A number of 24 h integrated PM2.5 samples were collected in four seasons from November 2015 to November 2016 (i.e., from 30 November to 31 December 2015 for winter, 19 April to 19 May 2016 for spring, 1 to 31 July 2016 for summer, and 9 October to 15 November 2016 for fall) on the campus of the Institute of Earth Environment, Chinese Academy of Sciences (IEECAS; 34.22∘ N, 109.01∘ E), in Xi'an, China. The sampling site is an urban background site surrounded by residential areas and has no obvious industrial activities. A total of 112 samples were collected on pre-baked (780 ∘C, 3 h) quartz-fiber filters (20.3×25.4 cm; Whatman, QM-A, Clifton, NJ, USA) by a high-volume PM2.5 sampler (Tisch, Cleveland, OH, USA) operating at 1.05 m3 min-1. The filter samples were stored at -20 ∘C until laboratory analysis.

The concentration of organic carbon (OC) was measured by a thermal–optical carbon analyzer (DRI, Model 2001, Atmoslytic Inc., Calabasas, CA, USA) with the IMPROVE-A protocol (Chow et al., 2011). A total of 10 NACs and 19 organic markers (see Table S1 in the Supplement) were quantified by a gas chromatograph mass spectrometer (GC-MS) using a well-established approach (e.g., Wang et al., 2006; Al-Naiema and Stone, 2017) and the details are described in Yuan et al. (2020). At least one blank filter sample was analyzed for every 10 ambient samples. Baseline separation with symmetrical peak shapes was achieved for the measured NACs (Fig. 1). The linear ranges, instrument detection limit (IDL), instrument quantitation limit (IQL), extraction efficiency, and regression coefficients for the measured NACs are shown in Table S2. The response of calibration curves for the NACs was linear (R2≥0.995) from 10 to 5000 µgL-1. The IDL ranged from 2 to 20 µgL-1 except for 5-nitrosalicylic acid (53 µgL-1). The IQL ranged from below 10 to 70 µgL-1 except for 5-nitrosalicylic acid (>100 µgL-1). The IDL and IQL are comparable to those in Al-Naiema and Stone (2017) (2.7–14.9 µgL-1 for IDL and 8.8–50 µgL-1 for IQL) and are sufficient for the quantification of our samples.

A UV-Vis spectrophotometer equipped with a Liquid Waveguide Capillary Cell (LWCC-3100, World Precision Instruments, Sarasota, FL, USA) was used to measure the light absorption of methanol-soluble BrC and NAC standards, following the method established by Hecobian et al. (2010). The absorption coefficient (Absλ; M m-1) can be obtained from the measured absorption data by Eq. (1): 1Absλ=Aλ-A700VlVa×Lln⁡(10), where A700 is the absorption at 700 nm used to correct for baseline drift, Vl is the volume of methanol used for extracting the filter, Va is the volume of sampled air, L is 0.94 m for the optical path length used in the LWCC-3100, and ln(10) is used to convert the absorption coefficient from log base 10 to natural logarithm.

The mass absorption efficiency (MAE; m2 g-1) of NAC standards in the methanol solvent at a wavelength of λ can be calculated as in Laskin et al. (2015): 2MAENAC,λ=Aλ-A700L×Cln⁡(10), where C (µgmL-1) is the concentration of the NAC standards in the methanol solvent.

The light absorption contribution of NACs to BrC at a wavelength of λ (ContNAC/BrC,λ) can be obtained using Eq. (3): 3ContNAC/BrC,λ=MAENAC,λ×CNACAbsBrC,λ, where the CNAC (µgm-3) is the atmospheric concentration of NACs, and the AbsBrC,λ is the absorption coefficient of BrC at a wavelength of λ.

The sources of the NACs were resolved by the PMF receptor model, which was performed by the multilinear engine (ME-2; Paatero, 1997) through the Source Finder (SoFi) interface encoded in Igor Wavemetrics (Canonaco et al., 2013). The input species include 5 to 10 NACs (as the number of NACs detected varies among seasons) and 19 additional organic tracer species (see Table S1 in the Supplement), with relative standard deviations (RSDs) < 10 %. These include phthalic acid for secondary formation, picene for coal combustion, hopanes for vehicle emission, fluoranthene, pyrene, chrysene, benzo(a)pyrene, benzo(a)anthracene, benzo(k)fluoranthene, benzo(b)fluoranthene, benzo(ghi)perylene, and indeno(1,2,3-cd)pyrene for combustion emission and vanillin, vanillic acid, syringyl acetone, and levoglucosan for biomass burning. To separate the source profiles clearly, the contribution of those markers unrelated to a certain source was set to 0 in the respective source profile (see Table S3).

To better understand the source origins of the NACs, air mass origins during the sampling period were derived from backward-trajectory analysis. This method was used in trajectory clustering based on the GIS-based software TrajStat (Wang et al., 2009). The archived meteorological data were obtained from the National Center for Environmental Prediction's Global Data Assimilation System (GDAS). According to the lifetimes of the different secondary species (Wojcik and Chang, 1997; Chow et al., 2015), in this study, 72 h backward trajectories terminated at a height of 500 m above ground level were calculated during the study period. The trajectories were calculated every 12 h, with starting times at 09:00 and 21:00 local time.

The concentrations of the NACs show clear seasonal differences, with the highest mean values in winter, followed by fall, spring, and summer (see Fig. 2). The concentration ranges of total NACs were 1.4–3.4 ng m-3 (spring), 0.1–3.8 ng m-3 (summer), 1.6–44 ng m-3 (fall), and 20–127 ng m-3 (winter). The average concentrations were 2.1 ± 0.6, 1.1 ± 0.8, 12.9 ± 11.6, and 56 ± 23 ng m-3, respectively (see Table S4). Nitrophenols (4-nitrophenol, 2-methyl-4-nitrophenol, 3-methyl-4-nitrophenol, 2,6-dimethyl-4-nitrophenol) and nitrocatechols (4-nitrocatechol, 3-methyl-5-nitrocatechol, 4-methyl-5-nitrocatechol) show the highest concentrations in winter and the lowest in summer, while nitrosalicylic acids (3-nitrosalicylic acid, 5-nitrosalicylic acid) show the highest concentrations in winter and the lowest in spring. The average ratios between wintertime and summertime concentrations are a factor of about 40 for nitrophenols, 175 for nitrocatechols, and 21 for nitrosalicylic acids. The large seasonal differences in NAC concentrations might be due to the differences in sources, emission strength, and atmospheric formation processes, as discussed below. Table 1 summarizes the NAC concentrations measured in this study together with those measured in Europe, the USA, and other places in Asia. In general, the NAC concentrations in winter are higher than those in summer, and the observed concentrations of different species are higher in Asia than in Europe and the USA. The only exception is a study in Ljubljana, Slovenia, which shows that in winter nitrocatechol concentrations are higher than those in Asia, likely due to strong biomass-burning activities (Kitanovski et al., 2012). The elevated concentrations of NACs in Asia suggest that NACs may have a significant impact on regional climate and air quality in Asia due to their optical and chemical characteristics, as discussed below.

As shown in Fig. 3a, among all measured NACs, 4-nitrophenol, 2-methyl-4-nitrophenol, 3-methyl-4-nitrophenol, 4-nitrocatechol, and 5-nitrosalicylic acid were detected in four seasons; 3-methyl-5-nitrocatechol and 4-methyl-5-nitrocatechol in fall and winter; and 2,6-dimethyl-4-nitrophenol, 3-nitrosalicylic acid, and 4-nitro-1-naphthol only in winter. In general, 4-nitrophenol and 4-nitrocatechol had elevated concentrations in all seasons, which is consistent with other observations (Chow et al., 2015; Ikemori et al., 2019) and might be related to their larger emissions or formation and longer atmospheric lifetime than other NACs (Harrison et al., 2005; Chow et al., 2015; Finewax et al., 2018; Wang et al., 2019; Lu et al., 2019a). For example, Lu et al. (2019a) measured the emission of NACs from coal combustion and found that the emission factors of 4-nitrocatechol were about 1.5–6 times higher than those of other NACs. Wang et al. (2019) quantified the concentrations of 4-nitrophenol and 4-nitrocatechol formed under high-NOx and high-anthropogenic-volatile-organic-compound (VOC) conditions and found that they are about 3–7 times higher than those of other NACs. The concentration of 2-methyl-4-nitrophenol was higher than that of 3-methyl-4-nitrophenol in all seasons, which is similar to previous studies (Kitanovski et al., 2012; Chow et al., 2015; Teich et al., 2017; Ikemori et al., 2019) and likely due to the efficient formation of 2-methyl-4-nitrophenol from photochemical oxidation of VOCs in the presence of NO2 (Lin et al., 2015; Wang et al., 2019). It should be noted that the contribution of 5-nitrosalicylic acid (27 %) to total NAC mass in summer is much higher than in other seasons (4 %–13 %), suggesting that 5-nitrosalicylic acid is mainly produced by secondary formation, for example, through nitration of salicylic acid (M. Li et al., 2020) and photochemical oxidation of toluene in the presence of NOx (Jang and Kamens, 2001; Wang et al., 2018).

Correlation analysis was conducted among the NACs measured in this study (Table S5). The four nitrophenols were positively correlated with each other (r2=0.52–0.98), and the three nitrocatechols were also highly correlated with each other (r2=0.94–0.96), indicating that the different nitrophenols and nitrocatechols might have similar sources or origins. Previous studies showed that 4-nitrophenol was mainly from primary emission of biomass burning (Wang et al., 2017), and 3-methyl-5-nitrocatechol and 4-methyl-5-nitrocatechol were identified as secondary products from biomass burning (Iinuma et al., 2010). Positive correlations were also observed between nitrophenols and nitrocatechols (r2=0.59–0.90), suggesting that they were partly of similar sources or formation processes. For example, both nitrophenols and nitrocatechols can be emitted through biomass burning (Wang et al., 2017) and coal combustion (Lu et al., 2019a) and can be formed by photochemical oxidation of VOCs in the presence of NO2 (Wang et al., 2019). However, for nitrosalicylic acids, the correlation between 3-nitrosalicylic acid and 5-nitrosalicylic acid was weak (r2=0.29). This is because 5-nitrosalicylic acid is mainly from secondary formation by nitration of salicylic acids, while 3-nitrosalicylic acid is mainly from combustion emission (Wang et al., 2017; M. Li et al., 2020). The correlations of nitrosalicylic acids with nitrophenols (r2=0.01–0.13) and with nitrocatechols (r2=0.04–0.25) were also weak, suggesting that they may have different sources or formation processes. Nitrosalicylic acids were dominated by 5-nitrosalicylic acids, which are mainly from secondary formation (Andreozzi et al., 2006; Wang et al., 2018). On the other hand, nitrophenols and nitrocatechols were dominated by 4-nitrophenol and 4-nitrocatechol, respectively, which are mainly from primary emissions (Wang et al., 2017; Lu et al., 2019a).

To identify and quantify the sources of the NACs observed in Xi'an, the PMF model was employed, and four major factors were resolved with uncertainties < 15 %. The factor profiles are shown in Fig. S1. The first factor, vehicle emission, which is characterized by high levels of hopanes, shows large relative contributions to NACs in spring and summer. Direct traffic emissions of NACs have also been verified in laboratory studies (Tremp et al., 1993; Perrone et al., 2014). The second factor is considered to be coal combustion for residential heating and cooking, which is characterized by the higher loadings of picene, benzo(a)pyrene, benzo(b)fluoranthene, benzo(k)fluoranthene, indeno(1,2,3-cd)pyrene, and benzo(ghi)perylene. This factor accounted for ∼ 40 % of the NACs in winter. The emission of NACs from coal combustion for residential usage was reported by Lu et al. (2019a), which showed emission factors of 0.2 to 10.1 mg kg-1. It is worth noting that with the emission control of residential coal burning after 2017, the contribution feature of coal burning to NACs could be different. The third source is identified as secondary formation because of the highest level of phthalic acid and its highest contribution in summer. The formation of secondary NACs is also supported by both field and modeling studies (Harrison et al., 2005; Iinuma et al., 2010; Yuan et al., 2016). The last source factor, with high loadings of levoglucosan, vanillic acid, vanillin, and syringyl acetone, was identified as biomass burning, which has higher contributions in fall and winter. The emission of NACs from biomass burning was reported by field studies and was considered to be an important source of NACs (Mohr et al., 2013; Lin et al., 2016; Teich et al., 2017).

The source contributions for NACs in Xi'an are shown in Fig. 4, which shows obvious seasonal differences. In spring, vehicular emission (41 %) was the main contributor to NACs. Secondary formation (26 %) and biomass burning (20 %) also contributed significantly. In summer, secondary formation had the highest contribution (45 %), which was likely due to enhanced photochemical oxidation leading to the formation of NACs. Besides, vehicular emission also contributed significantly (34 %) in summer. In fall, biomass burning (45 %) contributed the most, while secondary formation (30 %) and vehicular emission (23 %) also had significant contributions. In winter, coal burning (39 %) and biomass burning (36 %) were the main contributors, which can be attributed to emissions from residential-heating activities. It is worth noting that the absolute concentrations of NACs attributed by vehicle emission (see Table S6) were higher in winter than those in spring and summer, yet these differences of less than a factor of 20 are not as significant as the differences (spring and summer vs. winter) for NACs attributed by other primary emissions (> 80 times for coal burning and > 40 for biomass burning). These results indicate that anthropogenic primary sources are the main contributors to NACs in Xi'an, suggesting that control of anthropogenic emissions (biomass burning and coal burning) is important for mitigating pollution of NACs in this region. Secondary formation also contributes significantly to NACs, especially in summer. Further comprehensive field studies are necessary for understanding the formation mechanisms of NACs under different atmospheric conditions.

To reveal the source origins of the NACs, the concentrations of the NACs were grouped according to their trajectory clusters that represent different air mass origins, as shown in Fig. 5. In general, the air masses from local emissions (Cluster 1 in spring and fall and Cluster 2 in summer and winter), which showed the features of small-scale and short-distance air transport, caused significant increases in NAC concentrations. As for regional transport, the air masses from the neighboring Gansu Province across Baoji city before arriving at Xi'an presented higher concentrations of NACs in fall and winter (Cluster 2 and Cluster 3, respectively). In addition, air masses from Xinjiang across Gansu caused increased concentrations of NACs in spring and summer (Cluster 2 and Cluster 1, respectively). A small proportion of air masses from the northwest (Cluster 3 in spring and Cluster 1 in winter), the south (Cluster 3 in summer), and the west (Cluster 3 in fall), which showed long or moderate transport patterns, are related to the lowest concentrations of NACs. This may be due to the long-distance transport or relatively clean air from those regions. In the same season, the source origins of air masses were different between clusters, thus causing the difference in concentrations of NACs. However, the composition of NACs was similar between clusters, which is comparable to the results of Chow et al. (2015).

The correlations between NAC concentration and AbsBrC,365 for each season are shown in Fig. S2. The correlations are stronger in fall (r2 = 0.68) and winter (r2= 0.63) compared to those in spring (r2 = 0.15) and summer (r2= 0.40). These results indicate that NACs are important components of BrC chromophores in fall and winter.

Figure 6 shows the contributions of NACs to BrC light absorption at wavelengths from 300 to 500 nm (AbsBrC,300-500) as well as the carbon mass contributions of NACs to OC. The contributions of NACs to AbsBrC,300-500 are wavelength-dependent and vary largely in different seasons. High contributions at wavelengths of 350–400 nm were observed in fall and winter, but the contributions in spring and summer were mainly at wavelengths shorter than 350 nm. These results may be due to the high proportion of nitrocatechols in fall and winter (see discussion above), which have strong light absorption at wavelengths above 350 nm (see Fig. S3). The seasonal-average contributions of NACs to AbsBrC,365 were highest in winter (0.91 ± 0.30 %), followed by fall (0.36 ± 0.22 %), spring (0.14 ± 0.04 %), and summer (0.09 ± 0.06 %) (see Table S4). These contributions are comparable to a previous study where eight NACs were measured (Teich et al., 2017). The contributions of NACs to AbsBrC,365 in winter were about 10 times higher compared to those in summer, which could be due to the high emissions of NACs in winter. Alternatively, enhanced atmospheric oxidizing capacity in the summer can lead to enhanced formation of secondary NACs or the degradation and/or bleaching of certain NACs (Barsotti et al., 2017; Hems and Abbatt, 2018; Wang et al., 2019), which might eventually reduce the contributions in summer. The fractions of NACs to total OC also show obvious seasonal variation, with average contributions higher in winter (0.14 ± 0.05 %) and fall (0.05 ± 0.02 %) and lower in spring (0.02 ± 0.01 %) and summer (0.01 ± 0.01 %). The contributions of NACs to BrC light absorption at 365 nm are, however, 6–9 times larger than their carbon mass contributions to total OC. Our results echo previous studies showing that even small numbers of chromophores can have a non-negligible impact on the optical characteristics of BrC due to their disproportional absorption contributions (Mohr et al., 2013; Zhang et al., 2013; Teich et al., 2017; Xie et al., 2017).

The daily contributions of the individual NACs to light absorption of total NACs at wavelengths of 300–500 nm are shown in Fig. 7. Similar to the concentration fractions in NACs, nitrocatechols were the main contributors in winter and fall, with contributions of 38 %–65 % and 18 %–62 %, respectively. On the other hand, nitrophenols dominated in spring and summer, with contributions of 61 %–96 % and 27 %–100 %, respectively. As for nitrophenols, 4-nitrophenol was the most important chromophore, followed by 2-methyl-4-nitrophenol, 3-methyl-4-nitrophenol, and 2,6-dimethyl-4-nitrophenol (only observed in winter). As for nitrocatechols, 4-nitrocatechol was the main contributor in all four seasons, while 3-methyl-5-nitrocatechol and 4-methyl-5-nitrocatechol also contributed significantly in fall and winter. For nitrosalicylic acids, 5-nitrosalicylic acid contributed in all four seasons but contributed the most in summer, while 3-nitrosalicylic acid was only observed in winter, which could be attributed to their different sources, as discussed above.

The seasonal contributions of individual NACs to total light absorption of NACs at a wavelength of 365 nm are shown in Fig. 3b. The relative-contribution trends of 4-nitrophenol > 4-nitrocatechol > 2-methyl-4-nitrophenol > 5-nitrosalicylic acid > 3-methyl-4-nitrophenol, 4-nitrophenol > 4-nitrocatechol > 5-nitrosalicylic acid > 2-methyl-4-nitrophenol > 3-methyl-4-nitrophenol, 4-nitrocatechol > 4-nitrophenol > 4-methyl-5-nitrocatechol > 3-methyl-5-nitrocatechol > 5-nitrosalicylic acid > 2-methyl-4-nitrophenol > 3-methyl-4-nitrophenol, and 4-nitrocatechol > 4-methyl-5-nitrocatechol > 4-nitrophenol > 3-methyl-5-nitrocatechol > 2-methyl-4-nitrophenol > 4-nitro-1-naphthol > 5-nitrosalicylic acid > 3-methyl-4-nitrophenol > 3-nitrosalicylic acid > 2,6-dimethyl-4-nitrophenol were observed in spring, summer, fall, and winter, respectively. These trends are different from their concentration fractions in OC, which may be mainly due to the differences in light absorption ability (see Fig. S3). For example, 4-nitrocatechol has lower mass concentration but higher light absorption contribution compared to 4-nitrophenol. These results suggest that mere compositional information of NACs might not be directly translated into impacts on optical properties because they have startlingly different absorption properties.