Organic tracers of fine aerosol particles in central Alaska: summertime composition and sources

PM2.5 aerosols were collected at Fairbanks (64.51 N and 147.51W) in central Alaska during the summer of 2009 and analyzed for organic tracer compounds using a gas chromatograph–mass spectrometer. The organic compounds were grouped into 14 classes based on their functional groups and sources. Concentrations of the total organics measured ranged from 113 to 1664 ng m−3 (avg 535 ng m−3). Anhydrosugars (avg 186 ng m−3) and nalkanoic acids (avg 185 ng m−3) were 2 major classes among the 14 compound classes. The similar temporal trends and strong positive correlations among anhydrosugars and nalkanoic acids demonstrated that biomass burning (BB) is the major source of organic aerosols (OAs) in central Alaska. The dominance of higher molecular weight n-alkanoic acids over lower molecular weight homologs and their carbon preference index (5.6–9.8) confirmed that they were mostly emitted from plant waxes during BB in central Alaska. The mass concentration ratios of levoglucosan to mannosan denoted that softwood is the main biomass burned. The rainfall event distinctly enhanced the levels of mannitol and arabitol due to the growth of fungi and active discharge of fungal spores in the subarctic region. Molecular compositions of biogenic secondary organic aerosol (BSOA) tracers inferred that isoprene is a crucial precursor of BSOA over central Alaska. Our results suggest forest fires and plant emissions to be the crucial factors controlling the levels and molecular composition of OAs in central Alaska. We propose that PM2.5 laden with OAs derived in central Alaska may significantly impact the air quality and climate in the Arctic via long-range atmospheric transport.


Introduction 34
Atmospheric aerosols can absorb and scatter solar radiation and alter the radiative forcing of 35 the atmosphere (Seinfeld and Pandis, 1998;Wilkening et al., 2000). Fine aerosol particles 36 have a diameter size close to the wavelengths of visible lights and thus are expected to have a 37 stronger climatic impact than coarse particles (Kanakidou et al., 2005). They can also be 38 transported far away from the source regions and thus their climatic and environmental effects 39 are delocalized compared to the emission areas. Aerosol particles that are hydrophilic can act 40 as cloud condensation nuclei (CCN) and have an indirect climatic effect through modification 41 of cloud properties (Novakov and Penner, 1993;Novakov and Corrigan, 1996). 42 Organic aerosols (OAs) that are comprised of thousands of organic compounds 43 contribute about 20 to 50% of total mass of fine particles in the continental mid-latitudinal 44 atmosphere (Saxena and Hildemann, 1996) whereas it is around 90% in tropical forest areas 45 (Crutzen and Andreae, 1990; Andreae and Rosenfeld, 2008). They are derived from 46 anthropogenic and natural sources. They can alter the physical and chemical properties of 47 atmospheric particles depending on the meteorological conditions. OAs are highlighted for 48 the past decade because they are related to the changes of global and regional climate and 49 chemical composition of the atmosphere as well as public health. Primary organic aerosols 50 (POA) are directly emitted as particulate forms whereas secondary organic aerosols (SOA) 51 refer to particulate organic matters that are transformed to aerosol-phase via gas-phase 52 oxidation of organic precursors. Emissions of POA particles and SOA precursors can be 53 released from numerous sources near the ground surface and subsequently mixed in the 54 boundary layer and to a lesser extent in the free troposphere. The dry depositional removal of 55 OAs mainly depends on the sizes of the aerosol particles. 56 The molecular composition of OAs can be used as tracer to better understand the 57 sources and formation pathways. Advances were made during the last decade to better 58 understand the formation of OAs and their precursors in the atmosphere. On a global scale, 59 the emission of biogenic volatile organic compounds (VOCs) is one order of magnitude 60 higher than that of anthropogenic VOCs (Seinfeld and Pandis, 1998). It is notable that 61 biogenic VOCs are comprised of unsaturated hydrocarbons with double bonds and are more 62 reactive towards the atmospheric oxidants such as hydroxyl (OH) radical and ozone (O 3 ) than 63 anthropogenic VOCs that are largely comprised of aromatic hydrocarbons. This specific 64 feature of biogenic VOCs further enhances their significance as a conceivable supplier to the 65 global burden of OAs in the atmosphere. Laboratory and chamber experiments have also 66 containing 1.43 ng µl -1 of a C 13 n-alkane internal standard (40 µl) was added into the 165 derivatives before injection of the sample into a GC-MS. 166 The separation of compounds was performed on a 30 m long DB-5MS fused silica 167 capillary column (0.25 mm inner diameter and 0.25 µm film thickness). Helium was used as a 168 carrier gas at a flow rate of 1.0 ml min -1 . The GC oven temperature was programmed from 50 169 °C for 2 min to 120 °C at 30 °C min -1 and then 300 °C at 6 °C min -1 with a final isotherm hold 170 at 300 °C for 16 min. The sample was injected on a splitless mode with the injector 171 temperature of 280 °C. The mass detection was conducted at 70 eV on an electron ionization 172 mode with a scan range of 50 to 650 Daltons. The organic compounds were determined by the 173 comparison of the GC retention times and mass fragmentation patterns of a sample with those 174 of authentic standards and National Institute of Standards and Technology library data. The 175 mass spectral data were acquired and processed using HP Chemstation software. GC-MS 176 relative response factor of each compound was calculated using authentic standards or 177 surrogate compounds. The recoveries of authentic standards or surrogates were above 80% 178 for target compounds. The data reported here were not corrected for recoveries. The relative 179 standard deviation of the measurements based on duplicate analyses was within 10%. The 180 field blank filters were analyzed by the procedure described above. The target compounds 181 were not detected in the blank filters. site. The daily mean temperature was in a range of 2.0 to 33 °C with an average of 13.9 °C 185 whereas the daily average relative humidity ranged from 19 to 99 % with a mean of 63 %. 186 The mean wind speed was 5.2 km h -1 and the total rainfall was 122 mm during the sampling 187 period. The 5-days air mass backward trajectories at the height of 500 m above the ground 188 level were computed from Hybrid Single Particle Lagrangian Integrated Trajectory model 189 (Draxler and Rolph, 2013). The air mass backward trajectories arriving over the observation 190 site during the collection of aerosol samples is presented in Figure 3. 191 3 Results and discussion 192

Overview of the molecular composition of organic aerosols 193
A total of 96 organic compounds were detected in PM 2.5 samples collected at Fairbanks 194 during the sampling period. We grouped them into fourteen compound classes as listed in 195 Table 1 together with the mean concentrations and ranges. Figure 4 shows the chemical 196 The emission strength of BB products and their long-range atmospheric transport 231 influence the atmospheric levels of anhydrosugars. The backward trajectories reveal that air 232 masses mostly came from the ocean during the campaign (Fig. 3). This result shows that 233 anhydrosugars present in the Alaskan aerosols were mainly associated with the local and 234 regional BB during the campaign. The higher level of levoglucosan in Fairbanks than other 235 sites in the Arctic implies a possible effect of BB on the air quality and climate in the arctic 236 region. Stocks et al. (2000) and Grell et al. (2011) proposed that the frequency of boreal forest 237 fires recently increased in summer due to global warming. Figure 6a-c show the temporal 238 trends of anhydrosugars in the Alaskan aerosols. The levels of anhydrosugars expressively 239 alter during the campaign period. The lower levoglucosan levels were found at the beginning 240 of the campaign whereas they became very high (241 to 463 ng m -3 ) in 4-23 July (Fig. 6a). 241 Another peak of levoglucosan was found in 30 July to 4 August (169 ng m -3 ). The 242 concentrations of levoglucosan decreased towards the end of the campaign (23 to 50 ng m -3 ). 243 Forest fires smokes were seen during 4-23 July and 30 July to 4 August over central Alaska. 244 This observation demonstrates that levoglucosan levels became high due to the local forest 245 fire in central Alaska. Mannosan and galactosan presented similar temporal variations with 246 levoglucosan (Fig. 6b and c). The chemical reaction of anhydrosugars with OH radicals could 247 also influence their concentrations in the atmosphere. Although previous studies have 248 reported that levoglucosan can remain stable in the atmosphere for around 10 days with no 249 substantial degradation (Fraser and Lakshmanan, 2000;Schkolnik and Rudich, 2006) experiment and reported the lifetime of atmospheric levoglucosan to be 0.7 to 2.2 days when 254 exposed to 1 × 10 6 molecules of OH cm -3 . This lifetime is within the range of 0.5 to 3. an atmospheric lifetime of levoglucosan to be 26 days when exposed with OH level of 2 × 10 6 260 molecules cm -3 that is much longer than other predictions. 261 It is notable from the above discussion that the degradation of levoglucosan is mostly 262 induced by photochemical aging via oxidation by OH radicals during long-range transport. 263 Therefore, the degradation of levoglucosan could be insignificant if the receptor site is close 264 to the source region. As discussed previously, anhydrosugars detected in Alaskan aerosols 265 during the campaign originated from local and regional BB, we consider that the degradation 266 of anhydrosugars may not be important to explain the low levels of BB tracers in the samples 267 collected at the beginning and end of the campaign. The low concentrations of anhydrosugars 268 during the beginning and end of the campaign might be caused by the decreased emission rate 269 of BB tracers due to lower BB activities in the source region. Wet deposition may be another 270 cause to lower the level of anhydrosugars in aerosol samples collected at the beginning and 271 end of the campaign because we observed rainfall especially in 5 June to 3 July and 6 August 272 to 17 September in Fairbanks (Fig. 2). Although the concentrations of both mannosan and 273 galactosan are much lower than levoglucosan (Fig. 5a), we observed strong positive 274 correlations (r = 0.94-0.97) among these tracers (Table 2). This result indicates that they   Bases on their data, we calculated the L/M ratios to be 3.4 to 6.7 for softwood burning and 0.24. The concentration ratio of syringic to vanillic acid can therefore be used as a marker to 318 distinguish the type of vegetation burned. We found that syringic to vanillic acid ratios in 319 Fairbanks aerosols ranged from 0.02 to 0.5 (ave. 0.2), suggesting that softwood is more 320 important biomass burned in central Alaska during the campaign. This conclusion is 321 consistent with the observation on the L to M ratios as discussed above. The temporal 322 variation of 4-HBA is very similar to that of anhydrosugars whereas vanillic and syringic 323 acids presented rather similar temporal trends with DHAA in Alaskan aerosols ( Fig. 6d-g). 324 Simoneit et al. (1993) proposed that the emission of DHAA is different than those of lignin 325 and cellulose burning products and therefore it is a more specific molecular marker of the 326 burning of conifer trees. The concentrations of DHAA ranged between 0.9 and 19 ng m -3 327 (ave. 6.1 ng m -3 ), which are higher than those of lignin pyrolysis products (Fig. 5b). This 328 result suggests that the burning of conifer is a common source of OAs in central Alaska. 329

Lipids: tracers of leaf waxes and marine sources 330
Series of lipid class compounds, including n-alkanes (C 21 to C 33 ), n-alkanols (C 8 to C 30 ) and 331 n-alkanoic acids (C 12 to C 32 ) were detected in Alaskan aerosols. n-Alkanoic acids are the 332 major lipid class compounds in Alaskan aerosols (ave. 185 ng m -3 ), which is several times 333 higher than those of n-alkanols (ave. 46 ng m -3 ) and n-alkanes (ave. 24 ng m -3 ) ( Table 1). 334 plant-derived n-alkanes in total n-alkanes ranged from 53 to 70 % (ave. 61 %), implying that 370 leaf wax is a major source of n-alkanes in the Alaskan aerosols. 371 The average molecular characteristics of n-alkanols and n-alkanoic acids displayed 372 even-carbon-number predominance ( Fig. 7b and c). n-Alkanols presented maxima at n-alkanoic acids are in the range of 3.0 to 10 (ave. 6.2) and 5.6 to 9.8 (ave. 7.9), respectively, 384 suggesting a large contribution of plant waxes to lipid class compounds in central Alaskan 385

aerosols. 386
The concentrations of n-alkanes and n-alkanols slightly decreased from June 05-12 387 to late June samples (June 25 to July 04) and then dramatically increased in July 04-06 388 sample ( Fig. 6h and i). The concentration peaks of n-alkanes and n-alkanols were also 389 observed in sample of July 14-23 whereas their concentrations constantly decreased from July 390 30 to the end of the campaign. The levels of n-alkanoic acids were low at the beginning of the 391 campaign and then increased drastically in July 04-06 sample and remained high in two 392 samples collected in July 06-23 (Fig. 6j). Concentrations of n-alkanoic acids decreased from 393 July 30 to September 21. Fascinatingly, the temporal variations of lipid class compounds were 394 similar to those of anhydrosugars (Fig. 6a-c and h-j). We also detected unsaturated n-alkanoic acids in Alaskan aerosol samples. Oleic showed strong positive correlations with fructose (r = 0.91) and sucrose (r = 0.82) ( Table 2). 426 Fructose also presented a strong correlation with sucrose (r = 0.94) ( Table 2) reported in airborne pollen grains produced from blooming plants (Pacini, 2000), surface soil 435 and associated microbiota (Simoneit et al., 2004b) and dehydrated plant materials 436 (Ma et al., 2009). We found that glucose shows moderate correlation (r = 0.48) with 437 levoglucosan (Fig. 8d). Shafizadeh and Fu (1973)  where it can be stored in a large amount as deposited or dissolved free molecules. The 443 temporal trend of glucose showed a peak in the sample collected during July 14-23 (Fig. 6k). 444 Interestingly, the same sample showed a high loading of DHAA that is a unique tracer of the 445 burning of conifer trees (Fig. 6g). This result suggests that the burning of conifer plants is the 446 source of glucose in central Alaska. August 08 and then increased towards the end of the campaign when rainfall occurs in central 451 Alaska (Fig. 2 and Fig. 6m). This result shows that the major source of trehalose might be the 452 fungi in the surface soil of central Alaska that was emitted after the rainfall event. Terrestrial 453 plants and marine phytoplankton as well as soil dust particles and associated microorganisms 454 release xylose into the atmosphere (Cowie and Hedges, 1984). Although xylose is a minor 455 primary sugar in Alaskan aerosols (ave. 1.1 ng m -3 ), its temporal trend is very similar to that 456 of anhydrosugars (Fig. 6a-c and n). This result together with a strong positive correlation of 457 xylose with levoglucosan (r = 0.92) implies its BB origin in central Alaska (Fig. 8g). This Sugar alcohols presented the predominance of arabitol (ave. 6.6 ng m -3 ) and mannitol 461 (ave. 6.2 ng m -3 ) (Fig. 9b). The concentration levels of erythritol (ave. 1.0 ng m -3 ) and inositol 462 (ave. 0.3 ng m -3 ) are much lower than those of arabitol and mannitol in Alaskan aerosols. 463 Arabitol and mannitol concentrations were higher during the beginning and end of the 464 campaign than those during the middle of the campaign (Fig. 6o and p). We found that  The ambient concentrations of total phthalate esters ranged from 0.4 to 6.6 ng m -3 506 (ave. 1.7 ng m -3 ), which are slightly higher than those from the North Sea to the high Arctic 507  (Fig. 12). The dominance of PA over PNA in 571 summertime can be explained by the much lower vapor pressure of PA than that of PNA. 572 However, this pattern is different from those found in summertime aerosols at the summit of 573 Mt. Tai in China  and other sites in Europe (Kavouras and Stephanou, 2002)  reported that biogenic VOCs could also be emitted from biomass burning. Our result showed 601 a high level of β-caryophyllinic acid in the samples that were affected by BB in central 602 Alaska. Ciccioli et al. (2014) proposed that sesquiterpenes could be accumulated in leaves and 603 wood because of low volatility and then abundantly emitted upon heating. The temporal trend 604 variation of β-caryophyllinic acid is similar to those of anhydrosugars (Fig. 6a-c and 11l). 605 Interestingly, we found a strong correlation (r = 0.98) of β-caryophyllinic acid with 606 levoglucosan (Fig. 8j), again indicating that forest fire largely contributes to the formation of 607 β-caryophyllinic acid in central Alaska. 608

Aromatic and polyacids: tracers of SOA 609
We detected benzoic acid in the Alaskan aerosol with the concentration range of 0.1 to 0.9 610 (ave. 0.3 ng m -3 ). Benzoic acid is produced from several anthropogenic sources. It is a 611 primary pollutant in the automobile emission and smokes derived from burning of biomass 612 and biofuels (Rogge et al., 1993;Kawamura et al., 2002). It is also a secondary product of 613 photochemical degradation of toluene emitted from anthropogenic sources (Suh et al., 2003). 614 It can play an important role to enhance the new particle formation in the atmosphere 615 (Zhang et al., 2004). The temporal variation of benzoic acid is similar to anhydrosugars 616 detected in Alaskan samples (Fig. 6a-c and 11m). We also found a strong positive correlation 617 (r = 0.95) of benzoic acid with levoglucosan (Fig. 8k), demonstrating that BB is the source of 618 benzoic acid in central Alaska. Alaskan aerosols (Table 2). These results imply that polyacids may have similar sources or 624 formation pathways in central Alaska. We found that polyacids showed no significant 625 correlations with benzoic acid (r = 0.17-0.53), which is mostly of BB origin in Alaskan 626 samples as discussed above (Table 2). These correlations and different temporal trends of 627 benzoic acid and polyacids suggest that forest fires are not the main source of polyacids in the 628 Alaskan samples (Fig. 11m-p). This remark is further supported by the insignificant 629 correlations of polyacids with levoglucosan (r = 0.29-0.47) (Fig. 8l-n). Claeys et al. (2004) 630 suggested that SOA tracer such as tartaric acid is produced by the photochemical oxidation of 631 isoprene. Interestingly, significant positive correlations (r = 0.67-0.78) of polyacids were 632 found with total isoprene-SOA tracers detected in Alaskan samples ( Table 2), suggesting that 633 they may be produced by photooxidation of isoprene in the Alaskan atmosphere. 634

Contributions of compound classes to aerosol organic carbon 635
The contributions of each compound class to organic carbon (OC) in the Alaskan aerosols are 636 given in Table 3 OAs in central Alaska. The concentration ratios of levoglucosan to mannosan (2.2 to 6.8) and 670 syringic to vanillic acid (0.02 to 0.5) suggest that burning of softwood is common source of 671 OAs. The higher levels of HMW n-alkanoic acids and n-alkanols than their LMW 672 homologues together with high CPI values of n-alkanes (5.2 to 9.9), n-alkanols (3.0 to 10) and 673 n-alkanoic acids (5.6 to 9.8) further suggest that they were emitted by the thermal ablation of Data availability. The data set of this paper is given in Table S1 in the supplement file.    13.8 a All the organic compounds quantified were converted to carbon contents and then divided by OC. See Deshmukh et al. (2018) for OC and dicarboxylic acids and related compounds. b The results of lignin and resin acids were combined due to the very low contribution of resin acid to OC.