An Agilent model 7890A GC equipped with a 5975C mass selective detector was applied to identify and quantify organic compound classes. A split/splitless injector and a DB-5MS fused silica capillary column (30m×0.25mm i.d. and 0.5 µm film thickness) were equipped. The samples in the fused silica capillary column were analysed using the following GC temperature programme: 50 ∘C for 2 min, increasing to 120 ∘C at 15 ∘C min-1, before increasing to 300 ∘C at 5 ∘C min-1, and finally being held at 300 ∘C for 16 min. The carrier gas was helium. The MS detection was undertaken in electron ionization (EI) mode at 70 eV, scanning from 50 to 650 Da. Most of the recoveries for authentic standards or surrogates were over 80 %. The quality and quantity of a single compound were acquired using the ChemStation software. Moreover, authentic standards were employed to achieve GC-MS response factors. The results in this study were corrected for the field blanks, but not for the recoveries.

Concentrations of OC and EC were measured using a thermal/optical carbon analyser (model RT-4, Sunset Laboratory Inc., USA). The analytical errors were detected within ±10 % via a duplicate analysis of each filter. In winter, the blank levels were in the ranges of 1.52–2.84 and 0–0.03 µgC for OC and EC respectively. The summertime ranges of blank levels were 1.17–1.50 and 0 µgC for OC and EC respectively.

In this study, the contributions of POC and SOC to total OC were evaluated based on organic tracer method. An experimentally derived factor of 13 pg C per spore was used to calculate the contribution of fungal spores to OC (Bauer et al., 2008). Plant debris was evaluated using the tracer of cellulose and an experimentally derived factor (Puxbaum and Tenze-Kunit, 2003). The tracer mass fraction (fSOC) factors of 0.155±0.039, 0.231±0.111, and 0.023±0.005 for isoprene, α-pinene, and β-caryophyllene respectively were applied to evaluate the contributions of BSOAs to OC (Kleindienst et al., 2007).

The abundances and seasonal variations of n-alkanes during the two seasons are shown in Fig. 3a. Concentrations of n-alkanes (C18-C35) were 343±227 ng m-3 (daytime) and 499±307 ng m-3 (night-time) in winter, which were roughly 2–3 times higher than the summer average loadings of 141 ng m-3 during both day- and night-time. In general, the molecular distributions of n-alkanes for most samples were characterized by an odd carbon number predominance with a maximum at C23 (Fig. 3b) in winter versus high values at C27 and C29 in summer (Fig. 3c). The carbon preference index (CPI; concentration ratios of odd carbon to even carbon n-alkanes) ratios for C18–C35 n-alkanes of all samples were calculated, which is often used to identify the contributions of anthropogenic and biogenic sources (Simoneit, 1986). In winter, the CPIs were 1.21±0.11 at daytime and 1.19±0.09 at night, which is comparable to those reported in Beijing and other urban aerosols in China (1.0±0.43) (Wang et al., 2006). The average CPIs were 1.39±0.40 at daytime and 1.36±0.44 at night in summer. The CPIs of terrestrial higher plant waxes are usually ∼5–10, whereas CPIs close to unity are attributed to marine sources and/or petroleum residues (Simoneit et al., 1991a; Hsu et al., 2006). Such molecular distributions indicate that aerosols in Tianjin may be mainly derived from the incomplete combustion of fossil fuels/petroleum residue and/or marine sources, which tend to have similar CPIs in both seasons.

High-molecular-weight n-alkanes (HMWalk; C25–C36) are mainly derived from terrestrial higher plant waxes, in which C27, C29, and C31 are the dominant species. Low-molecular-weight n-alkanes (LMWalk; <C25) are usually emitted from biomass burning and fossil fuel combustion (Kawamura et al., 2003b; Freeman and Collarusso, 2001). The ratios of low-molecular-weight to high-molecular-weight n-alkanes (LMW / HMWalk) were 1.03±0.30 at daytime and 1.04±0.32 at night in winter and 0.83±0.50 at the daytime and 0.56±0.32 at night in summer, and indicate more important contributions from biomass burning and fossil fuel combustion in winter than in summer. In winter, the concentrations of both LMWalk and HMWalk at night were higher than those during the daytime (Fig. 3b and Table S1 in the Supplement), which may be related to enhanced anthropogenic activities (e.g. house heating) and the lower nocturnal boundary layer height. However, the concentrations of LMWalk were higher at daytime (39.1±14.1 ng m-3) than at night (34.9±23.3 ng m-3) in summer (Fig. 3c, Table S1). A distribution such as this might be due to the significant land/sea breeze circulation, which could bring large amounts of terrestrial higher plant waxes to Tianjin at night and transport marine organic matter to the mainland during the day. The percentage of wax Cn is the contribution of biogenic n-alkane that is derived from high plant waxes (Ren et al., 2016). On average, the plant wax n-alkanes (WNA; C25–C34) accounted for 10.1 % of total homologues during the day and 9.01 % at night in winter, which is similar to those in summer (average contributions of 10.2 % at daytime and 9.91 % at night; Table S2), indicating that higher plant waxes made a minor and stable contribution to n-alkanes in both seasons.

The abundances of and seasonal variations in normal fatty alcohols in organic aerosols are shown in Fig. 5a. n-Alcohols (C12–C31) were detected in the aerosol samples with concentrations of 1310±811 ng m-3 (daytime) and 1520±1010 ng m-3 (night-time) in winter and 621±367 ng m-3 (daytime) and 572±438 ng m-3 (night-time) in summer; these concentrations were apparently higher than the concentrations of n-alkanes and fatty acids. The molecular distributions showed strong even carbon number predominance, particularly at C16 and C18 (Fig. 5b, c). High-molecular-weight (HMWalc, >C19) n-alcohols are mainly derived from higher plant waxes and biomass burning (Wang et al., 2006). Low-molecular-weight (LMWalc, C12-C19) n-alcohols are related to marine and soil microbes (Fu et al., 2008). The total concentrations of LMWalc and HMWalc were in the respective ranges of 223–2830 ng m-3 (850 ng m-3) and 84.8–1400 ng m-3 (455 ng m-3) during the daytime, which were lower than those of 142–2910 ng m-3 (972 ng m-3) and 92.2–1730 ng m-3 (549 ng m-3) at night in winter respectively. In contrast, the concentrations of LMWalc (89.0–1320, 506 ng m-3) and HMWalc (28.7–238, 115 ng m-3) during the daytime were higher than those (139–1627, 467 and 19.2–474 ng m-3, 105 ng m-3 respectively) at night in summer (Fig. 5a, Table S1). Compared with HMWalc, the concentrations of LMWalc were significantly higher during the daytime than those at night, which may be due in part to the more significant sea breezes during the daytime and the land breezes at night. Therefore, we infer that the fatty alcohols in winter may mainly be derived from biomass burning and soil resuspension particles, whereas the contributions of marine/biogenic emission and biomass burning can possibly explain the molecular distribution of fatty alcohols in summertime aerosols.

The relative abundances (%) of HMWalk, HMWfat, and HMWalc in the Tianjin aerosols are illustrated in a triangular diagram (Fig. 6). The average abundances of HMWalc were 54.5 % (wintertime) and 44.6 % (summertime), and they were the most dominant species among aliphatic lipids. The percentage of HMWalk was lower in winter (22.5 %) than in summer (42.5 %). The relative abundances of HMWalk, HMWfat, and HMWalc in the aerosols collected over the East China Sea (Kang et al., 2017), on Chichi-Jima Island in the northwest Pacific (Kawamura et al., 2003b), at Mt. Tai (Fu et al., 2008), and in urban Beijing (Ren et al., 2016) are plotted as categories A, B, C, and D respectively. It is worth noting that the results from this study overlap with the four other categories, which indicates that the aerosols may have similar source contributions to some extent and also highlights that the coastal city of Tianjin is influenced by a mixture of terrestrial and marine air masses.

It was found that most of the wintertime data overlapped with aerosols collected at Mt. Tai (category C), and a fraction of samples also fell into category D that represents the Beijing aerosols. Fu et al. (2008) reported that the field burning of wheat straw largely contributed to the Mt. Tai aerosols, which could be further transported to the Pacific Ocean under the influence of westerly winds from the mainland. The Beijing aerosols in category D were significantly affected by incomplete fossil fuel combustion and biomass burning as well as biogenic emissions (Ren et al., 2016). However, most of the summertime aerosols in Tianjin fell into category A, suggesting that they may share similar sources with aerosols collected over the East China Sea (Kang et al., 2017), which may be significantly influenced by biogenic/marine emissions (under the marine air masses) and BB via long-range transport.

A total of 14 sugar compounds including 3 anhydrosugars, 5 primary saccharides, and 6 sugar alcohols identified in this study (Table S1) are water-soluble, and are known to contribute to water-soluble organic carbon (WSOC) (Graham et al., 2003; Elbert et al., 2007; Fu et al., 2008). In addition, they can affect aerosol hygroscopicity (Mochida and Kawamura, 2004; Fu et al., 2008) and regulate climate to some extent (Kanakidou et al., 2005). The average concentrations of total sugars were 371±208 ng m-3 (daytime) and 496±247 ng m-3 (night-time) in winter and 61.2±21.2 ng m-3 (daytime) and 96.9±94.0 ng m-3 (night-time) in summer.

Levoglucosan, a dominant tracer of biomass burning (Graham et al., 2002; Sheesley et al., 2003; Hays et al., 2005; Iinuma et al., 2007; Fu et al., 2008), was the most abundant compound among anhydrosugars and total sugars, with average concentrations of 205±122 ng m-3 (daytime) and 296±153 ng m-3 (night-time) in winter and 12.8±6.97 ng m-3 (daytime) and 34.4±46.0 ng m-3 (night-time) in summer. The contribution of biomass burning was much more significant in winter, especially at night in both summer or winter. Levoglucosan has two isomers, namely galactosan and mannosan, which were detected in all aerosol samples (Fig. 7, Table S1). These anhydrosugars are formed via the pyrolysis of cellulose/hemicellulose in different types of biomass such as grasses (Iinuma et al., 2007), woods (Graham et al., 2002), and agricultural residues including wheat (Fu et al., 2008) and rice straw (Sheesley et al., 2003). These three anhydrosugars showed similar seasonal patterns (Fig. 7a, b, c) with significantly higher wintertime concentrations compared with summer.

The ratios of levoglucosan to mannosan (L / M) and mannosan to galactosan (M / G) were applied to discriminate between different categories of biomass burning. Figure 8 shows the isomer ratios of crop residues, and soft and hard wood from a number of regions in different countries that have been reported in other literature. All of the values from previous studies were related to PM2.5 aerosol samples. Average L / M ratios were reported to be in the ranges of 3.0–5.80, 12.9–35.4, and 40.0–41.6 for smoke emitted by the burning of softwood, hardwood, and crop residue respectively (Sheesley et al., 2003; Fine et al., 2004; Oros et al., 2006; Engling et al., 2006). In addition, the average ranges of M / G ratios were reported to be 3.60–7.0, 1.2–2.0, and 0.30–0.60 for the burning of softwood, hardwood, and crop residue respectively (Sheesley et al., 2003; Fine et al., 2004; Oros et al., 2006; Engling et al., 2006). The relative contributions of levoglucosan and mannosan in aerosols can be used to indicate the apportionment of cellulose and hemicellulose in biomass fuels. Low M / G values suggest that the smoke may be mainly derived from the burning of biomass briquettes, crop straw, and grasses (Sheesley et al., 2003; Oros et al., 2006; Fu et al., 2008). In this study, L / M ratios (4.88–19.8, 7.38) were high and M / G ratios (1.0–5.51, 1.59) were low in winter, whereas the ranges of summertime aerosols were found to be 2.74–15.8 (5.68) and 1.15–2.84 (2.12) for L / M and M / G respectively. These results suggest that the soft- and hardwood are the main types of biofuels used in both seasons in Tianjin, and that the use of hardwood and/or crop residue was enhanced in winter.

Primary saccharides (fructose, glucose, xylose, sucrose, and trehalose) with substantially similar seasonal distributions (Fig. 7d–h) have been widely employed to indicate the sources of the resuspension of surface soils and unpaved road dust containing biological material (Simoneit et al., 2004a; Fu et al., 2008). The total concentrations of primary saccharides were 46.8±20.9 ng m-3 at daytime and 49.9±23.0 ng m-3 at night-time in winter and 20.1±11.1 ng m-3 at daytime and 21.2±21.5 ng m-3 at night-time in summer, showing no obvious diurnal variation. These results imply that the resuspension of surface soils and unpaved road dust containing biological material was quite a stable source of organic aerosols in Tianjin. We also determined six sugar polyols, consisting of arabitol, mannitol, glycerol, erythritol, xylose, and maltose. Glycerol was the most abundant sugar alcohol in both seasons. The average concentrations of glycerol were 49.8±32.0 ng m-3 (daytime) and 52.8±27.1 ng m-3 (night-time) in winter and 14.8±10.8 ng m-3 (daytime) and 25.0±31.9 ng m-3 (night-time) in summer. The levels of glycerol were notably enriched in winter, were higher than the concentrations of primary saccharides, and were positively correlated with levoglucosan (R2=0.73, p<0.01, N=85; Table S3), implying that there were potential emissions from biomass burning contributing to glycerol in winter. In summer, however, the low correlations (R2=0.32, ρ<0.05, N=60; Table S4) reported between glycerol and levoglucosan suggest that most of the glycerol was possibly derived from the metabolism of soil microorganisms (Simoneit et al., 2004b; L. J. Li et al., 2018) and photooxidation. Moreover, arabitol and mannitol were detected with similar seasonal trends (Fig. 7i–j), representing the contribution of fungal spores (Bauer et al., 2008; Zhu et al., 2016), which are prevalent in both land and marine areas (Elbert et al., 2007; Zhu et al., 2015a).

In Fig. 7, it is obvious that the concentrations of levoglucosan (a specific tracer of biomass burning), saccharides, and sugar alcohols were higher in winter than in summer. In addition, saccharides and sugar alcohols showed high levels, especially when the levoglucosan reached peaks in both seasons, implying that they may share similar sources to some extent. To verify the contribution of biomass burning to levoglucosan and sugar alcohols, we downloaded the fire maps for winter and summer during the sampling periods (Fig. S1). It can be clearly seen that there was high anthropogenic activity in winter (Fig. S1b). It is well known that the combustion of fossil fuels and biofuels is widely used for house heating in China in winter. Thus, biomass burning could contribute to levoglucosan and sugar alcohols. In summer, there were relatively dense fire spots distributed across the North China Plain (NCP) and in some southern agricultural provinces, such as Jiangsu, Anhui, Henan, and Shandong (Fig. S1a). Fu et al. (2008) reported that there was large-scale burning of wheat straw across the country during May–June. Meanwhile, the wind direction was mainly southerly during winter- and summertime in Tianjin (Fig. 2). Therefore, in addition to the contribution of local biomass burning, the southern wind carried a large amounts of biomass burning particulate matter from the NCP and the southern agricultural provinces to Tianjin, which could be a source of levoglucosan and sugar alcohols.

In this study, we also calculated the contributions of different saccharides to total sugars. It was found that the average percentages of anhydrosugars to total sugars (Table S2) were 0.67±0.05 (daytime) and 0.73±0.05 (night-time) in winter, which were roughly 3 times higher than those (0.27±0.08 during the daytime and 0.37±0.15 at night) in summer. In contrast, the percentages of primary saccharides and the two main tracers of fungal spores (arabitol and mannitol) had average ratios of 0.16±0.04 (daytime) and 0.13±0.02 (night-time) in winter, which were about 3 times lower than those of summertime aerosols (0.43±0.14 during the daytime and 0.30±0.11 at night). Meanwhile, the contributions of anhydrosugars were high at night, especially in winter, while a large proportion of primary saccharides and sugar alcohols were identified in summer, especially during the daytime (Figs. 9, 10, Table S2). These results suggest that biomass burning made a significant contribution in winter, accounting for 68.3 % and 74.0 % of total sugar sources for the day- and night-time respectively, whereas the summertime aerosols were apparently influenced by biological sources, especially during the daytime. Primary saccharides and sugar polyols were responsible for 30.4 % and 42.1 % of total sugars during the daytime and 20.0 % and 37.4 % of total sugars during night-time in summer respectively (Fig. 10).

The biosphere–atmosphere hydrocarbon exchange is usually subject to global biogenic emission with an isoprene loading of 600 Tg yr-1 (Sharkey et al., 2007). The reactive double bonds of isoprene can easily be oxidized by oxidants such as OH, NO3, and O3 in the atmosphere. Six molecular markers were identified as isoprene SOA tracers, i.e. 2-methylglyceric acid (2-MGA), C5-alkene triols (cis-2-methyl-1,3,4-trihydroxy-1-butene, trans-2-methyl-1,3,4-trihydroxy-1-butene, and 3-methyl-2,3,4-tri-hydroxy-1-butene), and two 2-methyltetrols (MTLs, 2-methylthreitol and 2-methylerythritol) (Table S1). The wintertime concentrations of total isoprene SOA tracers were 1.03–10.6 ng m-3 (4.13 ng m-3) during daytime and 0.53–11.7 ng m-3 (4.54 ng m-3) at night, which were much lower than the summertime samples that had values of 3.63–83.9 ng m-3 (29.6 ng m-3) at daytime and 5.40–106 ng m-3 (25.3 ng m-3) at night. In winter, 2-MGA was the most abundant species, followed by C5-alkene triols and MTLs. However, the levels of 2-MGA were lower than the concentrations of C5-alkene triols and MTLs in summer.

The average concentrations of MTLs were detected as 0.43±1.17 ng m-3 (daytime) and 0.40±1.03 ng m-3 (night-time) in winter and 12.1±8.70 ng m-3 (daytime) and 10.6±11.2 ng m-3 (night-time) in summer. C5-alkene triols are specific isoprene SOA tracers under low-NOx conditions (Surratt et al., 2010), accompanying similar seasonal variations as 2-methyltetrols in summer (Fig. 11b, c), whereas 2-MGA is a further oxidation product of isoprene under high-NOx conditions (Surratt et al., 2010). The concentrations of 2-MGA were 2.13±1.81 ng m-3 (daytime) and 2.32±2.15 ng m-3 (night-time) in winter and 5.76±3.89 ng m-3 (daytime) and 4.27±3.98 ng m-3 (night-time) in summer. In general, C5-alkene triols are recognized as significant terrestrial tracers (Fu et al., 2014), and they showed a good correlation (R2=0.83, p<0.01, N=60, Table S5) with MTLs in summer, suggesting that they may have similar terrestrial sources such as biomass burning and higher plant waxes. However, there was no correlation in winter (Table S5), implying different sources of these two species in winter. Meanwhile, the average 2-MGA over MTLs (2-MGA/MTLs) ratios were 17.0 at daytime and 22.0 at night-time in winter, which are much higher than those (0.54 and 0.50 for day- and night-time respectively) in summer; the higher values in winter are possibly due to the higher concentrations of NOx and acidity on SOA formation (Surratt et al., 2007), which can also explain the phenomenon that the concentrations of 2-MGA were higher in winter than in summer.

Four oxidation products of α-/β-pinene were detected in aerosol samples, including 3-hydroxyglutaric acid (3-HGA), 3-methyl-1,2,3-butanetricarboxylic acid (MBTCA), and pinonic and pinic acids. The total average concentrations of α-/β-pinene SOA tracers were 12.2±7.69 ng m-3 (daytime) and 11.4±6.02 ng m-3 (night-time) in winter and 23.4±13.6 ng m-3 (daytime) and 22.4±23.7 ng m-3 (night-time) in summer. Positive correlations were found between isoprene and α-/β-pinene SOA tracers as well as with T in summer (Table S6), whereas correlations in winter were weak, indicating that they may have similar sources or influencing factors such as T and RH in summer and a large number of different origins in winter.

Among the α-/β-pinene SOA tracers, the most predominant compound was pinonic acid, followed by pinic acid. The concentration of 3-HGA was higher than MBTCA in winter, whereas it was opposite in summer (Table S1). Pinonic and pinic acids are first-generation products of α-/β-pinene oxidation, which can be further photo-degraded into products such as MBTCA (Claeys et al., 2007). Therefore, the ageing level of α-/β-pinene could be evaluated using values of pinonic and pinic acids to MBTCA: (pinonic + pinic) / MBTCA (Gómez-González et al., 2012; Ding et al., 2014). In our work, ratios of (pinonic + pinic)/MBTCA were 17.6±15.9 (daytime) and 31.0±51.9 (night-time) in winter, which were much higher than the summertime aerosols (6.8±7.0 at daytime and 7.3±9.7 at night), suggesting that the summertime aerosols were more aged than those in winter due to strong photooxidation in summer.

Isoprene emissions are more susceptible to higher temperature with larger contributions in summer (L. J. Li et al., 2018). The ratio of isoprene tracers to α-/β-pinene tracers (Riso/pine) can also be used to evaluate the relative contributions of isoprene and α-/β-pinene oxidation to biogenic SOA formation. In the present study, the average Riso/pine ratios were 0.39 (daytime) and 0.52 (night-time) in winter, which were lower than summertime samples (1.85 at daytime and 1.77 at night), suggesting that isoprene oxidant products were more abundant in summer than those in winter. In addition, the ratios were higher than aerosols in Hong Kong (average 0.46) (Hu et al., 2008) during summertime, but lower than those of mountain aerosols at locations such as Mt. Tai, which has reported averages of 4.9 and 6.7 for the day- and night-time in summer (Fu et al., 2010a), and the Changbai Mountain region (3.7 at daytime) (Wang et al., 2008). This is reasonable because mountain aerosols contain more biogenic SOAs than urban aerosols.

β-Caryophyllene is one of the most abundant sesquiterpene compounds and is frequently reported in previous studies due to its high reactivity and relatively low vapour pressure (Fu et al., 2010b); it has been said to be emitted from plants such as pine and birch trees (Helmig et al., 2006; Duhl et al., 2008). β-Caryophyllinic acid is an ozonolysis or photooxidation product of β-caryophyllene (Jaoui et al., 2007), which was detected at concentrations of 10.7±9.33 ng m-3 (daytime) and 10.3±8.41 ng m-3 (night-time) in winter – 5 times higher than summertime aerosols (1.99±1.81 ng m-3 at daytime and 2.21±4.53 ng m-3 at night). Previous studies in Okinawa (Zhu et al., 2016), India (Fu et al., 2010b), and Beijing (L. J. Li et al., 2018) also reported that concentrations of β-caryophyllinic acid peaked in winter. There are large amounts of sesquiterpenes attached to woods and leaves due to their low volatility, which could be emitted from smoke during biomass burning, especially in winter (Zhu et al., 2016). The seasonal variation in the sesquiterpene SOA tracer was likely controlled by air masses from Southeast Asia, which carried more oxidized β-caryophyllinic acid via long-range transport (Fu et al., 2010b; Zhu et al., 2016). It is interesting to note that β-caryophyllinic acid, levoglucosan (a specific tracer of biomass burning), and 2,3-dihydroxy-4-oxopentanoic acid (DHOPA) had similar seasonal variations (Figs. 7a, 11h), with a high concentration peak occurring during a severe haze episode (28 May 2017). Moreover, there were strong positive correlations between these species in both seasons (Tables S5–6), which suggest that elevated β-caryophyllinic acid concentrations could be attributed to biomass burning during the haze periods (Fig. 11h).

OC concentrations detected in this study were in the ranges of 5.33–79.8 µgm-3 (23.7 µgm-3) in winter and 1.43–6.64 µgm-3 (3.78 µgm-3) in summer. The Asian summer monsoon may bring clean marine air masses to Tianjin and lower the atmospheric levels of OC in summer (Mao et al., 2017). In detail, the concentrations of BB-derived OC were the most abundant in winter (2.49±1.48 µgm-3 and 3.61±1.86 µgm-3 at day- and night-time respectively), and were significantly higher than those in summer (0.16±0.08 and 0.42±0.56 µgm-3 at day- and night-time respectively) based on the ratio of levoglucosan to OC (L / OC) of 8.2 % (Andreae and Merlet, 2001; X. Y. Zhang et al., 2007). Although, the L / OC ratio was just one of the experimental simulation results, we think it is reasonable in this study. Furthermore, there is plenty of literature that shows that L / OC ratios are variable depending on the sources of biomass (e.g. straw and wood) and burning conditions. On average, the BB contribution to OC was 12.1 % (daytime) and 16.0 % (night-time) in winter and 4.14 % (daytime) and 9.62 % (night-time) in summer, which were roughly comparable to the values of 1.0 %–35 % (9.9 %) reported for a whole year in the nearby megacity of Beijing (Li et al., 2018).

Mannitol and arabitol are generally employed to evaluate the numbers of fungal spores (Elbert et al., 2007; Bauer et al., 2008). This work uses the values of 1.7 pg mannitol per spore and 13 pg OC per spore (Bauer et al., 2008) in order to identify the contributions of fungal-spore-derived OC. Meanwhile, plant-debris-derived OC was estimated based on the relationship of glucose and plant debris as well as the OM / OC ratio of 1.93 (Puxbaum and Tenze-Kunit, 2003). Here, the estimations based on molecular tracers indicate that fungal spores and plant debris have higher contributions to OC during summertime than in wintertime (Figs. 13a–b, 14) due to high biological activity in summer. The contributions of fungal-spore-derived OC were 0.88 % (daytime) and 0.79 % (night-time) in summer, which were lower than the values in Beijing (2.79 % in summer; L. J. Li et al., 2018) and much lower than values found for forest aerosols in Japan (22 % during the daytime and 45 % at night; Zhu et al., 2016) and in tropical rainforest aerosols on Hainan Island (12.1 %; Zhang et al., 2010). Plant-debris-derived OC accounted for the lowest proportion (0.22 %) of primary OC during both the day- and night-time in summer, which was lower than summertime aerosols in Beijing (1.05 %; L. J. Li et al., 2018).

SOAs formed from biogenic and anthropogenic precursors are important contributors to PM2.5. It has been reported that global SOAs mostly originate from BVOCs, based on atmospheric modelling studies of SOA (Hallquist et al., 2009). Global BVOCs emissions can reach 760 TgC per year, from which isoprene, monoterpenes, and sesquiterpenes account for 70 %, 11 %, and 2.5 % respectively (Sindelarova et al., 2014). Anthropogenic SOCs from toluene and naphthalene were estimated from 2,3-dihydroxy-4-oxopentanoic acid (DHOPA) and phthalic acid respectively, with respective mass fractions of 0.0026 and 0.0199 as well as an OM / OC value of 1.93 (Kleindienst et al., 2007; Fu et al., 2014). Vehicle emissions may be the main sources of toluene in Tianjin, and solvent usage in the electronic and chemical industries are also potential sources (Widiana et al., 2019; Fu et al., 2016). McFiggans et al. (2019) demonstrated that there is a substantial overestimation of SOA production due to the simple linear addition of SOA mass yields from the individual yields of components in a VOC mixture. The measurement uncertainties and personal errors, such as wall losses (Lee et al., 2006), and the accurate determination of the mass of semi-volatile materials should also be considered (Ehn et al., 2014; McFiggans et al., 2019).

For biogenic SOAs, isoprene and α-pinene SOCs were more abundant in summer, whereas β-caryophyllene contributed more to SOCs in winter; this may be related to different sources such as biomass burning and other factors including the volatility of organic compounds (Figs. 13c–d, 14). Similarly, the contributions of SOC derived from isoprene (4.47 % and 3.95 %) and α-pinene (2.56 % and 2.44 %) to OC during daytime and night-time were apparently elevated in summer (Table S1, Fig. 13c–d), whereas their contributions in winter were low (Fig. 14a, b). In contrast, the concentrations of β-caryophyllene SOC were high in winter (averages of 0.47 and 0.45 µgm-3), accounting for 1.89 % and 1.80 % of the OC at day- and night-time respectively, which were comparable values to those (1.95 % and 1.82 %) in summer.

The concentrations of anthropogenic SOCs were 2–5 times higher than those of biogenic SOCs in both seasons. The average concentrations of anthropogenic SOCs were 2.12 µgm-3 (daytime) and 2.03 µgm-3 (night-time) in winter, which were higher than the values of 1.0 µgm-3 (daytime) and 0.73 µgm-3 (night-time) in summer. Toluene SOC was predominant with concentrations of 1.68±0.95 µgm-3 and 1.65±0.85 µgm-3 for day- and night-time in winter respectively, and with low levels of 0.82±0.58 µgm-3 and 0.64±1.02 µgm-3 in summer. In addition, the levels of naphthalene SOCs were 0.44±0.27 µgm-3 (daytime) and 0.38±0.21 µgm-3 (night-time) in winter, and 0.18±0.08 µgm-3 (daytime) and 0.09±0.08 µgm-3 (night-time) in summer. The high anthropogenic SOC concentrations in winter may be attributed to elevated biomass/biofuel combustion. Although the concentrations of summertime aerosols were lower, the contributions of anthropogenic SOCs were higher (Table S1, Fig. 13). Anthropogenic SOCs contributed 10.1 % (daytime) and 9.12 % (night-time) of the total OCs in aerosols in winter, and 24.7 % (daytime) and 13.6 % (night-time) in summer. Anthropogenic SOCs in summer may not only be related to fossil fuel combustion, but also to increased plastic emissions in summer (Fujii et al., 2003; Simoneit et al., 2005; Wang et al., 2006; Kong et al., 2013), which warrants further study.

The total average concentrations of primary OC were 2.55±1.51 µgm-3 (daytime) and 3.67±1.89 µgm-3 (night-time) in winter and 0.20±0.09 µgm-3 (daytime) and 0.46±0.57 µgm-3 (night-time) in summer, corresponding to 12.4 % and 16.3 % of OC for the day- and night-time in winter and 5.24 % and 10.6 % of OC in summer respectively. Additionally, BB-derived OC was detected as the most abundant primary source, followed by fungal spores and plant debris. SOCs including biogenic and anthropogenic sources were estimated to be 2.66±1.52 µgm-3 (daytime) and 2.56±1.26 µgm-3 (night-time) in winter and 1.38±0.81 µgm-3 (daytime) and 1.09±1.51 µgm-3 (night-time) in summer, accounting for 12.4 % and 11.3 % of OC during the day- and night-time in winter and 33.7 % and 21.8 % in summer respectively (Table 1, Fig. 14a–e).

Apparent differences in seasonal characteristics and diurnal variations in organic aerosols were observed between the two seasons in Tianjin. It is worth noting that the contributions of SOC to OAs in summer were roughly twice as high as those in winter, especially for the toluene SOC. Biomass burning OC and β-caryophyllene SOC were more abundant in winter, whereas fungal-spore- and plant-debris-derived OCs, as well as biogenic SOCs contributed more significantly in summer (Fig. 14e). These results are in accordance with previous studies in many megacities in China (Ding et al., 2017), including Beijing (L. J. Li et al., 2018). In total, the average contributions of primary and secondary OCs using the tracer-based methods discussed above were 24.8 % (daytime) and 27.6 % (night-time) in winter and 38.9 % (daytime) and 32.5 % (night-time) in summer. In addition, there are many other species that could contribute to OCs in PM2.5 such as aliphatic lipids, dicarboxylic acids, polycyclic aromatic hydrocarbons, and other complex compounds (e.g. proteins, amino sugars, and organosulfates) in the ambient atmosphere. GC-MS can only detect small molecules, which account for a small proportion of the total organic matter in the aerosols (Rogge et al., 1993). Thus, the realization of the full analysis of fine particles requires a combination of various analytical techniques such as 2D-GC-MS, HPLC-MS, and FT-ICR MS (Nozière et al., 2015), which are summarized in Table S8.