and Arey, 2003; C. Wu et al., 2020). Figure 11 illustrates the correlation of MEK with C8 aromatics in stack emission, workshops during working hours and non-working hours in the furniture coating industry, and ambient measurement near the industry. Positive correlations between MEK and C8 aromatics are observed in both emission sources and ambient measurements, indicating a shared source for these compounds. Additionally, the observed ratios of MEK to C8 aromatics in ambient measurements are also comparable with the ratios of MEK to C8 aromatics measured in emissions from the furniture coating industry (0.97 ppb ppb-1 for stack emissions and 0.75 ppb ppb-1 for workshop emissions), suggesting that industrial VCP emissions (specifically furniture coating) may account for the enhancement of MEK and C8 aromatics in this industrial area. The peak concentration of MEK exceeding 200 ppb from the ambient measurements are among the highest in the literature (Fig. 11). Therefore, we conducted a comparison of MEK and C8 aromatics concentrations in this study with those in clean environments (urban, rural, forest, and coastal sites) from previous studies (Fig. S13) (C. Wu et al., 2020; Coggon et al., 2024; Yuan et al., 2012; Seco et al., 2011; Acton et al., 2016; Tan et al., 2021; He et al., 2022b). The results indicate that ambient measurements in industrial areas have been significantly impacted by industrial VCP sources, and the MEK / C8 aromatics ratio can serve as good evidence by using high-time-resolution ROG measurements from PTR-ToF-MS.