Lipids can be effectively used as molecular markers of terrestrial and marine sources of atmospheric aerosols (e.g., Gagosian et al., 1981). In the terrestrial environment, different organisms contain a variety of long-chain alcohols (fatty alcohols; FAs) because of the diversity of pathways used by the biota for the synthesis of molecules and other compounds. Most fatty alcohols occur in biota as waxes covering parts of plants, which are useful for preventing desiccation, protection from bacterial attack, UV screening, and so on (e.g., Dahl et al., 2005). Because the main function of plant waxes is to reduce water loss through evaporation, waxes originating from this source tend to have longer chain lengths. The most prominent chain lengths are C26 to C30 (Tulloch, 1976). Once emitted into the atmosphere, FAs in atmospheric aerosols can act as tracers of primary biological aerosol particles (PBAPs) (Simoneit et al., 2004). Previous studies estimated global emissions of PBAPs mostly based on the abundance of fungal spores (e.g., the sugar mannitol as a biotracer) and bacteria (Carslaw et al., 2010). Among possible compounds of PBAPs, however, our knowledge is very limited on the abundance, emission mechanisms, and atmospheric behavior of FAs originated from plants. Possible emission drivers, such as temperature, light intensity, and local wind speed, likely vary with types and sources of PBAPs.

FAs are known as major components of surface lipids (waxes) with chains varying between C20 and C34 carbon atoms (Mudge, 2005). Fatty alcohols are composed of long-chain hydrocarbons and hydroxyl (-OH) groups, which can congregate at the aqueous surface and can act as surface-active organic aerosols (OAs). These aerosols can cause gases to transfer into the aqueous phase as atmospheric particles and prevent evaporation, which influences chemical reactions, particle lifetimes, fog/cloud droplets, and ice nuclei (IN) (Gill et al., 1983; Cantrell and Robinson, 2006; Knopf et al., 2018). In particular, biogenic particles such as fungal spores and bacteria can serve as a source of IN macromolecules, implying that IN macromolecules are therefore derived from a variety of biological or biogenic particles (Pummer et al., 2015). Long-chain alcohols show appreciable IN activity if they are crystallized into well-defined monolayers. This depends on chain length, the position of the OH group, and any substitutions on the side chains (Popovitz-Biro et al., 1994).

Oros and Simoneit (2001a) observed n-nonacosan-10-ol, a secondary FA (SFA), in smoke samples from conifers subjected to controlled burning. In addition, an alkanediol n-nonacosan-5-10-diol was identified in ambient aerosol samples collected in the western North Pacific during the ACE-Asia campaign (Simoneit et al., 2004). Some common alkanols were also found within ambient aerosols at urban and remote island sites and are thought to have been emitted to the atmosphere via biomass burning (Oros and Simoneit, 2001a, b). These studies also indicate an input of waxes from forests consisting of both soft and hardwoods. However, studies are very limited on the composition and sources of FAs in atmospheric aerosols obtained in the vicinity of source regions, such as forest environments.

In this study, we used gas chromatography–mass spectrometry (GC-MS) to identify five SFAs found in atmospheric aerosols, both in submicrometer particles and total suspended particulate matter (TSP), sampled throughout 1 year within canopies of two different temperate forests in Japan. We discuss their possible sources based on their seasonal variations in mass concentration found at the two forest sites.

Figure 2 presents an example of a GC-MS total ion chromatogram (TIC) of the TMS extract of the TSP sample collected at the FHK site. In previous studies, sugar compounds have been identified with the same method of derivatization, such as arabitol, fructose, glucose, and sucrose (e.g., Miyazaki et al. 2012; Müller et al., 2017). In the present study, P1–P5 refer to peaks which were identified to be TMS derivatives of SFAs with carbon numbers of 27 and 29, the peaks of which appear in the latter part of the TIC.

Figure 3 shows an EI mass spectra of P1–P5 obtained for the TMS derivatives with their molecular structures. On the basis of the detailed interpretation of the EI mass spectral data together with their comparison to literature data (Yamamoto et al., 2008) as well as the exact mass measurements by GC-TOF-MS, the derivatives were assigned to (a) n-heptacosan-10-ol-TMS (P1; C27), (b) n-heptacosan-5,10-diol-diTMS (P2; C27), (c) n-nonacosan-10-ol-TMS (P3; C29), (d) n-nonacosan-10-13-diol-diTMS (P4; C29), and (e) n-nonacosan-5-10-diol-diTMS (P5; C29). The spectra of n-heptacosan-10-ol-TMS is characterized by the presence of ion peaks at m/z 73, 229, and 341 (Fig. 3a). Similarly, the mass spectra of n-nonacosan-10-ol-TMS showed that ion peaks at m/z 73, 229, and 369 are significant (Fig. 3c). The ion at m/z 73, corresponding to [Si(CH3)3+], is characteristic for TMS derivatives containing one or more derivatized OH groups. Indeed, the exact mass measurements by GC-TOF-MS confirmed the estimated composition of each peak. For example, the composition of P3 with M-15 was identified as C31H65OSi, where its theoretical and experimental m/z are 481.4805 and 481.4826, respectively, with an error of 4 ppm.

The spectra of P2 (Fig. 3b), P4 (Fig. 3d), and P5 (Fig. 3e) were identified as the TMS derivatives of secondary fatty diols. The mass spectra of n-nonacosan-10,13-diol-diTMS (Fig. 3d) is characterized by intense signals at m/z 73, 229, 269, 327, and 367, whereas the spectra of n-nonacosan-5-10-diol-diTMS shows ion peaks at m/z 159, 317, 369, and 527 (Fig. 3e). Spectra with significant ion peaks at m/z 73, 159, 317, and 341 are produced by n-heptacosan-5,10-diol-diTMS (Fig. 3b), which was newly identified and is partly similar to the characteristics of the P1 and P5 spectra.

Figure 4 displays the seasonal variations of the identified SFAs in the TSP samples at the TMK and FHK sites in 2015. Here seasonal categories of spring, summer, autumn, and winter refer to the periods March–May, June–August, September–November, and December–February, respectively. The average concentrations in each seasonal category are summarized in Table 1. Among the SFAs identified, n-nonacosan-10-ol is the most abundant compound, followed by n-nonacosan-5,10-diol at both sites. The SFAs exhibited distinct seasonal changes, with the largest concentrations observed in the spring (growing season), both at the TMK (101.7±100.9 ng m-3) and the FHK (211.7±231.9 ng m-3) sites. The maximum concentrations of n-nonacosan-10-ol reached 268.6 and 442.1 ng m-3 at the TMK and FHK sites, respectively. The SFA with the second largest concentration, n-nonacosan-5,10-diol, showed a similar temporal variation to n-nonacosan-10-ol, suggesting that n-nonacosan-5,10-diol originated from identical or similar sources to those of n-nonacosan-10-ol. It is noted that the seasonal variation in concentrations of n-nonacosan-10-ol in 2016 was similar to that in 2015 at FHK (Fig. 4b). This similarity indicates that this seasonal trend is likely to be representative of the forest environment.

A primary plant origin of these SFAs is most likely, because the SFAs are known to be present in plant leaves; n-nonacosan-10-ol has been identified as a major compound in epicuticular waxes found in gymnosperm species (Tulloch, 1976; Schulten et al., 1986), such as Sequoiadendron giganteum (Yamamoto et al., 2008). Moreover, secondary fatty diols have been reported in other plant species, such as Pisum sativum (Vioque and Kolattukudy, 1997). In the atmosphere, n-nonacosan-10-ol was observed in smoke aerosol samples from conifers subjected to controlled burning (Oros and Simoneit, 2001a). Moreover, some of the common alkanols found in urban and remote island sites in spring are thought to have been emitted to the atmosphere by biomass burning (Oros and Simoneit, 2001a, b; Simoneit et al., 2004), while they also indicate an input of waxes from forests consisting of both soft and hardwoods.

In our study, a primary origin of biomass burning is unlikely because the SFAs showed no significant correlation with tracer compounds of biomass burning, such as levoglucosan with R2 of <0.02 (data not shown) at both sites. Rather, the seasonal trends of the SFA concentrations are similar to that of sucrose, which is an important primary saccharide of pollen grains (Pacini, 2000) and has been used as a tracer for pollen (e.g., Miyazaki et al., 2012). Specifically, the concentration of n-nonacosan-10-ol showed significant positive correlations with that of sucrose both at the TMK (R2=0.70) and FHK sites (R2=0.61). The average concentrations of n-nonacosan-10-ol were significantly larger than those of sucrose in the growing season at the two sites. These results suggest that the increase in the mass concentrations of n-nonacosan-10-ol and the other SFAs is due to large emissions of plant waxes or components thereof (i.e., fatty alcohols) from the forest leaves into the atmosphere. Our analysis also suggests that the SFAs identified here can act as tracers for primary biological aerosol particles (PBAPs) which are surface-active.

It is noted that the peak concentration of SFAs at FHK appeared in April (Fig. 4b), approximately 1 month earlier than at TMK (Fig. 4a). The difference in the appearance of the peaks is attributable to the difference in the growing season at the two sites: the growth of forest vegetation generally starts about 1 month earlier at FHK (35∘ N) than at TMK (42∘ N) (Nakaji et al., 2011; Takahashi et al., 2015), which is also evident from the seasonal changes in the net ecosystem production (Saigusa et al., 2008). The primary emissions of these organic compounds clearly depend on the activity of the forest vegetation and, more specifically, plant leaf phenology at these forest sites. Considering that the aerosol sampling duration of about 1–2 weeks is much longer than the timescale of transport of particles in the forest regions, the aerosol particles collected in this study were likely influenced by biogenic emissions from the whole area of each forest region. In contrast to the growing season, seasons in which the lowest concentrations were observed are not clear from Fig. 4. Large standard deviations of the average values in winter, summer, and autumn at the two sites (Table 1) mean that the difference in the average concentrations is mostly insignificant for these seasons. This is attributable to various factors (e.g., the difference in the emission strength and photochemical activity in the atmosphere) which need to be further investigated in a future study.

Figure 5 presents seasonal averages of the mass concentrations and particle size fractions of n-nonacosan-10-ol observed at the TMK site. The average values and the fractions of their mass to the TSP mass at the TMK site in 2015 are also summarized in Table 2. In spring, the mass of n-nonacosan-10-ol in the submicrometer size ranges accounts for ∼68 % of the TSP mass, the fraction of which is much larger than that in summer (∼10 %) and autumn (∼28 %). Similarly, the average mass fraction of submicrometer n-nonacosan-5,10-diol in TSP is 65 % in spring, whereas the mass fraction is 1 % and 15 % in summer and autumn, respectively. Our result is similar to that of Kavouras and Stephanou (2002), who also measured particle size distributions of n-alkanols (<C26) at Eucalyptus and conifer forests in summer to show clear bimodal size distributions with ∼40 % of the mass residing in particles with diameters smaller than 0.96 µm. In general, the majority of the mass of PBAPs resides in supermicrometer particles. Indeed, the mass fractions of sucrose in the submicrometer size range were 29 % in spring, showing that the majority (∼70 %) of the mass resides in the supermicrometer size range in the growing season (Table 2). In contrast to the typical PBAPs tracers, our result suggests that a significant fraction of the SFAs can be attributed to the submicrometer particles in the growing season.

Secondary fatty alcohols with carbon numbers of 27 and 29 were identified in atmospheric aerosol samples collected throughout the year at two different deciduous forest sites. As far as we know, this study is the first to identify fatty diols, such as n-heptacosan-5,10-diol, in atmospheric aerosol samples and to show the seasonal variations in their concentrations in the forest atmosphere.

Among the identified SFAs, n-nonacosan-10-ol was the most abundant molecular compound, followed by n-nonacosan-5,10-diol. The SFAs showed pronounced peaks in the growing season at both forest sites. In spring, the concentrations of n-nonacosan-5,10-diol were much larger than those of sucrose. Moreover, the concentrations of the identified SFAs showed significant positive correlations with those of sucrose at the two sites. These results indicate that the SFAs originated mostly from plant wax. The difference in the peak appearance of the SFA concentrations at the two sites indicates that the primary emissions of these organic compounds clearly reflect the activity of the forest vegetation such as plant leaf phenology at the forest sites.

In the growing season, the fraction of n-nonacosan-10-ol in the submicrometer size range accounted for ∼70 % of the TSP mass. Similarly, the average mass of submicrometer n-nonacosan-5,10-diol in TSP accounted for 65 % of the mass of TSP in spring. These fractions were much larger than those in the other seasons, suggesting that a significant fraction of the SFAs was attributed to the submicrometer particles in the growing season. This study demonstrates that the SFAs identified, particularly n-nonacosan-10-ol and n-nonacosan-5,10-diol, can be used as effective tracers for PBAPs with surface-active organic matter.