Source Apportionment of VOCs and O 3 Production Sensitivity at Coastal and Inland Sites of Southeast China

Urbanization and industrialization levels have a significant impact on the pollution of O 3 and its precursors. However, current studies of VOC sources and O 3 formation sensitivity in the regions with different urbanization and industrialization levels remained limited. Therefore, offline and online VOC observations were conducted at coastal sites (CS) and inland sites (IS) of southeast China to analyze spatial-temporal characteristics, source apportionment of VOCs based on the positive matrix factorization (PMF) model, and their effects on O 3 formation using the observation-based model (OBM). The results showed that the average concentrations of TVOCs at CS and IS were 49.1 ± 14.4 and 28.4 ± 9.6 ppb, respectively, with higher levels in autumn compared to those in spring and summer. Alkene species contributed the largest to the OH radical loss rate (L OH ) and ozone formation potential (OFP). The contributions of vehicle exhaust and combustion sources, industrial sources, solvent usage, biogenic sources, and fuel evaporation at CS and IS were 42.2% and 34.5%, 22.4% and 18.2%, 12.7% and 4.5%, 11.7% and 19.4%, and 11.0% and 23.4%, respectively. Meanwhile, vehicle exhaust and combustion sources (31.8%), fuel evaporation (21.3%) were the major contributors to O 3 formation in Ningde. The results of sensitivity analysis indicated that O 3 formation at CS was mainly VOC control in spring and autumn, and controlled by both VOCs and NO x in summer, but the VOCs is the key factor for the O 3 formation at IS, and the emission control of alkenes and aromatics was conductive to decrease O 3 levels. The scenario analysis suggested that the 20% reduction of VOC concentrations and 3% reduction of NO x concentrations could realize the 5% reduction of O 3 concentrations. This study might enhance the understanding of O 3 and its precursors in southeast China with different urbanization levels, as well as the emission reduction strategies.


2
Text S1 The introduction of the positive matrix factorization (PMF) model and the rationality analysis of the result.
The VOCs data set can be considered as a matrix xij shown in equation (1), where, i, j, p stands for the number of samples, species, sources, respectively, f is the species profile of every source, g is the contribution of k source in the i sample, and e is the residual matrix.
Besides, the Q value is used to evaluate the uncertainty (uij) of the PMF result, as shown in equation (2). (2) where, uij represents the uncertainty of j specie in the i sample.
The PMF model needs to input two files, including a matrix of the concentrations (Conc.) of VOC species and a matrix of concentration uncertainties.
The uncertainty (Unc) is calculated by equations (3) and (4) where, Unc is the uncertainty, EF is the error fraction, cij is the concentrations of VOC species, and MDL is the method detection limit.
In this study, the VOC data at different sampling sites (including 3 sites for the inland site and 5 for the costal site) were combined together to form the input data into the PMF model because the samples were simultaneously collected at eight sampling sites and the source category at CS and IS differed little. 23 VOC species were selected and they are all typical indicators of VOC sources. The rationality of the PMF result was usually evaluated by the Qtrue/Qexp value and the bootstrap 3 method. 4-9 factors were selected to seek for the optimal factor number. In this study, the Qtrue/Qexp value decreased gradually with the increase of the factor number, and its decreased amplitude became slight when the factor number increased to 6 (Table S2). Therefore, the factor number was determined as 6 after comparing the Qtrue/Qexp value. The bootstrap of 6 factors was performed with the minimum correlation r 2 (0.6), and the mapping of 6 factors was more than 85% (Table S3), indicating that the result was accepted.

Text S2 Photochemical reactive activity analysis
The loss rates of VOCs that react with the OH radical (L OH) can evaluate qualitatively the reactivity of VOC species. LOH was calculated by multiplying the VOC species concentration by its corresponding rate coefficient of the reaction with the OH radical (k OH) (Atkinson and Arey, 2003;Atkinson et al., 2006), as shown in equation (5).
The ozone formation potential (OFP) is used to evaluate the contributions of different VOC species to O3 production, which is calculated by multiplying the VOC species concentration (VOCi) by its corresponding maximum incremental reactivity (MIRi) (Carter, 2010;Zhang et al., 2021), as shown in equation (6).

Text S3 Descriptions of Online VOCs measurement
To determine the sensitivity of O3 production in coastal regions, online VOC measurements were carried out in the Ningde Meteorological Bureau using a gas chromatography system equipped with a mass spectrometer and flame ionization detector (GC-MS/FID). After water and CO2 were removed, VOC samples were firstly captured and then thermally desorbed. The temperature program duration was 24.5 min, with an initial temperature of 40 °C and a final temperature of 180 °C.
The low-carbon (C2-C5) and the high-carbon (C5-C10) VOCs were quantified using the FID and MS, respectively. Daily calibration was performed each day at 23:00 using 4 ppb standard gases to test the sensitivity and accuracy of the measurement system. The standard deviation was less than ± 10% for VOCs.

Text S4 The introduction of different VOC sources
Factor 1 was identified as solvent usage, characterized by high percentages of aromatics, such as toluene, ethylbenzene, m/p-xylene, and o-xylene (Yuan et al., 2010). Factor 2 is characterized by high percentages of n-hexane, styrene, and chloroform. Generally, n-hexane, styrene, and chloroform are mainly used as an important raw material used for organic synthesis. Factor 3 has high proportions of n-butane and n-butene, typical indicators for petrochemical industries (Kwon et al., 2007). Therefore, factor 2 and factor 3 were regarded as industrial production.
Factor 4 has high percentages of i-butane, propane, and propene, so it is considered as vehicle emissions (Cai et al., 2010;An et al., 2014). Besides, high percentages of chloromethane are associated with combustion sources. Therefore, source 3 was identified as vehicle exhaust and combustion sources. Factor 5 was assigned to fuel evaporation, characterized by a high proportion of i-pentane, n-pentane, 3methylpentane, i-butane, and propene (Chan et al., 2006;Liu et al., 2017). Factor 6 was designated as biogenic sources because it has high percentages of isoprene, a typical indicator for plant emissions (An et al., 2014).