Characteristics, Sources of Atmospheric VOCs and Their Impacts on O₃ and Secondary Organic Aerosol Formation in Ganzhou, Southern China

Abstract

Driven by factors such as meteorology, topography, and industrial structure, the concentrations of volatile organic compounds (VOCs) exhibit significant spatial heterogeneity. Investigating the characteristics and sources of VOCs in different regions is therefore crucial for formulating targeted strategies to mitigate their contributions to fine particulate matter (PM_2.5) and ozone (O₃) pollution. This study comprehensively investigated—for the first time—the concentration characteristics, sources, and contributions to secondary organic aerosol (SOA) and O₃ formation of VOCs at an urban background site in Ganzhou, a southern Chinese city, based on hourly observations of VOCs during 2023. Analyses included ozone formation potential (OFP), secondary organic aerosol formation potential (SOAFP), and positive matrix factorization (PMF) source apportionment. The influence of photochemical loss was assessed using a photochemical age parameterization method. The results showed an annual average total VOC concentration of 22.6 ± 13.17 ppbv, with higher levels in winter and lower in summer. Alkanes were the dominant species (45.76%). After correcting for photochemical loss, the initial concentration of VOCs (IC-VOCs) was approximately 60% higher than the observed concentration of VOCs (OC-VOCs), with alkenes becoming the dominant group in IC-VOCs (≈72%). OFP analysis indicated that the OFP calculated using initial VOC concentrations (IC-OFP) was substantially higher (by 320 μg/m³) than the values calculated using observed VOC concentrations (OC-OFP), primarily due to the increased contribution of alkenes. SOAFP was higher in spring and winter, and lower in summer and autumn, with aromatic hydrocarbons being the dominant contributors (>85%). PMF results based on month-case studies identified combustion and industrial process sources as the major contributors (>20%) in August, while combustion and vehicle exhaust dominated in January. Photochemical loss significantly influenced source apportionment, particularly leading to an underestimation of biogenic emissions during a warm month (August). These findings underscore the necessity of accounting for photochemical aging and offer a scientific basis for refining targeted VOC control measures in Ganzhou and similar regions.

1. Introduction

Since the implementation of the Air Pollution Prevention and Control Action Plan in 2013, China has achieved remarkable success in reducing fine particulate matter (PM_2.5) levels. However, surface ozone (O₃) concentrations have continued to increase across most regions [1,2]. Strengthening the coordinated control of PM_2.5 and O₃ has become essential for the sustained improvement of air quality [3,4]. Volatile organic compounds (VOCs), as critical precursors in atmospheric photochemistry, are key drivers in the formation of both O₃ and secondary organic aerosols (SOAs). Furthermore, some VOC components possess significant carcinogenic potential, and long-term exposure poses serious threats to human health [5]. Therefore, systematically elucidating the characteristics and sources of VOCs is of great scientific significance for effectively controlling O₃ and SOA formation and for protecting public health.

In recent years, atmospheric VOCs have become a hot research topic worldwide (e.g., [6,7,8,9]). Particularly in China, extensive studies on VOCs have provided a robust foundation for understanding emission characteristics and implementing targeted strategies for complex pollution. Major anthropogenic emission sources of VOCs in Chinese cities include vehicle exhaust, fuel evaporation, natural gas (NG) and liquefied petroleum gas (LPG) use, industrial processes, solvent utilization, and combustion emissions [10,11]. Among these, vehicle emissions are identified as the dominant contributor in most regions [12]. However, significant spatial heterogeneity exists in the concentration levels and source profiles of VOCs across different regions of China. Typically, total VOC (TVOC) concentrations are higher in the North China Plain (NCP) and the Yangtze River Delta (YRD), likely influenced by industrial structure and emission intensity, followed by the Chengdu–Chongqing (CC) and the Pearl River Delta (PRD) regions [13,14]. Moreover, variations in industrial structure leads to differences in the composition and proportions of VOCs among different cities [15,16]. Regarding source contributions, combustion sources contribute an average of 23.2% in northern regions, significantly higher than in other areas, while solvent usage shows the highest average contribution (19.5%) in eastern China. These disparities highlight the necessity for developing region-specific emission reduction strategies, for which a clear understanding of VOC source composition and contribution is a critical prerequisite [12].

On the other hand, due to the high photochemical reactivity of VOCs, the atmospheric lifetimes of some highly reactive species range from tens of minutes to a few hours [17,18]. Consequently, concentration levels measured at monitoring sites have already been subject to atmospheric photochemical consumption, making it difficult to accurately reflect their initial emission levels. This limitation constrains the formulation of refined prevention and control strategies. In recent years, the photochemical age-based parameterization method has been increasingly applied to estimate the initial concentrations of VOCs (IC-VOCs), offering a new perspective for characterizing their true emission features and deepening the understanding of regional pollution causes [19]. However, the methodological concern has been widely recognized in previous studies: reactive losses during transport can systematically depress short-lived tracers in the observed mixture and thereby bias receptor-model source apportionment and source-oriented interpretation, motivating the growing use of photochemically adjusted analyses [20,21].

Currently, most existing VOC studies in China have been predominantly focused on economically developed eastern coastal regions (e.g., YRD, PRD), while relatively less attention has been paid to less developed provinces and their cities, leaving the research foundation in some areas still relatively weak [22]. Ganzhou City, as the second-largest economy in Jiangxi Province, has experienced prominent air environmental issues accompanying its rapid economic development over the past decade. Its maximum daily 8 h average ozone (MDA8-O₃) concentration has shown an upward trend in recent years [23]. More importantly, Ganzhou is situated at the junction of South and East China, serving as a transitional zone connecting developed regions like the PRD, and is characterized by a typical subtropical monsoon climate. The relatively high annual average temperature and intense solar radiation provide favorable conditions for photochemical pollution. However, to the best of our knowledge, the concentrations and sources of VOCs in Ganzhou City are largely unexplored. This lack of data significantly impedes the development of evidence-based and effective control policies.

Therefore, this study utilizes online high-time-resolution VOC observations at an urban background site in Ganzhou in 2023 to systematically analyze concentration and compositional characteristics. The photochemical age parameterization method is employed to reconstruct the IC-VOCs. Unless otherwise specified, VOC concentrations discussed in this study refer to observed concentrations (OC-VOCs). By comparing the differences in ozone formation potential (OFP) and secondary organic aerosol formation potential (SOAFP) before and after this reconstruction, and combining positive matrix factorization (PMF) model for source apportionment, this research aims to quantify the impact of photochemical consumption on environmental effects and source identification of VOCs, and to identify key reactive species. The findings are intended to provide a scientific basis for the precise prevention and control of O₃ pollution in Ganzhou and similar regions. The layout of this study is as follows: Section 2 describes the sampling site and observation data, as well as the introduction of employed methods. Section 3 presents the results of VOC characteristics, the impact of photochemical loss on SOAFP and OFP, and PMF source apportionment in Ganzhou. Section 4 provides the summary and conclusions drawn from this study.

2. Data and Methods

2.1. Sampling Site and Observation Data

As illustrated in Figure 1, the VOC observation site was located on the rooftop of the main building (5th floor) of Hongqi No. 2 primary school in Zhanggong District, Ganzhou City (114.95° E, 25.85° N). The site was established and operated by the Jiangxi Provincial Environmental Monitoring Center in accordance with relevant technical guidelines for urban air quality and VOC monitoring (HJ 664–2013; HJ 194–2017) [24,25], and the rooftop setting meets standard requirements for open surroundings, minimal obstruction, and avoidance of direct emission plumes. The sampling site is adjacent to a major road on its east side, with public service facilities such as residential areas, schools, and hospitals distributed within a 500 m radius. No significant large-point pollution sources are present in the immediate vicinity, allowing the site to effectively represent typical urban background pollution levels in Ganzhou. Measurements may also reflect local micro-scale influences, such as nearby traffic and routine human activities. Continuous measurements were conducted from 1 January to 31 December 2023. The monitoring period was designed to capture one full annual cycle of seasons and photochemical conditions. Such setups are common and effective practices in studies investigating regional characteristics of VOCs (e.g., [15,26]). The prevailing wind direction during 2023 exhibited a distinct north–south distribution pattern (Figure 2). Specifically, southerly winds dominated in summer, while northerly winds prevailed in autumn and winter, and both directions coexisted in spring, which was consistent with the typical subtropical monsoon humid climate.

VOC measurements were performed using an online analyzer (XHVOC6000, Hebei Sailhero Environmental Protection Technology Co., Ltd., Shijiazhuang, China). Ambient air samples were pre-concentrated via electronic refrigeration at −45 °C and subsequently thermally desorbed at a rate exceeding 1000 °C/s before analysis by a coupled gas chromatography–mass spectrometry/flame ionization detector (GC–MS/FID) system. The instrument is capable of quantifying over 100 VOC species in a single injection, encompassing C₂ to C₁₂ hydrocarbons, aromatic hydrocarbons, halogenated hydrocarbons, chlorinated benzenes, oxygenated VOCs (OVOCs), sulfides, and semi-volatile organic compounds.

In accordance with the Chinese environmental standard [27] and other relevant standards, rigorous instrument calibration and data quality control (QC) procedures were implemented to ensure the reliability of the data. The calibration and QC protocol included (1) weekly zero-air blank checks and single-point QC checks to verify the absence of VOC carryover and ensure measurement accuracy; (2) monthly inspections of sampling flow rates; (3) quarterly establishment of calibration curves, mandating a coefficient of determination (R²) greater than 0.98; and (4) annual verification of method detection limits (MDLs), requiring that MDLs for ≥90% of species be ≤0.15 ppbv, alongside a 7-day continuous operation test to confirm instrument precision and accuracy were within ±20%. To further ensure data reliability, a robust Z-score method was applied to identify and remove potential outliers [28].

Simultaneous observational data for conventional pollutants (O₃, PM_2.5, NO₂, and SO₂) were obtained from the national air quality monitoring station (Meteorological Bureau station) in Ganzhou, which was located within 500 m of the VOC sampling site. Meteorological parameters, including air temperature (TEM), relative humidity (RH), wind speed (WS), wind direction (WD), boundary layer height (BLH), and downward shortwave radiation (DSR), were derived from the fifth-generation atmospheric reanalyses (ERA5) dataset provided by the European Centre for Medium-range Weather Forecasts (ECMWF). The ERA5 dataset features a high temporal resolution of 1 h and a spatial resolution of 0.25° × 0.25°. In this study, data from the grid cell encompassing the monitoring site were extracted and used for analysis, covering the entire calendar year of 2023.

2.2. Estimation of Photochemical Loss of VOCs

OC-VOCs are often lower than their actual emission concentrations due to atmospheric photochemical consumption. To reconstruct this lost information, the photochemical age parameterization method [29] was employed to estimate IC-VOCs. The calculation is performed as follows [19]:

$${\left[IC\right]}_i={\left[OC\right]}_i\times exp\left(k_i\times \left[OH\right]\times ∆t\right)$$

(1)

where $[IC{]}_i$ and $[OC{]}_i$ are the initial and observed concentrations (ppbv) of VOC species i, respectively; $k_i$ is the rate constant for the reaction of the ith VOC with OH radicals (cm³ molecule⁻¹ s⁻¹); $\left[OH\right]\times ∆t$ denotes the cumulative exposure of OH radicals (molecule cm⁻³ s), which is the product of average OH radicals concentration $\left[OH\right]$ and photochemical reaction time $∆t$.

The value of $\left[OH\right]\times ∆t$ can be obtained using Equation (2):

$$[OH]\times ∆t={\displaystyle \frac{1}{k_E-k_X}}\mathrm{l}\mathrm{n}({\displaystyle \frac{(C_E/C_X{)}_{initial}}{(C_E/C_X{)}_{current}}})$$

(2)

where $k_E$ and $k_X$ are the rate constants for the reactions of OH radicals with ethylbenzene and m/p-xylene, respectively, which can be obtained from studies by [30,31]. $(C_E/C_X{)}_{initial}$ and $(C_E/C_X{)}_{current}$ represent the concentration ratio of ethylbenzene to m,p-xylene in fresh emission and at the current time t, respectively.

The core assumptions underlying the photochemical age method are that the two tracer species share similar emission sources and transport pathways, and their atmospheric removal is dominated by reaction with OH radicals, while losses from other processes (e.g., NO₃, O₃) are negligible during the estimation period. Given that $(C_E/C_X{)}_{initial}$ is not directly measurable at the receptor site, it can be determined from measurements under minimally processed conditions. As shown in Figure S1, the 00:00–05:00 LST period was selected to represent fresh emissions with minimal photochemical processing based on the observed stability and low values of the ethylbenzene/m,p-xylene (E/X) ratio during these hours. This approach follows previous studies (e.g., [32,33]) that used similar nighttime windows to approximate the initial ratio.

Within the selected 00:00–05:00 LST window, the lowest 10% of $(C_E/C_X{)}_{current}$ data points were used for linear regression to approximate $(C_E/C_X{)}_{initial}$. Given the substantial seasonal variability, the fitting procedure was performed separately for the four seasons. The resulting $(C_E/C_X{)}_{initial}$ values were 0.38 for spring, 0.31 for summer, 0.35 for autumn, and 0.38 for winter, respectively. It should be noted that nighttime chemical losses due to NO₃ radicals and residual OH may still occur, especially for reactive alkenes. Although the current method primarily corrects for daytime OH-driven losses, neglecting nocturnal oxidation could lead to an underestimation of total VOC loss [34]. Moreover, regional transport of aged air masses or mixing with chemically processed pollutants could affect the initial ratio determination and introduce uncertainty in $∆t$ and IC-VOCs [35,36].

To quantify the uncertainty associated with the fitted reference ratio $(C_E/C_X{)}_{initial}$, a sensitivity test was conducted by perturbing $(C_E/C_X{)}_{initial}$ within a range of ±10%, and the corresponding IC-VOCs were recalculated under each scenario. The results show that increasing $(C_E/C_X{)}_{initial}$ by 10% led to an average 12% decrease in the reconstructed IC-VOCs, whereas decreasing $(C_E/C_X{)}_{initial}$ by 10% resulted in an approximately 21% increase. These variations indicate that uncertainty in $(C_E/C_X{)}_{initial}$ mainly affects the absolute magnitude of reconstructed IC-VOCs. However, within this perturbation range, the relative magnitude and seasonal patterns of reconstructed IC-VOCs remain unchanged, supporting the use of the photochemical age method for robust comparative interpretation of daytime photochemical processing.

By calculating the IC-VOCs, the theoretical photochemical loss of ambient VOCs (PL-VOCs) at time t can be further estimated as follows:

$$P{L}_{i,t}=I{C}_{i,0}-O{C}_{i,t}$$

(3)

where $P{L}_{i,t}$ is the photochemical loss amount of VOC species i in the atmosphere at time t.

2.3. Ozone Formation Potential (OFP) and Secondary Organic Aerosol Formation Potential (SOAFP)

To identify the key reactive species influencing O₃ formation, the ozone formation potential (OFP) was calculated using the maximum incremental reactivity (MIR) method [31], as follows:

$${\mathit{OFP}}_i=[VOC{]}_i\times {\mathit{MIR}}_i$$

(4)

where ${\mathit{OFP}}_i$ is the ozone formation potential of VOC species i (units: μg/m³). ${\left[VOC\right]}_i$ is the concentration of the species i converted from ppbv to μg/m³ under standard conditions (0 °C, 1 atm, 22.4 L/mol), and ${MIR}_i$ is the maximum incremental reactivity coefficient (g O₃ per g VOC), with specific values derived from [31].

The secondary organic aerosol formation potential (SOAFP) was used to assess the contribution of VOCs to secondary aerosol formation, and was calculated according to [37]:

$$SOAF{P}_i={\left[VOC\right]}_i\times SOA{P}_i\times f_{TOL}$$

(5)

where $SOAF{P}_i$ is the SOAFP of VOC species i (units: μg/m³), $SOA{P}_i$ is the dimensionless SOAP of species i relative to toluene, and $f_{TOL}$ represents the toluene aerosol coefficient, fixed at 5.4% in this study [38].

Note that MIR- and SOAP-based OFP/SOAFP are applied as relative, diagnostic metrics to compare VOC species/sources. While using fixed MIR and SOAP coefficients may bias the actual OFP/SOAFP values under Ganzhou’s local chemical and meteorological conditions, this widely adopted standard approach ensures comparability with numerous other studies.

2.4. Source Apportionment of VOCs with the PMF Model

The PMF receptor model is widely used for VOC source apportionment, benefiting from its advantage of not requiring a detailed source emission inventory [12]. The model decomposes the receptor sample matrix $X$ into a product of source contribution matrix $G$ and source profile matrix $F$, and a residual matrix $E$, expressed as follows:

$$X=G\times F+E$$

(6)

The PMF solution is achieved by minimizing the objective function $Q$, defined as follows:

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

(7)

where $U_{ij}$ is the uncertainty for species j in sample i, and the uncertainty for each concentration value is calculated as follows:

$$U_{ij}=\left\{\begin{array}{cc}\hfill 5/6\times MDL,& C_{ij}<MDL\hfill \\ \hfill \sqrt{{\left(EF\times C_{ij}\right)}^2+{\left(0.5\times MDL\right)}^2},& C_{ij}\ge MDL\hfill \end{array}\right.$$

(8)

where $C_{ij}$ is the concentration of the VOC species j in sample i, and $EF$ is the assigned analytical error fraction (set to 20% in this study).

PMF analyses were performed separately using both OC-VOCs and reconstructed IC-VOCs. The IC-VOCs was constructed by combining the estimated initial concentrations for periods with significant photochemical activity (06:00–23:00 LT) and the original observed concentrations for periods with minimal photochemical loss (00:00–05:00 LT). The resulting factor contributions and profiles from the observed-concentration PMF (OC-PMF) and the initial-concentration PMF (IC-PMF) were compared to assess the influence of photochemical loss on source apportionment.

3. Results and Discussion

3.1. VOC Characteristics in Ganzhou

Figure 3 shows the time series of VOCs, conventional air pollutants (O₃, PM_2.5, SO₂, NO₂), and meteorological parameters in Ganzhou during 2023. Overall, TEM, DSR, and BLH exhibited similar seasonal patterns, with their average values all peaking in summer (TEM: 29.0 °C, DSR: 208.9 W/m², BLH: 537.8 m,) and reaching their lowest points in winter (TEM: 10.2 °C, DSR: 123.1 W/m², BLH: 402.6 m). The seasonal variation in RH was relatively modest, with the highest average observed in autumn (78.8%) and the lowest in winter (74.8%). PM_2.5, SO₂, and NO₂ concentrations also showed higher levels in winter and lower levels in summer due to the combined influence of meteorological conditions and anthropogenic emissions (Figure 3b,c). However, the maximum ozone (MDA8-O₃) concentration peaked in spring (180.0 μg/m³), which is consistent with observations from other cities in southern China, mainly attributed to O₃-photochemistry favorable springtime meteorological conditions [39].

The annual mean TVOC concentrations at the urban background site in Ganzhou during 2023 were 22.6 ± 13.17 ppbv. Similarly to other pollutants, TVOCs exhibited a distinct seasonal variation, with the highest average concentration occurring in winter (30.0 ppbv), followed by spring and autumn (20.4 and 22.4 ppbv, respectively), and the lowest in summer (16.8 ppbv). A regional comparative analysis indicates that the TVOC concentrations in Ganzhou were significantly lower than those observed in typical industrial cities and regions in northern China, such as Beijing (winter/summer: 50.0/34.4 ppbv) [40], Shijiazhuang (summer: 23.2 ppbv) [41], and the Fenwei Plain (36.2 ppbv) [26]. It was also lower than levels in western cities like Chengdu (winter/summer: 53.3/26.8 ppbv) [42], central cities like Wuhan (winter/summer: 44.1/22.1 ppbv) [43], and southern cities like Guangzhou (winter/summer: 27.2/21.0 ppbv) [44]. In contrast, the difference was smaller compared to some coastal cities, such as Lianyungang (winter/summer: 23.7/12.0 ppbv) [22], Tianjin (spring–summer: 19.4 ppbv) [33], Shanghai (spring–winter: 21.4 ppbv) [45], and Qingdao (summer: 19.4 ppbv) [46]. In summary, the TVOC concentrations in Ganzhou rank in the medium-to-low range among the cited Chinese cities. Despite being an important industrial city in Jiangxi Province, its VOC levels remain lower than those in traditional large cities (e.g., Beijing), reflecting the constraint of regional economic development level on emission intensity. Furthermore, compared to earlier observational data, the generally lower concentrations seen in recent years across various locations also demonstrate the effectiveness of China’s strict emission control policies.

Figure 4 further presents the proportion of VOC components in Ganzhou. During the whole study period, alkanes were the dominant group of ambient VOCs (44.7%), followed by alkenes (16.1%), halogenated hydrocarbons (16.0%), oxygenated volatile organic compounds (OVOCs, 11.5%), acetylene (5.9%), and aromatic hydrocarbons (5.9%). Ethane (12.5%) and propane (10.8%) were the predominant species among alkanes, while ethylene (7.3%) was the most abundant alkene. Previous studies suggest that ethane and propane in urban environments primarily originate from natural gas usage, with the proportion of alkanes correlating with the level of urban fuel consumption demand, whereas alkenes are largely attributed to transportation emissions [40,42]. A high proportion of alkanes have also been observed in other cities such as Tianjin (58.2%), Wuhan (45.9%), and Chengdu (44.5–54.8%) [33,42,43]. However, the second-most abundant VOC group varies among cities, potentially being alkenes, aromatic hydrocarbons, or OVOCs [13,47]. Notably, the PRD and YRD regions show significantly higher proportions of aromatic fractions (20–30%) [44], which markedly differs from the profile observed in Ganzhou in this study. This discrepancy is likely attributed to the stronger industrial emission intensities in those developed coastal regions.

The seasonal variations of six VOC categories are shown in Figure 4a–d. Alkenes, OVOCs, and halocarbons generally exhibited higher fractions in summer and lower abundances in winter (or spring). Their respective average proportions in summer were 17.8%, 14.9%, and 17.1%, which decreased to 14.9%, 9.7%, and 14.1% in winter. Conversely, alkanes and alkynes reached their highest proportions in winter (48.3% and 6.4%, respectively) and their lowest in summer (39.9% and 4.9%, respectively). Aromatics displayed a distinct seasonal pattern, peaking in spring (6.9%) and reaching their lowest level in winter (3.6%). Further analysis suggested that species such as isoprene, 1-butene, and isopropanol might be key contributors to the increased proportions of alkenes and OVOCs in summer. Isoprene, a recognized biogenic tracer, exhibits emission levels highly dependent on temperature and solar radiation; elevated temperatures and intense radiation enhance plant physiological activity, consequently increasing isoprene emission flux [48]. On the other hand, high summer temperatures significantly enhance the volatilization of VOCs (e.g., 1-butene, isopropanol, cyclohexane) with relatively high saturation vapor pressures from liquid media such as solvents and fuels [49], thereby increasing their relative contributions to the VOCs during summer and autumn. Furthermore, the concentrations and relative rankings of ethylene and ethane rebounded in spring and winter, likely associated with increased fuel consumption for heating and industrial combustion activities during these colder seasons [12].

3.2. Photochemical Loss of VOCs and Their Impact on SOA and O₃ Formation Potentials

3.2.1. Seasonal Variations in Photochemical Loss of VOC Concentrations

Figure 5a shows the seasonal variations in IC-VOCs, OC-VOCs, as well as PL-VOCs in Ganzhou. It is seen that averaged IC-VOCs (~36.2 ppbv) were significantly higher than those of OC-VOCs (~22.6 ppbv) by 60.1% during the study period. Seasonally, IC-VOCs displayed a pattern of winter (43.9 ppbv) > autumn (35.8 ppbv) > summer (33.7 ppbv) > spring (31.6 ppbv), which differs from the OC-VOC pattern of high winter and low summer concentrations. This discrepancy is primarily due to the largest magnitude of PL-VOCs in summer (16.9 ppbv), reflecting how the meteorological conditions play an important role in photochemical loss. The fraction of PL-VOCs accounted for approximately 50.1%, 37.4%, 32.9%, and 31.7% of IC-VOCs in summer, autumn, spring, and winter, respectively, highlighting the strong influence of photochemical processes during warm season, which is consistent with other studies [50].

Regarding changes in species composition (Figure 5b), the season-averaged proportions of alkenes, OVOCs, and aromatic hydrocarbons in IC-VOCs increased compared to their proportions in OC-VOCs. The most notable increase was observed for alkenes (rising from 3.7 ppbv (16.1%) in OC-VOCs to 13.9 ppbv (37.3%) in IC-VOCs), followed by OVOCs (from 2.7 ppbv (11.5%) to 4.5 ppbv (12.2%)) and aromatic hydrocarbons (from 1.4 ppbv (5.9%) to 2.4 ppbv (6.4%)). In contrast, the proportions of alkanes, alkynes, and halogenated hydrocarbons decreased after reconstruction, with minimal changes in their absolute concentrations (alkanes: 44.7→30.1%; alkynes: 5.9→3.9%; halogenated hydrocarbons: 16.0→10.1%). This shift is primarily attributed to the discrepancy in photochemical reactivity. The reaction rate constants with OH radicals for alkenes are typically higher with range of 30–100 × 10⁻¹² cm³ molecule⁻¹ s⁻¹, accounting for over 70% of the total photochemical loss, whereas those rate constants for other species are generally lower (e.g., alkanes: approximately 6 × 10⁻¹² cm³ molecule⁻¹ s⁻¹; aromatic hydrocarbons: about 20 × 10⁻¹² cm³ molecule⁻¹ s⁻¹). As a result, the proportion of alkenes increased significantly in IC-VOCs after correcting for photochemical loss.

Seasonally, alkanes exhibited the most pronounced fluctuation (40.0–48.2%) in OC-VOCs (Figure 5b). After reconstruction, however, alkenes showed more prominent seasonal variation in IC-VOCs, with the highest contribution in summer (47.3%) and the lowest in spring (31.2%), while the differences for other species in IC-VOCs and OC-VOCs were relatively minor. Further analysis of the top ten species contributing to PL-VOCs in each season identified 1-pentene, 1-butene, and trans-2-butene (all ranking within the top 4 species in IC-VOCs) as the major influencing species. Particularly, isoprene additionally contributed to elevated PL-VOCs in summer due to its enhanced biogenic emissions from plants. The significantly increased concentrations of these highly reactive species after reconstruction indicate that IC-VOCs better reflect the true seasonal characteristics of emissions.

Compared to other key cities, the average summer PL-VOCs in Ganzhou was higher than that reported for Guangzhou summer (5.1 ppbv) [44] and Zhengzhou summer (10.2 ppbv) [51], and was similar to the level in Tianjin from spring to summer (17.8 ppbv) [33]. It is important to note that direct comparison of the absolute PL-VOC values between different cities requires caution due to differences in estimation methodologies and geographical conditions. However, the species composition of PL-VOCs across these cities is relatively consistent, with alkenes contributing approximately 44–87% of PL-VOCs. In this study, the contributions of alkenes to PL-VOCs in Ganzhou across all seasons fall within this reported range.

3.2.2. Ozone Formation Potential (OFP)

Figure 6a presents the calculated OFP based on observed, initial, and photochemical-loss VOC concentrations (namely OC-OFP, IC-OFP, and PL-OFP) in four seasons. On average, a significant disparity was observed between the OC-OFP and IC-OFP, with the former being approximately 70.8% lower. This discrepancy is driven by the substantial contribution of photochemically consumed VOCs to ozone formation. The OC-OFP demonstrated relatively modest seasonal fluctuations, with its maximum in winter (159.7 μg/m³) and minimum in summer (114.4 μg/m³). In contrast, the reconstructed IC-OFP was highest in summer (528.7 μg/m³) and lowest in spring (345.0 μg/m³), primarily due to the larger PL-OFP in summer (414.4 μg/m³).

Figure 6b further shows the seasonal OFP contributions by VOC groups, with detailed top 10 species-level information summarized in Table S1. For the OC-OFP, alkenes were the primary contributors (average: 48.3%), followed by OVOCs (17.1%), alkanes (17.0%), and aromatic hydrocarbons (15.8%). Despite their lower concentration than alkanes, the higher MIR coefficients of alkenes (average: 11.0) than alkanes (1.2) resulted in their dominant OFP. After reconstruction, the contribution of alkenes to the IC-OFP was even more pronounced (71.1%), aligning with the fact that they experienced the most significant photochemical loss (Figure 5b). OVOCs (13.0%), aromatic hydrocarbons (9.5%), and alkanes (5.8%) followed in contribution to IC-OFP.

From a practical perspective for ozone mitigation, targeting high-contributing VOC species within the photochemical loss fraction may be key to suppressing ozone formation. Neglecting photochemical consumption and relying solely on OC-OFP for assessment would lead to a substantial underestimation of the IC-OFP by approximately 65.8–78.4%. Strikingly, this underestimated portion (i.e., PL-OFP) was about 300% of the OC-OFP, highlighting the limitations of previous studies that overlooked the analysis of key species influencing photochemical loss.

Comparisons of the top ten OFP-contributing species were conducted between the OC-VOCs and PL-VOCs fractions (

Table 1. Comparisons of the top ten OFP and SOAFP contributing species for OC-VOCs and PL-VOCs in Ganzhou.

	OC-VOCs				PL-VOCs
Rank	Species	OFP (μg/m3)	Species	SOAFP (μg/m3)	Species	OFP (μg/m3)	Species	SOAFP (μg/m3)
1	1-Pentene	20.1	Toluene	9.99	trans-2-Butene	78	Styrene	27.85
2	Ethylene	16.33	Benzene	6.04	1-Pentene	44.48	m,p-Xylene	2.36
3	Propanal	14.85	m,p-Xylene	3.35	1,3-Butadiene	32.93	Toluene	1.77
4	1-Butene	10.22	Ethylbenzene	2.71	1-Butene	30.35	trans-2-Butene	1.11
5	Toluene	7.4	o-Xylene	2.05	Propanal	18.97	o-Xylene	0.99
6	Isoprene	6.85	Styrene	1.78	cis-2-Butene	17.71	Ethylbenzene	0.63
7	m,p-Xylene	6.64	Benzaldehyde	0.58	cis-2-Pentene	16.81	1,3,5-Trimethylbenzene	0.39
8	Propylene	6.3	1-Ethyl-2-methylbenzene	0.39	trans-2-Pentene	16.36	Benzaldehyde	0.28
9	Cyclohexane	4.76	n-Propyl benzene	0.24	Isoprene	11.55	cis-2-Pentene	0.27
10	Isopentane	4.42	p-Diethylbenzene	0.19	Propylene	10.83	trans-2-Pentene	0.26

). The analysis revealed that isoprene, propylene, 1-butene, propanal, and 1-pentene are identified as shared high-contributing species, indicating their strong ozone formation potential even after undergoing substantial photochemical loss. Conversely, the PL-OFP spectrum featured unique, highly reactive species like trans-2-butene, 1,3-butadiene, cis-2-butene, cis-2-pentene, and trans-2-pentene. Their exceptionally high reaction rates mean they are rapidly consumed in the atmosphere and thus rarely observed at high concentrations at monitoring sites. However, their prominence in the PL-OFP profile strongly suggests they act as key precursor drivers in the photochemical ozone production mechanism.

3.2.3. Secondary Organic Aerosol Formation Potential (SOAFP)

Figure 7a shows the seasonal variations in SOAFP derived from observed, initial, and photochemical loss VOC concentrations (namely OC-SOAFP, IC-SOAFP, and PL-SOAFP). Both OC-SOAFP and IC-SOAFP exhibited distinct seasonal variations, albeit with different patterns. The OC-SOAFP peaked in winter (35.6 μg/m³), followed by spring (31.8 μg/m³) and autumn (27.7 μg/m³), with summer showing the lowest value (19.2 μg/m³). In contrast, the reconstructed IC-SOAFP was highest in spring (82.1 μg/m³), followed by winter (75.8 μg/m³) and summer (55.6 μg/m³), and lowest in autumn (48.5 μg/m³). This pattern is primarily attributed to the larger PL-SOAFP in spring (50.3 μg/m³) and winter (40.2 μg/m³), which aligns reasonably well with the seasonal variation in PM₂.₅ (Figure 3b), suggesting that secondary organic formation processes contribute significantly to wintertime particulate pollution. However, it is noteworthy that the observed PM_2.5 concentration was lowest in summer, whereas the PL-SOAFP was lowest in autumn. This discrepancy arises because ambient PM_2.5 levels are also strongly influenced by primary emissions and meteorological conditions.

The contributions of different VOC groups to the SOAFP are presented in Figure 7b, with detailed top 10 species-level information provided in Table S2. Aromatic hydrocarbons overwhelmingly dominated both IC-SOAFP and OC-SOAFP (>90%), while contributions from other species were minor. This is a direct consequence of the higher SOAP coefficients of aromatic species. For example, the average SOAP coefficient of aromatic hydrocarbons is 0.9, whereas that of other VOC groups (mainly alkanes and OVOCs) is only 0.03. This compositional feature contrasts sharply with OFP contributions. Although aromatic hydrocarbons are less photochemically reactive towards ozone formation (Figure 6b), their higher molecular weights and oxidation products—such as nitrocatechols and dinitrophenols, which have low vapor pressures—are efficient SOA precursors that readily partition into the particle phase, thereby promoting SOA formation [37,52]. An analysis of the top ten species contributing to OC/PL-SOAFP (Table 1) identified toluene, m/p-xylene, o-xylene, ethylbenzene, styrene, and benzaldehyde as common high contributors. Notably, under nighttime conditions where NO₃ radicals become the dominant oxidant and RO₂ radical reactions are more active, these aromatic species retain significant potential for conversion, contributing to nocturnal SOA formation [53].

A comparison of OFP and SOAFP reveals that the PL-SOAFP was generally 100–150% of the OC-SOAFP, a proportion markedly lower than the corresponding ratio for OFP (~300%), indicating a lesser relative role of photochemically consumed VOCs in SOA formation compared to ozone formation. This difference can be explained by two key factors. First, SOA formation typically involves prolonged multi-step photochemical oxidation pathways (often requiring 2–4 days to approach theoretical maximum yield), which is considerably longer than the timescale for ozone production. Second, the VOC categories associated with high SOA potential differ from those driving high OFP. PL-VOCs are predominantly composed of alkenes and OVOCs, which are key drivers for O₃ but less so for SOA. Furthermore, the relationship between SOA yield and atmospheric oxidant exposure (e.g., OH radical) is non-linear; both excessively high- and low-oxidation conditions can suppress SOA formation [53,54]. This nonlinearity prolongs the timescale for VOC-to-SOA conversion, allowing SOA-relevant species to remain more abundant in the OC-VOCs.

3.3. Source Apportionment of VOCs in Ganzhou

3.3.1. Species Selection and Factor Identification

Given the distinct prevailing wind directions in summer and winter (Figure 2), source apportionment analyses were conducted in January and August using the PMF model based on OC-PMF and reconstructed IC-PMF in Ganzhou. It should be emphasized these two months were selected as representative seasonal case studies to illustrate source characteristics under contrasting meteorological conditions, rather than to quantify annual average source contributions.

Species selection for the model followed these principles: (1) priority given to species with higher concentrations; (2) inclusion of species indicative of specific VOC sources; (3) exclusion of species with a signal-to-noise ratio (S/N) below 0.5; and (4) careful inclusion of highly reactive species, ensuring it did not hinder the factor matching between OC-PMF and IC-PMF solutions. Multiple runs were conducted testing 2–8 factors and 20–40 species. The optimal solution was determined based on factor interpretability, the distribution of scaled residuals, and the Q_true/Q_exp trend with increasing factor number (Table S3). Accordingly, we retained a 3-factor solution for January and a 6-factor solution for August for both OC-PMF and IC-PMF. Model diagnostics indicated stable results: approximately 70% of species showed an R² > 0.6 between observed and modeled values (Table S7), most species residuals were within ±3, and repeated random-start runs supported solution stability (Table S4). Rotational ambiguity was evaluated using F_peak and DISP diagnostics; within |F_peak| ≤ 1 the changes in Q_robust were modest and DISP showed no factor swaps (Tables S5 and S6). The factor profiles resolved by IC-PMF and OC-PMF were highly consistent, facilitating subsequent comparative analysis.

The final resolved sources with OC-PMF are shown as an example in Figure 8. Six sources were identified in August (Figure 8a): Factor 1 was identified as vehicle emissions, characterized by significant contributions from propane, n-butane, isobutane, ethene, and acetylene. The propane/n-butane ratio close to 1 further supports its association with gasoline vehicle emissions rather than LPG [55,56]. Factor 2 was identified as industrial process emissions, featured by high loadings of isopropanol, ethyl acetate, and some alkanes. These OVOCs are commonly associated with industries such as printing, coating, and electronics [57,58]. Factor 3 was recognized as solvent use source, enriched in chloromethane, dichloromethane, ethylbenzene, and m/p-xylene, as these are typical components in solvents [18,26]. Factor 4 was identified as biogenic emissions, overwhelmingly dominated by isoprene (a typical biogenic tracer emitted by vegetation) without co-emitted combustion tracers. Factor 5 was identified as gasoline evaporation emissions, marked by isopentane and C₂–C₆ alkanes, which were closely associated with fuel evaporation and vehicle-related sources [59]. Factor 6 was determined as combustion source, characterized by contributions from alkanes, halogenated hydrocarbons (e.g., chloromethane, 1,2-dichloroethane), ethene, and acetylene, which those species possibly related to biomass or fossil fuel burning [47]. In the January PMF results (Figure 8b), the identified factors were fewer in number than those in August. Factor 1 was identified as combustion emissions, Factor 2 corresponded to vehicle exhaust, and Factor 3 was identified as an industry-related factor based on pronounced OVOC and aromatic contributions. Note that in August, solvent use and industrial process emissions were resolved as two distinct factors, while in January we report it as an aggregated industry-related mixed factor since several indicative species are shared between these two categories.

A comparison between the January and August month-case studies reveals that August exhibited additional contributions from biogenic emissions (10.6%) and gasoline evaporation (14.0%). In January, the contributions from vehicle exhaust and combustion increased substantially. The contribution of combustion rose from 25.1% in August to 57.2% in January, while vehicle exhaust increased from 17.0 to 31.0%, reflecting enhanced fuel consumption and traffic emissions during the colder season. In contrast, the combined contribution of solvent use and industrial process factors decreased from 33.3% in August to 11.8% in January (industry-related mixed factor). This indicates a stronger industry-related VOC influence in August. According to Ganzhou’s 2023 statistical yearbook [60], furniture manufacturing accounts for approximately 20% of the local industrial structure. Solvents used in this industry release halogenated and aromatic hydrocarbons, which have relatively long atmospheric lifetimes and can accumulate in the background environment. Consequently, based on the January–August month-case studies, control strategies in Ganzhou may prioritize traffic and combustion sources during the cold season, while targeting highly reactive VOCs from gasoline evaporation and industrial fugitive sources during the warm season.

3.3.2. Influence of Photochemical Loss on Source Apportionment

Figure 9 presents the comparison between IC-PMF and OC-PMF results. During August, the IC-PMF results showed increased contributions from biogenic, gasoline evaporation, and industrial process emissions, while the shares of vehicle, solvent use and combustion sources decreased. Specifically, the contributions of industrial process, gasoline evaporation and biogenic sources in OC-PMF were 20.6%, 14.0%, 10.6%, respectively, increasing to 22.7%, 17.7%, and 16.1% in IC-PMF. In contrast, the contributions from combustion, vehicle exhaust, and solvent use in IC-PMF decreased by 6.8%, 2.4% and 2.0%, respectively (Figure 9a,b). This redistribution stems from the varying reactivities of VOCs from different sources. As shown in

Table 2. Comparisons of factor concentrations, photochemical loss, and overall factor loss rates between IC-PMF and OC-PMF.

Month	Factor	OC Average Concentration (ppbv)	IC Average Concentration (ppbv)	Photochemical Loss Rate (%)	Overall Factor Loss Rate (%)
August	Biogenic emissions	1.41	2.78	49.41	23.25
	Gasoline evaporation	1.85	3.05	39.46
	Process emissions	2.75	3.91	29.86
	Vehicle emissions	2.25	2.52	10.92
	Combustion emissions	3.33	3.17	−0.05
	Solvent use	1.69	1.85	8.66
January	Combustion emissions	12.44	13.85	10.2	12.44
	Vehicle emissions	6.73	7.39	8.88
	Industry-related mixed factor (solvent + process)	2.56	3.58	28.47

, the photochemical loss of VOC emissions showed a clear August–January contrast, with a loss rate of 23.3% in August, significantly higher than the 12.4% observed in January. In August, biogenic emissions (49.4%), gasoline evaporation (39.5%), and industrial process emissions (29.9%) showed the highest loss rates, while vehicle emissions (10.9%), solvent use (8.7%), and combustion sources (−0.05%) had relatively lower loss rates. The slightly negative photochemical loss rate calculated for combustion sources in August is likely within the margin of uncertainty of the PMF model solution and the photochemical age method, rather than indicating net photochemical production. In January, except for combustion sources, the loss rates of all factors were generally lower than those in August. Among them, the industry-related mixed factor (solvent + process) still exhibited a relatively high loss rate (28.47%), and its contribution increased from 11.8% in OC-PMF to 14.4% in IC-PMF, reflecting a minor shift in the source contribution structure (Figure 9c,d).

Figure 10 further shows the diurnal variation in IC-PMF and OC-PMF source factors’ contribution. In August, the difference between IC-PMF and OC-PMF diurnal profiles for biogenic emissions was most pronounced between 11:00 and 17:00. The IC-PMF diurnal profile for industrial process emissions showed a distinct inverted U-shaped pattern from 06:00 to 23:00. Conversely, the differences between IC-PMF and OC-PMF diurnal profiles were relatively small for combustion, vehicle exhaust, and solvent use sources, which are dominated by low-reactivity VOCs such as alkanes. The diurnal profile for gasoline evaporation emissions exhibited an overall upward shift in the IC-based results, with slight enhancements during morning and evening rush hours, indicating superimposed peak characteristics on the overall increase. This suggests that enhanced emissions likely lead to greater photochemical loss related to IC-VOCs (Figure 10a–f). In January, the diurnal variations for all sources primarily showed a uniform upward shift in IC-PMF relative to OC-PMF, lacking the pronounced midday-to-afternoon peak features observed in August (Figure 10g–i).

These observations can be reasonably explained by differences in VOC-component reactivity. A typical PMF factor contains both highly reactive species (e.g., isoprene, OVOCs, and some alkenes) and low-reactivity or inert species (e.g., alkanes and some aromatic hydrocarbons). After photochemical correction, the concentrations and relative contributions of the reactive species within each factor significantly increased (e.g., the biogenic emission factor rich in isoprene showed a marked increase). For factors dominated by highly reactive species (e.g., industrial processes rich in OVOCs or biogenic emissions), intense photochemical activity around midday leads to rapid consumption of these species, resulting in significant divergence between the IC-PMF and OC-PMF diurnal profiles (Figure 10c,e). Conversely, for factors dominated by inert or low-reactivity species, VOCs undergo slower photochemical loss, and their IC-PMF/OC-PMF diurnal profiles primarily exhibit an overall upward shift.

Previous studies show that climate warming and increased CO₂ levels may enhance biogenic VOC (e.g., isoprene) emissions from subtropical vegetation [61]. Our finding implies that this future increase in biogenic precursor potential could be even more severely underestimated by conventional observation-based methods, leading to a potential “double underestimation” of future ozone risks (both from increased emissions and from the unaccounted-for reactivity of those emissions). This underscores the critical need to incorporate photochemical-loss correction into long-term air quality planning in warming subtropical regions.

4. Conclusions

4.1. General Findings

(1)

The annual average VOC concentrations in 2023 were 22.6 ± 13.17 ppbv, substantially lower than that in typical industrial cities. Concentrations were highest in winter and lowest in summer, driven predominantly by emission intensity and meteorological conditions. Alkanes were the dominant group (44.7%), followed by alkenes (16.1%), halocarbons (16.0%), and OVOCs (11.5%). The average concentration of IC-VOCs (36.2 ppbv) was approximately 60% higher than OC-VOCs, confirming significant atmospheric photochemical consumption. PL-VOCs peaked in summer (16.9 ppbv), accounting for 50% of the contemporary IC-VOCs. Alkenes were the dominant component undergoing photochemical loss, constituting 72% of PL-VOCs. Their true contribution in IC-VOCs (37%) was substantially higher than their observed share in OC-VOCs (16%), demonstrating that highly reactive alkenes are severely underestimated in conventional observations.

(2)

OFP analysis revealed that IC-OFP (453 μg/m³) was significantly higher than OC-OFP (132 μg/m³). PL-OFP (321 μg/m³) was about 2–3 times the OC-OFP, with alkenes contributing 81%. Short-lived, highly reactive species like isoprene, propylene, and trans-2-butene were key precursors. Neglecting photochemical loss would lead to a 67–78% underestimation of OFP. SOAFP followed the order IC-SOAFP (65.4 μg/m³) > PL-SOAFP (36.9 μg/m³) > OC-SOAFP (28.6 μg/m³), with higher levels in spring and winter. Aromatic hydrocarbons were the overwhelmingly dominant component (>85%), with toluene, xylenes, styrene, and benzaldehyde being the key contributors.

(3)

Based on the January (winter) and August (summer) month-case studies, the PMF model identified six major sources in August—combustion, vehicle exhaust, industrial processes, gasoline evaporation, solvent use, and biogenic emissions—whereas only three sources—combustion, vehicle exhaust, and industrial processes—were resolved in January. Source contributions showed a pronounced contrast between the January and August cases: biogenic emissions (10.6%) and gasoline evaporation (14.0%) were prominent in August, whereas combustion (57.2%) and vehicle exhaust (31.0%) dominated in January, reflecting increased fuel consumption and traffic activity during colder months. Photochemical loss plays a critical role in reshaping VOC source apportionment, particularly in August. The IC-PMF results revealed that highly reactive sources—such as biogenic emissions, gasoline evaporation, and industrial processes—were underestimated in the OC-PMF analysis. Conversely, sources dominated by less reactive species, including vehicle emissions, solvent use, and combustion, were slightly overestimated in the OC-PMF. These results underscore the importance of accounting for photochemical loss in VOC source apportionment, particularly during periods with strong photochemical activity.

4.2. Policy Implications

After accounting for photochemical loss, our OFP and SOAFP results identify priority VOC species for emission control. For ozone mitigation, priority could be given to reducing propene and 1-butene, which contribute strongly to OFP in both OC-VOCs and PL-VOCs. Due to the consistently high OFP contributions from such anthropogenic alkenes (e.g., propene, 1-butene) and aromatics, coupled with the typical urban NO_x levels in Ganzhou (Figure 3c), suggest that the city’s ozone production is likely in a VOC-sensitive regime, especially during high-O₃ episodes. This directly implies that controlling emissions of the identified key VOC species (from traffic, evaporation, and solvents) should be a high-priority and effective strategy in Ganzhou. For SOA mitigation, controls may focus on toluene, xylenes, and ethylbenzene, the dominant aromatics with the highest SOAFP.

Because individual VOC species can originate from multiple sources, we translate these species-level priorities into source-oriented directions using the PMF results. The January–August month-case comparison suggests that wintertime efforts may place greater emphasis on vehicle exhaust and combustion, whereas during warm season more attention may be warranted for gasoline evaporation and industry-related sources resolved in August, which are more likely to be underestimated when photochemical loss is ignored. Practically, these priorities point to strengthening controls on traffic/combustion emissions, reducing evaporative losses along the fuel supply chain, and improving management of industry/solvent-related emissions using locally applicable regulatory measures.

4.3. Limitations

This study has several limitations that should be considered when interpreting the results. First, the analysis relies on measurements from a single urban background monitoring site. While the site is representative of mixed residential-commercial areas, the findings may not fully capture spatial heterogeneity or be directly transferable to distinct functional zones (e.g., industrial parks or remote residential areas) within Ganzhou. Second, we assessed the uncertainty associated with reconstructed IC-VOCs (e.g., the choice of $(C_E/C_X{)}_{initial}$) using sensitivity tests, and we further propagated this perturbation to the IC-based OFP and SOAFP calculations to quantify its influence on these reactivity/yield-weighted metrics. Results showed that a +10% change in $(C_E/C_X{)}_{initial}$ decreased IC-OFP and IC-SOAFP by ~26% and ~22%, respectively, whereas a −10% change increased them by ~46% and ~39%. Despite significant downstream propagation of uncertainty (22 ~ 46%), it does not affect the relative contributions or rankings among species. However, we did not attempt an end-to-end uncertainty propagation to PMF-derived source contributions, because PMF uncertainty is additionally dominated by model-structure factors such as uncertainty/error estimation, factor-number selection, and rotational ambiguity. Accordingly, source apportionment results are mainly interpreted in terms of relative contrasts and source/species prioritization rather than as precise absolute values. Finally, the PMF source apportionment was conducted for January and August as seasonal case studies; these results illustrate key mechanisms but do not constitute a quantified annual average source contribution assessment.

4.4. Future Research

Building on the findings and limitations of this work, several promising directions for future research are proposed. First, expanding the observational network to include multiple sites across different functional zones (e.g., industrial, traffic, and suburban background) would be crucial for assessing the spatial representativeness of VOC characteristics and sources city-wide. Second, conducting long-term multi-year measurements is essential to distinguish typical seasonal patterns from interannual variability and to evaluate the impact of emission control policies over time. Third, future efforts should focus on developing and applying locally derived or adjusted reactivity parameters (e.g., MIR, SOAP) that reflect the specific NO_x regimes and oxidation environments of subtropical Chinese cities to refine OFP and SOAFP assessments.