Molecular markers of biomass burning and primary biological aerosols in urban Beijing: Size distribution and seasonal variation

Biomass burning and primary biological aerosol particles account for an important part of urban aerosols. Floods of studies have been conducted on the chemical compositions of fine aerosols (PM2.5) in megacities where the haze pollution is one of the severe environmental issues in China. However, little is known about their size distributions in atmospheric aerosols in the urban boundary layer. Here, size-segregated aerosol samples were collected in Beijing during haze and clear days from 15 April 2017 to January 2018. Three anhydrosugars, six primary saccharides and four sugar alcohols in these samples were identified and quantified by gas chromatography/mass spectrometry (GC/MS). Higher concentrations of a biomass burning tracer, levoglucosan, were detected in autumn and winter than other seasons. Sucrose, glucose, fructose, mannitol and arabitol were more abundant in the bloom and glowing seasons. Particularly high level of trehalose was found in spring, which was largely associated with the Asian dust outflows. Anhydrosugars, xylose, maltose, inositol and erythritol are mainly existed in 20 the fine mode (<2.1 μm), while the others showed the coarse mode privilege. The concentrations of measured tracers of biomass burning particles and primary biological particles in the haze events were higher than those in the non-hazy days, with an enrichment factor of 2–10. Geometric mean diameters (GMD) of molecular markers of biomass burning and primary biological aerosols showed that there was no significant difference in the coarse mode (>2.1 μm) between the haze and non-haze samples, while a size shift towards large particles and large GMDs in the fine fraction (<2.1 μm) 25 was detected during the hazy days, which highlights that the stable meteorological conditions with high relative humidity in urban Beijing may favor the condensation of organics onto coarse particles.The contributions of reconstructed primary organic carbon (POC) by tracer-based methods from plant debris, fungal spores, and biomass burning to aerosol OC in the total mode particles were in the ranges of 0.09–0.30% (on average 0.21%), 0.13–1.0% (0.38%) and 1.2–7.5% (4.5%), respectively. This study demonstrates that the contribution of biomass burning was significant in Beijing throughout the whole 30 year with the predominance in the fine mode, while the contributions of plant debris and fungal spores dominated in spring and summer in the coarse mode, especially in the size of >5.8 μm. Our observations demonstrate that the sources, abundance, https://doi.org/10.5194/acp-2019-841 Preprint. Discussion started: 11 October 2019 c © Author(s) 2019. CC BY 4.0 License.

Abstract. Biomass burning and primary biological aerosol particles account for an important part of urban aerosols. Floods of studies have been conducted on the chemical compositions of fine aerosols (PM 2.5 ) in megacities where the haze pollution is one of the severe environmental issues in China. However, little is known about their size distributions in atmospheric aerosols in the urban boundary layer. Here, size-segregated aerosol samples were collected in Beijing during haze and clear days from April 2017 to January 2018. Three anhydrosugars, six primary saccharides and four sugar alcohols in these samples were identified and quantified by gas chromatography/mass spectrometry (GC/MS). Higher concentrations of a biomass burning tracer, levoglucosan, were detected in autumn and winter than in other seasons. Sucrose, glucose, fructose, mannitol and arabitol were more abundant in the bloom and glowing seasons. A particularly high level of trehalose was found in spring, which was largely associated with the Asian dust outflows. Anhydrosugars, xylose, maltose, inositol and erythritol are mainly present in the fine mode (< 2.1 µm), while the others showed the coarse-mode preference. The concentrations of measured tracers of biomass burning particles and primary biological particles in the haze events were higher than those in the non-hazy days, with enrichment factors of 2-10. Geometric mean diameters (GMDs) of molecular markers of biomass burning and primary biological aerosols showed that there was no significant difference in the coarse mode (> 2.1 µm) between the haze and non-haze samples, while a size shift towards large particles and large GMDs in the fine fraction (< 2.1 µm) was detected during the hazy days, which highlights that the stable meteorological conditions with high relative humidity in urban Beijing may favor the condensation of organics onto coarse particles.The contributions of reconstructed primary organic carbon (POC) by tracer-based methods from plant debris, fungal spores and biomass burning to aerosol OC in the total-mode particles were in the ranges of 0.09 %-0.30 % (on average 0.21 %), 0.13 %-1.0 % (0.38 %) and 1.2 %-7.5 % (4.5 %), respectively. This study demonstrates that the contribution of biomass burning was significant in Beijing throughout the whole year with the predominance in the fine mode, while the contributions of plant debris and fungal spores dominated in spring and summer in the coarse mode, especially in sizes > 5.8 µm. Our observations demonstrate that the sources, abundance and chemical composition of urban aerosol particles are strongly size dependent in Beijing, which is important to better understand the environmental and health effects of urban aerosols and should be considered in air quality and climate models. detrimental to human health due to the ability to deposit into deep parts of lungs (Nemmar et al., 2002). In addition, due to distinct sources of fine and coarse particles, particle sizes of organic species could act as a tool to identify OA sources and to assess their health effects.
Primary biological aerosol particles (PBAPs) and biomass burning (BB) are two major sources of atmospheric primary organic aerosols (POAs). PBAPs are those directly emitted into the atmosphere including bacteria, plant debris, fungal spores, and fragments of plants and animals Fang et al., 2018;Fu et al., 2012;O'dowd et al., 2004;Samaké et al., 2019). These PBAPs not only affect the hydrological cycle and climate change by means of acting as cloud and ice nuclei but also jeopardize human health Humbal et al., 2018;Rajput et al., 2018). Although recent studies have revealed the abundance and size distributions of PBAPs, studies regarding their emission sources and relative contributions remained poorly known (China et al., 2018;Di Filippo et al., 2013;Huffman et al., 2010;Jia et al., 2010;Tobo et al., 2013). Primary saccharides have been used as tracers for PBAP Medeiros et al., 2006;Wan et al., 2007). Sucrose, glucose and fructose are predominantly derived from plant materials, such as pollen, fruit and leaves (Pacini, 2000;Puxbaum and Tenze-Kunit, 2003). Amongst sugar alcohols, mannitol and arabitol are the primary ones. Mannitol, arabitol and trehalose have been recognized as tracers of airborne fungi (Bauer et al., 2008;Fu et al., 2016;Samaké et al., 2019;Verma et al., 2018). Sparse data about erythritol and inositol can be found in the literature. Recently, these two species were reported as having similar sources to other sugar alcohols, such as arabitol and mannitol .
As a highly developed economic zone, Beijing has been facing serious air pollution. Though the government has made great efforts to improve the air quality in Beijing, haze events still occur frequently, especially in winter Zheng et al., 2015). Moreover, air pollution, ac-companied with high abundances of fine particles, has become a serious environmental problem in China's cities and adjacent areas. There have been many studies on the composition and sources of aerosols in Beijing . However, few studies have been conducted regarding the size distribution, contribution and processes of POAs in different seasons in Beijing. Thus, research on chemical compositions and size distributions of PBAP and BB in Beijing is vital in terms of regional air quality.
In this study, size-segregated aerosols were sampled in Beijing from April 2017 to January 2018. Molecular characteristics, size distributions, and seasonal trends of anhydrosugars and other saccharides were investigated during haze and clean days. Concentrations and contributions of plant debris, fungal spores and biomass-burning-derived OC were estimated according to the tracer methods. The variations in size distributions of organic species in aerosols in different seasons were also assessed.
2 Experimental methods

Sample collection
All samples were collected using a nine-stage cascade impactor sampler (Andersen, U.S.A.) at a flow rate of 25.8 L min −1 from April 2017 to January 2018. The sampler was put on a second-story building rooftop in the Institute of Atmospheric Physics (IAP), Chinese Academy of Sciences in Beijing, China (39 • 58 28 N,116 • 22 16 E). The IAP is located on the north side of Beijing, which is considered to be a representative urban urea for a mixed district of traffic, residential and commercial areas. Urban aerosol samples collected here are influenced by both anthropogenic (e.g., traffic and cooking emissions) and natural sources (e.g., soil dust, plant debris and microorganisms). Before sampling, the quartz fiber filters were combusted at 450 • C for 6 h. The cutoff sizes are in the following sequence: > 9. 0, 9.0-5.8, 5.8-4.7, 4.7-3.3, 3.3-2.1, 2.1-1.1, 1.1-0.7, 0.7-0.4 and < 0.4 µm. Four sets of size-segregated aerosol samples per season and thus 16 sets (i.e., 144 samples) were obtained throughout the year. Blank filters were obtained for each season for quality control. Specific information including weather conditions, sampling periods and sample numbers was summarized in Table S1 in the Supplement. All samples were stored at a −20 • in freezer prior to analysis.

Sample extraction and derivatization
Using a disinfected scissor, each filter aliquot (∼ 6.4 cm 2 ) was cut and extracted three times with a dichloromethane / methanol mixture (2 : 1, v/v) under ultrasonic agitation for 10 min. After filtering, the solvent extracts were concentrated with a rotary evaporator and a stream of filtered dry nitrogen gas. The extracts were converted to their trimethylsilyl derivatives using mixture of 25 µL of Note that the samples are divided into spring (April-May), summer (June-July), autumn (October-November) and winter (December-January).

Gas chromatography/mass spectrometry measurement
An Agilent model 7890 GC coupled to an Agilent model 5975C mass-selective detector (MSD) was used to analyze organic compounds of these samples. A total of 2 µL of pretreatment sample was injected and separated in a fused silica capillary column. The column temperature program was optimized with a series of procedures: start and hold for 2 min at 50 • C, increase to 120 • C with a pace at 15 • C min −1 , then raise to 300 • C at 5 • C min −1 , and keep isothermal at 300 • C for 16 min. The mass spectrometer was operated on electron ionization (EI) mode at 70 eV and scanned from 50 to 650. Mass spectral data were acquired and processed with the Chemstation software. Individual compounds were identified by comparing mass spectra with those of authentic standards. The data reported here were corrected for the field blanks.

Measurements of OC and EC
Organic carbon (OC) and elemental carbon (EC) were measured using a thermal/optical carbon analyzer (model RT-4, Sunset Laboratory Inc., USA). A punch (17 mm diameter) of each aliquot or blank filter was used for the determination. The concentrations of OC and EC here were all corrected for field blanks.

Backward trajectories and fire spot counts
To better characterize the features and origins of air masses in Beijing, 3 d backward trajectory analyses were conducted using the HYSPLIT4 model (Draxler and Rolph, 2013) for all the dust storm and haze periods in our study. Backward trajectories were calculated at an altitude of 500 m above ground level. Influence of BB in urban Beijing is demonstrated by fire spots acquired from the MODIS website (https://earthdata.nasa. gov/earth-observation-data/near-real-time/firms, last access: 16 July 2019). The crowding level of the fire spots here represented the scales and intensity of BB. Cluster analyses were applied for typical dust storm and haze events in the present study ( Fig. S1 in the Supplement). Levoglucosan, a typical tracer of biomass burning, is formed from the thermal decomposition of cellulose during vegetation burning activities, which has been detected in many urban as well as rural areas Chowdhury et al., 2007;Fu et al., 2010;Wang et al., 2006;Yttri et al., 2007). As additional important types of anhydrosugars, mannosan and galactosan, the isomers of levoglucosan, are exclusively derived from the pyrolysis of hemicellulose (Fraser and Lakshmanan, 2000;Simoneit et al., 1999;Simoneit, 2002). The concentrations of and temporal variations in anhydrosugars (i.e., levoglucosan, mannosan and galactosan) are shown in Table 1 and Fig. 1. Among the three anhydrosugars, levoglucosan was the predominant compound (80.1-1230 ng m −3 , 365 ng m −3 ), which was 1 order of magnitude higher than mannosan (8.41-168 ng m −3 , 46.8 ng m −3 ) and galactosan (5.09-108 ng m −3 , 27.6 ng m −3 ). The concentration of levoglucosan was the highest in winter, followed by autumn, spring and summer. Similar seasonality has been documented in several studies in Beijing Liang et al., 2016). The abundances in summer and winter were comparable with previous studies Zhang et al., 2008). Mannosan and galactosan displayed a similar seasonal trend. Similar to levoglucosan, total anhydrosugars peaked in winter ( Concentrations of individual anhydrosugars in haze days were much higher than those in non-haze days in each season (Table 2). Moreover, the level of anhydrosugars (2310 ng m −3 , average) in winter was 1 order of magnitude higher than the other seasons (302, 171 and 798 ng m −3 for spring, summer and autumn, respectively). Such trends indicated that the haze samples were largely influenced by biomass burning emissions in Beijing, consistent with previous studies in Beijing (Zhang et al., 2017) and other regions such as New Delhi (India), Tasmania (Australia) and Kath- mandu (Nepal) Reisen et al., 2013;Wan et al., 2019).
During dust storm periods in spring, air masses primarily originated from middle-to-north Asia (MNA) ( Fig. S1a-b). Specifically, 71 %-75 % of air masses originated from westto-central Russia, passed through Mongolia and finally arrived in Beijing. Such a phenomenon is likely attributed to the influence of Asian outflow. Compared to 17 to 19 April, fewer fire spots were found in the movement route of air masses in 4 to 5 May, which can explain the lower levels of anhydrosugars in the aerosols. As time came to summer, the influence of air masses from central China was strengthened. In the meantime, frequent straw burning events were found in these regions due to the harvest season (Fig. S1cd). However, this emission had an insignificant impact on the anhydrosugars considering the relatively low concentrations. In winter haze periods, 53 %-100 % of the air masses originated from northern China, and the rest originated from Mongolia. Interestingly, much fewer fire spots could be observed ( Fig. S1g-h) compared to summer and autumn. In winter, domestic burning and small-scale open-field burning are much more intensive than in other seasons. However, these two burning activities are hardly detectable by satellite, which could account for the much higher abundances of anhydrosugars.
Pearson correlations among anhydrosugars and OC and EC were also studied (Table S2). Levoglucosan displayed a positive correlation with mannosan (r = 0.99, p < 0.001) and galactosan (r = 0.98, p < 0.001). Such consistent trends suggested that the anhydrosugars were emitted from similar sources. EC is considered to be closely related with BB (Akagi et al., 2011). Positive correlations between EC and levoglucosan (r = 0.45, p < 0.001), mannosan (r = 0.43, p < 0.001) and galactosan (r = 0.43, p < 0.001) indicate that BB was a major source of the anhydrosugars in urban Beijing aerosols.

Ratios of L/M, M/G, L/OC and L/EC
Previous surveys found that levoglucosan / mannosan (L/M) ratios vary with biofuel burning sources; those from hardwood, softwood and crop residues are often in ranges of 13-27, 2.5-5.8 and 25-55.7, respectively (Engling et al., 2006;Fine et al., 2004;Křůmal et al., 2010;Sheesley et al., 2003). The L/M values are shown in Table 1 and Fig. 2. The annual L/M ratios varied within a narrow range over the sampling periods (7.62-8.64, 8.27), indicating a mixture contribution of softwood burning and hardwood burning. The L/M ratios were relatively higher in winter (7.30-11.8, average 8.64) and spring (6.34-11.8, 8.49), consistent with those obtained in previous studies (Zhu et al., 2015). This is probably associated with open-field biomass burning, e.g., agricultural residues and deciduous leaves. The L/M ratios were smaller in summer and autumn, along with a higher level of M/G ratios than those in winter and spring. These results were likely attributed to the higher proportion of softwood burnings and the degradation of levoglucosan (Arangio et al., 2015;Mochida et al., 2010). The ratio of mannosan to galactosan (M/G) is used as an auxiliary means to distinguish the burning substrates because galactosan is more abundant than mannosan in combustion smokes of crop straws, grasses and briquettes (Fabbri et al., 2009;Oros et al., 2006;Vicente et al., 2018). The M/G ratios during all the periods were in a range of 1.35-2.08 with an average of 1.70 (Fig. 2b). The M/G ratios reached a maximum in autumn (1.68-1.97, 1.88) and a minimum in sum-  (Engling et al., 2006). The higher ratios in typical summer time in July (14 to 16 July and 21 to 23 July) were consistent with previous results in the North China Plain and western United States (Fine et al., 2004;Fu et al., 2008). The M/G ratios ranged from 1.49 to 1.89 (average 1.62) in winter, consistent with the higher abundance of galactosan over mannosan associated with the BB products from crop wastes in the North China Plain (Fu et al., 2008).
The L/M, L/OC and L/EC ratios in this study were compared with those reported for other Asian cities in recent studies. The L/M ratios showed different seasonal and spatial distributions among different cities. Those reported in Nepal (16.3 ± 5.96) were much higher than our study (Wan et al., 2019). A previous study in Okinawa found a L/M ratio of 10.7 ± 6.1 and exhibited a similar seasonal trend to our study (Zhu et al., 2015). Lower ratios of L/M suggested that more burning substrates were softwood in Beijing, while in previous studies in Beijing lower L/M ratios in winter were observed . Higher ratios of L/M in our study may be ascribed to the reduced use of coal in recent years in northern China. Coal combustion is also a potential source of levoglucosan, leading to a lower ratio of L/M (Yan et al., 2017). In comparison to annual mean L/OC ratios in Guangzhou and Zhaoqing in the Pearl River Delta (PRD) region (10.8 × 10 −3 and 27.5 × 10 −3 , respectively) (Ho et al., 2014), the lower values of L/OC in Beijing were potentially associated with the less extensive burning activities under the prohibition of open agricultural residue burning. Although the seasonal trend is similar to that of Okinawa, the L/EC ratios in Beijing were much higher than those in Okinawa. Besides the high abundance of levoglucosan in Beijing, these results may also imply that EC was associated with other sources, such as local transport emission and coal combustion, while EC in Okinawa might represent a regional background level (Zhu et al., 2015).
The L/M, M/G, L/OC and L/EC ratios according to the particle size are presented in Fig. 3. Higher L/M values were observed in fine particles during non-haze days, which were 1.02-1.71 times higher than those in coarse modes. In contrast, the L/M ratios did not vary significantly according to particle size in the haze days. The L/M values in fine modes during haze periods were slightly higher than those in coarse modes. The ratios between fine and coarse modes were close to unity. The elevated L/M ratios in fine modes suggested that hardwood was potentially the source of the burning substrates, while the relatively lower L/M ratios in coarse modes indicated the mixture impact of hardwood and softwood burning. The M/G ratios were also higher in fine modes than coarse modes with a factor of 1.0-2.5, suggesting a large impact of biomass burning from crop straws and grasses in fine modes. Moreover, such activities were more pronounced in haze days than non-haze days, implying the increasing contributions from crop straws burning in haze days.
The L/OC ratios showed a uniform dependence on particle sizes much higher in fine modes during all sampling periods. The average ratios of L/OC (fine modes) to L/OC (coarse modes) for four seasons were 6.31, 4.49, 11.3 and 9.47, respectively. Such a pattern was consistent with the finding that levoglucosan is mostly present in fine modes (Simoneit and Elias, 2001). The L/OC ratios in summer were a magnitude lower than in other seasons, which is likely attributed to the fewer biomass burning emissions in the urban areas of Beijing.
The L/EC ratios also exhibited a clear pattern with high values in fine modes and low values in coarse modes, similar to the L/OC ratios. Cao et al. (2005) found that the EC in PM 10 was more abundant than in PM 2.5 , while most levoglucosan was in fine modes, resulting in low L/EC ratios in coarse modes. Nevertheless, a high L/EC ratio was observed in the coarse modes during 4-5 May when Beijing encountered the most severe dust storm in that spring. Unlike the other samples, levoglucosan accounted for an unexpectedly high proportion in the coarse modes (discussed in Sect. 3.3.1). Meanwhile, the concentrations (23-57 × 10 −3 µg m −3 ) of EC in coarse modes were much lower in the dust period than in non-haze periods (30-211×10 −3 µg m −3 ) during spring. As a consequence, a low level of EC and high level of levoglucosan lead to the high L/EC ratio in the coarse mode.

Abundances and seasonal variations in primary
saccharides and sugar alcohols

Primary saccharides
The concentrations of primary saccharides and sugar alcohols in four seasons are shown in Table 1 and Fig. 4. The concentrations of primary saccharides and sugar alcohols averaged 620 and 139 ng m −3 , respectively. Except maltose and xylose, other primary saccharides exhibited higher abundance in spring and summer compared to autumn and winter. Among these compounds, sucrose (30.8-1590 ng m −3 , 267 ng m −3 ) was the most abundant one, followed by glucose (38.9-502 ng m −3 , 162 ng m −3 ) and fructose (7.97-72.5 ng m −3 , 30.1 ng m −3 ). Fructose and glucose are mainly derived from plant fragments (e.g., bark, leaves), pollen, lichen and soil dust (Dahlman et al., 2003;Fu et al., 2012;Pacini, 2000;Vesprini et al., 2002;Zhu et al., 2015). Glucose reached a maximum in spring (107-502 ng m −3 , 255 ng m −3 ), followed by summer (184-279 ng m −3 , 234 ng m −3 ), winter (60.0-139 ng m −3 , 86.4 ng m −3 ) and autumn (38.9-93.3 ng m −3 , 71.2 ng m −3 ). Fructose varied with a similar trend. High concentrations in spring likely coincided with agricultural tilling activities  and comparably active vegetation in the warmer season. The significant correlation between glucose and fructose (r = 0.81, p < 0.001) further indicated that these two compounds originated from com-mon sources (Table S3), consistent with the results of previous studies in East Asian areas Fu et al., 2012;Zhu et al., 2015). Previous studies found good correlations between fructose and glucose with OC in biomass burning aerosols over an Amazonia pasture site (Graham, 2002), in contrast to this study (Table S3), which suggested that biomass burning made minor contributions to glucose and fructose in Beijing.
Sucrose, a major constituent of pollen, peaked in spring during the blossom booming season, in agreement with previous studies (Fu et al., 2012;Pacini, 2000;Yttri et al., 2007;Jia et al., 2010). A previous study reported that sucrose, like glucose and fructose, could abundantly emit into the air along with the increased biological activities (Rogge et al., 2007). Accordingly, sucrose was much more predominant in spring (266-  rather than in summer. This is likely attributed to the agricultural activities in summer. In June and July, tilling activities frequently exposed crop (e.g., wheat) roots to the air, leading to the release of sucrose from the roots into the atmosphere.
Trehalose is present in various microorganisms (e.g., fungi and bacteria), some higher plants and invertebrates (Medeiros et al., 2006). In this study, trehalose exhibited a high abundance in spring (51.8-559 ng m −3 , 199 ng m −3 ), peaked in summer (101-378 ng m −3 , 232 ng m −3 ), and sharply reduced by nearly 75 % in autumn and winter. Trehalose is a useful tracer for soil input to the atmosphere in regions related to agricultural activities and dust storms (He et al., 2001;Simoneit et al., 2004). Thus, the concentration of trehalose was considerably high in a dust period during 4-5 May (559 ng m −3 ) and haze days during 30 June to 2 July and 12-14 July (Fig. 4h).
Xylose and maltose exhibited a totally different seasonal trend from other primary saccharides. Xylose showed high levels in autumn (10.2-29.7 ng m −3 , 22.2 ng m −3 ) and winter (11.0-32.7 ng m −3 , 20.7 ng m −3 ). Xylose is present in most woods and straw as a major constituent of hemicellulose Sullivan et al., 2011). The biomass burning activities in Beijing likely caused the elevated abundance of xylose, given the positive correlation between xylose and levoglucosan (r = 0.86, p < 0.001). Maltose, a microbial degradation product of starch, showed a similar seasonal pattern with xylose (Medeiros et al., 2006) and positively correlated with levoglucosan (r = 0.89, p < 0.001). These results suggest that biomass burning was a major source of xylose and maltose.

Sugar alcohols
The sum of sugar alcohols (29.8-418 ng m −3 , 139 ng m −3 ) was much lower than primary saccharides and anhydrosugars (Table 1). In the detected compounds, mannitol showed the highest level  arabitol reached a minimum in winter (5.52-33.8 ng m −3 , 14.1 ng m −3 ). Mannitol significantly correlated with arabitol during the whole sample period (r = 0.83, p < 0.001), indicating they are generally related to common sources. Mannitol and arabitol primarily derive from airborne fungal spores Medeiros et al., 2006;Zhang et al., 2010) and detritus of mature leaves (Pashynska et al., 2002). In summertime, fungal spores can be derived from multiple sources (e.g., plants, vegetation decomposition and agricultural activities), while in autumn and winter fungal spores are mainly suspended from exposed surfaces (Liang et al., 2013). Moreover, a previous study indicated that a higher leaf area index and atmospheric water vapor would facilitate the emission of mannitol (Heald and Spracklen, 2009). In agreement with that study, mannitol and arabitol in our study may be closely related to the activities of the terrestrial biosphere.
Erythritol and inositol were the less abundant of the calculated sugar alcohols. The abundance of erythritol was lowest in spring (3.16-15.8 ng m −3 , 9.92 ng m −3 ) and highest in summer (13.8-44.1 ng m −3 , 30.2 ng m −3 ), and it remained at a relative high level in autumn and winter (11.1-33.0 ng m −3 , 22.3 ng m −3 and 10.7-41.6 ng m −3 , 24.9 ng m −3 , respectively). Temporally, erythritol correlated well with levoglucosan (r = 0.50, p < 0.001), indicating a potential contribution from BB (Table S3). Results on inositol in aerosols were seldom reported . In this study, inositol showed maxima (20.7-69.2 ng m −3 , 41.5 ng m −3 ) in spring and minima (5.05-26.0 ng m −3 , 12.4 ng m −3 ) in winter. Inositol was positively correlated with trehalose (r = 0.65, p < 0.001) and exhibited a similar temporal variation with other sugar alcohols, suggesting developing leaves and plant debris may be a potential source.

Anhydrosugars
The size distributions of anhydrosugars for each sampling period are shown in Fig. 5. The concentrations of anhydrosugars in fine (< 2.1 µm) and coarse (≥ 2.1 µm) mode as well as whole size ranges are listed in Table 3. Among the class of anhydrosugars, levoglucosan was approximately an order of magnitude higher than other compounds in the fine and coarse particles. Levoglucosan is initially emitted from the flame and subsequently condenses onto preexisting particles as temperature decreases, and it is thus present mostly in fine particles (Simoneit and Elias, 2001). In this study, levoglucosan showed a unimodal size distribution with a peak at the ranges of 0.4-2.1 µm during most sampling periods (Fig. 5ad). The concentrations of levoglucosan in the fine mode were 1.8-9.3 times those in the coarse mode. Mannosan and galactosan also showed a size distribution pattern similar to levoglucosan during all sampling periods. Table 4. Geometric mean diameters (GMDs, µm) of anhydrosugars and other saccharides in the total-mode particles in Beijing from April 2017 to January 2018. GMD: logGMD = ( C i logD pi )/ C i , where C i is the concentration of a compound in size stage i and D pi is the geometric mean diameter of particles collected on stage i. Interestingly, all three anhydrosugars exhibited a unimodal size distribution in the coarse mode during the most serious dust storm period (4 to 5 May). Such distribution behaviors highlighted the significant contribution of dust storms to the loading of organic aerosols (OA) in Beijing. In a dust storm, strongly windblown coarse particles from a large-scale resuspension from crustal, soil, roads or other unpaved areas, along with long-range transport of particles from north and northwest desert regions, could induce a remarkable size shift of OA from the fine to the coarse mode. A similar result has reported such a phenomenon (Herner et al., 2006). The total-mode GMDs (geometric mean diameters) of these species, corresponding to the dust storm periods in spring, showed larger values (1.12-1.47 µm) than those in the nonhaze periods (0.62-1.04 µm). The typical summer haze period (30 June to 2 July) with a minor peak at 3.3-5.8 µm in the coarse mode was likely due to the meteorological conditions with higher humidity and lower wind speed. Previous studies found that the size of the ambient particle can be modified by chemical reactions; condensation/evaporation; coagulation with other particles; and activation during high humidity, wet and dry deposition (Herner et al., 2006). Anhydrosugars are highly water soluble compared with fossilfuel-derived particles, whose sizes increase with increasing relative humidity (Reid et al., 2005). During the haze period, the concentrations of all species were comparatively higher, especially in autumn and winter. As mentioned above, a larger scale of BB in cold seasons could lead to massive emission of these species. Moreover, autumn and winter were characterized by a low mixing layer and stagnant conditions in haze days, which benefit the accumulation of the pollutants (Balducci and Cecinato, 2010;Bigi and Ghermandi, 2011;Carbone et al., 2010;Perrone et al., 2012;Pietrogrande et al., 2014). The total-and coarse-mode GMDs of anhydrosugars in spring and summer were larger than in autumn and winter, while the fine-mode GMDs showed smaller values (Tables 4,  S4 and S5). Such altering of particle size has been observed in other places (Herner et al., 2006;Wang et al., 2011).

Primary saccharides
The size distributions of all primary sugars are shown in Figs. 6-7. Both glucose and fructose are present mostly in coarse modes (Fig. 6). Though the concentrations of glucose were much higher than fructose, these two compounds showed similar trends with a bimodal size distribution, peaking at 0.7-1.1 and > 5.8 µm for spring, 0.4-1.1 and 3.3-5.8 µm for summer and autumn, and 0.4-1.1 and 4.7-9 µm for winter. Compared to the smaller contribution in the fine mode in spring and summer (5.9 %-8.5 % and 14.5 %-26.7 % for glucose and fructose, respectively), the fractions of these two compounds were more abundant in fine aerosols during autumn and winter (15.0 %-24.4 % and 32.6 %-45.8 % for glucose and fructose, respectively). How-S. Xu et al.: Size distribution of saccharides in urban aerosols Figure 6. Size distributions of glucose, fructose, sucrose, mannitol and arabitol in urban aerosols in Beijing. ever, the origins of glucose and related sugars in fine particles remained unclear. Former studies suggested that fragmented pollen grains, either in cytoplasmic vesicles or dissolved in the cytosol, could be sources of sugars in the fine aerosol (Pacini, 2000;Yttri et al., 2007).
The size distribution of sucrose is shown in Fig. 6i-l. Sucrose showed a bimodal size distribution with high concentrations in the coarse mode during the sampling periods. A minor peak was found in the size ranges of 0.4-1.1 µm and a major peak was found at sizes > 5.8 µm. Compared to the other three seasons (approximately 14.9 %, 18.8 % and 19.1 % for summer, autumn and winter, respectively), sucrose was not as prevalent in the fine mode (3.3 %) in spring. Such a phenomenon could be ascribed to the prevailing of pollen grain emission during the spring bloom season (Xu et al., 2012). Previous studies reported that most pollen of grasses actually remains in the anthers, and the cycle of wetting-drying causes the pollen to rupture from the anthers, with the pollen in the size range of 120 nm to 4.67 µm (Taylor et al., 2002). Meanwhile, furious windblown coarse aerosols from resuspension of dust in spring could further lead to higher abundance. Glucose, fructose and sucrose exhibited the largest total-and coarse-mode GMDs in dust storm aerosols, suggesting the effect of dust storms on par- ticles which could contribute a large number of sugars (Tables 4, S5). Trehalose resembles arabitol and mannitol more closely, which is mostly present in the coarse mode with a proportion of over 90 % and showed a nearly unimodal size distribution during the whole sampling periods (Fig. 7a-d). This result indicates that trehalose originated from fungal spores, consistent with former studies (Lewis and Smith, 1967;Medeiros et al., 2006). The same as sucrose, trehalose was greatly influenced by the dust storm in spring. The concentrations of trehalose in dust storm were 4-10 times higher than in nonhaze periods (Fig. 7a), indicating the essential impact of dust storms. Also, trehalose showed the largest total, fine-and coarse-mode GMDs in dust storm periods (Tables 4, S4-S5). This phenomenon was in good agreement with the observation in Seoul (Jeon et al., 2013). Noticeably, the totaland coarse-mode GMDs of trehalose in winter non-haze days were much larger than those in haze days. We assumed that soil resuspension by wind input a substantial amount of large particles into the air, but the mechanism needs to be further discussed.

Atmos
Different from other sugars and sugar alcohols, xylose and maltose showed a bimodal size distribution (Fig. 7). Except for summer, xylose and maltose were observed with a major peak at 0.4-1.1 µm and a minor peak at 3.3-5.8 µm in the other three seasons. In summer, xylose and maltose were mostly present in coarse aerosols and even peaked at 3.3-9 µm during 30 June to 2 July. The distribution patterns of xylose and maltose were similar to those of anhydrosugars, suggesting biomass burning may be a dominant sugar source in the fine mode. Xylose and glucose were positively correlated well with each other (r = 0.83, p < 0.001) in coarse modes, whereas the origin of maltose in coarse modes remained unclear in this study. The correlations between maltose and glucose and between maltose and sucrose were poor (r = −0.04 and r = 0.01, respectively), in contrast to former studies that showed that maltose is particularly important in developing flower buds (Bieleski, 1995), indicating that maltose may be derived from other sources.

Sugar alcohols
The size distributions of sugar alcohols are illustrated in Figs. 6-7. Mannitol and arabitol showed similar trends while erythritol and inositol exhibited a different distribution pat-tern. Both mannitol and arabitol showed a unimodal size distribution in spring and summer, while they showed bimodal size distributions in autumn and winter. Mannitol and arabitol peaked at > 9.0 µm and 3.3-9.0 µm in spring and summer, respectively, while they exhibited more fine aerosols in winter than in warm seasons. Such a phenomenon was attributed to various sources of mannitol and arabitol. In summer and autumn, fungal spores in the atmosphere can be derived from complex sources (e.g., plant leaves, agricultural activities), which may contribute more to coarse particles than those in winter (Yttri et al., 2007), while in winter there are fewer plant cover and agricultural activities, resulting in a higher fraction in the fine mode. Former studies suggested that mannitol was present in aerosol samples influenced by wildfire smoke, and mannitol and arabitol might be emitted by thermal stripping during wildfires Simoneit et al., 2004). These sources can also effectively elevate the abundance of mannitol and arabitol in the fine mode in winter. Compared to other sampling periods, mannitol and arabitol in the dust storm exhibited the largest GMD values, which was ascribed to large emissions from the resuspension of fungal spores from the surface soil. Previous studies also found that elevated concentrations of trehalose, mannitol and arabitol were generally related to resuspended soil and the outflow of dust storms (Liang et al., 2013;Rogge et al., 2007). In summer, autumn and winter haze days, the total-and coarse-mode GMDs of mannitol and arabitol were generally larger than those in non-haze (Tables 4 and S4). Inositol showed a bimodal size distribution in four seasons, while the pattern varied with season. In spring, the concentrations of inositol were high in the coarse mode with a peak at > 9.0 µm, while in other seasons the concentrations of inositol peaked at 0.4-1.1 µm. The total-mode GMDs of inositol in the dust storm samples were much larger than those in other non-haze and haze samples, indicating the substantial influence of dust. The size pattern of inositol was similar to those of arabitol and mannitol, potentially responding to fungal spores in the local terrestrial biosphere. In this study, inositol was positively related to levoglucosan in the fine mode (r = 0.77, p < 0.001), which could explain the fine-mode preference in cold seasons. Such a correlation was further ascertained by the total-and fine-mode GMDs of inositol, for its similarity with that of anhydrosugars.
Erythritol presented a bimodal size distribution for all sampling periods. In summer, erythritol is mostly seen in the coarse mode with a major peak at 3.3-5.8 µm and a minor peak at 0.4-1.1 µm, while in other seasons the distribution patterns reversed with the prevailing of the fine mode and peaked at 0.4-1.1 µm. The source of erythritol in the fine mode was different from that in the coarse mode. The positive relationship between erythritol and levoglucosan (r = 0.90, p < 0.001) in the fine mode suggests the dominant contribution from biomass burning to erythritol. However, erythritol covaried with arabitol (r = 0.59, p < 0.001) in the coarse mode, suggesting that erythritol is partly of local origin.
Larger GMDs of primary carbonaceous particles in the fine mode in haze days than in non-haze days have been observed in both previous studies and the present study. One reason is the enhanced hygroscopic growth of airborne particles under relatively high humidity. As shown in Table S1, haze days were characterized by high relative humidity. Previous studies found that all the fine modes of WSOC (water-soluble organic carbon) showed a larger GMD in hazy days, and the fine-mode GMDs of aliphatic hydrocarbons were generally larger in haze samples due to higher relative humidity (Kang et al., 2016;Wang et al., 2011). In addition, the increased concentrations of airborne aerosols could increase the overall coagulation rate, resulting in a faster rate of collisions between particles and the formation of larger particles (Herner et al., 2006). Table 2 shows that the concentrations of anhydrosugars, primary saccharides and sugar alcohols were much higher in haze days, which were responsible for the larger GMDs in the fine mode and a size shift towards large particles.

Abundance and contributions of OC from biomass burning, plant debris and fungal spores
Based on the tracer methods, the quantitative estimates of OC from biomass burning (BB), plant debris and fungal spores are shown in Table 5 and Figs. 8-9. In a combustion test in fireplaces conducted by Fine et al. (2004), the emission ratio of OC/levoglucosan mass was 7.35 on average, and high ratios of 8.0 %-8.2 % are probably typical for open fires of burning grasses, woods and agricultural residues (Andreae and Merlet, 2001;Zhang et al., 2007). Considering the burning of agricultural wastes, an L/OC ratio of 8.2 % was applied in this study. Except for summer, BB-OC was the most dominant and showed an order of magnitude higher than the other two OC contributors. The level of BB-OC reached a maximum in cold seasons (6.81 µg m −3 for winter, 6.11 µg m −3 for autumn) and a minimum in summer (1.39 µg m −3 ). The contributions of BB-OC reached the highest level (7.52 %) in winter, followed by autumn (5.51 %), and the lowest (1.21 %) in summer (Table 5 and Fig. 8b). Apart from the elevated abundance of levoglucosan, stable meteorological conditions would also lead to the accumulation of BB-OC in cold seasons (He et al., 2001). Compared to BB-OC, the contributions of plant debris and fungal spores exhibited totally different patterns. Based on the relationship between glucose and plant debris (Puxbaum and Tenze-Kunit, 2003), the plant-debris-derived OC was estimated. The plant-debris-derived OC showed a seasonal trend with higher concentrations in spring (0.34 µg m −3 ) and summer (0.31 µg m −3 ) but lower in autumn (0.09 µg m −3 ) and winter (0.11 µg m −3 ). Likewise, the contributions of plant debris to OC were relatively larger in spring and summer (both 0.30 %), but they only accounted for 0.01 % in autumn and 0.16 % in winter. Mannitol, a tracer for fungal spores, was applied to estimate the contribution of fungal spores to OC using an experimentally derived factor of 1.7 pg mannitol and 13 pg OC per fungal spore (Bauer et al., 2008;Carvalho et al., 2003;Graham et al., 2003). In general, the abundance of fungal-spore-derived OC exhibited a similar trend with plant-debris-derived OC, which reached a maximum in summer (1.15 µg m −3 ) and reduced in autumn and winter (0.16 µg m −3 ). The contributions of fungal spores to OC reached a maximum in summer (1.00 %) and a minimum in autumn (0.13 %).
The contributions of OC from BB, plant debris and fungal spores according to particle size are shown in Fig. 10. BB-OC accounted for more than 85 % of the three species in the fine modes. It was more pronounced in autumn and winter at approximately 100 %. The contributions of BB-OC in coarse modes were of two patterns. In the size range of 2.1-9.0 µm, the relative contributions of OC were higher in spring, autumn and winter, with a proportion over 60 %, and exhibited the smallest fraction in summer, while in the size range larger than 9.0 µm, the smallest fraction of BB-OC appeared in spring, and the relative contributions were below 60 % in most periods. Plant-debris-derived OC was mostly present in the coarse modes. In the size range larger than 9.0 µm, the plant-debris-derived OC was the dominant one and peaked in spring (Fig. 9). Similar to plant-debris-derived OC, fungal-spore-derived OC also showed a high contribution in the coarse mode. The annual contributions of fungal spores to OC ranged from 0.04 % to 2.73 % (0.56 %) in the coarse mode, while those in the fine mode were just 0.03 %-0.14 % (0.06 %). Burge (2002) found that fungal spores in the atmosphere were present predominantly in the size range of 2-10 µm. Previous studies also found similar results showing that mannitol mainly occurs in the coarse size fraction (Carvalho et al., 2003;Graham et al., 2003;Samaké et al., 2019).
The relative contribution decreased in the smaller size ranges and was much higher in summer compared to other seasons. The contributions of fungal spores to OC were particularly high (1.48 %) in the coarse fraction in summer, indicating the multiple sources of fungal spores in summertime. Both plant debris OC and fungal spore OC were present in fine modes, especially in spring and summer, which has been observed in several sites (Carvalho et al., 2003;Graham et al., 2003;Yang et al., 2012). It was worth mentioning that, though the contributions of plant debris and fungal spores in the whole range were insignificant in cold seasons, their contribution in the coarse mode could not be ignored.
Compared to the contributions of BB, plant debris and fungal spores in non-haze day, the total contribution to OC showed a larger fraction in haze days for each season with enriched factors of 1.2-9.1. However, the reasons for elevated contribution in hazy days varied with season. In summer, the enhanced concentration of fungal-spore-derived OC was the primary causation ( Fig. 9), especially in the size range of 3.3-9.0 µm. The levels of plant-debris-derived OC in haze days were about 2.4-4.0 times higher than that in non-haze days, while BB-OC and fungal-spore-derived OC showed no significant differences (Figs. 8-9). Hence, the contribution to POC in haze days was seen more in the coarse mode than in non-haze days, whereas in autumn and winter the elevated total contribution is attributed to the increased concentration of BB-OC. Figure 9f-i shows that BB-OC in the size range of 0.4-2.1 µm of haze samples was much higher than that in non-haze samples. As a result, the prevalence of BB-OC lead to more total contribution in the fine mode and more in haze days.

Conclusions
Atmospheric concentrations; seasonal variations; and size distributions of anhydrosugars, primary saccharides, and sugar alcohols were investigated in urban aerosols from Beijing, China. High concentrations of anhydrosugars in the cold seasons were associated with the enhancement of biomass burning activities and adverse meteorological conditions. High levels of primary bioaerosol tracers such as sucrose, fructose, glucose, mannitol and arabitol were found in bloom and glowing seasons. The predominance of trehalose, a known tracer for Asian dust, was found during the most severe dust storm event (4-5 May). Anhydrosugars, xylose, maltose, inositol and erythritol were mostly present in the fine mode, while other primary sugars and sugar alcohols dominated in coarse particles. Dust storm greatly enhance organic aerosol concentrations and induce a remarkable size shift towards coarse sizes. Fine-mode GMDs of biomass burning aerosols and primary biological aerosol particles were larger in the haze days than in the non-haze days, probably due to the stable meteorological conditions with higher humidity, which favor the condensation of organic matter onto aerosol particles, especially in the coarse mode.
Based on the tracer-based methods, the contributions of biomass burning, plant debris and fungal spores to OC were calculated with a percentage of 1.21 %-7.52 %, 0.09 %-0.30 % and 0.13 %-1.00 %, respectively. The contribution of biomass-burning-derived OC predominated in the fine mode throughout the year, although the contribution was not negligible in the size range of 2.1-9.0 µm. The contribution of plant-debris-derived OC dominated in the coarse mode and decreased with the decreasing size. The contribution of fungal-spore-derived OC dominated in the size range of 3.3-5.8 µm in summertime samples. Though the sources of plant debris and fungal spores prevailed in spring and summer, small contributions in the coarse mode were also found during cold seasons due to long-range atmospheric transport from the southern regions. To better characterize sources and formation processes of haze particles and atmospheric chemistry in the North China Plain, further investigations on the size distributions of biogenic and anthropogenic secondary organic aerosol (SOA) tracers are strongly needed in the future.
Data availability. The dataset for this paper is available upon request from the corresponding author (fupingqing@tju.edu.cn).
Supplement. Information on sampling parameters from April 2017 to January 2018 (Table S1). Linear correlation coefficients among anhydrosugars and OC, and with EC in Beijing aerosols from April 2017 to January 2018 (Table S2). Linear correlation coefficients among primary saccharides and sugar alcohols, and with levoglucosan in Beijing aerosols from April 2017 to January 2018 (Table S3). Geometric mean diameters (GMDs, mm) of anhydrosugars and other saccharides in the fine-mode (< 2.1 µm) and coarse-mode (≥ 2.1 µm) particles in Beijing (Tables S4 and S5). Clusters of 3 d backward trajectories of air masses arriving at Beijing during dust storm events and haze periods from April 2017 to January 2018 (Fig. S1). The supplement related to this article is available online at: https://doi.org/10.5194/acp-20-3623-2020-supplement.
Competing interests. The authors declare that they have no conflict of interest.
Author contributions. PQF designed this research. Field campaigns were organized by PQF, SJH and YCL. Aerosol samples were collected by SJH. Laboratory analyses were performed by SFX, LJR, HR and SJH. The paper was written by SFX and PQF in consultation with and with editing from all other authors.
Financial support. This research has been supported by the National Natural Science Foundation of China (grant nos. 41625014, 41581130024 and 41807303) and the State Key Joint Labora-tory of Environment Simulation and Pollution Control (grant no. 18K02ESPCT).
Review statement. This paper was edited by Qiang Zhang and reviewed by two anonymous referees.